Coastal and Estuarine PDF

Coastal and Estuarine Studies
ManagingEditors:
Malcolm J. Bowman Richard T. Barber
ChristopherN.K. Mooers John A.
Coastal
and Estuarine Studies
40
David Prandle, Editor
Dynamics and Exchanges

in Estuaries and the
Coastal Zone
American Geophysical Union

Washington
Managing Editors
Malcolm J. Bowman
Marine Sciences Research Center, State Universityof New York
Stony Brook, N.Y. 11794, USA
Richard T. Barber
Duke Marine Laboratory
Beaufort, N.C. 28516, USA
ChristopherN.K. Mooers
Ocean ProcessAnalysisLaboratory
Institutefor the Studyof the Earth, Oceansand Space
Universityof New Hampshire
Durham, N.H. 03824-3525, USA
John A. Raven
Dept. of BiologicalSciences,Dundee University
Dundee, DD1 4HN, Scotland
Editor
David Prandle
ProudmanOceanographicLaboratory
BidstonObservatory,Birkenhead
MerseysideL43 7RA, United Kingdom
Library of Congress Cataloging-in-PublicationData
Dynamicsand exchangesin estuariesandthe coastalzone/ David

Prandle, editor.
p. cm.- (Coastaland estuarinestudies)
Proceedings
of a 5th conference,
whichwas held9-13 July 1990 at the
Gregynogstudycentre.
Includesbibliographical
referencesand index.
ISBN 0-87590-254-5
1. Estuarineoceanography--Congresses. 2. Coasts•Congresses.
I. Prandle, David. II. Series.
GC96.5.D96 1992
551.46'09 dc20 92-6818
CIP
ISBN 0-87590-254-5
Copyright1992 by the AmericanGeophysicalUnion2000 FloridaAvenue,NW, Washington,DC

20009, U.S.A.
Figures,tables,and shortexcerptsmay be reprintedin scientificbooksand journalsif the source

is properlycited.
Authorizationto photocopyitemsfor internalor personaluse, or the internalor personaluse of
specificclients, is granted by the AmericanGeophysicalUnion for librariesand other users
registeredwith the CopyrightClearance Center (CCC) TransactionalReportingService,
providedthat the base fee of $1.00 per copy plus $0.10 per page is paid directlyto CCC, 21
CongressStreet, Salem, MA 10970. 0733-9569/87/$01. + .10.
This consentdoes not extendto otherkindsof copying,suchas copyingfor creatingnew
collectiveworksor for resale.The reproduction
of multiplecopiesand the use of fullarticlesor the
use of extracts,includingfiguresand tables,for commercialpurposesrequirespermissionfrom
AGU.
Printed in the United States of

Contents
Acknowledgement
Contributors List
Part I. Introduction
Introduction
D Prandle ......................................................... 3
Part II. BaroclinicDynamics

Theinfluence
of coastally
trapped
waveson thecirculation
in JervisBay,New SouthWales
P D Craig and P E Holloway ............................................ 9
Tidal mixing near the sill of a Scottishsealoch

A J Elliott, P A Gillibrand and W R Turrell .................................. 35
A topographically
inducedinternalwaveandmixingin theTamarEstuary
D R Sturleyand K R Dyer .............................................. 57
Turbulenceandshearinducedmixingprocesses in estuaries
E J Darbyshire& J R West ............................................. 75
Dynamically-active
modelsin the predictionof estuarinestratification
J H SimpsonandJ Sharples...................................... 101
Part III. Circulation
Residualflow in NaplesBay andits effecton constituent

concentration,
constituent flux and residence time
J Van de Kreeke .................................................... 117
Thestratified
hydrodynamics
of thePalmiet- a prototypical
bar-built
estuary
J L Largier,J H SlingerandS Talijaard ...................................
viii
Salinitystructureof a shallow,tributaryestuary
W W Schroeder, S P Dinnel and W J Wiseman Jr ............................. 155
On meteorologically
inducedsubtidalmotionin HangzhouBay
J L Su and W Chen .................................................. 173
Water level fluctuationsin the AtchafalayaDelta, Louisiana:

tidal forcingversusriver forcing
E M Swenson and C E Sasser ........................................... 191
Modellingof low-frequencysalinityvariationsin the St. LawrenceEstuary

K T Tee .......................................................... 209
On the estuafinecirculationwithin the Kattegat

N Winkel-Steinberg, J 0 Backhausand T Pohlmann ....................... 231
Part IV. Sedimemation
Observations of fine-sedimentconcentrations
and transport
in the turbiditymaximumregionof an estuary
R J Uncles,J A Stephensand M L Barton .................................. 255
Velocity asymmetries in frictionally-dominated

tidal embayments:
longitudinaland lateral variability
C T Friedrichs,D R Lynchand D G Aubrey ................................ 277
Effectsof sea-levelrise on muddycoastalmargins

R Kirby .......................................................... 313
Acousticmeasurements
of suspended sedimentover sandwaves
P D Thorne,R L Soulsbyand P J Hardcastle................................ 335
Someobservations on fluid mudresponse to waterwaves

F Jiang and A J Mehta ............................................... 351
The reflection of waves off beaches

J Darbyshire ................................................
ix
Part V. Modelling (Sediment)

Dispersionin tidally-averagedtransportequation
R T Chengand V Casulli .............................................. 409
Effect of bends on dilution rates

R Smith .......................................................... 429
Modellingthe verticaldistributionof suspended

sedimentin combinedwave-currentflow
A G Davies ....................................................... 44!
Someconsiderations
on mathematical
modellingof morphological
processes
in tidalregions
Z B Wang ......................................................... 467
A three-dimensional
transportmodelfor dissolvedandsuspended
matter
in estuaries and coastal seas
G C van Dam and R A Louwersheimer .................................... 481
An estuarineand coastalsandtransportmodel
B A O'Connor and J Nicholson .................................... 507
Part VI. Applied Studies

Currentand densitystructurein the Netherlandscoastalzone
W P M de Ruijter,A van der Giessenand F C Groenendijk...................... 529
On thedistribution
of suspended
matterandthe densitydrivencirculation
in the Dutch coastal area
M Visser ......................................................... 551
Coastaldynamicsalonga ruggedcoastline
B King and E Wolanski ............................................. .. 577
Transportof hypoxicwaters:an estuary-subestuary

exchange
A Y Kuo and K Park ............................................... .. 599
Interdisciplinary
studyon the tidalfxontin theBungoChannel,Japan
T Yanagi,O Matsuda,S TanabeandS Uye ................................. 617
Hydrodynamic
modellingfor a tidalpowerproject
T L Shaw ...................................
Acknowledgements
The organisingcommitteefor the GregynogConferencewas as follows:
R.T. Cheng,U.S.G.S., SanFrancisco,USA

A.G. Davies,U.C.N.W., Bangor,UK
C.B. Pattiaratchi, C.W.R., Perth, Australia
D. Prandle, P.O.L., B irkenhead, UK
C. van de Kreeke, U. Miami, USA
As organiserof the precedingconference, Ralph Cheng'sadviceand encouragement was

particularlyhelpful. The success of theconferencewas,of course,primarilydueto the efforts
of the participants.In particularI shouldlike to mentionDr. Cath Allen, who, makinglittle
of a long-standingillness,both presentedand discussedat length her importantwork on
random-walktechniques.Sadly,Cath died in March 1991.
In thepreparationof thisvolume,thepainstakingeffortof therefereesis duly recognised,

likewisethe help and assistanceof the publishers.
Finally, neitherthe conferencenor this volumewould have happenedwithoutthe efforts

of JackieHardcastleandMike Blackleyof theProudmanOceanographic Laboratory.I have
beenwarnedendlesslyby, seemingly,everyonewho haseverorganised sucha conferenceand
volulne that it involvesa minimum of "one year out". Since my contributionhas been
minimal, I can only guessat the extentof Mike and Jackie'sefforts- sincerethanks.
David Prandle
ProudmanOceanographic Laboratory
BidstonObservatory,
Birkenhead
Merseyside
L43 7RA, United
List of Contributors
D G Aubrey P D Craig
225 Clark Lab CSIRO Division of Oceanography
WoodsHole Oceanographic
Institute GPO Box 1538
Woods Hole Hobart Tas. 7(}01
MA O2543 Australia
USA
G C van Dam
J O Backhaus Rijkswaterstaat
Instiut fuer Meereskunde Tidal Waters Division
der UniversitaetHamburg PO Box 20907
Troplowitzstr.7 25(}0EX The Hague

D-2000 Hamburg54 The Netherlands
Germany
E J Darbyshire
M L Barton Schoolof Civil Engineering
Departmentof Civil Engineering Universityof Birmingham
Universityof Birmingham BirminghamB 15 2TT
BirminghamB 15 2TT United Kingdom
United Kingdom
J Darbyshire
V Casulli Unit for Coastal and Estuarine Studies
MathematicsDepartment Marine Science Laboratories
Universityof Trento Menai Bridge

Trento GwyneddLL59 5EY
Italy United Kingdom
W Chen A G Davies
SecondInstituteof Oceanography School of Ocean Sciences
PO Box 1207 UniversityCollegeNorth Wales
Hangzhou Menai Bridge
Zheijiang310012 GwyneddLL59 5EY
China United Kingdom
R T Cheng W P M De Ruijter
US GeologicalSurvey Universityof Utrecht
Menlo Park Instituteof MeteorologyandOceanography
California Princetonplein5
USA 3584 CC Utrecht
The
xii
S P Dinnel F C Groenendijk
Center for Marine Science Rijkswaterstaat
Universityof SouthernMississippi Tidal Waters Division
StennisSpaceCenter Koningskade4
MS 39529 2696 AA Den Haag
USA The Netherlands
K R Dyer P J Hardcastle
Institute of Marine Studies ProudmanOceanographic
Laboratory
PolytechnicSouthWest BidstonObservatory
Drake Circus Birkenhead
PlymouthPL4 8AA MerseysideL43 7RA

United Kingdom United Kingdom
A J Elliott P E Holloway
Unit for Coastal and Esmarine Studies Departmentof Geographyand
Marine Science Laboratories Oceanography
Menai Bridge AustralianDefenceForceAcademy
GwyneddLL59 5EY CampbellACT 2600
United Kingdom Australia
C T Friedrichs F Jiang
MIT-Woods Hole Ocean Institute Coastaland Oceanographic
Woods Hole Ocean Institute EngineeringDepartment
Woods Hole, MA 02543 Universityof Florida
USA Gainesville
Florida 32611
A Van der Giessen USA
National Institute of Public Health
and Environmental Protection B King
(RIVM), Centre for MathematicalMethods Australian Institute of Marine Science
PO Box 1 PrivateMail Bag No 3
3720 BA Bilthoven Townsville Mail Centre 4810
The Netherlands Australia
P A Gillibrand R Kirby
Departmentof Agriculture,Fisheries Ravensrodd Consultants Ltd
and Food 6 QueensDrive
Marine Laboratories Taunton, Somerset TA1 4XW
PO Box 101 United
Victoria Road
Aberdeen AB9 8TB
United Kingdom
J van de Kreeke A J Mehta
Rosensfiel School of Marine and Coastaland Oceanographic
AtmosphericScience EngineeringDepartment
Division of AppliedMarine Physics Universityof Florida
4600 RickenbackerCauseway Gainesville
Miami, Florida 33149-1098 Florida 32611

USA USA
A Y Kuo J Nicholson
School of Marine Science Departmentof Civil Engineering
Virginia Instituteof Marine Science Universityof Liverpool
Collegeof William and Mary Brownlow Street
Gloucester Point VA 23062 PO Box 147
USA LiverpoolL69 3BX
United Kingdom
J L Largier
Center for Coastal Studies, 0209 B A O'Connor
ScrippsInstitutionof Oceanography Departmentof Civil Engineering

La Jolla, CA 92093 Universityof Liverpool
USA Brownlow Street
PO Box 147
R A Louwersheimer LiverpoolL69 3BX
Rijkswaterstaat United Kingdom
Tidal Waters Division
PO Box 20907 K Park
2500 EX The Hague School of Marine Science

The Netherlands Virginia Instituteof Marine Science
Collegeof William andMary
D R Lynch Gloucester Point VA 23062
BedfordInstituteof Oceanography USA

PO Box 1006
Dartmouth T Pohlmann
Nova Scotia Insfiut ruer Meereskunde
Canada B2Y 4A2 der UniversitaetHamburg
Troplowitzstr.7
O Matsuda D-2000 Hamburg54
Facultyof Applied BiologicalScience Germany
HiroshimaUniversity
Higashi-Hiroshima724 D Prandle
Japan ProudmanOceanographic
Laboratory
BidstonObservatory
Birkenhead
MerseysideL43 7RA
United Kingdom
xiv
C E Sasser R Smith
CoastalEcologyInstitute Loughborough
Universityof Technology
Center for Wetland Resources Loughborough
LouisianaStateUniversity Leicestershire LEll 3TU
BatonRouge,LA 70803 United Kingdom

USA
R L Soulsby
W W Schroeder HydraulicsResearchLtd
Marine ScienceProgram Wallingford
The Universityof Alabama Oxfordshire OX 10 8B A
PO Box 369 United Kingdom
DauphinIsland,AL 36528
USA N Winkel-Steinberg
GKSS-Forschungszentrum
J Sharples Max-Planck-Str
School of Ocean Sciences D-2054 Geestacht
Marine Science Laboratories Germany
Menai Bridge
GwyneddLL59 5EY J A Stephens
United Kingdom PlymouthMarine Laboratory
PlymouthPL1 3DH
T Shaw United Kingdom
Sir RobertMcAlpine
40 Bernard Street D R Sturley
London WC1N 1LG DalhousieUniversity
United Kingdom Departmentof Oceanography
Halifax, Nova Scotia
J H Slinger Canada B3H 4J1
Division for Earth, Marine and Atmospheric

Scienceand Technology JLSu
CSIR SecondInstituteof Oceanography
PO Box 320 PO Box 1207
Stellenbosch,7600 Hangzhou
South Africa Zheijiang310012
China
J H Simpson
School of Ocean Sciences E M Swenson
Marine Science Laboratories CoastalEcologyInstitute
Menai Bridge Center for Wetland Resources
GwyneddLL59 5EY LouisianaStateUniversity

United Kingdom BatonRouge,LA 70803
S Taljaard S Uye
Divisionfor Earth,Marine andAtmospheric Facultyof AppliedBiologicalScience
Scienceand Technology HiroshimaUniversity
CSIR Higashi-Hiroshima724
PO Box 320 JAPAN
Stellenbosch,7600
South Africa M Visser
Instituteof Meteorology
andOceanography
S Tanabe Universityof Utrecht
Departmentof EnvironmentConservation Princetonplein5
EhimeUniversity 3584 CC Utrecht
Matsuyama790 The Netherlands
ZB Wang
K T Tee Delft HydraulicsLaboratory
Physicaland Chemical ScienceBranch PO Box 177
Departmentof Fisheriesand Oceans 2600 MH Delft
BedfordInstituteof Oceanography The Netherlands
Dartmouth, Nova Scotia
Canada J R West
Schoolof Civil Engineering

P D Thorne Universityof Birmingham
ProudmanOceanographic
Laboratory BirminghamB 15 2TT
BidstonObservatory United Kingdom
Birkenhead
MerseysideLA3 7RA W J Wiseman, Jr
United Kingdom Coastal Studies Institute
LouisianaStateUniversity
W R Turrell BatonRouge,LA 70803
Departmentof Agriculture,Fisheries USA
and Food
Marine Laboratories E Wolanski
PO Box 101 Australian Institute of Marine Science
Victoria Road PrivateMail Bag No 3
Aberdeen AB9 8TB Townsville Mail Centre 4810
United Kingdom Australia
R J Uncles T Yanagi
PlymouthMarineLaboratory EhimeUniversity
PlymouthPL1 3DH Matsuyama790
United Kingdom Japan
Part I
Introduction
Dynamics and Exchanges in
Coastal and Estuarine Studies Estuaries and the Coastal Zone Vol. 40
Introduction
D Prandle
In establishingthepeculiarnicheof a volumeof conference proceedingswithinthescientific

literature,theeditorhasa responsibility
(beyondthecullingandprocessing) to informreaders
of the background to the conference,the subjectareasand specificfoci. More tentatively,
underlyingdirections andcommonthemesshouldbe discerned to perhapsindicateareasof
research likelyto be fruitfulandtopicalin thenextfew years.Sincemanyreaderswill only
studyselected contributions,thevolumeis dividedintofive sections namely:-II. Baroclinic
Dynamics;III. Circulation;IV. SedimentDynamics;V. SedimentModelling;VI. Applied
Studies.This introduction incorporatessub-titlesto facilitatethe sameselectivity.
1.1 The GregynogConference

The conferenceseriesbeganin Hamburgin 1978 with the aim of encouraging cooperation
betweencoastalengineers andcoastaloceanographers. Proceedings havebeenpublishedby
Springer-Verlagin theirsectionon CoastalandEstuarineStudies.Gregynog(9-13 July 1990)
markedthe 5th, in what is now a biennialseries,to be followedby Perth,Australiain 1992
and the east-coastUSA in 1994. Some 70 scientistsattendedwith 40 oral and 20 poster
presentations.While participation
isprimarilyby directinvitation,nooneis excludedandcare
is takento ensurea balancein internationalrepresentation,betweenphysicists,engineersand
mathematicians,and acrossthe spectrafrom 'grey beards'to researchstudents.
Likewise,residentialvenues(Greygnogis a remotestudycentrekindly madeavailableby

theUniversityof Wales)havebeencarefullyselectedto promoteinformalcontactsacrossthe
dividesof bothage and discipline.
1.2 The Proceedings

All papershavebeenreviewedby at leasttwo refereesincludingsecondreviewswheremajor
revisionswere indicated. Theserefereeswere urgedbothto maintain'Journalstandards'and
to ensurecoherencewithin the scopeof the volume. As a resultof adheringstrictlyto these
requirements,the volumeis slimmerbut of a uniformlyhigherstandard.
Copyright American Geophysical Union

4 D Prandle
Part II- BaroclinicDynamics
Theorderof thefive papers in thissectionconvenientlyfollowsa sequenceof (firs0some

largescaleobserved featuresof suchdynamics thence in thenextthreepapers studiesof
successively
freerscaleprocesses. A final paperthenattempts to modelthe associated
turbulent
dynamics at a specificlocation.CraigandHollowaydescribe howoscillationsin
thethermalstructure
of a bayonAustralia'seastcoastcanbeassociatedwithcoastally trapped
wavesextending to theedgeof theadjacent shelfseas.Elliottet al develop
a numerical
model to simulate the influence of a sill in a Scottishsea loch on the saline stratificationo
similardynamics
maybe foundin manyfjords.SturleyandDyerobserved
themotions
of
internalwavesin a smallUK estuaryand estimatedtheir contribution
to verticalmixing.
Darbyshire
andWestdescribe directmeasurements
of turbulent
motions andsaltfluxesin the
sameestuaryandinferthatadditional
mechanismsmustcontributeto theobserved vertical
mixingrate. Finally,Simpson
andSharples
present
a modelstudythatincorporatesa Mellor
and YamadaLevel 2 turbulenceclosureschemeto reproduceverticaldensitydifferences
observedin LiverpoolBay.
Part HI - Circulation
This sectionincludes7 papersconcerned with circulationbothhorizontal andvertical.The

scaleof circulation
variesfrominternallywithina specificestuary, betweenestuaries (orbays)
to circulationbetweenadjacentseas. The first paperby Van de Kreekedetermines the
flushingrateof a smallFloridabayusinglimitedcurrentmeterobservations in conjunction
with dataon freshwaterinflowsand salinitydistribution.Largieret al describethe saline
exchange in a smallSouthAfricanestuarywheregravitational circulationis constrainedby
a hydrauliccontrolcreatedby therestricted crosssectional areaof themouth. Schroeder et
al useextensive observationsof salinitydistributionsto determinetheinfluenceof tide,wind
andfiverdischarge on thesalinitybalance between a sub-bay andconnecting bayin theGulf
of Mexico. Su and Chen use spectralanalysisof sea level oscillationsto separatethe
frequency domains withinwhichlocalwindforcingversus externalsealevelspredominate in
a largeshallowbayin China. Swenson andSanerusesealeveldataextending aslongas 15
yearstorelatetheinfluence of tides,windsandfiverflowonflooding(andthereby vegetation)
in a LouisianaDelta. Tee developes an x-z numericalmodelof the St Lawrenceestuaryto
explain40-50dayaxialoscillations observed in bothcurrentmeterandsalinityobservations
fromthislong(400kin) anddeep(300m) system).Winkel-Steinberg et al describenumerical
modelresultsof exchanges betweena deepandshallowseawithan estuarine-like baroclinic
circulation.
Part IV - SedimentDynamics
The extentof the emphasison sedimentation withinthisconference reflectsthe increasing

requirementto studyestuarine
physics for application
to waterqualityproblems withinwhich
the motionof fine sediments
is oftencrucial. Uncleset al presenta comprehensive studyof

D Prandle 5
the turbiditymaximumin the Tamar; using a numericalmodel in conjunctionwith

measurements
of velocity, salinity and suspended
sedimentconcentration
to indicatethe
controlling mechanisms.Friedrichs et al examinethe interaction
betweendynamics and
morphology in short(<20km),stronglytidalembayments,emphasisingthepredominantrole
of frictionin determiningebbor flooddominance in netsediment
movement. Reflectingthe
character of theconference,theseresultsfroma complexanalytical
studymaybecontracted
witha relatedstudybasedonanalysis of observational
data.Kirbyexaminestheprofilesof
inter-tidalmuddy-shores, notingtheinfluenceof bothtideandwavesandextrapolating to
considerthe impact of any future rise in mean sea level. Thorne et al describerecent
developmentsof anAcoustic
Backscatter
Probeformeasuringsandmovement, comparing data
frombothfieldandlaboratorywiththeory.FianandMehtadevelop a modelto simulatethe
propagation
of surfacewavesover a layerof softmud,resultsexplainobservationsmade
withinthemudlayerandemphasise theimportanceof longerperiodside-bands
in thewave
spectra.Darbyshireshowshowa triangular
arrayof bottompressure gauges
canbe usedto
estimate wave reflectance on a beach.
Part V - SedimentModelling
Thissection includessixpapersthataddress variousaspects

of modellingdispersion
with
particularapplication
to sedimentmovement.Chengand Casulliexaminethe accurate
definition
of parametersin thetidally-averaged
dispersion
equation,
theirhypotheses
arethen
rigorously
tested
usinga numerical
modelembracing
themouth
of SanFrancisco
Bay. Smith
extendshis seriesof theoretical
paperson dispersion
in estuaries
to considerthe effectof
bendson particlesof varyingdensities.Daviesusesan x-z modelto examinethe influence
of surface
wavestransported
by a (tidal)currentindicating
increases
of upto twoordersof
magnitude. Wangappliesa mathematical modelof sediment transport
in theYantzeestuary
in an attemptto simulatemorphological development.Van Dam and Louwersheimer
incorporate
analytical
functionsforbothvertical
andhorizontal (sub-grid
scale)motionswithin
a verticallyaveraged
coastalzonemodelto simulate particledispersion.O'Connorand
Nicholson
incorporate
withina depth-averaged
tidalmodel,theinfluence
of surface
wavesto
estimate
sediment
motionfollowing
dredging
in an estuary.Thisengineering
application
remindsus thatsomepeoplemustprovideanswersnow.
Part VI- AppliedStudies
Thesixpapers
in thissection
emphasise thediversityof regimes,
withobservations
fromfour
continents
andfroman estuarywith (almost)the world'slargesttidesto onewheretidal
exchanges
aresecondary. De Ruijteret al describe
thecoastal
plumefromtheRhinewhich
persists
for up to 80kmalongshore
andfor 25kmoffshore;boththelocalisedstratification
and
theoverallgeographic
configurationareshownto varystronglywithbothwindandfiver
discharge.
Visserdevelops
a linearised
analytical
modelof thecross-shore
circulation
within
theRhinePlume,thereby
explainingsurface
observations
of a minimum in suspended
sediment
concentrations
alongitsoffshore
edge.KingandWolanskiusea vertically-averaged

6 D Prandle
tidal modelof Australia'sCentralGreatBarrierReef to studythe exchangeof coastalwaters.

Kuo and Park use measurements of currents,salinity and dissolvedoxygen to calculate
exchangeratesbetweenChesapeake Bay and a sub-estuary concludingthat tidal motionsare
of secondary importance.Fromobservations of waterqualityparameters closeto theedgeof
a (temperatureand salinity) front in Japan'sSeto Inland Sea, Yanagi et al illustratethe
complexityof the interactions betweenphysical,biologicaland chemicalprocesses.In the
final paper,Shaw describesthe rangeof studiesrequiredfor the planningof a tidal power
scheme in the Bristol Channel.
1.3 Conclusions
Hopefullytheabovesinglesentence summaries helpin bridgingthegapbetweenthetitlesand

abstractsthataccompanyeachpaper.Thepublicationdeadlineforcedme to readall 30 papers
in one (extended)weekend. While not recommending this, it doesyield a usefuloverview.
The niche that commendsthis volumeis, perhapsinadvertently,suppliedby the form of
presentationassumedby almostall theauthors.Specifically, theyhavewrittenprimarilyfor
a broad spectrumof readershipreflectingthe conferenceaudience. Thus, in contrastto
specialisedpapersfor 'narrow' journals,the authorsinclude clearer introductions,more
thoughtful
justifications
of theiraimsandbroaderinterpretations
of theirresults,
whilerelying
on referenced
papersfor technicaldetails.Consequentlythepapersaremoreeasilydigested
andmorereadilyprovidebothan overviewandan insightinto thespecificdiscipline.
Lest anyonehasmissedthe discernible trendin estuarineresearchawayfrom traditional

engineeringapplications towardswaterqualityissues,thequestionnowis to detectanyloss
of momentum in thisveritableband-wagon.Havingspentthe lastfive yearscollabomting
with biologists,chemistsand sedimentologists in the U.K. North Sea Project, my
overwhelming conclusion is for the needto adoptprotocolsto facilitatesuchco-operation.
Theinternationalnatureof thecommercial marketforbothinstruments andcomputer hardware
and softwareeffectivelyenforcessuchprotocolsin theseareas. Likewisethe establishment
of bothnationalandinternational datacentres avoidsmuchduplication of effort,simplifies
modelling simulationsandreduces thecomplexities of inter-comparison
exercises. Whilethe
conference seriesdeliberatelyavoidsover-representation of narrownumericalmodelling
studies,
approximately 20 discrete modelling modules canbeidentifiedwithinthese30papers.
Almostall of thesemodules andmanymore,needto be incorporated in anycomprehensive
simulationof waterqualityparameters in estuarine
systems.Establishing theprotocols for
easyinter-connectionsbetweensuchmodules (andtherelateddatainputsandoutputs) poses
a majorchallenge.It is important thatthelinkingframework adoptedis sufficiently
flexible
to minimiseconstraintsandencourage individuality
withinco-operativeresearch programmes.

Part II
Baroclinic Dynamics
The influence of coastally trapped waves

on the circulation in Jervis Bay, New
South Wales
P D Craig and P E Holloway
Abstract
JervisBay is a small(15 km x 8 km) coastalembaymentin New SouthWales,Australia.As

partof a baselinestudy,threecurrentmetetingandthreetide gaugestationswere established
in thebay for onemonthin the summerof 1988/89.The bay wasstronglythermallystratified
and the dominantsub-inertialdynamicsappearto have been a responseto the large-scale
baroclinicmotionover the continentalshelfand slope.Many of the observedfeaturesof the
dynamicscan be reproducedwith a simplerectangularquasigeostrophic model of the bay,
forcedby temperature(and thuspressure)oscillationsimposedat the mouth.
2.1 Introduction
JervisBay is an embaymenton the eastcoastof Australia,locatedapproximately150 km

southof Sydney(Figure 1). It measuresabout 15 km in the north-southdirection,and 8 km
east-west,with only a narrow, 3.5 km, mouth to the oceanin the southeast. The depth
increasesfrom the coastlineof the bay to a maximumof 35 m in the mouth,with an average
depth,in the bay, of lessthan 20 m. Under normalcircumstances thereare no significant
sourcesof freshwaterinto the bay although,duringtimesof heavyrain, runoff from the land
may dramaticallyinfluencethe salinitystructure.
Despiteitsproximityto Australia'slargestcity,JervisBay andits immediatehinterlandare

in a relativelypristinecondition.Thereareno majortownsor industrialdevelopments on its
shoreline.There was, however,a federalgovernmentproposalto establishhome-porting
facilitiesin the bay for half of Australia'snavalfleet. In anticipation
of this possibility,a
three-year,principallybiological,baselinestudyof the bay was commenced in late 1988. A
smallcomponent of the studywasan investigation of the bay's dynamics,whichincludeda
month'sfield measurement programconducted in December1988andJanuary1989.

10 P D Craig andP E Holloway
150o40'E 42' 44' 46' 48' 50'

I I I
Crookhaven Bight
35ø00'S Hare Bay
&Green
&Blue
(Isson
Bay
obart
ired
100
,range
Metoo
'ological
s/
ß Currentmetermooring
It Tide gaugemooring
0 CTD station
Figure1: Chartof JervisBay, showinglocationsof themonitoringstations
The long-termaim of thephysicalstudieswasto understandtherole of thecirculation

in
distributing
naturalandanthropogenic
speciesaroundthebay.In theshorterterm,theobjective
wassimplyto identifythedominant
forcingmechanisms, andthenatureof thebay'sresponse
to them.
Duringthe measurement program,watersof the bay werestronglythermallystratified.

Tidesin thebaywereweak,aswasthelocalwind.The circulation appearedto beprincipally
a response to the larger-scale
dynamicsoverthe continental shelfandsloperegionexternal
to the bay. This particularstretchof the Australiancoastlineis renownedfor its coastally
trappedwaves.It wasthenon-isostatic sea-levelsignature
of theregionthatledHamon(1962,
1966)to firstpostulatetheexistenceof suchwaves.Then,in 1983/84,thecoastalwaterswere
instrumented for the AustralianCoastalExperiment(ACE) that led to descriptions of the
structureandpropagation of trappedwavesalongthe shelf(Freelandet al., 1986;Churchet

P D Craig andP E Holloway 11
al., 1986a,b; Huyer et al., 1988).
The JervisBay measurement programincludedno provisionfor identificationof coastally

trappedwaves.However,theftpresencecanbe inferredfrom sea-levelandtemperature records
collected in the mouth of the bay. In the present paper, we will examine, with the
measurements, and with a simple conceptualmodel, the responseof the bay to these
shelf-scale motions.
2.2 The field program

In all, eightcurrentmeters,threetidegaugesanda windstationweredeployedin thebay.The
currentmetersweremooredat threestations,designated "Red","Blue"and "Green",with four
meterson thefirstmooringandtwoon eachof theothers.Tide gaugestationsweredesignated
"Black", "Yellow" and "Orange",and the meteorologicalstationwas locatedon Governor
Head, on a 5 m tower, approximately50 rn above sea level. Locationsof the stationsare
shownon Figure 1, and detailsof the instrumentationare listedin Table 1. All instruments
were deployedon 7 December1988,and recoveredon 11 January1988.
Table 1. Detailsof instrumentdeployments
Station Water depth Instrument Instrument Data

depth
(m) (m)
Red 22 6.5 S4 c/m good
10.5 S4 c/m good
14.5 S4 c/m good
18.5 S4 c/m good
Blue 14 5.4 S4 c/m 8 daysonly
11.5 Aanderaa c/m good
Green 12 5.4 S4 c/m good
9.5 Aanderaa c/m temp only
Black 11 11 Aanderaat/g good
Yellow 20 20 Aanderaat/g good
Orange App.Micro.Sys good

t/g
* c/m - currentmeter; t/g - pressure-sensing

tidegauge

In additionto the fixed measurements, west-easttemperature

and salinitytransectswere
conducted on 5 December1988and12 January1989,fromthewestcoastof thebay,through
the mouth,and out to the edge of the continentalshelf. Thesetransectsare part of a
longer-termsurvey,andhavebeenconducted, on a monthlybasis,sinceFebruary1988.By
now, theyprovidea usefulrecordof the annualcycleof verticalandhorizontalstratification
acrossthe bay and continentalshelf(Hollowayet a/.,1990).
2.3 Field results
Data from the measurement programhave,to date,beenpresented only in limited-circulation

reports(e.g.Hollowayet al., 1989, 1990;Craig,1989).In thissection,we will presentonly
thedatathatarerelevantto thepresentdiscussion. In particular,
we will be concentrating on
the subinertialdynamicsof JervisBay. However,the pointshouldfirst be madethattidal
amplitudes in thebayareverysmall.Tide-height amplitudes are0.5 and0.2 m fortheM2and
K, amplitudes, respectively,
andM2 tidalcurrentamplitudes (semi-major axis)are0.02 ms"
at Red, 0.01 at Blue and0.005 at Green.K• currentamplitudes weresimilar,but may have
beencontaminated by a sea-breeze
response.TheS2amplitudes arearoundone-fifththeM2,
and O, abouthalf the K•.
Winds duringthe measurement period were light. The region is subjectto a distinct
sea-breeze
cycle,and yet the maximumrecordedhourly-averaged wind speedwaslessthan
10ms"With diurnalvariations
filteredout,the maximum
windspeedis about7 ms"
(Figure2). Thewindsignalin Figure2 hasbeentreatedwitha least-squares
filterof thetype
describexl
by Thompson(1983)anddesigned, in thiscase,to suppress
oscillationswithperiods
lessthan1 day,andpassthosegreaterthan2 days.
Alsoshownin Figure2 is thecurrentspeedrecorded at theuppermost

currentmeterat the
Red mooring,treatedwith the samefilter, togetherwith a modelsimulationof the current
speedfor thesamelocationandthesameperiod.The simulated currentwasproduced by a
conventional,three-dimensional,
constant-density,
constant-eddy-viscosity
modelof thebay,
drivenby thewindsin Figure2a. Thewindspeedandthemodelled currentspeedindicatethe
level of correspondenceto be expectedwith strictlywind-drivendynamics.The measured
current,ontheotherhand,showslittleresemblance to theothertwocurves.Whilethissimple
modelcomparison is notregardedasdefinitive,thelackof correspondencebetween peaksin
thewindandpeaksin themeasured currents
suggests thatthewinddoesnotexerta significant
directforcinginfluenceon the bay.
Figure3 showsthe othertime-series collected

at the meteorological
station,thatof the
atmospheric pressure.
Themw recordin Figure3 is shownagainin Figure4 aftertreatment
by a filterchosen
specifically
to isolate
thedominant longer-term
signalin thepressure.The
filtersuppresses
periodslessthan4 days,andpasses thosegreater
thanfive.Thepressure
signalin Figure4 shows,specifically, threecompletepeaksand threetroughs,with an

10'0
•I1"
IIIIIIIIIIIIIIII[IIIII1IIIII (a) Wind
5.0
0.0
360
I I•,[• I I I I I I I I I.i i i i i i i i i I i I''11 I;• I I
ß .
270 _ (b) Wind

ß
o
ß•- 180
- \
,r-- 9O
;
I•'"•1 I I ' I I I I I ! I I I I '1 I I
0.10 I I I I I I I I I I I I I I I I I I I I I I [ I I I I I
• (c) Modelled
current
"-' 0.05
0,00
0.10
0.05
-IIIl.IIIIIIIIIIIIII[IIIIIIIIIII
--
(d) •easured current
0.00 I I I I I I I I • IV
8 10 15 20 25 3031 1 5 7
DEC., 1988 JAN 1989
Figure2: Windspeed
(a)anddirection
(b)forthemeasurement
period,
together
withsimulated
(c)and
measured
(d) currentspeeds
from6.5 m depthat theRedstation(Records
arefilteredto removediurnal
and higherfrequencies).

14 P D Craig and P E Holloway
oscillatory
periodof 7 to 10 days.Fig. 4 alsoshows,for comparison,
filteredrecordsof the
sea-surfaceelevationfrom the Black station,and watertemperatures
from 11 m at Black and
20 m at Yellow.The correspondence betweenthefourcurvesis obvious.The surfaceelevation
signalwasrecorded by a submerged pressure
sensor,and so doesnot containthe "inverse
barometer"signal.Therangeof surface movement indicated
in thesignalis approximately
0.2
m. The Yellow thermistoris 9 m deeperthat that at Black, and consistently recording
temperaturesabout1ø cooler.On thisbasis,the temperatureoscillations
of around2ø canbe
convertedto a verticalisothermmovementof about20 m, or approximately 100 timesthe
amplitudeof thesurfaceoscillation.
Theisotherm movement is out-of-phasewiththesurface:
as the surfacerises,the isothermsfall, characteristic
of baroclinicmotion.The atmospheric
pressure
lagsthesurfaceelevationslightly,with a delayof approximately
2 days.
1030.0
1 I I I I I I I I I I I I I I I I I I I I i I I I I I I I' I I I I I I I I /
1020.0
1010.0
----' 1000.0
990.0
980.0
-
IIIIIIIIIIItIIIIIIIIIIII!IIIiiii!II!
6 10 15
DEC 1988
20 25 3031 1
JAN 1989
5 10 12
Figure3: Atmospheric
pressure
recordfor 6 Decemberto 12 January
Despitethestrongcorrelationin Figure4, theatmospheric

pressureis unlikelyto be directly
responsiblefor the motionindicatedby the temperature
andsurfaceelevationcurves.On the
scaleof JervisBay, the only responseto changingatmospheric pressureis likely to be an
inverse
barometer
effect,withsmallvelocities
(of order10'3ms4) andnobaroclinic
signal.
The final curveon Figure4 is thenorthwardwind component, filteredin the sameway as

the other parameters.At theselow frequencies,there is a close correlationbetweenthe
north-southwind andthe atmospheric pressure,with theformerleadingthelatterby abouttwo
days.This wind componentappearsto be the majoragentforcingthe responseseenin the
otherparameters in Figure4. Thereis no obviousreasonwhy JervisBay, whosenorth-south
and east-westdimensionsare similar,shouldrespondpreferentiallyto windsblowingalong
the north-southaxis. However,the northwardwind componentis the longshorewind for the
open continentalshelf outsideJervis Bay. The open shelf waters can be expectedto be
significantlymore responsiveto longshorethanto cross-shore winds(e.g. Gill and Clarke,
1974).

P D CraigandP E Holloway 15
Themostlikelyexplanation
for thecorrelation
evidentin Figure4 is thatit is indicative
of coastally
trapped
waveactivity
overtheshelfandslope region external
toJervisBay.
DuringtheAustralian
Coastal
Experiment in 1983/84,
threemooring linesweredeployed
acrossthesouth-eastern
Australian
continentalshelfandslopetodocument thepropagation
ofcoastally
trapped
waves
(Freelandetal.,1986;Church etal.,1986a,b;Huyeretal.,1988).
The firstthreewavemodeswerein factfoundto account for approximately
half the
sub-inertial
currentenergyalongthispartof thecoast.The correlation
betweensurfaceand
isotherm
displacement,
asseenin Figure4, is anestablished
property
of thewaves, andthe
displacement
ratioof 1:100iscompatible
withtheory (e.g.GillandClarke,1974).
| J I I i I I I I I i i I I I J J J J I i i I i i i i I i i i I I J J J I
1030
._.1020
• lOlO
17.1
(b) Blacksurfaceelevation
17.0
16.9
19.0 -
18.0 -
17.0 -
16.0 -
15.0 -
14.0 -
(e) northwardwind
5.0
o.o
-5.O
ß I ! ! ! I I I ! I I ! I J I I I f I I f I I ! ! I I I ! I I ! ! I I I I
6 10 15 20 25 30 I 5 10
DEC 88 JAN 89
Figure
4:Filtered
records
of(a)atmospheric
pressure,
(b)surface
elevation
from
theBlack
station,
(c)
watertemperature
at 11 •n depth
fromBlack,(d) watertemperature
at 20 m fromYellowand(e)
northwardwindcomponent.

Thereis ongoingdiscussion aboutthe originof the coastallytrappedwavesin the ACE

region, with much of the attentionfocussedon the role of BassStrait (e.g. Churchand
Freeland,1987;Clarke, 1987;MiddletonandViera, 1991).Localwindwill alsogeneratewave
activity.The coastallytrappedwavesare generallydescribed as "free"or "forced"depending
uponwhetherthey are generatedremotelyor locally(e.g. Churchet al. 1986b).Thereis no
point pursuingthis distinctionin the presentcontext. However,in Figure 4, it is clearthat
northward,that is, downwellingfavourable,windsare relatedto rising water temperatures,
while southwardwinds relate to falling temperatures. There is a consistent
phasedelay of
approximatelytwo daysfrom the time of peakwind speedto the corresponding temperature
maximum or minimum.
The wavelengthof the coastallytrappedmotionis of the orderof a thousandkilometres,

muchwider than the mouthof JervisBay. From the perspectiveof the bay, then,thereis a
stronglybaroclinicsea-surfaceandthermaloscillationat the mouth,whichforcesa response
inside the bay. Figure 5 shows the temperaturesignal recordedby thermistorsat
approximately 10 m depthon eachof the current-meter moorings.At leastaccording to the
GreenandBlue moorings,theresponse acrossthebay is closelycorrelatedto thatat theBlack
station.The response throughthewatercolumnis demonstrated by theRed mooring(Figure
6), wherethe deeperthermistors, particularly,showthe sametemperature signature. In both
Figures5 and 6, the Red 10.5 m recordappearsto be anomalous,but we haveno indication
whetherthe anomalyis real, or due to instrumentmalfunction.
I I t I I [ I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
22.0 - - 22.0
21.0 - - 21.0
20.0 ..... ß - 20.0
19.0 ß ,,•.•h.,•..." /
- / ,•, - 19.0
18.0 _ /•-
'"'Bl.u..e...
/:":i •r ,on•'/',"'"'/'
/'• / "-'"
17.0 /.,.z
-, • .... •
....... ";,,..• //
- t•,, x, '- Z'"'..'",.,.-"/ - 17.0
/,: Blac,
k
16.0 _ /.:, ----J - 16.0
,
ß
o
15.0 - - 15.0
14.0 - - 14.0
I6 I I I I10 I I I I 11
5I I I I I
20 I I I I I
25 I I I I 1301
I1 I I I I5 I I I I I10 I
DEC 88 JAN 89
Figure5' Filteredtemperature
recordsfroInBlackat 11 m depth,Blueat 11.5m, Red at 10.5m and
Green at 9.5 m.

i I I I I I I J I '1 I I I I I I I I t i I I I I I i i I i I I I I I I I I
2.2.0- -
21.0
20.0
19.0
18.0
17.0
16.0
15.0
14.0
I ! I I I ! I I I I I I I I I 1 [ t I I ! ! J J I I I I t I I I I I t I
6 10 15 20 25 30 I 5 10
DEC 88 JAN 89
Figure 6: Water temperatures

from the Red stationat depthsof 6.5, 10.5, 14.5 and 18.5 m.
Measuredcurrentshave a distinct,thoughnot so clean,response,similar to that of the

temperatures.In Figure7, currentcomponents areplottedtogetherwith temperatures,
for each
stationin the bay. For the Green and Blue stations,the currentcomponents are rotated
45ø clockwise, and 30ø anticlockwise,respectively,from north, to give a direction
approximately parallel to the bathymetry.For the Red station,the eastwardcomponentis
plottedfor metersbothat 6.5 rn and 18.5rn depth,thatis, at boththe topandbottomof the
water column.As is to be anticipatedwith baroclinicmotion,the two tracesare 180ø out of
phase,that is, almostequalin magnitudebut oppositein sign.The Greencurrentsand 6.5 m
Red currents,both measurednear the surface,are approximatelyin-phasewith the
temperatures, while thecurrentsat Blueandat 18.5 rn at Red arecloseto 180ø out-of-phase
with temperature. For eachstationin Figure7, thetemperature plot hasbeenmovedon the
time-axisto maximise, approximately, thevisualcorrelation
betweenthetemperatures andthe
currents.Thesizeof thisdisplacement in timeprovidesan indicationof thephaserelationship
betweenthe signalsand is listedin the figurecaption.

] I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
0.10 - - 19
0.05 -
Red r• 18'5m
_18
18.5m 17
E 0.00
"---6
5•16
-0.05 -
-0.10 - 15
0.10 -
0.05 _ 20
Gre 5.4m
•7
19
0.00
-0.05 18
-0.10
0.10
0.05
-
-
• •_•11.5m
_20
18
v
E 0.00 -""-"• 115
-0.O5
-0.10 •6
I I I I • I I I I I I I I ! I I I I I I I I I I I I I ! I • • • • • I I
6 10 15 20 25 30 1 5 9
DEC 88 JAN 89
Figure7: Overlaidtemperature (thinline) andcurrent(heavyline) recordsfor theRed,GreenandBlue

stations,at theindicateddepths.Velocitycomponents areat 90ø,-45ø and30øclockwisefromnorth,and
temperature recordsare displaced2 daysforwardfor Red, and oneday for eachof GreenandBlue.
2.4 A quasigeostrophic
model
In this section,we examinea simplemodel to determinewhetherthe observedcurrentand
temperature oscillations
arecompatible
withthehypothesis thatJervisBayresponds principally
to longshorewindsas a resultof remoteforcingthroughthe mouthof the bay. The simplest
setof equationsthatcouldreasonablybe expectedto describethestratifiedcirculationin Jervis
Bay is linear,hydrostatic,Boussinesq
and frictionless:

u, - fv - -P=, (2.1)
New' 0 , (2.3)
u•+ v• + w,- O, (2.4)
(e.g.Churchet al., 1986a;Gill andClarke,1974)wherex, y, z areCartesiancoordinates, with

z verticallyupwardsandy northward,u, v andw are corresponding velocitycomponents, P
is theperturbationpressuredividedby themeanwaterdensity,N(z) is thebuoyancy frequency
andf is the Coriolis frequency.The unknownsin (2.1) to (2.4) are u, v, w and P, time is
represented by t and subscriptsrepresentdifferentiation.
Equations(2.1) to (2.4) are appliedto an idealisedrepresentation

of JervisBay, that is
assumed to be bothrectangular andflat-bottomed(Fig. 8). The domainis thendescribed
by
0 < x < X, 0 < y < Y, -H < z < 0. The mouthof thebayis modelled as a channelgivenby
x = X, y• < Y <Y2.
On the verticalwallsandthebottom,the boundaryconditionis simplythatthe normal

velocityis zero.At the surface,z = O, the (linearised)
boundaryconditions
are
p- g• (2.5)
and
w = •, (2.6)
whereg istheacceleration
duetogravity,and• isthesurface
displacement.
Using(2.3),(2.5)
and (2.6) togethergive
P•- /•:
8
Patz-O. (2.7)
Scalingz asH, theratioof theleft to therighthandsideof (2.7)is thesquareof theratio

of thephasespeedof internalwaves(NH) to thatof surface gravitywaves,(gH//2.ForH~20
m andN4).01s
'•, thesquared
ratiohasa valueof4 x 10'4,sothatthefighthandsideof (2.7)
canbe comfortablyneglectedto give the surfacecondition

Pt" 0 at z"O. (2.8)
This is, of course,the rigid lid conditionwhich is a statementthat, on the time-scaleof

internalwave propagationaroundthe bay, the surfacegravity-wavespeedis effectively
infinite.
N
0 km
I 2.0 I I
' ' ' ' I , , , , I , , , , i , , • , I , , , , i ,
Figure 8: Diagram of the rectangularmodelledbay, showingthe computationalgrid and the model

station locations.
A final boundaryconditionis requiredon theopenboundaryat themouthof thebay, where

the solutionmustbe statedin a way thatrepresentsthe (known)movementof the surfaceand
isotherms.The exactspecificationof this conditionwill be postponed
for the present.
We are consideringmotionforcedat the mouthwith periodsof the order of five to ten

days, For simplicity, we will restrictattentionto a single frequencyto, and assumea
time-dependence, for the variablesin (2.1) to (2.4), of the form exp(-itoO.With this
time-dependence, all time derivativesin (2.1) to (2.4) canbe replacedby the multiplicafive
factor-ira.It shouldbe notedthat m will be an orderof magnitudesmallerthanf.

Equations(2.1) to (2.4) can be expressed

as a singleequationin P,
(2.9)
in which•o2hasbeenneglected
by comparison
withif. Thesurface
boundary
condition
onP
is (2.8), and the bottomboundarycondition,by (2.3), is
P,•-Oonz'-H. (2.10)
On the walls,the zeronormal-flowconditions

are expressed
usingthe relationships
1 .io _
and
to
(-7-
P•P•)' (2.12)
By the hydrostatic
approximation,
implicitin (2.3), therelationship
betweenthe pressure
P and temperatureperturbations
T is
P•wg
otT , (2.13)
whereg is theaccelerationdueto gravityandotis thecoefficientof expansion of seawater.

The relationship
(2.13) assumes thatchanges in densityare duestrictlyto temperature,
a
reasonableassumptionfor JervisBay in the summer.
The nominalpositionof theBlackstationon therectangular

basin(Figure8) is x = X, y
= y•. Using (2.5) and (2.13), we can specifya pressurestructureat this location,basedon
observations of the sea-surfaceelevationandthe water-column temperature structure.
The
response to a specifiedsea-levelat Blackis immediately obvious.If • in (2.5) is givenat
x = X, y = y•, thenthe solutionsatisfying(2.9) andall theboundaryconditions is
P=g• , u-v=w=O
over the wholedomain.That is, becauseof the rigid-lidassumption,

the bay responds
uniformlyand instantaneously
to changesin sea-level.We will thus restrict attention,
henceforth,
onlyto thedynamics
induced
by changesin thetemperature
structure
atthemouth.
We can predictthe response to temperaturechanges with someaccuracysimplyby

inspection.In (2.11)and(2.12),the•o/ftermsaresmall.If theyareneglected,
thepressure
behaves asa streamfunction.
In particular
withthisassumption,thepressure
at anysetdepth
will be constantall thewayaroundtheboundary, takingthesamevalueasthatspecifiedat

the Black station.The pressurecontours,and thusthe streamlines, will form closedloops

insidethebay, with the maximumnumericalvalueon the wall, decayingawaytowardsthe
centreof the bay. If the verticalscaleis takenas H/;t (see (2.14) below)),the horizontal
exponential
decayscale,by (2.10),is NH/tf[m Underthisapproximation,
therewill be no
flow throughthe mouthof thebay,andthe motionwill be in phaseacrossthewholebay.
Neglectof theo)/ftermsin (2.11)and(2.12)is similarto therigid-lidapproximation in the

sensethatit assumesthat,on thetime-scaleof theprescribedoscillationsat themouth,internal
waveseffectivelytransmitinformationinstantaneously acrossthe bay. Retentionof the •o/f
termswill introducephasedelays.
If N is constant,the solutionto (2.9) will have the form
p=Q(x,y)cos
mtz
H '
(2.14)
wheren is an integer,representing
the modenumber.Taking n = 1, the equationfor Q is
fit :O'0 (2.15)
On an opencoastline,thesystemof equations developed thusfar describesinternalKelvin

wavesof the type treatedin detailby Gill and Clarke (1974). On a long straightcoastline
givenby, say,y = 0, the free-wavesolutionshavethe form
Q=exp(•:ikx
- •y) , (2.16)
where
k = (an (2.17)
NH
and
(2.18)
NH
The expressions
for k and l• are derivedusingthe requirementthat v, givenby (2.12), is
zeroonthecoast,andagainmakeuseof theassumption that•0:< <2e. Thewavesdescribed
by (2.16)to (2.18)havevelocities strictlyparallelto thecoast.Fortypicalvalues,•0= 10'5s'•,
[fl = 10"*s'l, N = 10-:s'l, H = 20 m, theoffshore length-scale1/Itis about1 km,whilethe
longshorewavelength,2•:/k, is about40 km. Accordingto (2.15), the formal pictureof the
density-drivendynamicsin JervisBay is that of a wave propagatinginto the bay alongthe
southcoastof the inlet channel(Figure8), clockwisearoundthe bay, andthenoutalongthe

northerncoastof the channel.The total solutionin the channelis givenby
Q-A exp(-ikx - •(y - Yl)) +'Bexp(ikx- •(Y2- Y)) , (2.19)
whereA is the (complex)amplitudeof the incomingwave,andB thatof theoutgoingwave.

Equation(2.19) which can be appliedat x = X, for y• _<y _<y:, in fact becomesthe final
boundaryconditionfor the problem.
The totalproblemis specifiedby the field equation(2.15) subjectto boundaryconditions
f O•+
Q•,
=0 (2.20)
on the walls at x = 0 and x = X, and
(2.21)
on the walls y = 0 and y = Y. The boundarycondition(2.19) is appliedas follows. The

amplitudeof the incomingwave is set so that A exp (ikx) = 1. Equation(2.19) is then
differentiatedwith respectto x, and the unknowncoefficientB is eliminatedbetweenthe
differentiatedand undifferentiatedversionsof (2.19) to give
O,,-ikQ- -•kexp
•(y•- y) (2.22)
on x = X, y• < y < y:. After solvingthe systemfor Q, the total solutioncanbe scaledto give
the correctamplitudeat the Black station.
The solutionto (2.15) was obtainedby discretisingthe equationusingstandardcentral

differences.The numericalgrid is shownon Figure8, with Ax = Ay = 500m. The boundary
conditionswereapproximated with one-sided,ratherthancentraldifferences,
becausecentral
differencesin the longshoredirection causeda 2Ax-wavelengthinstability around the
boundary.The full discretised
problemformsa matrixequationwith non-zeroright-handside
arising from (2.22). The matrix is amenableto inversionusing either direct or iterative
techniques.
2.5 Model results
As we havealreadynoted,JervisBaywasstronglystratifiedduringthemeasurement program.
Figure9 showstemperature profriescollectedbeforeand after the programat the central
stationshownin Figure 1. We will concentrate on the December5 profile.The average
buoyancyfrequency,calculated by fittinga straightline betweenthe surfaceand bottom

Temperature (øC)
I i 1
10- 5 Dec 1988
14-
18-
Jan 1989
22-
26
Figure9: Temperature
profiles
collected
in thecentralbayon5 December
1988and12January
1989.
temperatures,
is0.022s'•. Themaximum
value,calculated
forthetemperature
gradient
in the
centre of the water column is close to double this value.
The factor•:/NH in (2.15) is an eigenvaluederived,for the caseof constantN, from

solutionof the eigenvalueproblem
a ( • az)+•z..o
dz t•(z) dz '
with

dZ
Oil z-0,-H.
With N = 0.022 s4 andH = 20 m, theeigenvalue

is
•
•:=--=7.1m -Is.
NH
For theactualtemperature
profile(Figure9), theeigenvalue
(calculated
numerically)
is •:
= 4.6 m4s.Thisis thesameeigenvalueaswouldoccurfor a constantbuoyancyfrequency of
N= • = 0.034s-•
•:H
althoughthestructure of theeigenv•torsfor theconstant andvariableN caseswill obviously

be different.As a demonstrationof the sensitivityto density,modelresultswill be produced
for constantN values of both 0.022 and 0.034 s'1.
If P hastheverticalstructure
givenby (2.14) then,by (2.13), the temperature
at theBlack
stationwill be givenby
T._Tosianz
H
whereTois the amplitude,and the minussignis includedbecausez takesstrictlynegative

values.By (2.13), the pressureat the stationis then
p---H
•
g• To
cOS
•HZ
The amplitudeTOandfrequencyto be usedin themodelsrunsweredetermined asfollows.

The temperature time-series
fromtheBlackstationwaslinearlyaletrended,
thenleast-squares
fitted with a singleharmonicof the form
T-To cos(o
t-Sp), (2.23)
whereq)is the(constan0phase.The correlation

betweentheactualandharmonictime-series
wastested overa rangeof frequenciesat intervals
of 0.5 x 10'6s'l, andshowed
a clearbest
fit at o•= 9 x 10'6s'1,corresponding
toanoscillationperiodof 8 days.Thebest-fitamplitude
was To= 0.56øC.Figure 10 showsa comparisonbetweenthis harmonicand the detrended
temperature
signalfor the Black station.

A full list of the parametervaluesusedin the modelrunsis shownin Table2. The results
of themodelsimulation
withN = 0.034s'• areshownin Figures
11and12.Figure11shows
the (normalised) pressure
field bothin-phaseandout-of-phase with the temperatureforcing;
that is, at the time of maximumtemperature at Black, and quarterof a periodlater.For this
value of N, the offshoredecayscale(baroclinicradiusof deformation),l/g, is 2 km and the
longshorewavelengthis 136 km, approximatelythreetimesthe circumference of the basin.
Figure 12 showsthe surfacevelocity vectorssimultaneous with the pressurefields in
Figure 11. By (2.11), (2.12), and(2.14), thevelocityvariesverticallyascosrcz/H.
Thusbottom
velocities,for example,will havethe samemagnitudebut oppositedirectionto thoseshown
in Figure 12.
1.0 - //•lacktemperature •
0.0
-1.0 -
l I I I I I I I I I I I I I I I .I I I I I I I I I I I I I I I I I I I I I
6 10 15 20 25 30 1 5 10
DEC 88 JAN 89
Figure10: Comparison
betweenthe linearlydelxended
temperature
signalfrom Black,andthebest-fit
harmonic,
havingamplitude
0.56øCandfrequency
9 x 10-6s-•
Table 2: Parameters used in model simulations
Name Symbol Value
buoyancyfrequency N 0.034 s'•

0.022 s'•
Coriolisfrequency f _10- s-•
imposedfrequency 9x 10'• S'1
accn.due to gravity 9.8 ms'"
imposedtemp.amplitude 0.56øC
depth H 20m
bay width X 8x 103m
bay length Y 15 x 103 m
grid spacing Ax 500 m
mouth coordinates Y• 2.5 x 103 m
Ye 6.0x 103 m

The modelresultsall confirmanticipatedbehaviourin themodelbasin.The signalsfor both

temperatureand velocitydecayin magnitudeaway from the walls. There is a progressive
phasedelay clockwisearoundthe bay, with peak temperatures, for example,occurring
approximatelyone day later at Yellow, on the north side of the entrancechannel,than at
Black, where the signalis initiated.
0.800
(a) (b)
Figure11: Themodelled
(normalised)
surface
pressure
fieldfor N = 0.034s-l, (a) in-phase
and(b)
out-of-phase
with the specifiedtemperature
at Black.
Fortheweaker
density
structure,
N = 0.022s'l, thesolution
structure
is notdramatically
altered(Figure13). Becausethe horizontaldecayscaleis shorterin thiscasethen,for the
sameamplitude temperature
forcing,currentsnearto thecoastaremarginallystronger,
butthe
decayis morerapidtowardsthecentreof thebay.

., . ,,
4 ,
'k ß
', 'k
(a) (b)
5 m/s
Figure12:Modelled
surface
current
fieldsforN = 0.034s-1,(a)in-phase
and(b)out-of-phase
withthe
specifiedtemperature
at Black.
2.6 Discussion
To enablea quantitative
comparison
with themodelresults,themeasured
time-series
were
eachleast-squares
fittedwitha single
harmonic
having
a fixedfrequency
to- 9 x 10'6s'l, as
describedfor theBlacktemperature in theprevioussection.The time-series
analysedin this
way werethosefromstationdepthsshownin Figs5 and7, exceptthat,for theRedstation,
the 14.5m temperature(Fig. 6) andthe6.5 m currentwereused.The analysis determinesan
amplitudeandphasefor eachtime-series.

(a) (b)
5 m/s
Figure13: Modelledsurfacecurrentfieldsasin Fig. 12, butfor N = 0.022s4.
Results for both the model and measurementsare summarisedin Tables 3 and 4. Table 3
listsamplitudesof bothtemperaturesandcurrents. Thecunentamplitudes quotedarethose
in the directionof the semi-majoraxisof the currentellipse.The modelamplitudes are
maximaoverthewatercolumn,thatis, mid-watervaluesfor temperature andsurfacevalues
for currents.
Theagreement betweenthemodelandthemeasurements indicated
by Table3
is well within a factor of 2, exceptat the Red station.The measuredvelocitiesand,
particularly,
temperatures at Redsuggestlittle,if any,decayin signalstrength
withdistance
from the coastof the bay.

Table 3: Comparisonof modelledandmeasuredtemperature

and currentamplitudes
Station Temperature
(øC) Current(ms-1)
Model Measured Model Measured
Black 0.56 0.56

Red 0.16 0.67 0.003 0.015
Green 0.34 0.44 0.013 0.020
Blue 0.33 0.53 0.014 0.020
Yellow 0.40 0.66
Table 4 lists phasedelaysin the temperaturesignalbetweenadjacentstations,moving

clockwisearoundthe bay. There is goodagreementin the phasedelaysbetweenmostpairs
of stations,with the measurements confirmingthe model predictionof clockwisesignal
propagationaround the bay. As with the amplitudes,however, there is one fundamental
difference.The modelis formulatedsothatthetemperature oscillationentersthebayat Black,
propagates around,and exitsat Yellow. For thisreason,the Yellow to Black phasedelayin
Table 4 is negativeratherthanpositive.The measured signal,however,appearsto beginat the
Green stationand propagateprogressivelythroughBlue, Black and Red. In contrastto the
model,thereis no phasedifferencebetweenthe signalat Yellow andBlack,but thereinstead
appearsto be someform of signaldiscontinuity betweenRed and Green.Interestingly,if the
Greentime-serieshadbeenusedto calculatethe forcingamplitudeToin (2.23), the accuracy
of the modelledamplitudesin Table 3 would improveconsiderably.
Table 4: Comparisonof modelledandmeasuredtemperature

phasedelaybetweenadjacentstations
Stations Phasedelay (Days)
Modelled Measured
Black to Red 0.7 0.7

Red to Green 0.2 -2.3
Green to Blue 0.4 0.3
Blue to Yellow 0 0.7
Yellow to Black -1.2 0
In summary,then,the modelappearsto predictapproximately the correctamplitudeand

phaserelationshipin the bay, but the observedlack of decayaway from the coast,and the
apparentappearance anddisappearance of signalsin the north-western
regionareanomalous.
Given the simplicity of the model, however,the level of agreementwith the data is
encouraging.

An obviousshortcoming of the modelis the formulationof the forcingat the mouthof the
bay. As we found,the prescriptionof a temperatureoscillationat the nominalBlack station
(Figure 8) leads to an internal Kelvin wave propagatingclockwisearoundthe bay. For
consistency,to ensurethatenergydoesnot accumulate withinthe bay, it is thennecessary
to
allow the wave to exit, by specificationof the radiationcondition(2.22) at the mouth.This
formulationignoresall complicationassociatedwith matchingthe bay motionto the larger
scalemotionover the shell As just discussed, it also leadsto inconsistent phasebehaviour
across the mouth.
In termsof the behaviourat the mouth,an apparentlymore consistentmodelis the even

simplerone discussed briefly in section2.4. This modelignoresthe to/f termsin (2.11) and
(2.12), so that the bay boundary,includingthe mouth, becomesa streamline.While the
amplitudeof the modelleddynamics(Table 3) is not greatly affectedby this additional
approximation, the modelprovidesno informationaboutphase,sincetemperature oscillations
acrossthe bay are exactlyin phase.This is not a badapproximation,
sincethemeasuredphase
delay from Greento Red is only 1.7 days(Table 4), lessthana quarter-period.
Retentionof the to/f termsin (2.11) and (2.12), however,allowsfor internalKelvin waves,
and indicatesthat someaspects,at least,of thephasestructurein the bay are consistent with
wave propagation.Since,in reality,the wavedoesnot appearto propagatein andout through
the mouth,the model needs,at the next stage,to incorporatedissipationto ensurethat the
wave energyin the bay doesnot grow withoutbound.
Amongstotherobviousomissions from the model,realisticbathymetrywouldmodify the

structureand propagationratesof the wavesinsidethe bay. Direct surfacewind stressis
probablycapableof generatingsignificantcurrentsat thesefrequencies(Figure2). Further,
lateral densitygradients,probablycausedby surfaceheating,may significantlymodify the
dynamics.The temperature plot, Figure 5, indicatesa temperature rise of up to 2ø from the
Blackto the Greenstation.Giventheseomissions, themodelhasgivena usefulinsightinto
JervisBay'sdynamics.Furtherinvestigation is likely to requireconsiderablecomplication of
the model.
Acknowledgements
We wishto thank:GrahamSymondsandRick NunesVaz for theircontributions to thiswork;
Marc JeffreyandJohnMathiasfor theftassistance
with the field work;JohnAndrewarthafor
preparation
of thedata;JohnChurchfor hiscomments on themanuscript;and
CathyFlanagan
andNikki Pullenfor seeingthe manuscript
into typescript.
Thisprojectwaspartiallyfundedby theAustralian
Department of Defence,asa component
of a baselinestudysupervisedby the CSIRO Divisionof Fisheries.

2.7 References
Church,J. A., H. J. Freeland and R. L. Smith, 1986a: Coastal-trappedwaves on the east Australian
continentalshelf.Part I: propagationof modes.Journalof PhysicalOceanography,16, 1929-1943.
Church,J. A., N.J. White, A. J. Clarke,H. J. FreelandandR. L. Smith,1986b:Coastal-trappedwaves

on the eastAustraliancontinentalshelf.Part lI: modelverification.Journalof PhysicalOceanography,
16, 1945-1957.
Church,J. A. and H. J. Freeland, 1987: The energy sourcefor the coastal-trapped

waves in the
AustralianCoastalExperimentregion.Journalof PhysicalOceanography,17, 289-300.
Clarke, A. J., 1987: Origin of the coastallytrappedwavesobservedduring the AustralianCoastal

Experiment.Journalof PhysicalOceanography, 17, 1847-1859.
Craig, P. D., 1989: Water circulationin JervisBay. In JervisBay BaselineStudies:Third Progress

Report,December1989, CSIRO Divisionof Fisheries,305-328.
Freeland,H. J., F. M. Boland,J. A. Church,A. J. Clarke,A.M. G. Forbes,A. Huyer, R. L. Smith,R.

O. R. Y. Thompsonand N.J. White, 1986: The AustralianCoastalExperiment:a searchfor
coastal-trappedwaves.Journalof PhysicalOceanography, 16, 1230-1249.
Gill, A. E. and A. J. Clarke, 1974: Wind-inducedupwelling,coastalcurrentsand sea-levelchanges.

Deep-SeaResearch,21, 325-345.
Hamon,B. V., 1962:The spectrums

of meansealevel at Sydney,Coff's HarbourandLord Howe Island.
J. Geophys.
Res.67, 5147-5155(Correction,1963J. Geophys Res.68, 4635)
Hamon,B. V., 1966: Continentalshelfwavesandthe effectsof atmospheric

pressureandwind stress
on sealevel. J. Geophys.Res., 71, 2883-2893.
Holloway, P. E., G. Symonds,R. Nunes Vaz, M. Jeffrey and J. Mathias, 1989: Oceanographic
measurements in JervisBay, December1988 to January1989.Workingpaper1989/2,Departmentof
GeographyandOceanography, AustralianDefenceForceAcademy,44 pp.
Holloway,P. E., G. Symonds,

R. NunesVaz, andM.Jeffrey,1990:Temperature andsalinitysurveys
in
JervisBay andon theadjacentshelf,.February1988to April 1990.Workingpaper1990/1,Department
of GeographyandOceanography, AustralianDefenceForceAcademy,59 pp.
Huyer, A., R. L. Smith, P. J. Stabeno,J. A. Church and N.J. White, 1988: Currentsoff south-eastern
Australia:resultsfromtheAustralian
CoastalExperiment.
AustralianJournalofMarineandFreshwater
Research, 39, 245-288.

Middleton,J. F. andF. Viera: 1991: The forcingof low frequencymotionswithin BassStrait.J. Phys.
Oceanogr,21, 695-708.
Thompson,R. O. R. Y., 1983: Low-passffitersto suppress

inertialand tidal frequencies.
Journal of
PhysicalOceanography,13, 1077-1083.

Tidal Mixing near the Sill of a Scottish

Sea Loch
A J Elliott, P A Gillibrand and W R Turrell
Abstract
Currentand CTD measurements

were madeduringthe summerof 1989 near the inner sill of
Loch Sunart, a sea loch on the west coastof Scotland. The inner basin, which is around 7km
in length andcontainswaterdepthsof around90m, is separated
from the outerbasinsby a
shallow sill that reaches to within 10m of the surface. Current observations from the inner
basinshowa striking180øphasedifferencebetweenthesurfaceandnear-bottomtidalcurrents.
This has been interpretedin terms of the vertical modal structureof the internal fide.
Circulationandmixingin thelochis being modelledby thelaterally-averaged semi-implicit
estuarinemodelof Wang and Kravitz (1980) whichhasbeenmodifiedto allow an irregular
vertical grid spacingand to permit the sea surfaceto move verticallythroughthe grid.
Numericalexperiments haveshownthatthe baroclinicmotionis generated as a resultof the
sill beinglocatednearthe depthof the halocline. The modelresultssuggestthat an increase
in the near-surfacestratificationfollowing a run-off event may causethe internaltide to
strengthen
andgeneratean increased
shearacrossthehaloclinethusmaintaining
the vertical
mixing betweenthe surfaceand bottomwaters.
3.1 Introduction
Fish farmingin Scottishsealochshasexpandedrapidlyduringthepastdecade.The annual

productionof salmonhasincreasedfrom 500 tonnesin 1979 to 28,500 tonnesin 1989, with
a turnoverof œ100million. This expansion, however,hasraisedconcernover possible
ecologicaldamagecausedby the increasing quantifiesof wastematerialsproducedby the
farms. A significantproportionof the food dispensedsinksand settleson the seabed, and
the uneatenfood and materialexcretedby the fish may enhancethe levelsof nutrientsand
organicsubstancesin the watercolumnand sediments (Duff, 1988). Secondly,the useof
toxicchemicalsto treatthe cagedsalmonstockfor sealice hasbeencontroversial (Anon,
1989). A thirdproblemrelatesto bacterialinfectionof thecagedsalmoncausedby water
bornepathogens. Thesecan be carriedbetweenadjacentcagesby the tidal currentsand

36 A J Elliott, P A Gillibrand and W R Turrell
residualflow andthusspreadthe diseasefrom an infectedcageto its neighbours.
-[ 1i i i i i i i I I i I I I I I I [.•."L.I•I
IIIIIIIi
LOCH SUNART
5 10 15 20
r IL•i i • i i t iJ i i i i ! i I I I! III I ! I I'd I

1 I I I I Inner basin
Figure 1' Loch Sunartshowingthe entixesystem(top) and innerbasin(bottom),the locationof the

mooringis markedby a circle andbottomdepthsare shownin metres.
Suchproblemshavepromptedstudiesof the dynamicsandflushingcharacteristics of the

fjordic sea lochs on the west coast of Scotland,there being particularinterestin the
mechanisms that causetherenewalof the deepwaterof the innerbasinsand whichinfluence
the retentiontimesof pollutantswithin the system.One particularfjord selectedfor studyis
Loch Sunart(Figure1); a 35kinlonglochthatpossesses two major sills,the outersill being
at a depthof 30m while the innersill is only 10mbeneathmeansealevel. Maximumwater
depthsarearound120mand90m in themainandinnerbasins,respectively.A tidalamplitude
at springsof about2m at the inner sill generatesstrongtidal currentsand turbulencewhich
is evidentas a 'boiling' of the surfacewaterabovethe sill. A surfaceconvergence front,
similarto thosefoundin shallowerestuaries(Turrell and Simpson,1988), is observednear
the sill duringa springflood fide. Hydrographic surveysduringthe summermonthshave
shownstrongstratification in the innerbasin,with low salinitywaterin a surfacelayerabout
5m thick overlyingthe deepermoresalinewater. However,dissolvedoxygenvalueswere

A J Elliott, P A Gillibrand & W R Turrell 37
comparablein the deep water to thosefoundat the surfacewhich suggeststhat vigorous

verticalmixing takesplacethroughoutthe summer(DAFS, unpublisheddata).
In contrastto Sunart,otherlochsshowan intermittentrenewalof the deepwaterdue to the

effect of fresh water ran-off in controllingthe salinitynear the sill. For example,in Loch
Etive (Failwards and Edelsten,1977) renewalof the deepwater within the basinis prevented
at timeswhenthe densityof the waternearthe sill, which is controlledby ran-off, is lessthan
thatof thedeepwater. For manypracticalproblems,suchasthoseassociated with fish farms,
it would be useful if a criterioncould be developedby which sea lochscould be easily
classifiedinto thosefor which the deepwaterrenewaloccurseithercontinuouslythroughout
the year or elseintermittently.Sucha classification is likely to requireconsideration of the
work doneby the tidal streamsnearthe sill in relationto thepotentialenergyof the salinity
inducedstratificationwithinthebasin(EdwardsandSharples,1986). Detailsof thetidal flows
and stratificationlevels are unknownin many lochs and would need to be estimatedby
modelling. In suchcases,therefore,parameterization of the verticalmixing within the loch
may be more readily obtainedby extendingan existingtidal modelthan by resortingto an
argumentbasedon energyconsiderations.One goal of the Sunartstudywas to assessthe
ability of an estuarinemodel to reproduceobservedcurrentand salinityfeatureswithin the
inner basin of a sea loch.
3.2 Observational Results
For 6 weeksfrom July 8 to August24 (days 189-236)of 1989 a currentmetermooringwas

placedaboutlkm to the eastof the narrowsat the entranceto the innerbasin(Fig.l). The
sill is locatedat the westernendof thenarrowswherethedepthis lessthan 10mbeneathchart
datum.Waterdepthat themooringwasabout60m, thedeepestpartof thebasinbeinglocated
about lkm to the eastof the mooring. The mooringsupportedtwo Aanderaacurrentmeters,
oneat a depthof 5m andthe otherat 50m. A pressuregaugeto recordsealevel changeswas
deployednearthe bank of the loch to the southof the mooring. A bottommountedADCP
instrumentwasalsodeployedin the loch;thisinstrumentwasplacedto the southof the main
mooringandcollecteddataduringa 10 day periodof intensivemeasurements from August
15-24 (days227-236). A CTD surveyof the loch was madeduringAugust14-15 (days
226-227) and againduringAugust24-25 (days236-237).
Figure2 showsthe salinitydistributions

observedalongthecentreline of theloch,theinner
basinbeinglocatedbetween25-32kmfromthemouth.(Thetemperatures weresimilarduring
each of the surveys,with typicalvaluesof about 13.8ø(2. The horizontaltemperature
gradientswere weak and verticalcontrastswere generallylessthan0.5øCin the innerbasin,
the verticalstratification
beingdominatedby salinity.) Duringthe first surveythe surface
waterwithin the basinhad a valueof around20 psu,while valuesof around33 psu were
recordedat a depthof about20m. The stratificationwas weakerin the outerbasins,but still
reachedvaluesof 4 psuper 20m in the near-surface layers. Therewas heavyrainfall and

strongwindsdirectedinto the loch betweenthe two surveys,and Figure2b showsthe

resultingmarkedreductionin thesurfacesalinities
neartheheadof theloch. Surfacevalues
of around5 psuwererecorded in theinnerbasinduringthesecondsurveywhenthestrongest
haloclinewas at a depthof about5m. Temperature was morewell-mixedverticallyat the
time of the secondsurvey,howeverthis may be a resultof the reduceddatacoverage(the
CTD failedduringthe secondsurveyandthe resultsshownare due to bottledata).
I,
[+ + + + ----.-___ -' + :,• <-+,.+ 3• _•/:•-•--'.•-,-';•

•,.
o,,1 + + + + • + + ••++ -••

•/+ + + +•+ + • + + •;• ++++
/•••_ +• .... + + • + "•++++ •'•••g
+ >:+ ' .
.• -- •.;• .;• f + '

80 • •, •5 ' ' +•
• •.•.•+ .- •.
100
5 10 15 20 25 30
Distance from mouth (•)
Distance from mouth (kin)
Figure2: Observedsalinitydistributions
(psu):(a) August14-15(days226-227),Co)August24-25 (days
236-237).

The mooringdata have beenanalysedto extractinformationon the propagationof tidal

energyinto the innerbasin. The data serieswere also low passfilteredto removethe tides
and, althoughthe resultsare not presentedhere,it is appropriatethat somecommentsshould
be madeaboutthe non-tidalfluctuations.Sealevel wasrecordedby measuring pressurenear
the sea bed and then convertingthe pressuresignal to water height. Consequently,the
elevationsignalwasautomaticallycorrectedfor the inversebarometereffectand the non-tidal
elevationrecordshowedrelativelylittle energy. The dominantfluctuationswereof magnitude
0.2m and occurredover timescales of 2-5 days.Significantlymoreenergywas shownby the
low passcurrentat a depthof 5m. This containedpronounced energyat the 2-day timescale,
although
thefluctuations
hada magnitude of onlyabout0.05ms4. Thefluctuations
at 50m
wereevenweaker,reachingvaluesof only0.02ms'•, however
the2-daytimescale
wasagain
apparent.The long-termmeanflow wascloseto zero at bothinstrumentdepths. In contrast
to shallowestuarinesystemswhichcanbe dominatedby wind forcing,e.g. Chesapeake Bay
(Wang, 1979),therewas no obviouscoherence
betweenthe low frequencyvariabilityin the
sea level, near-surface and near-bottom currents.
Figure 3 showsexamplesof the mooringdata from the periodJuly 28 to August8 (days

210-220). The non-linearcharacterof the tidal currentsis apparent,and resultsin the near-
surfaceflood currentsbeing strongerand of shorterdurationthan the ebb. In contrast,the
near-bottomcurrentwasstrongerduringtheebbwhichnearthebottomwasof shorterduration
than the flood.
A
2-
vvvvv'vvvvvvvvvvv
212 214 216 218 220 212 214 216 218 220
?IHE IN DAYS ?IHE IN DAYS
15.00 B
14.75' ø
14 50'
14 25'
1400'
13 75'
13 50'
13 25' -30
212 214 216 218 220 212 214 216 218 220
TII•IE IN DAYS TI[VlE IN DAYS
Figure3: Examplesof thedatafromtheperiodJuly29- August8 (days210-220):(a) sealevelelevation,

(b) near-surface
(5m) and near-bottom
(50m) temperatures,
(c) eastcomponent of near-surface(5m)
current,(d) eastcomponentof near-bottom(50m) current.

Table 1 showsthe resultsof a tidal analysisof the elevationand currentdata. If the tide
has the characterof a progressivewave as it entersthe inner basinwe would expectthe
elevationand currentsto be in phase. However,Table 1 showsthat the surfacecurrentled
the elevationby a phaseof about 55ø at the M,. frequencywhile the S,. currentled the
elevationby nearly80ø.
Table 1: Resultsof a tidal analysisof the elevationandcurrentdata.Amplitudesare in cm (elevation)

andcm s4 (curten0,phases
arewithrespect
to theequilibrium
fideat Greenwich,
orientation
of the
ellipsemajor axis is in degreesanti-clockwise
from east.
Elevation
H G
M2 136.6 178.0
S2 54.8 232.5
M4 5.4 201.8
MS4 4.7 331.6
5m
East North Major Minor Orient. Phase

H G H G
M•. 9.1 124.5 1.9 130.0 9.3 0.2 11.8 124.7

S,• 4.5 155.4 1.0 172.9 4.6 0.3 12.3 156.2
lVln 1.3 236.7 0.5 233.5 1.4 -0.0 22.4 -123.8
MS4 1.4 284.3 0.5 280.0 1.5 -0.0 18.4 -76.2
50m
East North Major Minor Orient. Phase

H G H G
M,. 2.6 317.9 0.5 308.6 2.7 -0.1 11.2 -42.5

S2 1.5 276.9 0.3 286.0 1.5 0.1 12.4 -82.7
M4 1.6 63.4 0.4 55.7 1.6 -0.1 13.2 63.0
MSs 0.9 90.7 0.3 103.5 1.0 0.1 17.9 91.9
This suggests
thata significantamountof tidalenergywasbeingreflectedfromthehead
of the loch,causingthe tidal waveto havesomeof the characterof a standingwave. Both
surfaceandbottomM,. currentswererecfilinearanddirectedtowards11ø (with respectto
magneticeast)which impliesa true bearingof about23ø anti-clockwise from east. This
agreeswell with the orientationof thebasinaxis(Figure1). A strikingfeatureof the tidal

analysis
is thenearly180øphase
difference
between
theM•_currents
nearthesurface
and
bottom
(Table
1). Figure
4 shows
a short
portion
ofthedata,
taken
fromaperiod
ofspring
tides,thatillustrates
thephase
difference
in thecurrents
andthenon-linear
character
of the
currentsand temperaturesignal.
30
20'
214. 215 0 215.5 216.0

TIME IN DAYS
14.6
14.4
14
I • I • I •
214.5 215.0 215.5 216.0

TIME IN DAYS
Figure
4:Current
andtemperature
details
during
aperiod
ofspring
tides
(August
2-3):(a)5m(solid
curve)and50m(dashed
curve)currents,
(b)5mtemperature.

42 A J Elliott, P A Gillibrand and W R Turfell
Furtherevidenceis providedby Figure5 whichshowstheverticalstructureof thecurrents

recordedby the ADCP duringtwo tidalcycles;thesedatasuggest
thatthedirextionof the tidal
currentreversedbetweenthe depthsof 20-30m at the semi-diurnaltime scale.
0.'75 1.'00 1.'25 1.'50 1.'75 2. O0

Tidal Cycles
Figure5: Depth-time
contour
plotof theADCPcurrentprofile,ebbflow is shownby shading(thedata
showncoverthe periodfrom 1900on August18 (day230) to 2(X)0on August19 (day231)).
Table1 showsthattheshallowwaterconstituentsM4andMS4contributed significantlyto

thenon-linearcharacter
of thefide.Thequarter-diurnal
M4 wasparticularly
strongin relation
to theotherconstituents
at the50mlevelwhereit exceeded theamplitude
of S,.. As withM,.,
theM4 constituent showeda 180øphasedifference at thetwodepthlevels.
Theverticalmodalstructurecanbecalculated if it isassumedthatthemotionis oscillatory

in thehorizontal
withfrequency •0andhasanamplitude thatdepends onlyondepth,i.e. the
horizontaland verticalvelocitiescanbe writtenin the form u = U(z).cos(kx-o•0and w =
W(z).sin(kx<00where•0is theknowntidalradianfrequency andk is an unknownhorizontal
wavelength(Cushman-Roisin andSvendsen,1983). Determination
of theverticalmodesthen
requiresfindingthe solutionof
(3.1)

where N=N(z) is the Brunt-V/iis/il/ifrequency,subjectto the boundaryconditions

W(0)=W(-H)=0 whereH is thewaterdepthwhichis independent of x. The verticalstructure
of the amplitudeof the horizontalvelocity,U(z), canthenbe determinedfrom
U- 1 dW
ß (3.2)
k dz
If thesecond
derivative
in (3.1)is writtenin finitedifference
formthena matrixequation
for the valuesof W at eachdepth canbe obtainedin whichthe unknownwavenumber
scale
k is relatedto the eigenvalues of thematrix. A moreintuitivesolution
canbe foundby
writing(3.1)in finitedifference
form,setting
W(0)=0andthenmarching thesolution
forW(z)
from the surfaceto the bottom. If W(-H)=0 thena modalsolutionhasbeendetermined,
otherwisea new valuemustbe selected for k andtheprocedure
repeated.The methodcan
be madeto convergerapidlyby carefulchoiceof the parameterk, the numberof zero
crossingsof the solutiondeterminingthe orderof the mode.
VERTICAL MODES
20' 20'
40' 40'
50' $0'
60 6(
Figure6: Verticalprofileof sigma-t

at themooring
position
andthecomputed
verticalmodes.
Figure6 shows
a representative
profileof sigma-t
fromthemooring position
duringthe
survey
of August
14-15.Thiswasusedtocalculate
theN2(z)profilethatappears
in (3.1)with
to setto the frequencyof the M2 fide. The figurealsoshowsthe verticalstructure
of the
amplitude
of thehorizontalcurrent,
thesolidcurveshowing
thefirstmodeandthesecond
modebeingshown by thedashedcurve.Thefirstandsecond
modes hadhorizontal
length
scalesof27kinand12kin,respectively
(Table2). Bothmodes
show theobserved
180øphase
difference
betweenthedepths
of 5mand50m.It shouldbenotedthatthe27kmlengthscale
of the firstM2 modeis aboutfour timesthelengthof theinnerbasinandthata resonant
response
maytherefore
exist.Thecalculations
werealsomadeforthefrequency
of theM4
tidalconstituent
andtheresultsare again summarized
in Table2.

A J Elliott, P A Gillibrand and W R Turrell
Table 2: Resultsof an analysisof the verticalmodalstructure.The lengthscaleL is givenby L=2•

where k is the horizontal radian wave number.
Calculated
mode 1 mode 2
L (kin) 27.3 13.7 12.2 6.1
4.2 4.2 8.1 8.1
Observed
3.5 0.9
[u_,l
Althoughtheverticalshapes of themodesweresimilarat thehigherfrequency, thelength

scaleswerereducedto around14kmand6km for the first two modes. Althoughthe 6km
length scaleof the Dinsecondmodeagreeswell with the7km basinsize,the 8:1 ratioof the
surface:bottomcurrentspeeddoesnot agreewith the observed1:1 ratio. In contrast,the
computedfirst mode ratio of 4:1 for the M2 frequencyagreesreasonablywell with the
observedratioof 3.5:1. Bothcomputed modes(Figure6) predictstrongverticalshearin the
horizontalflow beneaththe haloclinein the depthrange5-15m,suggestingthatthe internal
tidecangenerateanenhanced shearbeneaththehaloclinewhichmayleadto increased vertical
mixing rates.
3.3 Modelling Results

The laterally-averaged
equations
canbe writtenin the form (WangandKravitz,1980):
Equationof continuity
u +(wa9-
0 (3.3)

A J Elliott, P A Gillibrand & W R Turfell 45
(3.4)
Momentumequation
(3.5)
+8Bigtl+ pdz
•- 0
cgx po Ox
Salt conservation
(3.6)
Equationof state
p - po(4•+ ps) (3.7)
wherex and z are the longitudinaland verticalcoordinates; u and w are the corresponding
velocitycomponents; B is the width;•1 is the surfaceelevation;H is the meanwaterdepth;
CDis thebottomfrictioncoefficient;K•, N• are thelongitudinaldiffusivityandviscosity;and
I•, Nzaretheverticaldiffusivityandviscosity;
P0is thedensityof freshwater,and[• (7.6 x
10'4)represents
thedependence of density
on thesalinity,s.
The dynamicalboundaryconditionsare givenby
Nz3u/3z= windstress
at thesurface
andN• 3u/3z= I u I at thebottom.Thecondition
thatthereshouldbe no verticalflux of saltat the surfaceandbottomis satisfiedprovidingthat
I• 3s/3z= 0 at z = •1 andz = -H.
At the headof the loch the fiver flow is specifiedand the salt flux is set to zero, while at
the mouththe sea level elevationis specified. When inflow occursat the mouth the ocean
salinity is specified,otherwisean advectivebalanceis assumed and horizontaldiffusionis
neglectedon outflow.

The setof equations(3.3)-(3.7) and the givenboundaryconditionsare standard.One of

the f'u-stsolutionswas obtainedby Hamilton (1975) who useda two time stepintegration
schemeand allowed the free surfaceto move verticallythroughthe grid. The model was
appliedto the RotterdamWaterway,andthento the Merseyestuary(BowdenandHamilton,
1975). A similar modelwasdevelopedby Blumberg(1977) in a studyof the circulationof
Chesapeake Bay. His modelusedtheleapfrogtimesteppingschemeandsolvedequations that
had beenintegratedwithin eachverticallayer. A furthermodelwas formulatedby Elliott
(1976) who did not integratethe equations withineachlayerbut evaluatedthe system(3.3)-
(3.7) at the grid points,and allowedthe estuaryto branchso that interactions within an
estuarinesystemcouldbe studied.Wang and Kravitz (1980) then developedan implicit
solutionto the equations(3.4) and (3.5), whichallowed themto reducethe computational
time by an orderof magnitude.An implicitsolutionof estuarinedynamicshadalsobeen
derivedby Hamilton(1977).
The codeof theWangandKravitz (1980) modelhasbeenappliedto Loch Sunartafterthe

followingmodifications:
(a) The verticalgridspacingcanvarywithdepth. Thisis an importantmodification

for loch
applicationwhichrequiresclosegrid spacingin the near-surface
layersand abovesills with
lessresolutionin the deeperwater of the basins.
(b) Thecodehasbeenmodifiedto allowthesurfaceto movethroughthecomputational grid

(Figure7). Thisfacilitywasincludedin themodelof Hamilton(1975)andis necessary
when
applyingmodelsto Europeanwaterswheretidal rangesare large. Withoutthismodification
the near-surface
verticalgrid spacingcannotbe lessthanthe tidalrange.
(c) The hybrid advectionschemeof James(1986) has been used to treat the horizontal
advectionof salt. A purelycenteredschemegenerated
negativesalinitiesin theregionsof
strongadvectionnear the head of the loch.
The WangandKravitz(1980) modelcouldnotbe runwith a timestepverymuchgreater

thantheoneusedin a purelyexplicitmodel.In initialtests,theexplicitcodeof Elliott (1976)
providedrealistictidalbehaviour
whenusedwith a timestepof 15 s whilethelongesttime
stepthatcould be usedwith theimplicitcodewasaround45 s. The rapiddepthchanges and
the stronglynon-linearcharacter
of the lochcirculationareobviouslyin contrastwith those
in shallow,dissipative
regionssuchastheChesapeake Bayin whichthenon-linear effectsare
lessimportant.
Threeformulations
for theverticaleddycoefficients
were tested(in cgsunits):
(i) Wang(1983)allowedtheturbulent
fluxesto dependondepthby usinga Munk-Anderson
parameterization:

A $ Elliott, P A Gillibrand& W R Turrell 47
•v=- •Vo+ •v,O+•o •3 -•t: (3.8)
Kz- Ko + K•(1+3.33R0-at2 (3.9)
inwhich
Ri isthelocal
Richardson
number
given
by
(3.10)
I u3 u3
i U3 US U3 US U3
I
U
l 1• UI.•S UIS UlS i
i
z--,o
ui
!
!
s ui s
-- I
I
I ui S ui S ui $ u• •..,•u • s
I U3 U3
I
I
U3 U3 US US •3 I
u• S u• s u• s u• s Ul S Ui S UI S
i
i
]./)//U3/ / /Us
.//./U$-=•
Us ui S ui ui
us
S ui
us
S
i
i
I
I
U3 I
u! s Ui Ui S Ul S
,
i
Figure
7:Grid
scheme
used
inthe
model,
the
surface
isallowed
tomove
vertically
through
the
grid
(ii)The
model
ofBowden
and
Hamilton
(1975)
wasunstable
whenthemixing
coefficients
were
depth
dependent
sothey
related
theeddy
coefficients
toanoverall
Richardson
number
defined
in terms
ofthetoptobottom
current
anddensity
differences.
However,
with
the
present
integration
scheme
their
formulation
remains
stable
when
theRichardson
number
is

48 A J Elliott,P A GillibrandandW R Turrell
made
depth
dependent,
consequently
thefollowing
expressions
were
tested
inwhich
Riwas
givenby (3.10):
N•nlul.(1+7
N,' No+ 10 R0-• (3.11)
aHlu
Kz' Ko+ 10 l.(1+R0_7 (3.12)
It should
benoted
thatthese
expressions
areempirical
andwerederived
fromdatacollected
in theMersey
Estuary
(Bowden
andGilligan,
1971).
(iii) Using
theresults
of KentandPritchard
(1959),
Blumberg
(1977)
represented
the
dependence
of K• ontheRichardson
number
as
(3.13)
where
Ricisacritical
value
oftheRichardson
numberabovewhichmixing
issuppressed.
By
assuming
thatthegradient
Richardson
number
(Ri)andthefluxRichardson
number
(Rf)are
relatedby
Rf. 1
Ri l+Ri
he derivedanexpression
for N,.of theform
N• = Kz(I+Ri)forO<Ri<Ri• (3.14)
andN=- K=- 0 whenR•• R/•
Figure
8 shows
theexternal
forcing
thatwasused
when
simulating
theobserved
salinity
section
shown
inFigure
2b. Themodel
wasspun-up
for3 tidalcycles
byspecifying
and
holding
constant
thesalinity
observed
onAugust
14-15
(Figure
2a)while
forcing
with
thetidal
elevation
atthemouth,windstress
andrun-off.Theintegration
wasthenstepped
forward
for21tidalcycles,
over
theperiod
August
14-25,
withthesalinity
distribution
beingupdated
dynamically
using
(3.6).Thefinaltidal
cycle
wasthen
averaged
toproduce
thedistributions
showninFigure9. Following
thestrong
run-off
eventonAugust23theobservations
showed
averythin(less
thanlmthick)
surface
layerwithasalinity
ofabout
5psu.Thislayer
could
notberesolvedbythe2mgridspacingused
inthemodel;thelowest
salinities
predicted
by

49
A J Elliott,P A Gillibrand& W R Tunell
themodel
inthesurface
layeroftheinner
basin
wereoftheorder
of 10psu.
August 1989
14 15 16 17 18 19 20 21 22 23 24 25
-1
-2
20
10
-10
-20
280
240
200
160
120
80
40
22l' 2.10' 2.12' 2.14' 2.16'

DAYS
Figure
8: External
forcing
used
todrive
themodel
when
simulating
thesalinity
distribution
observed
on August24-25.
TheMunk-Anderson
formulae
givenby (3.8)-(3.9)
produced
goodagreement
withthe
observations
(Figure
9a),bothinthenear-surface
layers
oftheinner
basin
andinthedeep
waterof themainbasin.However,
thelowvalues
of thecoefficients
thatweren•essaryto
reproduce
thesharp
near-surface
gradients
intheinner
basin
resulted
inreduced
mixing
at
mid-depth
inthemiddle
basin.
Forexample,
the33psu
contour
wasreproduced
atadepth
of
about
15mwhich canbecomparedtotheobserved
valueofabout25m(Figure
2b).Forlarge
values
ofRi,theMunk-Anderson
formulaehave
theeff•t ofreducing
theeddycoefficients
tovalues
lessthanthose
givenbytheothermethods.Thebestsimulations
wereobtained

usingNo= Ko- 0.05, N• -- 35.0 andK• -- 20.0.
20-
- - i.•]:5.!'i•...:
' "-:--33.5----!,. •:• •
40'
60"
80'
100-
5 10 15 20 25 30
---• Distance from mouth (km)
0.4 m/s
• 4o . •'" t ......
'' - 'm
!:•i....
- •.'.-i"•-;'.½-tl.' '"• - -
1oo' 5 10 15 20
;--•--.'
25 30
Distance from mouth (km)
0.4 •/s
Figure9: Salinitysimulations
for different
mixingparameterizations:
(a) Wang(1983), (b) Bowden
andHamilton(1975), (c) Blumberg(1977).

A J Elliott, P A Gillibrand& W R Turfell 51
The formulae (3.11)-(3.12), adaptedfrom Bowden and Hamilton (1975), allow the
maximumvaluesof the mixingcoefficients to dependon waterdepthandthe instantaneous
currentspeed.The eddydiffusivity,Kz,givenby (3.11)-(3.12)for a chosen Ri, is higherthan
the valueobtainedusing(3.8)-(3.9) due to the strongtidal currentswithinthe loch. In the
innerbasinthishasthe effectof reducingthe salinityvaluesin the deepwater,althoughthe
overallagreementwiththeobserved salinitiesis good(Figure9b). Thebestsimulations were
obtainedusingNo = Ko = 0.05, N• = 0.20 and K• = 0.10.
Realistic salinitiesin the inner basin could not be reproducedwhen using expressions
(3.13)-(3.14) due to Blumberg (1977). Although depth resolvingmodels are usually
insensitive
to thevalueschosen
forK• andN• provided
theyarelessthan10s - 10• cm2 s-1,
themodelwouldnotremainstableunlesstheywerebothsetto 10 cm2 s4 when(3.14)was
used. The schemewas also unstableunlessthe criticalRichardsonnumber,Rio was set to a
valueof 5, eventhoughBlumberg(1977) suggests a valueof around10 for the upperlimit.
With theserestrictions,the near-surfacesalinitiesin the inner basin reacheda minimum value
of the order of 20 psu and the observed strongnear-surfacestratificationcould not be
reproduced (Figure9c). This suggests thatthe formulationgivenby (3.14), whichproduced
realistic salinitiesin the weekly stratifiedwaters of the ChesapeakeBay, may not be
appropriatein highly stratifiedfjord regions. However,in contrastto the surfacewaters,
reasonablesalinityvalueswere computedfor the deepwater of the outer basins. The best
simulationswere obtainedusingKo = 0.05, k• = 0.035 and Ri• = 2.
Initial testsaimedat reproducing

theverticalphaselag observedin thecurrentsshowedthat
near-surface stratification and the shallow sill at the entrance to the inner basin must both be
includedin the simulations.Changingthe lengthscaleof the inner basin,while holdingthe
sill depthand stratificationat chosenlevels,did not influencethe verticalphasedifference.
This supportsthe conclusionmade earlier that the internal tide does not result from a
resonanceof the secondmodeof the M4 constituentwhich has a lengthscalecomparableto
thelengthof theinnerbasin,andsuggests thatresonance doesnotplay a part in the baroclinic
M2 response.
Figure 10 showsthe resultsfrom a seriesof testsin which the sill at the entranceto the
innerbasinwassetat differentdepthsduringhighrun-offconditionsthatgeneratesignificant
near-surface stratification.To maketheseteststhefiver flow wassetat 65 m3 s4, a realistic
initialsalinitydistribution
wasassigned to thegridpointsandthentheoceansalinitywasheld
fixed while the model was spun-upfor 40 tidal cycleswithouttidal forcing- and withouta
sill at the entrance to the inner basin. This allowed the internal circulation to establish and
for realisticsalinitystratificationto develop. At the endof the 40 tidal cycles,by which time
thesolutionwasapproaching steadystate,the tidal forcingwasappliedat themouth,a sill was
inserted at the entrance to the inner basin and the simulation was continued for a further 7
tidal cycles,the last two of which were usedto generatethe figures. (The resultsof the 40
tidal cyclesspin-upwas storedon disk so that the final 7 tidal cyclesof simulationcould
continuewithoutduplicatingthe initial calculations.)With the sill at 52m (Figure 10a) the
simulatedtidal cunenmat the mooringpositionare fairly barotropicwith maximumvaluesat

springtidesbeingreduced
to around10cms". Thesurface
flowebbsthroughout
thetidal
cycle becausethe tidal currentsduringflood are weakerthan the river inducedsurface
velocity. Raisingthe sill to 20m and 14m(Figure10b,c)createsa jet centeredat 10mon the
flood tide and introducesthe observedverticalphasedifference. In Figure 10d are shown
resultsfrom the simulationwith the sill at its actualdepthof 8m, and the strongverticalphase
signalis in goodqualitativeagreement
with theADCP velocitiesshownin Figure5.
HW LW HW LW LW
A '6 ' 1............................

:--a''-::;•::--"::• .1 ..............
-::•½•-:.:L-::•:•:•½•:::::
:::
.:.:.:..:::::::-
•:
•:•::::-(•:::::':"
"::'"
...........
:'6'
,&....:.:...•
' ........
-'"':
•:-!•..........................
..... ;i•
•:•:•:•:• •i:::
•-•::-.•:
.......
•1•:•:•:•:•:•:-:.::.:.:..-:;.:.:.:...::•;½;::-:-'•½:.-:
........
-•-
50 0.5 1.0 ß 2.0 0.5 1.0 1.5

,
2.0
•
Tidal Cycles Tidal Cycles

HW L,W HW LW
D--'•'
:"
•'•;.•:•i•"-'
'::•
•. ':'•'•'-•
•-'•.!-•
........
•'•:•;
1
..
10
*'•:•i•-.'•.
.... • - ''•"•'"
'•:':
"•'
•2o .20
30
4O
50' , , , i , , ,
0.5 1.0 1.5 2.0 0.5 1.0 1.5 2.0
Figure 10: The influenceof sill depthon the tidalcurrentsin the innerbasinduringa periodof high
run-off: (a) 52m, (b) 20m, (c) 14m, (d) 8m.
A further seriesof testswas made to investigatethe shearinducedmixing due to the

internalfide. Figure 11 showsthe tidal currentsat the mooringpositionsimulatedduringa
period
oflowrun-off(10mas'•) whichresults
in modest
near-surface
stratification.
Although
the vertical shear is weak (Figure l lb), well-mixed water in the depth range 10-20m is
advectedinto the basinnearthe end of the floodtide (Figure11c)whichresultsin the
predictedRichardson
numberreachinga valueof lessthan 1 (Fig.1ld). In contrast,highrun-
off (65 ma s") leadsto significant
shearin the simulated
currents
(Figure12a).

A J Elliott, P A Gillibrand& W R Turrell 53
HW LW HW LW • L•W I•W L,W
t -o.••m• --- o.•.•.•?;• •

•o•
• o.•o.,
• •2 ••;/I
0.2•) I
L
• 02 0.2•
•0•
t
s0•
0.5 1.0 1.5 2.0 0.5 1.0 1.5 2
Tidal Cycles T&•aZ CycZe•
l-'-• •••.; .•"'-•'-t"• • •

• o
•'- / 6.•-o.• • o.• • .•
• '' Y•o.• -o.•- •o.••o.•

50
0.5 Z.O • .5 2.0 0.5 1.0 1.5 2.0
T•daZ CyaZes Tidal Cycles
Figure11' Predictedcurrents,shear,stratification
andRichardson
numberduringa periodof low run-off:
(a) velocity,(b) 104.(au/az)
2,(c) -g/p(ap/az).104,
(d) Richardson
number.
(The currentsshownin Figure 12aare slightlydifferentto thoseshownin Figure 10d because

the distributionsin Figure 12 were obtainedafterallowingthe modelto spin-upfor the entire
45 tidal cycleswith the sill in positionat a depthof 8m.) Figure 12b showsthat significant
verticalshearis generatedin thedepthrange10-15mat themiddleof theflood. Althoughthe
salinitystratificationis stronger(Figure 12c) thanduringlow run-offconditions,theincreased
sheargenerates Richardsonnumbersthatareaslow as 3 in thedepthintervalbetween15-25m.
Thusalthoughthe verticalstabilityat 15m is increased by a factorof at least20 (Figures1lc
and 12c), the minimum Richardsonnumberonly changesby a factor of about 3 due to the
enhancedsheargeneratedby the baroclinicmotion.
3.4 Discussion
The observational datacollectedat the mooringpositionwere characterized by the strongly

non-linearnatureof thetidalcurrentsandtemperature fluctuations.The moststrikingfeature
of the semi-diurnalcurrentswas the 180ø phasedifferencebetweenthe surfaceand bottom.
An analysisof the verticalmodes,basedon the observeddensityprofile, showedthatboth the

first andsecondmodebaroclinicmotionscouldproducetheobserved phasedifference.The

shallow water M4 tidal constituentwas strongnear the bottom,being of comparable
magnitudeto the M• and S• constituents,
and the lengthscaleof the computedsecondmode
at the M4 frequencywas comparable to the scaleof the innerbasin. However,it is thought
unlikely that a resonancewas occurringsincethe observedratio of the surfaceto bottom
currentspeedwas closerto thatpredictedfor the semi-diurnaltime scale. Inspectionof the
dataseries(Figure4) suggeststhatthe verticalshearwasgenerated by the semi-diurnalfide.
HW LW HW
40
50
0.5 1.0 5 2.0 0.5 1.0 1.5 2.0

H• L• H• L• • L• •W L•
c
lO
•20
• •o
40
50•o 2•
. • , , , ---0.2,--. •n , ,• , , 50
0 5 1.0 1.5 2 0 0.5 1.0 1.5 2.0
Figure12: Predicted
currents,
shear,stratification
andRichardson
numberduringa periodof highrun-
off: (a) velocity,(b) 104.(3u/3z)
2,(c) -g/p(3p/3z).104,
(d)Richardson
number.
A seriesof numericalteslshave shownthat the baroclinicmodeis generated by the

barotropic tideforcingat the sill of theinnerbasin,thedepth of which is locatedcloseto
the strongnear-surface halocline.The verticalphasedifferenceis onlypredictedat timesof
highrun-offwhenthe near-surface stratification
is well developed.This suggests thatthe
internalmodemayact to maintainverticalmixingwithintheinnerbasinby increasing the
shearbeneaththehaloclineat timesof increased verticalstability.Numericaltestsmadefor
low andhighrun-offconditionsshowed that,although theverticalstabilitymayincrease by
a factorof 20 at timesof highfiverflow, theRichardson numberonlyincreases by a factor
of 3 due to the verticalsheargenerated beneaththe haloclineby thebaroclinicmotion.

A J Elliott,P A Gillibrand& W R Turrell 55
Thehighlynon-linear character
of thelochdynamics limitsthesizeof thetimestepthat
canbe usedwith theimplicitmodel. Thisis in contrast to theshallow,dissipativeregions
suchas theChesapeake Bay, wherethe modelhadbeenpreviously applied.The mixing
parameterizations of BowdenandHamilton(1975)andBlumberg (1977),whichhadbeen
developed for applicationto the MerseyandPotomac estuaries,
respectively,werenot
successful in simulatingthe strongnear-surface
stratification
observed in the loch. This
suggests thatthemodelcouldbeusefullyextended through theuseof a closure scheme for
theprediction of theturbulent mixing(SmithandTakhar,1981). A Richardson number
probe,consisting of twoS4 currentmetersseparated
by a CTD wasprofiledverticallynear
thesillduringthefieldworkandthedataarenowbeinganalysed to deriveestimates of the
verticalshearandRichardson numberthroughout
thetidal cycle. Thisshouldleadto a better
understanding of theshearinducedmixingandenableits effectiveparameterization in future
modellingstudies.
Acknowledgements
We thankourcolleagues
in theHydrography
Sectionof theMarineLaboratory,
Aberdeen,
in
particular
R D AdamsandR Payne.We alsothankthecaptainandcrewof theRV SeolMara
for their help duringthe field work.
3.5 References
Anon, 1989: Call for pesticidecontrols.Mar. Poll. Bull., 20, 102.
Blumberg,
A.F., 1977:Numerical
modelof estuarine
circulation.
J.Hydraul.Div. ASCE,103,295-310.
Bowden,K.F. andR.M.Gilligan,1971: Characteristic

featuresof estuadne
circulation
asrepresented
in
the Merseyestuary.Limn. Oceanogr.,16, 490-502.
Bowden,K.F., andP. Hamilton,1975: Someexperiments with a numericalmodelof circulationand

mixingin a tidal estuary.Estuar.Coast.Mar. Sci., 3, 281-301.
Cushman-Roisin,
B., andH. Svendsen,1983: Internalgravitywavesin sill fjords:verticalmodes,ray
theoryand comparison
with observations.CoastalOceanography,
H.G. Gade,A. EdwardsandH.
Svendsen,Eds., Plenum Press, 373-396.
Duff, A., 1988: Tighteningcontrolson fish-fanning.Mar. Poll. Bull., 19, 300.
Edwards,A., andD. J. Edelsten,1977: Deepwaterrenewalof LochEtive: a threebasinScottishfjord.

Estuar. Coast. Mar. Sci.,5, 575-593.

Edwards,A., andF. Sharples,1986: Scottishsealochs:a catalogue.

ScottishMarineBiological
Association/Nature
Conservancy
Council,110 pp.
Elliott,A.J., 1976: A numericalmodelof the internalcirculation

in a branching
tidal estuary.
Chesapeake BayInstitute,Spec.Rep.54,TheJohnsHopkinsUniversity,85 pp.
Hamilton,
P., 1975: A numerical
modelof theverticalcirculation
of tidalestuaries
andits application
to theRotterdamwaterway.Geophys.J. Roy. Astron.Soc.,40, 1-21.
Hamilton,P., 1977:Onthenumericalformulation
of a time-dependent
multi-level
modelof anestuary,
withparticularreference
toboundary
conditions.
Estuarine Processes
II, M. Wiley,Ed.,Academic
Press,
347-364.
James,IX)., 1986: A front-resolving

sigmacoordinate
seamodelwith a simplehybridadvection
scheme.AppliedMath. Modelling,10, 87-92.
Kent,R.E.,andD.W. Pritchard,
1959: A testof mixinglengththeories
in a coastal
planeestuary.J.
Mar. Res., 8, 67-72.
Smith,TJ., andH.S. Takhar,1981: A mathematical

modelforpartiallymixedestuaries
usingthe
turbulenceenergyequation. Estuar.,Coast.and Shelf Sci., 13, 27-45.
Turrell,W.R., andJ.H. Simpson,

1988: The measurement andmodellingof axialconvergence
in
shallowwell-mixedestuaries.
Physical
Processes
in Estuaries,
J. Dronkers
andW. Van-Leussen,
Eds.,
SpringerVerlag, 130-145.
Wang,D.-P.,1979:Wind-driven
circulation
in theChesapeake
Bay,winter1975.J.Phys.Oceanogr.,
9, 564-572.
Wang,D.-P.,1983:Two-dimensional
branching
saltintrusion
model.J.Waterways,
Port,Coastal
and
OceanEng. ASCE, 109, 103-112.
Wang,D.-P.,andD.W. Kravitz,1980:A semi-implicit,

two-dimensional
modelof estuarine
circulation.
J. Phys.Oceanogr.,10, 441-454.

A topographicallyinduced internal wave

and mixing in the Tamar Estuary
D R Sturley and K R Dyer
Abstract
The stratifiedtidal flow overa depression in the TamarEstuaryhasbeenexaminedon a neap

tide. An internalwave formedfrom thisinteractionin the mannerdescribedby Maxworthy
(1979). CTD and velocitymeasurements from a fixed station,andobservations
madeusing
a 200 kHz echosounder and the EstuarineThermistorSpar,have revealedf'medetailsin the
wave structure. There is mostly excellentcorrespondence betweenregionsof acoustic
backscatter andtemperature gradients,
or fine structure.An analysisof thetemporalevolution
of the wave heightshowsthe formationmechanismto be as Maxworthy(1979) and Lee &
Beardsley(1974) proposed,and hashighlightedthe importanceof nonlinearsteepening as an
evolutionaryprocess.
The onsetof vigorousverticalmixing in the seawardend of the wave, after the time of
maximum barotropiccurrent,has indicatedthat this hydraulicphenomenonmay be an
importantcontributorto overallestuarinemixing. Theseobservations are the first of theft
kind for a smallpartiallystratifiedestuary,andmay be evidenceof an importanttransitional
phenomenon in estuarinemixing.
4.1 Introduction
Internalwavesproducedby the interactionof stratifiedtidal flowswith topography, over the

continentalshelf-breakand in fjords, have long been consideredimportantintermittent
contributors
to verticalmixing and circulation(e.g. Huthnance,1989). Their importancein
estuaries,
especiallysmallpartiallystratifiedones,hasso far beenleft relativelyunexplored.
Intensemixingevents,whichhavebeenbelievedto be causedby hydraulicjumpsandinternal
waves,havebeenreportedby Gardner& Smith(1978), Partch& Smith(1978), Partch(1981)
and New et al. (1986; 1987). Direct observationsof the interactionof stratifiedtidal flows
with topography in estuarieshavebeenrare (e.g. Gardner,1984; Dyer & New, 1986; New
& Dyer, 1987; Pietrzak, 1990). Partch& Smith (1978) have shownthat in the Duwamish

58 D R Sturleyand K R Dyer
Estuaryobservedintense
mixingperiods,
presumably produced by internalwaves,occupied
only20% of thetidalcycle,butaccounted
for 50% of theverticalturbulent
saltflux.
So far accurateestimatesin estuariesof the efficiencyof conversionof barotropictidal

energytointernalwaveenergy,andthentoincreased
potential
energythrough
verticalmixing
hasbeenlacking. The preliminaryresultsdescribed
haveshownhowthe formation of an
internalwave occurson an ebbingtide in the TamarEstuary,and how subsequent
vertical
mixing ensues.
New & Dyer (1987)identifieda depression in thebedat Cargreen, in theTamarEstuary,

asa promising sitefor theproduction
of internalwaves. Figure1 showsthebathymetry of
thebedat Cargreen.The depression is anelongated ellipticalgroove,whichis centralto the
mainchannel. A surveywasconducted on 6 November1989,on theebbstreamof a neap
tide (tidalrange2.6m),aftera shortperiodof intenserainfall(andriverflow),followinga
prolonged periodof drought. Thesurveyincluded CTD measurements at a fixedstationand
observations
taken on a longitudinaltmnsectof the estuaryusingan echosounder
and the
EsmarineThermistorSpar, (Sturley& Dyer, 1990). It was thoughtthat theseconditions
would be most conducive to wave formation.
4.2 Survey
The timeof predictedhighwaterat Cargreen, on 6 November1989,was 10 a.m. A launch
was mooredat Station"45" (seeFigure 1) from an hourand twentyminutesafter predicted
high waterHW+l'20 to HW+5:40, takingverticalprofilesof salinity,temperature, current
velocityanddirection,at twentyminuteintervals. Theseprofilesarepresented in Figures2
to 6. The temperature (T) and salinity(S) measurementswere madewith an MC-5 in-situ
T-S bridge(accurateto +_0.1 salinityunitsand_+0.1øC). The currentvelocityanddirection
were measured with an NBA (DNC-3) impellercurrentmeter(accurateto + 2% in velocity,
and+ 10ø in direction). The depthwascalculated from wire lengthout andangle,with a
typicalerrorof 5% to 10%.
FromHW+1:20 to HW+6:45 a secondlaunch(RV Catfish)madelongitudinaltransects over

thebeddepression with a 200 kHz echosounder andtheEstuarineThermistorSpardeployed.
The latterinstrumentis capableof fine temperature
measurements(resolutionapprox0.005øC,
accuracy + 0.1øC)at a verticalspacingof 0.25m to a depthof 4m, at a rateof 10Hz (see
Sturley& Dyer (1990) for precisedetailsof calibration,
designspecificationandlimitations
on use). SelectedSpardataand echogramsare presentedin Figures7 to 12. The Eulerian
measurements of temperatureand salinity at "45" were usedto infer salinity,and hence
density,from the Spardata,usingthe Millero-Poissonequation(Pond& Pickard,1983).

D R Sturleyand K R Dyer 59
RIVER TAMAR. CARGREEN AREA

GN
METRES
TQ--
FIX
MARKS
Figure1: The depressionin theTamarEstuarybedat Cargreen.The Tamarformsa naturalboundary
betweenDevonandCornwall. It is nominallypartiallymixedandhasa meanneaptidalrangeof 2.3m.
Cargreenis 20km fi'om the tidal limit and 10km from the mouthof the estuary. The isobathsare
plottedevery lm referencedto chartdatum. Station"45" was wherethe Eulerianmeasurements
were
taken.
4.3 Eulerian Measurements
The salinity(Figure2) andtemperature

(Figure3) profilesat "45"showthattheestuarywas
stratified
intotwo layers. Theywereseparated
by a regionof stronggradient
in temperature
and salinity. This gradientweakenedtowardthe endof the exerciseas the thermocline,and
halocline,thickened vertically,withlesssalinewaterbeingincorporated
intothethermocline,
by entrainment from theupperlayerin the seawardendof thewave. However,the difference
in density
between
thetwolayersremained
at about16kgm'athroughout
thestudyperiod,
despiteeachlayerbecoming marginallylessdense,astheebbtideprogressed.T andS were
verywellcorrelated in theEulerianmeasurements
andverticalprofries of T, S anddensityp
(notshown)arevirtuallyidenticalin shape.Figures2 and3 showthattheverticalmigration
of thethermocline, whichwasloweredandthenraisexl, in association
with thegrowthand
evolutionof the internalwave,greatlyinfluenced the T andS profiles. The positionof

60 D R SturleyandK R Dyer
station"45" wasselected
to represent
the "upstream"
conditions
of theflow but,asFigures
2 to 5 show, the internal wave distorted the level of the thermocline at this location. The
velocityprofilesshowa similar,butmoresubtleeffect. Figure6 showsthedepthaveraged
currentspeed,•, at "45". It depictsan ebbstreamwith a doublepeakin velocity. This
behaviouris commonin themiddleandupperTamar(George,1975).
o - 30 ppt
om
lm
2m
3m
4m
11:20 11=40 12:00 12=20 12:40 13:00 13:20 13:40 14:00 14:20 14:40 15:00 15:20 15:40
Time in
houre (GIRT)
Figure2: The verticalprofilesof salinityat "45"on 6 November1989,madeevery20 minutes, as

indicatedonthehorizontal timeaxis. Thedepthis in metres,andthescalefor salinityis indicated
in
salinityunits,(ppton theT-S bridge). The timeof predicted
highwaterwas10:00(GMT).
10 - 14'c
Om
D•pth
in
mettee
11:20 11:40 12:00 12:20 12:40 13:00 13:20 13:40 14:00 14:20 14:40 15:00 15:20 15:40
Time in
houre (GMT)
Figure3. Thevertical profiles

of temperatureat "45"on6 November 1989,madeevery20 minutes,
asindicated on thehorizontal
ti•neaxis. Thedepthis in metres, andthescalefor temperature
is
indicated
in thetoplefthandcorner in celsius.Thetimeof predicted
At thecentreof thedepression
theorientationof themainchannelwasestimated
to be 142ø
fromChartNorth.Thevelocity measuredat"45"wasdividedintovector
components, along
the orientation
of the mainchannelat "45"(longitudinal
velocity-positive
velocitiesare
seaward in Figure4), andin the cross-channel
direction
(transversevelocity- positive
velocities
aretowardtheCornishbankin Figure5). Figure4 indicates
maximum seaward
flow at mid-depthat, or slightlyabove,the level of the thermocline. The level of this
maximum flowevolvedin timein exactlythesamewayasshown by theT andS profiles.
Figure5 showsa strong
transverse
component oftheflowin theupper
layerswithnegligible
flow below the thermocline. The level at which the transverse
flow becamezero coincided
with thebottomof the thermocline,
andevolvedwith it in time. The transverse
flow reversed

D R SturleyandK R Dyer 61
towardsthe end of the tide. Figures4 and 5 togethershow a stronglyvariable flow

associated
with the growthand evolutionof the lee wave.
o - I m/s
om
I
Depth
in
metres
.......
. . .
11120 11140 12:00 12120 12140 13100 13120 13:40 14100 14=20 14140 15100 15:20 15:40•
Time in
hours (GNU)
Figure4: The verticalprofilesof longitudinalvelocityat "45" on 6 November1989, madeevery 20

minutes,as indicatedon the horizontaltime axis. The depthis in metres,andthe scalefor velocityis
indicatedin thetop left handcomer,in metresper second. The timeof predictedhighwaterwas 10100
(GMT). Positive velocities are seaward.
Calculationsof the longitudinalgradientRichardsonnumber,Ri, due to verticalshearsin

the longitudinalvelocityu, where
g tip 6u -2
showthat the thermoclineat "45" wasstable(Ri > 0.25) until 13120,despitelargeshears,but

Ri is closeto 0.25 after thistime. (The verticalsamplingintervalfor p and u was0.5m).
Usingthefull velocityvector0 = u + v in Ri mightmaketheseestimates evenlower,showing
ample opportunityfor sheardriven instabilitiesacrossthe thermocline.
o - 1 m/s
Om
Depth
in
metres
11:20 11:40 12:00 12:20 12:40 13:00 13:20 13:40 14:00 14:20 14:40 15:00 15:20 15:40
Time in
hours (GNT)
Figure5: The verticalprofilesof transverse

velocityat "45"on 6 November1989,madeevery20
minutes,asindicatedon thehorizontaltimeaxis. Thedepthis in metres,andthescalefor velocityis
indicated
in thetopleft handcomer,in meterspersecond.Thetimeof predicted highwaterwas10100
(GMT). Positivevelocities are toward the Cornishbank.

62
D R SturleyandK R Dyer
Depth
averaged
velocity
in m/s
0'6
0'5-
0'4 -
0'3-
0'2-
O' 1 -
0'0 ' I I I I 1
11 12 13 14 15 16 17
Time in hours (GMT)
Figure6: Themagnitude
of thedepthaveraged
velocity,
fi, at "45"on6 November
1989. Thetime
of predicted
4.4 Echograms
andSparData
TheSpardataandsynchronous echogramspresentedin Figures
7 to 12 weremadeon
longitudinal
transects
in theseaward
direction.Theyareinsequence
andareannotated
with
fixedpositions
(yachtbuoysandlandmarks).Table1 shows
someof thedistances
between
fixes in metres.

Table 1' Distancesbetweenfixed stationsshownin Figure 1.
LQ- BR BR -45 45 - CH CH- VA

87m 105m 234m 144m
The Spardatais contouredas isothermsin thermalMaps, representing longitudinalvertical

sectionsof the estuary. The contoursare plottedat 0.1øCintervals.
A few generalpointscan be madeaboutthe observations:
There is a strongresemblance betweenthe thermalmapsand the echograms. The exact

causeof acousticbackscatter is uncertain,but densitygradients,
microstructureandparticles,
are all goodscatterersand are oftenassociated with the thermoclineregion.Regionswhere
isotherms stronglydivergevertically(e.g.Figure10 at "VA". from 1 to 3m belowthe surface),
and have associatedbackscatterin the echograms,are identifiedas being composedof
finestructure,causedby turbulentdestratification.
HW + 2:45
J BR GP 45 VI $W WP Clt VA
0m
Landward Seaward
Figure 7: Transect5 at 12:45 on 6 November1989.
In Figure9 (HW+4:45), halfwaybetween"CH" and "VA" the thermoclinehad an undular

structure,which on closerinspectionrevealedbillows.
Propellermixing in the thermoclineis evident(e.g. in Figure8) seawardof "RH" the

isothermsare tangledandthereis a patchof associatedbackscatter.This occurredat a time

when RV Catfish was known to have crossed its own wake.

HW + 3:15
I-• .tu D g
P
T
2u H
t t t t t t t t t - 3u
j DR GP 45 SW WP CH VA RH
0•
2•
........ 4m E
P
T
H
6m
45
La ndwa rd Seaward
Figure 8: Transect7 at 13:15 on 6 November1989.
HW + 4:45
BLQ I.• J BR GP 45 vI SW WP CH VA
0m
8m
Landward
Seaward
Figure9: Transect11 at 14:45on 6 November1989.

HW + 5:30
0m
2m
4m
,• TU BLQ 45 VII -
Landward Seaward
Figure 10: Transect13 at 15.30 on 6 November1989.
HW + 6:00
Om
lm
2m
- 3m
TU BLQ LQ J BR GP 45 Vl SW WP CH VA RH
0m
4m
TU
I
Landward
Seaward
Figure 11' Transect15 at 16:00 on 6 November 1989.

HW + 6:25
-- ,• ,ldwm
,, lm
2m
TU• S DQd SR GP 45 VI SW C VA
0m
6m
8m
BLQ
Landward Seaward
Figure 12: Transect17 at 16:25on 6 November1989.
At aboutHW+4:45 (Figure9) the approximatetime of maximumstream,the thermocline

seawardof the wave thickenedverticallyreducingthe temperaturegradientacrossit. This
thickeningincreasedwith time, suggestingactivemixingin, or across,the thermocline. The
regionin which this thickeningtook placespreadlandwardat subsequent times(Figs. 10, 11
and 12), finally reachingthe wave trough.
The isothermsoften take on a steppedstructure,with jumps in the heightsof contours

corresponding
to the spacingbetweensensors.
4.5 Results
Figures7 to 12 depict the formationof an internalwave, in the mannerdescribedby

Maxworthy (1979) and Lee & Beardsley(1974), in the lee of the landwardslopeof the
depression.Figure13 showsthe stagesof development of a lee wave,in a conceptual
model
basedon Maxworthy'sandLee & Beardsley'sexperiments andobservations.In thismodel
a lee wave growsas a downwarddisplacement of a densityinterface(in a two layer fluid)
over a bed depressionin an acceleratingstream,(a), (b) & (c). Justafter maximumtidal
streamthe lee wave,whichhaddevelopeda largeupstreamphasevelocityin orderto remain
stationary
at maximumstream,nowbeginsto propagate upstream againstthe slackeningtide
(d). In doing so it undergoesnonlinearsteepening and wave growthdue to the nonlinear

advectiveaccelerationof the internaltide, by both the barotropictide and the internaltide

itself. There is strongevidencefor this steepeningmechanismin numericalmodelsof two
layertidalflowsovertopography (Pingreeet al., 1983; Sturley,1990). Finally,asthe tidal
streamslackensbeyondsomecritical value,if steepening and dispersiveeffectsbecome
comparable,fission into solitary waves may occur, (e), (f), but this behaviourwas not
observedin the data presentedhere.
Figures7 to 12, themselves,

do notallow a comparison
with thisconceptual
model,but an
analysisof the wave heightshowsthe behaviourmoreclearly.
4.5.1 Wave Height
The maximum thermoclinedisplacement,or wave height, has been measuredfrom each

transectand theseare presentedin Figure 14. It was necessaryto define the undisturbed
thermoclinelevel in each case. Our experience,and numericalmodelling(Sturley, 1990),
have shownthat the downstream(seawardin this case) level is a good indicatorof the
undisturbexl level. The verticalpoint of maximumgradient,or alternatively,verticalmid-
point of the thermocline,was chosenas the boundarybetweenthe upperand lower layers.
There is goodagreementbetweenthe maximumthermoclinedisplacementobtainedfrom

the echograms andfrom the thermalmaps. Comparisons
betweentransects
suggestthatsome
of the variation is due to the 3-dimensional nature of the wave, and the deviations of the
launchfrom the axis of the estuary,but in generalthe estimateis reasonablefor the whole
wave.
Figure 14 showsthat the wavewasalreadyquitewell developedat HW+2:45, in response

to the increasingebb stream,with a displacement
of about 1.75m. BetweenHW+2:45 and
the endof the studythe depthof the wavetroughvariesin responseto the tidally varyingebb
stream.This is shownby thecorrespondence betweenthemaximumthermoclinedisplacement
and the magnitudeof the depthaveragedvelocity, a, measuredat station"45". To remain
stationaryin the accelerating
seawardflow the wave developeda landwardphasevelocity,
continuallyevolvingto matchtheebbstream. BetweenHW+l:20 andHW+3:00, a increased
and the wave height increasedin response. From HW+3:00 to HW+3:20, • decreased
slightly,but the wave continuedto grow throughnonlinearsteepening, as predictedby the
conceptual model. • increasedagainfrom HW+3:20 to HW+4:20, but the responsewas a
decreasein wave heightover thisperiod. This may havebeendue to increasedattenuation
at the wave crest,throughdiffusionand dissipation. Fig. 14 suggeststhat theremay be a
"critical"heightfor thiswave,of about2m. FromHW+4:20to HW+5:20a decreased again.
The initialresponse, fromHW+4:45 to HW+5:15,wasan increase in height,againasa result
of nonlinearsteepening.From HW+5:20 to HW+5:40 • increasexl by a smallamount,and
thereis a suggestion thatthewaveheightalsoincreased
at thistime. The generaltrendfrom
HW+5:15 to HW+6:15 was a declinein the wave heightas it propagated landwardlosing
energy.

Free surface
Density• (a). No tidal

stream
interface
. u
(b). Before
maximum
stream
(c). Maximum
stream
( u
(d). Just after

maximum
stream
< u
(e). After
maximum
stream
•-- u
(f) . No tidal
stream
ß u
Figure13:Thestages
whichformtheconceptual
modelofleewaveformation.
(a),(b),(c): Thewave
growsoverthelm•dward
slopeof thebed&pression
in theaccelerating
tidalstream.(d): After
maximum tidethewavesteepens
nonlinearly.(d),(e): Thewavepropagates
intotheslackening
stream.
(e),(f): Thewavemayfissionintosolitary
waves through a balance
between
dispersion
andnonlinear
steepening.
At HW+6:45thethermocline
stillappeared
to be displaced
by about1.3m- 1.4m,butthe
wave structureand the level of the undisturbedthermoclinewere difficult to define.
Theerrorincurredin measuring
thesizeof thewavetroughis estimated to beof theorder
+0.2m. This erroris roughlyequivalent
to a 10% uncertaintyin the largestthermocline

displacement,
andis a resultof makingessentially
subjective
judgements
aboutthelevelof
thethermocline.
It ignores
instrumental
errors,
whicharefelttobesmallin comparison.
Maximum
thermocline
displacement
in metres
&
Depth averaged
velocity in m/s
2.5
1 '5'
ECHOGRAMS•
T•HERtdALitdAPS
_
0'5 -
11 12 13 14 15 16 17
Time in hours (GMT)
Figure14: Thesizeoftheobservedmaximum thermocline

displacement,
inmetres,
plotted
together
withthedepth
averagedspeed,
fi, at"45"inmetres
persecond.
Thetimeofpredicted
highwater
was
10:00(GMT). Theobservations
arederived
frombothechograms
andthermal
maps.
4.5.2 Wave Energy
Thetotalwaveenergy,
perunitlength
inthey (across
estuary)
direction,
Er,wascalculated
from (Phillips,1977):

Er-f •(x)'X
pg•x
overthedomaindef'med by thestations "VA" and"BLQ". Thisenergyrepresents thatof the

internalwaveonlyandassumes equipartition
of potential
andkineticenergy,whichis truefor
linear waves,and good to first order for nonlinearwaves. No representation of the
barotropictidal energyis includedin this equation,and it assumes the wave is on a thin
interfacein a twolayerfluid, andis 2 dimensionalin nature.•l(x) is theverticaldisplacement
of the densityinterface,x the longitudinaldirection,Ap the densitydifferencebetweenthe
layersandg = 9.8ms
'2.
In makingthecalculations
it wasnecessaryto definea fixedhorizontal
domainoverwhich
to calculatetheenergy. Sincethesurveysdid notextendfar enoughin a landwarddirection
to definecompletelytheupstream thermoclineposition,"BLQ"waschosenas thelandward
limit of the waveform. In the cases when the transectdid not extend to "BLQ" the
thermocline
levelwasextrapolated
fromthemostlandwardpointof thetransect.In all cases
"VA" was more seaward than the seaward end of the wave. The undisturbed thermocline
levelwasidentifiedin theway alreadydescribed.Theintegration
wasperformed in sections,
usingthe fixed stationpositions,and measuringa meanthermoclinedisplacementbetween
section
ends.Apwasassumed
as16kgm
-a,whichwasthemeandifference
in layerdensities
over the periodHW+2:45 to HW+6:45.
This definitionof the boundariesof the wave would,certainlyfor the latter slagesof the
study,be restrictiveof the landwardedgeof the wave. However,it is anticipated thatthe
effects of this are not severe, and that the largest uncertaintyis in the thermocline
displacement,
(_+0.2m
or +10%), leadingto an estimated
errorof about20-25%in thelargest
energyestimates.
Figure 15 showsthedepthaveragedspeedwith theenergymeasurements derivedfxomthe

echograms andthermalmaps,andtheseareroughlyconsistent.The energyfollowsa similar
patternto the maximumthermoclinedisplacement, but is moreexaggerated
in its fluctuations,
reflectingthe dependency
on the squareof the displacement.
Thetotalwaveenergyisrelatively
high,around160- 180kJm '•, atHW+2.45,beforefalling
to 120-140kJm
-• at HW+3:00in linewitha sharpincrease
in theebbstream.Thereason for
thisis unclear. Otherwise,the energyincreased
with an increasing
ebbstream,andcontinued
to increasefor a while when the ebb slackened,presumablydue to the nonlinearsteepening
mechanismdiscussedearlier. The energy diminishedwhen the ebb streamdiminished
significantly,especiallyafter HW+4:45.
A peakwaveenergyof about2201rim '• wasmeasured.FromHW+4:45to HW+6:45the

waveenergydropped by about85kJm '•. Permetreof thelengthof thewave(about500m)
thisamounts
to a fall in energyof 170Jm -2.

Total wave
energy in
Kilojoules
per metre
&
Depth averaged
velocity in cm/s
250 -,
ECHOGRAIdS
200 -
150 -
100 -
TH•:RIdAL:MAP<•
50-
0 I I I I I
11 1:2 1• 14 1• 1• 17
Time in hours (GMT)
Figure
15'Theobserved
total
wave
energy
inkilojoules
permetre,
with
thedepth
averaged
speed
at
"45"incmpersecond.
Thetimeofpredicted
highwater
was10:00
(GMT). Theobservations
are
derived
frombothechograms
andthermal
maps.
4.5.3 Mixing
Between
HW+4:45
andHW+6:45radical
changes
inthevertical
distribution
ofdensity
took
place
at"RH"and"VA". Thedistortion
ofthethermocline
duetothewave presence
enhanced
theshear
intheseaward
endofthewave,
near"RH",
causing
fresher
water
tobe
mixed
from
above
thethermocline
down intoit. Water
wasthen
redistributed
vertically
in
thethermocline
through
mixing.Thethermocline
thickened,
expanding
downwards,
atthe
expense
ofthelowerlayer.Withtimethisthermocline
thickening
wasmanifest
alongthe

wholelengthof thewave. At HW+6:25(Figure12) thethermocline from"RH"to "SW"was

affected. Mixing in the thermoclinehadtheeffectof intensifying
thetemperature
gradient
at 2.3m,givingtheimpression of a "new"thermocline
formingin theechograms(Posmentier,
1977). Little mixingappeared to havetakenplacebetweenthebaseof the thermocline
and
the lower layer.
4.6 Discussion
This setof observations has indicated,qualitativelyat least,that the formationof an internal

wave over topographyin a partially mixed estuaryis capableof radicallyalteringthe
stratificationlocally, throughenhancedvertical mixing. This mixing is inducedby the
increasedshearacrossthe thermocline,producedby the distortionof the thermoclineby the
internal lee wave.
Nonlinearsteepening
in the slackeningtidal streamappearedto play a largerole in the
wave formation and evolution. There is clear evidence for this role in theoretical studies:
Lee & Beardsley(1974), Pingteeet al (1983) and Sturley(1990). In particular,non linear
advectiveaccelerationof internal waves in a slackeningtide can producesteepeningand
growth. Twice, whenthebarotropic tidal streamslackened,
aftera maximum,the wavegrew
in heightas a response. The heightof the wave, in relationto the total waterdepth,itself
suggeststhatnonlinearities mightbe importantin thebehaviourof thebaroclinicresponse of
the estuaryat this location.
A reductionin the wave growthoccurredas the wave height approached2m, and this
appearedto be a limitingheight. Wave energywaslostthroughmixinganddissipation and
this causeda reductionin wave height. The positionat which mixing in, or across,the
thermoclinewas first observed,was in the downstream(seaward)end of the wave. This was
also the positionat whichthe shearacrossthe thermoclinewould havebeenmostenhanced.
The regionof increased mixingmigratedupstream(landward)to meetthe troughof thewave,
suggesting that it had a largerpropagationvelocitythanthe wave itself.
The transferof energyto mixing was almostcertainlythe causeof the lee wavesdecline
in heightandenergy,in the latterstagesof the tide (Figures9 to 12). Futurestudiesaim at
improvingthe quantificationof thisprocess.
4.7 References
Dyer, K.R., andA.L. New, 1986: Intermittencyin Esmarinemixing. Esmarinevariability- proceedings

of theEighthBiennialInternational
EstuarineResearchConference,Universityof New Hampshire,1985.
Ed. Wolfe, D.A., 321-339, AcademicPress,Inc. pp. 509.

Gardner,G.B., 1984: Internal hydraulicsand mixing in highly stratifiedestuaries. PhD Thesis,

Universityof Washington,pp241.
Gardner,G.B., andJ.D. Smith,1978: Turbulentmixingin a saltwedgeestuary. Hydraulicsof estuaries

andfjords,ed. Nihoul, J.CJ., ElsevierOceanography
Series23, 79-106.
George,KJ., 1975: Tidesandtidal streamsin the TamarEstuary. PhD Thesis,Universityof London.
Huthnance,J.M., 1989: Internaltidesandwavesnearthe continentalshelfedge.Geophys.& Astrophys.

Fluid Dyn., 48, 81-106.
Lee, C., and R.C. Beardsley,1974: The generationof long nonlinearinternalwavesin a weakly
stratifiedshearflow. J. Geophys,Res., 79, 3, 453-462.
Maxworthy,T., 1979. A noteon the internalsolitarywavesproducedby tidal flow overa 3-D ridge.
J. Geophys.Res., 84, C1-6, 338-346.
New, A.L., andK.R. Dyer, 1987:Internalwavesandmixingin stratifiedestuarineflows. Proceedings

of InternationalSymposiumon PhysicalProcesses
in Estuaries,Noordwijkerhout,1986. Ed. Dronkers,
J., Van Leussen, W., 239-254.
New, A.L., K.R. Dyer, andR.E. Lewis, 1986: Predictionsof thegeneration andpropagation
of internal
wavesand mixingin a partiallystratifiedestuary. Est. Coast.SheffSci., 22, 199-214.
New, A.L., K.R. Dyer, andR.E. Lewis,1987: Internalwavesandintensemixingperiodsin a partially

stratifiedestuary. Est. Coast.Sheff Sci., 24, 15-33.
Partch, E.N., 1981: Time dependentmixing in a salt wedge estuary,PhD Thesis,Universityof

Washington,165pp.
Partch,E.N., andJ.D. Smith,1978: Time dependent

mixingin a saltwedgeestuary. Est.Coast.Shelf
Sci., 6, 3-19.
Phillips,O.M., 1977: The Dynamicsof the UpperOcean,2nd.Ed. CambridgeUniv. Press.
Pietrzak,J.D., C. Kranenburg,
andG. Abraham,1990: Resonantinternalwavesin fluid flow. Nature,
344, 6269, 844-847.
Pingree,R.D., D.K. Griffiths,andG.T. Mardell, 1983: The structureof the internaltide at the Celtic
Sea shelf break. J. Mar. Biol. Assoc. U.K., 64, 99-113.
Pond,S., andG.L. Pickard,1983: Introductory

DynamicalOceanography.
2ndEd. Pergamon
Press,
pp.329.

Posmentier, E.S., 1977: The generationof salinity finestructureby vertical diffusion. J. Phys.
Oceanogr.,7, 298-300.
Sturley,D.R.M., 1990: Topographically

inducedinternalwavesand enhanced
verticalmixing in an
estuary. PhD Thesis,PolytechnicSouthWest (CNAA), pp.222.
Sturley,D.R., andK.R. Dyer, 1990: The EstuarineThermistorSpar: an instrument

for makingthermal
prof0esin shallowwater. The Hydrographic Journal,55, Jan.90, 13-21.

Turbulence and shear induced mixing

processesin estuaries
E J Darbyshireand J R West
Abstract
During the summerof 1988 datawerecollectedin the upperreachesof the Tamar estuaryto
investigateverticalturbulentmixing of soluteduringneaptides. Measurements were made
of the vertical variationof turbulentmean longitudinalvelocityand salinity and of the
turbulentperturbations of the verticalandlongitudinalcomponentsof velocityand salinity.
On theearlyebbtidea largeverticalsalinitygradientdeveloped,themeasured turbulentfluxes
were found to be small and an extensive internal wave field was observed. Consideration of
the temporalevolutionof the verticalsalinityprofilessuggested
thatthemeasuredfluxeswere
an underestimateof therequiredverticalexchange andit wasconcludedthateitherundetected
smallscaletransportassociated with internalwavebreakingor Wansport due to a transverse
secondarycirculationwas takingplace.
5.1 Introduction
The movementof water and solutesin estuariesis governedby the turbulentmean flow
structureandturbulentdiffusionprocesses.Themeanflow is causedby fluvial,tidalandwind
influencesandcontrolledin detailmainlyby turbulentdiffusionprocesses. The production
of turbulencewhich causesthe diffusionis causedby bed drag characteristics, the mean
velocityanddensitygradients,internalandsurfacewavecharacteristics
andacceleration.Thus
thereis strongfeedback
betweentheturbulence
andmeanfields. Giventhatthebathymetry
of estuarinechannelsis usuallycomplexit may be anticipatedthat esmarineflows and the
associatedscalarfieldsare complexfunctionsof spaceandtime.
In thelasttwo decades
extensive
measurements
havebeenmadein boundary
layersin air
and water in laboratories, some measurementshave been made at sea and a few have been
madein estuaries.The paucityof thelatterdatais surprisingconsidering theanthropogenic
pressures
on natureleadingtonumerous mathematicalmodelling studiesof esmarinetransport
processes.
Thereare,however,considerable physicaldifficultiesin obtaininggoodfield data

76 E J DarbyshireandJ R West
and it is not easyto developan understanding

of temporallyvaryingthreedimensional
flow
fieldsfrom a limitednumberof pointmeasurements whenmomentumandsolutetransportare
influencingthe flow.
This paperconsiderssometurbulencedata collectedin the Tamar estuaryin July 1988.

The dataare novel in that the turbulentfluctuationsof velocityand salinityhavebeenmade
over a wide rangeof relativedepthsover two ebbtideswith rangescloseto neapconditions.
The resultingturbulencefield in a channelwhosebathymetryis mainlycontrolledby themuch
greatershearstressesinducedby springtide, stormand fluvial floodprocesses, is relatively
weak but the measurements are shownto be adequatefor the developmentof hypotheses
concerningtransportmechanismsand the partial substantiation of someof theseideasfor
estuarine channels.
5.2 Previous work
Due to the interactionof verticalvelocityshearandlongitudinaldensitygradient,ebbtidesin

estuariesare frequently associatedwith significantvertical gradientsof salinity and
consequently density. Vertical turbulenttransportin suchenvironments dependson the
maintenance of turbulencewhich,in stratifiedflows dependson the relativemagnitudes of
shearproductionand verticalbuoyancyflux whichmay be expressed by the local gradient
Richardson number,
Ri- aplaz (5.1)

p (au/&)
whereU is themeanlongitudinal velocity,p is themeandensity,g is the acceleration

dueto
gravityandz is positiveupwards.Theoreticalinvestigationsby Miles (1961) showedthatthe
criticalvalueof Ri in stratifiedshearflow was0.25 and that instabilitywaspossiblebelow
this value.
The verticalturbulenttransport
of momentumor solutemay be relatedto the gradientof
the meanflow via the mixing lengthgivenby
(5.2)
for momentum,and by
(5.3)

E J Darbyshireand J R West 77
for a solute. Here S is the meansalinity,u, w and s are perturbations

from the meanvalues
of longitudinaland verticalvelocityand salinityrespectively and an overbarindicatesa
temporalmeanvalue. The subscripts of 1, the mixinglengths,are m, momentum,s, salinity,
z, verticaltransport,and O, homogeneous conditions.Hereafterverticaltransportwill be
assumedand the subscriptz omitted.
Theoretical
relationships
relatingtheturbulent
mixinglengthsin homogeneous
andstratified
flows to Ri have been proposedby Rossby& Montgomery(1935) and Kent & Pritchard
(1959) with
1. - 1.o(1* •iRO"O's (5.4)
and
1.,- 1.,o(1+Ill Ri)-1'0 (5.5)
respectively, where[t• and[t2areconstants.Valuesof theseconstants havebeendetermined

by Odd & Roger(1978) by considering a two dimensionalmassbudgetanalysisandby West
et al. (1985) by directmeasurementof the verticalturbulentflux. In boththesecasesRi was
evaluatedfrom data taken over a period of about 1.0 minute and the interpretationof the
resultswasbasedon the assumption of turbulencegenerationin theregionof maximummean
velocityshear,i.e. nearto the bed. Geyer& Smith(1987) showedthatevaluationof Ri over
muchshortertime scalescauseda significantincreasein the numberof subcriticalvaluesof
Ri. Thisreductionwasbroughtaboutby theadditionalshearcausedby internalwavevelocity
fluctuationsallowinglocal turbulenceproductionremotefrom the solidboundary.
The suppression of turbulence

by densitygradientsis alsodependant uponthelengthscales
of the turbulenceinvolved. Largelengthscalesmaybe almostcompletelysuppressed while
motionson much smaller scalesmay be practicallyunaffected. This can be seen in the
steepening of the slopeof thesalinityvariancespectrum in thepresenceof a densitygradient
fromthevalueof-5/3 whichis observed in homogeneous conditions(Turner(1973)). Various
valueshavebeenobtainedfor this steepened slopesuchas by Dyer (1981) who observeda
relationshipbetweenthe magnitudeof the slopeand the layer Richardsonnumber.
Internalwaveshavebeencitedas thecauseof increased turbulentmixingin estuaries

by
Partch& Smith(1978) and New et al. (1987) althoughtherehasbeensomedoubtas to the
exactmechanisms of wave generationandbreaking. In boththeseinstancesthe waveshave
beenobservedin estuarieswherethe densitystructurecan be approximated to a two layer
situationandthewavespropagatedalongthepycnocline acrosswhichthemixingtookplace.
A furtherfeatureof moststudiesof estuarinemixinghasbeenthattheyare confinedto a

twodimensional view. On floodtidesin theConwyestuaryNunes& Simpson(1985) showed
that significanttransverse
currentsoccurwhichnecessitatethe presenceof verticalcurrents
within the flow which in this case were downwards towards the centre of the channel and

78 E I DarbyshireandI R West
upwardstowardsthe banks. The effectof suchcurrents

wouldbe to transportmassand
momentum withouttheneedfor turbulent
exchange.Upwardverticalcurrents
wereobserved
duringebb tideson the Great Ouseby Odd & Roger (1978) who commented that the
momentum exchange dueto themcouldbe asgreatasthatdueto turbulence.Theydidnot,
however, comment about the mass transfer involved.
(a.) Depth mean values

Velocity •, Salinity
I I ! I
16.00 ca
--1
(b.) Velocity profiles O. 75 ms
I I
11. O0 12. O0 13. O0 14. O0 15. O0
(c.) Salinity proœiles •o. o ppt

i i
I I I
11. O0 12. O0 13. O0 14. O0 15. O0 16 O0
Time
Figure1' Verticalprofilesanddepthmeanvaluesof velocityandsalinity,

Cotehele
417/88.LSWat
17:45.
5.3 Data collectionandanalysis

Thedatawerecollected
nearto thecentre
of thechannel
of a straight
reachof theupper
Tamarestuary
neartoCotehele
Quay.Thechannel
widthat thispointis about100mandthe

depthvariesfrom 1.0m to 5.0m dependingon the stateof the tide. The measurements were
takenon the 4th and 8th of July 1988 whenthe daily meanfluvial dischargesof 6.8 cumecs
and 5.6 cumecsrespectivelywere aboveaveragefor the time of year but low comparedwith
winter valueswhich can exceed100 cumecs. The 8th was a neap tide and the 4th was an
intermediatebetweenneapand spring.
•_
(a.)Depth
o Velocity
meanvalues
z• Salinity _
•-- I I I I I I I I I I I
c5 13.O0 14.O0 15.O0 16.O0 17.O0 18.O0
-1
(b.) Velocity profiles o. 75 ms
I
[ I I I
13. O0 14. O0 15. O0 17.00 18.00
(c.) Salinity profiles 5. o ppt

I !
I I I
13. O0 15. O0 16. O0 18. O0
Time
Figure2: Verticalprofilesanddepthmeanvaluesof velocityandsalinity,Cotehele8/7/88.HSW at

13:15.

dS/dz (O.õtSm) + d$/dz (l.•4rn)• 4/?'/88
d50 pp•t
d$/dz (O.õSrn) •- d$/dz (I .•4rn) • 8/•'/88
Figure3: Profiles
andverticalgradients
of salinity
computed
fromfittedquadratics.
Turbulent
meanvelocityprofilesweremeasured usingeightBraystoke
impellerheads
mountedona 2.0mmastat intervals
of 0.25m. Themeanvalueswerecomputed over50s.
Turbulentmeansalinity
profiles
weredeterminedfromwatersamplespumpedtothesurface
via pipeswhoseinletswereattachedto the samemast. The salinitiesweremeasured
with an
MC5 salinometer.
Turbulent
perturbations
oflongitudinal
andvertical
velocity
weremeasured
using5.5x10'2m
diameter
electromagnetic
flowmeters(Coinbrook
Instrument
Development
Ltd.)andthoseof
salinity using an inductivesalinometerwith four sensorheads. These instrumentswere
mounted
on a streamlined
mastin sucha way thatthehydraulic
interference
betweenthe
sensorheadswasconsideredminimal. The flowmeterheadsweresituatedat 0.40m and 1.12m
withthesalinometer
heads
at0.56m,1.24m,
2.06mand3.07mabove
thebed.Theseparation
of theflowmeterandsalinometer
headsis of relevanceto thelaterdiscussion.
The mastwas
mounted on a weighted
bedframefittedwithorthogonal inclinometers
to ensurethatit was
level on the bed.
Turbulent
meanvelocitiesandsalinities
areshownasverticalprofiles
in figs.1 & 2 forthe
4thand8thofJulyrespectively
together
withthetemporalvariation
ofthedepth meanvalues.
Thevelocity
profiles
areplotted
ona logarithmic
scale.Thepointsdefiningeachprofilehave

beenfitted with a third orderpolynomialcurvein the caseof velocitiesand connectedwith

straightlinesin thecaseof thesalinities.Verticalprofileshavebeenplottedasdeviationfrom
the depthmean value. The ordinatefor eachprofile is plottedas a verticalline at the time
the profile datawere collectedand with a lengthequalto the flow depth. The scalesfor the
deviationfrom the depth meanvaluesare shownas horizontalbarsfor eachparameter.
Heightabove•he bed (m)

0.00 0.50 1.00 1.50 • 00 2.50 • 00 3.50 4.00
00.79
00.66
oO.
6J
o0.71
00.97
ol.
95
o15.50
oO.
/
JJo0.45
00.54
o0.
7,«o,.16
04.56/
o0..•1
o0.50
o0.?• o1.05 /
o.
II
Figure4: GradientRichardsonNumbers,4/7/88.

Height,abovet,hebed (m)
0.00 0.50 1.00 1.50 P. 00 •.50 3.00 3.50 4.00
I I I I
o21.29
o5.18
ol.
82o1•
,'•o0.9<3
o0.95
027.68
ø2.
31/•9000.6400.58
00.63
oo.
00.78
o0.
• 00.42
0.•
00.47
o0• 00.43 00.42 o 0.•
0.27 o 0.• o 0.•
0.27 o0.• 00.49
0.32 o0.• o0.• o0.•
Figure5: Gradient
Richardson
Nmnbers,
8/7/88.
Therecordsof theturbulent
perturbations
of thevelocity
andsalinity
records
weredivided
intorunsof 13.6minutesanddigitized
at 10s
-•. Largespikeswerethenremoved anda four
point
smoothing
applied
togivetmeffective
digitization
rateof2.5s
'•. Inorder
toeliminate
longer
period
non-turbulent
contributions
tothefluxes
therunswerethenfurther
divided
into
sub-sections
of 51.2sandthe lineartrendremovedfromeachrecord. Eachsub-section
was
thenFourier
analysed
using
a standard
F'FFroutine
andthevalues
of theturbulent
intensifies
andturbulentfluxesdetermined
for eachone. High andlow frequency
cut-offerrorswere

computedand appliedaccordingto Soulsby(1980). Finally an ensemblemeanof the results

for eachsub-sectionwas evaluated. To assessthe lower frequencydue to internalwavesthe
spectraof the complete(13.6 minute)salinityrecordswere also computed.
ß ß ß
o L! v W .-I-
/ \ z=O.40m
W t 9:15
lOI oo I 11. OO I•. OO 13. OO 14. OO 15• OO
4/?/88
z=l. 11•m
__ II. O0 I•-ø.O0 13. O0 14. O0

W'
at.
9:15 151O0
8/7/88
z=O.40m
HSW at 13:15
i I I
141O0 ' 151O0 161O0 19. O0 61
8/7/88
z= 1. l•'m
HSW at 13:15
i i I i
14.00 15.00 16.00 I• O0 16.00
Time
Figure6: Turbulentintensities
of velocityandsalinity.

84
E J DarbyshireandJ R West
4/?/•
z=O.40m
H$•r at 9:15
' 111O0 ! 1•.I O0 ! 131

.00 ! 141O0 I I . O0
4/?/•
z=l.
H$•r at 9:15
• •o. oo I .00 1•I.O0 13.O0 I . O0 I . O0
•/?/•
z=O.40m
HSW at 13:15
z--1.118rn
HSW at 13:15
__
I i i i
241oo 2•I oo z61oo 27. oo •doo •
Time
Figure7: Shearstress
andturbulent
fluxesof salt.
Computation
oftheturbulent
meansalinity
fromtheinductive
salinometer
data
provided
anadditional
representation
ofthevertical
density
structure.
Although
thisviewisless
detailed
inaspatial
sensethe
temporal
resolution
ismuchbetter
thanforthe
pumped
sample
profiles.
Profiles
computed
fromthese
data
areshown
infig.3 with
quadratic
curves
fitted
tothefour
points.
Also
shownisthetemporal
variation
ofthevertical
gradient
evaluated

from the quadraticsat z=0.56m and z=l.24m. Each inductivesalinometerhead had been
calibratedagainsta large MC5 salinometerfield data set which includedthe data which is
under considerationhere. The correlationcoefficientsfor this linear calibration were 0.9894,
0.9873, 0.9899 & 0.8825 for heads1, 2, 3 & 4 respectivelywherehead 1 is nearestthe bed.
The local gradientRichardsonnumbersevaluatedfrom the turbulentmean profiles are

shownin figs.4 & 5 for eachday. In orderto obtainthe requiredverticalgradientsvertical
interpolationsweretakenfromtheprofilesto therequiredheightsandtheninterpolations were
madein time assuminglinearvariation. Valuesof Ri are not shownwhereextrapolationof
the velocityprofrieswouldhavebeennecessary for theftcomputation whichaccounts for the
scarcityof valuesat z=0.50m. Thelinearinterpolation is reasonablein thecaseof thevelocity
data becausethe temporalresolutionof the profilingis good. In the caseof the salinity
profilesthe resolutionis only oneper hour. Howeverconsideration of the gradientsshown
in fig. 3 showthe assumption of linearvariationof the gradientsis not unreasonable.
The rootmeansquarevalues(turbulentintensities)
of thevelocity(u',w') andof thesalinity
(s') perturbations
aboutthe lineartrendin eachrecordare shownin fig. 6 at two depthsfor
theearlyebbtideof eachday whensalinitywaspresent.The associated valuesof shearstress
andverticaland longitudinalturbulentsaltfluxesare shownin fig. 7.
5.4 Observations
On bothdaysthe ebb tide flow beginsat the surfaceandthereis considerabledelaybefore

thebarotropicforcingdueto the surfaceslopefully overcomesthebaroclinicforcingdueto
the longitudinaldensity gradientin the lower layers. The velocity profiles remain
non-logarithmicforthedurationof thesalineperiod.Acceleration
of thedepthmeanvelocity
to a maximum
valueof about0.65ms
'2is gradual
buttheprocess
is notuniform
throughout
theflow withlayersat successively
lowerdepthsbeingbroughtintomotionrelativelyrapidly
thusproducinglocal shortperiodsof acceleration.
Significantverticalgradients
about1.0kgm 4 are presentat the commencement of data
collection and theseare observedto increase,particularlynearto the bed, up to about
8.0kgm'4. Fig. 3 showsthe character
of the profilesto changefrom 'concave',
where
ß S/dz2>0to'convex',
where•S/dz2<0,overtheperiod of datacollection
passing
through
a
periodwhenthe profile is nearlinear.
The gradientRichardson numbers, figs.4 & 5, arealmostalwaysgreaterthanthecritical

valueof 0.25 but alsorarelygreatlyexceed1.0. Thefew valueswhichareverylargearea
result of the use of vertical gradientswhich are close to the limit of resolutionof the
instrumentation.
The resultingvaluesof Ri are undoubtedly
largebut their accuracyis
debatable.

5.5 Discussion
The turbulentmeanprofilesof salinityare observedto changeover the observation period.

This changeis a combinationof advectionof the entireflow structurepastthe measurement
positionand the actionof the velocitygradientwhich can havethe effectof increasingthe
verticalgradientandverticaltransportof salt. The latterwill tendto reducetheverticalsolute
gradient. The fact that Ri remainsabove 0.25 for most of the time indicatesthat turbulent
exchange will be severely
inhibited.ForRi--0.5and,using[•--160obtainedby Odd& Roger
(1978) for eqn. 4, the value !m is reducedto about 10% of its value in homogeneous
conditions.However,Ri hasbeenevaluatedovertimescales of 50sandlengthscalesof 0.25m
and conditionsfor instabilitymay exist on shortertime scalesand lengthscalesthanthese.
If this is so then turbulenceproductionmustoccurat positionsremotefrom the bed. The
salinitygradientis large throughoutthe flow depthand showsno signsof erosiondue to
verticalmixing,evenat the lowestmeasurement position. This suggests
thatbed generated
turbulencehaslittle, or no, influenceon the thisstageof the tide.
The observedvaluesof the vertical turbulentflux of salt (fig. 7) are extremelysmall

especiallyat 0.40m whichis consistent with the suggestionof the previousparagraph.In
orderto comparethe effectsof velocityshearandverticalturbulenttransport the following
calculationwas performed. Ignoringthe effectsof advectionof the entireflow structurethe
rateof changeof verticalsalinitygradientdueto theactionof verticalvelocityshearis given
by,
(5.6)
Usingtheobserved valuesof U, aU/bzandaS/•x thispred_icted

valuewascomputed for
eachof thetimesfor whichtheobserved rateof changeof_•S/bzwa_s
available.Thevalue
of the longitudinal
gradientwas obtainedby assuming aS/•x=(1/U)(•S/•), the temporal
gradientobtainedfrom the salinityperturbation
records.Thesevaluesof thepredictedand
observedratesof changeof the verticalgradientare shownin columns3 & 4 of table 1.
Theseresultsshowthat muchof the time the predictedrateof steepening exceedsthat

whichwasobserved to occur.Thissuggests thatsomeverticaltransport
of saltis takingplace.
If thisis entirelydueto theverticalturbulent fluxthena minimumvalueof thisfluxmaybe
calculated. It wasassumed thatall the salttransported
by sucha flux into a layerof finite
thickness d wasaccumulated withinthelayerandthattheverticalgradients of salinityand
velocitywerelinearacrossthislayer. This givesa requiredflux of,

$W I (5.7)
2dt•z
The differencebetween the flux recordedby the sensorsand the values of the flux are
shownin column5 of table 1 wherea positivevalue is shownif the predictedflux exceeds
that which was observed.Althoughthe numberof positivevaluesis not muchgreaterthan
the numberof negativeonestheir magnitudes are generallygreaterleadingto the conclusion
that the verticalturbulentflux is frequentlybeingunderestimated.There are threepossible
reasons for this observation.
Table 1: Predictedand observedratesof changeof vertical salinity gradient,vertical turbulentflux

deficit and values of W needed for mass balance.
Predicted Observed Flux deficit

Date Run •__O•kgm'4s
'1 A(/}Sfi)z)kgm-4s
'1 -• kgm'2s
'1 Wms'1
z /}t (/}z) At
x 10'3 x 10-3 x 10'3 x 10'3

4/7/88 9 0.25 2.00 -0.39 1.50
0.40m 10 1.20 1.40 -0.47 1.02
11 0.25 0.39 0.07 -0.14
4/7/88 4 0.02 0.50 -0.16 -1.23
1.12m 5 2.00 0.17 0.52 2.54
6 1.10 1.00 -0.01 -0.74
7 3.10 1.60 -0.23 5.58
8 1.50 0.46 0.01 -0.49
9 1.60 1.60 -0.26 0.82
10 4.00 1.00 0.90 -1.42
11 1.50 0.10 0.47 -0.71
8/7/88 23 0.64 0.80 -0.04 0.18
0.40m
8/7/88 19 3.50 -0.21 1.23 4.26
1.12m 20 4.10 0.91 0.82 6.64
21 3.10 0.93 -0.18 -5.49
22 2.10 1.10 0.32 -1.89
23 0.63 0.57 -0.14 0.49
It is possiblethat the assumptionregardingthe effectsof advectionis incorrectand that

theycannotbe ignored.Longitudinal sectionsof estuariesusuallyshowthat/}S/i)zincreases
in a landward directiontowards thelimit of salinityintrusion
andtherefore simpleadvection
of thisstructure
duringebbtideswouldresultin anadditional steepening
of theprofrieswhich

88 E J Darbyshireand J R West
is not observed.
The secondpossibilityis thatthe verticaltransport

is occurringbut on very smalllength
scalesandthereforeremainsundetected by the turbulencesensorsanda thirdis thattransport
takesplacevia someotherturbulentmeanmechanism.
Date 4 7 88 Time 1248 Run 9
• z= 3. 07111
•_ z= 2. o6m
s• z= 1. 24m
• o •o •o •o •o • •o •o •o •o •
Time (seconds)
Figure 8: Salinityperturbations
at four heights.
Considerfirst small scaleturbulencegenerationand the possibilitythat this is associated

with internalwaves. It is apparentfrom the salinitytime seriesdata shownin fig. 8 that
internalwavesare presentthroughoutthe flow period. It is alsoapparentthat thereare many
frequenciespresentand that individualtrainsof wavesdo not persistover long periodsof
time. The frequencyspectraof thesetime seriesare shownin fig. 9 whereinputsof energy
dueto intemalwaveactivityareapparentparticularlyfor thespectrum computedfor z=0.56m.
In order to conducta simple analyticalinvestigationof the wave field assumean

approximate
density
structure
of theflowsuchthat3N/3z=0,whereN2= (-g/p)(3p/3z).
Over
the depthsbeingconsideredthisapproximates to thedensityprofilebeinga straightfine. Fig.
3 showsthat this is a realisticfirst approximation for mostof the time. Also ignorethe
effectsof the verticalvelocity shear. If the flow is irrotationaland the fluid incompressible
and is movingpastthe observerwith a velocityU thenit canbe shown,after makingfurther
assumptions (appendix1.), that the first modeinternalwave will havethe dispersion relation,

0 2•2]c4_2G•D2]c
3+(G•2D2+ •2 •g2
_•2 )•2) •.2_2G•
07g
2]C
+G•
2•g' 0 (5.8)
which canbe solvednumericallyfor k, the wavenumberwhereD is the flow depthand o• is
the observedwave frequency.
Nun
9 •
Time
Da •e 4
• Slope
o!
s•roigh•
lines=-5/$
z=O. 56m o
z= 1. •4m
z=•. 06m +
z=3. OTm
--• •o -1. 40 -o. eo o •o
Log•o Fre q u en cy
Figure9: Normalizedsalinityvariancespectra.
On solvingfor k it is foundthattworealrootsexist,onerepresenting
propagation
with the
flow andoneagainst.The resulting phasevelocities
aresuchthattheupstream
propagating
solutioncannotpropagateagainstthe flow and is thereforenot realistic. The waves are
thereforeassumedto be propagatingin the samedirectionas the meanflow.
Theparticlevelocities
associated
withthewavepropagation arecomputedbyassumingthat
theobserved
varianceof thesalinitysignalis dueto a singlefrequency
sinusoidally
varying
wave. Thedetailsof assumptions
andcomputation aregivenin appendix1 andtheresulting
maximumlongitudinal andverticalparticlevelocitiesare givenby

1¾
o - (SS/&)
-• 2xa$o (5.9)
and
(5.10)
where/5"
= to2(N2- to2)/lc
2. Theresults
of thesecomputations
foroneof thesensor
headsare
shownin table2 where7-o,So,U0 andW0aretheinternalwaveverticaldisplacement, salinity,
longitudinaland vertical velocity amplitudes. Referenceto fig. 8 showsthat while the
sinusoidalassumption is not very realisticfor the z= 1.24mtime seriesit alsoshowsthat the
predictedvalueof So(table2; run 9) is not a largeoverestimate and is sometimes exceeded
on timescalesof a few wave periods.
Table 2: Salinity,DisplacementandVelocityamplitudes,
z=l.24m, 4/7/88.
Run So(PP0 7-ox 10g(m) Wox 10•(ms

") Uox 10•'(ms
")
1 0.03 1.6 0.6 0.7
2 0.03 1.4 0.3 0.8
3 0.14 6.1 1.7 3.3
4 0.16 6.8 1.9 3.5
5 0.27 8.5 4.7 3.9
6 0.36 9.9 5.3 4.7
7 0.51 13.5 8.0 5.8
8 0.47 12.7 5.2 5.5
9 0.41 10.7 7.1 3.5
10 0.60 16.5 11.8 4.2
11 0.58 20.3 13.3 3.7
Thesevelocitiesmay thenbe comparedwith the observedvelocityperturbations.In order

that the wave effectscanbe seenmoreclearlythe time serieshavebeensmoothedover 6.4s.
Simultaneous variationsof w,u ands are shownin fig. 10. The firstfeaturewhichis clearly
apparentfrom the latter part of the run is that positivevalues of u correlatewell in a
qualitativesensewith positivevaluesof s whichis consistentwith a wavepropagating in the
samedirectionas the meanflow. The effectof advectionof the relativelyslowlyvarying
wavefieldpastthe sensorpositionis thattheobserved phaserelationship is changedfromthat
whichwouldbe observedin still waterandpositivevaluesof w lag behindpositiveu and s.
The observedmagnitudesof the velocityperturbationsare abouthalf thosewhich were

predictedby the theoryand it remainsa matterof speculation
whethersuchwavesmight
break.The presenceof verticalvelocityshearmeansthat the onsetof wavebreakingwill

occur at lower valuesof U0 than it would in a non-shearedflow. It is clear however,that

shouldbreakingoccur, that the resultingturbulencewill be on lengthscalesless than the
wave displacement amplitudewhichis generally<0.15m.
Date 4 7 88 Time 12 7 Run 6
Vertical
Velocity
Longitudinal Velocity
Salinity
Time (seconds)
Figure 10: Salinityandvelocityperturbations

at z=l.121n.
[] []
o
o 0
[] A
% [] A 0
0 0 /x /• rn•
0.56m o
1.24m /x
2.06m []
3.07m v
Figure11' Salinityspectrum
decayexponent
vs.verticalsalinitygradient.

92 E J Darbyshireand J R West
Thefactthatperturbations
of frequencies
lessthanaboutls'• aresuppressed
canbe seen
from examinationof the salinityspectra.A regionwherethe spectrumdeclinesmoresteeply
than-5/3 canbeseenextending
fromfrequencies
0.0ls'• to 1.0s
'•. Thereis a suggestion
that
abovethese frequencies the rateof decrease becomeslesssteepalthoughthereare ratherfew
pointsuponwhichto basethisspeculation.The magnitudes of theseslopeshavebeenplotted
againstthe verticalsalinitygradientin fig. 11. There is a clear trendto steeperslopesat
steepergradientsalthoughthereis considerable scatter.The observedvalueswhichare <5/3
are generallyconfinedto periodsof very low salinitywhenthepresence of randomnoisemay
have affectedthe spectra.
Furtherevidence of dampingasthedensitygradient increasesmaybe gainedfroma plot

of s' vs./}S//)z(fig. 12). Althoughnotconclusivethereappears to be a valueof/}S/i}zabove
which s' no longerincreases.The situationis complicatedby the presenceof internalwave
perturbations whichcontributeto s' andwill not be affectedby buoyancyin the sameway as
turbulent perturbations.Fig. 12 also showsthat s' tendsto increasewith distancefrom the
free surfaceand thebed whichis alsodueto the fact thattheperturbations are largelydueto
vertical wave perturbationswhich are smallernear to the flow boundaries.
0.56m o
1.26m /x
2.06m []
3.07m v
[]
D •x [] •
0 V zx 0 o o
zxO• [] 0
o •
-dS/dz (kg m -'•)
Figure 12: Salinityturbulentintensityvs. verticalsalinitygradient.
Finally, considerthe possibilityof the existenceof a flow structure'•,.,hich

resultsin a
turbulentmean velocity in the verlical and transversedirections. Althoughno direct
measurements weremadeit wasobserved thattheaxialconvergence whichwasobserved Io
form on flood fidesdoesnot survivebeyondthe early ebb and that the accumulated debris
appear to end up on the banks. This is the nature of a transverseflow consistentwith
upwellingtowardsthecentreof thechannel.Sucha flow maybe anticipated duringebbtides

E J DarbyshireandJ R West 93
due to the interactionof transversevelocityshearand longitudinalsalinitygradient. This

shouldproduce,at any givenheight,lower densitywater in the main channeland higher
densitywaternearto the banks. This transverse densitygradientcouldprovidethe forcing
functionfor the necessarycirculation.
In order to assesswhethersuch a circulationwould be realisticthe necessaryvertical

velocityto balan_cethe observedverticalflux deficitwascomputed in eachcase. It was
assumed that /}W/3z was smalland that the rate of changeof the salinitygradientdue to
verticalturbulent
mean flowis givenby W(•S/3z2). Thecomputed
necess_asy
vertical
velocitiesto balancethe flux deficit are shownin table 1. The valuesof •S/3z 2 were
obtainedfrom the quadraticrepresentationsof theprofiles(fig. 3) whichgivesa singlevalue
for thewholeflow depth.Thisvaluewasobservedto changesignduringtheebbtidesothe
necessaryvalueof W is notalwayspositive.Thevaluesof W donotexceed1.0x10 '2ms'1
whichis smallbut thereis no reasonwhy they shouldnot occur. Given a channelbreadthto
width ratio of about 20 this could restfitin transversevelocitiesof about 10.0x10'2 ms'1 which
may be detectableby surfacemeasurements.
5.6 Conclusions
1. Existingturbulencemeasuringsystemshave been shownto have somelimitationswith

respectto theinvestigation
of verticalturbulenttransport
processes
duringsolutestratifiedneap
ebb fidesin a shallowpartially mixed estuary. However,preliminaryestimatesof vertical
fluxeswere possibleand most featuresof the individualvectorand scalarparametershave
been determined.
2. The interactionof the longitudinalsalinitygradientandvelocityshearproducea turbulent

meanflow structure whichappearsto be influencedsignificantlyby localeffects.Thisis true
in botha spatialandtemporalsense.The mainevidencefor thislocalcontrolis thatin spite
of a tendencyfor the velocitygradientto increasewith time due to the velocityshearthe
gradientRichardsonnumberremainslessthan 1.0 over large sectionsof the flow due to
concomitant changes in thesalinitygradient.Thisleadsto theconclusion
thatthevelocityand
salinityturbulentmeandistributions arelinkedby oneor moremechanisms eventhoughsteep
verticalgradientsexist.
3. Measuredvertical fluxes of momentumand solutesusinginstrumentation with sensor

separation
of about0.15m gavemixinglengthsof 10%of the valuefoundin homogeneous
openchannelflows. The solutefluxeswere generallylessthanthoseexpectedfrom mass
balancecalculationsbasedon temporalevolutionof the turbulentmeanfield.
4. Continuous measurements of salinityshowedthatan incoherentinternalwave field was

present.Therms.valuesof theperturbations calculated
over13.6minutesinitiallyincreased
with salinitygradientbut thetrendbecamelessclearat highergradients.Theyalsoshowed

a tendencyto decrease in thevicinityof upperandlowerflow boundaries.An analysisof the

wave field for idealizedconditionssuggested that the waveswerepropagatingwith the flow
and the predictedparticlevelocitieswere not inconsistentwith thosewhich were observed.
The salinity frequencyspectrashowedan increasingeffect of buoyancyon the higher
frequencypart of the spectrumas the densitygradientincreased.
5. Previouslyother workershave suggestedthat for turbulentmean gradientRichardson

numberssomef'mescaleturbulencegeneration maybe expecteddueto internalwavebreaking
in the regionsof maximumshearwhere,on shorttime scales,Ri may be reducedbelow0.25.
This processis probablyintermittentand leadsto the periodicappearance
and decayof the
waves.
6. It is conceivable
thatthesecondary circulationcausedby a transverse
densitygradientmay
contributeto the vertical fluxes. In somecasesonly a weak circulationis neededfor a
significantcontribution.Sucha motionis not physicallyunrealisticdespitethe presenceof
steepdensitygradientsandthe absence of surfacedebrisaccumulatedduringthefloodtide is
circumstantial evidence for its existence.
7. Someprogress shouldbe possiblein provingsomeof theaboveconjecture by usingproven

techniques. Secondaryflow structuremay be investigated by droguetrackingusing
appropriatelydimensioned drogues.Continuous point sourcedye tracingcanalsoprovide
quantitativeestimatesof verticaland transversesolutemixing rates and hencediffusion
coefficients.
Notation.
d Layer thickness.
D Flow depth.
g Accelerationdue to gravity.
k Wave number.
1 Mixing length.
N Buoyancyor Brunt-Vaisalafrequency.
P Pressureor Spectrumdecayexponent.
Ri Local gradientRichardsonNumber.
s Perturbationof salinityfrom the turbulentmeanvalue.
s' Turbulentintensityof salinity.
s Salinity.
s Turbulentmean salinity.
So Internalwave Salinityamplitude.
t Time.
U,V,W Perturbations
of longitudinal,transverse
and
vertical velocities from the turbulent mean values.

E J Darbyshire
andJ R West 95
u',v',w' Turbulentintensitiesof the longitudinaltransverseand

vertical velocities.
U,V,W Turbulentmeanlongitudinal,transverseand vertical
velocities.
Uo,Wo Internalwave longitudinaland verticalvelocity
amplitudes.
x,y,z Longitudinal,transverseandverticalcoordinateaxes.
z Height abovethe bed.
Zo Internalwave verticaldisplacement amplitude.
•1,•2 Constants.
õ A parameter.
3, Wave length.
p Turbulentmeanfluid density.
c• Frequency.
to Angular frequency.
Subscriptsof 1.
m,s Momentum,salinity.
z In the vertical direction.
0 Under homogeneous
densityconditions.
Acknowledgements
The authorsgratefullyacknowledgethe financialsupportof the Scienceand Engineering
Research Council and the Natural Environment Research Council and the assistance of the
membersof the Tidal WatersResearchGroup,Schoolof Civil Engineering,in the collection
of data.
5.7 References
Dyer,K.R. 1981:Mixing causedby lateralinternalseichingwithina partiallymixedestuary.Esmarine,

Coastal and Shelf Science 15, 443-457.
Geyer,W.G. & Smith,J.D. 1987: Shearinstabilityin a highlystratifiedestuary.Journalof Physical

Oceanography17, 1668-1679.
Kent,P.E. & Pritchard,D.W. 1959:A testof mixinglengththeoriesin a coastalplainestuary.Journal

of Marine Research 18, 62-72.
Miles,J.W. 1961:On thestabilityof heterogeneous

shearflows.Journalof FluidMechanics10,496-508.

96 E J Darbyshire
andJ R West
New, A.L., Dyer, K.R. & Lewis,R.E. 1987:Internalwavesandintensemixingperiodsin a partially

stratifiedestuary. Estuarine,Coastaland Shelf Science24, 15-33.
Nunes,R.A. & Simpson,

J.H. 1985: Axial convergence
in a well mixedestuary.Estuarine,
Coastaland
Shelf Science 20, 637-649.
Odd, N.V.M. & Rodger,J.G. 1978: Verticalmixingin stratifiedtidal flows.Proc.ASCE 104(HY-3),

337-351.
Partch,E.N. & Smith,J.D. 1978:Timedependant

mixingin a saltwedgeestuary.
Estuarine
andCoastal
Marine Science 6, 3-19.
Rossby,C.G. & Montgomery,R.B. 1935: The layerof frictionalinfluencein wind andoceancurrents.

Pap. Phys.Oceanogr.Meterol. 3(3), 1-101.
Soulsby,R.L. 1980:Selectingrecordlengthanddigitizationratefor nearbedturbulencemeasurements.

Journalof PhysicalOceanography 10, 208-219.
Turner,J.S. 1973: Buoyancy

Effectsin Fluids.Cambridge
UniversityPress367pp.
West,J.R,Knight,D.W. & Shiono,

K. 1985:A noteonthedetermination
of verticalturbulent
transport
coefficientsin a partiallymixedestuary.Proceedings
of the Institutionof Civil
Engineers,Part 2 79, 235-246.
Appendix1 - Internalwave theory

Assuminga flow is irrotationaland incompressible, ignoringviscosityand frictionand
assuming thatanywaveswill beof a periodwhichis shortcompared withtheinertialperiod,
of the orderof the periodof the earth'srotation,its motioncanbe describexlby the four
followingequations,
LeBlond& Mysak(1978).
au aw
• + = 0 (5.A1)
ax az
ap+waPO
.0 (5.A2)
at az

E J DarbyshireandJ R West 97
au ap (5.A3)
øø3•' ax
aw.
at - Op
Po Oz_ pg (5.A4)
Here, P0 is the fluid densityat the equilibriumposition. Eliminatingp and p from these
equationsgivesthe followingequationfor the verticalvelocity,w
ilw N2
i5[iS2w
Oaw
ax: N
at• az• s'2• )+ •v
+
ax2
- 0 (5.A5)
•vhereN2= (-g/p0)0Po/bZ)
radss'•. If it is assumed thattheflowcanbe approximatedto a
layerof constant
buoyancyfrequency, i.e./}Na/bz-•,corresponding
toa density
profilewhich
is nearlinearover the waterdepth,thensolutionsexistof the form,
w- F,•z)e
•(•'•-'00 (5.A6)
Substitutingequation5.A6 in equation5.A5 gives

daW N 2 dW
+ , ;[2(•o2_N2)W - 0 (5.A7)
dz2 g dz
where3.2= k/to2,k = thewavenumber

in thehorizontal
direction
andto= angular
frequency
in rads s'•.
Solutionsto this equationare of the form,
W- W.ß-satt2$
sin(•z) (5.A8)
where•52= 3.2 (N2-to2)-N4/4g,

andN4/4gis smallcompared
to theothertermsandcanbe
neglected.Imposingtheboundaryconditionof no verticalmotionat thefree surface,i.e. W=0
at z=D, gives,
t•D- n• (5.A9)
wheren is an integer. Considering

the casewheren=l the followingdispersion
relationis
obtained,

a•z - (5.A10)
In a movingflow the frequencyseenby a fixed observeris, o• = • + Uk. In this equation

O•ois definedas the frequencyrelativeto the flow which has,until now, beenreferredto as
simplyo•._From
now on o• will be usedto denotethe frequency
observed
by a stationary
observer.U is the meanlongitudinalflow velocityand is positivein the directionof wave
propagation.Substituting
in equation5.A10 andrearranginggives,
•2Dak4
- 2•o•DZk
3+(½ozDz+•ff•a-D2Hz)Icz-2•o•2k+½o:•
'0 (5.All)
To calculateparticlevelocities'itis necessary
to returnto equation5.A8 andconsiderthevalue
of W, whichfrom now on will be referredto asWo, the verticalvelocityamplitude.For the
purposes
offurther
analysis
it isassumed
thate'•:saz
= 1 since
bothN andz aresmall.If
changesin time of the meansalinityare sufficientlysmallto be ignoredthen,
= (5.A12)
At At 0z
Thisallowscomputation
of thewavedisplacement
amplitudein theverticaldirection,Zo= So
(•}SfOz)
'•. If a singlefrequency
is considered,
thenequation
5.A12canbe written
dS
• = w• (5.A13)
dt
To obtainw, sinusoidalvariationof thesalinityo_ftheform,S(t)= S+Sosin(2not)isassumed

whereS(t) is thesalinityasa functionof time,S is themeansalinityandSois theamplitude
of the salinityvariation.This gives
(5.A14)
giving,
(5.A15)
thusallowingevaluationof w.
Wois a functionof z asdescribed

by eqn.5.A8 thesimplifiedformof whichis,

(5.A16)
Using eqns.5.A2, 5.A6 and4.A16 this leadsto
W• Wo•l(kz-•øt) (5.A17)
(5.A18)
u- -• wo-•tm(/5•)
giving,
(5.A19)
•rø'• tan(O•)
whereUo is the amplitudeof the longitudinalvelocity.

Dynamically-active models in the

prediction of estuarine stratification.
J H Simpsonand J Sharples
Abstract
Stratificationin estuariesand ROFI's (regions of freshwaterinfluence) exerts a critical

environmentalcontrol and frequentlyexhibits large variability. Predictionis inherently
difficult becauseof the 1ocalisednatureof the buoyancysourceand the complexsuite of
interactingprocesses that controlwatercolumnstability.
To advancebeyondthefirst orderaccountof thevariabilityof estuarinestratification

given
by the prescriptivemodel(Simpson, Sharplesand Rippeth(1991)), it is necessaryto have
recourseto a dynamically-activemodel in which the profiles of velocityand densityare
computedfrom forcingby surfaceslopesand densitygradients.The model describedhere
employsa Mellor andYamadalevel-twoclosureschemeto representthe interactionbetween
the density4riven and tidal flows via the frictional stressesand their dependenceon
stratification.
Theperformance of themodelis considered in relationto thatof theprescriptive

modeland
comparedwith recenttime seriesobservations from Liverlxx)l Bay. A singlecomponent
modelwithoutrotationwill reproducerealisticsemidiurnaland semi-monthly cycles,but the
latteronlywhenthestirringis artificiallyincreased. As with theprescriptivemodel,it is found
thattheinclusionof theeffectsof rotationin a twocomponent modelovercomes thisdifficulty
by reducingthe cross-isopycnal flow. Simulationsof the springs-neaps cycle with the
two-component T-C modelusingrealisticforcingfor the tidal flow and the densitygradient
showencouragingagreementwith the observations.
6.1 Introduction
Water columnstabilityis a key factorin determiningthe environmentbothin estuariesand

in adjacentregionso_ffreshwateri_nfluence(ROFI's). Observations
of its time-dependence

102 J H SimpsonandJ Sharples
suggestit is frequentlyhighlyvariablewith importantcycles on semi-diurnal,semi-monthly

andseasonal time scales(e.g.Haas(1977), Simpsonet al. (1990)) with moreerraticchanges
associated with episodicwind mixing (Goodrichet al. (1988)).
Stratificationinhibitsturbulentmotionsand thusrestrictsverticaltransferin a way which

is knownto be effectivebutis notunderstood in detailandhasproveddifficultto parametrise.
Episodesof stratificationmay lead to pronouncedhypoxiawith importantbiologicaland
chemicalconsequences. The level of stratification
may alsoact as a criticalcontrolon the
supplyof nutrientsto theupperpartof thewatercolumnfromre-mineralisation sourcesin the
seabedand influencethe light climatethroughits effecton the re-suspension of sediments;
both theseeffects may have a significantbearingon primary productionin the areas
concerned.
Understandingand, eventually,predictingthe developmentand decay of estuarine

stratificationmust, therefore,be an importantgoal for physicaloceanography requiring
complementary programmesof observation and modelling. The difficulty of the problem
arisesin partfrom thefact thattheprimarybuoyancyinputis localisedat oneor moresources
at the coastratherthanbeingdistributed uniformlyoverthe surfaceas in theheating-stirring
competition(which determinesseasonalthermalstratificationover muchof the shelf). In
additionthe suiteof processes involvedin the controlof stratification
interactin a non-linear
way generating variousfeedbackeffectswhichmay influencethe development.
In this paper,we describenew modelsof the evolutionof stratificationwhich allow for

someof the importantfeedbackprocesses thatare operatinganddiscussthe resultsin relation
to recentobservationsof the time historyof stratification
in a ROFI regime.
6.2 Prescriptivemodel
In two previouspapers(Simpsonet al. 1990,Simpson,Sharples
andRippeth(1991))we have
identifiedtheprincipalprocesses
operating topromoteanddestroystratification
andattempted
to representthese processesin simplemodelsof the time developmentof stratification.
Closelysimilarapproaches havebeenusedby NunesVaz et al. (1989)andvanAken(1986),
The stratifying influenceof freshwater buoyancyinputto an estuaryis manifestlocally

throughthe shearflow drivenby thehorizontaldensitygradient(theestuarine circulation).
In additionthereis a significant
periodicinputdueto theactionof tidalshearonthedensity
gradient(so-called tidalstraining)as well as a contribution
dueto surface heatingduring
springand summer.
Thesestratifyinginfluences
areopposedby stirringdueto the tidalflow andwindstress
with additionaleffectsfromconvective
coolingduringthewintermonths.

J H SimpsonandJ Sharples 103
In the prescriptive
model(Simpson,SharplesandRippeth),we specifiedtheseprocesses
usinganalyticalsolutions for thetidalflow (Prandle(1982))anddensitydrivenflow (Hansen
andRattray(1965) or Heaps(1972))andusedenergetics conditionsto describethemixingby
tidalandwind-stirring.The modelis forcedby a depthuniformdensitygradientandsurface
slopeswhichareconstrained by a no net flux condition(in the downgradientdirection).The
justificationfor the strongassumptionof a depth-independent densitygradient,which
represents an importantsimplification of theproblem,hasto be foundmainlya posteriorior
in reference to observations for the area concerned.
The individualstratifyingandmixinginputsareassumed to actindependently andsotheft

effectsarecombined in a simpleadditivefashion.This linearapproach givesa plausible
first-order
account of observations
of thetime-dependence
of stratification in LiverpoolBay,
reproducing boththesemi-diurnalandsemi-monthly cyclesof stability.It, however,explicitly
neglectswhatmaybe importantinteractions betweenthedensity-driven
andtidal flowswhich
occurthroughthe internalfrictionalstresses
andtheirdependenceon the stratification.
When theseinteractions are introduced,the problembecomesa muchmore difficult(and

interesting)one. For example,insteadof beingprescribed by thePrandleanalyticalsolution
with fixed eddy viscosity,the tidal flow would now be calculatedfor stresses
basedon an
Nz,whichvarieswith stratification, andvelocityshearwhichincludesthedensitydrivenflow.
Likewisethedensity-driven flow wouldbe modulated withthecycleof tidalstirringasin the
laboratory experimentsof Lindenand Simpson(1988).At the sametimethe efficiencyof
mixing, held constantin the presciptivemodel,would be free to vary with the level of
stratification and shear.
6.3 A dynamicallyactivemodel
We aimfor thesimplest
profilemodelthatcanadequately
represent
theseinteractions
between
the density-drivenand tidal flows.
Initiallywe considerflow in the x directiononly with no rotationso the dynamicsare

governedby the equation
a._Z av. (6.1)
wherewe useconventional
notationwith z positiveupwardsandz=Oat theseabedandz=h
at the surface.
The pressure
p is hydrostatic
andrepresented
by

gfs• pdz (6.2)

whereq = surfaceelevationand p(z) = density.
Assumingthatq/h << 1, thepressure

gradientis approximately
(6.3)
•xx' p ax ox
wherewe havefurtherassumed that3p/3x is independent

of depth,an assumption whichis
consistent
with the prescriptive
modeland hasthe advantage of simplifyingthe numerical
solution.
The shear stress is written in the usual form
au (6.4)
• - -Jr,paz
wherethe turbulenteddy viscosityN. is to be determinedfrom the closurescheme.At the
bottomboundarythe stressis setequalto
whereu• is the near-bottomvelocity.
The flow is forced in three ways:
(i) by the impositionof a depthuniformdensitygradientwhichmay alsobe (thoughnot

necessarily)constantin time.
(ii) by a meansurfacegradientof the form:

h Op (6.6)
Ox p 0x
in which the parameter7 may be chosento make the net transportzero.
(iii) by an oscillatingpressuregradientfor eachtidal constituent

givinga sumof the form:
N
g&l. gi• +y•A,cos(%

t+•,) (6.7)

The densityevolutionis controlledby horizontaladvectionand verticaldiffusionin the

equation:
ap . -u +• (6.8)
at Ox cgz
whereI• is the eddydiffusivity.Like N,, thediffusivityis to be determined

from closure
procedures.To do this we may make use of a level 2 formulationbasedon Mellor and
Yamada(1974)which assumeslocal equilibriumbetweenthe rate of productionof turbulent
kineticenergy(TKE) and its dissipation
in heatand work doneagainstbuoyancyforces.
The transfer coefficients are written as:
(6.9)
wheret is a mixinglengthwhichhasto be specifiedandthe stabilityfunctionsq>m

and• are
functions of the the flux Richardson number Rf-
(6.10)
whereRfc is a criticalvalueof Rf andRf• andRf2 are conslants.Pr is the turbulentPrandtl

numberfor homogeneous conditionsand is hereassumedequalto unity.
The flux Richardsonnumberis found from the local gradientRichardsonnumberRi

accordingto:
(6.11)
Rf- • A*Ri - . Ri - 4Rf•R
We have taken the the requiredconstants(Rf,,Rf•,Rf2)froma recentrecalibmtionof the

Mellor-Yamadaschemeby J Hamrick(personal communication) which gives the rounded
values:

I I I (6.12)
The mixinglengthI mustbe specifiedindependently

andaccordingto Melior andYamada
is the mostarbitraryaspectof the method. We haveemployedthe widely-usedform (e.g.
Davies (1990)):
•. Kz(l_z/h)• (6.13)
whichbehaveslike Kz near the bottomboundary(K =von Karmansk) and falls off rapidly
near the surface.
In this level 2 approach,thereis no needfor a separatedifferentialequationfor the TKE

which can be obtainedexplicitlyfrom the equilibriumrelation:
(6.14)
whereq = turbulentintensityandB• is a constant.
The model is readily extendedto two components

of flow so that the effectsof rotation
may be includedwhereappropriate. The two dynamicalequationsare now of the form:
all
- y,-x-xn-o au
ax pax az• •az]
(6.15)
a,,
a'•- -fu-xay-•$
(n-o a,,'
-pay +az[ •az,
andtheforcingincludestwo components of densitygradientandtidalpressure
gradient.We
retainthecondition
of no-net-fluxnormalto theisopycnals
andagainsetthebottomstress in
termsof thequadraticdraglaw now in the form:
x/•- -kp
UuI ; U- (Ul
2+v12)
m (6.16)
x• - -kp UvI
whereu• andv1 are the components

of near-bottom
velocity.

J H Simpsonand J Sharples 107
6.4 Results
6.4.1 Non-rotatingmodel
The unidirectionalmodelwas implementedusingan explicitfmite differenceschemeand a

time stepsatisfyingthe stabilitycondition
At< I (AO: (6.17)

2
Two independent versionswerewrittenin differentlanguages andtestedagainsteachother

andagainstanalyticalsolutions for constantviscosityanddiffusivity.(e.g.theHansen-Rattray
estuarineflow solution). The model was then run with driving forcescorresponding to the
locationof the observations in LiverpoolBay (Simpson,SharplesandRippeth).The result
wasthatthe stratification increasedcontinuously andindefinitely.To preventthisrun-away
stratification,
the tidal forcinghasbeenincreasedsomewhatandthedensitygradientreduced
to increasethe ratio of stirringto stratifyinginputs.(Suchan adjustment wasfoundnecessary
in a previousnon-rotatingmodel(Simpsonet al. 1990) whereit waseffectedby an increase
of Nz).
With thesemodifiedinputs, the behaviourof the modelin termsof the surface-bottom

densitydifferenceshows(fig lb) the samequalitativefeaturesas theobservations; thereis a
clear switchingbetweenstratifiedconditionsfor about8 daysfollowingneapsand almost
continuouslymixedconditionsfor the restof the fortnightlycycle. On top of thispatternis
superposed a regularcycle of semi-diurnalfluctuations(clearlyapparentin the data) with a
maximum
amplitude
of A6t= 0.15kg/m3. Theno-net-transport
condition
wasfoundto be
met with a value of •, closeto 0.3.
A comparison with the outputof the prescriptivemodelfor similarforcing(fig lb) shows

the two modelsbehavingin analagous fashionwith similarperiodsof enduringstratification.
There is, however,a significantdifferencein that the T-C model exhibitsa more rapid
increasein stratification
followingneapsandattainsa maximumstabilitymorethan4 times
thatof theprescriptivemodel,presumably becauseof thereducedefficiencyof mixingin the
stratifiedregime. Thereare alsosignificantdifferencesbetweenthe modelsin the timingof
the onsetandbreakdownof the enduringstratification.
This single-componentmodel will providethe basisfor predictingstratificationin the

interiorof narrowestuarieswherethe flow is effectivelyunidirectional.For the presently
availabledatafrom LiverpoolBay wheretheflow is certainlyinfluencedby rotation,weshall
proceedto a comparisonwith a two componentmodel in which the Coriolis forces are
included.

_iVVVVVVVVVVVVV
1.5
1.0
(b)
0.5
0.0
0 2 4 6 8 10 12 14 16 18 20 22 24 26
1.5
1.0
(c)
0.5
0.0
0 2 4 6 8 10 12 14 16 113 20 22 24 26
Figure1' Comparisonbetween
turbulence
closure
andprescriptive
models forsinglecomponent
turbulent
closuremodel.(a) Depthmeantidalcurrent
(m s-1)versus
time(days).(b) Turbulence
closure
model:
bottom-surface
density
difference
(kgm-3)versus
time.(c)Prescriptive
model:
bottom-surface
density
difference
m $-1. versus
time.
(N=--0.08
mes-l).
1D •x=2.5
x10
-8m-1,
depth
=40m,
surface
wind
speed
=5.2
6.4.2 Two componentmodel with rotation
A twocomponent
versionof themodelhasalsobeendeveloped
usingessentially
thesame
numerical
scheme
andtestedby comparing
a constant
eddyviscosity
formagainst
theresults
of theanalytical
calculation
of Heaps(1972)
for thecaseof density
gradient
drivingonly.

After the decayof inertialoscillations,

the modelgivesa steadystatesolutionvery closeto
the analyticalcaseif the parameter? = 0.56. (fig 2).
(a) (b)
; I I I I
•0. 04 -0. o2 0.04 o4 -o. 02 ß O. 04

' 55 ' :55
' 50 , :50
, 25
1
5
0 0
Figure2: Densitycurrentarisingfrom a steadydepth-uniform horizontaldensitygradientin a rotating

frame. (a) Heaps(1972) analyticalsolution.(b) Modelledsolution.Horizontalaxisis currentvelocity
m s4),verticalaxisis heightaboveseabed
0n). Solidlineis u component,
broken
lineis v component.
Nz=0.02
m2s4,1P •x=5.0
x10
-8m4,and
y=0.56.
This two-component model,includingthe turbulentclosureformulation,hasbeenusedin
an attemptto simulatethe springs-neaps cyclein LiverpoolBay for the periodMarch 29 to
April 12 1989,for whicha time-seriesof mooringobservations is available.Forcingof the
modelis by two components of surfaceslope(seetable1 for detailsof thesecomponentsand
otherconstants used)andthehorizontal,depth-uniform densitygradient.The timedependence
of the latterfor the periodconcerned (fig 3d) hasbeendeterminedfrom observation usinga
combinationof 3 quasi-synoptic CTD sectionstakenduringthe deploymentperiodand the
mooringtime-series from whichwe haveinferredthemeangradientoverthetidal excursion.
When run with theserealisticinputs,the model reproduces mostof the featuresof the
observations.The 8 dayperiodof enduringstratification
evidentin themodelresults(fig 3c)
closelyparallelstheobservations
(fig 3b) with goodagreementin thetiming andintensityof

11o J H Simpson
andJ Sharples
themaximumstratification.
Thevariation
inthemagnitude
ofthe semi-diurnal
oscillations
overthetwoweekperiodisalsoreasonably
well-represented
inthemodeloutput.
(a) 0
-1
(b) 1
0 1 2 5 4 5 6 7 8 9 10 11 12 15 14
2
(c) 1
0 1 2 5 4 5 6 7 8 9 10 11 12 13 14
1%
xlO-e lO
5
(d)
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Figure
3:Turbulence
closure
model
comparison
with
Liverpool
Baymooring
data.
(a)Depth
mean
tidal
current
(ms'l)versus
time
(days).
(b)Observed
bottom-surface
density
difference
(kgm'3)versus
time.
(c)Modelledbotto•n-surface
density
difference
versus
time.(d)Observed
variation
ofthedensity
gradient
1P •x(m-')

Table 1: Parameters
usedby the turbulence
closuremodelin figure3(c).
Tidalpressure
gradient
amplitudes
m s'2:
Constituent x-component phase y-component phase
M2 8.469x
10'• 2.415 7.113x10'• 3.749
S2 3.114x10
'• 0.383 2.364x10-• 1.669
N2 1.497x10
's 0.329 1.21lx10-• 1.529
Mean depth= 36.0m.
Depth variationamplitudes:
Constituent amplitude(m) phase
M2 2.52 1.977
S2 0.83 6.127
N2 0.45 6.082
Quadraticfrictioncoefficientk• = 0.003
Mean surfaceslopefactory = 0.55
Wind speedfrom meteorologicalobservations
Numberof depthbins= 10
Time step= 0.005 hr.
6.5 Discussion
In thispaper,we haveproposed singleandtwo component modelsincorporating a form of the

Mellor-Yamadaturbulentclosurescheme,which,becausetheyincorporateessentialfeedback
mechanisms,offer a more fundamentalapproachto the problem of predictingestuarine
stratificationthanourpreviousmodels.To demonstrate thegreaterskill of thesemodelsthan,
for example,theprescriptive model,a carefulcomparison with extensivedatasetson estuarine
stratificationin differentsituations
will be required.A particularpriorityin thisrespectis to
test the single componentmodel in the narrow estuarysituationwhere rotation can be
neglected.
Thereare, however,encouraging indicationsfrom the aboveresultsthatthe T-C modelis

able to convincinglysimulatethe time evolutionin LiverpoolBay. This is a situationwhere,
as wasapparentwith theprescriptive model,the effectsof rotationmustbe included in order
to limit the cross-isopycnal
flow which otherwiseinducesran-awaystratification.
It also appearson the basisof the model trials to date that it is importantto allow for
variationsin thehorizontaldensitygradientwhich,in previoussimplifiedapproaches, we had
treatedasconstantin timeas well asdepth. Estimatesfrom shipCTD sectionsandinferences

112 J H Simpsonand J Sharples
from mooringtime seriesusingthe tidal excursionindicatechangesby a factorof 2-3 over

a formight. One might expectsuchchangesin gradientto be relatedto changesin fiver
discharge,but there are indicationsthat the gradientsmay also be modulatedby the
springs-neaps cycle in responseto the enhanceddensitycurrenttransportat times of low
mixing (i.e.neaps).
The parameter¾,whichis setto give zeronet shorewards transport,accountsin a simple

way for the influenceof bottomfrictionalstresson the meanflow. The presentcondition
assumesan instantaneous responseof sealevel gradientto changesin densitygradientand
mightbe replaced,with advantage, by an adjustment with an appropriatetime delay.Thereis
alsothequestionof the extentto which¾dependson theforcingparameters; experiments with
the current model indicate that its value varies only slowly with changesin the density
gradient.
An obviouslimitationof the modelat presentis the lack of provisionfor surfaceheat

exchange.This omissionis not important for the periodof the LiverpoolBay observations
which is near to the equinoxwhen the net surfaceexchangeis small,but generallysurface
buoyancyfluxes will need to be includedwith separateequationsfor the temperatureand
salinityreplacingthe existingdensityequation.
Similarly, as we move towards models which aim for quantititivepredictionsof

stratification,it will be necessaryto utilisea full time historyof the wind forcingratherthan
a singlemeanvector. With suchan inputthe T-C modelshouldallow a properdescription
of theway windslress modifiestheverticalvelocityprofileandhencetheadvectionof density.
Acknowledgement
We are gratefulto JohnHamrickof theVirginiaInstituteof MarineSciencewho generously
providedguidanceandadvicein theapplicationof turbulentclosureschemesduringour stay
at VIMS.
6.6 References
van Aken H.M. (1986) The onsetof seasonalstratificationin shelf seasdue to differential
advectionin the presenceof a salinitygradient.ContinentalShelf Research,5 (4), 475-485.
DaviesA.G.(1990)A modelof theverticalstructureof thewaveandcurrentbottomboundary

layer. ModellingMarine SystemsVol.2, ed. A.M.Davies,CRC Press(inpress).

Goodrich D.M., W.C.Boicourt, P.Hamilton, and D.W.Pritchard (1988) Wind induced

de-stratification
in Chesapeake
Bay. Journalof PhysicalOceanography,
17 (12), 2232-2240.
Haas L.W. (1977) The effect of the spring-neaptidal cycle on the verticalsalinitystructure
of the James,York, and Rappahannockrivers. Estuarineand Coastal Marine Science,5,
485-496.
HansenD.V. and M.Rattray(1965) Gravitationalcirculationin estuaries.Journalof Marine

Research, 23, 104-122.
HeapsN.S. (1972) Estimationof densitycurrentsin the Liveqx)olBay areaof the Irish Sea.
GeophysicalJournalof the Royal AstronomicalSociety,30, 415-432.
LindenP.F. andJ.E.Simpson(1988) Modulatedmixingandfrontogenesis

in shallowseasand
estuaries.ContinentalShelf Research,8 (10), 1107-1127.
Mellor G.L.andT.Yamada(1974) A hierarchyof turbulenceclosuremodelsfor planetary

boundarylayers. Journalof AtmosphericSciences,31,1791-1806.
NunesVaz R.A., G.W.Lennon,and J.R.deSilva Samarasinghe (1989) The negativerole of

turbulencein estuarinemasstransport.Estuarine,Coastaland Shelf Science,28, 361-377.
Prandle D. (1982) The vertical structureof tidal currentsand other oscillatoryflows.

ContinentalShelf Research,1 (2), 191-207.
SimpsonJ.H.,J.Brown,J.P.Mathews, andG.Allen(1990)Tidal straining,densitycurrents,and

stirringin the controlof estuarinestratification.Estuaries,12, 125-132.
SimpsonJ.H., J.Sharples,and T.P.Rippeth(1991) A prescriptivemodel of stratification

inducedby freshwaterrunoff.EstuarineCoastaland Shelf Science,33, 23-35.

Part III
Circulation
Residual flow in Naples Bay and its effect

on constituent concentration, constituent
flux and residence time
J van de Kreeke
Abstract
NaplesBay is a shallow,partiallymixedestuaryin southwest Florida. At its mouth,the

estuarybranches intotwo channels,GordonPassandtheIntra-Coastal Waterway(ICW). Both
channelsconnectto theGulf of Mexicoandthereforeallow throughflowor residualflow. The
residualrateof flow is determinedusingrecordingcurrentmetersin a relativelyuniformand
straightsectionof the ICW. Measurements extendedover a period of one month. The
residualrate of flow is definedas the 40 hr low passfilteredinstantaneousrate of flow. The
monthlymeanvalueis 2.2 m3s
'•. Variations
withperiods
of 5-10dayscanbe aslargeas7
timesthe meanvalue. Variationsin the residualflow are stronglycorrelatedwith wind and
atmosphericpressure.
To evaluate the effect of the residual flow on the constituentconcentration,constituent

fluxesandresidencetimesassociated with contaminant
sources(marinas)in NaplesBay use
is madeof a tidallyaveragedsteadystateadvection-dispersion
model. Valuesof dispersion
coefficients
aredeterminedfromknownfreshwater inflowandlongitudinal
salinitydistribution
andrange
between
10-100m2s
'•. Modelresults
showthatwithintherangeof observed
values,
the residual rate of flow has little effect on constituent concentrations and residence times but
stronglyaffectsthe distributionof the constituent
fluxesover GordonPassandICW.
7..1 Introduction
NaplesBay, includingDollar Bay, is a smallestuarylocatedon the lower southwestcoastof

Florida;Fig. 1. It connects
to theGulf of Mexicovia GordonPassandconnects to Rookery
Bay via the Intra-CoastalWaterway(ICW). WhereasNaplesBay is an urbanestuary
surrounded by development anddredgedresidentialcanals,RookeryBay is still in a pristine
stateandhasbeendesignated asanestuarinesanctuary.
Naturally,thereisconcern
forwaterborne

118 J van de Kreeke
....
i::::
:;4I•SIDENTIAL
CANAL
DEVELOPMENT
I SECtiON
OF
I•SIDUAL
FLOW
MEASUREMENTS
• SECTION
OFWAIER-
AND
SAI.I
FLUXMEASUI•EMENTS
0
[]•'OME'r•
mESSU•
.....
Figure1' NaplesBay andlocationof measurement

stations
Figure2: Circulation
ill Estuarywifll MultipleMouths;PlanView

J van de Kreeke 119
pollutantsassociatedwith developmentto be carriedto Rookery Bay via the ICW. In

particular,thisstudyis partof theevaluationof a proposed
development, includinga 600 boat
marinaat the junctionof NaplesBay, GordonPassand Dollar Bay. Specificissuesto be
addressed are the concentrations in NaplesBay and the constituent
fluxestowardsRookery
Bay resultingfrom a steadycontaminantsourceat the siteof theproposedmarina. Also, from
a pointof view of contaminant concentration,
are therebettermarinalocationsin the general
Naples Bay area?
It seemsobviousthatpotentiallythe residualflow in the Intra-CoastalWaterwayplaysan

importantrole in determiningthe constituentconcentrations in NaplesBay and constituent
fluxes towardsRookeryBay. Therefore,the focusof this studyis on the magnitudeand
forcingof theresidualflow in thischannel,its importance as a transportmechanismrelative
to dispersionand its effect on concentrations,
constituentfluxesand residencetimes. The
problemat handis notunlikethatencountered in estuariesthathavemultiplemouths;seeFig.
2. Examplesof the typeof estuaryare the Hooghlyestuaryin IndiaandtheEmsestuaryin
The Netherlands.The circulationaroundtheislandis a potentiallyimportantmechanism for
the exchangeof waterbetweenestuaryandocean.
7.2 Physicalcharacteristics
Watersin thestudyareaare very shallowwith typicaldepthsrangingbetween2 m and5 m.
As a resultof dredging,NaplesBay hassomedeeperareasup to 8 m.
The primarywatermovementin the bay is forcedby the tide in the Gulf of Mexico. A
typicalrecordof waterlevelsobserved at a tidestationat thesouthernboundary of DollarBay
is presented
in Fig. 3. For the locationof the tide station,seeFig. 1. The fidesare of the
mixedtype. The averagetidal rangeand the averagespringtidal rangeoff GordonPass
respectively
are0.64 m and0.85 m. (NOAA tidetables).Whengoingtowardstheheadof
NaplesBay,tidalrangesincrease slightlyandreacha maximum in themiddleof NaplesBay,
andafterthatslightlydecrease (Simpson 1979). Thispatternis a resultof a sudden change
in geometryhalfway betweenGordonPassand the head.
•" J/'Vv',VV/V//V/•
!v
-500vV•VV-lV/•l
1989
v' •VV/•/VV
MaJ3 M•6
'!1'M•9 Mg12 •15 M218 M'21
• M'2a
• M• 23 M• 30
•
Apr
Figure3: Observed
Waterlevels
at Tide Stationin DollarBay

120 J van de Kreeke
Duringflood,waterentersGordonPassandflowsintoNaplesBayandDollarBay. During
ebb thispatternreverses.The tidal maximumof the currentspeedin GordonPassvaries
between 0.6 ms4 and1 ms4 andin NaplesBay,DollarBay,andtheIntra-Coastal Waterway
variesbetween 0.2 ms4 and0.5 ms'l, theexactvaluedepending onthephaseof thetide. For
averagetide conditions, a particleenteringGordonPassat the beginningof the floodcould
travelup to the middleof NaplesBay or whenenteringDollar Bay couldreachthe southern
boundaryof this bay.
The mainsourceof freshwaterinflow in NaplesBay is theGoldenGateCanal. Freshwater

entersthe estuaryvia a weir. For the period November 1, 1976 - November 1, 1977, the
monthlymeanvaluesof thefreshwaterinflow arepresented in Fig. 4. Duringthe wet season
therateof flowreachesa maximum of 23 m3/s.Typically,
duringthedryseason(November-
June)the rate of flow is 1 m3/s. The corresponding steadystatelongitudinal
salinity
distribution
is presented
in Fig. 5. Thesalinitydistribution
isbasedonobservations
in stations
alongtheaxisof theestuary.The salinitydistribution showsthe typicalS-shape.Duringthe
dry seasonthe estuarycan be classifiedas mixedto partiallymixed;on a tidallyaveraged
basis,differencesbetweenbottomand surfacesalinitiesdo not exceed2-3 ppt. (Simpson,
•979).
Processesresponsiblefor the salt intrusionand, for that matter, the transportof any
dissolvedconstituent,are dispersionand advection. The dispersionis characterizedby a
dispersioncoefficient. Its valuewill be estimatedusingthe observedsalinitydistributionand
knownfreshwaterinflow. Advectionis associated with the residualflow. For NaplesBay
properthe residualflow equalsthe freshwaterdischarge.For GordonPass/DollarBay/ICW,
the residualflow is not a priori known. Its valuewill be estimatedfrom simultaneous
current
and waterlevelmeasurements carriedout over a periodof one monthin cross-section I, see
Fig. 1.
7.3 Residual flows
7.3.1 Definitions
Assuminga relativeuniformdistributionover the cross-section

Q(O= A(OV(O (7.1)
in which
Q = rate of flow
A = cross-sectional
area(m2)
V = axialvelocityobservedby a currentmeterin thecross-section
(ms')
t = time

J van de Kreeke 121
Writing
A(O = Ao + wq(o (7.2)
in which
Ao = cross-sectional
area below mean sealevel
W = width of the cross-section
q = watefievel variation relative to Still Water Level
and substituting
in Eq. (7.1), it follows
<O> <qV>
(7.3)
In Eq. (7.3), h,v= AJW and< > designates

low passfilteringwith a periodof 40 hrs.
In the contextof this paperthe low passfiltereddatais referredto as residuals.The left

handsideof Eq. (7.3) representsthe transportvelocitywhereasthe first ands•ond termon
therighthandsiderespectivelyrepresent theEulerianmeanvelocityasobserved by thecurrent
meterand the verticallyaveragedStokesvelocity.
25
2O
OCT DEC FEB APRIL JUNE AUG OCT

1977
Figure4: Mean MonthlyDischargeof GoldenGateCanal.

122 J van de Kreeke
34-
'" 32-
DECEMBER
28,
1976
'" 313-
;>
>'- 28-
::d
@ 12' 26-
z
<•: 24-
>-z
• 22-
z
_o 20-
18-
• 16-
14-
DISTANCE TO WEIR (KM)
Figure5: ObservedSalinityDistributionin NaplesBay.
7.3.2 Measurements
The determinationof the residualflow from a singlepoint measurement is basedon the

assumption that the observedvelocitiesare representativefor the entirecross-section.
In
practice,thisimpliesthatthecross-sectional
distribution
of thetidalvelocities
andtheEulerian
meancurrenthasto be relativelyuniform. To satisfyasmuchaspossiblethiscondition,the
cross-section in which the measurements were carded out was selected in a uniform stretch
of the Intra-CoastalWaterway;seeFig. 1. A secondrecordingcurrentmeterwasplacedin
the samecross-section to gain informationon the cross-sectional
variabilityof the current.
In addition,measurementswith hand held currentmeterswere carded out on March 10, 16,
and 24, 1989. Measurementsextendedover threehoursand includedebb- and flood currents.
Velocityprofilesweremeasured in four slationsin the cross-section
locatedat 20 m, 35 m,
45 rn and60 m; seeFig. 6. Currentspeedwasmeasuredat 30 cm intervals,startingat 30 cm
below the surface to 30 cm above the bottom.
To evaluatethe Stokesvelocity,waterlevelsin the cross-section

were measuredusinga
pressuresensor. For cross-section
and locationof the two recordingcurrentmetersand
pressure
sensor,
seeFig.6. At themeasurement
site,thecross-sectional
areaAo= 120m"the
averagedepthh,v= 1.5 rn and thewidthW = 80 m. Measurements
extendedovertheperiod

J van de Kreeke 123
March 3 - March 31, 1989.
Currentmetersare NISKIN wingedtype currentmeters6011 MARK II. Meters were

operatedin the"vector-averaging
mode".A burstsampleconsisting
of 16readingsspaced4
secondsapartwastakenevery5 minutes.The16readings
werevectoraveragedandrecorded
as a northandsouthvelocitycomponent.Currentmeter#348 providedgooddatafor the
entiremeasurementperiod.Currentmeter#350provided
gooddatafor theperiodMarch3 -
March 7, 1989andMarch 16 - March 21, 1989. To determinethe axial currentspeedtheNS-
EW coordinate systemof the #348 currentmeterwasrotatedin increments
of 1ø until the
standarddeviationof one of the velocitycomponents
reacheda minimumfor a rotationof
336 ø. The line 336ø-156ø was taken as the direction of the channel axis. This is close to the
332ø-152 ø direction estimated from the nautical chart.
meters 0.0 35. 45. 80.0

! 1
O.O
154 . • 350 w348•nigmS0

20
Figure6: Measurement
SectionI; Locationof CurrentMetersandTide Gauge
Waterlevelswere derivedfrom the pressureobservations of a GeneralOceanicsModel

6041/51 pressuregauge. Pressurewas recordedevery 15 minutes. In convertingto
waterlevels,
pressureswerecorrectedfor atmospheric pressure usinghourlyobservations
from
the weatherstationat Naplesairport. For locationof station,seeFig. 1.
To determinethe possiblecorrelationbetweenresidualflow and wind, hourlyvaluesof

windspeedanddirectionasobserved by NOAA, NationalWeatherServiceat theNaplesPier
were used. Observations refer to an elevation of Mean Sea Level + 10 m. For location of
the weatherstation,seeFig. 1.
7.3.3 Results
Becausecurrentmeter#348 provideduninterrupted data for the entiremeasurementperiod,

V(0 in Eqs.(7.1) and(7.3) is takenequalto the axial currentspeedof thismeter. The time

124 J van de Kreeke
1oo
oo
-10.0
150
-15o
1.0
o.o
-1.0
10.0
-10.0
!
i i i i i i i i i i i
Mat 3 Max 6 Mar 9 Mar 12 Mar 15 Max 18 Mar 21 Mar 24 Mar 27 Mar 30 Apr '
1989
Figure
7. FromToptoBottom,
Eulerian
Me,'u•
Current,
Water
Level,
Stokes
DriftandTransport
Velocity.Positive
direction
is towards
Naples
Bay.
series
fortheEulerian
mean
current
obtained
bylowpass
filtering
V(t)ispresented
inFig.
7. In thesame
figure,
thelowpass
filtered
timeseries
forthewaterlevel
inthecross-section,
theStokes
velocityandthetransport
velocityarepresented.
All fourtimeseries
exhibit
vm'iations
witha period
of5-10days.Timeseriesaretooshort
todelineate
possible
longer
period
variations
inresiduals.
Becausetidalvariations
in waterlevel
andcurrent
speedare

J van de Kreeke 125
closeto 90ø out of phase,the contributionof subtidalvariationsto the Stokesvelocity is

substantial.Euleriancurrentspeedis roughly5 timesas largeas the Stokesvelocityand is
the majorcontributorto the transportvelocity. The meanoverthe time seriesof the transport
velocityis 1.8 cm s" towards
NaplesBay. The maximum valueof thetransport
velocity
towards NaplesBayis 13 cms-' andthemaximum valuetowardsRookeryBayis 5 cm s'•.
Witha cross-sectionalareaAo= 120m:, thecorresponding
valuesof theresidual
flow are,
2.2 m3s
'• forthemean,15.6m3s'• forthemaximum flowtowardsNaplesBayand6 m3s-• for
the maximumflow towardsRookery Bay.
The 5-10 day variationsin the residualflow of the Intra-CoastalWaterway suggest

meteorologicalforcing. Most likely wind and/or atmosphericpressuregradientscause
differencesin mean sealevelbetweenboth end of the waterway. Direct wind forcing is an
unlikely contributoras the waterwayis protectedby mangroveson both sides. The two
availablemeteorological parameters, localwind velocityandlocalatmospheric pressureshow
a strong(negative)correlation. Taking the axial wind speedto be representativefor
meteorological forcing,it follows from Fig. 8 that a strongcorrelationexistsbetweenthis
parameterand the transportvelocity. A linear regressionanalysisshows a regression
coefficientof 0.85 suggesting that85% of thevariancein thetransportvelocityandtherefore
in the residenceflow canbe explainedby meteorological forcing.
7.4 Effect of residual flow on residence time, constituent

distribution and constituent flux
The time a particlereleasedat a location(x,y) residesin an estuaryis referredto as the transit

time. The transittime dependson the phaseof the tide at which the particleis released.
Whendealingwith a continuous injectionof constituent,theaverageoverthetidalperiodof
the transittimesof the particlesis referredto as the residencetime, %. In general,% is a
functionof x andy. Assumingsteady-state conditions,therelationbetweentheresidencetime
% andthetidallyandbasinaveraged
concentration
• is
x,- <V>
M
•' (7.4)
where<V> is thetidallyaveraged
basinvolumeandM (kg/sec)is thestrength
of thesource.
(van de Kreeke, 1983).
In caseof freshwaterinflow, it followsfrom Equation(7.4)

126 J van de Kreeke
< V• So-S
(7.5)
O•T s•
in whichQsisthefreshwater
discharge,
T isthetidalperiod,
Soisthesaltconcentration
of the
receiving
watersand<V> (So-S-)/So
represents
thevolumeof freshwaterin theestuary.From
the longitudinalsalinity distributionobservedon December28, 1976, see Fig. 5, and
informationon thebathymetrypresented in Simpson(1979), a valueof thetidallyandspatially
averaged
valueof • = 30.6pptwascalculated.
With<V> = 8.73 10• m3 (= tidallyaveraged
volumeof NaplesBay and GordonPass),T = 44,712 secand So-- 34 ppt it followedthat xr
= 19.3 tidal periodsor approximately10 days.
7.0
-7.0
-10.0
i i ß i i i i i ' ' ßi i i i
Mar 3 Mnr 6 Mar 9 Mar 12 Mar 15 M• 18 Max 21 Max 24 M• 27 Mar 30 Apr '
1989
Figure8: From Top to Bottom. Axial Wind Speedand TransportVelocity. Positivedirectionis

towardsNaplesBay.
To evaluate the effect of the residualflow on the residencetime, constituentconcentration

and constituent
flux, a link-nodemodelis used;seeFig. 9. For eachlink the transport
equationis
<O>c - AD d•c
dx
=K (7.6)
in which

J van de Kreeke 127
<Q> = residualrate of flow

c = cross-sectionallyaveragedand low-passfiltered constituentconcentration
A = tidal mean cross-sectional area
D = longitudinaldispersioncoefficient
K = constituent flux
x = longitudinalcoordinate;for positivedirectionseeFig. 9
Figure9: Link-NodeSchematization.
Arrowindicates
positivex-direction.
For eachlink, <Q>, A, D andK are assumed

constant.SolvingEq. 7.6 it follows
<Q>L
(7.7)
c,= cbe Az• + (1-e )
<Q>
in which
ce = constituentconcentrationat end of link

c = constituentconcentration
at beginningof link
L = lengthof link
Making use of the conservationof water and constituentat each node, the valuesof the
concentrations
at the nodesfollow from Eq. (7.7). Valuesof concentrationsin the Gulf of
Mexico,node1, andin RookeryBay, node7, are setequalto zero.

128 J van de Kreeke
Table 1: Link Characteristics
Link A(m2) L(m) D(m2s

4)
5-4 150 4000 10
4-3 600 2500 30
3-2 500 2000 100
2-1 450 1500 100
2-6 300 2800 100
6-7 120 4200 50
Table 2: VolumesAssignedto Nodes
Node <V> (m3)

0.40 106
1.30 106
2.35 106
2.75 106
1.40 106
0.85 106
0.12 106
Takingc in Eq. 7.6 to represent thesalinityandsettingK = 0, estimates of thelongitudinal

dispersion coefficientsfor NaplesBay were madeusingthe observedsteadystatesalinity
distributionof December28, 1976. Basedon theseestimates,valuesof the dispersion
coefficientsselectedfor eachlink are presentedin Table 1 togetherwith valuesof A andL.
Valuesof thedispersion coefficients for GordonPassandDollarBay weretakenequalto the
valuefor lowerNaplesBay. As canbe inferredfrom a studyon the waterandsaltfluxesin
cross-sectionII, seeFig. 1, therelativelylargedispersion
coefficientsin thelinks3-2, 2-1, and
2-6 canbe attributedto trappingof waterin theresidualcanals,phasedifferences in velocities
nearthejunctionandtheproximity(lessthanhalf the tidalexcursion)of theGulf of Mexico
(van de Kreeke, 1982). Similar high valuesof longitudinaldispersioncoefficientswere
reportedby Wolanskiand King, (1990) for HinchinbrookChannel,Australia. This channel
hasroughlythe dimensions of NaplesBay andtrappingis associated with storageof waterin
the mangroveswamprather than in residentialcanals.
As a checkon the valuesof the dispersioncoefficientsthe salinitydistributionwas

recomputed usingthelink-nodemodel. Startingfromthecomputed salinitydistribution,
the
massof freshwaterin NaplesBay wascalculated by assigningthe volumeto eachnodelisted
in Table2. The residence time wasthencalculatedusingEq. 7.4. Thisresultedin the exact
same value as obtainedfrom the observedsalinity distributionand known freshwater
discharge,i.e., x• = 10 days.

J van de Kreeke 129
The model was then used to calculate the constituentconcentrations,constituentfluxes and

residencetimesfor continuous releasesof dye at variousnodes.To testthe sensitivityof these
parameters to the valueof the residualflow, calculationswerecarriedout with <Q> ranging
between-2.2 m3s
-1and6 m3s 'l. <Q> = -2.2 m3s
'• is themeanvalueoverthemeasurement
periodand<Q> = 6 m3s
'l isthemaximum observed residual
flowtowards
RookeryBay. The
resultsof the calculationsfor a continuousdye releaseat node 5 and for residualflows of
-2.2m3s
-1 and6 m3s
-1arepresented
in Tables3 and4. Similarly,fora continuous
dyerelease
at node2 theresultsarepresented
in Tables5 and6. In all threeexperiments,
therateof dye
releasewas1 kg s'l andthedischarge
of theGoldenGateCanalwas1 m3s
'•.
Concentrating
on a constituentdischargeat node5 it followsthata changein residualflow
from-2.2 m3/s
-1to 6 m3/s
'• haslitfieeffecton concentration
distributions
andresidence
times;
see Table 3. However, the distribution of the constituent fluxes over the Intra-Coastal
WaterwayandGordonPassdiffersconsiderably;
seeTable4. For a residualrate of flow of
-2.2 m3s-1theconstituent
flux in theIntra-CoastalWaterwayis 0.02kg s'l andfor a residual
rateof flowof 6m3s
'l theconstituent fluxis 0.18kg s'l. In bothcases
thefluxesaretowards
RookeryBay. Similar conclusions
holdfor a constituentdischargeat node2. Concentrations
andresidencetimesare notvery sensitiveto theresidualflow in theIntra-Coastal
Waterway;
seeTable 5. However,the constituent flux towardsRookeryBay stronglydependson the
residualflow; seeTable 6. Comparisonof the concentrationdistributionsin Tables3 and 4
showsthat a constituentdischargeat node 2 leads to much smallerconcentrations and
residencetimesthana constituentdischargeat node5.
Table3: Concentrations
andResidence
Times;Sourceof 1 kg s4 at node5.
Res. Flow in ICW is -2.2 m3s'• Res. Flow in ICW is 6 m3s'•
node c Residence time c Residence time
5 0.944 0.943
4 0.189 0.188
3 0.068 0.067
2 0.031 10.24days 0.030 10.26days
1 0.000 0.000
6 0.024 0.030
7 0.000 0.000
,
Concentrations
arein kg m-3

130 J van de Kreeke
Table4: Constituent
Fluxes;Sourceof 1 kg s4 at node5.
Res. Flow in ICW is -2.2 m3s4 Res. Flow in ICW is 6 m3s4
Link <Q> <Q>c -A• K <Q> <Q>c -A• K

5-4 1.0 0.74 0.26 1.00 1 0.79 0.21 1.00
4-3 1.0 0.18 0.82 1.00 1 0.13 0.87 1.00
3-2 1.0 0.05 0.95 1.00 1 0.05 0.95 1.00
2-1 3.2 0.05 0.93 0.98 -5 -0.07 0.89 0.82
2-6 -2.2 -0.06 0.08 0.02 6 0.18 0.00 0.18
6-7 -2.2 -0.01 0.03 0.02 6 0.16 0.02 0.18
Fluxesin kg s4. Advective

anddispersive
fluxespertainto middleof reach
Table5: Concentrations
andResidence
Times;Sourceof 1 kg s4 at node2.
Res. Flow in ICW is -2.2 m3s4 Res. Flow in ICW is 6 m3s4
node c Residence time c Residence time
5 0.002 0.002
4 0.025 0.025
3 0.029 0.029
2 0.030 2.36 days 0.030 2.40 days
1 0.000 0.000
6 0.024 0.030
7 0.000 0.000
concentrations
arein kg m4

J van de Kreeke 131
Table6: Constituent
Fluxes;Source
of 1 kg s-• at node2.
Res. Flow in ICW is -2.2 m3s'l Res. Flow in ICW is 6 m3s'•
Link <Q> <Q>c -A• K <Q> <Q>c-A•x K

,
5-4 1.0 0.010 -0.01 0.00 1 0.00 0.00 0.00

4-3 1.0 0.025 -0.025 0.00 1 0.03 -0.03 0.00
3-2 1.0 0.029 -0.03 0.00 1 0.03 -0.03 0.00
2-1 3.2 0.050 0.93 0.98 -5 -0.07 0.89 0.82
2-6 -2.2 -0.050 0.07 0.02 6 0.18 0.00 0.18
6-7 -2.2 -0.013 0.03 0.02 6 0.15 0.03 0.18
Fluxesin kg s4. Advective

anddispersive
fluxespertainto middleof reach
7.5 Discussion
The success
of measuringresidualflow with a singlecurrentmeterdependson theuniformity
of the cross-sectional distributions of the tidal velocities and Eulerian mean current.
Experimental
dataon thedistribution
of thesevelocitiesin a homogeneous
tidalchannelcould
not be found. Theoretical studieson the vertical distribution of tidal velocities, Eulerian mean
current,Stokesvelocityand transportvelocityfor a sealevelcanalare presentedby Huanget
al. (1986) and Wang (1989). Huanget al. (1986) usinga 2DV numericalmodel calculated
the tidal- and mean currentsfor the Chesapeakeand Delaware canal. Wang (1989) presents
an analyticalsolutionfor the mean currents. The resultsof both studiessuggestthat for a
typicalSouthFloridacoastalchannelwith a lengthon theorderof 10 km anda depthof 2-3
m, the mean currentsdo not reversedirectionin the vertical.
In thepresentstudy,thevelocitymeasurementscarriedoutwiththehandheldcurrentmeters
showa practicallyuniformvelocitydistribution
overthe measurementinterval,i.e., between
the free surface -30 cm and the bottom +30 cm. Velocities of the stations at +35 rn and +45m
differ less than 5%. Velocities of the stations at +20 rn and +60 rn are 10-15% lower than
thoseof the two stationsin the middle part of the cross-section.Values of the mean of the
axialcurrentspeedfor theperiodMarch3-7, 1989are6.5 cms'• for currentmeter#350and
5.8 cms'1for currentmeter#348. Themeansof theaxialcurrentspeedfor theperiodMarch
16-21, 1989 are 3.1 cm s'• for currentmeter#350 and 2.3 cm s'• for currentmeter#348.
Althoughdefinitelynotproof,theforegoingtheoreticalandexperimental resultsprovidesome
supportfor theassumption of a relativeuniformvelocitydistribution
andat leastsuggestthat
the orderof magnitudeof the observedresidualflow and the periodof its fluctuationsare
correct.

132 J van de Kreeke
The insensitivityof the constituent

concentrations
andresidencetimesto theresidualflow
betweenGordonPassand RookeryBay in part stemsfrom the fact that in a largeportionof
the estuary,dispersionis the dominanttransportmechanism;seecolumns3, 4, 7 and 8 in
Table 4 and 6.
The selectionof a steadystateratherthana time dependent transportequation,Eq. (7.6),

is in part basedon easeof computation and in part is basedon obtaininga worstcasewith
regardto the constituent flux towardsRookeryBay. It is realizedthat the fluctuations
in the
residualflow are oftentoo shortto reachsteadystate. Therefore,the calculatedconstituent
fluxof 0.18kg s'• towards
Rookery
Bay,seeTables4 and6 should
beconsidered
anupper
limit.
7.6 Conclusions
Basedon a one monthtime seriesof currentspeedand directionit is estimatedthat the

residualrateof flow in theIntra-Coastal
WaterwaybetweenNaplesBay andRookeryBayhas
a valueof 2.2 m3s
'• towardsNaplesBay. Maximum values
duringthemeasurement
period
are 15.6m3s
'• towards NaplesBay and6 m3s
'• towards
RookeryBay. Theresidual
rateof
flow exhibitsvariationswith periodsbetween5 and 10 days. Thesevariationsareassociated
with meteorological forcing. The timeseriesof thecurrentvelocityis tooshortto delineate
the forcingof possiblevariationshavingperiodslongerthan10 days.
The contribution of the Eulerian mean current to the residual rate of flow is about five
timesthecontribution
of the Stokesvelocity.
The influenceof theresidualrateof flow in theIntra-Coastal Waterwayonconcentrations

andresidence timesis small. Thisis a resultof therelativelylargedispersive
transport.The
residualflow stronglyaffectsthe distributionof theconstituentflux betweenGordonPassand
Rookery Bay.
Residualratesof flow are derivedfrom onepointmeasurements of currentspeedand

direction. The paperpointsout the shortcomings
in this methodand the needfor a more
sophisticatedmeasurementtechnique.
From a waterqualitypointof view, i.e., low constituent

concentrations,
it is betterto locate
the proposed
marinanearthejunctionof GordonPass,DollarBay, andNaplesBay than
furtherlandwardin NaplesBay.

J van de Kreeke 133
7.7 References
Huang, P.S., D.P. Wang, and T.O. Najarian, 1986. Analysis of Residual Currentsusing a Two-
DimensionalModel In: Physicsof ShallowEstuariesandBays,J. van de Kreeke(ed). Lecturenoteson
Coastal and Estuarine Studies No. 16, 71-80.
Simpson,B. L., Ed. 1979. The NaplesBay Study.CollierCountyConservancy.

P.O. Box 2136, Naples,
Florida.
van de Kreeke,J. 1982. Salt and Water Flux in NaplesBay, Florida. UniversityMiami Rosenstiel
Schoolof Marine and AtmosphericScience. TechnicalReportTR82-2 pp. 28.
van de Kreeke,J. 1983. Residence

Time:Applicationto SmallBoatBasins,Joumalof Waterway,Port,
Coastaland OceanEngineering.Vol. 109, No. 4, pp 416-428.
Wolanski,E. andB. King, 1990. Trappingin Hinchinbrook Channel.A GiantMangroveSwamp.To

be publishedin proceedings
of 5th InternationalBiennialConference
Physicsof EstuariesandCoastal
Seas.
Wong, K.C., 1989. Tidally GeneratedResidualCurrentsin a Sea Level Canal or Tidal Strait with
ConstantBreadthand Depth. Journalof GeophysicalResearch,Vol. 94, No. 6, 8179-8192.

The stratified hydrodynamics of the

Palmiet - a prototypical bar-built estuary
J L Largier, J H Slinger and S Taljaard
Abstract
Tidal inlrusionandmixingin thehighlystratifiedPalmietEstuaryarediscussed asan example

of a typicalsmall bar-builtestuary. The flux of salt throughthe estuaryis dominatedby
advectionowing to the strong gravitationalcirculation. This verticalcirculationpatternis
characterized by four processes.The densesalinewaterentersthe estuaryas a hydraulically
controlledflow througha constricted entrancechannel. Beyondan internalhydraulicjump,
landwardof the flood tidal delta, the salineintrusionpropagatesas a non-mixingbottom
densitycurrentand eventuallyreplenishes deepbasalreservoirs of salinewaterin the upper
estuary. On the ebb tide this lower salinelayer is preventedfrom flowing seawardby the
presenceof the sill at the mouth(upstreamblocking)- upwardentrainmentof haloclinewater
occurs via subcritical shear and the sub-halocline water is not removed. The entrained salt is
exportedto seaby way of the outflowingbuoyantsurfacelayer.
This patternof circulationis expectedto be presentin countlesssmall bar-builtestuaries

aroundthe world. While thesesystemsare generallycharacterized by micro-tidaland high
wave-energy coasts,theyare observed in a varietyof othersituations.In thePalmietEstuary
thispatternoccursduringboththe drier summerand the wetterwinter,exhibitinga certain
robustness to seasonalchanges.Theseestuaries are generallyclassifiedas highlystratified,
although theactualstratificationvariesasa functionof season, tideandpositionin theestuary.
Of consequence is thelongresidence timesexpected, andobserved, for thedeepsalinewater
found in the scourholes of the upper estuary. With decreasingfiver flow and more
constrictedconnections with the sea,tidal mixingdecreases andthisdeepbasalwatercanbe
residentlong enoughto becomeanoxic.
8.1 Introduction
The tidalflow throughthemouthof manyestuaries is restricted

by thepresenceof a sill or
a contraction.Most commonis the combinedsill andcontraction whichis providedby a

136 J L Largier,J H Slingerand S Taljaard
wave-builtsandspit (the "bar-built"estuary). The longshoretransportof sandseeksto close

the mouth and attainsa dynamicequilibriumwith the tidal scour,which increasesas the
mouthnarrows. A linearrelationshiphasbeenfoundbetweenthe cross-sectional areaof the
mouthand the volumeof the tidal prism (O'Brien, 1969). The restrictionof the tidal flow
resultsin shorterflood tideswhich are characterized by largervelocities(in the presenceof
low fiver flow) anda greatertransportof sedimentinto the estuary.This is enhancedby the
suspension of sedimentsin the surf-zone. There is a net accumulationof sandin the estuary
duringtheseperiodsof low fiver flow. A floodtidal delta(deposition region)formslandward
of the narrowmouthchannelandfloodandebbchannelsdevelopowingto the strongresidual
transportof sand. Scouringof theestuaryandmouthoccurduringperiodsof strongfiver flow
when ebb tidal currentsexceedthoseon the flood fide. This hydrodynamically balanced
controlof sedimentation is attainedin a varietyof tidal channels
whichare subjectto scouting
and shoaling,bothnaturaland man-made.Given the correctsupplyof massandbuoyancy,
stronglystratifiedflow andstructureoccurin thebasinlandwardof thisrestrictedflow through
the mouth,as is observedin the PalmietEstuary.
The Palmiet Estuaryis a small, drownedriver valley on the southcoastof Africa. Its
mouthis a narrowchannelthrougha largewave-builtsandspit(Figure1). The heightof the
sill increasesandthe widthof the contraction decreases throughthe dry summerseason,both
beingscouredopenduringthe winterrains. The strongfiver flow duringwinterstormsalso
scoursdeeperregionsin the bendsof the upperestuary(Figure 1). Althoughthe mouthis
usuallyopen,on averageit closesfor abouta monthat the end of everyothersummer.As a
resultof its small size and owing to its undisturbed,
naturalcharacter,this estuaryhasbeen
the siteof a numberof investigations in ecology(BranchandDay, 1984;Taljaardet al., 1986;
CSIR, 1987) and hydrodynamics (Largier, 1986; Slingerand Largier, 1989; Largier and
Taljaard, 1990). The basinis small (Figure 1) with immediateconnections to the sea and
fiver. The fiver entersthe estuaryas a fast-flowingmountainstream,low in suspended and
dissolved matter.
Thispaperaddresses the natureof theflood-tidalintrusionof seawaterandits subsequent

propagationup the estuaryas a gravity-drivendensitycurrent.On encountering the scour
holes,this denseinflow formsa deepwater masswhichmay exhibitvery long residence
times. The paperalso addresses the upstreamblockingand sub-criticalentrainment of the
lowerlayerduringthe ebbtidal outflow. Seasonal examples are discussed andcompared -
illustrating
thegeneralnatureof theintrusion anderosion mechanisms in thePalmietEstuary.
It is arguedthatthesedynamicsshouldbe commonto manyof the estuaries on highwave-
energycoasts(e.g., southernAfrica, southernAustralia,westernAmerica,etc.). Someof the
implicationsof this stratificationand circulationare discussed
in the final section.
8.2 The Winter example

Followingheavyrain, typicallydueto the winterpassage
of mid-latitudefrontalstorms,the

J L Largier,J H Slingerand S Taljaard 137
riverflowsstrongly
(e.g.,80m3s
4 on14August
1986)andthePalmiet
Estuary
iscompletely
flushedof anysalt. As thefiver flow decreases
and/orthetideincreases,
newseawater may
entertheestuary
onthefloodtide(Figure2), asit didon 19August(riverflow30 m3s
'•, tidal
range1.7 m in the estuary). Approaching
the subsequent
neaptide (tidal range0.5 m), the
riverflowincreased
againto 40 m3s
4 on26 August
andall theseawater
wasremoved.Of
interesthereis that periodbetween19 and 26 Augustduringwhichthe estuarywas salt-
stratified.
We discuss threeprocesses:
thetidalintrusion
front(plungingflow);thegravity-
drivenbottomdensitycurrent(saltwedge),anderosionof thehaloclineby ebbtidaloutflow.
Figure 1: The morphologyof the PahnietEstuary,locatedon the southcoastof Africa. These

bathymetricdatawerecollectedduringwinter(August1986)- isobaths in metresbelowmeansealevel.
In summer theshapeis essentially
thestoneexceptfor theappearance of a largerfloodtidaldeltasouth
andwestof station0 anda tidalscourholearoundstation0. The presence of fluvialscourholesof 3
to 4qn depthwereconfirmedby a later surveyof the curvesin the chm•nellandwardof station3.
TheTidalIntrusion Front: Whenthenetoutflow(fiveroutflowminustidalinflow)drops

belowa certainvalue,whichisdependent
onthedensity difference
of thewatertypesandthe
sizeof theconstricted
mouth(Largier,
1990),thesalinewatercanentertheestuary asa dense
underflow.Thisoccurswhilethelessdenseestuarine wateris stillflowingout. Thistwo-
layer
exchange
flowpersists
through
zero
netflow
toanetinflow
characterized
byF4=0.3,
where
F4isaninflow
Froude
number

138 J L Largier,J H SlingerandS Taljaard
/•r3D 2
•0•0
andQois thenetinflowat themouth,g' is thereducedgravity,Hois thedepthandBois the

widthat themostconstricted
section(Largier,1990). The flow is hydraulically
controlled
by
atwo-layer
Froude
composite
numbers
Froude
number
foreach
G2=
layer
and
F•+F22,
F/2
where
= u/2
F12
andF22
are
/ gh;,where
the
densimetri
uisthefluid
velocity
andhisthe
layer depth(Armi, 1986). For inflow Froudenumbersin excessof 0.3 the ebbingsurface
layer is blocked(Ul = 0) and a plungeline appears.A surfaceslickor foam-linemarksthe
surfaceconvergence characteristic
of a tidal intrusionfront.
Station No.
0 1 2 3 4 5
I I I I I
Figure2: A longitudinal
sectionof thesalinitydistribution
in thewaterchannel
duringfloodtideon
19 August1986 (seeFigure 1 for stationlocations).
While thefirstintrusionof seawaterwasobservedon 19 August(Figure2), we obtained

themostdetailedmeasurements on 20 August.In thepresence of a 15 m3s
4 fiverflow, the
tidetumsat 1235,threehoursafterlow tidein theocean(0941). Meanwhile,thedensesaline
water startedits subsurfaceintrusionat about 1200 when the net outflow was 3.4 m3s4. Weak
surfacefoamlines(tidelines),whichappearto markthesubsurface convergenceat thehead
of the salinityintrusion,werefirst noticedaroundstation0 (Figure1) at 1210. In the

J L Largier,J H SlingerandS Taljaard 139
narrowestsectionof the mouth,at about1230, the upperlayer (13.8øC,1.2%o) was flowing

outat 0.3 to 0.4 m s'• - varyingasa functionof surf-zone
dynamics
(Figure3). The lower
layer(15.2øC, 26%0) wasflowingin at 0.1 m s'•. Theinterfaceoccurred
at mid-depth (0.5
m), but roseand fell with fluctuationsin the net flow as a resultof surfacegravity and
infmgravity
wavesentering
themouth.Thetwo-layer
Froude
number
G2is of orderunity,
indicative
oftheinternal
hydraulic
control
at,this
section.
Atstation
0 (at1245)
the0.5m
basal
flowissupercritical
(F•= 2.3)with
g = 0.14m2s
4 and
u2--0.4ms4. Somewhere
betweenstations0 and 1, where the basal flow is subcritical(Figure 3), the intrusion
experiencesa hydraulicjump with the attendantmixing. Turbulentmixingis alsoobserved
in the mouthand at station0 as a resultof supercriticalshear.
1.0
• 1625
(m)
1610 •1310 ;- 1320
1335
..-r 1445
'•u•.
1355/17•0
(1445)
<•
IL.... (1730)
½ l
-1245
M.S.L.
--'-•1•5)•
<-•
-I.0
uthl
'
1320
-2O
l•O
-3 0
Sta. 0 Sta. 1 Sta. 2 Sta. 5
.40L , M• I
0.2m/s 30%o
I i I I I Stla'
31 I [ st:'
4I • I I
Figure3: Verticalprofilesof salinityandvelocityduringthe floodtide on 20 August1986. The dots

andarrowsdisplaythe actualmeasureddata. The cloud-likefeatureson theprofilesindicateturbulent
mixing.Arrowspointingto theleft correspondto seawardflow. A circlearounda salinitydotrepresents
a near-zero
velocity(< 0.02 m s'•).Timesarenotedalongtheprofilesandat thelevelof thewater
surface.
By 1300 strongfoam lines,characteristic

of plunging,are seenat the seawardend of the
mouthchannel- upstreamof the controlsection. This indicatesa net inflow of between6.7
and12.2m3s '•, asobserved.
However,asQoincreases,
sodoesHoandstronginflowis only
attainedfor severalminutes
between1300and 1330whenQoexceeds15.7m3s4 andHo =
1.3m. This stronginflow is characterizedby plungingdownstream(landward)of the most
constrictedsection- theplungeline is foundat thelandwardendof themouthchannel(Figure
4). This plungeline, which separatesthe singlelayer seawaterinflow from the two-layer
estuaxineflow, is markednot only by the accumulation of foam and flotsambut alsoby an
obviouscolourchange. Theseplungingflows havebeenreferredto as tidal intrusionfronts
(SimpsonandNunes, 1981) and havebeenreviewedby Largier(1990). The V-shapeseen
in Figure4 is a characteristic
shapeseenin a numberof strongtidal intrusionfronts- the sides

of the flow plungefirst(nearthecriticalsection)whereastheflow in thecentreaccelerates

owingto inputfromthesides.Thecentreplunges furtherdownstream (landward).Thestrong
horizontalshearacrossthe front resultsin shearinstabilitieswhichpreventa singularityfrom
developingat the apexof the front.
Figure4: The V-shapedfoam-lineassociated

with thetidalintrusionfrontduringstronginflowon 20
August1986. The neckof thecontraction
liesjustto theleft of thepicture.The frontis alsomarked
by a colourchangebetweenthe brown(tannicacid) river wateron the fight and the foamygreen
seawater on the left.
The BottomDensity Current: Lm•dwardof this tidal intrusionfront, beyondthe internal

hydraulicjump,theseawater continues
to propagate asa bottomdensitycurrent.It arrivesat
station3 (Figure1) at 1334,producinga salinityincrease
of 18%oin 4 minutes(Figure3).
Themostrapidchangeof 16%oin 100s indicatesa horizontal
gradientof 0.8%0m4 through
thehead,givena speed
of advanceof 0.2 m s4. As thisbasalintrusion
moveslandward,a
weak surfacefoam line is observedto mark the positionof the head. This foam line canbe
usedto estimate
thepropagation
speedof the intrusion.
It advancedat about0.15 m s4
betweenstations
0 and3 andaccelerated
to about0.20 m s4 on.theslopingbedbetween
stations3 and 4. Similar valuesof about 0.15+0.05 m s4 are obtainedfrom the arrival times
of the salinityfrontat stations3 and4. The empiricalresultof Simpsonand Bfitter (1979)

that u = 1.15 c where u is the fluid velocityand c is the propagationspeed,can be usedto

obtainsmallervaluesof 0.10 to 0.15m s'1for theimmature
densitycurrentat stations
0 and
1 at about1315. Largervaluesof 0.15to 0.25 m s'1wereobtainedat stations2 and3.
Theoreticalexpressions(Benjamin,1968) and empiricalrelations(Wilkinson, 1983) have

beenobtainedfor the speedof advanceof the headof sucha gravitycurrentas a functionof
g', h2andthefractionaldepthh2/H. The speedobtainedfor an inviscidintrusionon a flat bed
is reducedsignificantlyby viscousdragfor smallh•/H (Wilkinson,1983)andby the presence
of a "headwind," an opposingflow in the upperlayer (Simpson,1982).From measurements
of g', h2 and H at stations1 (1310), 2 (1320) and betweenstations3 and 4 (1420), the
predicted
speeds
are0.20m s'l, 0.22m s'1and0.18m s'l, respectively.
Theobservations
from
the waterchannelseawardof station3 are lessthanpredicted.It is suggestedthatthisshould
be expectedowing to our strategyof alwaysmeasuringin the deepestpoint of the section.
Towardthe sidesh2and, particularly,h2/H are smallerso that c is reduced. The laterally-
averagedspeedof advanceof thisbasalflow mustaccountfor the three-dimensional nature
of the spreadingof this densewater - c shouldvary as the width and shapeof the channel
varies. In contrast,beyondstation3 the theoryappearsto underpredict,
thisbeingdue to the
downwardslopeof the bed.
On a sill betweenstations4 and5, thisbasalintrusionis againsubjectto internalhydraulic

control.
Theflowaccelerates
to0.23ms'1andF22
- 1.0sothattheseaward
flowing
freshwaterand the landwardflowing salinewater are coupledas an exchangeflow. An
interestingfeature is the velocity maximum observedin the halocline,not near the bed as
describedby Simpson(1987). This velocitymaximum,which is alsoobservedat stations0
and 1 (Figure 3), is due to the absenceof turbulence(and momentumdiffusion) in the
pycnoclineas comparedwith the viscousdynamicsof the flow nearerto the bed (Geyer,
1988).
VerticalExchange:AlthoughthePalmietis essentially two-layered,thisviscousintrusiondoes

exchangewaterwith theoverlyinglayer. Duringtwo-layerexchangeflow, theturbulentsaline
flow entrainsthe lessdenseambientestuarinewaterand is about20% dilutedbeyondthe
hydraulicjump (salinity decreasesfrom 25.5 to 19.5%o). The initial dilution from 35.5 to
25.5%0 is a resultof ebb-plumeand surf-zonedynamics.Later, duringstronginflow and
plunging,water of 31%o is introducedto station0 which is beyondthe plungeline. On
flowingfromstation0 to 2, thissalinelayeris diluteda further13%(to 27%0) via themixing
associated
withthehydraulicjump. WhilebulkRichardson numbers of theorderof unityare
observedat the head of the salineintrusion,the underflowis dilutednegligibly(4% to
25.7%0) en route to station4. The productsof this turbulenceat the headare containedin
a thirdstratifiedlayerbetweenthebasalandsurfacelayers. A bulk Richardson numberof the
orderof unity is also observedat plunging.
Even after the tide hasturnedat the mouth,and after the seawaterintrusionis halted,this
basalunderflowcontinuesto draininto thedeepestsections
of theestuary.The saltcontent
of thewatercolumnat station4 reaches a maximumbetween1600and1700,by whichtime

thesurface
layeris flowingoutat0.3 m s4 (Figure3) - upward
entrainment
intotheturbulent
surfacelayer beginsto removesomeof this salt. Richardsonnumberswere near-unitybut
supercritical,consistentwith the observationthat the haloclinedoesnot breakdownbut is
rathereroded
slowlyfromabove.Thelocalentrainment
velocityat station
4 is 4.2 x 10'5m
s4 on themorningebbtide(between
0930and1330),witha maximum of 8.3 x 10'5 m s'1
obsewedbetween0958 and 1008. In the afternoon(between1700 and 1900) the entrainment
velocity
is 3.5x 10'5m s'1. Following
Turner(1973),theobserved bulkRichardsonnumbers
of about3 correspondto entrainmentvelocities
of about4 x 10'5m s4. Thisentrainment
in
theupperestuary,wherethesurfaceflow is containedandstrong,is muchstronger
thanin the
lowerestuary
- theoverallentrainment
velocity
is 1.5x 10'5m s4. Thesalinity
of thesurface
waterincreases
from 0 to 1.5%oby station4 but only increases
to 2.0%o by the mouth.
Duringmostof the year,whenfiver flow in the Palmietis not strongenoughto produce

subcriticalRichardsonnumbers,the sillsin the estuaryare equallyimportanton the ebb fide.
The stratifiedfluid belowthe sill heightandupstream(landward)of the sill is blocked. This
water experiences very long residencetimesconsistent with slow verticalentrainmentrates.
The waterbelowthe haloclineis unaffected by thisentrainment (e.g.,below2 m at station4,
Figure 3). The salinewater remainingin the deeperregions(stations4 and 5) on the
followingmorning(21 August)is evidencethatthisdeepwateris residentfor longerthana
tidalcycle. Theaverageresidence timeof thesalinelayer,basedonoutflowrateversushigh-
fide volume(LargierandTaljaard,1990),is 1.6 tidal cycles. The deepwateris expectedto
be residentfor substantially
longer. The low salinitysurfacewateris effectivelynon-resident
with an averageresidencetime of 0.4 tidal cycles.
8.3 The Summerexample

As summerapproaches
therain stormsare smallerandlessfrequent,thefiver flow weakens
and the estuarybecomesmore saline. For much of the summerthe Palmiet River flows at
about1 m3s4, withoccasional
increases
afterlightrains.Tidalflowdominates
theestuary
with maximumvelocitiesoccuringduringflood. Stratificationremains,however,with the
thermalgradients
enhancingthe salinitygradients
in densityas thefiver warmsfrombelow
15øCto over 25øCwhile the oceangenerallyremainsbelow 15øC. At the sametime the
reduced fiver flow allows sand to accumulate in the mouth - the channel becomes narrower
andshallower.Thissewesto protecttheestuaryfromexcessive tidalflow andthus,together
with thecontribution
fromthermalstratification,
thePalmietcontinues to be highlystratified
mostof the time duringsummer.In February1985 we obtainedsomeobservations of the
typicalsummerversionof the stratification
structures
andprocesses.
Duringspringtideswhichoccurin the dry periods(fiver flow lessthan1 m3 s4), the

shallower
channelupto station3 canbecome
partiallymixedfollowingshear-driven
ebbtidal
mixing. Thisebb-fidedestratificafion,
dueto subcriticalRichardson
numbers,is well described
by comparison
withtheebb-fide
breakdown
of two-layerstructure
in theFraserRiver(Geyer,

1988). Mixing occurredat about 0830 on 6 February,4 hoursbefore slack water in the
mouth. This partially mixed characterremainsuntil new salinewaterof 34.7%0 and 17øC
intrudesbeneaththe 32.5%0 and22øCambientestuarinewaterat about1400 (Largier, 1986) -
this corresponds to a dilutionof seawater(35.5%0) of about25% througha combinationof
surf-zoneand intrusionprocesses.This 2%0 and 5øCstratificationremainsfor the 3 hour
flood tide and is erodedagainon the ebb tide after 1700 at station1. The basalintrusionis
seenat stations2, 3 and4 by about 1430, 1515 and 1630, respectively- corresponding to a
realistic
speedof propagation
of theorderof 0.1 m s'• andnegligible
furtherdilution.A
similarbasaldensitycurrentwas observedon the previousday (5 February).
Times of data collection
'•.....
......
'"'"
ß' _-.".:X:• ........
'•5•
''•'"'.':5:'
.i/t•:'-•"'- ....
•-'::z..:'::•'
":':':":':"
4 .;t:' "- .":'• .:':':'-:':':':
_ ":'-"
'-'.."..."........'•...-........,`
'.-0._:-./..
:.'::.'i
½! .ii:.:.½:i:ii:i:'.½:::i:½.:.'
"':'"':'
':".
;-]
ß :"'. .'".'•'.'i'.".': X \
5:!:• )."'
t / ß"
"':'
'-'
':""
'"
..•.:.:....:....?•.......... ..:.....•.:.:........:........:.
•.•. - j ...... .
I I :.•.,..-z...:.-.:, -.•:oz-.:-:.:;:r.':.'..,z...[.
,-..;.,.
,,-.,
o,-.,::,,--•:•-.½:
•:.-.,'.-..:;..
>,.:-.: ...,..
,,.:.,,.;_
0700 0900 1100 1300 1500 1700 1900
TIME OF DAY (hi
Figure5: Laterallyaveraged
watervelocitiesat a channelcross-section
throughstation1 on 12February
1985. Velocities
arein unitsof cm s'• andnegative
valuescorrespond
to landward
flow.
During neap tides, or during slightlyenhancedriver flow, the two-layerstratification

remains
through
ebbtide. Forexample,
on 12February
(riverflow3 m• s'•) bottomsalinities
of 30%0 andsurfacesalinitiesof 3%0 wererecordedat station1 at about1330,afterpeakebb
flow (Figure5) - an interfacewasfoundat 0.5 m. Further,duringflood tide,curvedandV-
shapedplungelines were notedat the landwardend of the flood tidal delta. On the earlier
flood tide, when inflow was confinedto the channelalongsidethe delta,the plungeline
adopteda U-shapewhich evolvedinto a V-shape(similarto Figure4) as the denseinflow
strengthened and the centreof the plungeline waspushedfurtherdownstream (landward)-
thesearereferredto as typeC andD tidalintrusionfrontsby Largier(1990). At about1855,
a saline
underflow
(33.6%0and0.5m s'•) wasrecorded
belowstationary
freshwater
atstation
0 whichwaslandwardof theplungeline (corresponding to a dilutionof lessthan10%before
the hydraulicjump). This supercriticalbasalflow appearsto exceedthe critical turbulence

velocities
required
for suspension
trodto scourthesandbedbetweentheplungelineandthe
hydraulicjump. A closedareaof over3 m depthis recorded
aroundstation
0 in summer.
As thewaterlevel rises,thenatureof the tidal intrusionfrontchanges.The incomingseawater
flowsoverthefloodtidalshoalsanda newcurvedplungelinedevelops
alongtheedgeof the
shoalasthedensewaterspillsdownintothedeeper channel
landward
of theshoal(Figure6) -
thisis a typeA tidalintrusion
front(Largier,1990).
Figure6: Thecurved
foam-line
andcolour
change
associated
withtheplunging
flowalongtheedgeof
thefloodtidaldeltaduringlatefloodon 12February1985.Station0 is to therightof thefoamlineand
the boatis at station1. The U- andV-shapedplungelines,whichoccurred on the earlierflood,were
locatedin theleft foreground of thepicture.The photographer
wasstanding at thewesternendof the
sandspit, lookingnorth.
AlthoughFigure5 clearlydemonstrates
thistwo-layervelocitystructure,
it doesnotdisplay
the presence
of a transient
thirdlayerwhichis observed as a smallbasaldensitycurrent
intrusionin summerconditions.It does,however,displaya velocitymaximumin the lower
halocline,as notedfor the winter intrusions.The third layer developsthroughthe tidal
intrusionprocesst3irlylateon thefloodtide. Initiallytheinflowingwaterplunges
belowthe
estuarinewaterand is substantiallydilutedby the strongshear,producinga salinelayer of

about 32.5%0 which fills most of the estuary. It is suggestedthat as the flood tide
strengthens,
sotheplungelineis pushed backintodeeperwaterand,eventually,
theseawater
flowingin alongthebottomis sufficiently
removedfromtheinterfacial
shearthatthebillows
do not dilute it to the sameextent as the rest of the intowing seawater. This appearsat
station1 (Figure7) asa basalintrusion
of 33.5%0and 13.5øC;it maycontinueto draininto
thedeeperchannelat station4. In additionto thetemperature-salinity
character
displayedin
Figure7, thisbasalintrusion
wasobserved to haveenhancednutrientlevelscharacteristic
of
the recentcoastalupwellingin the ocean.
SURFACE
lWARMING
25-
ß ', 1/,February_1985
12h50
river
water
•!• ,•AINMENTS•'.
I
1 2
3 4'
,
20'
TEMP.
(o[)
15'
',
sea •ater
100.... 10
SALINITY { %0 )
Figure 7: Temperature-salinity data collectedat about 1300 (high tide) on 14 February 1985 (data
plottedfor surface,depthincrements of 1 m andbotto•nat eachstation).River waterwas0%o and23øC
whereasseawaterwas35%0 and 13øC.The separation of waterintotwo typesis clear. A is thesurface
water, exhibitingsurfacewinning and increasingsalinityowing to upwardentrainmentof halocline
water.B is the old salinewater at middledepthand D is the old basalwater in the scourholesof the
upperestuary.C is the newly introducedsalinewaterwhich is noticeablycolderandin the processof
intrudingunderthe olderwaterresidentat stations 4 and5. The coldest,•nostsalinepointis the bottom
densitycurrentinitiatedlate on fl•e flood fide.

Throughthecombined effectof supercritical

Richardsonnumbers andupstream
blocking
by the sills,thedeeperregionscontinueto play therole of trappingandaccumulating
the
densest water that reaches them. Denser water intrudes beneath the old bottom water and
replacesit, similardensitywatermayintrudeat somemid-depth andlessdensewaterintrudes
alongthebaseof the pycnocline, allowingthe old waterto remaintherefor anothertidal
cycle. It is expectedthatthe residencetimeof thisdeepwaterwill be substantially longer
thanthe 1.5and5.0 tidalcycleswhichwerecalculated for theseawaterfractionof theestuary
for the dry, springcaseand the wet neapcase,respectively.In particular,35%0 water
recordedat stations4 and 5 on 6 Februaryprobablyremainedthere until the cold water
intrusionon 14 February(note,however,thatthedeepest
pointsplottedin Figure7 werenot
fromthedeepestspotsin the upperestuary,andthatsome35%0 watermaystill havebeen
residen0.Average
upward entrainment
velocities
of between0.5 and1.0x 10'• in s4 are
calculated
duringebbtide,corresponding
to a netestuarine
circulation
of some0.9 ma s4
(Largier, 1986).
In comparison,
thefreshwater
fractionof theestuaryhasresidence
timesof lessthana tidal
cycle(asfor winter). Typicalebbtidesurface
velocities
average
0.1 in s4 or morefor6
hours,yieldinga tidal excursionof 2200 m whichis clearlylongerthanthe estuary(Figure
1). There is thus no residentfleshwaterlayer. This outflow of surfacewater appearsto
inducea sympathetic seawardflow in thesalinelowerlayer(Figure5). However,thisdenser
wateris blockedby the sill acrossthemouthandverticalvelocitiesare negligiblesincebulk
Richardsonnumbersare in excessof 10. It is suggested that the interfaceis in the process
of beingset up with an opposingbaroclinicpressuregradientgrowinguntil it matchesthe
imposedstressof the overflowinglayer.
8.4 The generalcase

The abovetwo observations of stratifiedflow in the PalmietEstuarysuggestthatthe summer
and winterpatternsare variationson a commontheme. The significantseasonal changesin
fiver flow and bathymetryserveto counterbalance each other and resultin a gravitational
circulationpatternwhichis fairly robustagainstseasonal effects. The actionof wind on the
smallsurfaceareaof the Palmietappearsto be of little consequence in thisstratifiedflow but
remotewind effects,throughforcingof the coastalocean,canprovidedramaticincreasesin
waterlevelandseawaterintrusion- for example,on 7 February1985the estuarinewaterlevel
was 0.39 m higherthanfor an equalhigh tide on the previousday.
The verticalgravitationalcirculationaccountsfor mostof the tidal saltflux in the Palmiet.

It is characterizedby four processes: hydraulicallycontrolledinflow of densewaterthrough
the constricted mouth; basalintrusionof a non-mixingdensitycurrent;shear-driven upward
entrainment into the surfacelayer and removalby this outflowinglayer of buoyantwater.
Thiscollectionof processes appearsto be presentin differentproportions in a varietyof tidal
inlets- particularlyin micro-tidalestuaries,
whicharecharacterized by salt-wedge stratification

andfloodtidaldeltas.Tidal plungelineshavebeennotedin a varietyof similar"bar-built"

estuariesalongthe southern Africancoast(e.g.,Sout,Qora,Umgeni),theAustralian coast
(e.g.,PortHacking- Huzzey,1982),theCaliforniacoast(e.g.,Mad - Largier,1991),andthe
Oregoncoast(Pistol- Cliftonet al., 1973). Tidal plungelineshavealsobeennotedin
Scottishlochs(e.g.,LochCreran- Booth,1987),in someunconstricted macro-tidalWelsh
estuaries(e.g.,Seiont-Simpson andNunes,1981)andalongthecoasts of somebroaderbays
(e.g., HumboldtBay, California,and Chesapeake Bay, Maryland). Field observationsof
hydraulicallycontrolledexchange flow havealsobeenreportedfor smallboatharbors(e.g.,
Hilary Harbourin Australia- SchwartzandImberger,1988;HamiltonHarbourin Ontario-
Hamblinand Lawrence,1990). The point is that this Palmiet-typesystemis a common
feature,particularlyin the countless
small"bar-built"estuaries
alongthe highwave-energy
coasts of the world.
For the purposesof comparingthis PalmietEstuarywith othersystems,it is usefulto

attemptto place it on estuarineclassificationschemes.We have already noted the
geomorphologicalclassifications
of thePalmietasa drownedfivervalley,a "bar-built"
estuary
andas a micro-tidalestuary(tidal rangelessthan2 m) whichhasa morphology determined
primarily by fiver flow and surf-zonedynamics. In the following hydrodynamic
classifications,
stratificationand flow are compared. Fischer (1972) presentsa Estuarine
RichardsonNumberR which expresses the ratio of buoyancyinput to tidal mixing. This
dimensionlessnumber is intended as an index of stratified versus mixed structureand, thus,
of a saltbalancedominatedby densitycurrentadvection(e.g.,Palmie0or tidal diffusion. In
general,thehighlystratifiedPalmietexhibitsvaluesof theorderof 10 (Figure8) - valuesin
excessof unity taken as representative of stratifiedestuaries. Smallervaluesare found
towardsthe mouth(greatertidal velocities)duringsummerspringtidesand low fiver flow.
For example,on 6 February1985,a smallbuoyancyinput(fiver flow) andlargetidalcurrents
(springfide) conspireto reducethisRichardson numberbelowunity - consistent with the
partiallymixed conditionsobservedon the ebb tide. Similar resultsare found from an
in•tion of the EstuarineStratificationNumberof Stigebrandt (1988), in which the tidal
mixingis scaledby the frictionvelocity. The numericalvaluesare uncertainbut generally
muchlessthanunity,indicativeof strongstratification.Thisdimensionless numberis location
dependent andindicatesan increasein stratification towardsthefiver inflowandin thedeeper
regionsas well as a decreasetowardsthe mouth and in the shallowerregions. This
longitudinalvariationin stratificationis a commonfeatureof these"bar-built"estuaries.Over
the shallowsof the flood tidal delta (and over the intertidalsandbanks)the water columnis
oftenmixed(singlelayer salineinflow duringflood tide), in the main channelit is stratified,
in thedeeperregionstowardstheheadthe structure is extremelystablystratified(oftenmulti-
layered)andbeyondthe headof the salt wedgethe structureis againmixed (freshwater).
The two-criteriaclassification
diagramsof HansenandRattray(1966) and Jay and Smith
(1988) are more illustrative. Hansen and Rattray (1966) use indices of tidal-average
stratification
AS/•qandcirculation
U,/U!(Figure
8),whereas
JayandSmith
(1988)useinternal
(FB) and barotropic(Fr) Froudenumbers. This latter schemeseeksto classifythe estuary
throughevaluationof barocliniccirculationandbarotropictidalforcing. Palmietvaluesof FB

148 J L Largier,J H SlingerandS Taljaard
areof theorderof, but greaterthan,one- indicatinga highly-stratified

systemandsignificant
baroclinicnonlinearity.In the high-flowwintercase(e.g.,25 August1986),at stations1, 2
and 3, the saline water is completelyremoved on the ebb tide and FB goes to one,
corresponding to themostextremebaroclinicnonlinearity.In contrast,at thedeeperstations,
Fa takessmallervalues(goingto zerowhenthe salineintrusions are discontinued owingto
increasedfiver flow or closingof themouth)- thisindicatesincreasingstratification
strength.
Thevalues
of Fr, generally
greater
than10'•, represent
thenon-linear
nature
of thebarotropic
tide.
10•01•4
I I Type4
.D400A4
10
-1
1•l Type
2
10-2
1 10 102 103
Us
Figure8: The classification

schemeof HansenandRattray(1966). The Palmietis classifiedashighly-
stratified(type 4) and compareswith the classification
of the Mississippiand ColumbiaEstuaries.
Values for the salt flux diffusion/advection
ratio v andfor the EstuarineRichardsonNumberR (Fischer,
1972)areplotted
asfunctions
of thecirculation
U•/U.tandstratification
AS/Sparameters.
Points
A• and
An are for stations2 and4 duringmoderateriver flow and springtide on 20 August1986. PointB is
for bothstations1 and4 duringweakflow andneaptideon 12 February1985. RegionC2 is for station
2 duringvery weak flow and springtide on 6 February1985 - this corresponds to partiallymixed
conditions. PointDn is for station4 whenthemouthwasclosedandtheriver flowedveryweakly(1
m3s'l) in March1988.Region Enisforstation
4 whenonlya thinlayerof saline
waterremained
during
strong flow(30 m3 s'l) on24 August1986.

In Figure 8, the winter and summerdata are plottedon Hansenand Rattray's scheme.
While mostof the pointslie in a smallregionin the highly-stratified upperleft corner,their
similarityis a bit deceptivein that this is alsoa regionof largegradientsin criterialike the
saltflux diffusion/advection ratio v. It doeshoweverprovidethe essenceof the classification:
the Palmiet is highly stratifiedmost of the time and the advectivesalt flux, due to the
gravitationalcirculation,is at least as importantas the diffusive fluxes (v -- 0.5) - and
generallymore importantin the upperestuary(station4). Even whenthe mouthis closed
(next section),the datafall into thistype4. The one exceptionis the datafrom station2 on
the springtide duringvery weakfiver flow on 6 February1985. In thiscasethe estuaryis
partiallymixed(type2) andit is predictedthatdiffusiondominates the saltflux in the lower
estuary.
8.5 A future scenario
Many of the estuarieswith constrictedmouthsalso exhibit sills and scourholes further

landward.It is thiscomplexbathymetry whichis mosteffectivein trappingdensesalinewater
for longperiodsof time. With increasing depthin thewatercolumn,residencetimesincrease
(from zerotidal cyclesat the surfaceto severaltidal cyclesat the baseof the deeperregions).
With thisincreasein residencetime is a parallelincreasein primaryproductivityand general
biomass. However, if the deep water remainsresidentfor too long, the oxygen will be
depletedandanoxicconditions mayoccur- asis reportedby CSIR (1982) in themarinacanals
of the SandvleiEstuary. Further,decomposition of organicmatteralsoresultsin high levels
of hydrogensulphideand carbondioxide (e.g., Swartvlei - CSIR, 1983). This is toxic for
animaland plant life and may alsobe unpleasantfor the local society.
Concerned with thepotentialeffectsof anthropogenic decreases in fiver flow (e.g., dueto

dammingof the fiver) in this sortof system,a field studywas conductedin the summerof
1988 when the fiver flowed so weakly that the mouthclosed. We were concernedwith how
longthewaterin the deepregionswouldremainwithoutoxygenventillationfrom the surface.
As the mouth becomesmore constricted,the strengthof the tidal mixing in the estuary
decreases until it is zero whentidal inflow is prevented.However,the residentbuoyancy
remainsand furtherbuoyancyis input throughsurfaceheatingand weak fiver flow - the
estuarineclassificationsindicatestrengtheningstratification.In theabsence of substantial wind
forcingor fiver inflow, would this stratifiedstructureremain,decayingslowly owing to
molecularprocesses?
Followingclosureof themouthin January, we visitedthePalmiet8 timesfromFebruary

to April. The temperature-salinity
character of thewaterat station5 is plottedin Figure9 for
the first 6 visits,evenlyspacedover a periodof 55 days. The fiver and surfacewaterscool
by 4øCand7øC,respectively.Thiscooling,however,is balanced by a continualinputof new
freshwaterandstrongstratification remains.On thefirst3 daysthisnewcoolerwater(mixed
with somesalineestuarywater)is seento intrudebeneaththe warmsurfacewateralongthe

top of the halocline. On the fourthday, the 'affectof a recentfreshetis seenin coolerand
moresaline(owingto increased entrainment of haloclinewater)surfacewater. At aboutthis
time the estuarybeginsto overflowthebar acrossits mouthandlargershear,associated with
this throughflowof surfacewater,accountsfor morerapiderosionof the haloclineby days
5 and6. However,themixingcontinues to exhibitsupercriticalRichardson numbersandonly
the top of the haloclineis removed. As the haloclineis erodeddeeperit retainsmuchof its
strength(Figure 9).
26
// ?,._/
24
22
2O
x5
6.• 6 ,Surface
/
+
ß
o
2m
3m
4m
i.6 X 5m
0 10 20
/ ,
Salinity (%o)
Figure9: Temperature-salinitydatafromstation5 for thefirst6 fielddays(separated

by about11 day
intervals)whenthe mouthof the Palmietwasclosedin February-April 1988.River temperatures
are
plottedon the zerosalinityaxis. The thin connecting
linesindicatethe maximumgradientsectionof
thehalocline.The ellipseencloses thedeepwaterdatain whichdiffusionproceeded at ratesconsistent
with thetotalabsence of watermove•nent.By connecting filedatafrmnthestonedepthanddifferent
day,onecm•obtainm•impression of howthetemperature-salinity character
changedwith time.
What is particularlyinteresting
is the apparentabsence of mixingbelowthis halocline.
Slingerand Largier(1989) calculateddiffusionratesbasedon an assumption of Fickian
diffusion.Bothsalinityandtemperature ratesindicatethatturbulenceandadvectionis absent
at depthandthatdiffusion occursvia molecular action(thermaldiffusivityKt '-' 10'3cm2 s'•,
salinediffusivity
Ks- 10'scm2s'•). Thesurface
layerturbulence,
dueto surface
windstress
and fiver inflow (and intrusion),
cannotpenetrate
the strongpycnocline.The deeper

observations, whicharecircledin Figure9, arerepresentative

of thisslowmoleculardiffusion.
This sub-pycnocline wateris thusresidentfrom whenit wasintroduced in Januaryuntil the
mouthre-openssufficientlyto allow new intrusionsof densesalinewater(beforeday 8 in
May) - this is a residencetime of about3 months(at only 3-m depth).The accumulation of
nutrientsand detritusand the depletionof oxygenin this basalwatercan havesevereeffects
after 3 months. While no such measurementswere obtainedfrom the Palmiet, this has been
observedin the Sandvlei(CSIR, 1982), Swartvlei(CSIR, 1983), Mtamvuna,MsikabaandLake
Sifungwe(Day, 1981)estuaries
alongthesouthernAfricancoast.It canalsobe expected in
manymorecomparablesystemsaroundthe world. Of specialconcernis the potentialrole of
upstream
reservoirs
andwaterusein creatingtheseadverseconditionsin thePalmietEstuary
and its kin.
Acknowledgements
We gratefullyacknowledge the supportof the NationalResearch
Institutefor Oceanology
(CSIR) andthe SouthAfricanCommitteefor Oceanographic Research(FRD, CSIR).
8.6 References
Armi, L., 1986: The hydraulicsof two flowinglayerswith differentdensities.J. Fluid Mech., 163,
27-58.
Benjamin,T. B., 1968: Gravitycurrentsandrelatedphenomena.

J. Fluid Mech.31 , 209-248.
Booth,D. A., 1987: Someconsequences

of a floodtide frontin LochCreran. Estuarine,Coastaland
Shelf Science,24,363-375.
Branch,G. M. andJ. A. Day, 1984: Ecologyof southern Africanestuaries.

XIII. The PalmietRiver
estuaryin the southwestern
Cape. S. Afr. J. Zool., 19 (2), 63-77.
Clifton,H. E., R. L. Phillips,andR. E. Hunter,1973: Depositional

stmcmres andprocesses in the
mouthsof smallcoastalstreams, southwesternOregon,p. 115-140. In D. R. Coates(ed.),Coastal
Geomorphology,
Publications
in Geomorphology,
New York.
CSIR, 1982: Estuaries

of theCape:Sandvlei
CSIRResearch
Report413, Stellenbosch,
SouthAfrica.

of theCape:Swartvlei
CSIRResearch
SouthAfrica.

of theCape:PalmietCSIRResearch
SouthAfrica.

Day, J. H. (ed.), 1981:EsmarineEcologywith ParticularReferenceto SouthernAfrica, CapeTown,

Balkema,411 pp.
Fischer,H. B., 1972: Masstransportmechanisms

in partiallystratifiedestuaries,J. Fluid Mech.,31 (2),
209-248.
Geyer,W. R., 1988: The advanceof a saltwedgefxont:Observations anddynamicalmodel.In Physical

Processesin Estuaries(Dronkers/vanLeussen,eds.), Springer-Verlag,Heidelberg,pp. 181-195.
Hamblin, P. F. and G. A. Lawrence, 1990: Exchangeflows betweenHamilton Harbour and Lake

Ontario, Proc. 1990 AnnualConf. CanadiaSoc. for Civil. Engr.
Hansen,D. V. andM. Rattray,1966:New dimensions

in estuaryclassification,Limnol.Oceanogr.,11
(3), 319-326.
Huzzey, L. M., 1982: The dynamicsof a bathymetricallyarrestedesmarinefront. Estuarine,Coastal

and Shelf Science, 15, 537-552.
Jay, D. A. and J. D. Smith, 1988: Residualcirculationin and classificationof shallow,stratified

estuaries.
In PhysicalProcesses
in Estuaries(Dronkers/vanLeussen,eds.),Springer-Verlag,Heidelberg,
pp. 21-41.
Largier,J. L., 1986: Structureandmixingin the PalmietEstuary,SouthAfrica. S. Afr. J. Mar. Sci.,4,

139-152.
Largier, J. L., 1990: Tidal intrusionfxonts.Submittedto Estuaries.
Largier,J.L., 1991:Estuarine
circulation.Proceedings
of theEstuaryandOceanProductivity
Workshop
- 1990NE PacificChinookandCohoSalmonSymposium(C. Tuss,ed.),HumboldtStateUniversity,
Sept. 1990, (in press).
Largier,J. L. andS. Taljaard,1990:Wintertimecirculation

andstratification
in thePalmietEstuary,
South Africa. Submitted to Estuarine, Coastal and Shelf Science.
O'Brien,M.P., 1969:Equilibriumflow areasof inletsonsandycoasts,J. Waterways

andHarborsDiv.,
ASCE, No. WWI, Feb. 1969, 43-52.
Schwartz,R. A. andJ. Imberger,1988:Flushingbehaviourof a coastalmarina, Proc.21stInter.Coastal

Engr. Conf.
Simpson,
J. E., 1982:Gravitycurrents
in thelaboratory,
atmosphere
andocean,Ann.Rev.FluidMech.,
14,213-234.

Simpson,J. E., 1987: Gravity Currentsin the Environmentand the Laboratory, Ellis Harwood,
Chichester,244 pp.
Simpson,J. E. andR. E. Britter,1979:The dynamicsof the headof a gravitycurrentadvancing

over
a horizontal surface, J. Fluid Mech., 94,477-495.
Simpson,J. H. and R. A. Nunes, 1981: The tidal intrusionfront: An estuarineconvergencezone.

Esmarine, Coastal and Shelf Sci., 13,257-266.
Slinger,J., andJ. L. Largier,1989.The evolutionof thermohaline

structure
in a closedestuary.S. Aft.
J. Aqua. Sci., (in press).
Stigebrandt,A., 1988: Dynamiccontrolby topography

in estuaries,Hydrodynamics
of Estuaries,B.
Kjerve (ed.), CRC Press,17-26.
Taljaard,S., G. A. Eagle and H. F.-K. O. Hennig, 1986: The PalmietEstuary: A modelfor water
circulationusingsalinityandtemperaturemeasurements overa tidalcycle, Water S. A., 12 (3), 119-126.
Turner,J. S., 1973: BuoyancyEffectsin Fluids,CambridgeUniversityPress,Cambridge,368pp.
Wilkinson, D. L., 1983: Studiesinto the structureand motion of densitycurrents, Water Research
LaboratoryResearchReport 160, Universityof New SouthWales,Australia,36 pp.

Salinity structure of a shallow, tributary

estuary
W W Schroeder, S P Dinnel, and W J Wiseman Jr
Abstract
Analysisof salinitydatafrom 14 field surveysin WeeksBay (Alabama,USA) indicatethat

the salinityregimewithin this shallow,tributaryestuaryvariessignificantlybothtemporally
and spatiallyduringperiodsof minimaltidal range. Horizonal salinitygradientsvary from
weak to strongand complex,bothin the longitudinalandlateraldirections.Vertical salinity
structurealsovariesfrom well mixedto ultra-stratified(up to 14 psuin two metersof water).
This variabilityresultsprincipallyfrom subtidalexchangeswith the main estuary(Mobile
Bay), both in the barotropicand baroclinicmodes,and flashy local fiver ranoff. At a
secondary level,salinityfieldscanbe influencedby bothbathymetry,particularlydeeperareas
which eitherchannelnear-bottomwater within the bay or impoundbottomwatersin scour
holesor depression areas,and local windsassociatedwith moderateto high energyevents.
Excessfreshwater in WeeksBay, relativeto Mobile Bay, canbe accounted for by local

fiver inflow. WeeksBay fill time is oneto four daysandis dependent on the lengthof time
sincethe last local freshet. A principalcomponentanalysisof stratificationresultedin two
dominanteigenvectors.The first describes 54% of the totalvarianceandoccurswhensurface
salinitiesare uniformlyfreshetthanhighersalinitybottomwaters. The second,limited to the
lowerbay,describes17%of thetotalvarianceandmayindicategravitational circulationdriven
by variationsin fivefine input to the bay.
9.1 Introduction
Resultsfrom fourteenmonthsof a two year field studyof WeeksBay, Alabama,provide

insightintothestructure
of the salinityregimein a shallow,microtidal,
tributaryestuary
1.
1Atributary
estuary
is defined
hereinasanestuary
located
withina largemainesmarine
system
where
the largerestuaryservesas the tributaryestuary's"coastalocean"salt source.

156 W W Schroeder, S P Dinnel and W J Wiseman
WeeksBay is locatedon the easternshoreof the Mobile Bay estuary(Fig. 1) in the northern
Gulf of Mexico. It is nearlydiamondshaped,with a 3.4 km north-south
longitudinalaxisand
a maximumeast-westextentof 3.1 km. A small, 5 to 7 m deepscourfeatureis locatedin
the narrowmouthof the bay and a similarscourfeature,3 to 4 m deep,occursin the Fish
River approximately200 m upstreamof whereit emptiesinto the bay. The bay hasa mean
depthof 1.4m anda surface
areaof 6.95x 10• m2 (Crance,
1971). Tidesareprincipally
diurnalwith a meanrangeof approximately0.4 m. Two riversdischargeinto the bay; the
Fish River at the northernend, accountingfor approximately73% of the total fresh water
input,and the MagnoliaRiver on the easternshore. Mean combineddischargeis estimated
at 9 m3s'1 , withfreshets
up to 4 timeslargeroccurring
throughout
theyear. Shortcurrent
meter recordsfrom the mouth of Weeks Bay are available(Schroederet at, 1990b). These
recordsare generallyfrom a singlemeterand,thus,do notresolvethe barocliniccomponent
of the flow. A singledeployment of two metersdoesindicatethatbaroclinicexchanges are
occasionallyimportam,but the barotropiccomponentgenerallyis dominant. Tidal currents
rangein amplitude
from0.15to0.3 ms'1andsubtidal
(40-hour
low-passed)
currents
from0.1
to 0.3 ms'1.
Previouswork in WeeksBay utilizingwater level, wind and ranoff data,haveshownthat

subtidalbarotropicexchangebetweenWeeksBay andthe Mobile Bay estuary,the adjacent
"coastalocean",is dominatedby a co-oscillation of the two bays(Schroeder et al., 1990b).
Furthermore, it hasbeendeterminedthatthe salinityof WeeksBay is stronglycoupledto the
sa!:nityof theMobile Bay estuary,whichis controlledby theextensiveseasonalrunoffof the
MobileRiversystem,andonly secandarily a functionof freshwaterinputfromthewatershed
of Weeks Bay (Schroederet at., 1989).
If the salinityin WeeksBay is secondarily

a functionof the localriversthentheresidence
time of freshwaterin WeeksBay, relativeto thatin thecoastalocean,shouldbe rathershort.
Schroeder et al. (1990b)suggestthatnormalfreshetswouldrefill the bay on the sametime
scalesas wind drivensubtidalexchangeand normaltidal exchange,and thesetime scales
wouldbe of the orderof a few days. In orderto estimatethe time scaleof WeeksBay
responseto fiver flow we use the parameter"fill time". Fill time is definedas the time
requiredfor freshwaterflow to accountfor the excessfreshwater volume,relativeto Mobile
Bay salinity,presentat the time of eachsurvey. Fill time is a functionof time-varying
processes,unlike residencetime, which is a functionof time-averaged processes.The
usefulness
of the fill time parameterhasbeendocumented for freshwateron the Louisiana-
Texas continentalshelf (Dinnel and Wiseman, 1986).
Schroederet at. (1990a) have shownthat stratification/destratification

eventscan occurin
thebroad,shallowMobileBay estuaryat variousfiver stages,
andthattheseeventsarerelated
to therelativestrengths
of thewindsandfiverdischarge. Verystrongstratification
variability
hasalsobeenobserved in WeeksBay: verticalsalinitydifferences
fromnear0 psuup to 14
psubetweensurfaceand 1.5 m depths,in 2.25 m of water. In an attemptto describethe
spatialvariationin WeeksBay stratification,
a principalcomponentanalysiswasperformed.

W W Schroeder, S P Dinnel and W J Wiseman 157
9 10
!•" .i,'
•, phi_?
I's ana Gulf of /•/exico I I
..
..
..
WEEKS BAY
USGS-WLil ...i::?ii??!
•....... •iiili
..
•:.::.
DEPTH IN METERS
--2
MOBILE
BAY i:•: 0I • 1
!
• Km
Figure1: Locationmapof WeeksBay,Alabama,showingbathymetry,andlocationof waterlevelgauge

(USGS-WL). InsertshowsMobileBay, andthelocationsof themeterological
station(DISL) andFish
River dischargegauge(USGS-R).

9.2 Observationsand analysistechniques

DuringtheperiodApril 1986to June1987 salinitydatawerecollectedin WeeksBay utilizing
a Hydrolab SurveyorII. All but one of the surveyswere made during periodsof low
amplitude,equatorialtides. Surveysof thebay weregenerallycompletedwithin six hoursor
less. Clearly,somealiasingresultedfrom the non-synoptic
natureof the sampling.Sincethis
regionof the coastis dominatedby diurnaltides,though,we expectthat the aliasingwill be
minimal. Hourly WeeksBay waterlevelswere obtainedfrom the United StatesGeological
Survey(USGS-WL), and hourlyvaluesof wind speedand directionwere obtainedfrom the
eastend of DauphinIsland (DISL) (Fig. 1). DISL windswere convenedto north-southand
east-westcomponents of wind stressusinga drag coefficientthat varies with wind speed
(Amorocho and DeVries, 1980, 1981). Water levels were demeaned,then both water level
and wind stresswere low-passfilteredwith a 0.6 cycleper day cutoff (Figs.2b and 2a).
Daily averagefiver dischargemeasurements for the Fish River were obtainedfrom the
USGS (USGS-R)(Fig. 1) startingin November1986. For theperiodMay to November1986
Fish River dischargewasestimatedfrom local precipitationdata. Precipitationwassummed
overthe threereportingstationsin theFishRiver drainagebasin,Bay Minette,Fairhope,and
Robertsdale (U.S. Departmentof Commerce,1986,1987,1988). A linear,multipleregression
model,with independent variablesof zero-,one-,two-, andthree-daylaggedprecipitation,was
fit to 480 days(December1986throughMarch 1988)of knowngaugedFishRiverdischarge.
In effect,we considerthecatchment basinto be a linearsystemwith FishRiver discharge as
the output. The impulseresponse of the systemintegratesa weightedversionof the runoff
over a 4-dayperiod. The impulseresponse decaysrapidly,with time. This simplemodel,
withanR valueof 0.815,adequately explainedlow andmoderate dischargevalues,buttended
to underestimate highFishRiver discharge.Never-the-less, it wasusedto estimatetheearly
missingFishRiverdischarge data. Daily averageFishRiverdischarge is depictedin Figure
2c. To extrapolatefrom daily averagedFishRiver dischargeto the totalfreshwaterflow into
WeeksBay,twofactorswereused.ForFishRiverdischarge, <2.83m3 s4, a factorof 2.76
wasused. For higherdischarges,
>2.83m3 s4, a factorof 3.65 wasused(J.L.Pearman,
USGS, personalcommunications).
Mapsof the horizontaldistributionof salinityweredrawnfor theseasurfaceanda depth

of 1.5 m belowthe surfacefor eachsurvey.When the waterdepthwaslessthan 1.5 m, the
bottomsalinitywasusedat that station.The mapson the deeperlevel are hereafterreferred
to as bottommaps. A contouringintervalof 2 psuwasemployed. In addition,north-south
longitudinalcross-sectionsfrom theFishRiver, alongthe channelon the easternsideof the
bay, throughthe mouth-of-the-bay and out into Mobile Bay were also constructed.An
exampleof suchsections is shownin Figure3. It is clearthatmostof thevariabiltybelow
a depthof 1.5 m is dueto pondingof high-salinity waterin scourholes. Thus,the mapsat
1.5 m depth are representative of the waters below the pycnoclinewhich are not
topographicallyisolatedfrom Mobile Bay.

W W Schroeder,
S P DinnelandW J Wiseman 159
I 6 7A 7B
' • ' "•!•'•I' •r-

0.0 0.02 0.04
0,6--
0.3-
-0.3-
-0.6
24-
18-
I 6 7A 7B 12
12
MAY 1JUL 1SEP I NOV 1 JAN 1 MAR 1 MAY 1JUL

1986 1987
Figure2: Timeseriesof low-passed

windstress
fromDISL (a) (top),demeaned,low-passed
WeeksBay
waterlevel(b) (center)anddailyaverage
FishRiverdischarge (c) (bottom).Numbers
indicatetimes
of individualsalinitysurveysillustrated
in subsequent
figures.
Forfill timecalculations,
WeeksBaywaspartitionedinto16segmentswithonestationper
segment; segment borderswereapproximatelyequidistant
between
adjacentstations.Two
stations
in theFishandMagnoliariverswereexcluded.The surfaceandbottomsalinities
wereaveraged andthisaveragesalinitywasassumedcharacteristic
for eachsegment.The
freshwaterfraction,F, for eachsegmentwas determinedas

WhereSi is the averagesalinityof the i-th segmentandSois the salinityof the Mobile Bay
stationimmediatelyoutsidethemouthof WeeksBay. The freshwaterfractionmultipliedby
the stationdepthandthe segmentsurfaceareagivesthe freshwatervolumefor thatsegment,
andsummingovertheentirebayproduces theexcessfreshwatervolumeof WeeksBay. The
total freshwaterflow was summedbackwards in time, beginningwith the day priorto the
survey,until the excessfreshwatervolumein thebay presentat the time of the survey,was
accountedfor. This lengthof time is the fill time.
MOBILE BAY WEEKS BAY FISH RIVER
Figure3' Longitudinal
salinity(psu)cross-section
fromMobileBaystation,alongthechannel
on the
easternsideof WeeksBay to theFishRiver station,survey1, on 20 May 1986.
A measureof the stratification

at eachstationin WeeksBay wasestimated
as thebottom
salinityminusthesurfacesalinitydividedby thestationdepth.Highvaluesindicate large,
verticalsalinitygradients,low valuesindicatesmall verticalgradients,and a zero value
indicatesa vertically
homogeneous watercolumn.Thisparameter is alsoindicative
of density
stratification
asWeeksBayis nearlyisothermal throughout theyear. A principal component
analysis wasperformed usingthisstratification
parameter for 16 stationsonthe14 surveys.
Only thosestationswithinthebayproperwereincluded,two stations,
onein boththeFishand
MagnoliaRiver mouthsand the stationin MobileBay wereexcluded.The principal
component
analysis
wasperformed
onthecovariance
matrixusingPRINCOMP(SAS,1985).
9.3 Horizontalandverticalsalinitypattems
In the followingsectionwe will hypothesize
thatspecifichorizontal
salinitypatterns
and

vertical salinitystructures,duringlow amplitudetidal periods,resultprincipallyfrom (1)

subtidalexchanges with Mobile Bay both in the barotropicand baroclinicalmodesand (2)
flashylocalrunofffrom theFishRiver. At a secondary level,salinityfieldscanbe influenced
by (1) bathymetry, particularlydeeperareaswhicheitherchannelnear-bottom waterwithinthe
bay or impoundbottomwatersin scourholesor depression areas,(2) local windsassociated
with moderateto high energyevents,and (3) high dischargefrom the MagnoliaRiver
fresheningtheeasternsideof thebay. The scenarios whichwe presentare serf-consistentand
consistentwith priormeasurements withinWeeksBay (Schroederet al, 1990b). Somecaution
is necessary.We are presentingscenarios of time-dependent eventsinferredfrom single
snapshots of the salinityfield and time seriesof the forcing. When the data might be
interpretedin differentfashions,we presentthat whichappearsmostreasonableto us.
9.3.1 -SubtidalExchange-
The salinitydistributions from surveys1 (Figs. 3 and 4) and 11 (Fig. 5) illustratetypical

patternsassociated with subtidalexchange events.Survey1 wasundertaken duringa falling
waterlevelperiodfollowinga majorintrusion event(Fig.2b) whichmovedhighsalinitywater
fromsoutheastern MobileBayintoWeeksBay. The salinitypatternin thelongitudinal section
(Fig. 3) suggeststhatat thepeakof theintrusion
bottomwaterswith salinitiesgreaterthan18
psuwereforcedup into themiddleof thebay asevidenced by the 18 psuwaterfoundin the
deepscourhole at the mouthof the bay (Fig. 3) and the elongateddomeof 18 psu water
locatedin the deepchannelon the eastsideof the lowerbay (Figs.3 and4b). In addition,
10 psu waterwould have extendedup into the lower reachesof the Fish River as evidenced
by the 10 psuwaterin the scourholenearthemouthof thefiver (Fig. 3). For manydays
prior to the survey,windswereweak,but persistent from the southeast.This wouldtendto
drive waterinto the bay becauseof Ekmanset-upat the coast. Water levelswithin Weeks
Bay werehigh. Immediatelypriorto thesurvey,windsshiftedaroundto blowfromthenorth,
causingwaterlevelsto drop. As freshsurfacewaterwasdrivenseaward, a deepreturnflow
may have beenmaintained.
As a resultof fallingwaterlevelsthehighsalinitybottomwatersmixed,spreadlaterally
to the westoverthe deeperportionsof the lowerbay andmovedseaward.At the sametime
lowersalinitywaterflowedseawardin thesurfacelayerandcoveredtheshallowbottomareas
adjacentto themouthsof bothrivers(Figs.4a and4b). Duringsurvey1, windswereweak
from the north(Fig. 2a) and fiver discharge wasjust comingunderthe influenceof recent
rainswitha moderate to highincreasein flow (Fig.2c) seenasverylow surfacesalinities
at
the headof the bay (Figs. 3 and4a).
Survey11 alsooccurred duringa fallingwaterlevelperiodfollowinga majorintrusion

event(Fig. 2b). Windswereveryweak(Fig.2a), butunlikesurvey1, survey11 followeda
10-dayperiodof modemeto highfiver discharge (Fig. 2c). High salinitybottomwaters
extended fromMobileBay intoWeeksBay (Fig. 5b). However,althoughwaterlevelsare
falling(Fig. 2b), thetightlypackedisohalinesin thebottomsalinityfield showno signof
mixingor spreading.Thisappears to be theresultof thestrongfreshwaterflow fromthe

162 W W Schroeder,S P Dinnel and W J Wiseman
FishRiverthatdominates theupperandwesternbay(Figs.5a and5b). Duringbothsurveys

salinitydistribution
patterns areconsistent
witha gravitational
circulation.
Becausethesalinity
surveys werecompleted within4 to 6 hours,duringlow-amplitude
equatorial
tides,we expect
thataliasingof the salinityfield waskeptto a minimum.
Figure4: Surface(a) andbottom(b) salinity(psu)for Week Bay for survey1, on 20 May 1986.
9.3.2 -Fish River Discharge-
The influenceof localrunofffromtheFishRivercanbe seenin thesalinitydistributions from

survey11 (Fig. 5) undertakentowardsthe endof a multipleday highfiver discharge period
(Fig. 2c) and from surveys1 (Figs. 3 and 4) and 6 (Fig. 6) madeat the time of short,
moderatefiver dischargeevents(Fig. 2c). During survey 11, high salinity waterswere
confinedto thedeepestportionof thechannelin thelowerbay (Fig. 5b), while surfacewaters,
exceptin the lower bay, andbottomwatersin the upperand westernbay were nearlyfresh
(Figs.5a and5b). As a resultof the largepulseof freshwater(Fig. 2c) the intrusionof high
salinityMobile Bay waterswas limited to the deepchannelin the lower bay wherea very
stronghaloclineformedwith a maximumverticalgradientof 14.2psubetweenthesurfaceand
1.5 m depths,in 2.25 m of water(Figs. 5a and 5b).
Also of interestis the obviouslack of influenceFish River dischargehason the surface

salinityat the Mobile Bay station(Fig. 5a). This is notedby the 4.1 psusurfacesalinityat
the mouthof the bay comparedwith the 15.1 psusurfacesalinityat the Mobile Bay station
(Fig.5a). However,low salinitysurface watersfromWeeksBaycouldmoveintoMobileBay
as a narrowplume, flowing along the shoreline. This conditionhas been observedin a

W W Schroeder,S P Dinnel and W J Wiseman 163
L,'mdsat
satelliteimage(not shown)whichrevealsa ribbon-likeplumeof turbidwaterexiting
WeeksBay andorientedto the westandnorthwestalongthe shorelineof Mobile Bay.
Figure5: Surface(a) andbottmn(b) salinity(psu)for WeeksBay for survey11, on 2 March 1987.
Figure6: Surface(a) andbottom(b) salinity(psu)for WeeksBay for survey6, on 15 October1986.

Duringsurveys1 and6, highsalinitybottomwaterswereobservedspreadoverthedeeper,

lowerandmiddlebay regions(Figs.4b and6b), while lower salinitysurfacewatersoccurred
throughoutthe bay. At the mouthof the Fish River during survey6 (Fig. 6a), very low
salinity waterswere measured. At the times of both of thesesurveysthe salinity fields
exhibitedwell developedhaloclinesovermostof thebay with thestrongest stratification
again
confinedto the region of the deepchannelin the lower bay (Figs. 4 and 6). Fish River
dischargewas insufficientto completelyflush salt from surfacewatersin the lower reaches
of the fiver (Figs. 3, 4a and 6a). It is possible,of course,that maximumfiver discharges
impactedthe bay after surveys1 and6 werecompleted.
9.3.3 -Bathymetry-
Bathymetrycontrolson the salinitydistributionmightbe expectedto be slightin watersas

shallowas WeeksBay (Fig. 1). This is not consistently the case. As notedabove,the deep
channelin lower Weeks Bay directsmuchof the subtidalexchangeof bottomwaterswith
Mobile Bay and permitsthe developmentof very stronghaloclines(seeFig. 5). In addition,
it is clearlyseenin Figure3 thatthedeepscourfeatureslocatedin themouthof thebayand
nearthe mouthof the Fish River canreadilyimpoundhigh salinitybottomwatersand,at least
temporarily,isolatethesewatersfromtheremainingbay-fiversystem.At times,muchsmaller
bathymetricvariationsalsoinfluencethe structureof bottomsalinityfields. For example,
tonguesof relativelyhigh salinitybottomwatershavebeenobservedin the deeper,eastern
portionsof the middle(Fig. 6b) andmiddle-upper (Fig. 8b) bay. Also,duringsurvey7A an
elongatedpool of 16.0 psu bottomwater (Fig. 7b) was found within the finger-shaped
depression whichlies betweenthenorthwestern shoreof WeeksBay anda submerged shoal
that extendsinto the bay from the north (Fig.l). This pool of bottomwater apparently
remainedintact,maintaininga smallsurfaceto bottomsalinitydifferenceof 2 psu,while the
remainderof the bay was nearlyverticallyhomogenized by moderatelystrongnorthwinds
associated with the beginningof a wintercold frontpassage(Fig. 2a).
9.3.4 -Local Winds-
Subtidalexchange processes areoftendrivenby Ekmanconvergences or divergencesforced

by non-localwinds(Schroeder et al, 1990b). At certaintimes,though,localwindforcingis
clearlyimportant. Survey7A was madejust as coastalAlabamawas comingunderthe
influenceof a wintercoldfront,while survey7B occurredfive dayslater,48 hoursafterthe
passage of thefront. The windsoverthisperiod(Fig. 2a) were(1) moderately strong(5 to
10m s'•) fromthenorthfourteen
hours
priortoandduring
survey7A, (2) strong
(8 to 16m
s'•) fromthenorth-northeast
tonortheast
for44 hours
aftersurvey
7A followed by38hours
of approximately
5 to 8 m s'• windsfromthenortheastto east,and(3) <5 m s'• windsfrom
thesouthwest
to west24 hourspriorto andduringsurvey7B. Comparison of Figures7 and
8 fromthetwosurveysindicatesa numberof significant
changes whichwerelikelydueto the
winds.

Figure 7: Surface(a) andbottom(b) salinitypsufor WeeksBay for survey7A, on 12 November1986.
Weeks Bay waterswere first nearlyverticallyhomogenized and alsowell mixed laterally

and longitudinally(Figs. 7a and 7b), presumablydue to wind mixing. Strongvertical and
horizontalgradientsoccurredonly in the upperbay at the mouthof the Fish River (Figs.7a
and 7b). After survey7A, at the time of the strongestnortherlywinds, bay waters were
rapidlyset-down(Fig. 2b) andwaterswithinWeeksBay werelikely completelymixed. Tl'.ere
wouldhavebeenvery strongmixingin Mobile Bay, also. When the windsrelaxed(Fig. 2a),
bay waterlevel rebounded (Fig. 2b) andWeeksBay refilledwith high salinityMobile Bay
water. Thesehigh salinitywatersmovedup into the deeperareasof Weeks Bay, appearto
have reacheda maximumintrusionm•d thensubsidedsomejust beforethe survey7B. This
can be seenin the distributionof bottomsalinities(Fig. 8b) wherea lens of 20 psu water in
the northeastcornerof the bay and 21.7 psuwaterat the mouthof the MagnoliaRiver have
bothbeenpinchedoff from the largepool of >20 psuwaterin the lower bay (Fig.8b). The
pool of 20 psuin the lower bay haslikewisebeendetachedfrom the highersalinity"coastal
ocean"in Mobile Bay (Fig. 8b).
High salinitywaterswerealsopooledon the westernsideof WeeksBay. This is seenin

both the surfaceand bottomsalinityfieldsand probablyresultedfrom upwellingalong the
upwindshoreof thisshallowbay. The effectsof the southwest and westwindsprior to and
duringsurvey7B canalsobe seenin thepositionof theFishRiver plumewhichwasoriented
along the eastern,downwindsideof the bay (Fig. 8a).

Figure8: Surface(a) andbottom(b) salinitypsufor WeeksBay for survey7B, on 17 November1986.
9.3.5 -Mobile Bay/Mobile River System-
In previouswork (Schroeder et al., 1990b),we haveindicatedthattheflushingtimeof Weeks

Bay is very short. Therefore,variationsin the salinityof Mobile Bay shouldbe readily
transmitted to WeeksBay. The salinityvariationsin Mobile Bay are normallyof long time
scale(weeksto months).Occasionally, whentheMobileRiver systemabruptlyfloods,shorter
time scalechanges(oneto two weeks)in salinitymay occur. Suchan eventresultedin the
low salinitiesobservedwithin WeeksBay duringsurvey12 (Fig. 9). Althoughthe discharge
from the Fish River was very low (Fig. 2c), the watersof WeeksBay are seento be nearly
fresh,as are thoseof Mobile Bay. Therefore,we assumethe bulk of the low salinitywater
withinWeeksBay originatedin MobileBay asa resultof MobileRiver Systemdischarge, not
from Fish River discharge.
9.3.6 -Magnolia River Discharge-
The MagnoliaRiver dischargeaccountsfor lessthan25% of the total freshwater flow to

Weeks Bay. Its plume is, consequently,
proportionallysmallerthan that of the Fish River.
Furthermore,sincethedrainagebasinsof bothriversarelocal,thesameweathersystems will
tendto forcebothfiver discharges.
An exampleof theMagnoliaRiverfreshening theeastern
sideof the bay canbe seenduringsurvey6 (Fig. 6a).

i •.'n
o.8'
0.9
.3.0 (
Figure9: Surface(a) andbottom(b) salinitypsufor WeeksBay for survey12, on 10 April 1987.
9.4 Fill time
FishRiver dischargeis flashy(Fig. 2c). Dischargesrespondrapidlyto localprecipitationthen

rerumto baselinelevelsin just a few days. Excessfreshwaterflow intoWeeksBay, relative
to Mobile Bay salinity, is assumedto be accountedfor by the total fresh water flow into
Weeks Bay. Most of the 14 surveysfollowed a noticeablefreshetand fill times were
generallybetweenoneandfourdays. The fill timewaszeroduringthetimeof survey10 (not
shown)whenthe entiresystem,WeeksBay andMobile Bay, hadvery low salinities,i.e. 1.5 -
2.5 psu,andtherewasno excessfreshwater. The longestfill time, six days,wasassociated
with survey11. A multiple-day,highdischarge periodhadoccurreda weekbeforeandWeeks
Bay appearedto havereceivedan influx of highersalinitybottomwaterfrom Mobile Bay a
few daysbeforesurvey11. A brief,butsharppeakin WeeksBay waterlevel suggests Mobile
Bay waterwasdrawnin alongthebottom(Fig 2b). Windsat thistime wereextremelyweak
andwe areprobablyseeingtheresultsof a significantgravitational circulation.A rapiddrop
in WeeksBay waterlevelthentrappedthehighersalinitywaterin thedeeper,lowerbay,and
high Fish River dischargecreateda fresh surfacelayer. Fill time for the excesswater in
WeeksBay is shortand is dependent uponthe lengthof time sincea locally-drivenfreshet.
The pulsesof freshwatercouldpossiblyremainin WeeksBay for morethanoneweek,but
no fill time neededmorethanoneprecedingfreshetto accountfor the excessfreshwater.

168 W W Schroeder,S P Dinnel and W J Wiseman
9.5 Salinity stratification

WeeksBay averagesalinitiesrangefrom 0 to 20 psu. Strongstratification, as well as strong
horizontalgradients,can occur within the bay when salinitiesare between5 and 15 psu.
Stratificationin Mobile Bay showsno significantcorrelationwith the stratificationin Weeks
Bay.
The principalcomponent analysisof thestratification empiricallypartitionsthevarianceof

the stratificationparameterinto orthogonalpatterns(or eigenvectors)known as principal
components.Eacheigenvector hasan associated eigenvaluethatdefinesthepercentage of the
totalvarianceexplainedby that eigenvector.The mostimportanteigenvectordescribes over
half the total variance,54%, of the stratificationparameter. This eigenvectorhas high
amplitudesof similarsign over muchof the bay indicatingthat mostof the salinitystations
respondcoherenflyto externalforcing(Fig. 10a). The salinitystationin the bay mouthand
thosestationsnearerthe fiver entranceshave low amplitudesand do not contributemuch
individualvarianceto thefirst eigenvector (Fig. 10b). The secondmostimportanteigenvector
describes 17% of the totalvariancein the stratification parameter.The highestamplitudes of
the secondeigenvector describes stratification
thatis localizedto the lowerregionof thebay.
Thereis little contribution
to thesecondeigenvector fromtheindividualstations fartheraway
fromthedeeper,lowerbay (Fig. 10d). Althoughtheremainingeigenvalues combineddescribe
39% of thetotalvariance,no additionaleigenvalue is statistically
differentfromrandomnoise
at the 95% level usingthe criteriaof Overlandand Preisendorfer (1982).
The firstprincipalcomponentcorrelates well with thearithmeticaverageof thestratification

parameterovereachsurvey,an R valueof 0.985. This impliesthatthe dominantstratification
patternis consistent
bay-wide. The first eigenvector is very similarto the patternof strong
stratification
whichoccurswhenthe surfacesalinifiesare uniformlyfresherthanthe higher
salinitybottomwater(for exampleseeFig. 4). The secondprincipalcomponent correlates
with one-andtwo-dayold FishRiverdischarge, with R valuesof 0.66 and0.65 respectively.
Higherfiverdischarge is associatedwithstrongerstratification
in thelowerbay. Thissuggests
that local dischargedoeshave someinfluenceon the stratificationin WeeksBay. This
principalcomponent is strongestfor surveyswith highsalinitybottomwaternearthe bay
mouthandlow salinityoverthe restof the bay (seeFig. 5b). It may indicategravitational
circulationdrivenby variationsin fivefineinputto the bay.
The timing of our surveyswas biasedaway from periodsof strongtropicfides. The

mechanical mixingof the surfaceandbottomsalinitiesin WeeksBay, because of weaktidal
currents,may be muchlower thanneededto breakdownthe observedstratification.In Mobile
Bay proper, stratificationappearsto break down when strong tidal currentsoccur in
conjunction with significant
waveactivity(Schroeder
et al., 1990a).It is notyetclearwhether
a similarrelationshipholdsin WeeksBay.

Figure 10: First and secondeigenvectors of stratification

parameter.Amplitudesfor first eigenvector
(a) whichhas an eigenvaluethatdescribes 54% of the totalvariance;percentage
of individualstation
variancecontributed to first eigenvector
is shownin panel(b). Amplitudesfor secondeigenvector (c)
which has an eigenvaluethat describes17% of the total variance;percentageof individualstation
variancecontributedto secondeigenvectoris shownin panel(d).
Acknowledgements
This work is the resultof researchsponsored
by the U.S. Departmentof Commerce,NOAA,
NOS, Officeof OceanandCoastalResource Management, MarineandEstuarine
Management

Division underGrant No. NA86AA-0-CZ017, The Universityof AlabamaMarine Science

Program,the MarineEnvironmental SciencesConsortiumof Alabamaandthe Mississippi-
AlabamaSeaGrantProgram(GrantNo. NA85AA-D-SG005;ProjectsR•R-16 andE/O-16).
Partialsupportwas alsoreceivedfrom the U.S. GeologicalSurveyWetlandsLossProject,
Contract No. 14-08-0001-23411. T.R. Matthews served as the Graduate Research Assistant
for theproject. J.L. PearmanandH.H. Jeffcoatcoordinated
cooperative
monitoringwith the
USGS.
ContributionNo. 162 of the Aquatic Biology Program,The Universityof Alabama;

ContributionNo. 193 of the Marine EnvironmentalSciencesConsortium,DauphinIsland,
Alabama; and ContributionNo. 0117 of the Center for Marine Science,University of
SouthernMississippi.
9.6 References
Amorocho, J. and J.J. DeVfies, 1980: A new evaluation of the wind stresscoefficient over water
surfaces.J. Geophys.Res., 85:433442.
Amorocho,J. andJ.J.DeVfies, 1981: Reply. J. Geophys.Res.,86:4308.
Crance,J.H., 1971: Descfiptionof Alabamaesmarineareas- cooperativeGulf of Mexico estuarine

inventory.AlabamaMar. Res. Bull., 6, 85 pp.
Dinnel,S.P.andWm. J. Wiseman,Jr., 1986: Freshwateron theLouisianaandTexasshelf. Cont.Shelf

Res., 6, 765-784
Overland,J.E. andR. W. Preisendorfer,

1982:A significance
testfor principalcomponents applied
to a cycloneclimatology. Mon. WeatherRev., 110, 1-4.
SAS Institute,Inc., 1985: SASUser'sGuide:Statistics,

Version5 edition. Cary,NC. 956 pp.
Schroeder,
W.W., S.P.DinnelandWm.J.Wiseman,Jr., 1989: Stratification
variabilityin a tributary
estuary. Abstracts,Tenth BiennialInternationalEstuarineResearchFederationConference,Baltimore,
Maryland,Oct. 1989. p. 105.
Schroeder, W.W., S.P. Dinnel and Wm.J. Wiseman, Jr., 1990a: Salinity stratificationin a river-
dominatedestuary. Estuaries,13(2):145-154.
Schroeder,W.W., Wm.J. Wiseman, Jr. and S.P. Dinnel, 1990b: Wind and fiver inducedfluctuationsin
a small, shallow,tributaryestuary. ResidualCurrentsand LongtermTransport,R.T. Cheng,Ed.,
Spfinger-Vedag,pp. 481-493.

U.S. Departmentof Commerce,1986: ClimatologicalData, Alabama,Volume 92. NationalClimatic

Data Center,Asheville, NC. 264 pp.

Data Center,Asheville,NC. 276 pp.

Data Center,Asheville,NC. 276 pp.

10
On meteorologicallyinduced subtidal
motions in Hangzhou Bay
J L S u and W Chen
Abstract
Multiple input spectralanalysisis appliedto subtidalsealevel oscillationsinsideHangzhou

Bay whichis largeandshallow.Amongthethreeforcingfactorsfor thesealeveloscillations:
The coastalsealevel is the dominantone; for periodslessthan 10 daysthe local along-bay
wind stresscomponentis alsoimportant;the localcross-baywind stresscomponent hasonly
minorimportance. One dimensional linearinviscidmodelis shownto be usefulto describe
the subtidalbay oscillations. The analysisshowsthat, whenthe inputsare highly correlated
multipleinputspectralanalysiscanencounterdifficulty.
10.1 Introduction
HangzhouBay is a largefunnel-shaped
bay alongthe EastChinaSeacoastof China (Fig.l),
whose width is about 100 km at the mouth and narrows down to around 20 km some 100 km
to the west. Severallargecitiesare locatednearthe bay and a nuclearpowerplant is being
built on the north shore of the bay. Consequently,there have been interests in the
understanding of the environmentalqualityof the bay water.
The bay is shallowwith an averagedepthof about10 rn andthe averagetidal rangeat the

baymouthis around2.5 m. Thus tidal flushingis probablythe dominantmechanismfor the
bay-shelf water exchange. Nevertheless,it is also of interest to understandhow
meteorologicallyinducedoscillationscontributeto this water exchange. In this study,
meteorologically
inducedsubtidalmotionsin Hangzhou
Bayareinvestigated,
usingmainlythe
tidal and meteorologicaldata.

174 J L Su and W Chen
120'E 122'E
32'N
oSS
• vsc
ZAP
•SNS
oo CT
30 'N
Figure1:Locationsof datastations:
tide(circle);meteorology
(solidcircle);wind(solidtriangle).Double
concentriccirclesdenotethe Shanghaicity.
10.2 The Data
The data stationsare shownin Fig. 1. The two wind stations,YSC and SNS, providethe
remote (over the shelf) wind data and the wind stationZAP gives the local (over the bay)
wind data. Throughout thisstudytheterm"windstress" refersto the quadraticwind velocity
partof the usualwind stressformula. For the wind stresscomponents the east(cross-shelf)
is takenas the positivedirectionof the x-axisand the north(alongshelf)the y-axis. The sea
level stationCT providesthe dataof the coastalsealevel whichalsoforcesoscillationsin the
bay. In thefollowingwe will usethetermscross-shelf andalongshelfwindswhendiscussing
the coastalsealevel oscillations.Methodsandformulasusedin dataanalysisare thestandard
ones(e.g., Mooresand Smith, 1968;Jenkinsand Watts, 1968).
The dataperiodusedspansfrom 12.1973to 12.1974. All thetime seriesareerrordetected,

low-passedand then decimatedwith a 24-hour intervalbeforeuse. A Lanczos-cosine f'fiter
(half-powerpoint of 40 hr) is usedto low-passthe hourlysealevel recordsand a weighted
symmetricfilter (half-powerpoint of 54 hr) to low-passthe 6-hourly wind stressand

J L Su and W Chen 175
atmospheric pressurerecords. All sealevel recordsare alsobarometricallyadjusted. Two

subsetsof the dataperiod,12.1973to 3.1974 and6.1974 to 9.1974, are chosento represent
the winter andthe summercharacteristics,
respectively.
Tukeywindowis usedin thespectral computations. For thecomplete recordthefrequency

resolutionis 0.011 cpd and the equivalentdegreesof freedomis 8.2, while for the seasonal
recordstheyare0.045cpdand10.7,respectively.Largerdegree-of-freedom Tukeywindows
have alsobeenemployedwith no effectson the main conclusions.
10.3 Multiple input system

It is knownthatthe effectof meteorological forcingon the sealevel oscillations
of a narrow
estuarycanbe resolvedinto the localandremotecomponents (WangandElliott, 1978). The
formeris forcedby localwindsalongthe thalweg,settingup a sealeveldifferencealong
the estuary. The latter is induced by the coastalsealevel oscillationsat the mouthof the
estuary, whichpropagates with little attenuation
into the estuary. Both components are
importantfor oscillationswithperiodsof a few days. For longerperiodoscillations, however,
the remoteeffectis expectedto dominate (Garvine,1985).
HangzhouBay is ratherwideandhasa gentlebottomtopography. Cross-bay windsmay

alsocontribute significantlyto thesealeveloscillations
in thebay. Thusfor oscillationsin
Hangzhou Bay with shortperiods,say,lessthan10 days,we havea three-input systemand
the inputsare the coastal(CT) sealevel andthe two components of the localwind stress.
For longperiodoscillations in HangzhouBay, however,only the coastalsealevel is needed
astheinput. As will be demonstrated later,coastalsealeveloscillations
withperiodsof a
few daysarecausedmainlyby thealongshelf winds,whereaslongerperiodoscillations are
caused by bothalongshelf andcross-shelf winds. Therefore,for thelongperiodcase, we
may alsotreat the bay systemas havingtwo inputs,namely,the two components of the
remotewind stress. Becauseof thelargeextentof a weathersystemthe localwindsandthe
shelfwindsarestatistically correlated.The twowindstress components at thesamestation
alsocorrelatewith eachother. Therefore, in bothfrequency rangetheHangzhou Bay sea
leveloscillationis theoutputof a multipleinputsystemwhoseinputsarecorrelated.
Forsubsequent discussions
we lista fewformulae of multipleinputspectral
analysis.For
simplicity,
considera linearsystemwith oneoutput(u) and two inputs(xI and x2).
Assuming thatXl andx: arecorrelatedit canbe shown(ChenandSu, 1991)thattheoutput
spectrumof this systemcan be written as
S,., = E•a + El2+ Ez•• + E21+ SR
whereE•a= I-I• Susa

= R:ul/2 Su•
a,

E•2 = I• (Su•- Su•a),
andlikewiseforEznandE2•. "S"denotes
eithertheautospectrum
or thecross-spectrum,
"H"
the frequencyresponsefunction and "R" the coherencesquared. The slashsymbolin the
subscriptof eithera spectrumor a coherence squared
indicates
a "partial"quantity.El/2is the
contributionof the uncorrelatedpartof x• to S•uandE•: is the contribution
to Su•dueto the
partof x• thatis correlated
with x:. We takethemodulusof (El/2+ E•2)asa measure of the
contributionof x• to S•. The meanings of E• andE,.•canalsobe explained accordingly.
We notethat the sum(F_q: + E:•) is a real number.
Similar formulaecan be derivedfor linearsystemswith threeor morecorrelatedinputs

(Chen and Su, 1991).
When a systemis studiedwith a multipleinputspectralanalysistwo problemswill arise

if theinputsarehighlycorrelated.Firsfly,unlesstherearecausalrelationsamongthehighly
correlatedinputsand unlesstheserelationsare also studied,interpretation
of partialcross
spectrumanalysisresultsmay be difficult.Secondly,a frequencyresponse functiondepends
on the ratioof a partialcrossspectrumovera partialautospectrum
(BendatandPiersol,1980).
When two inputsare highly correlatedwith eachotherand one of theseinputsis highly
correlatedwith the output,the varianceof the estimationof the correspondingfrequency
responsefunctionis likely to be large.
10.4 Results and Discussions
10.4.1 Sea level oscillations
Fig. 2 givesthespectraof thesealevel oscillations bothinsidethebay (GP) andat thecoast

(CT). Thegeneralformsof thetwospectra arequitesimilar,especially at thelow frequency
end.Insidethe bay the spectrum hasa prominentpeaknearthe 15 day periodwhichis not
evidentat the coast(Fig. 2). This is due to aliasingof the diurnaland semi-diurnaltides
whichhavenotbeenentirelyremovedby thefilterused(FangandYu,1985). Amplification
of thesemi-diurnal tidesby thefunnel-shaped baycontributed to thishighaliasedpeak. The
spectrum of thealong-baysealeveldifference betweenGP andCT is alsoplottedin Fig. 2.
For periodslongerthan 10 daysbut awayfroma narrowbandnearthe 15-dayperiodthis
spectrumis significantlylower than the sealevel spectrumof GP, confirmingGarvine's
suggestion(1985) that the remoteeffect dominatesthe sea level oscillationsin estuaries.At
all frequenciesthespectrum of thetransverse-baysealeveldifferencebetweenJSandHH was
foundto be lowerthantheJS sea-level
spectrumby a factoraround10, suggesting
thatthe
effectof thenorth-south
componentof thelocalwind-stress
isprobably
of minorimportance.

J L $u and W Chen 177
5
• •o•- • 105-
E
4
7. 10.- 7' lo,-
GI:' >.
I-- 3.
•
z
lO ...... I--10-
Z
',," "•;'v
• >'- 10•-_
• •o•
z '• z
lO- ! i
•7 i i
I LU 10 I I I
0.01 0,1 1 0,01 0,1
FREQUENCY (cpd) FREQUENCY (cpd)
Figure 2: Sea level spectrainsidethe bay (GP) and at the coast(CT) (solid lines), and sea level
difference
(GP-CT)spectrum (dashed line). Invertedtriangleontheabscissa
indicates
the15 dayperiod.
(The 95% confidenceinterval is •narked)
105.
104-
t
103-
102•.
10,
O. OI m O.Im • m
I 10- m
0.01 I J,
0 I I
Figure3: Spectraof thewind-stress

components
overtheshelf(YSC). (The95% confidence
interval
is marked)
The wind-stressspectraof YSC are given in Fig.3. The north-south (alongshelf)

wind-stress
spectrum
is in generalslightlygreaterthantheeast-west
(cross-shelf)
wind-stress

178 J L Su andW Chen
spectrum.
Spectral
peaks
atfrequencies
corresponding
tothepeaks
ofthesealevelspectra
are
alsofoundin oneor bothof thewind-stress
spectra,
including
thefortnight
frequency
(Fig.
2). At thehighfrequency
endthewind-stress
spectra
dropsoff moresharply
thanthesea
levelspectra.
Thisisbecause,
atthehighfrequency
end,thesealevelinside
thebayisalso
excitedby continental
shelfwaves(ChenandSu, 1991).
1.0
C? v$ %,
rD
w 0.5
0.0
0.01 0.1 1
FREQUENCY (cpd)
]-0-
CT vs %,
0.5,'
0.0-
0.01 o'.1
FREQUENCY (cpd)
Figure4: Squared
coherences
between
thecoastal
sealevelat CT andtheshelf(YSC)wind-stress
components.(The95% significance
levelis at thehorizontal
dashed
line)
The squared
coherences betweentheshelf(YSC) windstress componentsandsealevels
at thecoast(CT) andinsidethebay(GP) areshownin Figs.4 and5. Thefiguressuggest that
subtidalmotionsinsideandoutsidethebaywithperiodsup to 70 daysaredriven,directlyor
indirectly,
by meteorologicalforces,exceptfor oscillations
nearthefortnight
period.In the

followingwe will ignorethesealeveloscillations nearthefortnightperiodunlessotherwise

stated.At thecoast(Fig.4) thealongshelf wind-stress component seemsto beresponsiblefor
mostof the sealevel oscillations at periodsshorterthan 10 days,whereasfor longerperiods
bothalongshelf andcross-shelf windsareimportant.Insidethebay(Fig. 5) bothwindstress
components are influentialfor the sealevel oscillations. For periodslongerthan 10 days
influenceof the east-westwindson the bay oscillationsis mostlyindirectthroughthe coastal
sealevel oscillations(Figs.2and 5). For shorterperiodsthis influencealso comesvia the
localwindsdirectly,resultingin sealevel differences alongthebay (Figs.2, 4 and5). As
mentionedabovedirecteffectof the north-south components of the local windsis probably
of minorimportance.
1.0-
0.0'
0.01l 0.1 1
FREQUENCY (cpd)
,.o] tip vs 'rv
rr 0.5'
0
0
I I
O.Ol o.1 1
FREQUENCY (cpd)
Figure 5: Squaredcoherences
betweenthe bay sea level at GP and the shelf (YSC) wind-stress
components.(The 95% significance
level is at the horizontaldashedline)

104
WINTER SUMMER
GI:
103
GP
JS
I0
JS
IO
(GP
HH
IO
CT
(JS)
CT
10-
( HH
I0 I ! I I ,
(CT)0 0. I 0.2 0.3 0.4 0 5 0 0. I 0.2 0.3 0.4 0.5
FREQUENCY (cpd)
Figure 6: Sea level spectrainside the bay (GP, JS, HH) and at the coast(CT) during winter and
summer.
To seebetterthe spectralcharacteristics
at the highfi'equency
endtheseasonal spectraare
computed.Againthesealevel spectrainsideandoutsidethebay havesimilarforms(Fig.6).
The winter sea-levelspectrahavepeaksnearperiodsof 7, 4.5, 3 and 2.5 days,whereasin
summerthe peaksare closeto periodsof 10, 4.5 and 3.3 days.Comparedwith the summer
casethewintersea-levelspectraarenoticeably higherat periodsshorterthan5 days.Thisis
probablydueto theeffectof theshelfwaves(ChenandSu, 1991)whicharemoreenergetic
in winter(ChenandSu,1987).Peaksat periodscloseto positions of sealevelspectrapeaks
are also found in the respectiveseasonalspectraof one or more of the four wind stress
components(Fig. 7). The local windsare more attenuatedthanthe shelfwinds. However,
at the coastthe east-westwindsare moreenergetic,whereasinsidethe bay the north-south
windsare stronger.

10s] WINTER SUMMER
-
10
3
\ .._•/
10_ ., i-_ • • i I i I I -
0 0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
FREQUENCY (cpd)
Figure7: Wind-stressspectraover the shelf(SNS) andinsidethebay (ZAP) duringwinter andsummer.

(xx - solidline;•y - dashed
line)
Basedon the aboveresultswe will apply the cross-spectral

analysisseparatelyon two
bandsof thesubtidalmotionsin HangzhouBay, oneshorterthan10 daysandtheotherlonger.
10.4.2 One-dimensional model
AlthoughHangzhouBay is rather wide, its axis is alignedmore or less in the east-west

direction.A one-dimensional verticallyintegrated
modelwasconstructed to studythesubtidal
sea-levelresponses of the bay both to the east-westwinds over the bay and to the coastal
sea-leveloscillations(Chen and Su, 1991). Scale analysisshowedthat the advection,
time-dependent andbottom-friction termscanall be neglectedfrom the momentumequation.
The solutionshowsthat the variablebaywidthaffectsonly the transport.
Whenonly the sealevel at thebaymouthis allowedto oscillatewith certainfrequencythe

sealevel insidethe bay oscillatesin-phasewith the sameamplitude.When only a uniform
oscillatingeast-westwind is appliedover the bay the sealevel insidethe bay is foundto
oscillateat 180 degreesout-of- phasewith the wind. In this case the amplitudeof the
oscillationinsidethe bay variesas Cp;dxl/pgH. 'c is the east-westwind stress(dimension
(m/s)2),
C thewindstress
coefficent,
patheairdensity,
p thewaterdensity,
g thegravitational

acceleration,
H the averagewaterdepth,andx the distancemeasured from the bay mouth.
Therefore,as far as the effect of the local east-westwindsis concerned,the modelpredicts
that the amplitudeof the sealevel response at GP is abouttwiceas largeas thatat JS.
Equivalently,the sealevel insidethebayhasa frequencyresponse functionof unitywith

respectto thecoastalsealevelanda frequency
response functionof Cp•lxl/pgHwith respect
to the east-west wind stress.
10.4.3 Multiple input spectralanalysis:periodsshorterthan 10 days
Analysisgiven aboveof the shortperiodoscillationssuggests that the effectof the local

cross-bay,i.e., north-south,
windsis of minorimportance.To quantifythis statementwe first
considerthebay oscillationsas a systemwith threeinputs,namely,the coastalsealevel (qcr)
andthetwocomponents (%,Xy)of thelocalwinds.Applyingthethree-input
spectral
analysis
to thewinterdatathepercentage
contributions
of qcr,%,Xyandtheresidualto thesealevel
spectrum
(0.1-0.5 cpd)at eachbay stationhavebeencomputed
(Table 1). We recallthatthe
contributions
of qcr,% andxyaremeasured
by themoduliof complex
numbers.
Hencethe
sumof the four percentagecontributions
at eachbay stationcanexceed100 for eitherseason.
Table 1: Percentage
contributions
to the winterbay sea-levelspectra(0.1-0.5 cpd):three-input
model.
Station GP JS HH
Local winds
33 14 9
14 12 10
Coastalsealevel, qcr 78 93 96
Residual 11 3 2
Table 1 showsthatat eachstationthecontribution

fromqcr is overwhelming.
It alsoshows
thatcontributions
from'•, and'•yarecomparable,
contrary
totheearliersuggestion.
However,
the threeinputs
arehighlycorrelated
withoneanother
(Fig.8),especially
between
'•yandqcr,
and with the bay oscillations(e.g., Fig. 9). The compositionof the contributionsalsoreflect
the effectsof thesehigh correlations. At GP, for example,the ratio of the correlatedpart to
theuncorrelated
partof eachcontribution
isaround
5.4fora:y,1.8fora:,,and1.9forqcr. As
pointedoutearlier,for a highlyinter-correlated multipleinputsystemlargevariances maybe
associated with the estimationof the frequencyresponsefunctions,which in turn will affect
the estimatesof the contributions of therelatedinputs.The threewintergainfunctionsat GP
are given in Fig.10. The gain functionwith respectto qcr has large variationsfrom the
one-dimensional resultof unity. Similarresults are alsofoundat the othertwo stations.The
coastal
sealevelgradient
dueto meteorological
forcingis usuallyof theorderof 10'?in the
alongshelfdirectionso that the one-dimensional
model shouldbe a reasonablygood
approximation.Onepossiblereasonfor theselargedeviations
is dueto the highcorrelation

J L Su andW Chen 183
amongthe inputs.
I'01 WINTER
,, ,.-:, :':
I'\, I•, A t,, • '
• '•'11•/ : ..,. • •
,,z,
o.s•- •:'--•- •-,-• •'-/• .... •1-•- • -•--
0.0
FREQUENCY (cpd)
Figure
8:Thewintercoherence
squared
between
'riotandXy(dash-dotted
line),'riotandx, (dashed
line)
andx, andxy(solid
line).(The95%signific,'mce
levelisatthehorizontal
dashed
line)
r'r' 0ß 5.
UJ
o
o
0.0'
FREQUENCY (cpd)
Figure9: Thewintercoherence
squared
between
thesealevelat GP and(a) •lc•-(dashed
line),(b) x,
(dash-dotted
line)and(c)Xy(solid
line).(The95%significea•ce
levelisatthehorizontal
dashed
line)

FREQUENCY (cpd)
Figure10: The wintergainfunctionof thesealevelat GP withrespect

to •lcr (dashed
line),
(dash-dotted
line)andXy(solidline). (Three-input
modelresults)
Table2: Percentage
contributions
to theseasonal
bay sea-level
(0.1-0.5cpd):modifiedthree-input
system.
, ,
winter summer
Station GP JS HH GP JS HH
Local winds
•x 30 17 12 36 31 25
'cy 10 9 14 6 4 11
Coastalsealevel, qcr 64 85 79 65 75 72
Residual 18 7 7 15 11 8
Basedon the abovearguments

we construct
a modifiedthree-input
systemwherethe
frequency
response
function
withrespect
to•lcTisassigned
thevalueof unitya priori.Using
slightlymodifiedformulaefromthoseusedin computing
thevaluesin Table1 we find,for
themodified
system,
thepercentage
contributions
of •lcr,xx,xyandtheresidual
tothebay
sea-levelspectra
aslistedin Table2. Contribution
from•lcr remainsdominant
forall stations.
Except during winterat HH,contributionfi'om'lJy
is in generallessthanhalfof thatfromxx.
Station HH islocated in shallowwatersouthof anextensive tidalflat.Strong
northerlywinter
winds arelikelythereason for,atHH,thecontribution from'cytobecomparable tothatfrom
x•. At all stations
thecontribution fromtheresidual is largerthanthecorresponding value
computed fromtheoriginal system(Table1). Thisisexpected sincethefrequency response
function related to •lcris predetermined
andnotby formulae equivalentto a leastsquare

J L Su andW Chen 185
regressionmethod.
31 WINTER
SUMMER
2
z z
<
1
1 ...
I •r, i L_.' -" ", / -'•

0 ! ! i i
0.0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 O.Zl 0.5
FREQUENCY (cpd)
FREQUENCY (cpd)
/ f I L/ f II I I
,,,,/ ! •
:'U"l f f f I; I
/, f
_u.:] Ill
I l, I I I I
-9oJ !' , ' t ,
I:' ,--'..
_180:!, "/ ,
0.0 0.1 0.2 0.3 0.4 0.5
Figure11' Thegainandphasefunctions
of thesealevelat GPxvithrespect
to qcr(solidline)andxx
(dashedline). (Two-inputmodelresults)
Fromtheabovediscussions we therefore
ignoretheeffectof thecross-bay windstress and
treatthebay oscillation problemas a two-inputsystems. The two inputs,qcr and%, are
correlatedbutnottoo highly.The computed frequencyresponse functions
at GP andJS are
givenin Figs.11 and 12. The gainfunctions at GP showreduced oscillations
thanthe
previousresults obtainedfromthethree-input system(Fig.10).Themeanvaluesof thegains
over0.1 to 0.5 cpd are listedin Table3. The one-dimensional modeldiscussed before
predictsa constant unityvaluefor thegain withrespect to qcr and CpJxl/PgH for thegain
withrespect to 'c,. Thecomputed gainwithrespect toqCThasaverage valuesranging from
1.03to 1.24. If we takeC=l.5x10'3, p•/p= 1.25x10 '3, H=10 m andIxl =100 kin, the
distanceof GP fromthebay mouth,thepredicted valuefor thegainwithrespectto 'c,at GP
is 1.9x10
'3m/(m/s) 2,whichis one-third
to one-halfthevaluesgivenin Table3. Thepredicted

gainwithrespectto % at eitherJSor HH shouldbe onehalf thatat GP. The computed ratio

rangesbetween0.35 and0.83. As to the phasefunctionsthe modelpredictsa constant zero
phasewith respectto qcr anda constant180 degreephasewith respectto x,. The computed
results(Figs. 11 and 12) agreewell with thesepredictions.Therefore,overallspeaking,
the
one-dimensional modelcanbe usedto describetheresponseof HangzhouBay to the forcing
of qcr andx: for periodslessthan 10 days.
31 WINTER SUMMER
O1 ' 0
0 0.1 0.2 0 3 0.4 0 5 0 0.2i t
0.3 l
0.4 0.5
Figure 12: The gain functionof the sealevel at JS with respectto qcw(solidline) andx: (dashedline).
(Two-input modelresults)
Table 3: Mean valuesof the gainsof the bay oscillationsover 0.1 to 0.5 cpd.
Winter Summer
Station GP JS • GP JS •
Forcing
coastalsealevel, qcr 1.08 1.03 1.24 1.22 1.15 1.21
east-westwind stress,% 6.2 3.4 2.2 3.5 2.9 2.7
(10-3m/(m/s)
•)
If we usea largerdegree-of-freedom
for the spectralcomputations
the oscillations
in the
cross-speclral
resultswill becomesmaller. The conclusionsremain,however,unchanged.
10.4.4 Multiple inputspectralanalysis:periodslongerthan 10 days
Thecoastalsealevelis drivenby boththeshelfwindsandshelfwavespropagating downfrom

the north.The latterusuallyhaveperiodslessthan 10 days. As discussed
earlier,for short

period
coastal
sealeveloscillations
thealongshelf
windstress
component
dominates,
whereas
forlongperiod
oscillations
thecross-shelf
windsu'ess
component
becomesequally
important.
Thetwo-input
systemusedabove does notworkwellherebecause
thetwoinputs,
thecoastal
sealevelandeast-west localwindstress component,
arenowhighlycorrelated.However,as
discussedbeforeit is sufficient
to usethecoastal
sealevelastheonlyinputto studythelong
period
oscillations
inside
thebay.Alternatively,
wecanalsousethetwocomponents
ofthe
shelfwindstressas theinputs,alsodenoted
respectively
as x• andXyin thefollowing
discussions.
In thiswaywecanalsogainsome inforlnation
ontheeffectofeast-west
winds
over the bay.
o l0a_
i I
-2
FREQUENCY (10 cpd)
Figure 13' Contributionstothecoastal

(CT)sealevelspectrum
(heavy
solidline)fromx•dash-dotted
line),'r.y
(lightsolidline)andtheresidual
(dotted
line).
It wasfoundfromthe coherencesquaredbetweenXxandxythatthetwocomponents are

correlated
butnottoohighly. We firstapplythetwo-input
modelto thecoastal
sealevel
oscillations.
Contributions
fromXx,'cy
andtheresidual
tothecoastal
(CT)sealevelspectrum
areshown in Fig.13. Evidently
thecontribution
fromXydominates
overthosefrombothx•
andtheresidual
at periods
shorter
than12 days,whileat periods
longerthan20 daysboth
windstress
componentsareimportm•t,
confirming
thearguments givenabove.Nextweapply
thetwo-input
modelto the bay oscillations.
In Fig. 14 we showtheresultsat GP. At

periodsgreaterthanabout18 dayscontribution
from'cxto the sealevel oscillations
at GP is
actually
moreimportant thanthatfrom•y.Thissuggests
that,liketheshortperiodcasestudied
above,thecontribution
fromthenorth-south
component
of thelocalwindstressissignificantly
lowerthanthatfromtheeast-west
component.
Wenotethedominant
contribution
from
aroundthe15-dayperiod.Thisis anartifactsince,aswe commented onearlier,thepeakof
theGPsea-level spectrumatthe15-day
periodisduemainlytoaliasingfromtheincompletely
removed tides.In Table4 thepercentage
contributions
from'cx,'Cy
andtheresidual tothesea
level spectrum(0.01 to 0.06 cpd) at GP, JS, HH andCT are listed.
ß .
ß
ß
ß
I I I -1
3 6 9 12
FREQUENCY
(10-2cpd)
Figure
14:Contributions
tothebay(GP)sealevelspectrum
(heavy
solidline)fromxx(dash-dotted
line),
xy(lightsolidline)andfl•eresidual(dottedline).
Table4. Percentage
contributions
to thesealevelspectra
(0.01to0.06cpd): two-input
model.
Station GP JS HH CT
'c• 69 69 64 49
'Cy 28 25 33 45
Residual 12 15 13 16
,

2•
(a) (b)
o o
o o 3 6 9 12
FREQUENCY
(10-2cpd
) FREQUENCY
(10-2cpd)
Figure15. The gainfunctions

of thesealevelat (a) CT and(b) GP withrespectto % (dash-dotted
line)
and'•y(solidline).
Thegainfunctions
at CT andGPareshownin Fig. 15. Thetwogainswithrespect
to
are quite similar,indicatingthat the contributionfrom the cross-baywinds to the bay
oscillations
is of minorimportance.The gainwith respectto % at GP is higherthanthatat
CT by about10 m/(m/s)
2. Thisdifference
is abouthalf of the valuepredicted
by the
one-dimensionalmodelfor the gain at GP due to east-westwind stressover the bay. The
phasefunctionsat CT and GP were foundto vary closeto 180 or -180 degrees.This is
expectedsincebothnortherlyandeasterlyshelfwindswill lowerthe coastalsealevel andthe
easterlywindsover thebay will lowerthe sealevel at GP.
10.5 Summary
Applyingmultipleinputspectralanalysisto sealevel dataof HangzhouBay it is foundthat
a. Coastal sea level has the dominant influence on the subtidal sea level oscillations inside
thebay. The localwind stresscomponent alongthe axisof the bay is alsoimportantfor
the bay oscillationsat periodsshorterthan 10 days. The local cross-baywind stress
component is of minorimportancefor theseoscillations.
b. Linearinviscidone-dimensional modelcanbe usedto describethe sealevel response of
the bay to forcingby both the coastalsealevel and the alongbay winds.
This studydemonstrated that,whenthe inputsare highlycorrelatedmultipleinputspectral

analysiscan encounterdifficulty.

10.6 References
Bendat,J.S.,andA.G. Piersol,1980:Engineering
Applications
of CorrelationandSpectralAnalysis,John
Wiley & Sons.
Chen,D.K., andJ. L. Su, 1987: Continentalshelfwavesalongthe coastof China,Acta Oceanologica

Sinica, 6, 317-334.
Chen, W., and J. L. Su, 1991' Low-frequencysea level fluctuationsin the HangzhouBay, Acta
OceanologicaSinica, 10, (to appear).
Fang,G. H., andZ. W. Yu, 1985:On calculation

anduseof daily meansealevel,II. On useof daily
meansealevel data,J. Oceanogr.Huanghai& Bohai Seas,3(2), 1-9 (in Chinese).
Garvine,R., 1985:A simple modelof esmarine subtidalfluctuations

forcedby localandremotewind
stress,J. Geophys.Res. 90 (C6), 11945-11948.
Jenkins,G.M., andD.G. Watts,1968:Spectral

Analysisandits Applications,
Holden-Day,525 pp.
Moores,C. N. K., andR. L. Smith,1968:Continental

shelfwavesoff Oregon,J. Geophys.
Res.,73,
549-557.
Wang,D.P., andA. Elliott, 1978:Non-tidalvariabilityin Chesapeake

Bay andPotomac
River,evidence
for non-localforcing,J. Phys.Oceanogr.,8, 225-232.
Walters,R.A., and C. Heston,1982:Removingtidal-period

variationsfrom time-series
data using
low-passdigitalfilters,J. Phys.Oceanogr.,12, 112-115.

11
Water level fluctuations in the Atchafalaya

Delta, Louisiana: tidal forcing versus river
forcing
E M Swenson and C E Sasser
Abstract
The Louisiana coastal marsheswere built over the last 8,000 years by the successive
formation, abandonmentand subsequentsubsidenceand erosionof delta lobes from the
MississippiRiver. The AtchafalayaRiver Delta is one currentsite of depositionand land
buildingin thisprocess.Sedimentsof the AtchafalayaRiver havebeenfilling Atchafalaya
Bay since1950,and new islandsfirst emergedin 1973. This new land is beingformedin a
coastal
regionwhichis currently
experiencing
landlossat a rateof about155km"peryear.
The surrounding marshesare basicallyfreshdue to the presenceof the fiver but also havea
waterlevel signalthatis stronglyinfluencedby astronomical tides,atmospheric forcingand
the AtchafalayaRiver itself. The resultingwaterlevel regime(inundationand frequency)is
a key elementcontrollingthe marshvegetationdynamicswithin thisdeltaicsystem.
Data for the analysiscame from four recordingwater level stationslocatedwithin the
system.The generalwaterlevelpatternwassimilarfor all stations,showinga strongdiurnal
tidal signalsuperimposeduponotherlargerscaleevents. Theselarger scaleeventsincluded
bothfrontalpassage eventsaswell asthefloodingeventsof theAtchafalayaRiver. The fiver
flow explainsabout 86% of the variationin water levelson an annualtime scaleand about
45% of thevariationon monthlytimescales.On timescalesof severaldaystheatmospheric
forcingbecomes important,with thenorth-south
windcomponent explainingabout62% of the
variation in water levels.
The resultingmarshinundationregimeshowsa logisticshapedcurvewith low elevations

(< -25 cmNGVD) beingflooded100%of thetimeandhighelevations (>50 cmNGVD) only
beingfloodeda few hoursa year. The totalhoursthemarshis flooded,at themid-elevations
(~25cmNGVD), showssubstantial variationamongyears,withthedifference betweenlarge
flood years and small flood years being on the order of 3,000 hours. In addition, the
frequency of floodingshowschanges on theorderof 250 eventsper year. Thus,vegetation
growingwithinthissystemexperiences widerangesin bothfloodingfrequencyandduration.

192 E M Swensonand C E Sasser
Theseyearto yeardifferences
aredetermined
by theflowof theAtchafalaya
River.
The hydrologic regimeis themajorabioticfactorinfluencing thevegetationdistribution

withinthissystem.An understanding of therelationship of thisregimeto the vegetation
successionwill providekey dataon whichto basemanagement decisionsregardingthe
implementation of freshwaterandsediment diversionprojects.Presently,suchlargescale
diversion
projectsarebeingproposedfortherestorationof thedeteriorating
LouisianaCoastal
Wetlands.
11.1 Introduction
The Louisianacoastalmarsheswere formed,over the last 8,000 yearsby the successive

formation,abandonment, and subsequentsubsidence and erosionof delta lobesfrom the
MississippiRiver. The moderndelta,whichhasoccupied its presentpositionfor aboutthe
last700 years,consists
of two depositional areas,(1) themaindeltaof theMississippi fiver
and(2) theAtchafalayafiver delta. TheAtchafalaya deltais thecurrentsiteof deposition and
landbuildingin thisprocess,sincethesediments in themaindeltaarebeingdischarged into
thedeeperwatersof thecontinental shelfKolb andVan Lopik(1966);Van Heerden(1983);
Baumannet al. (1984). Since1963,theflow of theAtchafalayaRiver hasbeenstabilizedat
30% of the combinedflow of the Mississippiand Red riversby controlstructures locatedin
Simmesport, Louisiana.Since1983,sediments of theAtchafalayaRiver havebeenfilling
AtchafalayaBay Van HeerdenandRoberts(1980). New islandsfirstemergedin 1973,and
emergentplantsbeganto colonizeas soonas the sedimentshad accretedto intertidal
elevations.Thesemarshesare basicallyfresh,due to the presenceof the AtchafalayaRiver
but alsohavea waterlevel signalthatis stronglyinfluencedby the astronomical tidesas well
asatmosphericforcingAdamsandBaumann(1980). TheAtchafalaya Deltarepresents anarea
wherelandbuildingis occurringin a coastalzonewhichis otherwiseexperiencing landloss
at a rateof about155km:peryearTurner(1990).
Oneof thekeyelements
in thevegetation
dynamics
of coastalmarshes
is theroleplayed
by thewaterlevelregime. Of particularconcernis thedurationandfrequency of inundation.
It hasbeen shown,that the periodicfloodingof a marshsystemmay serveto enhancethe
productivityof theemergentvascularplantsTeal (1962), HopkinsonandHoffman(1984),to
enhancethe flux of nutrientsbetweenthe marshproperandadjacentwaters,andincreasethe
availablefeedinghabitatfor smallestuarine-dependent fishspeciesBoeschandTurner(1984),
Zimmermanand Minello (1984), Minello and Zimmerman(1985). Thesestudieshavebeen
largelyconfinedto saltandbrackishmarshsystems.Fieldstudiesinvestigating thehydrologic
regimeas it relatesto the vegetation,indicatethat the degreeof inundationmay alsobe an
importantparameteraffectingfreshmarshvegetation.Sasser(1977), Mendelssohn andMcKee
(1987). Recentwork on vegetationwithin the Atchafalayaislandsby Johnsonet al. (1985)
indicatedthat the environmentalfactorswhichvary mostconsistently with the vegetationare
thosewhich dependdirectlyon hydrology. Theft work however,did not includea detailed

E M Swenson and C E Sasser 193
analysisof the hydrologiccharacteristics

of the islands. In a modelingstudyof vegetation
dynamicswithin the AtchafalayaDelta Rejmfineket al. (1987) the predictedchangesin
vegetationdistributionwere assumedto be stronglyrelatedto the floodingregimewithin the
islands. An understanding of the hydrologicregimeof this areaand its relationshipto the
vegetationcommunitycan provideimportantmanagementinformationfor use with fiver
diversionprojectsbeing designedfor the large-scalecreationand restorationof Louisiana
coastal wetlands.
In thispaper,we describethe inundationpatternof the Islandsin the AtchafalayaDelta,

with emphasison the frequencyand durationof flooding. The relationship betweenthe
observedpatternsand the majorhydrologicforcingfunctionsis alsodescribed.
11.2 Study area

The AtchafalayaBay system(Figure1) is a shallowestuaryfringedby salt and brackish
marsh.The deltaproperis characterized by a seriesof lobateislands,whichfirst emerged
followingtherecordfloodof 1973. Theislands becamevegetated withintwoyearsandhave
sincebecomeimportantwetlandhabitatfor wildlifeandwaterfowl.In general,theislandsare
characterizedby an elevationgradientfroma highupstream tip, usuallyvegetatedby willow
(Salix nigra) to a low-lying,arrowhead(Sagittariaspp.)marsh. This low-lyingmarshis
characterized by extensivemud-fiatsandis vegetatedon a seasonal basis. The intermediate
elevationsare colonizedby a mixtureof vegetationwhich is alsoseasonalin nature. This
elevationgradient servesto def'mea floodinggradientwithlessfloodingat theupstream tip
of the islandand greaterfloodingon the mud fiats.
11.3 Data and methods
In connection with its mission of Flood Control and Channel Maintenance within the
Atchafalaya
Delta Complex,theU.S. Army Corpsof Engineers
(COE) routinelymonitors
waterlevelsat four locationswithinthissystem(Figure1). The stationsof interestincluded
DeerIsland(DI), in theLowerAtchafalaya River, the AmeradaHessPlatform(AH) in the
deltaislandsproper,and EugeneIsland(EI) an offshorestation. The fourthstation,Belle
Island(BI), is locatedin an oil fieldon a saltdome,andwasnotusedin thisstudy.
The recordsof water stageat thesestationswere obtainexlfrom the COE and were
machine digitizedat onehourintervals.TheycoveredthetimeperiodfromJanuary1979
throughDecember 1986. Thedataweredigitizedusinga "Numonics2400"digitizingtablet
interfaced
withan "IBM AT" computer.Duringdigitizing, notesweretakenregarding any
gaps,anomalousdata or time basedrift on the records. These noteswere usedlater for data
editing.

194 E M Swenson and C E Sasser
MISSISSIPPI
LOUISIANA
TEXAS
29 ø
N GULF OF MEXICO
94ø 92ø
/ 90o
I i I
Figure1. Basemapof theAtchafalaya DeltaAreashowingthelocations

of theCorpsof Engineers
tide
gagesusedin the analysis.DI = Deer Island,AH = AmeradaHessPlatformandEI = EugeneIsland,
BI = Bell Isle.
Dataanalyseswereconducted usingtheLSU Mainfrmnewith"Statistical AnalysisSystem"

SAS InstituteInc. (1985 a,b,c). After all of the digitized data had been transferred,a
preliminaryanalysiswasperformed.Duringthisanalysis,the datafrom all monthsfor any
givenstationwereconcatenated; anycorrections,baseduponthedigitizingnotes,weremade;
anda dateandtime variablewascreatedfor thedatasets. The raw datawerethenplotted,
andanyobviousspikeswereeditedout. In generalthedatawerequitegoodwith few data

pointseditedout, althoughthere were numerousgapsin the records.
In orderto averageout the slighterrorsin digitizing,three-hourmeanswerecomputedfrom

the original one-hourdata set. The three-hourmeanswere usedin all final analyses.
The final analysisof the datafocusedon the determination of floodingstatisticsfor the

deltaislands.Of particularconcernwerefrequency anddurationof floodingasa functionof
elevation,a major controllingfactorfor the vegetation.The frequencyof floodingwas
calculatedby countingthenumberof individualfloodingeventswithina givenelevationclass.
The durationwas calculatedby measuringthe length of each floodingevent. These
parameters werecomputedusing10 cm elevationclassintervals.The 5 cm flooding datawas
theninterpolated from the 10 cm floodingdata. The floodingstatisticswere determinedon
a monthlybasis, thensummarized as the annualvaluespresentedhere. Figure2 presents a
schematic representationof the definitionof floodingfrequencyanddurationas usedin this
paper. In thisexample,thereare threefloodingevents(1, 2 and3). Thesefloodingevents
are then subdividedby 10 cm elevationclasses. This subdivisionyields the following
breakdown;threeeventsgreaterthen0.0, threeeventsgreaterthan 10.0, one eventgreater
then20.0, andoneeventgreaterthen30.0. The durationof eacheventat a givenelevation
(in hours)is indicatedby the letters(A1, A2, A3, B 1, etc).
•E•2
EVENT 1 EVENT 3
A1 A2
-10,
TIME IN HOURS
A1 .) A2., A3. DURATION OF EVENTS ) 0 Cid

B1., B2., B3 - DURATION OF EVENTS ) 10 CM
C - DUI•TION OF EVENTS ) 20 CM
D - DURATK• OF EVENTS )
Figure2: Schematic of a typicalwaterlevelcurveillustrating themeasurementsmade. Eachof the

floodingevents(1, 2, 3) is subdivided into 10 cm elevationclasses.The floodingdurationfor a
particularelevationclassis indicatedby the letters(A1, A2, A3, B 1, etc.).

MAR 82
ß
8o
t
40-
:2 20
-2o ß i ß i ß i ß i ß i ß i ß I ß i ß i ß I ß i
ß
ß i
.
ß i
.
ß
80
MAY 1982
60
ß 40
:s 20
-20 ' i ß i ß i
ß
ß i ß i ß i ß i ß i
.
ß i ß i ß i
.
ß I ß i ß
80
JULY 83
ß 40
z
• 20
øt i
8
ß i
16
ß I
24
ß i
32
ß i
40
ß i
48
ß i
56
ß i
64
ß i
72
ß
80
i ß i
88
ß
96
i ß
104
i ß
112
TIME (HOURS/3)
Figure3: Plotof measured

waterlevels(solidline)fromAmerada
Hessagainst
waterlevelspredicted
fromDeer Islanddata(diamonds),for (topto botto•n)March1982,May 1982 andJuly 1983. The
horizontal
axisis time(hours/3)
andrepresentsthefirsttwoweeksof datafromthemonth.Thevertical
axis is in centimetersrelative to National GeodeticVertical Datum (NGVD).
In conducting
theanalysis,
it wasnecessm'y
to havethewaterleveldatafromthedelta
islandsproper(Amerada
Hess)asopposed to offshore
(EugeneIsland)or fromAtchafalaya

Bay (Deer Island). However,themwerequitea few gapsin the recordfrom AmeradaHess.

In orderto fill in thesegaps,datafrom the Deer Islandgagewasusedto predictthe missing
AmeradaHessdata basedupona relationshipobtainedwhen data was availablefrom both
gages. A total of eight monthsof data were usedin this analysis. A regressionanalysis
betweenAmeradaHess and Deer Island was conductedusing data from five of the eight
monthsto developa predictiveequation. This predictiveequationwas then testedby
predictingthewaterlevelcurvefor theremainingthreemonthsdata. The resultingpredictions
weresubtracted from the actualmeasureddatato obtainthe error(residual)of theprediction.
Figure3 presents examplesof the predictedandmeasured waterlevelsfor comparison.The
results,indicatedthatthe DI three-hourlyvaluescansuccessfullypredicttheAH three-hourly
valueswith meanresidualerrorsof lessthen 10%. However,for the final analysiswe were
primarilyconcerned with thecalculationof floodingfrequencyanddurationasopposedto the
actualpointmeasurements of waterlevels. In orderto furthercheckon the accuracyof the
prediction,the floodingfrequencyanddurationwerecalculatedusingboththe measureddata
andthepredicted data.Theresultsindicated excellentagreementbetweenfloodingfrequencies
and durationscalculatedfrom the measureddata and the predicteddata.
Figure4: Time series(A) of three-hourmeanwater levelsat the AmeradaHess.gagein centimeters

relativeto NationalGeodeticVertical Datum (NGVD) for Jmmarythru April 1983, and(B) time series
of three-hourmem•waterlevelsat the AmeradaHess.gagein centimetersrelativeto NationalGeodetic
Vertical Datum (NGVD) for Januarythru April 1984

11.4 Results and Discussion
Figure
4 presents
threehourmeanwaterleveldatafromAmerada
HessforJanuary
thruApril
for 1983and 1984. Figures5 and6 presentexamples of the daily meanwaterlevelsat
Amerada HessandAtchafalaya Riverflowat Simmesport for 1983and1984. TheAmerada
Hesswaterleveldatashownin Figure3 is typicalof thedatafromall years,whichshowa
strongdiurnaltidalsignalsuperimposed uponotherlargerscaleevents.Theselargerscale
eventsincludeboth frontalpassages that influencesealevel and floodingeventsof the
AtchafalayaRiver. Thistypeof signalis quitecommon to waterlevelsmeasured in coastal
LouisianaByrneetal. (1976),Chuang andSwenson (1981),Swenson andTurner(1987).The
fiverflowdatashows a typicalpatternwitha springflood,followed by a fairlyrapiddropin
flow duringthe summer.
•D A
OtJAN83 O1MAR83 OtMAY83 O1JUL83 015EP83 01DEC83
•o• B
600
n00
z00
zøø11
OtJAN83 01NAR83 OtNAY83 O1JUL83 01SœP83 01DœC83
Figure5: Timeseries(A) of dailymeanwaterlevelsat theAmerada

Hess.gage
in centimeters
relative
to NationalGeodeticVerticalDatum (NGVD), andmeandaily river flow (B) of the Atchafalayariver
at Simmsportin cubicmetersper second.Data is from 1983.
Figure7 presents a plot of meanmonthlywaterlevelsat AH andmeanmonthlyflow of

the Atchafalaya River. It canbe seenthatthe waterlevelsmimicthe fiver flow. Linear
regression betweenthe two variablesindicatethatthefiver flow accounts for about35% of
the variationin waterlevels,at thistime scale. Figure8, presents
theyearlymeanfiver flow
andwaterleveldata. Thereappearsto be a closerelationshipbetweenthe waterlevel and

•5•
•.
•oo
75
50
O.
0IJAN84
0•84 0m•¾84 0•J6•84 0•s•P84 01D•C84
O1JANB4 O1MAR84 0•MA¾84 01OUL84 0•SœP84 0•DœC84
Figure6: Timeseries(A) of dailymeanwaterlevelsat theAmeradaHess.gage

in centimeters
relative
to NationalGeodeticVerticalDatum (NGVD), andmeandaily river flow (B) of the Atchafalayariver
at Simmsportin cubicmetersper second.Data is from 1984.
MONTHLY RIVER FLOW AND WATER LEVELS
20000 50
ß !i ,•I WATER
LEVEL
15000
;I :I •, /! 40
ß /•i , ,;:, :• ..,•
::":
i¾' t. '. 30 •
>,
\. : : •,, , 20 z
10000
", ; ; ":•
x '.. ; ow
10
5000
0
ßi•7;., .•.9•., ßi9•iß! ß•.,• ß! ß;•8•.,.ßi•8•.! .•½•s

-•0
YEAR
Figure7' Time seriesof meanmonthlyAtchafalayaRiver flow (solidline) andmeanmonthlywater

level (dashedline) at AmeradaHess.

YEARLY RIVER FLOW VS WATER LEVELS
12000 ß 35
RIVER
FLOW • WATER
LEVEL 3O
10000 ',, / ',, 25

ß ',, i \
20
8000
15
10
6OOO
4000 .... , .... , .... 0

1970 1975 1980 1985
YEAR
Figure8: Time seriesof yearlymeanAtchafalaya

Riverflow (solidline)andyearlymeanwaterlevel
(dashedline) at AmeradaHess.
WATER LEVEL VS NS WIND STRESS

DECEMBER 1982
7O 1000
6O
5O ,A .o.,'ø•WATER
LEVEL 500
4O
..'.. A v\ ILl
Z
3O
2O
v: WIND STRESS -500
>-
10
ß I ' I ' I - 1000
0 10 20 30 4O
01 DEC82 00:00 HRS.
TIME IN DAYS
Figure9: Time seriesof daily meannorth-south

wind cmnponent(solidline) anddaily meanwaterlevel
(dashedline) at AmeradaHess for December1982.
the river flow. Althoughthere are only 6 data points (1979 - 1984), regressionanalysis
betweenthe two indicatedthat river flow may explainabout86% of the waterlevel variation

at time scalesof a year. At shorterti•ne scales(daily) the fiver flow explainsonly 20% of
the variation. At these shorter time scales, both the diurnal tide and the effect of frontal
passagesbecomeimportant.Figure9 presentsan exampleof waterlevelandnorth-southwind
component for December,1982. Linearregression indicatesthatthewindeventsexplainabout
62% of the water level variation,in this particularinstance.
AMERADA HESS' HOURS FLOODED BY ELEVATION

10000
-- 1979
8000 -' 1980
-- 1981
; 1982
6000
4000
2000
o
-25 25 $0 75
10000
= 1983
8000 ß 1984
ß 1985
= 1986
6000
4000
2000
0
-25 0 25 50 75
ELEVATION(CM NGVD)
Figure10. Yearlyfloodingduration
curvesfor AmeradaHessfor 1979'through
1986. Indicatedarethe
numberof hoursa givenelevationwasfloodedduringthe year.

AMERADA HESS: YEARLY HOURS FLOODED BY ELEVATION
10000
-10 CM
8000
0CM
10 CM
6OOO
20 CM
4000
30 CM
2000 40CM
50 CM
0 ' I ' I ' I ' I '
1978 1980 1982 1984 1986 1988
Figure11: Timeseriesof floodingduration

forvariouselevations
at AmeradaHess. Indicated
arethe
numberof hoursa givenelevationwasfloodedfor the year.
Table 1: Yearly floodingdurationfrom the AtchafalayaDelta at AmeradaHess. Indicatedare the

numberof hoursfloodedfor a givenelevation(in c•n relativeto NGVD).
Elevation(cm) 1979 1980 1981 1982 1983 1984 1985 1986

-25 8760 8703 8568 8722 8760 8724 8760 8725
-20 8760 8637 8454 8690 8760 8692 8745 8705
-15 8760 8495 8253 8601 8750 8653 8716 8630
-10 8742 8345 7941 8481 8730 8578 8659 8471
-5 8676 8032 7467 8216 8656 8436 8536 8196
0 8499 7477 6813 7776 8493 8227 8323 7798
5 8238 6878 6198 7204 8227 7908 7934 7165
10 7835 6146 5424 6470 7818 7423 7396 6303
15 7174 5304 4614 5702 7369 6721 6737 5461
20 6409 4520 3867 4709 6787 5992 5867 4419
25 5453 3537 3042 3612 5998 5097 4934 3426
30 4375 2669 2298 2647 5167 4199 3949 2469
35 3457 1940 1602 1740 4371 3268 2950 1682
40 2605 1298 1029 1212 3446 2393 2056 1044
45 1788 717 618 732 2608 1743 1219 656
50 1103 382 411 441 1773 1128 535 357
55 585 193 246 248 1081 659 311 148
60 252 131 192 124 605 296 214 83
65 116 96 144 68 326 147 167 38
70 88 60 132 42 166 95 120 24
75 73 47 67 29 66 44 95 0
80 47 25 0 17 27 13 69 0
85 45 12 0 7 7 13 62 0
90 15 8 0 4 7 13 55 0
95 15 3 0 4 7 13 51 0
100 12 2 0 4 7 9 51 0
125 5 0 0 0 0 0 26 0
150+ 0 0 0 0 0 0 3 0

2O3
E M Swenson and C E Sasser
AMERADA HESS: YEARLY FLOODING EVENTS

400
; 1979 EVENTS
ß 1980 EVENT8
= 1981 EVENT8
• 300 ß 1982 EVENT8
u. •00
0
ß lOO
z
0
-25 0 25 50 75 100 125 150
-- 1984 EVENT8
300 ß 1985 EVENTS
200
100
0
-25 0 25 50 75 100 125
ELEVATION CM NGVD
Figure
12: Yearlyflooding
frequency
plots
forAmerada
Hessfor1979through
1986.Indicated
arethe
numberof timesa givenelevationwasfloodedduringtheyes.
Theannual
flooding
duration
curves
for theDeltaIslands
(AH) from1979through
1986
arepresented
inFigure10. These
plotspresent
thecumulative
distribution
of hours
flooded
asa functionof elevation.The elevations
arein centimeters
relativeto theNationalGeodetic
VerticalDatum(NGVD). In general,
it canbe seenthatelevations
lessthanabout-25 cm
NGVD areflooded100%of thetime(therem'e8,760hoursin a commonyear),thecurvethen

exhibitsa logisticshapewith very few hoursof floodingaboveelevationsof 50 cm NGVD.

The variationamongyearsis large,with the differencesat mid elevationclasses(around25
cm NGVD) betweenlargeflood years(1979, 1983) and smallerfloodyears(1981) beingon
the orderof 3,000 hours. Thus vegetationgrowingat theselevelswouldexperiencea wide
rangein floodingfrom yearto year. This variationcanbe seenmoreclearlyin Figure11. and
Table 1 whichpresentfloodingdurationas a functionof elevation.
Figure 12 and Table 2 presentthe annualfloodingfrequencydatafor the Delta Islands

(AH) from 1979 through1986. Theseplotspresentthe distribution of floodingeventsas a
functionof elevation. The curvesare bell shapedwith the maximumnumberof events
occurring at valuesof 25 to 50 cm NGVD. As with the floodingduration,the variation
amongyearsis large, with the differencesat mid elevationclasses(around25 cm NGVD)
beingon the orderof 250 eventsper year. Thus vegetationgrowingat theselevelswould
experiencea very wide rangein both floodingfrequencyand duration.
Table 2' Yearly floodingfrequencyfrom the AtchafalayaDelta at AmeradaHess. Indicatedare the

numberof floodingeventsfor a given elevation(in cm relativeto NGVD).
Elevation(cm) 1979 1980 1981 1982 1983 1984 1985 1986

-25 1 1 1 1 1 1 1 1
-20 11 21 56 22 12 17 17 14
-15 13 37 82 36 15 26 26 26
-10 14 54 118 53 18 34 34 37
-5 31 80 172 103 39 49 56 61
0 49 107 226 153 60 65 77 85
5 72 134 268 210 96 93 114 126
10 96 161 309 268 133 121 152 161
15 136 181 315 308 172 171 180 180
20 176 201 321 349 212 170 208 200
25 190 177 280 324 241 169 214 189
30 204 153 240 299 269 147 221 178
35 182 117 187 236 262 113 201 151
40 182 117 187 236 262 113 201 151
45 124 58 97 122 216 80 134 94
50 88 34 61 72 178 80 88 65
55 61 23 39 47 96 56 58 44
60 34 13 18 22 79 31 29 24
65 22 10 9 13 55 22 20 16
70 10 7 2 5 32 11 11 7
75 8 5 1 4 20 8 9 0
80 6 3 0 3 9 4 7 0
85 4 2 0 2 5 3 5 0
90 2 1 0 1 1 1 2 0
95 2 1 0 1 1 1 2 0
100 2 1 0 1 1 1 2 0
125 1 0 0 0 0 0 2 0
150 0 0 0 0 0 0 1 0

11.5 Conclusions
Dueto theroleof thevariousforcingfunctions withinthedeltasystem,

vegetationgrowing
withintheislandsmustcopewitha widerangeof flooding frequency
andduration,depending
upontheelevationat whichit is growing.In addition,themajorhydrologic
forcingchanges
throughoutthe year. This resultsin a patternof emergence,growthand die-backof the
vegetationwhich followsthe yearlywaterlevel pattern.
The emergence of the vegetationin the springis controlledby thedurationof the flood
peakof theAtchafalaya River. OncetheRiverstagehasdecreased to a levellow enoughto
allowthemudflatsto havesomesub-aerial exposure, thevegetation emerges.Duringthe
summermonthswhen the vegetationis activelygrowing,the water level patternsare
controlledby the astronomicaltides. Recentresearch SasserandFuller(1988) indicated
grazingstresswithinthe islandsto be an important bioticcontrolof the vegetation.At
elevations wherethe inundation is high,the resulting floodingfrequency and duration
becomes a majorabioticfactorinfluencing the vegetationgrowthanddistribution.In these
instances,thegrazingstress, coupledwith thephysiological (flooding)stresscanleadto a
reductionin thevegetation cover,particularly for theSagittariaspecies.
Table3: Listing
of themajorhydrologic
forces
controlling
thewaterlevelswithintheAtchafalaya
Delta
Islands.Indicated
arethemajorforcing
functions,
byseason,andthehypothesized vegetation
response
to this forcingfunction.
SEASON MAJORHYDROLOGICFORCE VEGETATIONRESPONSE
WINTER FRONTAL PAS SAGES NON-VEGETATED(DORMANT)
SPRING RIVER INCREASE DROP OF RIVER DETERMINES

THEN DECREASE EMERGENCE OF VEGETATION
SUMMER ASTRONOMICAL TIDES VEGETATION DISTRIBUTION INFLUENCED

BY FREQUENCY AND
LENGTH OF INUNDATION
FALL ASTRONOMICAL TDES ONSET OF STRONG COLD FRONTS

FRONTAL PASSAGES DETERMINES DIE BACK OF
VEGETATION DUE TO SALINITY
INCREASE WHEN SOUTH WINDS
MOVE GULF WATER INLAND
In thefall, southerly
windsassociated
witha majorfrontcanmovemoresalineGulf of
Mexicowaterintothedelta,flooding
theislands.Thus,thevegetation
mustalsocopewith
periodic
saltstress.
Wehavehypothesized thatthissalinity
sixess
mayberesponsible
forthe

large-scaledie backof vegetation

observed duringtheFall andearlyWinter. Thisdie back
occursprimarily in the Sagittariaand seasonalvegetationalcommunities. The major
hydrologic forcesandtheirhypothesized influenceon thevegetationarelistedin Table3. In
addition,tropicalstormsand hurricanes,althoughrare,canserveas importantlandbuilding
forceswithin the deltaRejmfinek,M. et al. (1988).
This researchrepresentsthe first step in describingthe hydrologicregime of the

AtchafalayaDelta islands.We havethusfar limitedour analysisto a statistical description
of thefloodingcharacteristics
of theislands,themajorabioticfactorinfluencing
thevegetation
distributionin an areaof CoastalLouisianawith a largefreshwater and sedimentinput. A
morecompleteunderstanding of therelationship
betweenthehydrologic regimeandvegetation
succession withparticularemphasis onthehypo•es• salinitystressfromfall fxontalactivity
is the topicof proposedfuturestudies.In addition,furtherdetailedanalysisof the existing
water level data usingspectraltechniquesshouldbe conducted.The long-termtime series
waterlevelrecordsdiscussed in thispaperwhencoupledwith theexistingmeteorological data
shouldyield a more completeunderstanding of the forcingmechanisms within the delta
environment.Possiblysomesortof predictivemodelcouldbe developed.Suchinformation
will providea data baseon which to basemanagement decisionregardingfreshwater and
sedimentdiversionprojectscurrentlybeing consideredas a meansof restoringmarshesin
Coastal Louisiana.
Acknowledgements
The authorsthankthe U.S. Army Corpsof Engineers,particularlyMr. RobertCunningham,
for supplyingus with copiesof the raw datarecordsfrom the gagesusedin thisstudy. We
alsoacknowledge the assistanceof Ms JessicaKastlerwho spentcountlesshoursdigitizing
theraw recordswhilekeepingcarefulnoteson theprocess.Her diligencemadetransferring
and editingthe dataa muchsimplertask. We also acknowledge commentsmadeby two
anonymousreviewers.
This study was supportedthrough the National Marine Fisheries Service, Habitat
Conservation
Division,U.S. Departmentof Commerce Additionalsupportcamefrom the
LouisianaBoardof RegentsResearchandDevelopment ProgramandtheLouisianaSeaGrant
CollegeProgram,a part of the NationalSea Grant Collegeprogrammaintainedby the
NationalOceanicand AtmosphericAdministration,
U.S. Departmentof Commerce.
11.6 References
Adams,R. D. and R. H. Baumann. 1980. Land Buildingin CoastalLouisiana:Emergenceof the

Atchafalaya
Bay Delta. Centerfor WetlandResources,
LouisianaStateUniversity,BatonRouge,LA

70803. Sea Grant Publication LSU-SG-80-02.
Baumann,R. H., J. W. Day and C. A. Miller. 1984. MississippiDeltaic wetland survival:

Sedimentation
versuscoastalsubmergence.
Science224: 1093-1095.
Boesch,D. F. and R. E. Turner. 1984. Dependence

of fisheriesspecieson salt marshes:the role of
food andrefuge. Estuaries7(4A):460-468.
Byrne,P., M. Borengasser,
G. Drew, R. Muller, B. L. Smith,Jr., andC. Wax. 1976. BaratariaBasin:
Hydrologicand ClimatologicProcesses.Centerfor WetlandResources, LouisianaStateUniversity,
BatonRouge,Louisiana70803. SeaGrantPublicationLSU-T-76-012. 175 pp.
Chuang,W. S., E. M. Swenson. 1981. SubtidalWater Level Variationsin Lake Pontchartrain,

Louisiana.JournalGeophysical
Research86(C5):4198-4204.
Hopkinson,C. S. , Jr. andF. A. Hoffman. 1984. The estuaryextended-a-recipient-system

studyof
estuarine
outwellingin Georgia.Pages313-330in V. S. Kennedyed.TheEstuaryasa Filter. Academic
Press, Orlando, Florida.
Johnson, W. B., C. E. SasserandJ. G. Gosselink.1985. Successionof Vegetationin an Evolvingriver

delta, AtchafalayaBay, Louisiana. Journalof Ecology73:973-986.
Kolb, C. R. and J. R. Van Lopik. 1966. Depositionalenvironments of the MississippiRiver deltaic

plain - southeasternLouisiana. pp. 17-61 in Shirley,M. L. ed., Deltas, HoustonGeologicSociety,
Houston, Texas
Mendelssohn, I. A. and K. L. McKee. 1987. Experimentalfield and greenhouse

verificationof the
influenceof saltwaterintrusionand submergence
on marshdeterioration: mechanismsof action. ch.
8 in Turner, R. E. and D. R. Cahoon, editors. Causesof Wetland Loss in the Coastal Central Gulf of
Mexico. Volume II: TechnicalNarrative. Final report submittedto MineralsManagementService,
New Orleans,LA. ContractNo. 14-12-0001-30252. OCS Study/MMS 87-0120. 400 pp.
Minello, T. J. and R. J. Zimmerman. 1985. Differential selectionfor vegetativestnlcturebetween

juvenilebrownshrimp(Penaeusaztecus)andwhiteshrimp(P. setiferus),andimplicationsin predator-
prey relationships.Estuarine,Coastaland Shelf Science. 20: 707-716.
Rejmfinek,M., Sasser,C. E. and Gosselink,J. G. 1987. Modeling VegetationDynamicsin the

MississippiRiver deltaicplain. Vegetatio69, 133-140.
Rejm•mek,M. Sasser,C. E. andPeterson,G.W. 1988. Hurricane-Induced

Sedimentdepositionin a
Gulf Coast marsh. Estuarine, Coastal and Shelf Science. 27:217-222.
SAS Institute,Inc. 1985a. User'sGuide: Basics,Version5 Edition. Cary, NC: SasInstituteInc.,

1290 pp.

SAS Institute,Inc. 1985b. User's Guide: Statistics,Version5 Edition. Cary, NC: SasInstituteInc.,
956 pp.
SAS Institute,Inc. 1985c. SAS/GraphUser'sGuide,Version5 Edition. Cary, NC: SasInstituteInc.,

596 pp.
Sasser,C.E. 1977..Distribution
of vegetationin Louisianacoastalmarshesasresponse
to tidalflooding.
M.S. Thesis. LouisianaStateUniversity,BatonRouge,Louisiana,70803. 40 pp.
Sasser,C. E. andD. A. Fuller, eds. 1988. VegetationandWaterfowlUse of Islandsin Atchafalaya

Bay. Final reportsubmittedto LouisianaBoardof Regents,150 RiversideMall, Suite 129, Baton
Rouge,LA. ContractNo. 86-LBR/018-B04.
Swenson,E. M. and R. E. Turner. 1987. Spoil Ba16:s: Effectson a CoastalMarsh Water Level
Regime. Estuarine,Coastaland Shelf Science. 24:599-609.
Teal, J. M. 1962. Energyflow in the Salt MarshEcosystelnof Georgia. Ecology43: 614-624.
Turner,R. E. 1990. LandscapeDevelopmentandCoastalWetlandLossin the NorthernCentralGulf

of Mexico. Amer. Zool., 30:89-105.
Van Heerden,I.L. 1983. Deltaic Sedimentationin easternAtchafalayaBay, Louisiana. Centerfor

WetlandResources,LouisianaStateUniversity,BatonRouge.
Van Heerden,I. L. andH. H. Roberts. 1980. The Atchafalayadelta: rapidprodegradation alonga

traditionallyretreatingcoast(southcentral
Louisiana).Zeitschriftfur Geomorphologie.
38: 188-201.
Zimmerman, R. J. and T. J. Minello. 1984. Densities of Penaeusaztecus,Penaeussetiferusand other

natantmacrofaunain a Texas saltmarsh. Estuaries7(4A):421-433.

12
Modelling of low-frequencysalinity
variations in the St. Lawrence Estuary
K T Tee
Abstract
Low-frequency salinityoscillationswiththeperiodof 40-50 daysin theSt. LawrenceEstuary

were examined using current meter measurements in 1982, and a two-dimensional
(cross-sectionalaveraged)numericalestuarinecirculationmodel. Forcing of the model
includeslow-frequencyvariationsof wind, ice and freshwaterrunoff from Januaryto
September, 1982. Restfitsof themodelsimulationandfieldobservation consistently
showthat
the salinityoscillation(1) is forcedmainlyby the along-channel
wind stresswith maximum
salinitycorresponding to maximumwind forcingin downstream direction,(2) occursearlier
in the middleestuary,and (3) occurs earlierat deeperlocationsin the middle and lower
estuaries.The wind inducedupwellingand downwellingare the majorfactorproducingthe
salinity variations.
12.1 Introduction
The St. LawrenceEstuaryis a long and deepestuary(Fig. 1). Its lengthfrom QuebecCity,
wheresaltintrusionends,to Pointe-des-Monts,
wherethecoastlinesuddenlydiverges,is about
400 kin. An abruptrise in the LaurentianChannelnear Tadoussacin the middle estuary
separatestheupperestuary,with typicaldepthsof 20-50 m, andthelowerestuary,with typical
depthsof 200-300 m. The bottomtopography in the upperestuaryis fairly complexwith a
deepchannelin the northernsectionanda shallowchannelin the southernsection.
Studiesof subtidalvariationsin othercoastalestuarieswere carriedout mainly in shallow

water (generallylessthan 15 m), and for periodsbetween2 and 30 days(Weisberg,1976;
WeisbergandSturges,1976;Elliott, 1978;Elliott et at., 1978;Wang andElliott, 1978;Wang,
1979a,b;WongandGatvine,1984;Vieira, 1986;Smith,1977;Schroeder andWiseman,1985;
Walters,1982;Goodrichet at. 1987). Seasonal variations of sealeveldatawerediscussed by
Wang (1979a) andWalters(1982). The responseof the St. LawrenceEstuaryto atmospheric
andbuoyancyforcingis very differentfrom thoseof the shallowestuaries.This is because

210 K T Tee
the St. LawrenceEstuaryincludesthe deepLaurentianChannelwherefrictionaleffectsare

small and stratificationis strong. The responseis expectedto have a strongbaroclinic
component associated with the verticaldensitygradient.
7?_
ø 71ø 70 ø 69 ø 68 ø 67 ø 66 ø 65 ø
i i i i i
50 ø
50 ø ST. LAWRENCE
ESTUARY
'. . .
0 km 50 MANICOUAG•N•.'."2
ß'
ßRIVEjR.:i•:'
BAIE-COMEAU
' I•TE'DE•.
49 ø 49 ø
MONT-JOLI
RliViOUSKI
ß
ß..
ß
.....
48 ø LAURENTIAN 48 ø
CHANNEL
ß '...". '
BAlE-SAINT-PAUL
ß ..
48*
4.• o
ß... .... ,,,,,,,,
ß i ß.:.'.5 " ..':'.:."
.... .. .-; •. ' 44 o
72 ø 71ø 70 ø 69 ø 72 ø 68 ø 64 ø 60 ø 56 ø 52*
Figure 1: The locationandthe nmnericalgrid systemof the St. LawrenceEstuary.The currentmeter

mooringsat stationsA, B., C., D, E andF are indicatedby soliddots.
For the St. LawrenceEstuary,thereexist two setsof long-term(> 100 days)currentmeter

dam,onenearthe mouthof the estuarycarriedout in 1979 (E1-Sabhet al., 1982;Mertz et al.
1988, 1989);andanother,carriedoutin 1982,involvessimultaneous observations
in theupper,
middle and lower estuaries(Tee, 1989). Both observationsshow subtidaloscillationswith
the periodsof 10-15 daysand80 days(Tee, 1990). The 10-15 daysoscillationis inducedby
meteorological forcing. The forcingof the 80 daysoscillationis unclear.
In addition to these oscillations, the 1982 observationsalso show an oscillation with the
periodof 40-50 days. A simpleanalysisindicatedthatthe oscillationin the middleestuary
occurredearlier than that in the upperm•dlower estuaries.Becauseof shortcurrentmeter
records,an accuratedescription of the oscillationusingspectralanalysiscannotbe cardedout.

K T Tee 211
Oneof theobjectivesof thisstudyis to conFinnthisinitialoccurrence of theoscillationin the

middleestuaryusinga two-dimensional (cross-sectionally averaged)estuarinecirculation
model. The observedverticalvariationof the oscillationwill alsobe examinedand compared
with the model result.
Section12.2 describesthe observedlow-frequency salinityvariationsin the St. Lawrence

Estuary.The formulationand forcingof the numericalmodelare described in section12.3.
Thecomparison betweenobserved andcomputed low-frequency salinityvariationsis examined
in section 12.4.
12.2 The observationof low-frequencysalinityvariations

In 1982, fourteencurrentmeterswere mooredfrom April to Septemberat six stationsalong
the St. LawrenceEstuary:stationsA and B near Pointe-des-Monts in the lower estuary,
stationsC and D near Tadoussacin the middle estuary, and stationsE and F near
Baie-Saint-Paul in theupperestuary(Fig.l). Thereweretwo AandemaRCM-5 currentmeters
at eachstationexceptstationC, wheretherewere four AanderaaRCM-5 currentmeters. At
the stationsin the middle and lower estuaries(A-D) the meterswere mooredat 20 and 30
meters from the surface. Another two meters at station C were moored at 40 and 50 m. In
the upperestuary,the metersweremooredat 14 and 19 m at stationE andat 16 and21 m
at stationF. All the currentmetersfunctionedproperlyexceptthoseat stationF, wherethe
lower currentmeterwas lost and the uppercurrentmetercontaineda shortrecord(62 days)
which is not usedin this study.
The data were collected at 30 minute intervals. The currents were resolved into
along-channel and cross-channel components.The along-channel axis is directedfrom the
head to the mouth of the estuary,and the cross-channel axis from northwestto southeast
shores. To studythe low-frequencyoscillations,the time seriesof salinity,temperatureand
currentcomponents were first low-passedwith a Cartwrightfilter (half power at 35 h) and
subsampled at 6 h intervals. The low-passed datawere thenfitted (leastsquaresmethod)to
10thorderpolynomials.The low-frequencyvariationof the temperature is closelycorrelated
with the salinityvariation,with high (low) temperatureassociated with low (high) salinity.
Becausethe low-frequencyvelocity variationhas strongsignalsof spring-neapvariation
generatedby tidal rectification,and/orseasonalvariationgeneratedprobablyby freshwater
ranoff(Tee, 1989),the40-50 dayssignalin thevelocityvariationcannotbe identifiedclearly.
In thefollowing,onlythe low-frequency oscillationin salinityis discussexlandcomparedwith
the model prediction.
Figure2 showsthe low-frequency components of the along-channel

wind velocityandthe
atmospheric
pressureat Mont Joli (Fig.1), andthe measuredsalinityat currentmeterstations.
An exampleof the 40-50 day oscillationis the salinitymaximaarounddays115, 150, 200 and
250 andminimaarounddays130,180and220 at stationC. We can seefrom the figurethat

212 K T Tee
a. FORCING
lO
Uw
1035
•-o
1015
r-
•995 •
27
C. MIDDLE
33
•'31
d. UPPER
o• 25
23
O0 140 180 220 260
t (julian day)
Figure2: The low-frequency

oscillations
of the atmospheric
forcingandthe salinityfor the upper
currentmetersin 1982. (a) The along-channel wind velocityuw,(solid)andthe atmospheric pressure Pa
(dashed).(b) The salinityat stationsA (solid)and B (dashed)in the lower estuary.(c) The salinityat
stationsC (solid) and D (dashed)in the middle estuary.(d) The salinity at stationE in the upper
estuary.PositiveU,• indicatesthe wind blowingtowardthe mouthof the estuary. The forcingand
propagation of the40- to 50-dayoscillation
is shown.The windvelocityUwin (a) hasa maximumat
days 100 whichcorrectsthe correspond day of 110 shownpreviously(Fig.ll of Tee, 1989). This
correctionis dueto the useof longerwind recordin thepresentstudy(seeFig.5).

K T Tee 213
thereare someapparentcorrespondences betweenthe salinityvariationand the atmospheric

forcing. The correlationbetweenthesalinityandwind velocityis positive:maximumsalinity
corresponds to maximumwind velocity. The salinitymaximaor minimaat stationsC and
D in the middleestuaryoccurearlierthanthoseat stationsA andB in the lower estuaryand
at stationE in the upperestuary.
The verticalvariationof low-frequencysalinitydatain the St. LawrenceEstuarywas not

examinedin the previousstudies(Tee, 1989). Here, it was found that the oscillationsat
stationsA, B, C, and D occurredearlier at deepercurrentmeters,and thoseat stationE
occurredslightly laterat deepermeter.The magnitudes of thevariationdecrease
with depths.
An exampleof the upwardpropagationof the maxima or minima, and the decreasesof the
low-frequencysignalsat stationC is shownin Figure 3.
50m
40m
30m
20m
lOO 140 180 220 260
T(DAYS)
Figure3: Theverticalstructure
(20, 30, 40, 50 m) of theobserved low-frequency salinityvariationat
stationC in themiddleestuary.The upwardpropagation of the salinityvariationis shown.
12.3 A numericalmodel and forcingfunctions

A two-dimensional (cross-sectional
averaged)numericalmodelwasusedby Tee andLim
(1987)to studythefreshwater
pulsein theSt.LawrenceEstuary.The modelis modifiedhere
to includewind forcingand ice formationand melting. For detaileddescriptions
of the

214 K T Tee
numericalmodel, the readeris referredto Tee and Lim (1987). Only thoseaspectsthat are
relateddirectlyto this studyare outlinedhere.
12.3.1 Governingequations
In the Cartesianco-ordinates,
the governingequations
of the model,averagedoverthe width
of the channel, are:
1 a(Bu) + aw = 0 (12.2)
B ax az
--.- (12.3)
wherex is the horizontalco-ordinatemeasuredseawardfrom the headof the estuary;z is the

vertical co-ordinatemeasuredupwardfrom the mean sea level; u and w are the velocity
components in thex andz directions, N andK are the coefficients of verticaleddyviscosity
and diffusivity,B is the widthof theestuary,t is time,s is salinity,Pois the densityof fresh
water, andp is pressure.
The density(p) is relatedto the salinityby:
p = Po(1+ 7.57x10'4s) (12.4)
wheres is expressedin termsof partsper thousand(ppt). The pressureis derivedfrom the

hydrostaticapproximation,
(12.5)
p(z)
- p,+• gpdz
wherep, is the pressureat surface,and g is gravity.
The St. LawrenceEstuaryis subdividedinto 49 sectionswith conslanthorizontalgrid

spacing(Ax) of 9.5 km. There are twelve levels with variablespacingsin the vertical.
Beginningfrom the surface,the verticalgrid spacing(Az) are 10 m each for the first six
spacings, 20 m eachfor the nexttwo spacings,
30 m for the followingspacingand40 m each
for the last threespacings.

K T Tee 215
The verticaleddy coefficients,N and K, are taken to be a functionof the Richardson

number:
(12.6)
and
K- (•/2)a + bO + R) -7n (12.7)
where
- _ g(apl&)l(aul&) (12.8)
is the Richardson number, and a and b are constants. These forms of N and K were chosen
by BowdenandHamilton(1975) for theftstudiesof the estuarinecirculation.The constants
a andb arechosen
to be3x10'4m2s
4, thevaluesfoundto givethebestfit between
computed
andobservedsalinityin FebruaryandMay of 1963in the St. LawrenceEstuary(Tee andLim,
1987).
The bottomboundaryconditionsare:
u• = 0 (zero normal velocity) (12.9)

NOu/•}z= •,u• (linearbottomfriction) (12.10)
Os/Oz= 0 (no saltflux) (12.11)
where u, is the normal velocityat the bottom,ubis the horizontalvelocitynear the bottom
takento be the valueat the deepestlevel, and•, is the linearfrictioncoefficienttakento be
0.001ms4. At thesurface(z=0),theboundary
conditions
are:
w = 0 (rigid-lid condition) (12.12)

pN•}ul•}z
= 'c,(wind stress) (12.13)
pKOs/•}z= -[p,•si- s)Oh/•}t]
(saltflux) (12.14)
where% is the along-channel

component
of the wind stress,h is the ice's thickness,Pi =
0.0009kg m'3is theicedensity,
andsiis theicesalinity
takensimplyto be zero. Thesalt
flux is negativeif the ice is meltingandpositivewhenit is forming. The simpleboundary
conditionsgivenin (12.13) and(12.14) do not takeinto accountcomplexinteractionbetween
ice and water,whichprobablyaffectssignificantly the circulationin the estuaryduringpart
of JanuaryandFebruarywhena largeareaof the estuarywascoveredwith ice. Howeverthe
inclusionof saltflux (12.14) in the modelprobablyproduces the first ordereffectof ice on
the salinityvariationin the estuary.
Althoughthe comparisonof observedand model salinityis carriedout in the ice-free

season,the ice formationand meltingare includedin the modelbecausethe observationin

216 K T Tee
Aprilcanbe affected
by icemelting
in March.Notethattheresponse
of theestuary,
estimated
byusing
theratioof length
scale(500km)tovelocity
scale(0.1ms'l),isabout
two
months.
30-
20-
LU10-
Z
,,,,,
I I I I I
6 26 46 66
T (DAYS)
Figure
4: Theicethickness
in theSt.Lawrence
Estuary
fortheyear1982estimated
fromicecharts
produced
byIceForecasting
Central,
Ottawa,
Canada.
Theestuary
issubdivided
intofivesections
of
equalhorizontal
spacing.
Thesection
number
increases
fromoneneartheheadto fivenearthemouth
oftheestuary.
Solid:
section
1; dotted:
section
2;dash-dotted:
section
3; dashed:
section
4;long
and
short
dashed:
section
5. Thelow-frequency
variation
of theicethickness
usedin themodel
isobtained
byaveraging
thethickness
over
allthesections,
andthen
fitting
(least
squares
method)
theaveraged
data
intoan1l thorderpolynomial.
Theresultisplotted
asa thicksolidcurve.

K T Tee 217
ALONG- CHANNELWIND
lO
1040
C) ATMOSPHEREPRESSURE
1020
lOOO
98O
o •o •o •o •o •o
t (DAYS)
Figure5: The(a) along-channel (u•),(b)cross-channel

(Vw)component of thewindvelocity,
and(c)
theatmospheric pressure
(p•)atMont-Joli.
Solid:low-passeddata;Dashed:low-frequency
dataobtained
by fitting(leastsquares
method)thelow-passed datato a 20thorderpolynomial.

218 K T Tee
12.3.2 Forcingfunctions
The ice thickness andcoveragefor the year 1982 wereestimatedfrom ice chartsprovidedby
Ice Forecasting Central,Ottawa,Canada. The chartwasproducedeverytwo to threedays.
Most areasof the estuaryexceptthosenear the head were ice free after March 12. The
modellingarea is subdividedinto five sectionsof equalhorizontalspacing. The average
thicknessis computedfor each sectionand shownin Fig.4. The sectionnumberincreases
from one near the head to five near the mouth of the estuary. The thicknessat the two
sectionsnearthe headof the estuary(solid-anddotted-curves) are basicallythe sameexcept
for theperiodbetweendays18 and25 andnearday 65 whenthetwo thicknesses are slightly
different. As expected,the ice near the headof the estuarywherethe salinityis low is
generallythickerthan downstream.The exceptionare thosebetweendays47 and 58 when
theice nearthemouthof theestuaryis significantly thicker thanthatinsidetheestuary.The
low-frequencyvariationof the ice thickness(thicksolidcurve)usedfor computingsaltflux
on the surface(F_xt.12.14) is obtainedby averagingthe thicknessover all the sections,and
thenfitting (leastsquaresmethod)the averageddatainto an 1lth orderpolynomial.
Hourly wind andatmospheric pressure(Po)datain 1982 at Mont-JoliandBaie-Comeauin

the lowerestuaryandQuebecCity in the upperestuarywereobtainedfrom theEnvironment
CanadaWeatherOffice. It wasfoundthattheoscillations in theatmospheric
dataat the three
stationswere similarand occurredmoreor lesssimultaneously.This is becausethe weather
systemin the Gulf of St.Lawrencearea generallyhas a large horizontalscale(order of
thousand kilometers)andpropagates rapidly(orderof thousandkilometerper day). In the
following,only the dataat Mont-Joliare appliedin the modelandanalysis.
The hourly wind data were resolvedinto along-channel (uw) and cross-channel (vw)
components. The resolveddatawerelow-passed with a Cartwrightfilter (half powerat 31
hours)andsubsampled at 6 hourintervals.Figure5 showsthe low-passed (solid-curves)
and
low-frequency (dashedcurves)windvelocityandatmospheric pressure.The low frequency
data are obtainedby fitting (least squaresmethod)the low-passeddata to a 20th order
polynomial.As indicatedby thefigure,the low-frequency dataof u• andv• are inversely
correlatedwith thatof Po; maximain u• andv• correspond to minimain Po,or vice versa.
Withintheperiodof observation (betweendays105and250), thealong-channel component
of thewindvelocitywasmuchlargerthanthecross-channel component. In the2-D modelling
carriedout in this study,only the formercomponent is needed.
The freshwater
runoffinto the St. LawrenceEstuaryis contributed
mainlyfrom the St.
LawrenceRiver (78-80%, Tee and Lim, 1987; EI-Sabh, 1988). The contributionsfrom
Saguenay
Riverin themiddleestuary
andtheManicouagan
RiverSystem
in thelowerestuary
are respectivelyabout 10 to 13%, and 9 to 10%. The runoff from St. LawrenceRiver is
characterizedby seasonal peaksin springandfall, whilethatfromSaguenay hasa seriesof
shortpulses(10-20days)between May andJuly(Tee,1989). TherunofffromManicouagan
River system,becauseit is highlyregulatedfor hydroelectric
power,is moreor lessuniform
throughoutthe year. Figure6 showsthe 1982 freshwater ranoffthroughMontreal(solid

K T Tee 219
curve),240 km upstreamof QuebecCity. The dashed-curve in thefigureis thelow-frequency

curve obtained by fitting (leastsquaremethod)the runoff data to a 20th orderpolynomial.
The runoff throughQuebecCity nearthe upstreamend of the model(dashdotted,Fig.6) is
estimatedfrom the meanratio(1960-1972)of monthlydischarges throughQuebecCity to that
throughMontreal. A phaselag of 7 daysis appliedfor the runoff to arrive at QuebecCity
from Montreal (F. Jordan, Bedford Institute of Oceanography,Canada, Private
Communication).In thepresentform of two-dimensional model,only therunoffthroughthe
headof the estuary(QuebecCity) canbe incorporated.
16
14
6 -
0 ½0 100 150 2•0 2•0 3•0

t (DAYS)
Figure6: Freshwater runoffthroughMontreal(about240 km upstreamof QuebecCity) andQuebec

City. Solid:dailyrunoffthroughMontreal;Dashed:low-frequencyrunoffthroughMontrealobtained
by fitting(leastsquares
method)thedailydatato a 20thorderpolynomial;
dash-dotted:
low-frequency
runoffthroughQuebecCity.
12.4Comparison
betweenobserved
andcomputed
salinityvariation
In the numerical
model,a steadystateestuarine
circulationwasfirstobtainedby usingthe
observedwind,ice andrunoffon January7, 1982. The computationwasthencontinuedfor
ninemonthsusingthe longperiodforcingdescribed in theprevioussection.
Figure7 showsthecomparison between computed (dash-dotted

curves)andobserved (solid
and dashedcurves)salinityvariationfor the uppercurrentmeterat stationsA, B, C, D and

220 K T Tee
I I I i
35
b) MIDDLE
c) LOWER
MODEL '-•
3O
T(DAYS)
Figure
7: Thecomparison
between
computed
andobserved
low-frequency
salinity
variation
attheupper
current
meters
at(a)station
E in theupper
estuary,
(b)stations
C andD in themiddleestuary,
and(c)
stations
A andB inthelowerestuary.
Solid:
observed
salinity
atthesouthernstations
(A,C,E);dashed:
observed
salinity
atthenorthern
stations
(B,D);dash-dotted:
computed
salinity
withalltheforcings;
dotted:
computed
salinitywiththewindforcingonly.
E. At station
E in theupper
estuary,
wecanseefromFig.7athattheobserved
long-period
salinity
variation
canbereproducedapproximately
bythenumerical
model.Notethatbex:ause
of thesimplicities
of themodel(e.g.cross-sectional
averaged)
andthecontamination
of the
observed
40-50dayssalinity
variation
by spring-neap
signals
generated
by strong
tidal

K T Tee 221
rectificationprocessin the upperestuary(Tee, 1989), the observedlow-frequencyvariation

of salinity is expectedto be reproducedonly qualitatively. Figure 7a indicatesthat the
observedmagnitudes of the variation,andoccurrence
of two majormaximaarounddays200
and 250, and two majorminimaarounddays 130 and 220 are simulatedreasonablywell by
the model.
30 • , , ,
35 •
I I
I •C,D
30 •I
,
I I I
I
I
25- ',
20-
L
0 0 100 150 200 2 0 300
t (DAYS)
Figure8: The wind forcingandthe modelresultsof 40- to 50-daysalinityoscillationat 20 rn in the

St.LawrenceEstuary.The along-channel windvelocityu•,is shownin theuppercurve.The namesof
thestations
(A, B, C, D andE) areindicated
on theright-hand sideof thesalinitycurves.Theforcing
andpropagation
of the 40- to 50-dayoscillationis shown.

222 K T Tee
At stationsC and D in the middleestuary,the observedand simulatedsalinityvariations

duringthe laterperiodbetweendays190and250 havecomparable magnitudes [- 0.8 ppt for
the observeddata and - 0.6 ppt for the model;Fig.7b]. However,the observedsalinity
variations with maximumat days 150, and minimaat days 130 and 180 duringthe early
periodare underestimated significantlyby the model. This may be becausethe runoffby
Saguenay River nearthestations(Fig.1), whichhassignificant
outflowduringthisearlyperiod
(maximum
valueof 2.8x10
am3s
'• at days140;seeFig.14of Tee,1989),is notincluded
in the
model. The absenceof this low salinitywaterin the modelduringspringresultsin a small
verticalsalinitygradientnearthe surfacein the middleestuary,whichthenproducesa small
salinityvariationfrom wind-inducedverticaldisplacement of halocline.
At stationsA andB in thelowerestuary,thesimulatedlow-frequencysalinityvariationwas

foundto be significantlysmallerthan thosedetectedin the observeddata (Fig.7c). This
disagreement maybe becausethefleshwaterrunofffromthe Manicouagan River systemnear
these stations(Fig.l) is not includedin the numericalmodel. The runoff from this fiver
systemhasbeenfoundto reducesignificantlythesurfacesalinityin the lowerestuaryduring
springand summer(Therriaultand Levassiur,1985).
Althoughsomeof the amplitudeof the salinityvariationin themiddleandlowerestuaries

cannotbe simulatedquantitatively,the time lagsof the salinitymaximaor minimabetween
the stationsshownin Fig.2 are reproduced very well. Usingthe modelsalinityat 20 m at
stationsA, B, C, D andE, the evidenceof time lagsfor the salinityoscillationis shownin
Fig.8. The uppercurveis the low-frequency variationof the along-channelwind. The two
middlecurvesare the salinityvariationfor stationsA andB in the lowerestuary,andC and
D in themiddleestuary.The salinityvariationfor stationE in theupperestuaryis shownin
the lowercurve. By comparingFigs.2and 8, we can seethatthe occurrences of the observed
salinitymaximaandminimawhichare earlierat stationsC and D, andlaterat stationsA, B
and E are simulatedwell by the numericalmodel.
Figure9a showsthe quantitativecomparison of relativetime lagsat 20m betweenthe

stations.Theobservedtimelagof 11.2daysbetweenstations in thelower(A, B) andmiddle
(C, D) estuaries
compareswell with the computedtimelag of 9.2 daysfromthe numerical
model.Similarly,theobservedtimelagof 5.2 daysbetweenthestations in theupper(E) and
middle(C, D) estuaries
agreeswith the computedtimelag of 3.2 daysfrom themodel.
The computedverticalvariationof salinityoscillationat stationsC and D in the middle

estuaryis shownin Fig.10. The numberon the fight-handsideof the curveindicatesthe
depthsfrom the surface. We can see from the figurethat the magnitudes of oscillation
decreasewith depthandthe oscillationoccursearlierat deeperlocations.The sameresults
wereobserved in thewatercolumnof theupperandlowerestuaries.By comparing Figs.3
and10,wecanseethatthenumerical modelalsoreproduces thedecreaseof magnitude toward
deeperwaterandtheupwardpropagation of salinityoscillation
observedat stationC. Figure
9b showsthequantitativecomparison of relativetimelagsat variousdepthsat thisstation.
Betweenthe salinityvariationat 20 and50 m, theobserved timelag of 2.2 dayscompares

K T Tee 223
well with the modeltime lag of 2.7 days. The sameresultcanbe foundfor the salinity
variationat stationsA and B in the lower estuary(Fig.9c),wherethe observedtime lag of
3.5 daysbetweenthe salinityat 20 and30 rn compares well with thecomputed timelag of
5.2 days. At stationE in theupperestuary,thesalinityoscillation
at thelowercurrentmeter
(at 19 m) lagsthoseat theuppercurrentmeter(at 14 m) by 0.8 days. This smalltimelag
disagreeswiththemodel'sresultswhichshowsthatthe salinityvariationin thelowercurrent
meterleadsthatof the uppermeterby 1.8 days(Fig.9d).The causeof the disagreement is
unclear. It may be becausethe observed time lags(0.8 days)is too smallto be estimated
accurately.
2O (a) (b) 0 5 10
E•. 20
LU 40
6o
i
O0 200 36o 4OO

x (km)
T- To (days) T- To (days)
-10 0 lO 0 5 10
, ,
(d)
•
T
F-
2o
(c) •'1o
I
Q_
Q_
LU 40 c3 20
60 3O
Figure9: The comparison of relativetime lags(T - To)betweenobserved(soliddots)andmodel(solid

curves)resultsof the salinityvariationin the St.LawrenceEstuary.(a) The horizontalvariationof (T -
To) at 20 m, whereTo is the time lagsat stationsC andD. (b) The verticalvariationof (T - To) at
stationsC andD in the middleestuary,whereTo is the time lagsat 20m. (c) The verticalvariationof
(T - To) at stationsA and B in the lower estuary,whereTo is the time lags at 20 m. (b) The vertical
variationof (T - To) at stationsE in the upperestuary,whereTo is the time lagsat 14 m.
The modelresultsindicatethat the contribution of ice on the salinityvariationis quite

small,lessthan5% of thetotalsalinityvariationin mostof thearea,exceptneartheheadof

224 K T Tee
the estuary(near QuebecCity) where it contributesto a maximum of about 18%. The

comparisonbetweenobservedand simulatedlow-frequencysalinityvariationis foundto be
affectedinsignificantlyby the ice forcing.
The observed40-50 dayssalinityvariationsare inducedby wind forcing. This is shown

by theclosecorrespondence betweenwind forcingandsalinityvariations(Figs.2 and8). To
examine more accuratelythe importanceof this forcing on the low-frequencysalinity
variations,a computationwascarriedoutby varyingthewind-forcing,butkeepingtheice and
runoffat theirJanuary7 values.An exampleof thecomputation at 14 m at stationE is shown
in Fig.7aby a dottedcurve. By comparingdashdotted(all forcings)anddotted(wind forcing
only) curves,we canseethattheobserved long-period salinityvariationis inducedmainlyby
wind forcing. Similar resultswereobservedfor the otherstationsin the middleand lower
estuaries.
As the wind blowsdownstream (positiveUw),waterparticlesin the upperportionof the

watercolumnare advectedout to the Gulf of St. Lawrence.Becausethe salinityincreases
towardthemouthof theestuary,theadvectionof low salinityfromupstream reducesthelocal
salinityin thesurfacelayer. This horizontaladvectionof salinitycannotexplainthepositive
correlationbetweenwind forcingand salinityin the surfacelayer throughout the estuary
(Figs.5and 11). Horizontaladvection of salinitybelowsurfacelayerin thedeeperportionof
the estuary(includingstationsA, B, C and D) also cannotexplainthe salinityoscillation
because thesalinitybelow20 m in thisregionis moreor lessuniformalongtheestuary(see
Fig.10 of Tee and Lim, 1987).
The observedand simulatedpositivecorrelationbetweenwind forcingand salinityis

inducedmainlyby verticalmovementof halocline.As thetotaltransport in thewatercolumn
for the two-dimensionalmodelis givenby the freshwaterrunoff throughthe headof the
estuary,the wind induceddownstream flow in the upperlayer generatesthe upstreamflow
in thelowerlayerof thewatercolumn(Fig.12). Thistwo-layerestuarine circulationproduces
upwellingin the estuarywhichadvectshaloclinetowardthe surface.The salinityis thus
expectedto correspond with the downstream wind velocity(Figs.2and 8). The upward
movement of haloclineis particularlystrongnearstations C and D in the middleestuary
because thestations areneartheheadof theLaurentian Channelwherethesteepslopeof the
bottomtopography divertstheupstream flowfromdeepto shallow waters resultingin strong
upwelling.Thisstrongrespond of salinityto wind forcingindicatestheearlyformation of
salinitymaximumnearthesestations.More detailedexplanation of the salinityvariations,
includingthehorizontal andverticalpropagation of thesalinitymaximaandrninirna, requires
anextensiveanalysis of thecomplex andnonlinear interactionsbetween velocityandsalinity.
This analysisis the subjectof a futurestudy.

K T Tee 225
34-
32-
;
; ' / • "-15
;
tt
30-
28-
26-
0 I I
100 I •
2 0 I I
300
T (DAYS)
Figure 10: The modelresultsof low-frequencysalinityvariationat stationC and D in the middle

estuary.Thedecrease
of magnitudes withdepthandtheupwardpropagation of thesalinitymaximaand
mini•nais shownin the figure.

226 K T Tee
35
30
25
20
15
10-
5-
\ /
\
0 I I \ I .,,• I I I
0 50 100 150 200 250 3OO
T (days)
Figure11' The modelresultsof salinityvariations

at 5 m on varioussections
in theSt.Lawrence
Estuary.
Thesection
number
measured
fromthehead
oftheestuary
(Fig.
l) isshown
ontheright-hand
side of each curve.

K TTee 227
-4.0
-7.0
-8.0
-8.0
0.0
3O
i i i i I i
0 50 100 150 200 250 300
T (days)
Figure12: Themodelresultsof velocityvariationat stations

C andD. Thedepth(m) fromthesurface
is shownon the right-handsideof eachcurve.
12.5 Conclusion
BetweenAprilandSeptember1982,fourteen
currentmeters
weremooredat sixstations
along
theSt.LawrenceEstuary:
twostations
eachin thelower,middleandupperestuaries
(Fig.l).

228 K T Tee
The experimental datarevealthelow-frequency oscillationwith the periodof 40-50 days.

A simpleanalysis indicatedthattheoscillation
in themiddleestuaryoccurred earlierthanthat
in theupperandlowerestuaries.In theverticalwatercolumn,theoscillation occurred earlier
at deeperlocationsin themiddleandlowerestuaries, andslightlyearlierat shallower
locations
in the upperestuary.
Becauseof the shortcurrentmeter records(about 150 days), the time lags betweenthe
stationscannotbe estimatedobjectivelyusinga spectralmethod.To confirmtheexistence and
propagationof thisoscillationin theSt.LawrenceEstuary,a two-dimensional (cross-sectional
averaged)estuarinecirculationmodelwith the observedlow-frequency forcingof wind, ice
and freshwaterrunoff was developed.The observedsalinityoscillationin the upperestuary
is simulatedwell but that in the lower estuaryis underestimatedsignificantlyby the model.
In themiddleestuary,thesummer-time (days190to 250) salinityoscillationis simulated
well,
but the spring-time(days 100 to 190) oscillationis underestimated by the model. This
underestimationin the middle and lower estuariesis probably becausethe runoff from
SaguenayandManicouagan Rivers(Fig.l) are not includedin the model.
Althoughthe modelunderestimates someobservedmagnitudes of the salinityoscillation,

it reproduces manyimportantfeaturesof the oscillationincludingthe decreaseof magnitude
from surfaceto deeperwater,andtheearlieroccurrence of theoscillationin themiddleestuary
and at deeperlocations. The model also confirmsthat the most importantforcing of the
oscillationis the along-channel
wind stress. The wind-inducedupwellingand downwelling
are the major factor producingthe salinity variations. Resultsof this studyindicatesthe
importanteffectsof wind forcingon the formation,distributionand propagationof salinity
oscillations in coastal estuaries.
12.6 References
Elliott, A. J., 1978: Observationof the meteorologically

inducedcirculationin the PotomacEstuary,
Estua. Coastal Mar. Sci., 6, 285-299.
Elliott, D. P. Wang, andD. W. Pritchard,1978: The circulationnearthe headof Chesapeake

Bay, J.
Mar. Res., 36, 643-655.
E1-Sabh,M. I., H. J. Lie andV. G. Koutitonsky,1982: Variabilityof the nearsurfaceresidualcurrent

in the lower St. LawrenceEstuary, J. Geophys.Res., 87, 9589-9600.
Goodrich, D. M., W. C. Boicourt, P. Hamilton and D. P. Primhard, 1987: Wind-induced destratification

in ChesapeakeBay, J. Phys.Oceanogr.,17, 2232-2240.
Mertz, G., M. I. E1-SabhandV. G. Koutitonsky,1988:Wind-drivenmotionsat themouthof the lower

St. LawrenceEstuary,Atmosphere-Ocean, 26, 509-523.

K T Tee 229
Mertz, G., M. I. El-Sabh and V. G. Koutitonsky,1989: Low frequencyvariabilityin the lower St.
LawrenceEstuary,J. Mar. Res.,47, 285-302.
Schroeder,W. W. andW. J. Wiseman,Jr., 1985: Low-frequencyshelf-estuary exchangeprocesses in

Mobile Bay andotherestuarinesystemson the northernGulf of Mexico, in Proceedingsof the Eight
BiennialInternational
EstuarineResearchConference,editedby D. A. Wolfe, 355-367, Acadexic,San
Diego, California.
Smith, N. P., 1977: Meteorologicaland tidal exchangesbetweenCorpusChristiBay, Texas, and the

NorthwesternGulf of Mexico, Estuar. CoastalMar. Sci., 5, 511-520.
Tee, K. T., 1989: Subtidalsalinity and velocity variationsin the St. LawrenceEstuary,J. Geophys.
Res.,94, 8075-8090.
Tee, K.T., 1990:Meteorologically

andbuoyancyinducedsubtidalsalinityandvelocityvariationsin the
St.LawrenceEstuary,in "Oceanographyof a large-scaleestuarinesystem- the St.Lawrence"
editedby
M. I. E1-SabhandN. Silverberg,CoastalandestuarineStudies,39, 51-70.
Tee, K.T. andT. H. Lim, 1987: The freshwaterpulse- a numericalmodelwith applicationto the St.
LawrenceEstuary,J. Mar. Res.,45, 871-909.
Therriault,J. C. andM. Levasseur,1985:Controlof phytoplankton production

in thelowerSt. Lawrence
Estuary: light and freshwaterrunoff, NaturalisteCan., 112, 77-96.
Vieira, M. E. C., 1986: The meteorologically

drivencirculationin mid-Chesapeake
Bay, J. Mar. Res.,
44, 473-493.
Walters,R. A., 1982:Low-frequency

variationin sealevelandcurrents
in SouthSanFrancisco
Bay, J.
Phys.Oceanogr.,12, 658-668.
Wang,D. P., 1979a:Subtidalsealevelvariationsin the Chesapeake

Bay andrelationsto atmospheric
forcing,J. Phys.Oceanogr.,9, 413-421.
Wang,D. P., 1979b:Wind-drivencirculation

in theChesapeake
Bay,Winter1975,J. Phys.Oceanogr.,
9, 564-572.
Wang,D.P. andA. J. Elliott, 1978: Nontidalvariabilityin the Chesapeake

Bay andPotomacRiver:
Evidencefor non-localforcing,J. Phys.Oceanogr.,8, 225-232.
Weisberg,R. H., 1976:The nontidalflow in theProvidence

Riverof Narragansett
Bay: A Stochastic
approachto estuarinecirculation,J. Phys.,Oceanogr.,6, 721-734.
Weisberg,
R. H. andW. Sturges,
1976:Velocityobservation
in theWestPassage
of Narragansett
Bay:
A partiallymixedestuary,J. Phys.Oceanogr.,6, 345-354.

230 K T Tee
Wong, K. C. and R. W. Gaxvine, 1984: Observationsof wind-inducedsubtidalvariability in the

DelawareEstuary,J. Geophys.Res., 89, 589-597.

13
On the estuarine circulation within the

Kattegat
N Winkel-Steinberg,J O Backhausand T Pohlmann
Abstract
A three-dimensional, prognostic,baroclinic circulationmodel, based upon the primitive

equationsis appliedto the Kattegatand adjacentsearegionsin orderto providean estimate
of the estuarinecirculationunderforcingconditionstypicalfor theregion.The modeldomain
includespartsof the Skagerrakand henceit allows to studythe communicationbetweena
ratherdeepand a very shallowsearegionin the presenceof stratification.The circulationis
predominantly
governedby the influenceof windsand stratification.
Tidal dynamicsplay a
rather marginal role and were thereforenot consideredin the model. The discussionand
evaluationof the modelresultsis carriedout with the intentionto explainthe influenceof the
circulation and the entrainment of nutrient-rich bottom water on the distribution of nutrients.
Thesephysicalprocesses are considered
to be importantmechanisms for the preconditioning
of thefrequentlyoccuringoxygendepletioneventsin theKattegat.The modelresultscompare
ratherwell with the knownhydrographicalfeaturesof the Kattegat.Entrainmentratesderived
from the modelresultsare comparedwith estimatesobtainedfrom an integratedbox-model
for the Kattegat,developedby Stigebrandt(1983).
13.1 lntroduction
The North Sea and the Baltic Sea communicate via a transition-region with complex
topography (fig. 1). It consists
of theSkagerrak,theKattegat,theBelt SeaandtheSoundand
theWesternBaltic(Kiel Bigh0. In theKattegat,thecirculationandtransformation of thewater
massesis cruciallydeterminedby both,local processes (winds,stratification)and by the
influenceof the far-fieldNorth Seaand Baltic,respectively. Thereare weatherconditions,
which,dueto the combinedeffectof windsandair pressure anddueto theshallowness of the
sea,cancausestrongin- andoutflowevents.(Moreextensive descriptionsof thehydrography
and the dynamicsof the transitionregionare givenby Aure and Saetre(1981), Svansson
(1984),LarssonandRodhe(1979),Thorstensson (1978),Knudsen(1899),Dietrichand$chott
(1974).) Theseeventscausedby the meteorological and hydrodynamical far-field have
predominantly synoptic time-scalesof 1-5daysandtheycontributesignificantlyto theextreme
variability,whichis typicalfor theregion.As a resultthecurrentsmayvarybetweenzeroand

232 N Winkel-Steinberg,J O Backhausand T Pohlmann
(locally) up to five knotsand the salinitycan vary within a rangeof fifteen psu (practical
salinityunits).In the transitionregionan increasingabundanceof oxygendepletionsdue to
anthropogenic eutrophication is observedfor thepast,saytin years.Theseeventshavecaused
severeharm to the local fisheriesand to the marineenvironment.They are causedby local
nutrientinputsvia the atmosphereand via run-off from the mainlandand from numerous
islands in the domain and, on the other hand, from remote sourceslocatid in the southern
North Sea and the eastirn Baltic, respectively.For the contributionfrom the west the
circulationand advectionin the North Sea, in particularalong the Danishwest coast(the
Jutlandcurrent),playsan importantrole for the nutrientbalancein theKattegat.The nutrients
carriedwith the Jutlandcurrententerthe Katiegatat depth,via a deepchannelalong the
Swedishwestcoast,due to the strongdensitycontrastto the muchlesssalineBaltic run-off.
If not disturbedby the atmosphere thisrun-offis a geostrophic
buoyantplumewhichfollows
the Swedishwestcoasttowardsnorth.The extinsionof thiscoastalplumewithinandbeyond
the Skagerrakis called the Norwegiancoastalcurrent.The salinityand also the nutrient
balancewithin the Kattegatis hencestronglydeterminedby advection(North Seainflow and
Balticran-off)andby entrainment betweenthein- andoutflowingwatir masses, respectively.
The lattir is dominatedby local atmospheric disturbances.
In earlysummer1988a very excessive toxicandharmfulalgae-bloom, whichappeared in

the pressas the "Killer-Algae"episode,occured.Besidespoliticalreactionsit initiateda
numberof investigations Backhauset. al. (1988), Aksneset. al. (1989). Further,a reportof
an international
groupof experts,calleduponby theInternational Councilfor theExploration
of the Sea (ICES), was editedby Skjoldaland Dundas(1989). It comprises physicaland
bio-geochemical reasonsandintirpretationsfor the dramaticevent.It likewisedocuments the
still considerablelack of knowledgeabout anthropogenicdisturbancesof the marine
environment. Amongstotherquestions theseinvestigationsaddressed therelativeimportance
of local versus remote nutrient sources in the context of environmental accidents and of an
increasedtrendin the eutrophication
of the north-westEuropeanwaters.
Onlya few modellingstudies haveasyetbeencarriedout,for thetransition regionKatiegat

in particular,whichwouldhelpto answersomeof theabovequestions. Thereis a canalmodel
approach by Svansson (1972), thenthewell knownandwidelyaccepted integratedbox-model
by Stigebrandt(1983) and a three-dimensional (grid-box) circulationmodel by
Mueller-Navarra (1983). Only Stigebrandt's modelso far took into accountthe two-layer
densitystructurein conjunction with saltentrainment. His modelis carefullycalibratedand
in goodagreement withobservations.However,dueto thelaterallyintegrated boxes,it cannot
provideinsightinto thespatialvariabilityof thecirculationandof theadvectionof substances
andof saltandnutriententrainment. The modelof Mueller-Navarra is a diagnostic
oneand,
henceit doesnot allow the simulationof the space-timeevolutionof the baroclinicfields.
Furtherit doesnot containa tirm that explicitlydescribes the verticalmixingof salt and
temperature.

233
N Winkel-Steinberg,
J O Backhaus
andT Pohhnann
10'
$9'
N
0R
WlA
Y fiord
58'
Goto River
Fluden
NORTH SEA
57'
Laholm
KATTEGAT B,ght
DENHARK
GRENAA
ZONAL SECTION
•56'
SJAELLAND
GR. BELT
LT. BELT
BALTIC SEA
55'
KIEL BIGt•T
GERMANY
56'
Figure l: Map of the studyarea
This contribution is basedupona prognostic barocliniccirculationmodel,wherevertical

mixingratesaredetermined by a space-timedependent algorithm.It is therefore
anticipated
that the resultsprovideinsightinto the spatialstructure
of the baroclinicfieldsdue to the
effects of both, advectionand entrainment.As a whole this investigationaims at a better

234 N Winkel-Steinberg,J O BackhausandT Pohlmann
understanding of therole of the circulationin regardto thedispersionof nutrientsand,in view

of the oxygendepletions,in particularto the entrainment of nutrient-rich
bottomwatersinto
the euphoticnear-surfacezone within the Kategat. This paper is a shortand still rather
superficialdescriptionof the resultsof a preliminaryinvestigation phase,whichin the future,
in collaboration
with the SwedishMeteorological andHydrologicalInstitute,will be continued
and deepenedwith the availabilityof recentfield data from the SKAGEX '90 experiment.
13.2 A brief summaryof the numericalmodel scheme

The three-dimensional numericalmodelis baseduponthe schemedescribedby Backhaus
(1985) and by Backhausand Hainbucher(1987). Thosetermsin the nonlinearprimitive
equations,whichin theirdiscretenumericalapproximation maycausenumericalinstability,
areintroducedby implicitalgorithms, whicheffectivelyavoidinstability.The schemeas such
is semi-implicit,becausethe remainingtermsare treatedexplicitly.The densityis derived
from the standardnonlinearoceanographic equationof state.Verticaldiffusion(mixing)of
momentum,heatand saltis approximated by a space-timedependent approach proposed by
Kochergin(1987), who derived the mixing principlefrom the equationfor the turbulent
energy.Theprincipleconsiders shearproduction andbuoyancy; theequationis diagnostic
due
to theassumption of a localstationarity
of theturbulentkineticenergy.For furtherdetailssee
Kochergin(citationabove)or Winkel (1989). The verticalmomentum exchangecoefficient,
derivedfromtheKochergin-relation
wassetto varywithintherangeof 0.013x 10'4to 500
X 10'4 m2s'l.
The modelis definedon the Arakawa-Cgrid andit usesa two time-levelscheme.For the
approximationof the advectionof momentum,heat and salt a Lagrangian,"selective
vector-upstream
scheme"was used.This will be describedherein somemore detail,because
it is missingin theabovecitedpapers. The scheme is, dueto theupwindalgorithm,a priori
diffusive.However,comparedto a straightforward,classicalupwindschemeit has the
advantageof a small cross-flowdiffusivity,which is worsewheneverthe flow directionis
alignedwith or closeto thediagonalof thecoordinate system.Theimprovement wasobtained
by a Lagrangianextensionof the straightforwardupwindoperator.The ideaof the selective
vector-upstream schemeis readilyexplainedin a graphicalway by meansof thebelowsketch
(fig. 2). As an example,andalsoto comparethedifferentupwindalgorithms, we consider a
flow fromthesouth-west quadrant,which,in theLagrangian sense,advectsthepropertyP to
the centralgrid-pointC within one time-step(for the sakeof claritythe other,irrelevant
quadrants aroundC havenot beendrawn).The positionof P is uniquelyder'reed by the
advectionvelocity,i.e. the Lagrangianpathduringonetime-step,whichendsat C.
We consider the three cases:

a) component-upwind: it involvesthe grid-points:W, C, S
b) vector-upwind: it involvesthe grid-points:SW, W, C, S
c) selectivevector-upwind: it involvesthe grid-points:SW, C, S

N Winkel-Steinberg,
J O BackhausandT Pohlmann 235
+ w c
+SW + S
Figure2: Locationof grid-points

(denotation
according to compass-rose
indexing;i.e. SW meanssouth-west
fromthecentralpointC).
All threecasesbecomeequivalent,wheneverthe flow is alignedwith a coordinateaxis.It

should,however,be mentioned,thatneithercaseb) nor c) canbe backtracedto the analytical
form of the advectionterm, becauseof the Lagrangianform of the discreteadvection
approximation. This interpretesandconsiders thatthe propertyP (saltor momentumetc.) is
determinedby a spatialinterpolation (whichis truein eachof the casesa) to c)). At the end
of the time-stepthe propertyP arrivesat the locationC and causesa local advectivechange
at this point.The intentionof the versionsb) (seeHaltiner 1977) and c) is to improvethe
qualityof the spatialinterpolationof the propertyP. This is ratherbad for casea) whenthe
flow is diagonal.It is easilyverifiedthat the mostselectivecasec) musthave the smallest
cross-flowdiffusivity.In the aboveexamplethe interpolationfor casec) within the triangle
(SW, C, S) will give the mostprecisespatialestimateof P, becauseit usesthe smallest
distances betweenP andthefixed Euleriangrid-pointswithinthe south-west quadrant(related
to C). Experimentshaveshownthatversionc), whichwasusedin the followingsimulations,
givesthemostrealisticresultsof advectionprocesses. This is mainlydueto the fact, thatonly
in casec) theartificialdiffusionbehavesisotropic.Usinga suitablefine grid,formulationc)
helps to effectivelyavoid the applicationof complicatedand computer-timeconsuming
flux-corrected schemes. For the verticaladvectiona component-upwind approachis adopted
in the model.As we haveappliedan explicitschemefor all advectivetermsthe time-stepis
restrictedby the CFL-stabilitycriterion.
13.3 Adaptionof the model to the transitionregion

The modeldomainas a discreteapproximationof the topography is depictedin figure3. The
horizontalincrements
of the north-oriented
sphericalgrid are 5 and3 minutesin meridional

and zonal direction,respectively.This is approximately5000 m. The dimensionof the

horizontalmatrix is 115 times 61 grid-points.In the vertical, in order to obtain a decent
resolutionof the verticalstratificationin the Kattegat(maximumdepthabout90 m), thereare
eight equidistantlayers with 5 m thicknesseach. Towardsincreasingdepth the vertical
incrementis graduallyincreased, leadingto a totalnumberof 18 modellayersfor the model
domain.Within the deep portions(maximumdepth 700 m) of the Skagerrakthe vertical
resolutionis of the orderof 100 to 250 m. The internalRossby-Radius of deformationfor the
Kattegatis of the orderof five kilometers.Hencethe modelis not an eddy-resolvingone.The
time-stepof the semi-implicitmodelwas set to 600 s. For an explicit schemethe included
deep region of the Skagerrakwould have requireda time-stepof about 50 s. The initial
baroclinicfields(T, S) for an early summersituationwereinterpolated from a coarserNorth
Sea- Balticmodel,developedby BoehlichandBackhaus(1987). The spatialresolutionof that
modelis 12 nauticalmilesandhencethe initial fieldsare only a very crudeapproximation of
the typical densitystructureof the domainof interest.
Similarly sea levels at the open boundarieswere interpolatedfrom the coatset,

encompassing model.Due to theesmarinecharacterof thetransition
regiona considerable
sea
level differenceof about0.40 m existsbetweenthe openBaltic boundaryand the western
boundarywithinthe Skagerrak(fig. 4). The (mean)atmosphericforcingfor theearlysummer
situationwas obtainedfrom a meteorologicalatlas (Backhauset. al. 1985), which was
compiled
froma databasefor theyears1955- 1982.The6-hourly,gridded
data( 150x 103
m) was kindly providedby the NorwegianMeteorologicalInstitute.The mean wind field
appliedwas almosthomogeneous; the directionnearlyeverywhereis from the west and the
rathermoderate
windspeedvarieswithintherangeof 3.5 to 4.5 ms'• (fig. 5).
13.4 The mean Summer circulation
The "Killer-Algae"eventoccuredduringthe monthsMay to June1988. The initial model

simulationwas preparedfor a climatologicalmeanof an early summersituation,in orderto
providean estimateof a meansummerlyestuarinecirculationprior to futuresimulations of
hydrographic and meteorological anomalies,whichwere quitepronounced for the previous
winterandearlysummer1988 (Backhaus et. al. 1988).The modelwasstartedfromthe above
described initial and forcingfieldsand due to the prognosticequations for temperature and
salinitytheinitiallyverycrudebaroclinic fieldsareeventually adjusted to thetopography and
to the circulationby geostrophic adjustmentandserf-advection andmixing.After about40
dayssignificantchangesin the flow fields as well as in the baroclinicfields vanishedand it
wasthenassumed thatthe modelhadsettleddownto an equilibriumstateundertheapplied
forcingwhich was held constant.A secondexperimentwas conductedin which the wind
forcingwasomitted,in orderto quantifythe solelybuoyancydrivencirculation.The model
wasmn for the sameperiod.In thefollowingsomefew resultsof thesetwo experiments are
presentedwith the intentionto provideinsightinto the meancirculationand T, S-structure,
obtained withthemodel.At theendof bothsimulations thebaroclinic
fieldsarenotablymore

N Winkel-Steinberg,J O Backhausand T Pohhnann 237
structuredthanthe crudeinitial fieldsmid in very goodagreementto the meanhydrography

known from observations(citationsin introduction).
0 - 10
,.•
ß
10 - 35
.i:i.., :-:-35 - 50
•- •i•:i:
85 - 150
' I :;.•.
• 250
1so- 250
- SO0
.• •i•soo
•ooo
:•• "/'::'"':'::•:•:!*::•
""":••
..:...!. !
.....
•!i::
•
' '::J:i-. [
I -'•• •
...... • .... I ......... . •

I 10 11 12 13
Figure3: Modeldomain(discretized
topography,
depthsin m)

238 N Winkel-Steinberg,J 0 BackhausandT Pohlmann
Figure4: Seasttrfacetopography(m x 10) lbr •neanstnnmersituation

N Winkel-Steinberg,J O Backhausand T Pohhnann 239
Figure5: Atmosphericforcingfor meansummersituation

windstress
(Pax 10'2)
air pressure
(Pax 102)

240 N Winkel-Steinberg,J 0 Backhausand T Pohlmann
18.0
0 - 20.O
•-•71•18.0
'
- 20.0 20.0 22.0 20.0
.....................................22.0-
- 22
2•.0
0
2•.0 28.0
26.0 28.0
• _-
.-.
-•J•' -'.::.-':!:.::!*•:.::!:i:•!
•;.-';::
28.0 30.0 •:•; 28 0 30 O
• ' 30.032.5
[] 32.5 35.0
.:... ,,-;:,-,:..;ill
. ::
..:.-zT?
-•-'
: .. :
=.. ,.1'.
,.
9 10 11 12 13 ,,,•--_
.• ,•,..
9 1o 11 12
.• • 13
Figure6: Surfacesalinity(psu)for meansummersituation

a) withoutwind forcing
b) forcedby meansummerwind

N Wi]ficel-Steinberg,
J O BackhausandT Pohlmann 241
ß, I < 2 I
• % 2 - 5
% 5 - lO
% 10- 15
% 15 - 20
r-'-':5-,,•--) , •-
•, 20- 2s
,o -
% BO 70
J •r •/ •-.•½ •, IL' •L.
Figure 7: Circulationpatternat differentdepth

a) nearseasurface(0 - 5 m); withoutwind forcing
b) nearseasurface(0 - 5 m); forcedby meansum,nerwind

Figure 7: Circulationpatternat differentdepth

c) at 20 - 25 rn depth;withoutwind tbrcing
d) at 20 - 25 m depth;forcedby mem•summerwind

N Winkel-Steinberg,
J O Backhaus
andT Pohlmann 243
An inspection andcomparison of the computed surfacesalinitiesfrombothexperiments

(fig.6 a, b) revealsthatthebuoyantplumethatemerges intotheKattegatandflowsalongthe
Swedishcoastline is largelyretardedandalmostblockedby the effectof the appliedweak
wind forcing(cf. fig. 5). This also holdsfor the Skagerrak,whichis in agreement with
observations of AureandSaetre(1981),whoreportabouta blockingof thecoastalcurrentfor
westerly winds.Theratherdrasticchanges in themodelresults thatareinducedby theapplied
weakwind forcingsupporttheobserved highsensitivity of theregionto wind events.Along
the Swedishwestcoastthe salinityincreases by aboutten psuon its way from the northern
entranceof the Soundto the northernboundaryof theKattegat(latitudeof Skagen),whichis
a resultof the entrainmentof halinebottomwater in the deep channelalong the Swedish
coast.
The seasurfaceelevationfor the meansummersituation,depictedin figure4 showsthe

strongsealevelgradientbetweenNoah SeaandBaltic,whichis maintained by theestuarine
stratification.The near surfacecirculationfor both experiments(fig. 7 a, b) showsthat also
the flow is sensitiveto the appliedchanges.In particulartwo counter-clock-wise rotating
stationaryeddiesin the transitionSkagerrak/Kattegat, which occurfor the wind induced
blocking situation,supportthe assumptionthat the communication betweenthe two
topographically verydifferentregionsis verydependent onthewindforcing.The northward
(out-)flow throughtheGreatBeltandtheSoundis likewiseblockedby thewindforcing.The
flow regimeat greaterdepth(20 - 25 m) is lesssensitive to thewind disturbance and shows
theexpectedinflow of halinewaterwithinthe channelalongthe Swedishcoast(fig. 7 c, d;
notethat only half the flow informationwasdrawnfor the sakeof clarity).In the northern
entranceof the Kattegatthe stationaryeddy reachesdown to the bottom.
13.5 Model experiments with different boundary conditions

(forcing)
Apartfrom the meanestuarinecirculationduringearlysummerdescribed above,our model
configurationwas used to estimatethe influenceof changesin the forcing (winds and
in/outflow)on the stratificationwithin the Kattegat.Theseexperimentswere conductedby J.
Svenssonand his colleaguesof the SMHI at Gothenburg.Each of their experimentswas
initialisedusingthemeansummersituation.For a zonalsectionwithintheKattegat,just north
of the northernentranceof the Sound(cf. fig. 1), the resultingsalinitycontours(J. Svensson,
pers.communication) are shownfor the followingcases(fig. 8 a-d):
a) Mean summersituationincludingmeanwind (fig. 8 a) This sectionagreeswell with long

term means of the salinity, compiled by Thorstensson(1980) and Szaron (1979). It
complements figure 6 a (surfacesalinity) and servesas a referencemean value for the
followingcases.

b) Shorttermoutflowsituationwith strongnorth-easterly
wind (fig. 8 b) After threedayswith
a fairlystrong
north-easterly
windof 10ms'• andthesealevelattheBalticboundary
risenby
0.50 m only the local effect of the wind tendsto broadenthe coastaloutflow plume. As a
resultthe salinityincreasesat the Swedishcoastin conjunctionwith a local upwardmotion
(cf. the 28 psu contourin figure 8 a).
c) Long termoutflowsituationwith moderatenorth-easterly

wind (fig. 8 c) In contrastto the
caseb) themodelwasrunfor 26 dayswitha moderate
north-easterly
windof 5 ms"andthe
waterlevel in the Balticrisenby 0.50 m likewise.This situationallowsthatthe advectionof
low salinitywatermassesfrom the Balticcausea significanteffectwithintheKattegat.As a
resultratherlow salinityvaluesappearat the Swedishcoastandthe outflowingplumeis not
alignedwith the coastdue to the offshorewind forcing.Due to a re-circulationwithin the
Kattegatalso low salinityvaluesappearat the Danish sideof the section.
d) Inflow situationfor south-westerly

winds (fig. 8 d) The wind forcing was set to a
south-westerly
windof 10 ms'• andthemodelwasrun for a periodof threedays.At the
entrancetowardsthe North Sea the Jutlandcurrentwas increasedby a respectiverise in the
sealevel gradientnormalto the coast.Althoughthe simulationis rathershortan intrusionof
halinebottomwaterappearsat the westernflank of the sectionin the southernKattegat.The
coastalplumeat the Swedishsidehasnarrowedin comparision to the meansummerreference
situation(fig. 8 a).
Theseexperiments demonstrate thatthemodelreactsin a rathersensitiveway to changes

in theforcingconditions.Theyalsohelpto demonstrate whytheKattegatis characterized by
a very high variabilityof the circulationand hydrography.The modelresultsare in good
qualitativeagreementwith featuresobserved duringsimilarmeteorologicalconditions.
13.6 Salt entrainment
Nutrientsare rather non-conservative tracer due to biologicalactivity. Therefore,in the

discussion of the fate of nutrientsoftenthe salinityis takenas a tracer.Sincenutrientsenter
theseavia thefreshwater ran-offtheyarein a goodapproximation inverselycorrelated
to the
salinity.The entrainmentof saltin the Kattegatis hencealsoa measurefor the entrainment
of nutrients,whichare advectedinto the regionby the Jutlandcurrent.The verticaldiffusive
salt-flux(fig. 9) derivedfrom the modeloutputis a goodmeasureof entrainment. It shows
thatindeed,assupported by a watermassanalysis of observations,entrainment occurs along
the entirechannelat the Swedishwestcoast.A new featurethatemergesfrom the resultis
the strongentrainment at the northerncoastof the islandSjaelland,whichis the southern
coastalboundaryof the Kattegat.The saltentrainedthereis thenadvectednorthwardwith the
outflowingwater.In contrastto thebox modelof Stigebrandt (citationin introduction)this
modelprovides insightintothe spatialdistributionof theentrainment. Due to thecomplex
topography thisis far frombeinguniform.An inspection of theentrainment ratesfor theentire

N Winkel-Steinberg,
J O Backhaus
andT Pohlmann 245
Kattegatyieldsa five-timeslowervaluethanthatpublished by Stigebrandt (1983).Themain

reasonfor this discrepancy is that he forcedhis modelwith transientwinds,whereasour
estimatesare derivedfrom the mean (and rathercalm) summersituation.
GRENAA BAY
OFLAHOLM
GRENAA BAYOF LAHOLM
10 26
2o 29 >"
25
30
(m)
GRENAA BAY
OFLAHOLM
GRENAA BAY OF LAHOLM
10 26 26 2
15
15
27 20
20
25
25
30
30 •
(m
(m) Z
z
Figure8: Salinity(psuunits)on a zonalsectionthroughsouthern

Kattegat(for locationcf. fig. 1)
a) referencecase(meansummersituation)
b) short-term
outflow(3 days),north-easterly
wind10 ms'•
c) long-term
outflow(26days),north-easterly
wind5 ms-'
d) short-term
inflow(3 days),south-westerly
wind10 ms'•
Future model experiments,conductedwith transientforcing, shouldallow for a more

consistent
andrealisticcomparison.It is thenexpectedthat the mixing ratesproducedby the
model are much higher. It is interestingto note that especiallythe region with strong
entrainmentat the southerncoastof the Kattegatis knownfor its high abundanceof oxygen
depletions.
There secondary algae-blooms are favouredby the entrainmentof nutrient-rich
bottomwatersinto theeuphoticsurfacezone,which,in earlysummer,oftenhasa low nutrient
contentdue to the regularspringplanktonbloom.
13.7 A tracerexperiment
The import of water massesfrom the North Sea into the Kattegatis demonstrated by a
(passiveand conservative)tracerexperiment,where at the westernboundaryof the model

domaina continuousreleaseof arbitrarymatter(100 %) wasprescribed.

The dispersion
of this
tracerwascalculatedwith an Euleriantransportmodelwhichhasthe samegrid-resolution as
the circulation model. In this model the horizontal and the vertical advection as well as the
verticaldiffusionwas determinedaccordingto the schemeoutlinedin section13.2. The input
data for this simulationwas providedby the circulationmodeldescribedin section13.3.
Figure9: Saltentrainment
ratesat 15 m depth(ms-'x psu)
As known from water mass characteristics the waters of the Jutland current enter the
domainvia thedeepchannelalongtheSwedishwestcoast.Theychangethesignature of the
buoyantsurfaceplumeby entrainment.Thesefeatures
areshownin thefigures10 a-c,which
depicttracerconcentrations
(dissolved
arbitrary
matter)atthreedifferent
depthranges.
Clearly

N Wildeel-Steinberg,
J O Backhausand T Pohhnann 247
the tracerat depth proceedsfarthestsouthand the concentrations closerto the surfaceare

causedby (vertical)entrainlnent.This dispersionexperilnentis in perfectagreementwith the
interpretationof the Kattegatdynamics,as derivedfrom observations.
I !0 !! 12 13
Figure10:Concentrations
of passive
andconservative
tracers
in percent
a) surfacedistributionafter 106 days
b) distribution
in depthrange:10-15m after 1.06days

Figure10:Concentrations
of passiveandconservative
tracersin percent
c) distributionin depthrange:20-25 m after 106 days

N Winkel-Steinberg,
J O Backhausand T Pohlmann 249
13.8 Conclusions and Outlook
Some preliminaryresultson the adaptationand applicationof a barocliniccirculationmodel

on the stratifiedtransitionand communicationregionbetweenthe North Sea and the Baltic
have been presented.In collaborationwith Scandinaviancolleaguesthis work shall be
continuedwith the intentionto supporttheinvestigation
of theverycomplexandconsiderable
rangeof questions andproblemsin conjunction with anthropogeniceutrophicationof coastal
watersin northernEurope.Presentlythe performanceof the model is only confirmedin
qualitativeway. However, only the long term applicationof the model with a realistic,
transientforcingand in conjunction
with suitablysampledfieM datafor its validation,will
helpto gainconfidence in themodelandto supportits reliabilityin regardto its application
in environmentalproblems.
Acknowledgement
The authorswish to thankJ. Svensson
of the SMHI at Gothenburg
for conducting
all model
experimentsdescribedin sectionV.
13.9 References
Aure,J., R. Saetre,1981:Wind effectson the Skagerrak

outflow,in: The NorwegianCoastalCurrent,
Vol. I, ed.: R. Saetre, M. Mork.
Aksnes,D. L., J. Aure,G. K. Furnes,H. R. Skoldal,R. Saetre,1989:Analysisof theChrysochromulina

Polylepis
bloomin theSkagerrak in May 1988,environmental conditions
andpossiblecauses,
Bergen
ScientificCentre,Rep. BSC-89/1.
Backhaus,
J.O., 1985:A three-dimensional
modelfor thesimulation
of shelfseadynamics.
Dt. hydrogr.
Z., 38 H. 4, 165-187.
Backhaus, J. O., D. Hainbucher,1987:A finite differencegeneralcirculationmodelfor shelfseasand

its application
to low frequency
variabilityon the North-European shelf.In: three-dimensional
models
of marine and estaurinedynamics,221-244, ed.: J.C.J.Nihouland B.M. Jamart,Elsevier Science
publishersB.V., Amsterdam,NL.
Backhaus,
J. O., J. Bartsch,D. Quadfasel,
J. Guddal,1985:Atlasof monthlysurfacefieldsof air
pressure,wind stressand wind stresscurl over the North EasternAtlanticOcean:1955-1982,Techn.
Rep. 3-1985,UniversitaetHamburg,Inst.f. Meereskunde.

250 N Winkel-Steinberg,
J O BackhausandT Pohlmann
Backham,J. O., J. Bartsch,P. Damm, D. Hainbucher,T. Pohlmann,D. Quadfasel,G. Wegner,1988:

Circulation
andhydrographicconditions
in theNorthSeaduringwinterandspring1987/88:a physical
oceanographic
studyof theextremealgaebloomof spring1988,Techn.Rep.3-1988(Englishversion),
UniversitaetHamburg,Inst. f. Meereskunde.
Boehlich, M. J., J. O. Backhaus, 1987: Simulation windangeregterStroemungenim System

Nordsee/Ostsee
zurInterpretation
eutrophierungsrelevanter
biologisch-chemischer
Prozesse
in deutschen
Kuestengewaessern,
Forschungsbericht102 04 215/18UmweltbundesamtBerlin.
Dietrich, G., F. Schott,1974: Wasserhaushalt

und Stroemungen, in: Meereskunde
der Ostsee,ed.:
L.Magaard,G. Reinheimer, SpringerVerlag,Berlin,New York,Heidelberg,
3141.
Haltiner,G. J., 1977:Numericalweatherprediction,JohnWiley andSons,Inc.
ICES, 1989:Reporton: The Chrysochromulina

Polylepisbloomin the Skagerrakandthe Kattegatin
May-June1988:environmentalconditions,
possible
causes,
andeffects,ICES,C.M.1989/L:18,ed.:H.R.
Skjoldal,I. Dundas.
Kochergin,V. P., 1987: Three-dimensional

prognosticmodels,in:Three-dimensional
coastalocean
models,ed.:N. S. Heaps,Coastalandestuarine
sciences
4, AmericanGeophysical
Union,Washington,
D.C., 201-208.
Knudsen,M., 1899:De hydrografiske

forholdi de danskefarvandeindenforSkageni 1849-88,Komm.
for Vidensk. Unclers. i den Danske farvande, 2:2.
Larsson,A.M., J. Rodhe,1979: Hydrographical

andchemicalobservationsin the Skagerrak1975-77,
Reportno. 29, Oceanografiska
Intstitutionen,
Goeteborgs
Universitet,Goeteborg.
Mueller-Navarra,S., 1983:Modellergebnisse
zurbaroklinenZirkulationim Kattegat,im Sundundin der
Beltsee,Dt. hydrogr.Z., 36 H. 6.
Stigebrandt,
A., 1983:A modelfor theexchange
of waterandsaltbetweentheBalticandtheSkagerrak,
J. Phys.Oceanogr.,13, 411-427.
Svansson, A., 1972: Canal modelsof the seaand salinityvariationsin the Baltic and adjacentwaters.
Fisheriesboardof Sweden,SeriesHydrographyrep. no. 26.
Svansson,A., 1984: Hydrographicfeaturesof the Kattegat,National Board of Fisheries,Inst. of

HydrographicResearch,Gothenburg,Sweden.
Szaron, J., 1979: Preliminarytransportcomputationof water, salt and nutrientsthrough the

Goeteborg-Fredfikhavnsection(GF) in thenorthernKattegat,baseduponmeasurements,
1975-77,Medd.
Havsfiskelab.,Goeteborg,Sweden.

N Winkel-Steinberg,J O Backhausand T Pohlmann 251
Thorstensson,
B., 1978:Decademeanvaluesof hydrographical
parameters
fromMarstrand
to Skagen
(theP-section),Medd. Havsfiskelab.,
Goeteborg,Sweden.
Winkel,N., 1989:Salz-Entrainment
im Kattegatuntersommerlichen
Bedingungen
- eineModellstudie,
DiplomaThesis,UniversitaetHamburg,Inst.f. Meereskunde.

Part IV
Sedimentation
14
Observations of fine-sediment
concentrationsand transport in the
turbidity maximum region of an estuary
R J Uncles,J A Stephensand M L Barton
Abstract
It is shownthat a mobilestockof fine sedimentand a suspended

particulatematter(SPM)
turbiditymaximumcanbe maintained in the upperreachesof the TamarEstuaryby tidal
pumping mechanisms. Tidalpumping is associatedwithasymmetry in boththefloodandebb,
barotropictidal currentsand in the durations of the highand low-waterslackperiods.The
asymmetry betweenebbandfloodtidalcurrenm leadsto anasymmetry in theratesof erosion
of bedsediments, owingto thedependence of erosionratesonbedshearstress. Thedifference
in durationsof the highandlow-waterslackperiodsleadsto a differencein the amountof
suspended fine sedimentwhichis deposited duringslackcurrents.Thesebarotropic,tidal
processes are modifiedby the existenceof buoyancyeffects.Data from three typesof
experiments in the turbiditymaximumareaof theTamarareusedto illustratetidalprocesses
and their modificationsby buoyancy.
Verticalprofilingdatashowpronounced stratification
duringtheearly ebbfollowinghigh
water,with subsequent reductionof verticalmixing.It is notuntilthestratification hasbeen
erodedsignificantly in the faster currentsandwith the approach of the freshwater-saltwater
interfacethat SPM can be mixed into the water column.Vertical mixing is, as an average
overthe watercolumn,moreintenseduringtheflood. The mixingmaximizesastheinterface
passesthroughthe siteon the floodand SPM is mixedinto the near-surface waterswhere
considerableup-estuary transportcanoccur.Thesedata arecomplemented by a longitudinal
transectexperiment. In thestratifiedregiondown-estuary of thefreshwater-saltwaterinterface,
immediatelyfollowinghigh water,a considerable periodof time is availablefor settlingof
SPM. As the ebb currentsincrease,the stratificationerodesand vertical mixing increases,
leadingto high SPM loadsas the interfacepassesdown-estuary.Much of the SPM is
advecteddown-estuary,where it has little time to settleand depositat low water, and
subsequentlyadvectsbackup-estuary in closeassociation
with the interface.Thesefeatures
are also apparentin data obtainedfrom two longitudinallyseparated instrumentpackages
deployedat 0.25m abovethe bed in the turbiditymaximumregion. SPM concentrations,

256 R J Uncles,J A StephensandM L Barton
longitudinal
velocityand salinitydatarecordedat 5 minuteintervalsover a 10 dayperiod
demonstratethegrowthof theturbiditymaximumovertheneapto spring-fide periodandits
dependenceon bothlocalresuspension, and thereforethelocationof thestockof bed-source
sediment,andthepresence of the freshwater-saltwater
interface.
14.1 Introduction
A suspended particulatematter(SPM, or turbidity)maximumfrequentlyoccursin the low

salinityreachesof parfly-mixedandwell-mixedestuaries.It hasbeenknownfor manyyears
that thisphenomenon can be causedby localresuspension of sedimentsin the upperestuary
(Allen et al., 1980) or by gravitationalcirculationaccumulating
rivefineand coastalSPM
(Festaand Hansen,1978), or by a combinationof thesetwo factors(Officer and Nichols,
1980;Officer, 1981).Numericalsolutionshavebeenderivedfor turbiditymaximumformation
due to both density-drivencirculation(Festa and Hansen, 1978), and to tidally-induced
resuspension (Unclesand Stephens,1989).
Recently,it hasbeenshownquantitatively thatintratidalvariationsin water-column

salinity
stratification,
and thereforestabilityand vertical mixing,may alsoplay an importantrole in
turbiditymaximumformation(Hamblin,1989). It hasalsobeenshownthatstratification due
to resuspended sedimentcan dampturbulenceand verticalmixingnearthe bed,whichmay,
in turn,leadto the formationof a lutocline(Shengand Villaret, 1989;Wolanski et al., 1988;
Kirby andParker,1977) and affectthe behaviourof the turbiditymaximum. The supplyof
fine sedimentis alsoof crucialimportance;Grabemannand Krause(1989) concludedfrom
a studyof the turbiditymaximumin theWeserEstuarythattidally-induced resuspension
from
limitedsources, anddepositionat slackwater,are thedominantintratidalprocesses, while the
density-driven circulationprovideda long-termsupplyof free sedimentto the turbidity
maximumregion.
Thispaperhasthreemainobjectives. First,we demonstrate thattidalprocesses areableto

maintaina turbiditymaximumanda long-termsupplyof fine sediment to theupperreaches
of theTamar,a typicalmacrotidal, partly-mixedestuary(Figure1) andthatthepositionand
magnitude of thebedsedimentsupplyandturbiditymaximumdependon seasonal influences.
Second,we presentvertical-profileand longitudinal-transect data which demonstrate the
intrafidalbehaviour of theturbiditymaximumandtheeffectsof water-column stabilityon its
advectionand growth. Third, we presentdata on near-bedcurrents,salinityand SPM
concentrations at 5 minuteintervalsovera spring-neap cyclefor two longitudinally-separated
locationsin the turbiditymaximumregionof the Tamaranddemonstrate the importance of
both advectionand the modifying influenceof the freshwater-saltwater interfaceon the
resuspension processes andtransportof cohesive,fine sediment.
The possibleimportanceof theresuspension

of bottomsediments
by tidalcurrents
to the
formationof theturbiditymaximumin mesotidal
andmacrotidal
estuaries
hasbeenrecognised

R J Uncles,J A Stephensand M L Barton 257
for sometime (Officer 1981 review paper). The followingtidal processes can be important
(Allen et al., 1980; Uncles et al, 1985 a,b,c; Dronkers,1986): (a) asymmetryin the tidal
currentsin which flood currentsexceedebb currentsand high-waterslack periodsexceed
low-waterslackperiods,(b) suspension of erodedbottomsediments,and (c) the existenceof
an up-estuarymaximumin the tidal currentsand thusin the erosionof sediments.In this
paperwe providesomeevidencefor the importanceof tidal processes and demonstratethe
modificationof these processes by intratidalvariationsin water-columnstabilityand the
presenceof the fleshwater-saltwater interface.
14.2 Tamar sediments
14.2.1 Bed sediments
The Tamar is a muddyestuary;muchof the intertidaldepositedand suspended sediment

consistsof mud (that is, primarilycohesive,silt andclay admixtures,Delo(1988)). The terms,
fine sedimentor mud,are usedin this paperto describetheseadmixtures. Largeareasof
intertidalmudfiatsexistthroughout muchof theTamar'slength.Fine sedimentdoesnotoccur
on thebed within 3 km of the tidal limit at Weir Head (Figure1) underwinterconditionsdue
to scourby the slrongflesh-waterflows (Bale, 1987). Mobile bed sedimentvariesfrom 50%
silt (<63pmgraindiameter)at 3km fromtheweir to 90-95%at 13kmduringwinter. A sandy
regionthenoccursbetween13-16kmwith 30-50% silt. The silt contentincreases to about80%
nearthe mouthof the Tavy (Figure 1) and thenfalls to about60% at the mouthof the Tamar.
Silty, mobilebed sedimentoccurswithin lkm of the weir at Weir Head duringsummerand
the silt contentincreases
from 60-70% at the mouthto 90-95% at the head(Bale, 1987). The
Watercontentof bed sedimentis about70% by weight in the areasof high silt content.
Typicalparticlesizesin the silty bed sedimentare 20-30pmand the organiccontentis
typicallyabout 10%.
14.2.2Suspended
sediments
A turbiditymaximumexistsin theupperreachesof the estuaryandcomprises fine sediment

of around20-30pmmediandiameter,as determined by in-situlaserdiffraction(Westet al.,
1990). These maximum SPM concentrations
occuraroundmaximumflood and ebb current
speeds.Minimumconcentrations occuraroundhighwaterwhena widerangeof SPMparticle
sizesco-exist
butwithlarge,loosely-bound
riocsdominating
(mediandiameters
200-300pm).
It is shownin Unclesand Stephens (1989) that purelytidal processes can generatea

turbiditymaximumin theTamar.Tidalcurrents andelevations arecomputed usingnumerical
solutions of thestandard,
one-dimensional
hydrodynamic equations. Realistictopography is
incorporated. Boundary
conditions
aretakento bethepredicted tidaloscillations
in waterlevel

258 R J Uncles,J A Stephensand M L Barton
at the mouthof theTamarandtheobservedfresh-waterinputsat its head. The space-step

in
the modelis lkm and the variabletime-stepis of the orderof severalminutes.
O-•.We
irHead
5
Tamar
Tavy
lkm
Lynher
3O
Plymout
Sound
Figure1: TheTamarEstuary,showingitssub-division
into51onsegments
and thelocationsof Stations
CK andHQ in the turbiditymaximumregionwhich havebeenusedas deployment sitesfor the
long-terminstrumentpackages.
The one-dimensionalequationof sedimentconsel•,ation

is solvednumerically
on thesame
grid and with the sametime-stepas the hydrodynamicalmodel. All parametervaluesare
givenin Unclesand Stephens (1989). Computedcurrentsare usedto defineadvectionof

R J Uncles,
J A Stephens
andM L Barton 259
sediment
andlocalresuspension
anddeposition.
Localtermsarebased
onformulations
given
by OddandOwen (1972). OddandOwen(1972)assumed thattherateof erosion
is
proportional
totheexcess
shear
stress
(x-xc),
wherexcis thecritical
erosion
stress
andx the
bed shearstress.Our modelusesa quadratic
relationship
betweenbed shearstressand
cross-sectionally
averaged
velocity.Based
onlimitedprofiling
dataandmeasurementsof the
shearstrength
of intertidal
mudflats,whichindicate increasingstrengthfromheadto mouth,
we takethecriticalstress
for erosion
to increasefrom0.14N m'"neartheheadto 1.6N m'"
nearthemouth.An erosion rate"constant"of 3x10'• kg m'"s4 is used.
Fordeposition,OddandOwen(1972),following Krone(1962),assumed thattherateof

deposition
is proportional
to thedeficitshearstress.On thebasisof flumedataOddand
Owen,following Krone(1962),tookthe depositional
stress
tobe0.06N m"-.According to
Delo(1988),thepresentstateof knowledgeof thecriticalbedshearstressfor deposition
indicates
valuesbetween0.06 - 0.10 N m"-. We usea valueof 0.06 N m'". The rateof
deposition
is proportional
to the aggregate
fall velocity,whichin turnis takento be
proportional
to theSPMconcentration
witha constant
of proportionality
of 2x10
'6m s4
(ppm)
4.
Thefresh-water inputSPMconcentration, Pf,is relatedto thefresh-water

flowby a power
law deduced from regressional analysisof measured data (Harriset al., 1984). SPM
concentrations
in coastalwatersneednotbe specified on theebb. It is assumed thatcoastal
waters
lyingfurtheroffshore thana meanspring-tide tidalexcursion haveconcentrationsof
5ppm(consistentwithmeasured values)
andthata lineargradient in SPMconcentrationexists
in the offshore direction.
Figure2(a) showscomputed (dashed lines)andmeasured longitudinal

distributionsof
depth-averaged, high-waterSPM at springtides during 1985. The date on which
measurements weremade, followedbythefresh-waterran-offduringthatday(m3s'l) andthe
tidalrange(m) aregivenin thetoprightof eachdiagram.Withreference to intratidal
data
presentedlater,theSPMconcentrationstendto be lowestat high-water
slack.
Themodelsimulates
a high-waterturbidity
maximum in theupperreacheswhichisroughly
of thecorrectmagnitude
andin thecorrectposition.A comparison of dataon 11/03and
09/04clearlyshowstheeffectof highm-off for a similartidalrange(5.0 and4.9m);the
maximum
is displaced
down-estuary
at highran-off(60asopposed
to 9 m• s4) andhasa
muchsmallermagnitude. Thisbehaviouris reproduced
by themodel.A comparison of data
on 11/03and27/06clearlyshowstheeffectof tidalrangeon themagnitude of theturbidity
maximumfor a similarfresh-water ran-off(9 and8 m• s4);the maximumis locatedin
roughlythesameplacebutis muchreduced forthesmallertidalrange(3.2asopposed to 5.0
m). This behaviouris also reproduced
by the model.
A featureof thehigh-water
datawhichis notreproduced
by themodelis thenarrowness
of the turbiditymaximumpeak,the tidal modelsmearsthe SPM maximumover a tidal
excursion. It is shown later that this narrownessmay be associatedwith the

freshwater-saltwater
interfaceand the effect of water-columnstability down-estuaryof the
interface.
(A) (B)
SPRING TIDE SPM DISTRIBUTION (1985)
22/02
11,4 5
I f• 11/03 O9/04
60,4 9
r• NEAP
TIDE
SPM
\
DISTRIBUTI
(1985)
30/01 28/02 15/03
\\73,20 8,21 9,19
/-\\
/ \,,,
29/04 13/05 09/09
27/O6
8,32 10,2 0 6,2 0 14,1 6
12/12 21/11 20/12

17/09
\•12/11
// •. 26,4 7 12,2 2 18,2 3
13,5
2
,•. 08/10
,,._
DISTANCE FROM HEAD (0 - 30 km)
Figure2: Measuredandcomputed(dashedlines)distributions
of SPM at highwaterof (a): springtides
and(b): neaptidesalongtheestuaryduring1985. Eachverticalaxisrepresents
0-250ppmin (a) and
0-50ppm
in (b). Eachhorizontal
axisrepresents
0-30km.Thedate,run-off(m3s-•) andtidalrange(m)
for eachperiodof observation
are shownin thetoprightof eachdiagram.Reproduced
fromUnclesand
Stephens(1989),with permissionof the AmericanGeophysical Union.
Figure2(b) showscomputed(dashedlines)andmeasured longitudinaldistributions

of SPM
at neap fidesduring 1985. Concentrationsare much smallerthan thoseat springtides. A
comparison of measureddatafor 30/01 and 29/04, whichhavethe sametidal ranges(2.0m)
butverydifferent
valuesfor therun-off(73 asopposed
to 10m3 s-i),demonstrates
themuch
higherfresh-waterSPM concentrations
at high run-off. The pronounced turbiditymaximum
on 30/01 wasprobablydue to resuspension
in the strongebbcurrents;it is reproduced
by the
model, but is muchlesspronounced.
14.2.3 Fine-sedimenttransport
The tidal model and observations establishthat the turbiditymaximumwithin the Tamar
migrateslongitudinallyin responseto changingfiver flows. The sameoccursin the Gironde
Estuary(Allen et al., 1977, 1980);the maximummigratesup-estuaryduringlow fresh-water
discharge,andis locateddown-estuary duringhighdischarge.Thereis anaccompanying shift
in thesedimentshoalingpattern,whichconsists of sediment accumulation
in theupperestuary
in summerand autumn,and in the lowerestuaryin winterand spring. A similarseasonal
patternof sedimentmovementhasbeenobservedin theTamarEstuary(Baleet al., 1985)and
hasbeenmodelledusingpurely tidal processes (Uncleset al., 1988).

The sedimentsmodel was run for 12 years of simulatedtime. During the simulation,
environmentalforcingfunctionsof tideandfresh-waterrun-offrepeatedthemselvesthreetimes
with a four-yearperiodicity. This was done to ensurethat quasi-periodic
solutionsfor the
sediment-mass distributionscouldbe achieved. ObservedTamar forcing functionsfor the
period1982-1985wereused.Thedistribution of sediment massin theupper20 km of estuary
was essentiallythe samefor the secondand third four-yearperiods.
The estimatedflux of sedimentacrossthe headof the estuaryfor the period 1982-1985is

shownin Figure3(a). Thesedatahavebeensubjectedto a runningtwo-weekaverageto
facilitatepresentation.
Theestimated sediment inputsto theestuarycorrelatewithandamplify
the seasonalvariationsin flesh-waterrun-off. Inputsare very small duringlow run-off
summerperiodsandlargeduringhighrun-offwinterperiods.Thisis emphasized in Figure
3(b), whichshowsthecumulativefluvialinputof sediment to theestuaryduring1982-1985.
The majorchangesin sedimentloadoccurduring2-4 monthwinterperiods.
Figure3(c) showsthecalculated massof f'mesediment deposited

on thebedof theestuary
(kg per metreof esmarinelength)during 1982-1985at 2 km and 8 km from the head. The
datum of zero sediment mass at a location defines the situation in which all of the sediment
which is able to be erodedand transported, under the conditionsexperiencedduring
1982-1985,hasbeenmoved from the location. Figure3(d) providesthe sameinformation
for 2 km and 18km fromthehead. Dataarerunning-averages overa spring-neap(14.7 day)
periodto facilitatepresentation.
A pronounced
seasonalcyclein sediment
massoccurswithin
theupper20 km of ourmodelestuary,as observed in theTamarby Baleet al. (1985).
Fine sedimentaccumulates on thebedof the estuaryat 2 km from the head duringlow

run-off summermonths. Sedimentis transported down-estuary during highrun-off winter
months,sothatthisregionis subsequently denudedof sediment.The sedimentaccumulation
closeto the headis very sensitiveto fresh-waterflows. Increasingsummeraccumulation
from 1982 to 1984 (Figure 3(c)) is a consequence of longersummerperiodsof low
fresh-waterrun-off. The smallaccumulationduringthesummerof 1985is a consequence of
higher thanusualrun-offduringa summerperiod.
The accumulationof sediment at 2 km fromtheheadoccursduringtimeswhenthefluvial

sedimentinputis small(Figures3(a,b)). The calculatedmobilesedimentaccumulated
at 2 km
fromtheheadduring
thelowrun-offsummer
of 1984is2.2x 10akgm'•. If thefluvialinput
of sediment
duringthisperiodweredeposited
in theupper2 km of estuary,it wouldamount
to 102kg m'•. Therefore,
in agreement
withtheobservations
of Baleet al. (1985),the
accumulated sedimentmassmustbe derivedfrom up-estuary transportof materialand not
fromdeposition of fluvial-derived sediment.The removalof sediment duringhighrun-off
winter monthsmustbe due to erosionin the fast ebb currents.When ran-off is low, the tidal
flood currentsexceedthe ebb currents,and sedimentis erodedfrom down-estuary and
subsequentlytransported intotheupperreaches.Directobservations of theseprocesses are
given in Uncleset al. (1985 a, b, c).

262 R J'Uncles,J A Stephensand M L Barton
The patternof sedimentaccumulation shownin Figure 3(c) at 2 km from the headis

repeateddownto about6 km. The summeraccumulation of sedimentmaximizesat 4 km.
This patternbecomesless sensitiveto run-off further from the head. Accumulationof
sedimentbeginsearlierin the summer,and removalis delayeduntil later in the autumnand
winter. Daily fluctuations in run-off have little influenceon the accumulatedmassof
sediment.
(B)
(c)
2 kin, 8 km (- -) (D) 2 kin,18km(- -)
2
o
82 83 84 85
TIME IN YEARS (1982-1985)
Figure3: Esti]nated
fluvialsedimentinputsandsimulated
massacculnulationonthebedof theestuary
asfunctionsof timeduring1982-1985.(a): Fluvialsedi,nent
illpUtS(107kg month-i);
(b):Cumulative
fluvialsedimentinputs
(10skg);(c):Massof mobilebedsediment perunitlength
of estuary(103kgm-1)
at 2 kin (full line) mid8 kan(dashedline - plottedat twiceactualmagnitude
for clarity)fromhead;(d):
Massof mobilebedsediment
per unitlengthof estuary
(10• kg In-1)at 2 km (full line)and18 km
(dashedline - plottedat twiceactual•nagnitude)
fromhead.Reproduced
frmnUncleset a1.(1988),with
permissionof the EsmarineResearchFederation.
The seasonalpatternof sedimentaccumulationchanges furtherdown-estuary.At 8 km

from the head,sedimentis accumulated andremovedtwicea year(Figure3(c)). Sediment
is tr,'msported
down-estuary duringhighest
run-offuntiltheupperregionis depleted.Sediment

is tidally-pumped
up-estuary
earlyin theyear,whentherun-offis sufficiently
low thattidal
pumpingcandominate.Whenthe run-offis very low duringsummer,sedimentis tidally
pumpedup-estuaryfasterthan it is being suppliedfrom down-estuary, and sedimentis
depleted.Duringthe onsetof highrun-off,sedimentis transporteddown-estuaryfrom the
upperreaches.Someof thissedimentaccumulates at 8 km untiltherun-offachievesvery
highvaluesandthe sedimentis thentransported
furtherdown-estuary.
The seasonalpatternof sedimentaccumulation is againdifferentat 18 km from the head

(Figure3(d)). Sedimentaccumulation is stronglyandpositivelycorrelatedwith run-off and
fluvial sedimentinflux (Figure 3(a)). The calculatedmobile sedimentmassaccumulatedat
18kmduringthewinterof 1983/84is 103kg m'• (Figure3(d)). If thefluvialsediment
influx
duringthisperiod(Figure3(b)) were deposited
over the upper20 km of estuary,it would
supply103kg m'•. Therefore,the accumulated
sedimentcouldbe derivedfrom the
down-estuarytransportand depositionof fluvial inputs,any additionalmassbeing due to
transport
andaccumulation of mobileestuarinesediment.Thisestuarinesedimentis rapidly
accumulatedfrom up-estuaryduringthe onsetof high run-off, and is rapidly removedby
erosionand up-estuarytransportwhen the rate of supplyof mobile estuarinesediment
decreases.Very litfie up-estuarytransport
occursbelow20 km fromthehead(mosttransport
is down-estuary),so that only a singleperiodof sediment accumulationoccursduringthe
year.
14.3 Intratidalbehaviourof turbiditymaximum
14.3.1 Vertical profiling data
The intratidalbehaviourof salinity,velocity and SPM transporthas been observedon a

numberof occasionsin the upper reachesof the Tamar using direct-readingprofiling
equipmentfrom anchoredvessels.We presentone setof datafrom a stationcloseto CK in
Figure 1. The measurements were madeduringa springtidal cycleon 7 July 1982 with a
tidalrangeof 3.9m(smallsprings)
anda lowfresh-water
ran-offof 3.5 m3s'•.
Figure4(a) showscontourplotsof salinityon a depth-timediagram. The verticalscaleis

heightabovebed and a line depictingthe watersurfaceis drawnto representtotal depthof
water.The freshwater-saltwaterinterface(somewhatarbitrarybetween0 and 1 ppt but taken
in thisdiagramas 0.1 ppt) passesthroughthe stationon theebbat roughly11.5 hoursandon
the flood at roughly16.5 hours. The salinitystratification
on the ebb betweenhigh water
slackandpeakebb currentsat mid-wateris muchhigherthanduringthe flood and implies
inhibitedverticalmixingof saltand SPM duringthe ebb.
Figure 4(c) showscontourplots of SPM concentrationon the depth-timediagram.

Concentrations
arelow at highwaterand,whencompared
with thesalinitycontours,
showan

increaseas the ebb progresses and salinity stratificationerodes. A turbiditymaximumis

reachedin the freshwater roughly0.5-1 hour after the passageof the freshwater-saltwater
interface. Concentrationsdrop to 100-150ppmaroundlow water and then increaseas the
interfaceapproaches on theflood. The turbiditymaximumis reachedas
the freshwater-saltwater
interlacepassesthe station.The flood turbiditymaximumis mixed
more effectivelythroughthe watercolumnthanthe ebb maximum.
(B) ST1S, VELOCITY (cm/s, Ebb -ve)

(A) ST1S,SALINITY(ppt)
6 6
0
o i i !
7.5 9.5 11.5 131.5 15.5 17.5 19.5 7.5 9.5 11.5 13.5 15.5 17.5 19.5
TIME (hours) TIME (hours)
(C) STIS, SPM(ppm) (D) ST1S,SPM(ppm)

6-
• 4
0 i ! i
7.• 915 11.5 13.5 15.5 17!5 19.5 7.5 9.5 11.5 13.5 15.5 17.5 19.5
TIME (hours) TIME (hours)
Figure4: Contour plotsof salinity,velocity(ebbpositivehere)andSPMconcentrations ona height-time

diagramovera springtidalcycleat StationST1SnearStationCK (located in Figure1) on7 July1982.
(a): Salinitycontours
(0.1, 1, 2..... 12 ppt);(b): Velocitycontom's
(0, - 20..... -100 and0, 20.... 60 cm
s'l)ß(c): SPMconcentrations
(50, 100,150..... 600ppm);(d):Sameas(c).
The velocitycontoursam shownin Figure4(b) and the SPM concentration

contoursare
reproduced
againin Figure4(d). Ebbcurrents
havemaximum surface
speeds
of about0.6 m
s'• andnear-bed
shearmaximizes
at approximately
11.5hours.The SPMconcentrations
rapidlyincrease at thistimeandmaximizeroughly0.5-1 hourlaterwhenthenear-bed shear
is stillhigh butthedepth-averaged currents
havedecreasedsomewhat(Figure4(b)). Flood
currentshave maximumspeedsof about1 m s'•. Maximumnear-bedshearoccursat
approximately the sametime as the maximumsurfaceand depth-averagedcurrents(16.5
hours)whichalsocorresponds to thepassageof theturbiditymaximum
(Figure4(d)). The
flood turbiditymaximumis mixed into the near-surfacewaterswhere the SPM is carded
up-estuaryby the strongcurrents.

The differencein the behaviourof the turbiditymaximumbetweenebb andfloodportions

of the tide is clearerin Figure 5(a) where depth-averaged currentspeed,salinity and SPM
concentrations are plottedagainsttime. SPM concentrations lag depth-meancurrentson the
ebb and coincideon the flood. The estimatedverticaleddy diffusivity,I•, at 0.25m above
the bed and the depth-averaged SPM concentrations are plottedagainsttime in Figure 5(b).
The estimationof IC•at 0.25m assumes the validityof the mixing-lengthapproximation with
Munk-Andersondamping(Munk and Anderson,1948) and that an approximatelylinear
increasein I• occurswith heightabove,andcloseto, the bed;verticaldensitygradientsdue
to SPM are ignored.MaximumSPM loadsappearto occuraftermaximumbed-shearstresses
are reachedand correspondto maximum vertical mixing near the bed when near-bed
suspensions are mixed into the flow.
1500 ,
(A):
!
SPM(-), !
U(-*-), S(- -)
! ! !
•ooo-
500 -
0
6 20
60 !
(B):
!
0.1*SPM(-),
!
Kz(-*-)
! ! i
40-
20-
o
6 2o
TIME DURING DAY (HOURS)
Figure5: (a):Depth-averaged
SPM (-), speed(U, -*-) andsalinity(S, - -) and(b):Depth-averaged
SPM
(-) andthe estimatedverticaldiffusivity(K,, -*-) at 0.25m abovethe bed as functionsof time for the
contoureddata shownin Figure4.
14.3.2 Longitudinaltransectdata
Figures6(a,b)showlongitudinal tmnsectdataof verticalprofilesfor SPM concentrations

and
salinityduringa springtide on 8 July 1988in the turbiditymaximumregionof the Tamar.

The salinityis depictedby dashedcontours.The timesof beginningandendfor eachtransect

areshowntogetherwith the stateof thetiderelativeto highor low waterandthelongitudinal
currentat 0.25m abovethe bed near the down-estuary boundaryof the transect.
(A) .......... (B)

300
LW*2 +0 4m/s
!
lO'O
,, ,
7o /
250
HW-1.5 +015m/s
/
6 7 8 9 lO 11 12 13
HW-1 +0 lm/s DISTANCE FROM WEIR (kin)
300/ 2•0•
9 10 11 12 13
OISTANCE FROM WEIR Ikm)
Figure6: Longitudinal transect

databetween StationsCK andHQ locatedin Figure1 showing SPM
(-) andsalinity(- -) contours
duringa springtideon 8 July1988. Thelongitudinal
velocityat 0.25m
abovethe bedat 12.5kinis shownol• the top fight of eachplot andthe startandendtimesfor each
transect
on the left andrightof eachplot. (a): Datafor low water(LW) plus1 houruntilhighwater
(HW) minus1 hour;(b): Datafor highwater(HW) plus1 houruntillow water(LW) minus2 hours.
The salinitydatashowthefreshwater-saltwater interfaceadvectingintotheregionon the

floodwithstrongverticalmixing. At highwatertheup-estuary freshwateris ebbing,asis
thesurfacewaterin thedown-estuary, stratifiedregion. The near-bedwaterin thestratified
region'stalls', partly becauseof the opposinglongitudinal densitygradient,and the
stratification
becomes intenseasfresherwaterebbsin thesurlhce layers.Thenear-bed water
in the stratifiedregion beginsto ebb roughlytwo hoursbefore low water and the
interfacemovesdowntheestuary.
At thestartof thefloodtheSPMconcentration increases

downestuary,sothattheturbidity
maximum is,by implication,
associatedwiththefreshwater-saltwater
interface
region.Asthe
interfaceadvects intotheregionon thefloodthereis considerable
resuspensionof near-bed

mud at the interfacein the strongcurrents. Someof this SPM moveswith the interfacebut
much of it settlesand depositsas high-waterslack approaches.Some SPM remainsand
reachesthe limit of intrusionof the interface,where it forms a slowly erodingturbidity
maximumas particlessetfieand deposit. At the beginningof the ebb in the freshwater the
lower SPM concentrations advect down-estuary.In the stratifiedregionthe SPM ebbsin the
surfacelayers. Depositionoccursin the stagnant,near-bedwater. As the near-bedcurrents
increaseon the ebb and the interfacemoves down-estuary,strong vertical mixing and
resuspension occurin the unstratifiedwaterbehindthe interfaceandthe associated turbidity
maximumadvectsdown-estuary.
14.4 Long-termmonitoringof SPM concentrations
14.4.1 Instrumentation
A near-bed,instrumentpackagehasbeendevelopedto monitorlong-termbehaviourof the

SPM concentrations.
Measurementsof SPM concentration
aremadeusingPartechElectronics
(UK) light-transmittance
turbiditymeters(0-102and 0-104ppm).Theseinstruments
are
calibratedusing gravimetricdeterminations of suspendedsedimentconcentrations in the
measurement area. Flume-calibrated,CoinbrookInstrument(UK) E/M currentmetersare used
to measurehorizontalvelocity. ValeportMarine (UK) conductivity/temperature sensorsare
usedto determinetemperatureand salinity. Thesemeasurements are madeat 0.25m above
the bed and the pressuresensoris locatedin the top of the frameat 0.65m abovethe bed.
Batteries,electronicsboardsanddataloggerarehousedat thetopof theframeandhorizontal
stabilisinglegs are fitted to its base.Data are recordedevery 5 minutes. Details of the
instruments are givenin Uncleset al. (1989),althoughtheframeshavesincebeenredesigned
to accommodate the shallowwatersof the upperestuary.
14.4.2 Deployment
Twoof theseinstrumentpackages havebeendeployed in theturbiditymaximumregionof the

TamarEstuaryat StationsCK andHQ in Figure1 for 10 daysat a timefor severaloccasions
during1988and 1989. Stations CK andHQ are7 and12.5kmfxomthehead,respectively.
The datadiscussedherewereobtainedduringNovember1988 at a time of low fresh-water
flowsbutfollowinga periodof spates
duringOctober.We therefore expectfromFigure3(c)
thatmuchof the mobilemudhasmovedfromtheupperreaches, withina few kilometresof
thehead,andhasbeendeposited
between
HQ andCK andforsomeshortdistance
up-estuary
of CK.

14.4.3 Spring-neapbehaviourof SPM concentrations
SPM concentrations at 0.25m abovethebedfor StationCK are shownin Figure7(a), during

a 10-dayperiodin November1988. The dataat HQ arevery similar,althoughconcentrations
are somewhatlower,andare thereforenot presented.The near-bedSPM levelsfall to low
valuesduringthelow neaptide on day4 andsubsequently increaseon therisingspringfides.
Eachtidal cyclecomprises two mainpeaksfor bothstations. The SPM dataexhibit striking
similarityfor all tidalcycles.The increasing SPM concentrations with increasingtidalrange
demonstrate thatresuspension in thetidal currentsprovidesthesourceof water-column SPM
for theturbiditymaximum.This SPM is, in turn,derivedfromthefine sexliment depositson
the bed of the estuary.
Thefluxesof SPM(F, ppmm s'•) at 0.25mabovethebedfor StationCK areshownin

Figure7(b). The corresponding
dataat HQ are very similarandare thereforenotpresented.
The tidally-averagexl
(25 hourrunningmean)fluxes,<F>, are drawnas the bold line and
demonstratethat the residual fluxes are much smaller than the instantaneousvalues, so that
one wouldexpectthe net amountof mudtransported duringa singletide to be very small.
However, the magnitudeand directionof transport,basedon just one point in each
cross-section,
haslittle significanceto the estuarinemud budget.
Numerical solutions have been derived for turbidity maximum formation due to
density-driven circulation(Festaand Hansen,1978), and this circulationpatternmay be
expectedto contributeto SPM transportin the presenceof fidally-inducedresuspension.
Figures7(c,d) showtheresidual,longitudinal currentat 0.25mabovethebedfor StationsCK
and HQ, respectively. An up-estuary flow occursat HQ duringthe neapfidesand smaller
springtidesbutreverses duringthelargerspringfides(Figure7(d)) whenSPM concentrations
arelargest(Figure7(a)). At CK thenear-bed residualflow is directeddown-estuary
(Figure
7(c)), so that density-drivencirculationsalonecannotbe invokedas a causativemechanism
for turbiditymaximumformationin the upper estuary.
14.4.4 Intratidal behaviourof water depthsand velocities
We will focuson the singletidal cycle during9.0-9.5 daysat the end of the record,when
SPM levelsandfluxesarehighest(Figures7(a,b)). Thebehaviour of theturbiditymaximum
is similarduringthepreceding
springtides. FloodandebbSPM peaksalsooccurat neap
tides,butthesehavemuchlowerSPM concentrations dueto thegreatlyreducedinfluenceof
resuspensionprocesses.
Althoughthe 25 houraveraginghastheeffectof shortening the time series,as shownin

Figures7(c,d),a 12.5houraverageof thevelocitydatafor the lasttidalcycleconfirmsthe
trendin Figures7(c,d)whichshowsthatnear-bed
residualcurrents aredirected down-estuary
whenSPM levelsare largest.

R J Uncles,J A Stephens
andM L Barton 269
4OOO
(A): SPM, (CK) 4000
(B): F(-), <F>, (CK)
•3000
2000-
•ooo
0
• 1000
0 -2000
0 5 10 0 5 10
0.1 (C): ,(CK) 0.! (D): , (HQ)
....
-0.2 -0.2
0 5 10 0 5 10
TIME (DAYS) TIME (DAYS)
Figure7: Time seriesdataat 0.25m abovethebedfor StationsCK andHQ locatedin Figure1; times
0.0 to 1.0 daysrepresent
November14, 1988. (a): SPM concentrations at StationCK; (b): SPM fluxes
at StationCK (F, flood positive)wifl• a 25 hourrunningmean,<F>, drawn as the bold line; (c,d):
25-houraveragedlongitudinalcurrents(, flood positive)at StationCK (c) andHQ (d).
Backgroundphysicaldata are shownin Figure 8. Total water depthsat both sitesare

plottedin Figure8(a) andvarybetweenaboutlm and6m over the tidal cycle. The CK site
is roughly lm deeperthmi the HQ site, despitebeing furtherup-estuary. Low water is
approximately 0.5 hourlater at StationCK and high wateroccursat approximatelythe same
time. Salinitydataare shownin Figure8(b) anddemonstrate that the freshwater-saltwater
interfaceintrudesjust a shortdistanceup-estuaryof CK.
Longitudinalvelocitiesat 0.25m abovethe bedare shownin Figure8(c) for CK (dashed

line) andHQ. Rapidaccelerations occurbetweenmaximumebbandmaximumfloodcurrents
with only verybrief periodsof slackcurrentsat low water. Low waterslackoccursroughly
0.5 hourlaterat CK. Followingmaximumflood currentsat HQ thereis a slow d•eleration
until high-waterslack at roughly221 hours. This is followedby a long period of slow
currentswhichare theresultof shallowwaterovertidesanddensity-driven currentsopposing

thenear-bexl,ebb-directedtidalflow immediatelyfollowinghighwater.Therapiddecelemtion
following maximumflood currentsat CK and the long periodof near-zerocurrentson the
mid-flood appearsto be triggeredby resuspension of high SPM concentrations.This
behaviouris currentlynot understood, althoughit is consistently
observedon severalearlier
tidal cyclesand is possiblyrelatedto the dampingof turbulenceandreductionof bed shear
stressin the presenceof high SPM loads,as modelledby ShengandVillaret (1989).
8 (A): HQ(-),CK(--)
! ,
(B): HQ(-), CK(- -)
15 -
10 ' -
0 • '
215 220 225 230 215 220 225 230
(C)' !
HQ(-), CK(- -)
, 5000
(D)' HQ(-), CK(- -)
0.5
0 .....
-0.5
-1 0
215 220 225 230 215 220 225 230
TIME (HOURS)-9.0-9.5 DAYS TIME (HOURS)' 9.0-9.5 DAYS
Figure8' Timeseriesdataat 0.25mabovethebedfor Stations CK (- -) andHQ (-) locatedin Figure

1; times216 to 228 hourscorrespond
to times9.0-9.5daysin Figure7. (a):Totaldepthsof water;(b):
Salinities;(c): Longitudinalcurrents(floodpositive);(d): SPM concentrations.
14.4.5 lntratidal behaviour of SPM concentrations
Figure8(d) showsthe SPM concentrations at HQ (-) andCK (- -) asfunctionsof time for the
final tidal cyclecovering9.0-9.5 daysof the record. Flood $PM levelsare smallerthanebb
levelsat HQ, thereversebeingtrueat CK, despitesimilarnear-bed, peakcurrentvelocities
at bothsites (Figure8(c)). Thisimpliesa scarcity
of sourcemudon thebed,down-estuary
of HQ. Eachmainfloodandebb$PM peakcomprises several
sub-peaks,
someof whichmay
be theresultof patchiness
in thebed-sourcemud,whichin turnmayleadto patchiness
in the
resuspendedSPM.

Both sites exhibit low SPM levels around low-water slack (roughly 217 hours) and
extendedperiodsof very low $PM levelsaroundhigh-waterslack(roughly221 hours). The
mainfloodSPM peakat HQ leadsthatat CK andthemainebbpeakat CK leadsthatat HQ.
However,thisis not unequivocalevidencethatadvectionof $PM dominates localprocesses
becauseFigure8(c) showsthatfloodcurrents(andpossiblelocalresuspension) at HQ lead
floodcurrentsat CK, whereasebbcurrentsat CK (andpossiblelocalresuspension)
lead those
at HQ.
14.4.6 Intratidal behaviour at CK
SPM concentration (-), longitudinalvelocity(- -) andsalinity(-.-) at 0.25m abovethebedfor

CK duringthe tidal cycleat 9.0-9.5 daysare shownin Figure9(a). The salinityis zero for
mostof the tide andreachesonly 3 ppt at high-waterslack. Considering the ebbportionof
the tide afterroughly222 hours,the suddenincreasein $PM concentrations corresponds to
the increasein verticaldiffusivityfollowingthepassage of the freshwater-saltwaterinterface
oncethe criticalerosionshearstresshasbeenexceeded.The salinityis zero for mostof the
ebb.
A markedcorrelationbetweenebbvelocityandSPM concentrations at 0.25m abovethebed

is exhibitedin Figure9(b). This scatterplot impliesa criticalvelocityfor erosionof 0.2 rn
s4, or a criticalbedshearstress
of about0.15N m'"(assuming
thisto be 3.7u",whereu is
measuredat 0.25m abovethe bed, West et al., 1990). This is similarto critical shearstresses
for erosionof someestuarineandSanFranciscoBay muds(Villaret andPaulic,1986;Mehta,
1988). The strong correlationimplies that resuspension and advectionof mud from
immediatelyup-estuaryof CK are the dominantprocesses governingSPM loadsduringthe
ebb at CK.
The flood SPM levels (times greaterthan 217 hours in Figure 9(a)) show a dramatic
increaseas the criticalshearstressis exceeded.Salinityis zeroat thistime andthe enhanced
mixing on the flood tide associatedwith strongercurrentspeeds(implied from profiling
measurements but only marginallyreflectedin the near-beddatapresented)lead to elevated
$PM levelsat 0.25m abovethe bed. The near-bexl SPM-induceddensitygradientsappearto
'stall' the near-be• current and leads to a decline in SPM levels as SPM either settles or is
replacedby particlesmixed down from flood-directedcurrentshigherin the water column
where $PM levels will be smaller. The increasingsalinityimpliesthat a similar vertical
mixing is occurringfor salt. The passageof the freshwater-saltwaterinterfaceleadsto a
subsequent peakin $PM levelsas $PM associated with the interfaceadvectsthroughCK.
14.4.7 Intratidal behaviour at HQ
SPM concentration (-), longitudinalvelocity(- -) andsalinity(-.-) at 0.25m abovethebedfor

HQ duringthetidalcycleat 9.0-9.5 daysareshownin Figure9(c). The salinityis zeroduring
the peak ebb to peak flood and reaches15 ppt at high-waterslack. Consideringthe ebb

portionof thetide (afterroughly224 hoursfor currentsat 0.25m),thesuddenincreasein SPM

concentrations appearsto correspond to theresuspensionanddown-estuary advectionof mud
deposited overhighwaterbetweenCK andHQ, someof whichis cardedin the surfacelayers
of the stablewater column. An increasein vertical diffusivity follows the erosionof
stratificationand mixes near-bed SPM higherinto the water column. The passageof the
freshwater-saltwater interfaceresultsin a maximumin the verticalmixing and advection
throughHQ of SPM associated with theinterfacebecause of msuspension furtherup-estuary.
SPM levelsfall followingthepassage of theinterfacedespitestrongnear-bedcurrents, which
impliesan absence of localsourcemudandthepresence of sufficientverticalmixingthrough
the water columnto result in 'dilution' of resuspended SPM advectedfrom muchfurther
up-estuary.
U(- - ), S, (CK)
6000 (A)' SPM(-),
,
3000 (B)' SPM(EBB,CK)
4000 o o
2000-
2000 %%0
lOOO- øo o
0 :_•-
o%•
a-,-•000 o
215 220 225 230 -1 -0.5 0 0.5
4000 (C)' SPM(-),U(- - ), S, (HQ) 4000

(D): SPM (EBB, HQ)
o _
2000- •3000- o
•zooo- o•ø
cPc• -
.o%
0 • 1000-
oOgo o
o
•-2000 ' ' 0 • O• _ct:'"O

0 O•
215 220 225 230 -1 -0.5 0 0.5
TIME (HOURS)' 9.0-9.5 DAYS vElOCITY (•/s), E• -VE
Figure9: Timeseriesdataat 0.25mabovethebedfor Stations CK andHQ locatedin Figure1; times

216 to 228 hourscorrespond to times9.0-9.5 daysin Figure7. (a): SPMconcentration (-), longitudinal
velocity(U, - -) andsalinity(S, -.-) at CK; (b): Scatterplotof SPM concentration against longitudinal
velocityduringtheebb(-ve) flow at CK; (c): SPM concentration (-), longitudinal
velocity(U, - -) and
salinity(S,-.-) atHQ; (d):Scatter
plotof SPMconcenU'ation
against
longitudinal
velocity
duringthe
ebb (-ve) flow at HQ.

A correlationbetweenebb velocity and SPM concentrations

at 0.25m above the bexlis
exhibitedin Figure9(d). Thisscatterplotimpliesa criticalvelocityfor erosionof 0.35 m s'•,
or a criticalbedshearstressof about0.45 N m'"(assuming thisto be 3.7u",whereu is
measured at 0.25m abovethebed,Westet al., 1990). However,thecorrelationis clearlypoor
for currentspeeds in excessof 0.5 m s'l, whichimpliesthatadvection of SPMresuspended
from someconsiderable distanceup-estuaryof HQ is the dominantprocessgoverningSPM
loadsduringthe ebb at HQ.
The floodSPM levels(timesgreaterthan216.5 hoursin Figure9(c)) showan increaseas

soon as the near-bexlcurrent tums to flood. Therefore, SPM which has not had time to settle
duringthe brief slack-waterperiodreturnsthroughHQ. The SPM peakassociated with the
freshwater-saltwaterinterfacepasses
throughHQ whenthenear-bedsalinityis 3ppt,whichis
somewhathigherthan duringthe ebb and impliesa lag betweenthe movementof high
concentrationSPM nearthebexlandcurrents higherin thewatercolumnfollowinglow-water
slack. SPM levelsfall afterthe passageof the interface,indicatingthatthe bexl-source
mud
is exhausted.
14.5 Conclusions
Numericalmodellingshowsthat a mobile stockof fine sedimentcan be maintainedin the

upperreaches of theTamarEstuaryby tidalpumpingmechanisms associatedwithasymmetry
in boththefloodandebbtidal,barotropiccurrents(andratesof erosion)andin the durations
of thehigh andlow-waterslackperiods.Thetimeavailablefor deposition of SPM at high
water greatlyexceedsthat at low water. A turbiditymaximumin the SPM can also be
maintainedby purelytidal mechanisms in the upperreaches.
Thesemodelled
tidalprocesses
aremodifiedby buoyancy
effects.Datafromthreetypes
of experiments
in the turbiditymaximumareaof the Tamarhavebeenusedto illustratetidal
processes and their modifications
by buoyancy.Verticalprofilingdatashowpronounced
stratification
duringtheearlyebbfollowinghighwater,with subsequent reductionof vertical
mixing. Therefore,althoughbedshearstresses maybe high,it is not until the stratification
hasbeenerodedsignificantly withtheapproachof thefreshwater-saltwater
interface thatlarge
amounts of SPM canbe mixedintothewatercolumn.Verticalmixingis moreintense,asan
averageover the watercolumn,duringthe flood. The mixingmaximizesas the interface
passesthroughthe site and SPM is mixed into the near-surfacewaterswhere considerable
up-estuary transport can occur.
Thesedataare complementedby a longitudinal

transect
experiment,whichdemonstrates
the importanceof resuspension.In the stratifiedregion down-estuary of the
interface,
immediatelyfollowing
highwater,a considerable
periodof time
is availablefor settlingof SPM. As the ebb currentsincrease,the stratificationerodesand
verticalmixingincreases,
leadingto highSPMloadsas theinterface
passes
down-estuary.

Much of the SPM is advecteddown-estuary,

whereit haslittle time to settleand depositat
low water,andsubsequentlyadvectsbackup-estuary
in closeassociation with theinterface.
These featuresare also apparentin data obtainedfrom two longitudinallyseparated

instrumentpackagesdeployedat 0.25m abovethebedin the turbiditymaximumarea.SPM
concentrations,
longitudinalvelocityandsalinitydatarecordedat 5 minuteintervalsovera 10
day perioddemonstrate the growthof the turbiditymaximumover the neapto spring-tide
periodandits dependence on bothlocalresuspension, andthereforethelocationof the stock
of bed-sourcesediment,and the presenceof the freshwater-saltwater
interface.
14.6 References
Allen, G.P., J.C. Salomon,P. Bassoulet,Y. Du PenhoatandC. De Grandpre,1980: Effectsof tideson

mixing and suspended sedimenttransportin macrotidalestuaries,Sedimentary
Geology,26, 69-90.
Allen, G.P., G.Sauzay,P. CastaingandJ.M. Jouanneau,1977:Transportanddepositionof suspended

sedimentin theGirondeEstuary,France,In: Estuarineprocesses,
(ed.M Wiley), AcademicPress,N.Y.,
63-81.
Bale, A. J., 1987: The characteristics,

behaviourand heterogeneous
chemicalreactivityof estuarine
suspended particles.Ph D Thesis,Plymouth,UK, 216pp.
Bale,AJ., A.W. MorrisandRJ.M. Howland,1985:Seasonal

sediment
movement
in theTamarEstuary,
OceanologicaActa, 8, 1-16.
Delo, E.A., 1988:Estuarinemudsmanual.ReportSR 164. HydraulicsResearch,Wallingford,U.K.
Dronkers,J., 1986:Tide-inducedresidualtransport
of finesediment,
In: Physicsof shallowestuaries
and
bays, (ed. J. van de Kreeke),Lecturenoteson coastaland estuarinestudies,16, Springer-Verlag,
228-244.
Festa,J_g.andD.V. Hansen,1978:Turbiditymaximain partiallymixedestuaries:

A two-dimensional
numerical model, Estuarine and Coastal Marine Science,7, 347-359.
Grabemann, I. andG. Krause,1989:Transportprocesses of suspendedmatterderivedfrom time-series

in a tidal estuary,J. GeophysicalResearch,94, No. C10, 14373-14379.
Hamblin,P.F., 1989:Observations
andmodelof sedimenttransport neartheturbiditymaximumof the
Upper SaintLawrenceEstuary,J. GeophysicalResearch,94, No. C10, 14419-14428.

Harris,J.R.W., AJ. Bale, B.L. Bayne,R.F.C. Mantoura,A.W. Morris, L.A. Nelson,PJ. Radford,R.J.
Uncles,S.A. WestonandJ. Widdows,1984:A preliminarymodelof the dispersalandbiologicaleffect
of toxinsin the Tamar Estuary,England,EcologicalModelling,22, 253-284.
Kirby, R.A. and W.R. Parker, 1977: The physicalcharacteristics

and environmentalsignificanceof
fine-sedimentsuspensions in estuaries,in: Estuaries,Geophysicsand the environment,(ed. C.B.
Officer), 110-120, National AcademyPress,Washington,D.C.
Krone,R. B., 1962: Flume studiesof the transportof sedimentin estuarialshoalingprocesses,

Univ.
California,HydraulicEngineeringLaboratory.
Mehta, A. J., 1988: Laboratorystudieson cohesivesedimentdepositionand erosion.In: Physical

processesin estuaries(ed. J. DronkersandW van Leussen),Springer-Verlag,
427-445.
Munk, W.H. andE.J. Anderson,1948: Note on the theoryof the thermocline,

J. Marine Research,7,
276-295.
Odd, N.V.M. and M.W. Owen, 1972: A two-layermodelof mud transportin the ThamesEstuary,
Proceedings
of the Institutionof Civil Engineers,Paper 7517S, 175-205, 1972.
Officer,C.B., 1981:Physicaldynamicsof estuarinesuspended

sediments,
Marine Geology,40, 1-14.
Officer, C.B. and M.N. Nichols, 1980: Box model applicationto a studyof suspendedsediment
distributions
andfluxesin partiallymixedestuaries.In: EstuarinePerspectives,
AcademicPress,London,
329-340.
Sheng,Y.P. andC. Villaret, 1989:Modelingthe effectof suspended

sedimentstratificationon bottom
exchangeprocesses,
J. GeophysicalResearch,94, No. C10, 14429-14•.•.•..
Uncles,R.J., R.C.A. Elliott and S.A. Weston,1985a:Observedfluxesof water,salt and suspended

sedimentin a partly mixedestuary,Estuarineand CoastalSheffScience,20, 147-167.
Uncles,R.J.,R.C.A. Elliott andS.A. Weston,1985b:Dispersion

of saltandsuspended
sedimentin a
partly mixed estuary,Estuaries,8, 256-269.
Uncles, RJ., R.C.A. Elliott and S.A. Weston, 1985c: Lateral distributionsof water, salt and sediment
transportin a partlymixedestuary,In: Proc.of the 19thInt. Conf. on CoastalEngineering,
ASCE,
Chapter204, 3067-3077.
Uncles,R.J., J.A. Stephensand T.Y. Woodrow,1988: Seasonalcyclingof estuarinesedimentand

contaminanttransport,Estuaries,11, 108-116.
Uncles,R.J. andJ.A. Stephens,

1989:Distributionsof suspended
sediment
at highwaterin a macrotidal
estuary,J. GeophysicalResearch,94, No. C10, 14395-14405.

Uncles,R.J.,J.A. Stephens, R.H. Bruce,C.D. BarrettandN.J.Bloomer,1989: Observationsof temporal

variabilityin a partlymixedestuaryusinga newinstrument package,In: Developmentsin estuarine
and
coastalstudy techniques (eds.J. McManusandM. Elliott),OlsenandOlsen- InternationalSymposium
Series,Fredensborg, Denmark,151-158.
Villaret, C. and M. Paulic, 1986: Experimentson the erosionof depositedand placedcohesive

sediments in anannularflumeanda rockingflume.Rep.UFL/COEL-86KI07,CoastalandOceanographic
Eng. Dept.,Univ. Florida,Gainesville.
West, J.R., K.O.K. Oduyemi, A.J.'Bale and A.W. Morris, 1990: The field measurement
of sediment
transportparameters in estuaries,
Estuarine,CoastalandShelf Science,30, 167-183.
Wolanski, E., J. ChappeR,P. Ridd and R. Vertessy, 1988: Fluidizationof mud in estuaries,J.
GeophysicalResearch,93, No. C3, 2351-2361.

15
Velocity asymmetriesin
frictionally-dominated tidal embayments:
longitudinal and lateral variability
C T Friedrichs,D R Lynch and D G Aubrey
Abstract
Asymmetrictides play a centralrole in net near-bedtransportand, ultimately,in the

morphologic evolutionof small(length<-20 kin), well-mixed(tidalamplitude/depth >-0.1)
tidal embayments. In this studywe examinethe physicswhich causedifferentpeak tidal
velocitiesduringthe ebb and the flood, and relatethesepattemsof velocitydominanceto
geomorphologies commonto shallowtidalembayments. Underconditions appropriate
tomany
shallowembayments, frictionis one-to-twoordersof magnitudelargerthanacceleration
in the
cross-sectionallyintegratedmomentumequation.Tidal flow in thesesystemsthereforemay
be describexlby a zero-inertia diffusive model which capturesthe most important
non-linearities,
while greatlysimplifyingconceptualinterpretation.Linearizedsolutionsare
governed
by thefrictionallengthscaleL•, whichdetermines
boththepropagation
speedand
therateof amplitudedecay.Non-linearbehavioris represented
by variationin the magnitude
of L• duringthetidalcyclewhich,in tum,reflects
variation
in totalsystem
widthandmean
channeldepth. Non-linearnumericalsolutionsto the zero-inertiaequationrecreatepatterns
of velocitydistortionobservedin thefield andareconsistent
with conceptual
extensionsof the
linearized solutions. Shallower areas tend to be flood-dominant and enhance landward
near-bedtransport;deeperareastend to be ebb-dominantand enhanceseawardnear-bed
transport.Changesin morphologyare associated
with reversalsin velocity-dominance. The
feedbackbetweenhydrodynamics and embaymentmorphologyis stronglypositive. We
hypothesizethatantecedent
geometrymaydictatetheinitialoverallsenseof velocitydistortion
for an embayment,and that hydrodynamics
slowlymodulatethat morphologyin a generally
positivefashion.

278 C T Friedrichs,D R Lynch andD G Aubrey
15.1 Introduction
The studyof tidalnon-linearitiesis importantfor geologicunderstanding of coastalevolution,

especiallyin shallowembaymentswhich are ephemeralgeologically.Asymmetrictidal
velocitiescontributestronglyto themorphodynamics of small(length<-20 kin), shallow(tidal
amplitude/mean depth>-0.1) tidalembayments.Net bedloadsediment transportoftenoccurs
in tidalsystemshavingasymmetric tidalcurrents,
evenunderconditions of zeroresidualwater
volumetransport(Aubrey,1986). If bedloadtransportis assumed to be a geometricfunction
of velocity(e.g., Meyer-PeterandMuller, 1948), thenthe directionof net coarsesediment
transportwill tendto be in the samedirectionas the maximumcurrentvelocity. Residual
transportof suspended sedimenthasalsobeenrelatedto asymmetries in tidalvelocity(Postma,
1961; Aubrey, 1986; Dronkers,1986a). If a longerperiodof slackwateroccursnearhigh
water thannearlow water, moresedimentmay settleout of the watercolumnafter the flood
thanafter the ebb.Furthermore,asymmetries in maximumvelocitydirectlycontributeto net
suspended sedimenttransportunderconditions of significantscourlag; thatis, whena larger
velocityis requiredto erodecohesivesedimentthanthatat whichit settiesbackto the bed
(Postma,1967). Thusasymmetricvelocitypatternsin tidal embayments provideinsightinto
potentialpatternsof sedimenttransport and,ultimately,intotheevolutionof theirmorphology.
Peak bottom stressis a functionof maximum depth-averaged velocity for well-mixed

conditions.In thisstudywe examinethephysicswhichcausedifferentpeakvelocitiesduring
the ebb and the flood, and thereforecontributeto net near-bedsedimenttransport. We then
relate thesepatternsof velocity dominanceto geomorphologies commonto shallowtidal
embayments.Here we definea pointwithin an embaymentasbeingflood- or ebb-dominant
if the maximumcurrentspeedduringa tidal cycleat that locationoccursduringthe flood or
ebb, respectively. Asymmetriesin flood and ebb duration can be determinedfrom
observationsof tidal elevation alone, suggestingthe probablevelocity distortiontype.
However, to determineunambiguously local ebb- or flood-dominance, velocitiesmust be
analyzed. In this studywe investigatefive specificaspectsof velocitydominanceobserved
in shallowtidal embayments.The first threeconcerncross-sectionally averagedflow (i.e.,
one-dimensional), whereasthe last two considerdepth-averaged flow (i.e., two-dimensional):
1. Embayments characterized
by largechangesin meanchanneldepthduringthe tidal cycle
haveflood-dominant main channels.2. Embayments characterized by largechangesin total
width during the tidal cycle have ebb-dominantmain channels. 3. Systems with
flood-dominantmain channelstend to have ebb-dominanttidal inlets. 4. Systemswith
ebb-dominant mainchannelstendto haveflood-dominant shallowmarginsandintertidalareas.
5. Tidal inlet cross-sections
areoftencomposed of separatedeepor ebb-dominant andshallow
or flood-dominant sections.
To investigate spatialvariations
in velocityasymmetry
in a varietyof shallowembayments,
a zero-inertiadiffusivemodelof tidaldynamicsis developedfor well-mixedtidal flow. We
showthatunderconditions appropriate to manysmall,well-mixedtidalembayments, friction
is one-to-twoordersof magnitudelargerthanthe acceleration term in the cross-sectionally
integratedmomentumequation.Linearanalyticsolutions to thezero-inertiaequationareused

C T Friedrichs,D R Lynch andD G Aubrey 279
to identify non-dimensional parameterswhich govern non-linearbehavior. Non-linear

numericalsolutionsto the zero-inertiaequationare usedto recreatethe patternsof velocity
distortionobservedin the field. A physically-based
conceptualmodelfor velocitydistortion
and near-bedsedimenttransportresults.
We concludethat patternsof velocityasymmetryin thesesystemsreinforcemorphology

and likely play a centralfeedbackrole in its creationand maintenance.
Shallowerareastend
to be flood-dominant and enhancelandwardnear-bedtransport;deeperareastend to be
ebb-dominant andenhanceseawardnear-bedtransport.Changesin morphology areassociated
with reversalsin velocity-dominance. Also,because of thedependenceof channelgeometry
on tidal prism,we hypothesize that antecedent geometrymay dictatethe overall senseof
velocitydistortionfor an embayment,and, therefore,predetermine evolutionarytrends.
15.2 FrictionalDominanceof the MomentumEquation
15.2.1 Previous work
Throughscalingarguments, field measurement

and/ornumericalmodelingof the individual
termsof the 1-d momentumequation,many authorshave demonstrated the dominanceof
frictionover the inertialtermsin well-mixed,shallowtidal embayments and estuaries.A
surveyof the literature(Table 1) indicatesthat in systemsof interest(we!l-mixed,tidal
amplitude/mean
depth>4).1, tidal velocities
-0.5 ms'•),the frictionterm is typically
one-to-twoordersof magnitude largerthaneitherthelocalor theadvectiveacceleration term.
Furthermore,the local and advectiveacceleration termsare typicallyof oppositesignand
partiallycancel. The relativemagnitudeof frictiondependsdirectlyon the choiceof bottom
drag coefficient,which is 0.009, on average,amongthe first ten studieslisted in Table 1.
Other authorswho have emphasizedobservednon-linearities,but who did not measurethe
momentumtermsdirectly,may have underestimated the effectof frictionby assuminga
"classical"shelfvalueof ca= -0.002 to be appropriate for systemshavingmeandepthsof
only a few meters(e.g., Robinsonet al., 1983;Unclesand Kjerfve, 1986). The last two
entriesin Table1 demonstrate thelimitsof thefrictional-dominanceassumption. In theLower
ColumbiaRiver Estuarywheresalinityintrusionis present,the shearstressat the bed is
reducedby stratificationin thewatercolumn,partiallydecoupling theoverlyingflow fromthe
bottom(GieseandJay, 1989). Upriverbeyondthe intrusionof salinity,the tidal pressure
gradientis primarilybalancedby friction (Gieseand Jay, 1989). In this paper,the
frictional-dominanceapproximation is appliedonly to tidalembayments havingwe!l-mixed
watercolumns.In the final example(PingteeandMaddock,1978),the EnglishChannelis
simplytoodeepandbottomstress toosmallfor frictionto dominate themomentum equation.
We applyfrictional-dominance onlyto systems havinglargeratiosof tidalamplitude to mean
waterdepth,•/h > 4).1.

Table1' Magnitudes
of local(/.,4) andadvectiveacceleration
(AA)relativeto thefrictionterm(F), along
with relevant characteristic scales.
Location LA/F AA/F h •'o Uo Cd Source

(m) (m) (ms-l)
1. 1-dtrapezoidal 0.04 0.01 2 1 0.5 0.02 Speer(1984)

channel
2. 1-dmodelof Conwy 0.04 0.04 3 2.4 0.5 0.01 Miinchow andGatvine(1991)
Estuary,Wales
3. ConwyEstuary, 0.05 0.02 3 2.4 0.5 0.006 Wallis andKnight(1984)
Wales
4. FraserEstuary,BC 0.05 0.05 9 4.5 1 0.005 LeBlond(1978)
Canada
5. 1-dmodel,StonyBrk. 0.05 0.1 2 0.9 1 0.01 Park (1985)
Harbor, NY, USA
6. Macquarie
Harbor, 0.1 0.007 6 0.5 1 0.004 van de Kreeke(1967)
Tasmania
7. GreatBay,NH 0.1 0.02 7 1.3 1 0.03 Swift andBrown(1983)
USA
8. St.Lawrence
Estuary, 0.1 0.1 7 3.5 1 0.001 LeBlond(1978)
Canada
9. Ingram
Thorofare, 0.2 0.07 3 0.5 0.7 0.002 Weismanet al. (1990)
NJ, USA
10. DelawareEstuary, 0.4 0.03 6 0.7 1 0.0025 Parker (1984)
USA
11. *ColumbiaR. Estuary, 1 1 10 1 1 0.0008 GieseandJay (1989)
WA, USA
12. *EnglishChannel 5 2.5 40 2 0.8 0.0025 PingreeandMaddock(1978)
*Thesetwoexamples
areincluded
to illustrate
limitsof thefrictional
dominance
assumption.
Seetextfor
discussion.
It was only recentlythat the dominantrole of friction in the momentumbalanceof shallow

tidally-dominated estuarieswasdemonstrated to theestuarinecommunityby LeBlond(1978).
Despite the work of LeBlond, the more appropriatefrictionalinterpretation of highly
non-lineartidalflowsin shallowembayments hasreceivedrelativelylittle attention.During
the last decadeonly a few papershaveusedfield observations (Aubreyand Speer,1985;
FriedrichsandAubrey,1988),numericalmodeling(SpeerandAubrey,1985;Westerinket al.,
1989), or scalingarguments (Parker,1984; Munchowand Garvine,1991) specificallyto
address therole of non-linearfrictionaldistortionin shallowtidalembayments.The neglect
of non-linearfrictional effects may result in part from long-heldclassicalviews of
depth-averaged estuarinenon-lineardistortion.A century-and-a-half ago,Airy (1845) noted
thatin a frictionless
estuary
where• / h is large,thetidepropagates
asa shallowwaterwave
with depth-dependent
phasevelocity. Lamb (1932) combinedthe frictionlessnon-linear
advection
andcontinuityeffectsintoonenon-linear
expression
for frictionless
phasevelocity:

c- g• 1+-- _ +O ,
2 h
whereh is undisturbed meandepth,and•(t) is time-dependent tidalelevation.Becausetotal

water depthis significantlygreaterat the tide crestthan at the trough,the crestpropagates
more quickly than the trough,resultingin a shorterflood, a longerebb, and occurrenceof
highest velocity currents during the flood. Since the time of Airy, "shallow water"
non-linearityhasbeensynonymous with the non-lineareffectsof frictionlessaccelerationand
continuity.Discussion of Airy-typeshallowwaternon-linearity haslargelyovershadowed the
enhancement of frictionalnon-linearities
as•o/ h increases. In thefollowingsection, a sc_a_l_ar
analysisappropriateto shallow tidal embaymentsof interestis used to demonstratethe
dominanceof frictionover inertiain the 1-d momentumequation.
15.2.2 Scalaranalysis
I I
Figure 1: Schematicdiagramof a tidal embaymentcross-section:

b(x,t) is the total width of the
embayment
cross-section,
includingtidal fiatswhichactin a storagecapacityonly;bc is the widthof
themomentum-transporting
tidalchannel; • = surface
elevation
relativeto a verticaldatum;HW = high
water;LW = low water;Ref = datumlevel.
The cross-sectionallyintegrated,one-dimensionalequationsof motionfor well-behaved,

channelized
flow in a tidal embayment with intertidalfiats(Fig. 1) maybe expressed
as:
ax
+ at
- 0, contiauity (15.2)

282 C T Friedrichs,
D R LynchandD G Aubrey
mommmm (15.3)
whereQ is along-channelflux (confmed to thesub-tidal

channel),
x is horizontal
distance,A
is totalcross-sectional
area(includingtidalflats),A½andPcarethecross-sectional areaand
perimeterof the momentum transporting,sub-tidalchannel,and othersymbols are as
previously defined.
Restatedin termsof characteristic
scales,
(15.2)and(15.3)become
(15.4)
(]5.5)
Weassume
O(•SQ)
.• O(Q)andsolve
forfixin(14.4).Thensubstituting
fixinto(15.5)gives:
•Q +o ,•A •A•A••Q •(•A/ 0 z

0 •- A½
• A•A•1St 15tiaO}A•
(•.•)
If weassume O(•SA/A½)
< O(1)(which
isthecaseforallobservedand/or
modeledembayments
considered inthispaper),
thenthesumofthethreeinertia
terms isO(f)Q/•St).
Combiningthe
firstthreetermsof (15.6)anddividing
(15.6)bythescaleof thesurface
gradient
gives:
O + O(1) + O. -0. (15.7)

garbage gA••A• (
The ratio of the friction scale to the inertial scale is then
___F.
•t%e• (]5.8)
where
Uois thecharacteristic
scaleof velocity,
averaged
overthemomentum-transporting
portionof the cross-section,
and R½ = Ac/P½is the hydraulicradiusof the
momentum-transporting
channel.In shallow
tidalembayments
of interest,
Uoisof theorder
0.5ms
'•, c,••- 10':,thesemi-diurnal
period
& = 4.5x 104 sec,and1 m < Re< 10m.
Therefore
F will typicallybe one-to-two
ordersof magnitude
largerthanI in these

tidally-dominated embayments.SinceF/I is frequency-dependent, however,one shoulduse

the periodof the overtideof interestwhenconsidering stronglynon-linearflow. This is not
a seriouslimitation:F will still dominateI by one-to-twoordersof magnitude,even if one
scales(15.8) with the quarter-diurnal
tidal period.
AssumingF/I >> 1, thenfrom (15.7) for a singleembayment:
[,,) (15.9)
Thus ca is sensitive to the fraction of the cross-sectionthat is defined as

momentum-transporting. If one comparesdrag coefficientsfrom severaltidal embayments,
a consistentdefinitionshouldbe used. In Table 1 (entries1-10), the momentum-transporting
cross-sectionsusedin thebalancecalculations are all clearlydefined,sub-tidalchannelswith
negligiblechangein widthduringthetidal cycle. In caseswhereextensiveintertidalflatsare
immediatelyadjacentto subtidalchannels(Speer,1984;Park, 1985;MunchowandGarvine,
1991), the authorstreatedintertidalflatsin a storagecapacityonly, excludingextensiveflats
from the momentum balance.
15.2.3 Numerical comparison
We also solved (15.2) and (15.3) numerically,with and without the inertial terms, to
demonstrate
furtherthe validity of the zero-inertiaapproximation
in shallowregimes. The
modelembaymentconsistsof a sevenkilometerlong, wide rectangularchannel(i.e., Rc =
channel depth,Pc= bc= constant),
witha linearlysloping bottom.We let ca= (0.013mz/a)
R•'z/a,usingtheManningformulationfor bottomfriction. ForRc= 2 m, ca= 0.01. The
channel bottom rises from two meters below mean sea level at the ocean end to one meter
abovesealevel inland, and is forcedby a one meteramplitudeoceantide. Eqs. (15.2) and
(15.3) with the inertial termswere solvedexplicitlywith a 1-d finite differencemodel (At =
15 s, Ax = 250 m, from Speerand Aubrey, 1985). Omittingthe inertial terms,(15.2) and
(15.3) were solved implicitly using Crank-Nicolsonin time (At = 450 s) and 1-d finite
elementsin space(Ax = 250 m). Withoutthe inertialterms,the solutionis free of advective
time-stepping constraints,allowing stablesolutionat much larger At. This may be an
importantconsideration in the solutionof highlycomplexgeometries in two dimensions.
The zero-inertiaand fully dynamicsolutionsare essentiallyidentical(Fig. 2). Exceptat

slackwater,the pressuregradientand frictiontermsin the fully dynamicsolutiondominate
the inertialtermsby one-to-twoordersof magnitude.Furthermore,the local and advective
acceleration termsare typicallyof oppositesignandpartiallycancel. Clearlythe zero-inertia
assumption is inappropriate
at slackwaterwhenu = 0 andin the immediatevicinityof sudden
channelexpansions or fightchannelmeanders.However,duringmostof the tidal cycleand
for the majorityof the tidal embayment,the zero-inertiaassumptionis weB-justified.

284
C T Friedrichs,
0.5
a)
0
-1
0 5 10 15 20 25 30 35 40
0 7
0.35
o
•ß -0.35
I 0.3
:•
5//,•x 0
//
-0.35
-0 7
0 5 10 15 20 25 30 35
1. temporal
acceleration
2. advectiveaccelera•on
3. pressure
gradient
4. bottom friction
time (hours)
Figure 2: Comparisonof numericalsolutions

to theone-dimensional
equations
of motion
withand
without theinertial
terms. (i) Surface
elevation
and(ii)transport
perunitwidth
atfivetidalstations,
and(iii)momentum balance atmiddlestation:
(a)withtheinertial
terms;
(b)without
theinertial
terms.

C T Friedrichs,D R LynchandD G Aubrey 285
15.3 Analyticaltreatmentof the linearizedzero-inertiaequation
15.3.1 Formulation
The zero-inertiamomentumequationfor well-behaved,cross-sectionally

averagedflow may
be expressedas (e.g., Henderson,1966):
•Ox +nZRc
-4aIulu- O -toO( (15.10a,b)
Ox
whereu is velocityaveragedoverthemomentum-transporting
portionof thecross-section
(Fig.
1),n = RcU•(c,,,/g)
mistheManning
frictioncoefficient,
andtheothersymbols
areaspreviously
defined. Cross-s•tionally averaged continuity (15.2) may be re-expressed,with
momentum-transport
confinedto the channel,as
b c9•
c3t
+•Ox bchu
I 01 (15.11)
whereb andbcare as definedin Fig. 1, andh is the cross-sectionally

averageddepthof the
momentum-transporting
channel. Substituting(15.10b) into (15.11) gives
a• 1 a (Da• I 0 (15.12)
at b ax!• -•xx} '
whereD is a non-lineardiffusioncoefficientgivenby
(15.13)
Eq. (15.12) may be solvedto first order analyticallyif we assumethat D is finite and
slowlyvaryingin timeandspaceduringmostof thetidalcycle. Thisanalysisis inappropriate
for tidalflow at slackwaterbecause D is undefinedfor zerosea-surface gradient.However
this linearized analysis still provides valuable insight into the behavior of
frictionally-dominatedtidal propagation
duringthe majorityof the tidal cycle. Underthese
assumptions, (15.12) may be restatedas
a• I [ a2• (15.14)
Ot CgX
2'
with a "constant"
diffusioncoefficient,,c,defined(with Rc-- h) as

bc h• c3C[
-m (15.15)
b n axl
The boundaryconditionsfor theproblemare no-flowat x = L, whereL is thelengthof the

channel,andsinusoidal
forcingof amplitude
•o at x = 0. Theanalyticsolution
for l•(x,t)is
cx•(•Lj,-
•t- 9),-•ttt+
oos(•L•+
•0, •ttr*+cos(x4L•-
•0e-•t¾*
+cx•(•Lj,+
•t- 9),•ttt
2 (eo• 9 + cosh9)
(15.16)
where
t0isangular
tidalfrequency,
L/= (2r,/t0)
mand• = 2L/Le
15.3.2. Length scalesand zeroth-orderresults
Eq. (15.16) describes

an exponentiallydecaying,partiallyprogressive,
partiallystandingwave
governed bythefrictionallengthscaleLe L/determines boththepropagationspeed (c = LAo)
andtherateof amplitude decay(e-foldinglength= L/) of eachof thefourtermsin (15.16).
If L/L/<<I,then(15.16)reducestoanx-independent standing wave.If L/Lj•>I, then(15.16)
reducesto an equationfor a diffusivepropaFating
wave:
½(x,t)- CoC•(x/L•-
•Oe-•tty. (15.17)
Eq. (15.17) is analogousto the infinitetidal channelsolutionconsideredby LeBlond(1978).

The infinite channelsolutionof LeBlond,however,does not capturethe full range of
zeroth-order(i.e., amplitudedecayandphaselag) behaviorfoundin the small(L < - 20 km)
systemsof interestto the presentstudy. Significantvariationis seenin (15.16) over a small
rangeof L/L/(Fig.3).
Weestimate
L/asa function
ofknown
parameters
byfirstevaluating
la•fdxlatx = 0. From
(15.16), it may be shownthat
a½
0x
(x-0)- (sin•eos(•t-
n/4)+sinhq•sin(•t-
n/4)) . (15.18)
L(cos,
ForL/L/> 1.25,sinh•is at leastanorderof magnitude

largerthansin•, andsinh••- cosh•.
Thusfor L andL/of the sameorder,we maydetermine la[Jaxlusingthesesimplifying
approximations.
An estimate
for thetime-averaged
absolute
valueof (15.18)is thengivenby

C T Friedrichs,
D R LynchandD G Aubrey 287
(z=O) - 2meø
sin(•/4-
eO= (15.19)
(a) L//4' = 1/2
0.5
o 1 2 3 4 I 2 3
t/me (a•n) t/me (a•n)
(d) I///=4
.
1
1
0.5
-1
1 2 3 4 0 1 2 3
time(an/n) time
Figure3: Timeseries
of Eq.(15.16)during
twotidalcycles
recorded
atx/L= 0, 0.2,0.4,0.6,0.8and
1.0:(a)L/L/=1/2;(b)L/L/= 1;(c)L/L/=2; (d)L/L/=4.
Wesolve
directly
forL/bysubstituting
thisrelation
into(15.15)
along
withthedefinition
L/
= (2r,/co)a:
(15.20)
According
to(15.20),
propagation
speeddecreases
andamplitude
decay
rateincreases
asdepth
is reduced,
frictionis increased,
or totalembayment
widthis increased
relativeto thewidth
of thechannel.Eq. (15.20)is alsoconsistent
withobservations
thatindicateshallowtidal
embayments actaslow-pass filters(AubreyandSpeer,1985).

(a) SwashBay andChannel,VA, USA L/Lf= 0.38 (b) StonyBrook,NY, USA L/L/= 0.57
.... 1 ß
0.5
-0.5
0 5 10 15 20 25 0 5 10 15 20
time (hours) time (hours)
(c) SouthChannel,Nauset,MA, USA L/L! = 0.97 (d) The Fleet,EnglandL/L/= 3.3

0.6
1
0.4
0-5
0,' /,
-0.2
-0.4
_
-0.6
0 5 10 1'5 2'0 25 0 5 10 15 20 25
time (hours) ume (hours)
Figure4: Time seriesof meanobservedM2 surfaceelevationduringtwo tidal cycles:(a) SwashBay and

Channel(x/L = 0, 1); (b) StonyBrookHarbor(x/L = 0, 0.2, 0.4, 0.8); (c) SouthChannel,Nauset(x/L
= 0, 0.2, 0.3, 0.5, 0.7, 0.9); (d) The Fleet (x/L = 0, 0.2, 0.4, 0.6, 0.8, 1).
The zeroth-orderdecayand delay of the M2 tide in four tidal embayments (Fig. 4) are
consistent
withbehaviorpredictedby simple,linearsolutionsto thezero-inertiaapproximation
(Fig.3). Table2 summarizes
theparameters
usedin calculating
L/for fourrealsystems.
We
choosebc/bequalto embaymentareaat meanlow waterdividedby embaymentareaat mean
tide level (exceptfor The Fleet, for which Robinsonet al., 1983, providebcand b). The
valuesgivenfor Manning'sn (Table 2) are basedon the n or cavaluesusedby the authors
in 1-d numericalmodelsof the embayments.Robinsonet al. (1983) assumed a typicalshelf
valuefor the dragcoefficient(ca= 0.002), possiblyunderestimating the frictionaleffectsin
the tidal channel.They thenmatchedobservations of tidal elevationby tuningthe valuesof
bc/bin theftnumericalmodel,a modelwhichlimitsmomentumtransportto thecentralwidth,
bc. Eq. (15.20) demonstrates that decreasingbc/b is analogousto increasingn in a
frictionally-dominated
system.SinceRobinson
etal.tunedbc/b,theL/L/valuecalculated
for
The Fleet (Table 2) should still be accurate,even if Robinsonet al. underestimatedn.
Robinsonet al. notedthatlow valuesof bc/bstronglyincreaseamplitudedecayandphaselag,
but theymistakenlyassumedthatthetidein The Fleetpropagates asan Airy-typefrictionless
long-wave.With ca,bc, Q and h as givenby Robinsonet al., F/I = 40 for a semi-diurnal
tide
in The Fleet. By assuming thedominance of accelerationoverfriction,theyderiveda phase
speed
of (ghbJb)
m,whichbecomes
smaller
atlowbdb.Robinson
etal.alsoneglected
friction
in their discussion
of tidal non-linearityand attributedobservedflood-dominance
only to

Airy-type behavior(seeEq. 15.1). The followingdiscussionwill demonstrate,however,that

frictional-dominanceprovidesa moreappropriate, physically-basedmodelfor the non-linear
behaviorof shallowtidal embaymentsof interesthere.
Table2: Geometricandfrictionallengthscales(andtheparameters
usedfor calculation)for four shallow
tidal embayments.
Location
L h M2o bc/b Manning's
n L/Lf Sources
(km) (m) (m) (m-1/3s)
1. SwashBay andChannel,6.4 2.5 0.58 0.61 0.025 0.38 BoonandByrne(1983);

Wachapreague,
VA
2. StonyBrookHarbor, 5.2 1.7 0.86 0.77 0.035 0.57 Park(1985);

NY, USA Friedrichs
andAubrey(1988)
3. SouthChannel, 8.2 2.1 0.98 0.76 0.051 0.97 AubreyandSpeer(1984,1985);
Nauset, MA, USA Ffiedrichs
andAubrey(1988)
4. TheFleet,England 12.5 0.7 0.57 0.32 0.015 3.3 Robinson
et al. (1983)
15.3.3.Non-linearities
predictedby the linearizedequation
Simple
analytic
solutions
suchasthose
displayed
in Fig.3 assume
L/L•,to beconstant
throughout
the tidal cycle. However,non-dimensional
parameters
importantto non-linear
behavior
canstillbeidentified
byconsidering
variations
inthemagnitude
ofLj•during
thetidal
cycle.Sincethetidalwavepropagation speedis proportional
to thefrictionallengthscale(c
= LI c0),thetidalwavewill propagate
mostquicklyduringtheportionof thetidalcyclewhere
Lj•is greatest.
If Lj•is significantly
greater
around
highwaterthanlowwater(L•i•n>>L.
now),
thenthe propagation
speedwill be greaternearhigh water,and the crestof the tide will
propagatemorequicklythanthetrough.Theresultwill be a shorter
flood,a longerebb,and
occurrence
of highest
velocity
currents
during
theflood.Analogously,
if L•i•n<<L.
now,
the
troughwill propagate morequickly,andebb-dominant currents
will result.From(15.20)the
ratio of high- to low-waterfrictionallengthscalesis dependenton two non-dimensional
parameters,
bnign/btow
andhhign/htow.'

290 C T Friedrichs,
D R LynchandD (3 Aubrey
Thisconceptual
extensionof lineartheorymaybetestedby measuring tidallagsrelativeto
boundaryforcingin real systems.Tidal lag, •, is inverselyproportional
to the tidal
propagation
speed
and,therefore,
inversely
proportional
toL•, Thustheratioof tidallagsat
highandlow wateris givenby
(15.22)
Theratioof Oh•Sh
to Olo,•provides
a measure of thedegreeof tidalasymmetry and,by our
conceptual
extension
of lineartheory,a roughmeasureof theratioof frictional
lengthscales
nearhighandlowwater.Table3 compares
ratios
of Oh•s•
toOlow
observed
in fourshallow
embayments
withvalues
predicted
by (15.22).In all fourcases
thegeneral
trend(>1 or <1)
predictedby (15.22) is in agreement
with observations.
Table3' Predicted ratioof tidallags(andtheparameters

usedforcalculation)
alongwithmeasuredtidal
lag ratiosat innermostgauge,basedonM2 andM4 only. AD = riseduration- fall duration.
predicteA observed
hhighbhigh6}high• 6}high
-- -- 6} AD Sources
Location hlowblow6}low
(hours)
(hours)
6}low
1. SwashBay and Channel, 1.6 4.0 1.49 0.65 0.28 1.55 Byrne et al. (1975);
Wachapreague,VA BoonandByrne(1981)
2. StonyBrookHarbor, 2.8 1.6 0.44 1.7 -1.1 0.53 Park (1985);

NY, USA FriedrichsandAubrey(1988)
3. South Channel, 1.9 1.6 0.67 1.6 - 1.3 0.43 AubreyandSpeer(1985);

Nauset, MA, USA FriedrichsandAubrey(1988)
4. The Fleet,England 2.1 1.0 0.44 4.9 -1.6 0.72 Robinsonet al. (1983)
15.4Numerical
treatment
of thenon-linear
zero-inertia
equation
Linearsolutions
to thezero-inertia
equation
identify
parameters
(L•s•L•o,•,
b•s•b•o,,,,
hnisn/hlo•,)
whichlikelygovern
non-linear
behavior. We nowutilizefullynon-linear
solutions
toconfn'm
theimportance
ofthesenon-dimensional
parametersandtoinvestigate
fivespecific
patterns
of velocity
asymmetryobserved
in thefield:(a)Embayments characterized
bylarge
changes
inmeanchannel depth
during
thetidalcyclehaveflood-dominantmainchannels.
(b)

Embaymentscharacterizedby large changesin total width during the tidal cycle have
ebb-dominantmain channels. (c) Systemswith flood-dominantmain channelstend to have
ebb-dominanttidal inlets. (d) Systemswith ebb-dominantmain channelstend to have
flood-dominantshallowmarginsand intertidalareas. (e) Tidal inlet cross-sections
are often
composedof separatedeep or ebb-dominantand shallowor flood-dominantsections. The
followingsub-section addresses the first two observedpatternsof velocitydistortionand
discusses empiricalrelationshipsbetweentidal prismandchannelgeometry.Sub-sections 2
- 4 individuallydiscusspatterns(c) - (e).
Table4: Resultsof least-squares

linearregressions
for channelarea(A½),width (be)andmeandepth(h)
asexponential
functions
of springtidalprism(f), of theformZ = ctaQo,
whereZ is thedependent
variableando• andI5 are constants.The linearregressions areperformed on the naturallogsof the
variables,thereforethe linearequationis of the formIn • = •t + I3 In fl. 0 = standarderror.
Z In a ]•
1. Alongchannel(datafrom Chantler,1974) Ac - 6.8 0.87

(0.4) (0.02)
bc -3.2 0.56
(0.4) (0.02)
h -3.8 0.32
(0.3) (0.02)
2. At tidal inlet (datafrom Jarrett,1976) Ac -11 1.1

(0.9) (0.05)
bc -8.3 0.86
(1.4) (0.08)
h -2.8 0.25
(1.2) (0.07)
15.4.1 Velocity-dominance
of the main tidal channel
Linear solutionsto the zero-inertiaequationsuggestthat either flood- or ebb-dominant

velocitieswill be producedin the main tidal channelby non-linearsolutionsto (15.12)
depending
onthevalues
of bhish/blow
andhhis•/hlo
•. Therefore
wewished
toinclude
a rangeof
thesenon-dimensional geometricparameters.It waspracticalto investigate only a f'mite
numberof fully non-linearsolutions,and we wishedto limit investigations to realistic
morphologies. Thusit wasusefulto followempiricalrelationsbetweentidaldischarge and
channelgeometry.Two of the mostsystematic compilations of empiricaldataarethoseby
Jarrett(1976),whorelatedchannelcross-sectional
parameters at tidalinletsto springtidal
prism,gl, andby Chantlet(1974),whorelatedcross-sectional
parameters alongthelengthof

tidal channelsto maximumspringdischarge,Q,.. Here we utilize both setsof data by

approximating maximumspringprismasffl = 2Q,,fio,andfittingexponential
relationships
of
theforma.Cl
• (• and• constant)
to channel
cross-sectional
area,widthandmeandepthat
meantide level (Fig. 5; Table 4). Followingtheseempiricalrelations,channelwidth and
depthbecomeuniquefunctionsof tidalprism,andreasonably realisticmodelgeometries
can
be producedeasilyby first specifyingthe tidal rangeandareaof intertidalstorage.
(a) At tidal inlets (b) Along tidal channels

105 ........................................... 105
104
103
104
102
10• +
+•'•
ß
103 lOO
104
107 108 109 103 104 105 106 107 108 'i'09
spnng
prism
(m3) •ng •sm(m•)
Thames x Delaware
ß InletalongU.S.AtlanticCoast
with oneor nojetties Forth + Unnamedestuary
Savannah ß Wrecked Recorder Creek
Figure5' Cross-sectional
areaas a functionof springtidalprism:(a) Measuredat 37 tidalinletsalong
the Atlantic Coastof the U.S. with one or no jetties (data from Jarrett,1976); (b) Measuredat stations
alongthe lengthof six tidal channels(datafxomChantlet, 1974).
In solving(15.12)we chosemeantideconditions
of • = 1 m, L = 7000m andn = 0.036.
Manning's
n = RcU•(c,t/g)m
= 0.036corresponds
to caTM
0.01forRc= 2 m or ca= 0.006for
R, = 10 m, a roughapproximation of the rangeseenin Table 1. The distribution
of intertidal
storagefor the modelis shownin Fig. 6. We assumed maximumspringdischarge to be
one-and-a-half timesthatof a meantideandthendetermined channelgeometries on thebasis
of the empiricalrelationssummarizedin Table 4. Jarrett'sdata were used to determine
channel cross-sections at the inlet, whereas Chantler's data were used to determine
cross-sections
within the innerportionof the basin.A linearcombinationof the two relations
wasusedin-between.Basinwidthsrangingfrom 100m to 10 km resulted
in 10• m3 < fl <
108m3andthedistribution
ofembaymentcross-sections
shownschematically
inFig.7. Small
tidalprisms
areassociated
withshallow
channels
andnarrowintertidal
areas,
i.e.,hhign[hto
w>>
1. Large
tidalprisms
areassociated
withdeep
channels
andwideintertidal
areas,
i.e.,bh•sh/bto•
>> 1. For the rangeof tidalprismsexamined,
Fig. 8 displays
Lfhis•/L•o,•
(averaged
along-channel)as predictedby lineartheory. This extensionof lineartheoryto highly
non-linearsolutionsis accurateto lowestorderonly. Nonetheless, resultsare generally

C T Friedrichs,
consistent
withthefullynon-linear
solutions
tobediscussed.
Aslineartheorysuggests,
small
prismswithL•s, > Lj•o•areassociated
withflood-dominant
mainchannels,
andlargeprisms
withL•s, < Lj•o•areassociated
withebb-dominant
mainchannels.
maximumwidth, bmax,adjustable
intertidal
ai'½a
L=7km
constant
localwidthof embayment
= bmaxsin (n:x/L )
channelgeometry,
functionof prism
•'o= 1 m
constant
Figure
6:Schematicdiagram ofmodeltidalembayment. Tidalbasin
areaisspecified
bya sinefunction.
Withtheforcing
tidalamplitudeandembayment length
fixed,thechoice
ofmaximum embaymentwidth
determines
thespringtidalprism,
which,in turn,empirically
determines
thechannel
geometry.

Spring
prism
(m3) Cross-section Geomewi½
parameters
bhigh
= 150m
10• _ 106 h =2m htow

>> brow
T
bhig
h= 600m
106_ 107
n,••
....'• T• •' = 3m
hhig
hbhig
h
hlow= blow
bhigh
=3000
m "-I
hhigh bhigh
107-1 htow
<<brow
Figure7: Schematic
diagramof the relationships
(with •o andL fixed) betweenspringtidal prism,
embaymentcross-sectional
geometryandthe non-dimensional
geometricparameterswhichgoverntidal
asymmetry.
Severalobservationalstudiesof tidal embayments

indicatedeepmain channelsand large
intertidal flats are associated with ebb-dominance, whereas shallow channels and small
intertidalstorageare associated
with flood-dominance
(e.g., Boonand Byrne, 1981;Aubrey
and Speer, 1985; Friedrichsand Aubrey, 1988). Friedrichsand Aubrey examinedtidal
elevations from54 gaugesin 26 embayments. In systemscharacterized
by largechangesin
totalwidthduringthetidalcycle(bhigh
>> brow,,
hnign
• hto•,),
theyobserved
ebbduration
to
decreaseas V,/V, (volume storedin intertidalareas/volumein channels)increased. In
systems
withlargechanges
in meandepthduring_the
tidalcycle(bn•gn
-- brow,,
hn•gn
>> hto•,),
they observedflood durationto decrease
as •/h increased.Solutionsto the non-linear
zero-inertiaequationare in agreementwith thesefield observations (Fig. 9). Velocity
asymmetryis parameterized by the ratio of maximumflood speedto maximumebb speed,
calculatedalongthe axis of the embayment.Underconditionsof well-mixedtidal flow, this
ratio capturesthe differencesin peak bottomstresses whichare importantto net near-bed
sedimenttransport.Non-linearsolutions to (15.12)indicatethatlow waterpropagates
fastest
(Lfn•gn
< L.6o•,)
in ebb-dominant
channels
(Fig. 10a),whereas
highwaterpropagates
fastest

C T Friedrichs,D R LynchandD G Aubrey 295
(Lf•,i½,
> L3o,,)
in flood-dominant
channels
(Fig.10b).
1.2
0.8
0.6
(b)
bhigh1213
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
•
hhighllOD
th'-'•o•t
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
, .........
====================================================================
.•...,,. • • • 10/9
h• b high]-2131h
Lfhig._._.• high
I
107 108Lftow b'•o•
I •hlow
!
spring
tidal
prism
(m3)
Figure8: Thenon-dimensional geometric

parameters
whichgovern linearandnon-linear
tidalbehavior
asa functionof springtidalprism,averaged
overthelengthof theembayment (with• andL fixed):
(a)LILy;
Co)b•s•/b,o.,,
h•,,s•/h,o.,
andLf•¾/Lfto.,.
In Co),
shading
separates
flood-dominant
(Lf•,,s•/Lfto.
, > 1)
andebb-dominant
(Lfn•s•/Lfto.,
< 1) domains.
Eq.(15.12)wassolvedimplicitlyusingCrank-Nicolson in time(At= 450s) and1-dfinite

elementsin space
(Ax= 250m). Solutions for(15.12)werecompared tosolutionsfor(15.2)
and(15.3),whichinclude
theaccelerationtermsinthemomentum equation (Fig.10c,d).
Eqs.
(15.2)and(15.3)weresolvedexplicitlywitha 1-dfinitedifference model(At= 15 s, Ax=
250 m, seeFig. 2). The zero-inertia
solution is virtuallyidentical
to the solution
which
includesthe acceleration
terms,exceptimmediately
aroundslackwaterwhenthe surface
gradientis zero.

296 C T Friedrichs,D R LynchandD G Aubrey
Maximum flood speed/maximum

ebb speed
IL -40.95.......
:,xlo, ebb-dominant
__
__•0.9L' '•
.:::::::iiiii::•!::::::i•ii::::::i•i::::i!•::iiii::ii•::i::•ii!!::i•::!::iii::i::•ii?•:i:•:::iiii!i!i:..::•.::•:!iiii!
...........................
9
0 1 2 3 4 5 6
distancealong channel(km)
Figure9: Results
of 30 numerical
solutions
to(15.11).Theratioof cross-sectionally
averagedmaximum
floodto maximum ebbspeedduringa meantidalcycleiscontouredasa function
of localposition
along
thechannelandtotalembayment springprism.If maximum floodspeed/maximum ebbspeed is greater
thanone, the channelis flood-dominant;if the ratio is lessthanone, the channelis ebb-dominant The
dashedlines indicatetwo modelsexaminedin more detail, resultsfrom which were usedto force other
models(seeFigs. 10, 12, 13).
15.4.2 Ebb-dominant inlets to flood-dominant main channels
Ebb-dominant velocitiesare oftenobserved in the immediatevicinityof inletsto othenvise

flood-dominantembayments (e.g.,MurrellsInlet,SC,Perryet al., 1978;New ChathamInlet,
MA, Gieseet al., 1989). Solutions to thenon-linear zero-inertia
equation areconsistent with
theseobservations (Fig. 9), indicatingsystemswith flood-dominant main channelscontain
ebb-dominantcross-sectionally averaged velocitiesin theimmediate vicinityof theinlet. In
embayments with h•s•/hto•
>> 1, BoonandByrne(1981)attributed localebb-dominant
velocities
at inletsto continuity
effects. In frictionally-dominated
systems, changes in
elevation
withintheinnerembayment maylag seasurface forcingby up to severalhours.
According
to BoonandByrne,if inletvelocityis in phasewith discharge
fromtheinner
embayment,
whileinletcross-sectional
areais in phasewiththeseafide,averageinletarea

will be smallerduringebb thanduringflood. Continuitythenforcesstrongerebb thanflood

velocities to occur at the inlet.
(a) 0.4 (c)
0 0
,• -0.4
0 10 20
L• 0 l0 20
0.3
-0.3 , ,
0 !0 20 o lO •o
time (hours) time (hours)
Figure 10: Time seriesof tidal elevationsat the forcingend and 5500 m up the tidal channel:(a)
ebb-dominant mainchannel;(b) flood-dominant mainchannel(seeFig. 9). Tingeseriesof tidalvelocities
producedby thezero-inertiaequationanda geometrically identicalmodelwhichincludestheacceleration
terms,5500 m up thetidalchannel:(c) ebb-dominant mainchannel;(d) flood-dominant mainchannel.
It may be inappropriate, however,to consider

velocitiesat the inlet to be forcedby the
phaseof velocities
withintheembayment. Advectiveeffectsarenegligible in thesesystems,
sovelocities
at theinletcannotbe viewedasforcedby velocities elsewhere.An explanation
moreappropriate to frictionally-dominated
embayments resultsfromconsideration of set-up.
In systems
withL•ss> L•o,•,low tidedecays
morequicklythanhightide,i.e., differences
betweenoceanand embayment elevationsare significantly
largernearlow waterthannear
highwater(Fig.10a). Near theinleta muchsteepersurfacegradientoccursnearlow water
thannearhighwater. By solving(15.10) for velocity,

(15.23
we seethata muchsteepersurfacegradientnearlow waterleadsto highestvelocitiesat the
inlet duringtheebb(assumingtheeffectof theincreasein surfacegradientis largerthanthe
effect of the accompanying
decreasein hydraulicradius).
15.4.3Flood-dominantintertidalareasadjacentto ebb-dominantmainchannels
Observations of intertidalembaymentsindicatethat tidal channelswhich draincompletelyat

low tide are flood-dominant(Lincoln and FitzGerald, 1988; Speeret al., 1991). Intertidal
areasborderingrelativelydeepebb-dominant tidal channelshave also beenobservedto be
flooddominant.Swift andBrown(1983) presented harmonicanalysesof theGreatBay, NH,
that indicate a landward transition from ebb-dominance in the main channel to
flood-dominance
in the marginalshallowsand intertidalareas(Table 5). The Great Bay
Estuaryhas• = 1.3 m, meanchanneldepthson the orderof 7 m, and extensiveareasof
intertidalflats (Fig. 11). A conceptualexplanationfor the landwardtransitionfrom ebb- to
flood-dominanceobservedat Great Bay followsfrom a consideration of spatialvariationsin
thefrictional
lengthscale,Lein twodimensions.
In 2-d,(15.12),(15.13)and(15.20)become
O.(DO½}-
0½
c9t
. 0' (15.24)
L!,,,H• {••ts
(nto)-m, (15.26)
whereH(x,y) is totallocaldepth. Over all but theinnermost

portionof theGreatBay system,
deepchannels arefoundadjacentto largeintertidalareas.Thusaveragedepthsovermostof
the systemare shallowest aroundhighwater,andhightidepropagates moreslowlythanlow
tide(L/his•
< L•6o,,,).
Overtheinnermost
marginal shallows,however,
average
depths
aremuch
shalloweraroundlowwater(Lj•his•
> L/to,,,).
Increased
frictional
losses
aroundlowwaterslow
the propagation
of low tide, and the tide becomeslocallyflood-dominant.

T-16 • ,
T-14A
C-UNH
T-UNH
South Ehot
K•ttery
T-13
T-12
T. 11
Seavey
T-19
Stratham
km
Figure11: Locationmapof the GreatBay Estuary,New Hampshire,

indicating
locations
of tidal
elevationstations(*) andcurrentmoorings
(-). (After Swift andBrown,1983).
To investigatethe physicswhichrestfitin flood-dominance overmarginalshallows, we

solved(15.24)for a linearlyslopingintertidalflat, risingfrom onemeterbelowmeantide
levelat theforcedboundaryto onemeterabovemeantidelevelat thelandwardend,overa
totalhorizontaldistanceof 2500m. We performed twoexperiments, eachusingAt = 90 s and
Ax = 62.5 m, andeachforcedwithdifferenttime-series takenfromprevious1-dsolutions (see
Fig. 10). The first forcingtime-series had a shorterdurationfloodand the secondhada
shorterdurationebb. Thefirstexperiment resultedin flood-dominant currents
acrosstheentire
widthof the intertidalflat. In the secondcasecurrentsadjacentto the distortedforcingwere
ebb-dominant, but half way acrossthe flat the velocity-dominance reversedandthecurrents
becameflood-dominant (Fig. 12). Acrosstheintertidalflat averagedepthsaremuchshallower

300 (2 T Friedrichs,D R Lynch andD G Aubrey
around
lowwater(L•ish
> L•ow),
andhighwater
propagates
morequickly
thanlowwater.The
fasterpropagation of highwaterfurtherenhances flood-dominance whenthetideat theedge
of theflat is forcedby a shorterdurationflood(Fig. 13a,c). If theboundaryis forcedby a
shorterdurationebb,however,the fasterpropagation of highwaterfirst mustovercome the
pre-existing shorterebbbeforeeventuallyproducing a shorterfloodandflood-dominance
fartheroutacrossthe flat (Fig. 13b,d).The sharppeakin floodvelocitywhichdevelops with
distanceoverthe intertidalflat is consistent with observationsof sedimentresuspension
at the
GreatBayEstuary,NH. Anderson (1973)indicated thatfloodtidesediment resuspensionis
at at leasttwice that of the ebb over flats at the Great Bay.
Table5: Characteristics of theMe andMs tideobserved at tide gaugesandcurrentmetersthroughout

theGreatBay, NH tidalembayment (fromSwift andBrown,1983). Velocitydataarein bolditalics.
A shorterdurationebb,suggesting probableebb-dominance, is described
by a sea-surface
relativephase
rangingfrom 180ø-360 ø. Neglecting residualcurrents,
ebb-dominance is indicated
moredirectlyby a
velocityrelativephaseof 90%270 ø. Theremaining rangesof relativephasecharacterize
flood-dominance
(FriedrichsandAubrey,1988).
Gauge
ormeter distance M2 M2phase M4/M2 relative dominance
(km) (morms-l) (dog) phase(dog)
1. T-5 4.2 1.29 106 0.016 238 ebb

2. C 104 4.2 0.47 248 0.085 107 ebb
3. Seavey 5.5 1.20 114 0.017 207 ebb
4. T-11 7.3 1.12 117 0.026 243 ebb
5. C 119 7.8 0.70 248 0.029 129 ebb
6. T-12 9.3 1.00 128 0.030 231 ebb
7. T-13 10.7 0.95 133 0.042 235 ebb
8. T-14A 12.5 0.93 140 0.043 258 ebb
9. T-14 13.3 0.94 145 0.032 213 ebb
I0. C 124 13.3 1.48 262 0.027 222 ebb
11. T-16 15.0 0.83 166 0.012 156 flood
12. C 131 19.2 0.31 261 0.097 333 flood
13. T-UNH 20.0 0.87 171 0.034 260 ebb
14. T-19 26.0 0.92 176 0.033 103 flood
15.4.4 Inlets with distinct flood- and ebb-dominant sections
In a seriesof articles,Kjerfve andothers(Kjerfve, 1978; Kjerfve andProehl, 1979;Kjerfve

et al., 1981)discussedvariabilityof velocity-dominanceacross cross-sections
of tidalchannels
at NorthInlet, $C. NorthInlet is a well-mixedtidalembayment with • = 0.7 m, h = 2-3 m
andlittle freshwaterinflow (NummedalandHumphries,1978;FriedrichsandAubrey,1988).
Kjerfve and othersfound velocity-dominance to be related to a characteristicbimodal
distributionof cross-sectional bathymetrynear the tidal inlet. High velocity, strongly
ebb-dominant flow wastypicallyobserved in thedeeperof two sub-parallel
channels, whereas

C T Friedrichs,
theshallower
channeltypically
contained
lowervelocity,
weaklyflood-dominant flow. These
observations
areconsistent
withqualitative
descriptions
of tidalinletflood-andebb-channels
by Hayes(1975;1980). KjerfveandProehlspeculated
aboutthe observed
velocity
distribution,
including
theinfluence
ofchannelcurvature,
theshapeoftheflood-tide
delta,and
thenearby junction
of twotidalcreeks.
Fromtheconsideration
of thezero-inertia
equation,
however,it appears
thatthevelocity
pattern
observed
atNorthInletmaybea directresultof
thedistribution
of cross-sectional
bathymetry.
b-do nant
0 i I i I l
0 500 1000 1500 2000 2500
distanceacrossintertidalfiats (m)
Figure
12: Theratioof depth-averaged
maximumfloodtomaximum ebbspeedduring
a mean tidal
cycle,
plotted
asa function
ofdistance
across
anintertidal
flat. Theforcing
elevations
wereproduced
bythetwomodels highlighted
inFig.9,recorded
6500mintothetidalchannels. Thesolid(dashed)
lineis forced
bya timeseries
witha shorter
duration
flood(ebb).
If flowthrough
a channel
cross-section
is approximately
straight
andparallel,
it is
reasonable
to assume
sea-surface
elevation
andgradient
arerelatively
constant
across
that
givenchannel
cross-section
at anyonetime(e.g.,Henderson,
1966).Therefore,
by the
two-dimensional
equivalent
to (15.23)
I•[ - //2•[•½[m, (15.27)
depth-averaged
velocities
willconsistently
besmallerovertheshallower
portions
of the
channel
cross-section.
Furthermore,
thedifference
invelocity
between
thedeepandshallow

3O2 C T Friedrichs,D R Lynch andD G Aubrey
channelswill be enhancedat lower water levels. Since elevationchangeswithin the

embayment
atNorthInletlagbehind
ocean
forcing
(Nummedal
andHumphdes,
1978),depths
neartheinletareless,on average,
duringtheebbthanduringtheflood. Because
h,bdh•oo,t
<
1, frictional
dragis significantly
greater
in theshallow
section
duringthefloodthanduring
the ebb. Thereforeebb-currentsin the shallowsectionare muchweakerthan ebb-currentsin
thedeepsection,whereasflood-currents
in theshallow
section
areonlyslightly
weakerthan
flood-currents
in the deepsection.It followsthatthe deepandshallowportionsof the
cross-section
shouldbe ebb- and flood-dominant,
respectively,
if the total cross-sectionally
integrated
flowatNorthInletwereneither ebb-norflood-dominant. Considered
asa whole,
however,theembayment atNorthInletisa typicalebb-dominant
system withmoderately
deep
channels(relativeto tidalamplitude)andlargeareasof intertidalstorage(Nummexlal
and
Humphdes,1978;Friedrichsand Aubrey,1988).Thusthe deeperchannelis strongly
ebb-dominant,
whereasthe shallowchannelis only weaklyflood-dominant.
!
(a) 0.2
(c)
........
o 0
-0.2
• 0 1o 20 • 0 !o 20
.....
' 0.15 (d)
o 0
-0.15
-1
0 10 10 0 10 20
time(hours) time(hours)
Figure13: Timeseries
of anundistorted
sinecurve,tidalelevations
usedforboundary
forcing,andtidal
elevationsrecordedat 500 m intervalsacrossthe intertidalflat: (a) forcedby a time serieswith a shorter
duration
ebb;(b) forcedby a timeserieswitha shorterduration
flood(seeFig.9). Timeseriesof tidal
velocitiesrecordedat 500 m intervalsacrossthe intertidalflat: (c) forcedby a time serieswith a shorter
durationebb;(d) forcedby a time serieswith a shorterdurationflood.

UnclesandKjerfve (1986) attemptedto reproducetheresidualvelocitydistributionat North

Inlet with a 1-d numericalmodelthatincludedthe cross-sectional distribution
of bathymetry.
The model,whichdiscretizedthe cross-section into twelveelements,neglectedadvectionbut
includedtheeffectsof localacceleration.The modelwasforcedby threeperiodictime series
of axial surfacegradientandelevation,eachbasedon the M4 andM2 components of a single
tidalcycleobservedby KjerfveandProehl. UnclesandKjerfveobservedthatdepth-averaged
velocityfrom the modelwascloselyrelatedto topography andwaslargerin the deepwater
wheretheeffectof frictionwassmaller.Their modelpartiallyreproduced thedepth-dependent
distributionof residualvelocityobservedin the field. Howeverthe field observations were
generallymoredepthdependent thanthe modelresults. UnclesandKjerfve speculated that
theirerrorwas due to advectioneffectsneglectedby their numericalformulation.However,
we have arguedthat advectionis negligiblein shallow,relativelystraightchannelswithin
well-mixed embaymentssuch as North Inlet. Their error insteadmay be due to an
underestimate of the dragcoefficientappropriateto NorthInlet. UnclesandKjerfveapplied
an open-shelfvalue of ca = 0.0025, even thoughFinley (1976) had alreadypublished
observations from North Inlet that indicatedca= 0.01.
To investigatethe physicswhich result in a flood-dominantshallow sectionand an

ebb-dominant deepsectionwithina singleinlet,we solved(15.24) for a one-kilometer
length
of channelcomposed of distinctshallowanddeepsections.We performedtwo experiments,
eachusingAt = 450 s anda two-dimensional f'miteelementgridcomposed of approximately
1000elements(Fig. 14). In eachcasethe seawardendof the channelwasforcedby a sine
curvewith• = 1 m, andthelandward
endwasforcedby time-series
recorded
1 km intothe
previouslysolved 1-d channels. The mean depthand width of the 2-d channelswere also
identicalto thegeometryof thefirstkilometerof the 1-dchannels
whichproduced theforcing
time-series.The channeldepthsin thetwo casesdifferonlyby a constant
scalingfactor. The
first modelis narrowerand shallowerand was forcedat the landwardend by a time-series
witha shorterdurationflood. Thesecond modelhasa greaterwidthandmeandepthandwas
forcedat thelandwardendby a time-series with a shorterdurationebb. The firstexperiment
resultedin stronglyebb-dominant currentsin the deepsectionof the channeland strongly
flood-dominant currentsin the shallowsection.In the secondcasethe deepersectionwas
stronglyebb-dominant, and the shallowersectionwas weakly ebb-dominant (Fig. 14).
Consistentwith the conceptualdiscussion, the divisionof a singleinlet into ebb- and
flood-dominant
sectionsis enhanced as hhisn/h•ow
is increased.One can alsoexpect
enhancement
of thiseffectasL/Lsisincreased
becauseh,bJh• willdecrease
asthetidalwave
becomesmoreprogressive.

304 C T Friedrichs,D R LynchandD G Aubrey

(a) Finiteelementgrid
Mean depth= 2.1 m

Mean depth= 4.5 m
depth(m) depth(m)
2.5 4
3 5
6
3.5
3.5
6
3 5
2.5
--1.5 -- 2 3 4
, ! 2 •
0 2• 4• 600 8• I000 20O 4OO 600 8OO lOOO
• ' ß •o•v-:-o'•-•
....:.:.:.:::::::!:i:i:i:i:i:i:!:i::.:.:..i!.i•:i:.::::::i:i:i:i:i:i:•:•:i:•:$•i!ii•i•iiiiiiiii•
20 ..........
:::::::::::::::::::::::::::::::::::
'"'"'"'"' ":':':':':':'""i
........ ...::!'•iii
0 •:i-i-:-:•i-_:-i:i:i:i:i:i::_:•:•:..':?:v:'•!•_._..:
.......................
;.•.;
...............
.•.•.•.•
.........
.?:
......_._
...... •
0 200 400 600 800 ß 1000 2OO 460 6OO 800 1000
distancealong channel(m)
Figure14: Two-dimensional
finiteelementmodelsof tidalchannels
with distinctdeepandshallow
sections.The modelsareforcedby a sinecurveat oneendandelevationsrecordedfrom the two models
highlighted
inFig.9 attheotherend,a distance
of onekilometer
upstream: (a)Thefiniteelement
grid
usedforbothmodels;(b),(c) thebathymetry
of thetwochannels;(d), (e) theratioof depth-averaged
maximumfloodto maximumebbspeedin thetwo channels.

15.5Implicationsfor sedimenttransportandmorphologicevolution
From the precedingexaminationof velocity asymmetriesproducedby the zero-inertia

equation,systematicpatternsare identified which reinforce the underlyingembayment
morphology: Shallow and deep areasenhanceflood- and ebb-dominance, respectively.
Considering
theembayment
asa whole,shallow
systems
(hhis,
> > hlow,
b•,is•,
•- blow)
arelargely
flood-dominant,exceptin theft deepestportionsnear the tidal inlet, which are primarily
ebb-dominant.
Deepsystems
(h•,is•,
-- hlow,
b•,is•,
> > b•ow)
arelargelyebb-dominant,
exceptin
theftshallowest portionsoverintertidalflats,whichareprimarilyflood-dominant.Also, sharp
changes in morphologyareassociated with reversalsin velocitydominance.Acrossthe width
of tidal inlets, a local changefrom ebb- to flood-dominance is typically associatedwith a
markeddecreasein channeldepth. Similarly, the transitionfrom ebb- to flood-dominance
observedin the inner embaymentis associatedwith a morphologicbreak from deep tidal
channel to shallow intertidal flat. The velocity patternsreinforce the morphologies:
flood-dominanceenhanceslandward near-bed transport,channel shoaling, and furlher
reductions in depth;ebb-dominance enhances seawardnear-bedtransport,channelerosion,and
furtherincreasesin depth.
Severalauthorshaveseparated
tidalembayment
morphologies
intotwocategories.
Dronkers
(1986b)discussed two fundamental typesof basins:
Type 1 basinshavedeepchannels (• /
h << 1) andhigh ti_dalflat areas,on the averageabovemeansealevel.Type 2 basinshave
shallowchannels(h at mosta few timeslargerthan•) andlow tidalflat areas,whichat high
waterbecomepartof the streamcrosssection.Dronkers'Type 1 and2 systemsareequivalent
to typicalebb- and flood-dominant geometries discussedby others(e.g., Boon and Byrne,
1981; Speerand Aubrey, 1985; Friedrichsand Aubrey, 1988). Boon and Byrne, (1981)
suggested thatin a systemwithflood-dominantmainchannels andlackingextensivetidalflats,
high velocityfloodscouldfill an initially deepbasinto a leveljust floodedat high water,
increasethe areaof flats,and eventuallyproducean ebb-dominantmainchannel. But results
from SpeerandAubrey(1985) andFriedrichsandAubrey(1988) indicatedflood-dominance
of the mainchannelsdoesnot leadto eventualebb-dominance. Net bed-loadtransportmore
likely will causeshallow,flood-dominantchannelsto shoal,whereasunder ebb-dominant
conditions,deeperchannelsmay be maintained.
It can be misleading, however, to label entire embaymentsas ebb-dominantor

flood-dominant.As this papershows,differentportionsof a singleembaymenttypically
containdistinctpatternsof velocitydistortion. For example,local flood-dominance over
peripheralshallowsand intertidalareasmay occurat the edgesof otherwiseebb-dominant
systems.Duringthe evolutionof thesesystems, a morphologictransitionmay be enhanced
between the ebb-dominantcentral portionsof the embaymentand the flood-dominant
peripheralareas. Sedimentwill tendto be trappedalongtheedgesof theembayment dueto
highervelocityfloodcurrents,
whereas deepchannels will bemaintainedin thecentralportions

of theembayment by highervelocityebbflow. A relativelyabruptlateraltransitionfromdeep

channelto intertidalfiat, as characteristically
seenin suchsystems,will tendto develop. In
contrast,netsedimenttransportwill be more-or-lesslandwardthroughoutthe innerembayment
of systemswith flood-dominantmain channels. There is less tendencyfor abrupt
discontinuities
to developbetweensubtidalchannelsand intertidalfiats in systemswith
flood-dominantmain channels.Thus, dependingon the velocitydistortionof the main tidal
channel,distinctembaymentgeometries may conceivablyevolvealongthe linesof Dronkers'
(1986b) Type 1 and 2 basins.
Velocityasymmetries producedby thezero-inertiaequationalsoprovidedetailsof potential

near-bedtransportpathswithin the tidal channelcross-section.Across-channel patternsof
velocity distortionreinforce bathymetryand are consistentwith existing observationsof
morphology at tidalinlets(Hayes,1975;1980). Hayesobserved thatseawardtransportoccurs
in main ebb channels,whereaslandwardtransportoccursin marginalflood channels.The
divisionof the channelcross-section into ebb- and flood-dominantsectionsalsoprovidesa
mechanism for bedloadtransportin thedirectionoppositeto thatof cross-sectionally
averaged
velocity-dominance. For example,shallowsystems havingflood-dominant mainchannels tend
to have cross-sectionally averagedcurrentsnear the tidal inlet which are ebb-dominant.
Nonetheless, net landwardbedloadtransportcan still occurthroughthe tidal inletsof such
systems. If littoral transportsuppliesabundantsedimentto the tidal inlet, more sediment
transportmay occurlandwardalongthe marginalflood channelsthan seawardthroughthe
main ebb channel.
This paper utilizes empirical relationsbetween maximum spring tidal prism (or,
interchangeably,
maximumspringtidal discharge) and channelcross-sectional
geometryand
discussesthefeedbackbetweenflow andmorphology.For example,hydraulicengineers have
longrecognized the rapidfilling of tidal inletswhichoftenfollowsenlargementof inletsby
dredgingor reductionof prismby engineering works. Channelmorphodynamics respond on
much shortertime scalesthan changesin overallbasingeometrydue to sedimentation or
changesin relativesealevel. Gardnerand Bohn(1980) presented geologicevidencefrom
Holocenetidalchannels whichsupports thiscausalinterpretation.
Thuspre-existing
geologic
controlmay dictatedominancetype: Smallerinlet/baysystemsthat form with limitedtidal
prismswill tendto haveshallower channels,
whereas largersystems withrelictmorphology
that dictatesa largertidal prismwill have deeperchannels.In basinsforcedby similar
offshoretidalamplitudes,initiallylargeor smalltidalprismswill favorebb-or flood-dominant
mainchannels, respectively.Alonga barrier-beach coast-line
backedby a low-lyingcoastal
plain, the occurrence
of flood- or ebb-dominantembayments speculatively
canbe relatedto
tidalrangealone.Hayes(1979)andFitzGerald(1988)relatedthespacing of tidalinletsalong
barrierbeachcoastlinesinverselyto thetidalrange.Thereforelowertidalamplitude maylead
to widerspacing of inlets(bhigh
>> bto,,,),
largerprisms,
deeperchannels
(h•,ign
• hto,,,),
and
ebb-dominantmainchannels. Highertidalamplitudemayleadtocloser
inletspacing
(b•i•=
bt,,,,),
smaller
prisms, shallowerchannels
(h• >> hto,,,),
andflood-dominant
mainchannels.

Theserelationships
are beingexploredas ongoingresearch.
15.6 Summary
1. Underconditions appropriate to manysmall(L <-20 km),well-mixed(•o / h > 4).1) tidal
embayments, friction is one-to-twoordersof magnitudelarger than accelerationin the
cross-sectionally
integrated momentum equation.Thustidalflow in thesesystems maybe
described
by a zero-inertia diffusivemodel.Thisrepresentationcapturesthemostimportant
non-linearities
which contributeto asymmetricpeak tidal velocities(and, therefore,to net
near-bedsedimenttransport), while greatlysimplifyingconceptual
interpretation.
2. Thelinearanalytic
solution
tothezero-inertia
equation
describes
anexponentially
decaying,
partially
progressive,
partially
standing
wavegoverned
bythefrictional
length
scale
L•, L/
represents
boththepropagation
speed(c = L/•0) andtherateof amplitude
decay(e-folding
length= L). Propagation
speed
decreases
andamplitude decayincreases
asdepthisreduced,
frictionis increased,or total embaymentwidth is increasedrelativeto the width of the
channel.Thezeroth-order decayanddelayof theM2 tidein fourexistingtidalembayments
areconsistent with behaviorpredicted
by thelinearsolutions.
3. Non-dimensional
parameters
important
to non-linear
behavior
areidentified
byconsidering
variations
in the magnitudeof Lj,duringthe tidalcycle. Sincepropagation speedis
proportional
toL/,thetidalwavepropagatesmostquicklyduring
thatportion
of thetidalcycle
havingthegreatestLe An extensionof linearresults
indicates
thattheratioof frictional
lengths
nearhighandlowwateris givenbyL•sdL.
ao,,--(h•,•sdhto,,)•ø'S(b•sdbto,,)
'2a. In four
existing
tidalembayments,
non-linear
behavior
asindicated
by M2andM4tidalcomponents
is consistent
with conceptual
predictions.
4. Non-linear
numerical solutionsto thezero-inertia
equation
recreate
patterns
of velocity
distortion
observed in thefield: (a) Largechangesin channel
depthduringthetidalcycle
(hn•sn/h•o•,
>> 1) areassociatedwithflood-dominantmainchannels.(b)Largechangesin total
widthduring
thetidalcycle(bn•db•o•,
>> 1)result
inebb-dominant
mainchannels.
(c)Systems
withflood-dominant mainchannels tendto haveebb-dominanttidalinlets.(d) Systems
with
ebb-dominant mainchannelstendtohaveflood-dominantshallow marginsandintertidal
areas.
(e) Tidalinletcross-sections
areoftencomposed of separate
deepebb-dominant andshallow
flood-dominant sections.
5. Bothanalytic
andnumerical
solutions
indicate
thatvelocityasymmetry
in thesesystems
reinforcesmorphologyand likely plays a centralrole in its creationand maintenance.
Shallowerareastendtobeflood-dominantandenhance landward
near-bed
transport;
deeper
areastend to be ebb-dominant and enhanceseawardnear-bedtransport.
Changesin
morphology are associated
with reversals
in velocity-dominance.
The velocitypatterns
reinforce
themorphologies:
flood-dominanceenhanceslandward
near-bed
transport,
channel

308 C T Friedrichs,
shoaling,
andfurtherreductionsin depth;ebb-dominance
enhances
seaward
near-bed
transport,
channelerosion,and furtherincreasesin depth.
6. Geologiccontrolson tidal prismmay ultimatelydictatevelocitydistortionpatternsand

morphologicevolutionof theseshallowembayments. Largetidalprismsareassociated with
deepchannels
andwideintertidal
areas(bs•sdbto,•
>> 1), whereassmalltidalprismsare
associated
withshallow
channels
andnarrowintertidal
areas(hs•sdhto,•
>> 1). Duringthe
evolutionof systems havinglargeprisms,themorphologictransition
betweentheebb-dominant
centralportionsof theembayment andtheflood-dominant peripheral
areaswill be enhanced.
In systemswithsmallprisms,netsediment Wansport
will bemore-or-lesslandward throughout
theinnerembayment, leavingintertidalandchannelareasmorphologically lessdistinct.
Acknowledgments
Specialthanksis extended
to Dr. Ole S. Madsenwhoprovidedmanyhelpfulcomments onthe
manuscript.Thisworkwassupported by theOfficeof NavalResearch throughtheAmerican
Societyfor EngineeringEducation;by theNationalScienceFoundation, undergrantnumber
CEE-8352226; by theWoodsHoleOceanographic Institution's
CoastalResearch Center;by
theNOAANationalSeaGrantCollege ProgramOffice,Department of Commerce, undergrant
numberNA86-AA-D-SG090,WHOI Sea GrantProjectR/B-82;and by the A.W. Mellon
Foundation.The U.S. Government is authorized to produceand distributereprintsfor
governmental purposesnotwithstanding anycopyright notation
thatmayappearhereon.
15.7 References
Airy, G.B., 1845:TidesandWaves. Encyclopedia

Metropolitan,
London,Article 192.
Anderson,
F.E., 1973:Observations
of somesedimentary
processes
actingon a tidal flat. Marine
Geology,14, 101-116.
Aubrey,D.G., 1986:Hydrodynamic
controlson sediment
transport
in well-mixed
baysandestuaries.
Physics
of ShallowEstuaries
andBays,J. vandeKreeke,Ed.,Springer-Verlag,
245-258.
Aubrey,D.G., andP.E. Speer,1985:A studyof non-linear

tidalpropagation
in shallowinlet/esmarine
systems.Part I: observations.Esmarine,CoastalandShelf Science,21, 185-205.
Boon,J.D.III, andRJ. Byrne,1981:Onbasinhypsometry

andthemorphodynamic
response
of coastal
inlet systems.MarineGeology,40, 27-48.

Byrne,R.J.,P. BullockandD.G. Tyler, 1975:Responsecharacteristics

of a tidal inlet: a casestudy.
EstuarineResearch,Volume II: GeologyandEngineering,L.E. Cronin,Ed., AcademicPress,201-216.
Chantler,A.G., 1974: The applicabilityof regime theory to tidal watercourses.Journalof Hydraulic

Research, 12, 181-191.
Dronkers,J., 1986a:Tide-inducedresidualtransportof free sediment.Physicsof ShallowEstuariesand

Bays,J. van de Kreeke,Ed., Springer-Verlag,228-244.
Dronkers,J., 1986b:Tidal asymmetryandestuarinemorphology.Netherlands

Journalof SeaResearch,
20, 117-131.
Finley,R.J., 1976:Hydraulicsanddynamicsof NorthInlet, SouthCarolina,1974-75.US Army Coastal

EngineeringResearchCenter,G.I.T.I. Report10, 188 pp.
FitzGerald, D.M., 1988: Shoreline erosional-depositional

processesassociatedwith tidal inlets.
Hydrodynamicsand Sediment Dynamics of Tidal Inlets, D.G. Aubrey and L. Weishat, Eds.,
Springer-Verlag,186-225.
FitzGerald,D.M., and D. Nummedal,1983: Responsecharacteristics

of an ebb-dominated
tidal inlet
channel.Journalof Sedimentary
Petrology,53, 833-845.
Friedrichs,C.T., andD.G. Aubrey,1988:Non-lineartidal distortionin shallowwell-mixedestuaries:a

Synthesis.Estuarine,Coastaland Shelf Science,27, 521-545.
Gardner,L.R., andM. Bohn,1980:Geomorphic andhydraulicevolutionof tidalcreekson a subsiding

beachridgeplain, North Inlet, S.C. Marine Geology,34, M91-M97.
Giese,B.S., andD.A. Jay,1989:Modellingtidalenergetics

of theColumbiaRiverEstuary.Estuarine,
Coastal and Shelf Science,29, 549-571.
Giese, G.S., D.G. Aubrey and J.T. Liu, 1989: Development,characteristics

and effectsof the New
ChathamHarborInlet. WoodsHole Oceanographic InstitutionTechnicalReport,WHOI-89-19,33 pp.
Hayes,M.O., 1975: Morphologyof sandaccumulations in estuaries.EstuarineResearch,Volume II:

GeologyandEngineering,L.E. Cronin,Ed., AcademicPress,3-22.
Hayes,M.O., 1979:Barrierislandmorphology
asa functionof tidalandwaveregime.BarrierIslands:
From the Gulf of St. Lawrenceto the Gulf of Mexico, S.P. Leatherman,Ed., AcademicPress,1-28.
Hayes,M.O., 1980:Generalmorphology
andsediment
patterns
in tidalinlets.Sedimentary
Geology,
26,
139-156.
Henderson,
F.M., 1966:OpenChannelFlow, Macmillan,522 pp.

Jarrett,J.T., 1976:Tidal prism-inletarearelationships.

US Army CoastalEngineering
Research
Center,
G.I.T.I. Report3, 55 pp.
Kjerfve, B., 1978:Bathymetryasan indicatorof netcirculationin well mixedestuaries.

Limnologyand
Oceanography, 23, 816-821.
Kjerfve, B., and J.A. ProeM, 1979: Velocity variabilityin a cross-section

of a well-mixedestuary.
Journal of Marine Research,37, 409418.
Kjerfve, B., L.H. Stevenson,J.A. Proehl,T.H. ChrzanowskiandW.M. Kitchens,1981:Estimationof

materialfluxesin an estuarinecrosssection:a criticalanalysisof spatialmeasurement
densityanderrors.
Limnologyand•nography, 26, 325-335.
Lamb, H., 1932:Hydrodynamics.CambridgeUniversityPress,738 pp.
LeBlond,P.H., 1978: On tidal propagation

in shallowrivers. Journalof Geophysical
Research,83,
4717-4721.
Lincoln, J.M., and D.M. FitzGerald, 1988: Tidal distortionsand flood dominance at five small tidal inlets
in southemMaine. Marine Geology,82, 133-148.
Meyer-Peter,E., and R. Muller, 1948: Formulasfor bedloadtransport.Proceedingsof the Second

Meeting,InternationalAssociationof HydraulicResearch.Stockholm,Sweden,39-64.
Mfinchow,A., and R.W. Garvine,1991. Non-linear,barotropictidesand boresin shallowestuaries.

Tellus, 43A, 246-256.
Nummedal,
D., andS.M. Humphries,
1978:Hydraulics
anddynamics
of NorthInlet,1975-76.US Army
CoastalEngineeringResearchCenter,G.I.T.I. Report16, 214 pp.
Park,M., 1985:Predictionof tidal hydraulics

andsediment
transport
pattemsin StonyBrookHarbor.
M.S. thesis,StateUniversityof New York, StonyBrookMarineEnvironmental
Science
Program,146
pP.
Parker,B.B., 1984:Frictionaleffectson the tidaldynamicsof a shallowestuary.Ph.D. thesis,Johns

HopkinsUniversity,292 pp.
Perry,F.C., W.C. Seabergh

andE.F. Lane, 1978:Improvements
for MurrellsInlet, SouthCarolina.US
Army EngineerWaterways ExperimentStation,TechnicalReportH-78-4, 339 pp.
Pingtee,R.D., andL. Maddock,1978:The M• tide in the EnglishChannelderivedfroma non-linear

numericalmodelof the M: tide. Deep-SeaResearch,25, 53-68.

Postma,H., 1961: Transportand accumulationof suspendedmatter in the Dutch Wadden Sea.

Netherlands Journal of Sea Research, 1, 148-190.
Postma,H., 1967: Sedimenttransportand sedimentation

in the marineenvironment.Estuaries,GJt.
Lauff, Ed., American Association for the Advancement of Science, 158-186.
Robinson,I.S., L. Warren and J.F. Longbottom,1983: Sea-levelfluctuationsin the Fleet, an English

Tidal Lagoon. Estuarine,Coastaland Shelf Science,16, 651-668.
Speer, P.E., 1984: Tidal distortionin shallowestuaries.Ph.D. thesis,Woods Hole Oceanographic

Institution- MassachusettsInstituteof TechnologyJointProgramin Oceanography,210 pp.
Speer,P.E., andD.G. Aubrey, 1985: A studyof non-lineartidal propagationin shallowinlet/estuarine

systems.Part H: theory. Estuarine,Coastaland Shelf Science,21,207-224.
Speer,P.E.,D.G. AubreyandC.T. Friedrichs,1991:Non-linearhydrodynamics

of shallowtidalinlet/•y
systems.Tidal Hydrodynamics,B.B. Parker,Ed., Wiley, in press.
Swift, M.R., and W.S. Brown, 1983: Distributionof bottomstressand tidal energyin a well-mixed
estuary. Estuarine,Coastaland Shelf Science,17, 297-317.
Uncles,R.J., andB. Kjerfve, 1986:Transversestructureof residualflow in North Inlet, SouthCarolina.

Estuaries, 9, 39-42.
van de Kreeke, J., 1967: Water level fluctuations and flow in tidal inlets. ASCE Journal of the
WaterwaysandHarborsDivision, 43, 97-106.
Wallis, S.G., andD.W. Knight, 1984:Calibrationstudiesconcerning a one-dimensional

numericaltidal
model with particularreferenceto resistancecoefficients. Estuarine,Coastaland Shelf Science,19,
541-562.
Weisman,R.N., G.P. Lennonand F.E. Schuepfer,1990: Resistance

coefficientin a tidal channel.
EstuafineandCoastalModeling,M.L. Spaulding,
Ed., AmericanSocietyof Civil Engineers,
123-131.
Westerink,J.J.,K.D. Stolzenbach
andJ.J.Connor,1989:Generalspectral
computationsof thenonlinear
shallowwatertidal interactions
withinthe Bightof Abaco. Journalof PhysicalOceanography,19,
1348-1371.
Appendix 1. Symbolsdefinedin text

A totalcross-sectional
areaof embayment,
includingintertidalflats
Ac cross-sectional
areaof momentum-transporting
tidal channel

advectiveaccelerationterm of momentumequation
b totalwidth of cross-section,
includingintertidalflats
width of momentum-transporting portionof tidal channel
widthsaveragedover embaymentlength
bdb embayment
surface
area(a•) at meanlowwaterdivided
by a• at meanfidelevel
c phasespeed
c,• bottomdragcoefficient
D non-linear diffusion coefficient
Mean valueover the ebb for quantity•
•bb
F fiction termof momentumequation
__flooaMean valueover the flood for quantity•
g accelerationof gravity
_H localtotaldepth
h depthat meantide level, averagedover areaof embayment
h depth,averagedacrossmomentum-transporting portionof channelcross-section
__high high tide valueof quantity_
I inertiaterm of momentumequation
__inlet valueof quantity__ at inlet
L systemlength
localaccelerationtermof momentumequation
frictionallengthscale
low tide valueof quantity•
Manning'scoefficient
perimeterof entireembaymentcross-section, includingtidal flats
perimeterof momentum-transporting tidal channel
volume transport
maximumspringdischarge
hydraulicradiusof momentum-transporting tidal channel
time
depth-averaged velocity
characteristic
scaleof depth-averaged
velocity
voiume stored in intertidal areas/volume in channels
horizontal coordinate
constants
determinedby least-squares linearregression
characteristic
variationin quantity•
numericaltime step
surface elevation relative to a vertical datum
amplitudeof sea-surface
forcing
tidal lag
"constant" diffusion coefficient
dependentvariablein least-squares
linearregression
2 L/Lf
angulartidal frequency

16
Effects of Sea-Level Rise on Muddy

Coastal Margins
R Kirby
Abstract
The cross-sectional
shape,width and heightof muddyshoresis relevantto a wide rangeof
engineeringissues,but little is knownof how thesevary in nature. It hasbeenappreciated
for manyyearsthatbothpredominantly accretionary andpredominantly erosionalcoastsoccur
in natureand more recentlyit was shown,for a large numberof mud fiats, that shorecross-
sectionalshapevariessystematically, at leastover a narrowbandof tidal range.
Four coastalembaymentsin UK were selectedfor measurements of their height/arm

relationships. The sites were chosenon the basis of their sedimenttype, tidal range,
depositionalor erosionaldominanceand the availabilityof surveydata. The areaof shoreat
sequentialheightswas measuredto permitarea/height,(hypsographic), curvesto be plotted.
The datawere non-dimensionalised to permit shoresof widely varyingsize to be compared
directly.
The studyrevealsanapparentsystematictrendin hypsographic

curveshapefrommicrotidal
throughto hypertidalregimes. This is complicatedby a secondfinding, that accretion-
dominatedshoresappearto trend towardsa high and convex-upwards cross-section,
whilst
erosion-dominatedshoresappearto trendtowardsa low and concave-upwards cross-section
on coastsof the sametidal range. This may reflectthe combinedeffectsof relativewave
exposureand sealevel rise.
Arisingfrom thesefindingsit is suggested that changingtidal range,for exampleby

constructinga tidal power barrage,will resultin cross-sectionalshape,width and height
changingtowardsa new equilibrium.Furthermore, risingsealevel,whichmightbe expected
to changethe balancebetweenaccretionanderosion,is alsolikely to give rise to different
changes.Furtherwork is required,butif thisconfirmsthatshorecross-sectional shapedoes
varysystematically,suchempiricalknowledge providesoneapproach to predicting
long-term
change.

314 R Kirby
16.1 Introduction
Our perceptionof muddycoastalmarginshasdevelopeda greatdeal in lessthan 100 years.

In the early dayssuchareaswereoftentermed"wastelands", a termcarryingthe implication
that they were of no value otherthan for reclamationand perhapswildfowling. Today we
recognisea greatvarietyof engineeringand environmental significances,
rangingfrom an
initial defenceagainstcoasterosionto outletsfor landdrainage,stabilityof cables,pipesand
othershore-attached structures,throughto major nutrientsourcesin the coastalzone,nursery
areasfor fish,feedingareasfor migratorybirdsandthehabitatfor invertebrate animalswhich
providethe foodsource of theselatter. In parallelwith our relativelyrecentappreciation
of
the value of muddycoastalmargins,our knowledgeof themremainslimited. Unlike sandy
coasts,whichhavebeenextensivelystudied,relativelylimitedinvestigations of thedynamics
and morphologyof muddy coastshave been undertaken. We are furtherrestrictedby a
generallack of good topographicsurveysof intertidalzonesand also by the absenceof
recognisedtechniquesto evaluatechangeand to predicthow suchcoastlinesmay evolvein
the future.
Of thosefew investigations
of muddycoastswhichareavailable,the majorityinvolvereal-
time monitoringof short-termchanges,over periodsrangingfrom hoursto a year or more.
Mathematical models which consider one or a few ebb and flood fides are available. In
contrast,limited consideration hasbeengivento longerterm evolution,eitherin the pastor
asit is likely to occurin thefuture. Mathematicalmodelsareespeciallyproblematical
in areas
wherelongtermpredictionis required,whereunder-,normallyor over-consolidated substrates
are presentand where mixed or layered substratesoccur. It is these long term, real
complexitieswhich need to be assessed so that consequences
of changesin waterlevel on
varioustimescalescan be predicted.
The relative elevationof the land and sea can changesuddenly,for exampledue to
engineering workssuchas the construction of largedamsfor tidal powergeneration,or due
to rapid earthmovements.Relativechangescan alsooccuron muchlongertimescales due
to sea level rise or earthmovements.Each of thesewill have major effectson a muddy
coastline,which may take many yearsto adjust,even if the changeitself occursrapidly.
Changes arelikely to affecttheheightandwidthof thecoastalzone,its cross-sectional
shape
and whetherit is predominantlyerosionalor alepositional.
Knowledgeof the ambientheightand shapeof muddycoast-linesis oftenpoor. These

zonesareat thejunctionbetweenlandor air-based andmarine-based surveytechniques.Other
thanincidentallyin the courseof hydrographic surveysfor navigational purposes,
muddy
intertidalzonesof a highlyunstableandnon-significantnaturein navigational
termstendnot
to be surveyed with anydegreeof precisionor regularity.Consequently,shoreheight,width
andshapetendsnot to be well known. Only air photographic surveysflownat appropriate
tidal times,from low level, with appropriate
equipment cangive comparable dataandsuch
surveysare neveravailableon an incidentalbasis.

R Kirby 315
A preliminaryevaluationof long term shorestabilityand shapeand a tentativestartin

developinga methodology for predicting long-termevolutionof muddycoastalzoneshasbeen
madeduringinitial feasibilitystudiesinto the effectsof buildinga tidalpowerbarragein the
SevernEstuary,UK. The resultsapplynotonlyto suchengineering worksbut alsoto natural
changesin sealevel, as well as thosein prospectdue to man-inducedchangesin climate.
Havinga foreknowledge of suchchangesis an importantaspectof manyengineering and
environmentalissuesandmust,in future,be an importantaspectof coastalzonemanagement.
16.2 Muddy Shore Evolution

Whilst intertidaltopography
is frequentlysurveyedincidentallyandnot necessarily with the
precisionrequiredfor a comprehensive plot of its shapeto be derived,the sedimentary
provenance of muddyshoresis muchmorepoorlystudiedandknown. It is well established
thatsomeareasarelong-termdepositional sites,whereasothersareerosional.Studyingshort-
termchangein muddyenvironments presentsseverelogisticproblems,whereasfor long-term
change,wheredirectreal-timemonitoringcannotbe undertaken, the problemsare evenmore
severe.Thereare, nevertheless,a widerangeof physical,biologicalandchemicalindicators
of long-termchange.
Evidencefor changemaycomefromold mapsandcharts,whichin somepartsof theworld

are availablefor the last severalhundredyearswith steadilyimprovingaccuracy. More
recentlythesehavebeensupplemented by air photography
and satelliteimagery. Normally,
andarisingfrom the scalesof changes, all suchdocuments
revealarealchangemorereadily
thanheightvariation.
Sincetheearly1950s
and1960s
scientific
investigations
ofmuddy
shores
experiencing
long
termaccretionhavebeenavailable(van Straaten,1950,Evans,1965etc.). In suchregimes
a coast-aligned sequence of mud and sandfiats occursacrossthe shore,whichis closeto a
long-termequilibriumwith the hydrodynamicconditionsshapingthe shore. Physical
propertiesof thesediment, suchassediment grainsizeandgeotechnical propertiesareadjusted
to the hydrodynamics of the overlyingwatermass. Similarly,biologicalcharacteristics,
both
floralandfaunal,arein someformof dynamicequilibrium.Saltmarshes riseby deposition
andaregenerallytransitional withthetidalfiat witha seaward
limit closeto MeanHighWater
Neaps(MHWN). In thesameway, hard-shelled, burrowinganimalsremainin positionat the
baseof their burrowson deathand are steadilyentombedby the risingtidal fiat surface.
Finally,tidalfiat sediments containa widerangeof anthropogenic chemicals,reflectingboth
deposition of sediment containing adsorbed contaminantsandincorporation by animals.The
sediments consequently reflect,andarein balancewith,thechemistry of theoverlyingwaters.
In contrastto this situation,tidal flats experiencinglong-termerosionpresentmany

differences.The grainsizeandgeotechnical properties
of the sedimentmay notreflectthe
presenthydrodynamic regime.The erosionmaybe detected in geological
andarchaeological

316 R Kirby
indicators, suchastheover-consolidated
natureof thedeposits,
mudcracks,saltmarshor tidal
fiat cliffs andotherindicators
of erosion,peatbeds,fossilforestsandby a widerangeof
archaeological evidence,(Kirby,1990,Kirby,in press,andSELRC,1990). In additionto
thesenaturalindicators,measurements againstman-madestructures, especiallyof linear
features crossing
theshoresuchaspipelines andjetties,maybe important,(Kirby,1988).
In manytidalmudflatareas,biologicalindicators,
in theformof scoured-out
examples of
burrowing orboringbivalves,suchasMyaarenaria,Scrobicularia
plana,andPholasdactylus,
indicatethat an areais experiencing
long-termerosion. Largescale,short-termepisodic
changes needtobedistinguishedfromlongerterm,slower, evolution,
butfrequently
thereare
a rangeof indicators
of thetimescale
of change.Exposure of suchbivalveshasbeenusedin
a numberof areasto indicatethe sensebut nottherateof long-termevolution,(Kirby, 1990,
Kirbyet al.,in press,
andBradley,1957).Frequentlytheprogressive
lowering
of themudflat
surfacewhichexposes thebivalvespermitsepifaunalspecies
suchasbarnacles
andalgaeto
colonisethe bivalve surfacesin bands. Band age decreases towardsthe sedimentsurface.
Measurements of bandthickness, reflectedin yearlyspaffall,providesa directmeasurement
of erosionrate, (Kirby, 1990).
Biological
indicators
maybe complemented
by chemical
evidence,
suchastheabsence
of
radionuclides
from surficialsediments
underlyingcontaminated waters,(Kirby and Parker,
1980)or theabsenceof otheranthropogenic
chemicals,(Little, 1989).
Despitethesefeaturespermittingareasof long-termerosionand depositionto be

distinguished,
thereis no systematic
studyof theirdistribution.Moreover,thereasons
for
differentareasbeingdominatedby long-termdepositionor erosionremainsobscure. No
previousattempthasbeenmadeto usethesesedimentary criteriato distinguish
accretionary
and erosionalshorelinesand use this as a basis for assessingwhether such areas have
characteristic
and differingshapesand heights.
16.3 Muddy ShoreMorphology

If shoreshapeandheightwerecompletelyrandom,theprospects of findingany underlying
similarityor control,whichmightin futureserveasa basisforprediction, wouldbe negligible.
However,shoremorphology, in generaltermsat least,is notcompletelyrandom.In the 1960s
and 70s Hayescarriedout an assessment of the morphologicalfeaturesof soft sedimentary
coastsaroundthe world. He recognised thatcertainfeaturesoccurredregularlyandwasable
to usethecriteriaestablished
by Davies(1964),Table 1, to demonstrate a link betweencoastal
morphology andtidal range,(Hayes,1975),Figure1.
Figure 1 showsthat, whereasfiver and tidal deltas,barrierislandsand inlets are the

dominantshorefeatureof microtidalcoasts,linear sandridges,tidal flats and salt marshes
dominatemacrotidalshores.The fact thatthislink betweenbroadhydrodynamic conditions,

R Kirby 317
specified
in the tidalrange,andcoastal
morphology
is so general
in naturegivessome
confidencethat thereis somelink betweenthe changingdominanceof wave and current
processes
andthecoastalmorphology theygiveriseto.
Table1' Meantidalrangedivisions,
followingDavies(1964)andothers
Name Mean Tidal Range

MTR (m)
Hypertidal > 6.0
(High 5.0 - 6.0

Macrotidal
(Low 4.0- 5.0
(High 3.0 - 4.0

Mesatidal
(Low 2.0- 3.0
Microtidal > 2.0
River Barrier Tidal Linear T•dot Salt'

Deltas Islands Deltas Sand Inlets Fiats Marshes
R•dges
, :
Figure1:Variation
in morphology
of depositional
shorelines
compared
to tidalrange(Hayes1975).

318 R Kirby
+2,0
West'Heep HTR=1,80m I.-'

(Low
Hesohda[) HHW;
Ofzumer
Ba[jeHTR
=2.50m .'"•HHW
+1.0 (High
Hesohda[) ." l
..... Heldoffer
BuchfHTR
=3,30m -' I J•HHW
(Low
(HTRmec•
M•crofld•[)
hd•t
r•nge
) ''
'''' J
! /
NSL-
.ø
, ..9MLW•/
, , , , , , ,
0.1 0,2 0.3 0/, 0.5 0.6 0.7 0.8 0.9 1.0 1.1
relafiveconsfanf-[eve[
area(A/AMHW )
Figure2: Relationship
between
shapeandtidalrangeestablished
for barrier-island
tidalfiatsby
Dieckmannet al (1987).
Table2: Areasof tidalfiatsaboveMTL (fromDieckmannet al.)
High Mesotidal 62.5%

Low Mesotidal 33.0%
Microtidal 21.0%
Microtidal
regimes
arelargelydominated
by waveswhereas,
astidalrangeincreases,
the
relativeimportanceof tidal currentsmustincrease.
In addition,a furtherandapparently systematic

characteristic
of muddyshoreswas
identifiedby Dieckmann et al. (1987). Dieckmannet al. measured
the area•eight
(hypsographic)
curvesfroma largenumber
of inlets(37)ranging
in sizefrom95-260km
2and
involving
a totalof4,430km
2. Thesurveys
used
werealsotoa veryhighprecision,
oftena
fewcentimetres
in heightresolution.
Theyusedmeasurements
of areaatvarious
heights
on
the shoreto determinetidal flat shape. From thesemeasurements
non-dimensionalised
hypsographic
curves
wereproduced.
Non-dimensionalising
theplotspermits
areas
ofdiffering
sizeto be directlycompared.For thenarrowspread
of tidalrangevariations,
(microtidalto
highmesotidal),alongtheDanish,German andDutchcoaststheyshowed thattheareaof tidal
mudflatabove
MeanTideLevel(MTL)increases
withincreasing
tidalrange,
whilst
theshape

R Kirby 319
of theareacurvechangesfrom concave-upwardsin the caseof microtidalto convex-upwards

in the caseof mesotidal,Figure2, Table 2.
Theverylargenumberof measurements by Dieckmann et al., theirhighresolution andthe

extensiveareatheycomefromprovidea cornerstone leadingto confidence thatthedifferences
they reportare real and typify the balancebetweenwave and currentforcesin this barrier
islandcoastof the North Sea. Dieckmannand Pattensky(1985) have shownthat the main
factorgoverning morphological shapeof thetidalfiatsis thetidalrange.Thisconfransearlier
workby Hayes(1979), whoassessed therelationshipbetweencoastalmorphology, tidalrange
and wave climate. The criteriaof Hayes(1979) placesthe GermanBight tidal fiats clearly
withinthe groupingof typicalbarrierisland-typetidalfiats,"tidallydominatedmixed-energy
coasts".The resultsproducedby Dieckmannet al. thusled to the hypothesis that tidal mud
fiat shapesfrom higheror lower rangeareasmight fit logicallyinto the spectrumof types
establishedfor theserelativelylow tidal rangecoasts.The findingsraisedthe possibilitythat
hypsography might be applicable,in an empiricalsense,to predictthe consequences of
changedtidal range.
Shoulda tidal powerbarragebe built in the SevernEstuary,meantidal rangeinsidethe

enclosedbasinwill changefrom thepresent8.55mto about4.5m. It remainsto be established
preciselywhattherevisedhydrodynamic conditions will be withintheenclosed basin,although
a numberof importantattributesare well defined,as is their impactupon the shore. To
providecomparativedataon what the full spectrumof tidal flat shapescouldpotentiallybe
in the post-closure regimebehindthe barrage,a carefulappraisalwas carriedout to locatea
smallnumberof tidal mud fiat localitiesof comparablemeantidal rangeto act as analogs.
Sitesrepresenting exposedandsheltered, narrowandwide,erosionalanddeposifional regimes
were chosen.In eachcasea pre-requisite was thata goodqualitysurveyof the coastalzone
in questionwasavailable.All fourchosensitesarepredominantly or entirelymuddy,although
in theWashtheclassical,shore-parallel mudandsandflat environments of an accrefingshore
are developed. The Severnand Medway representsource-limited estuaries,in which fme
sedimentinputs are negligibleor small at present. The Humber and Wash have more
abundantinputs,especiallyfrom the sea. All four havea fiver wall formingtheir landward
margin.The sedimentary regimeof theSevernEstuarytidalfiatshasbeenspecifiedby Kirby
(1988), thatof the Mealwayby Kirby (1990), of theWashby Evans(1965) andof the Humber
to someextentby Wilkinsonet al. (1973). Some34 sitesfrom the Humber(MTR 4.8m at
Immingham)and Medway(MTR 4.2m at Sheerness) estuariesandThe Wash(MTR 4.8m at
Hunstanton) coastalembayment werechosen.The HumberandWashrepresent enclosedand
exposeddepositionalareasrespectively,whilstthe Medway is an enclosederosionalsite. It
wasconsidered thatthesesiteswouldembracethefull rangeof possibilities for Severnestuary
tidal flatsat equilibriumpost-closure.For comparison 9 siteswithin the Severn(MTR 8.55m
at Cardiff, 9.4m at Avonmouth)were also measured,giving43 in total. Siteschosenhere
rangedfrom enclosed(Cardiff Bay) to very exposed(Stert/Berrow)from both sidesof the
estuaryandincludedlocalitieswhichare bothoutsideand insidethe proposed barrageline.
Unlike otherlocalities,goodqualitysurveyswerenotavailablefromall sitesin the Severnand
specificaerialsurveysof 7 siteshadto be commissioned.

320 R Kirby
Almostincidentally,but consciously,
the siteschosenin the four localitiesaroundthe UK
coast,whenaddedto theGermanwork,permittedalmosta completespectrum of hypsogmphic
curvesfrom microtidalthroughto hypertidalto be assembled
for the first time.
Figure 3: Locality map showingthe four coastal embaymentsin which measurements were
undertaken.Boundaries
of numbered
andnamedsub-areas measured
areindicated.Notescales
vary.

R Kirby 321
16.4 Methods
Inputdatafrom 2 sites,HumberandMedway,wereconventional hydrographic surveys,whilst

surveysfrom the WashandSevernweremainlyor entirelyaerialsurveys,Figure3. In each
casethesurveyswerefirstcontoured at thenarrowest contourintervalpossible(generally0.25
- 0.5m) at a commondatum. Areasat particularheightswerethendetermined by planimeter
out to the channelthalwegor a naturallyoccurringbreak of slopeoffshore. Areas were
incremented cumulativelyandnon-dimensionalisedto permitplottingof hypsographiccurves,
whichwere in everyrespectcomparable to thoseobtainedby Dieckmannet al., (1987).
Table 3: Total AreasSurveyed,Tidal Flat only
Area km 2
(Spurn Bight 42.0

Humber
(Brough 31.9
Wash 162.1
(Slede Ooze 1.4

Medway (Deadman'sIsland 1.0
(StangateCreek 0.9
(Stert and Betrow Flats 38.4

Severn (Cardiff Bay 1.9
(SevernTidal PowerGroup
Areas 1-3, 5-8 (0.3 - 2.3) 9.0
GermanBight (Dieckmannet al.) 4430.0

(95.0- 260.0)
The curvesmeasured
from the eightrepresentative
sub-areas
presented
herediffer in a
numberof respects.
16.5 Results
A totalof 43 sub-areas
werecalculated andbothnon-dimensionalanddimensionalplotswere
prepared.Sub-areas wereamalgamated suchthataslargean areaascouldbe justdiedhas
beencalculated in orderthatresultsareasrepresentative
aspossible.The43 sub-areaswere
reduced to 15. The 10 Humbersiteswereamalgamated into2 hypsographiccurvesfrom

322 R Kirby
continuous
areas,14 sitesin the Medway werecompiledinto 3 curvesfrom continuous
areas
and 10 sites in the Wash were consolidated into 1 curve. The 9 sites measured in the Severn
couldnot be consolidated
furtherbecauseall are from non-adjacent areasand all aredifferent
onefromtheother. Only 8 of thecurvesarereproduced here. In additionto individualplots,
all Severn Estuary curves are individually compiled into a single plot. In addition,
representative
curvesfrom this studyare compiledtogetherwith the earlierDieckmannet al
datato producea compositespanningtheentiretidalrangespectrum from 1.8 to 9.4m MTR.
The arealcoverageof the curvesvariesconsiderably from site to site,Table 3.
Areas 1-3 combined
MHW *6 55m
MTL -"4 1 m
OD(N)
.39m •
MLW ß1 75 m •
0m CD
-0.1 m LAT
MEAN TIDAL RANGE 4.8m IMMINGHAM
0 " 0•.2 I 0•.4 • 0.6i

Relative Constant Level Area
Figure4: Hypsographic
curvefor areas1-3 SpurnBight,HumberEstuary.

R Kirby 323
16.5.1 Humber:SpurnBight
Theentireintertidalzoneis a mudflat. Thehypsographic

curve,Figure4, is linearto slightly
convexupwards.The upperpartof thetidalflat doesnotappearto riseas highas6.0m
aboveChartDatum(CD), whereasMeanHigh Water(MHW) is at 6.55m. As a resultthe
tidal flats must be coveredfight to the top on every tide (MHWN = 5.8m at
Immingham).Calculation of therelativeconstant
levelareafightoutto themainchannel axis
hasmadeit appear thattheareaaboveCD isrelatively small.If onlya narrowzoneoffshore
hadbeenconsidered the curvewouldhavebeenvirtuallyindistinguishable from thatin the
Wash. Asplotted,theonlydifferences arethattheoffshore subtidalcomponent in theWash
is smaller,whilsttheuppertidalfiat risesaboveMHW. SpurnBightdoesnothavea salt
marshat itsupperedge,whereas theWashdoes.Possibly this,in part,explains
thedifferent
levelsof theuppershore.Alsothehighermudflat in theHumberis poorlysurveyed, in
contrastto theWash. The formerwassurveyed froma boatandthelatterfromthe air.
Areas 1-10
MHW .6.5m
MTL +3.8m
OD(N)
MLW +1.7m
3.7m below OD(N) LAT= CD -3.7m Tabs Head
MEAN TIDAL RANGE 4.8m HUNSTANTON

-10-
o o:2 o:n o'.• o'.s •

curvefor areas1-10 The Wash.
16.5.2 Wash
Only themarginaluppermudflat anduppersandflat regionhasbeenmeasured.The shore-

attachedfingerbankstowardsthe axishavebeenomitted. All 10 sub-areas
into whichthe

324 R Kirby
tidal flat was dividedfor convenience

of measurement
have a similarhypsographic
curve.
Arisingfromthisit wasjustified
tocondense the10sub-areas ontoa singlecomposite curve
(Figure5). TheWashdoesnothavea pronounced offshorechannel,withtheresultthatthe
offshorelimit chosenwas ratherarbitrary. When the subtidalportionof the curveis
consideredtogetherwiththe intertidal
portion,thecurvehasa pronounced convexshape.
Evenneglecting thesubtidal
portionthecurveis lineartoslightly
convex.Theshape is most
similartoSpurnBightin theHumber andCardiffBayin theSevern both,liketheWash,areas
of ongoinglong-termnet accretion.
Areas 1-3
MHW ,-5.25m
OD(N) -'3.75 m
MTL*3lm j
MLW .1 05m
f 0m 'CO
+0.1rn LAT
• _
MEAN TIDAL RANGE 4 2m SHEERNESS

_
-8-
_
-12-
_
-14-
0 ' 01• ' 0'.• ' oi• ' o'.8 '

curvefor areas1-3 SledeOoze,MedwayEstuary.
16.5.3 Medway: Slede Ooze
This tidal flat is entirely muddyother than for a narrow zone with shellbanks. The
hypsogmphic curve(Figure6) hasa pronounced sinuositywith a verybroadflat zonejust
below MTL. The tidal flat is at a low elevation,onlyjust reachingaboveMTL closeto the
coast. Thereis no uppershore,the tidal flat only risingto +3.80m (closeto the national
datum,Ordnance Datum(Newlyn),(OD{N})), andthemid-shore sectionis narrowandsteep.
Thecurveis mostcomparable withthatfromDeadman's Islandin theMedwayandthecurves
for sections1,2,3, 5 and 6 in the Severn.

325
R Kirby
2-
MTL +6 45m
OD(N] •
•MLW
+2
25m
6.3m below OD(N) LAT (CD)
MEAN TIDAL RANGE 8 55m CARDIFF
-10 • , , '
o o'.• o'.• o'.•
curvefor CardiffBay,SevernEstuary.
,• .2
•-4
-8
.10
-].2
-].4
0 ' 0:2 ' 0:4 ' 0:6 ' 0•,8 ' 1.0
Relahve Constant Level Area
curveforStertandBerrowFlats,Bridgwater
Bay,BristolChannel.

326 R Kirby
16.5.4 Severn:Cardiff Bay
The intertidalsedimentsare entirely muddy. The hypsographic curve has a distinctive

pronounced "double-convexity" (Figure7) bothupperandlowershoresareconvex. The bay
itself hasa high proportionof its areain tidal flats. More than65% of the wholeareaand
70% of the intertidalzone is aboveMTL, indicatinga generallyhigh level tidal flat. The
near-shorezone is formedby a salt marsh.
Area 3
6-
MHW ,10.8m
MTL ,6.45m
OD(N)
MLW ,•2.25 m
• • 5.8m below OD(N] CD Newport

-6-
MEAN TIDAL RANGE 8.55m CARDIFF

-8-
-10 • , ,
0 0'.2 014 0'.6 ' 0. '8 ' 1.0
curvefor Gold Cliff, SevernEstuary.
16.5.5 Severn: Stert and Berrow Flats
Theinnerintertidalzone,possibly10-15%of thearea,is a sandbeachor hasscattered sandy

veneers.Theremainder is mudflat. Thisis a geographically
largeareain BridgwaterBay,
whichis possibly themostexposed localityfor anymudflat in theestuary.Definingthe
subtidallimit of theshoreprofilepresenteda problemowingbothto inadequate surveydata
as well asthedifficultyof recognisinganypronounced breakof slopeoffshore.

R Kirby 327
The hypsographic
curve(Figure8) showsa pronounced
andcontinuous
concavityextending
to the subtidal limit. The curve extends above both MTL and MHW but the inner section is
narrowand in relativeterms,steep. Only 20% of theprofile is aboveMTL. There is no salt
marshat the top of the tidal flat for muchof the coastalsectionand the salt marshbetween
Steart village and Hinkley occupiesonly 2% of the total intertidal zone. The section
consequently providesa markedcontrastwith thehypsographic curvefor CardiffBay, despite
the latterbeinga similardistancedown-estuary on the oppositebank.
Area 6
6-
MHW ,11.6m
2-
MTL ,6.92m
ODCN)
MLW +2.2m
6.5m below ODIN) CD
-8-
-10-
MEAN TIDAL RANGE 9.4m AVONMOUTH

-12-
-14-
0 ! 12
0. [ 0.4 ! ! 0.6 o18
Figure10: Hypsograpic
curvefor Clevedon,SevernEstuary.

328 R Kirby
16.5.6 Severn: Gold Cliff
The intertidalzoneis entirelymuddy,otherthanfor smallareasof exposedpeat. The profile

(Figure9) hasa complexshapewithanoverallconcavity butwithconvex"shoulders" above
MTL andbelow Mean Low Water (MLW). Thereare two gentlyinclinedshoreplatforms,
a relativelynarrowupperonebetween+l.2m and+2.2mOD(N) anda broader,lowerlevel
platformstartingat -3.5mandextending intothesubtidalzone. Only 20% of thetotalcurve
is aboveMTL, although70% is intertidal.
1 ..... Cardiff Bay

2-- -- Kingston Seymour
_
3--.-- Sand Bay
- 4........
Stert
&Berrow
Flats 4.•."/'"
5
7
.... Gold
Rumney
6......
Clevedon
Cliff •
•
. •' .? •
2y.,•3,5,7-9
9 ..... Avonmouth •. • "•' 7•'• ./
/" / ,.' _.•-

ß"/ /' -•-:"//'• ..4
...• _ .- ,?•;..'4'//•..•"
...' / /
•
/.' / •%-•'%'"'
•- _..•.• ß MHW
-
,. •'..•" .
........
? Referenced
toMTL
/" .
. ... /
.'
0 • 0. •2 • 0,4' ' '
0.6 • 0•.8 • '0
1. 1.2
Figure11' Superimposed
Hypsographic
curvesfor all nineSevernEstuaryandBristolChannellocalities.
16.5.7 Severn: Clevedon
The shoreis muddy,otherthan for an extensivenarrowbandof exposedrock at mid-shore

level. The profile (Figure 10) has a similarshapeto Gold Cliff with an overallconcave-
upwardshapeand two slightconvexities.An uppershoreplatformis presentbetweenMHW
and MTL but in this casecould be, in part, an artefactarisingfrom rock outcropsin some
partsof this zone. The flatter, outerplatformpresentin many other SevernEstuarycurves
is alsopresentbut herelies just aboveMLW, between-2.0m and -4.2m OD(N).

R Kirby 329
16.5.8 Severn: all 9 curves
Two distinctivesuitesof curvescan be recognised(Figure 11). The largergroupof 6 show

closelycomparable concave-upward curves.Mosthavehighor low levelshoreplatforms,or
both,althoughtheseoccurat differentlevels. The othergroupof 3 showhigher,convex-
upwardshapes andapparently occurin zoneswhichareshallower offshore.Theyalsoextend
further acrossthe graph.
1 •;F Severn
2 CB Severn
--Cardiff
Stert
Flats
Bayt
Hypert•dal MTR 8.55m
II 3 The
4
Wash
t
5......
Humber - Spurn Bight
Medway- SledeOoze)
Low macrotidal MTR
t4'8m
•4.8m
4.2m 1
1
/
II MTL
6.........
7.....
8..... Meldorfer
Bucht
West
Meep •,•',•
mesohdal
MTR
Otzumer
M,c'rohdal
MTR
1.8m3.3m
BelleLow
• •
mesotldal
.•5
2.5m J 4
,.•"....'•''"•'
.........
1 ß• •.-• • *• • ....
/ x 1
!
/,/
{,• , , , ,
.e,eence,o.,
, I
o o.• o'.• o.•6 o."8 •'.o

Figure 12: SuperimposedselectedHypsographiccurvesrepresenting

hypertidal,low macrotidal,high
mesotidal,low mesotidalandmicrotidalregimesin the UK andGermanyclassifiedaccordingto Table
1.
16.5.9 Hypsographiccurvesfrom Micro- to Hypertidalregimes
It wasinitially postulatedthat if the trendsapparentin the Germansuiteof curvescontinued

logically,thenmacrotidal andhypertidalhypsographic curvescouldbe expectedto be higher
with respectto MTL, moreconvexin shapeandshiftedto the left on the graph,implyinga
greaterproportionof the whole curve was intertidal. The comparisonin Figure 12 is

330 R Kirby
complicated
by thefact thatwhilsttheUK datais referenced
to MeanTide Level (MTL) the
Germandatais referencedto MeanSeaLevel(MSL). A strictintercomparison
is notpossible
because MSL is not known for the UK curves.
The resultsshowthat in orderof ascending

tidal range,andprovidingthatappropriate
hypsographic curvesarechosen,theMedway(SledeOoze)MTR 4.2m,the Humber(Spurn
Bight)MTR 4.8m,theWashMTR 4.8mandtheSevern(CardiffBay)MTR 8.55mdo show
the expectedshift. However,superimposition
of the curvefor StertandBerrowFlatsin the
Severn,or indeedcertainotherMedwayor Severnmainchannelcurves,showsthat the full
relationship
is notquitesosimpleasthedeliberatelyselectedcurveswouldmakeit appear.
16.6 Interpretation
It hasbeenpointedoutthatmuddyshorelines havebeenpoorlysurveyed andonlyveryfew
determinations
of shoreshapeareavailable.Arisingfromthisanyfindings fromthisstudy
mustbetreatedwithcaution.Nevertheless themeasurements of shoreshapegainin reliability
fromthefactthatthe long-termsedimentary regimein the areasunderscrutinyis sowell
established.
TheHumber,especially in theSpurnBightregion,is knownto be experiencing
infillordepositionoverthelongerterm,Wilkinson et al (1973),asistheWash,Evans(1965).
Boththesesitesapparently havean adequate sediment inputandbotharegainingsediment
fromseaward aswell aslandward.In contrast, theMedway,at leastin respect of its outer,
main channelcoastline, has beenshownfrom a varietyof evidence,includingstake
measurements,archaeologicalandmapevidence, aswellasbiological evidencefromMyaand
otherbivalves,to be erosional.Thisswingto anerosional regimehasoccurred sometime in
thelast300 years,priorto whichtheMedwaytidalflatswereaccretionary, Kirby(1990).
The Severnis a muchlargersystemthattheMedwayor Humber,comparable with or

biggerthantheWashandis notevolvingidentically
throughout
itsentireintertidalzone.
Until 1905CardiffBaywasco-extensive withRumneyandWentlooge andnodoubthad

comparableshapedfiats. It wasenclosed
earlyin thecenturyto cremea largedockandhas
consequently
shiftedto becomea zoneof rapidlong-term depositionwithan abundantfine
sedimentsupply.This siteprovidesa markedcontrastwith thetidalflatsof themainchannel.
Theseareexperiencing rapiderosionoverthe longerterm.
Although ratesof shoreretreatoverthelongertermarenotpossible to quantify,

thetidal
flatsat Bridgwater
Baylie in themostexposed locality,whilstthoseat Avonmouth havebeen
keyed-into
a channel
notched
intorockforseveral
thousandyears,ata sitewheretidalrange
andhencecurrents
aremaximal.Thesetidalflat areasareinterpretedasbeingthemost
severelydegraded.

R Kirby 331
Tidal flatsat Rumney,Wentlooge,GoldCliff andClevedonareknownfromgeologicaland

archaeological evidenceto haveexperienced largescaleretreat(a verticalerosionof 3.0m in
70 yearsat Rumney),althoughthey are lessexposedthanBridgwaterBay and Avonmouth.
KingstonSeymourandSandBay representsiteshemmedin by headlands on oneor bothsides
and henceprotectedto somedegreefrom somecomponentsof the wave climate and the
strongest tidal currents.Furtherprotectionis providedby an offshoresandbank. SandBay
hasa sandybeachat the top of the shore,but this represents10% or lessof the total shore
area,the remainderbeingmuddy. KingstonSeymouris entirelymuddy.
This understanding of the long-termdynamicsof the sedimentary regimemay assistin

interpretingthe shapesof tidal fiats which result. Firstly, tidal fiat shapesat the Humber
(SpurnBight),Wash and Severn(Cardiff Bay) are all sitesof long-termaccretionand self-
similarin shape(Curves2, 3 and 4, Figure 12). All the curveshavea convexprofile and
thereis a possibletendencyfor a largerproportionto be shallowerthanMTR as tidal range
increases.The Wash hypsographic curve is generallyat a lower heightthan the Humber,
althoughbothhavethesameMTR. Possiblythisis accounted for by thefrequentreclamations
in the Wash in the past. If so and the Humberis in equilibrium,the tidal fiats in the Wash
have still to rise to reachan equivalentheight. The uppertidal fiat in the Wash is, however,
higherand hassalt marshes.
The tidal flatson the mainchannelof the Medwayare self-similarand mostcomparable

to thoseof themainchannelof theSevern.Theyhavea pronounced sinuosity
anda generafly
concaveupwardshape,oftenwith two smallconvexities,oneat mid andtheothertowardslow
tide level. In generalthe bulk of the curvelies at a low elevationcomparedto tidal fiatsin
the Humber,Wash and Cardiff Bay and the innertidal flat zoneis narrowand steep. The
low-level,pronounced concavityand sinuousshapemay all be manifestations of the known
predominantly long-termerosionalregimeof the main channelcoastsof the Medwayand
Severn.
KingstonSeymourandpossiblySandBay areknownsitesof long-termerosion,but have

an overallconvex-upward shape. If the hypothesis beingproposed hereis correcttheseare
erosionalbut the leastdegradedof any erosionalprofriesin the main channelcoastof the
Severn. A possibleexplanationfor this is theft confinedlocationbetweenheadlandsand in
thecaseof KingstonSeymourtheprotection affordedby theLangfordGroundsBankoffshore.
By selectingparticularhypsographic curvesthe suiteof curvesspanningthe range

microtidalthroughto hypertidal canbe builtup andseemsto followa logicalprogression.
Curvesdo risewith respectto mid-tidelevel,becomemoreconvexupwardsin shapeand
moveto the left on the graph,implyinga greaterproportionof the curveis intertidaland a
greater
proportion at highlevel. Theexplanationfortheshiftpresumablyrelatesto theprecise
interplayof tidal currentand wave forceson the tidal fiat and remainsto be confirmedwith
more dataandexploredin fundamentalterms.

332 R Kirby
16.7 Conclusions
Despitesomeminor lithologicalvariationsbetweenthe varioussites,threebroad,though

tentative,deductionscan be drawn from the analysescarriedout during the study,one
independent and two linked one to the other.
The first is thata sequence of hypsographiccurvesrepresenting a largeproportionof the

spectrumof tidal rangesencounteredin naturehas been assembledfor the first time. In
simplesttermstheseappearto form a suiteof curvesbeatinga logicalrelationone to the
other. If curvesfrom elsewherein the spectrumof tidal rangesfall within, or add to this
spectrumthenits validitywill be enhanced.Possiblya closelyrelatedspectrumof curvesdoes
occurfor tidal flat hypsographic
curvesat equilibrium?Shouldthisproveto be thecasesuch
knowledgecan be usedto predictthe revisedshapeof tidal flat profilesin circumstances
wherethe tidal rangechangesin eitherthe shortor longerterm.
Secondly,the understanding of the sedimentaryregime in the sites studiedand the

hypsographic curveswhichhavebeenmeasured suggestthepossibilitythatcurveshapemay
differsystematically
dependinguponwhethertidalfiatsareaccreting or erodingin thelonger
term. Equilibriumshapesfor accretionaryregimes,possiblyindicatedby the Humber,Wash
andCardiffBaycurves,appearconsistently convex-upwards, whilst,in contrast,
hypsographic
curvesfor deeplydenudedtidal flatsin the mainchannelcoastsof the MedwayandSevern
tendto beconcave-upwards,implyinga wideandlow-lyingoutershoreandsteepinnershore.
Possiblyconvexshapes evolvethroughvariousintermediate curvesintoconcave shapes when
the sedimentaryregimechangesfrom accretionary to erosional?
Linkedto thisconclusion it is evidentfromtheMedway,whereovera 300 yearperioda

long-termaccretionary regimehasswungto an erosional regimeandfromCardiffBay in the
Severn,wherean erosional regimehasswungto accretionary overa 90 yeartime-span,
that
tidal fiat shapechangesthroughhistorictime. In the Medway,a convex-upward tidalfiat
shapehaschanged to a concave-upward shape,whilstin CardiffBaythereversehasoccurred.
In neithercaseis theold surveydataadequate to producing accurate
hypsographiccurvesto
illustrate this feature.
Thelattertwoconclusions haveimplications for sealevelriseanda suggested narrowing

of thecoastlineandenhanced tendency for erosionof muddycoastswhichwill materialise,
(Goss-Custard et al. 1990),dueto theGreenhouse Effect. They alsohaveimplications for
tidal powerschemes, for the relativeproportionsof intertidalto subtidalarea,for coastal
stability,fish and bird feedingtimesand the durationof waterinundation requiredfor
invertebratesto reachmaturity.

R Kirby 333
Acknowledgements
The authorwishesto thankthe SevernTidal PowerGroupfor permissionto publishtheresults
of contractwork. Surveydatawassuppliedby BritishPortsAssociation(HumberandCardiff
Bay), Binnieand Partners(Wash),MeAlwayPortsAuthority(MeAlway).Surveysfor Severn
Estuarymain channelcoastswere undertakenby contractaerial surveys. The surveyof
BridgwaterBay wasavailablein scientificliterature.The earlystagesof the studybenefitted
from guidanceprovidedby Dr-Ing ManuelaOsterthunof the FranziusInstitutein Hanover,
Germany.
16.8 References
BradleyW.H., 1957 Physicalandecologicfeaturesof the Sagadahoc

Bay tidal flat, Georgetown,Maine.
Geol SocAmerica,Memoir 67, p.641-682.
Davies J.L., 1964 A morphogenicapproachto world shorelines.Z. Geomorphol,8, p.27-42.
DieckmannR. andPattenskyH-W., 1985 Quantitativegeomorphology

of tidal flat areas. Symposium
on CoastalMorphology,Sedimentary
Budgets,CoastalandRiverHydraulics.Reykjavik,Iceland,p.52-
62.
DieckmannR., OsterthunM. and PartenskyH-W., 1987 Influenceof water-levelelevationand tidal

rangeon the sedimentation
in a Germantidal flat area. Progressin Oceanography,
Vol 18, Nos 1-4,
p.151-166.
EvansG., 1965 Intertidalfiat sediments

andtheirenvironments
of depositionin theWash. QJGSLond,
Vol 121, p.209-245.
Goss-Custard J.D., McGrorty S and Kirby R., 1990 Inshorebirdsof the soft coastsand sea-levelrise.
In: Expectedeffectsof climaticchangeon marinecoastalecosystems,
Editors:BeukemaJ.J.,Wolff, W.J.
and Brouns,J.J.W.M. Kluwer AcademicPub., p.189-193.
HayesM.O., 1975 Morphologyof sandaccumulations in estuaries.In: CroninL.E. (Ed) Esmarine

Research,Vol 2, GeologyandEngineering.AcademicPress,NY, p. 3-22.
KirbyR. andParkerW.R., 1980 Settledmuddeposits

in Bridgwater
Bay,BristolChannel.IOS Report
No. 107, 65pp (unpublished).
Kirby R., (in press) The evolutionof the fine sedimentregimeof the SevernEstuaryand Bristol
Channel.BiologicalJournalof theLinneanSociety.Proc.Symp.onevolutionandchangein theBristol
Channeland SevernEstuary.

334 R Kirby
Kirby R., 1990 The sedimentbudgetof the erosionalintertidalzoneof the MedwayEstuary,Kent.

Proc.Geol. Assoc.,Vol. 101, Pt 1, p.63-77.
Kirby,R., 1988 The stabilityof intertidalmudfiatsandtheirrelationship

to longtermsources
andsinks
for fine sediment: with particularreference
to the SevernEstuaryandBristolChannel.Reportto STPG
ref tasks3.1(iii)b and 3.7(i)a, 103pp,(unpublished).
Kirby R., BleakleyRJ., WeatherupS.T.C.,RavenPJ. andDonaldson N.D., (in press)Effectof episodic

eventson tidal mud flat stability,ArdmillanBay, StrangfordLough,NorthernIreland. Workshopon
nearshoreandestuarinecohesivesedimenttransport.SpringerVerlag.
Little D.I., 1989 The appraisalof contaminants

in sedimentsof the SevemEstuaryand Inner Bristol
Channel,Reportto STPG, 47pp, (unpublished).
SevernEstuaryLevelsResearchCommittee,1988 The archaeological

potentialof the SevernEstuary.
An initial assessment
for STPG, 64pp,(unpublished).
Straaten L.M.J.U., van 1950 Environment of formation and facies of the Wadden Sea sediments.
Tijdschr.Kon. Neded. Aardr. Gen., LXVII, No. 3, p.354-368.
WilkinsonH.R., de BoerG. andThunderA., 1973. A cartographic

analysisof thechangingbed of the
Humber. A reportto the HumberEstuaryResearchCommittee,Dept. of Geography, Hull University,
Misc. SeriesNo. 14, 68pp, (unpublished).

17
Acoustic measurementsof suspended

sediment over sandwaves
P D Thorne, R L Soulsbyand P J Hardcastle
Abstract
Bedformsarecommonfeaturesin ourestuarineandcoastalwatersandthedynamicinteraction
betweenthebedformsandtheflow is of considerable interest.In strongturbulenttidalcurrents
thesuspended loadis thepredominate formof transportandit is essential
thatthesedimentary
processes of thiscomponent be understood. The work reportedhereis partof a studyaimed
at clardyingthe role of suspended transportuponsandwavedevelopment and migration,and
the feedbackof thesemodificationson the suspended load. To discernthe mechanismsof
transport,measurements of the turbulentflow were takenat four heightsabovea sandwave
andrecordsof the response of the suspended sedimentconcentration wereobtainedusingan
AcousticBackscatter Probe,ABP. In conjunction with theABP themeansuspended sediment
concentration was measuredusingconventional pumpsampling.The ABP is increasingly
becoming a prime techniquefor measuringsuspendedconcentrationfluctuationsand
consideration is given here to the assessment of the acousticmethodfor making such
measurements.
17.1 Introduction
Over thepastdecadeinterestin sedimentmovementhasincreasingly focusedon the detailed

effectsof turbulentcurrentsand wave actionon the transportof sedimentary material.Of
particularinterestis the sedimenttransportover bedformsin estuariesand coastalwaters.
Frequentlysuspended sedimentandsandwaves occurtogetherin theseenvironments,however,
thereis a lackof knowledgeof the mutualinfluences betweenthe migratingbeAforms, the
suspended sedimentand the flow. To developour understanding of suspended sediment
transportprocesses undertheseconditionsrequiresdetailedmeasurements of the turbulent
currentstructureandthe fluctuating suspendedsediment concentrationprofiles.
For the flow measurements

normallyelectromagnetic
currentmeters,ECM, and acoustic
traveltimedevices
havebeenemployed.Forsuspended loadmeasurements miniature
optical
backscattersensors,
MOB, (Downinget al (1981)) and acousticbackscatterprobes,ABP,
(Haneset al (1988)) are underdevelopment.The ABP is potentiallythe morepowerful

336 P D Thorne,R L SoulsbyandP J Hardcastle
techniquesincea singleinstrument canbeusedto non-intrusively measure suspended sediment

concentration profileswith a highspatial(-lcm) andtemporal(-0.1s) resolution. However,
whenclaimingthispotentialacknowledgement needsto be madeof the statistical natureof
thebackscattered signal.The amplitudeof thereturnedechois incoherent andhasa Rayliegh
amplitudeprobabilitydistribution. Thereforea degreeof incoherent averaginghas to be
conducted on thepressure databeforeconvertingthe signalto concentration levels.Also the
interpretation
of thebackscatteredechoesrequiresanunderstanding of theinteractionof sound
with suspensions andan accuratedescription of thisprocessis still underdevelopment.To
datethe usualapproachhasbeento conducta seriesof laboratorycalibrationexerciseson
suspensionsof variousconcentrationandparticlesizesanddevelopanalgorithmfor translating
the acousticobservations to sedimentparameters(Haneset al). Suchdetailedexercisesare
unnecessary if the backscattering
from marinesuspensions canbe described theoretically.
To examinetheinteractionof soundwith marinesuspensions a seriesof measurements have

been taken at 3 MHz in the laboratoryand in an estuarineenvironment.The laboratory
observationsare usedto appraisea theoreticaldescriptionof the acousticscatteringprocess
and provide an estimatefor the accuracyof the acousticapproach.The resultsof the
laboratorywork are appliedto the field datato obtainacousticestimatesof the suspended
load. Thesevaluesare assessed usingthe pump samplemeasurements obtainedin the field.
The highspatialandtemporalresolution facilityof theABP is thenemployedto lookin detail
at thesuspended concentration
field andto examineturbulentsedimentfluxesoverbedforms.
17.2 Acousticanalysis
17.2.1 Theory
The geometryfor the scatteringproblemis shownin Fig 1. An acoustictransceiver transmits

a short(of the orderof 10•s duration),high frequency(typicallya few MHz) acousticpulse
whichpropagates downthroughthewatercolumn.The rangeof theseinstruments is normally
of the orderof a coupleof metres.If any materialis in suspension thenas the soundpasses
throughthe water a proportionof the soundwill be backscattered and the returnedechois
usedto estimatethe suspended sedimentconcentration profile.
The problemis approached by initially specifyingthe interactionof soundwith a single

particle.Employingthe formulationgenerallyapplied (Neubauer(1974))the backscattered
pressure,P•, from a particlein thebeamof the transceiver is givenby
/•s' as/•oro
D2
2r
2if. l (17.1)

P D Thome,R L SoulsbyandP J Hardcastle 337
Pois thetransmitted pressureat rangero(usually= lm), a, is theparticleradius,r is therange

to theparticle,D is the transceiverdirectivityfunction,o• is the attenuationdue to the water
andfmis knownas the form functionanddescribes the scattering propertiesof the particle.
Sidelobes Parhcle backscatteringsound

'r Sound
ranscelver
•)U.•lse•
• am
earn
acoustic
axis
•nearfleld I'arfleld
range dependence Inverse range

complex dependency
Figure 1: Schematicof the backscattering

from a suspended
particle.
For a cloudof scatterersequation(17.1) needsto be integratedover the volumeinsonified

by the acousticpulseas it propagatesdownthroughthe water.When multiplescatteringcan
be neglected theintegralcanbe readilyevaluatedandtheensemble averageroot-mean-square
backscattered pressure,<Prms>, can be written (Shengand Hay (1988) and Thorne et al
(1990a)) as
 - (17.2)
M is the massconcentration, r is the rangefrom the transceiver,

c•,is the attenuationdue to
the suspension,p, is the densityof the sedimentin suspension,c is the velocityof soundin
thewater,x is the pulsedurationandV is a termto modifytherangedependence within the
transceiver
nearfield(Thorneet al (1990b)). k is theincidentacoustic wavenumber,a•is the
radiusof the transceiver, 0 the anglesubtended to the acousticaxis and J• is a first order
Besselfunction(this describesthe beam pattern).Equation(17.2) specifiesthe forward
problem;knowingthesediment parameterswe canevaluatethebackscattered signal.We need
to know•x,,fro,a•,p, andM(r). Forlaboratory workthelatterthreeparameters
canbereadily
measured, however,• an.dfmare not generallyknownfor irregularlyshapedparticles.To
obtainestimates for theseparameters it hasrecentlybeenconjectured(ShengandHay) that
relativelysimpleexpressions for c•, and fmmay describethe backscatteringbehaviourof
sedimentary
materialand the suggested
appropriate
expressions
are

(17.3)
Where•candp arethecompressibilityanddensityrespectively andsubscriptss ando refer

to the suspension
and water. To evaluate% an integrationover the concentration
profile is
requiredgiving
if.. ½
, --I' dr (17.4)
The expressionfor •, theattenuation

constant,
hasbeeninvestigated,
by ShengandHay,
usingthe datasetof Flammers(1962) andresultedin
• -(lJpa,){15(kaj)nJ[l+(kaj)
2+(-•)[3(kaj)n]} (17.5)
2
Thevariationof fa and• withka•areshownin Fig 2. For fa thereis a sharpdecrease

in
the form functionbelowkay<0.5,while aboveit thereis a reductionin gradientleadingto a
uniformvalueof unity for kay>5. A similarresponseis seenfor the attenuationconstant.The
predictionof the backscatteredpressure,the forward problem,can thereforebe expressed
analyticallyand comparisons madewith experimentaldata to assessits validity.
For marine measurements

the inversesolutionis required,using <P•> the suspended
concentration
needsto be evaluated,
a knowledge
of a,, fm,a•.p, is againnecessary.
For a
relativelywell sortedbed a first approximationfor a, andp, is to usevaluesfor the surficial
layerof thebed (theseare nominallyknownusually)andassumean invariantvaluefor a• in
thewatercolumnabovethebed. Theparameters fa anda, arethenestimated fromequations
(17.3)-(17.5).Theconcentration profileis thenobtained by simplyrearrangingequation(17.2).
</>2>
r292
ee(%.,.)
16a,
p, ".2Jl(karqinO)
sinOdO
(17.6)
t•r• • 3xc kat•inO

P D Thorne,R L SoulsbyandP J Hardcasfie 339
fm= 1
100
I0ø' pOs•:
075 b
10-1
10-2
10-3
10-4 I I I I I I I I
10-2 10-1 100 101 102 I0ø i0• I02

ka s
kos
Figure2: a) Variationof theformfunctionwithkay. b) The attenuation

constant
withkay.
It canbeseenfromequation (17.6)thatto evaluate M(r) we needtoknowM(r) in equation

(17.4). Thistherefore requires
a computational procedurewheresequentially M(r) is evaluated
anda numerical integration
conducted to evaluateor,andthiscalculation repeatedstep-wise
throughout thewatercolumn.Thisprocess hasits problems; for exampleif M(r) is initially
overestimated thenthevalueof or,is increased, thishighervalueis fed backintoequation
(17.6), M(r) is furtherover estimatedat the followingrangeand the effectcan accumulate
throughtheprofile.Thereis a positivefeedback processcausing a divergent
solution.Thisis
particularlyacuteat highconcentrationswhenthereis a degreeof uncertaintyin a,, whichwill
invariablybe the casesincethe def'mifive valuefor 0t, for any particularsedimentand
conditions is notat presentaccuratelyknown.However,thecorollaryto thisis thatfor low
valuesof o• the problemis well defined.This is furtherconsidered whenanalyzingthe
experimentaldata.
17.2.2 Laboratorymeasurements
To examinethe applicabilityof equation(17.2) a seriesof laboratorymeasurementswere

conductedin theverticalsedimenttowershownin Fig 3. Thisconsisted of a hopperandan
impellermounted on a perspextowerwith a collection trayat the base.A homogeneous
suspensionof approximately 1.5m couldbe obtainedwith this arrangement.Absolute
estimates
of suspended concentration
weretakenusingpumpedsampling.

Vibrator.
Sediment
container
Sieve•
Impeller
mixer Hopper
Mixer
Hydrophone
beam
width
Pump
sampling• •.. '¾ Homogeneous
3m
suspension
Electronics
8roadband
Tronsducer
IH,ydrophone
Figure3: Experimentalarrangement
for laboratorymeasurements
of backscattering.
The 3 MHz acoustictransceiverwasmountednearthe baseof the towerand the envelope

of the transceiversignal,backscattered
from the sedimentin suspension,wasdigitizedandfed
directlyintoa microcomputer. The systemsampledthebackscattered signalat 256 contiguous
range bins to yield an acousticprofile of the returnedecho. Typically 200 profiles were
collectedto mitigatethe randomfluctuationsof the signalmentionedin the Introductionand
the ensembleaveragermsbackscattered pressureprofilecomputed.Thesemeanprofileswere
usedfor comparisonwith the theoreticaldevelopments. The materialusedfor the suspension
wasmaterialrecoveredfrom theestuarinesite. This wascleansandwith a suspended particle
diameterof approximately d•0= 160•m.

P D Thorne,R L SoulsbyandP J Hardcastle 341
To establish
experimentally
the valuesfor fmand • a numberof measurements
at a
selectionof concentrations
were carriedout. Measurements
of the attenuationfor a rangeof
concentrations
were obtainedby recordingthe signallevel measuredusinga hydrophone
placedatthetopof thetower.Themeasurements gave•--0.8+0.15,
thiscompares
witha value
obtainedfromequation(17.5)of • = 0.35. Theformfunctionwasestimated
fromreadings
takenat low concentration
whenthe effectof or,wasnot significant.This wascarriedout by
rearrangingequation(17.2) and makingf•, the dependent variableon the backscatered
pressure.
(17.7)
<P,.,>,r
3xcM
J [•.[ •a-•;• SinOdO
This resultedin a valueof fro=0.7+0.1,equation(17.3) givesfro=0.5.The form function
valuesarecomparable whilefor attenuationthedifferenceis significant.At presentwe are
not in a positionto resolvethe latter discrepancy definitivelyalthoughif a, is being
underestimated
by20%thereadjusted
calculated
values
wouldgivefro----0.58
and•=0.53which
wouldbe roughlywithinthe expecteduncertainties.
However,repeatedmeasurements of a,
showa variabilityof theorderof 10%.Alsotheanalytical
expression
for ( is a meanfit to
the data of Flammer's which does include a scatter about a mean of 10%. However, these
variationsare not sufficientto accountfor the discrepancy
in the attenuationconstantand no
convincingexplanation
is availableat presentto resolvethe inconsistency.
For the current
laboratory
analysis
themeasured valuefor • hasbeen.•mplemented.
The resultsof the variationin the backscattered

signalwith the rangefor a homogeneous
suspensionare shownin Fig 4a. Thereis seento be a steadydecrease in signallevel with
rangewhichis dueto a combination of spherical
spreadingfromthe source,ie the 1/r term
in thedenominator of equation(17.2) andthereductiondueto attenuation in the argumentof
the exponent.It canbe seenthatfor the higherconcentration the signallevel at aboutlm is
an orderof magnitudebelow the lower concentration valuesdue to the dominanceof the
sedimentattenuationon the rangedependence.
Computationsfor the rms backscattered

signalusingequation(17.2) are seento be in
excellentagreement
withtheobservations
overtheorderof magnitudechangein concentration,
a rangevariationof nearlytwo decadesand a signalamplitudecoveringgreaterthan two
ordersof magnitude.
Comparisonof the variationwith range of the ratio of the acousticestimateof the

concentration,calculatedusingequation(17.6), to the pump samplevalueare shownin Fig
4b. The resultswerecomputedby evaluating% usingthemeasured concentration. Ideallythe
valueshouldbe unityandconstantwith range.The resultsare closeto thiswith somescatter
aboutunity,thereis greaterdeviationwithin approximately0.2m of the transceiverand this
is consideredto be due to the uncertaintyin modellingbackscattering in the transceiver
nearfield.The generalagreementis good though.A collectionof all the observations are

shownin Fig 5. Thisshowstheresults

for Ms, averaged
overtherange0.42-0.84mfromthe
transceiver,
against
1• for theconcentration
rangecovered.
100[ • 0171
kgm
3
1011 11 b
10ø 0.1
589kgm3
101
10 2 mp
10 3
01
100
[• 3 10
1 217 kgm3
101[ 1
0.1
10 2 10 1 10o 05 10
R(m) R(m)
Figure4: a) Rangedependence of thebackscattered

signal.-Equation
(17.2). b) Ratioof theacoustic
estimateof concentration
to thepumpedsamplemeasurement.
• Ol ot i.o to.o
M,(kcjrn
-•)
Figure
5: Comparison
of acoustic
measurement
of suspended
loadM•,,withthepumped
sample
values
M..

The variation
aboutMA=M p is of the orderof 20%, someof whichis assigned
to
uncertainties
in Mpof 10%,andtheresttotheacousticmeasurements.
Theresults,
however,
do show that the theoreticaland measuredconcentrationsare in excellentagreementand
confinn the applicability of the acoustictechniquefor measuringsuspendedsediment
concentration.
17.2.3 Marine observations
The estuarinemeasurements werecardedoutduringa springtidalperiodoverfloodsalone,

on a dryingsandflatat YellandMarshin the fiver Taw estuaryin North Devon.The estuary
is sheltered
andno waveactivitywaspresentfor the durationof the experiment.Duringthe
periodhigh concentrations of suspendedsedimentwere observed.The bed consistsof fine
sandand the areais dominatedby sandwaves whichgenerallyhavea wavelengthof 15-20m
with a trough-to-crest
heightof approximately 0.8m. There is little suspendedclay or silt in
the waterduringthe floodtide andtemperature andsalinityare bothuniformwith depth,due
to strongmixing, so there is no thermohalinedensitygradientpresent.The site for the
experimentis shownin Fig 6 with insertsdisplayingthe experimental arrangement and the
particlesizedislribution.At low waterthe sitewasdry andreadilyaccessible for arranging
the equipment.
31dHz
transducer
•-•
'•F Electl-omagnet•C
current
meter I • (tOm)
Figure 6: Field site, location,sedimentdistributionand experimental

arrangement
for estuarme
measurements.

At highwaterthedepthwasapproximately 5m. Measurements weretakenabovethebed

at 0.1, 0.2, 0.4 and 0.8m using;5cm diameterElectromagnetic CurrentMeters (ECM),
ImpellerCurrentMeters(ICM), the 3MHz ABP andpumpedsampling.The ABP wasalso
usedto obtaincontinuous profilesdownto thebed.Pumpedsamples wereobtainedat 5min
intervals,thereforeeachheightwas sampledevery 20min,the ICM's were recordedover a
4min countperiodeach5min. Analoguerecordsof the ABP andECM's weretakenon an
instrumentation taperecorderandwerelaterdigitisedat 5Hz for analysis.
Thecontinuous ABP
profileswerecollected ona separateloggingsystem consisting of 256contiguous
bins(ranges)
with profilesbeingobtainedat a rateof 3.6Hz. The profilesweredigitisedat 7usproviding
a spatialsamplingintervalof 0.525cm.The spatialresolution in thiscasewasdetermined by
thepulselength,thiswas 12usgivinga resolution of 0.9cm.
Over the floodperiodsconcentration profileswereobtainedfrom the acousticdatausing

equation (17.6).To establishvaluesfor f= and( for analyzing
thefielddataa smallnumber
of pumpsamplemeasurements wereemployed.Usingtwo of thepumpsampleobservations
at 0.8m heightthe value for fa was calculatedusingequation(17.7). This is a first
approximation with the assumptionbeingmadethat for thisshortrange,0.2m, ignoring %
wouldnotintroducesignificanterrorsin the estimationof f=. Employingthisvalueof form
functiona singlehighconcentration pumpsamplevalueat 10cmabovethebedwasnotedand
( adjusted sothattheacoustic estimateof concentration,
computed usingequation(17.6),at
the samerangegave a corresponding value.This gave valuesfor the form functionand
attenuationconstant of fa = 0.83+0.15and• = 0.45+0.1respectively. The formfunctionis
similarin valueto the laboratoryobservations althoughthe attenuationis approximatelyhalf,
whichis closerto the theoreticalestimate.An avenueconsidered for the discrepancy
between
the laboratoryand field estimatesfor the attenuationconstantwasthe assumption of uniform
particlesizedistribution
with heightabovethebedwhenestimating•. However,analysis
of
earlierobservations
showedno significantdifferencein particlesizedistributionat 0.25mand
lm abovethe bed. We have still yet to satisfactorilyresolvethis differencein ( . For
analyzing
thefielddatatheinsituobservations
of f= and• wereemployed
for all theacoustic
estimates of concentration.
Estimatesof concentration
profiles,computedusingequation(17.6), obtainedfrom one
minuteaveragenns acoustic
pressure dataareshownin Fig 7a. The measurementsnearpeak
flow,around1700hours,showsignificant
concentration
levelsof greater
thana Kgm'3.The
suspension
is well mixedthroughthe watercolumnwith valuescloseto 1 Kgm'3 being
estimated at l m above the bed.
Towards the end of the flood, 1900 hours, when concentrationsare seen to be about two
ordersof magnitudedownon the peak thereis an almostuniformconcentration profile. At
thehigherconcentrations thereis thepossibility
thatuncertainties
in thevalue of ( canlead
to significanterrors in concentrationestimatesat the longer ranges,ie near the bed. As
mentioned previouslythestep-wisesequentialcomputation of M(r) is proneto accumulating
errors,as M(r) is usedto calculateor,and is then usedto continueevaluatingM(r). An
exampleof thevariationwhichcanoccuris shownin Fig 7b. Thereis a steadydivergence in

predicted
concentration asthebedisapproached
andtherefore theconfidence
in concentration
estimates
is relatively
lownearthebedforhighconcentrations whenthevaluefor • is not
precisely
known.However, havingtakencognisance
of this,theresults
showninFig8, where
a comparisonof theconcentration
estimates
obtained
usingtheABPwith theesmarine pump
sampledatacollected
overtwo consecutive
floodsshowmoderately
goodagreement.
0 102
100
1735
R(m) R(m)
101 101
•s = 0.6
10-2 10ø I • 101 102

103 10-2 10-1 10ø 101 102 10• 100
M(kgm3) M(kgm3)
Figure7: a) Acoustic
measurements
of suspension
profilesat differenttimesovera floodperiod.b)
Effect of varyingthe attenuationconstant.
,Olin
+02m
•04m
aOSm•+• oIm
o2m
o4m
08m
.•
M.(kg•s) MA(kgn•
3)
io-• io-2 io-i io0 ioI Id3 1(•2 i(• •0 I0•

Mp
(kgm-3) Mp
(kcjr•
•')
Figure8: Comparison
of thefieldmeasurements
of thesuspended
concentration
usingacoustic,
M•,, and
pumpsample,
Mp,dataontwoseparate
days.

346
P D Theme,R L Soulsby
andP J Hardcastle
•' 4 Transceiver Bed
• •,' ':t ,. ....

60
T(s)• '."--',.......
--'-,,•;?__
• ...g•:-=
R(m)
Figure
9:High
temporal
resolution
acoustic
measurements
ofsuspended
lead.
Values
below
0.2kg
m-3
havebeensetto zeroforclarity.
The
results
are
sufficiently
similar
toprovide
encouragement
forprogressing
and
examining
the
data
topresent
itshigh
temporal
resolution
characteristics.
Anexample
ofthe
preliminary
results
presentlybeing
calculated
arethose
shown inFig9.Thiscovers
oneminute
ofdata,
equivalent
tooneprofile
inFig7 andshowsthetemporal
andspatial
variability
ofthe
concentration
field.
Surges
inconcentration
arefollowed
seconds
later
byrelatively
quiescent
periods
ofsimilar
duration,
withtheprocess
being
extremely
variable.

This changeabilityin concentrationis beingstudiedto examinetheprocessesof sediment

transportsubjectto turbulenttidal current.However,as previouslymentionedcare mustbe
adoptedin the interpretationof such diagramsbecauseof the statisticalnature of the
backscatteredreturn. In Fig 9 an elementof averaginghasbeencarriedout althoughthe eye
has been left to conducta degreeof the smoothing. Further work is ongoingon such
representations.
17.3 Sediment observations
Theanalysisof theacoustic datato obtainconcentration

profrieswhichhavesynchronism with
the turbulentdata is still in progress.This is becausefor logisticreasonstwo independent
systemswere employedto record the acousticand turbulencedata. However, discreet
measurements, sampledfrom the acousticprofries,at the four ECM heights,was recorded
simultaneously with the ECM data. Employingthe temporalresolutionprovidedby the
acoustictechnique theopportunity wastakento consider thelongitudinal-• andvertical•
diffusivefluxes at five locationsover a sandwave(Atkins et al (1989)). Resultsfrom this
analysisare presentedin Fig 10.
The measurements wereobtainedoverfive floodperiods,no datawere collectedover the

ebb. To facilitatecomparisonof the observations at different stagesof the flood the
measurements were categorisedinto fourvelocitybandsderivedfromthevelocityat 0.4m
abovethebedon a referencemast approximately 50mupstream of theexperimental site.The
fourbandswere0.1-0.3,0.3-0.5,0.5-0.7and0.7-0.9ms'• with band1 beingthe lowest
velocityrange.Mean valuesof the flux wereobtainedby takingthe productof the mean
currentandmeanconcentration overtheperiodwhilethevelocitywasin theappropriate
band.
The diffusiveflux wasobtainedfromtheproductof the5Hz velocityandconcentrationdata
over the sameperiodof time.
Shownin Fig 10ais thelongitudinal diffusiveflux. Thisis positivealongtheflankof the

sandwave indicatingthatthediffusionis occurringin theupstream direction.
In thetroughthe
fluxis negative
at lowvelocities,whileasthevel•i• increases thedepthof thedownstream
flux reduces.
Valuesfor themeansediment flux, UC, havealsobeencomputed andthishas
allowedanestimate
of-•/UC to be calculated
which,apartfromthetrough,
gavea value
generallyless than 0.05. The longitudinaldiffusiveflux is thereforegenerallydirected
upstream and lessthan5% of the meansedimentmovement.Fig 10b showsthe vertical
diffusiveflux which is the mechanism by whichmaterialis lifted into suspension.The
negativevaluesof verticaldiffusivefluxmeasuredin andcloseto thetroughimpliesthatthe
diffusionis aidingthe settlingprocess.
At all otherobser2r_ational
positions
the vertical
diffusiveflux is smallerthanthe calculated
settlingflux, %C. Over mostof the flank the
ratioof •E/w,Cis typically
0.2 at a heightof 0.1m,increasing
withheightto about0.7 at
0.8m. Thisis an indicationof a net downwardsettlingof sediment
overtheentiresandwave
whichis difficultto explainat present,however,previousobservations
haveshownsimilar

results.Analysisof thedatais still in progress

at thisstage,althoughthepotentialof using
the temporalresolutionof the ABP to investigateturbulentfluxeshasbeendemonstrated.
•
0.0 oo
o.8rt•-
> 0.02
0.02
o.
ø'8
I
0.8
0.0
•-0.
• o0o•
• o oo•
-5. •
0.0o6
0.00•
SCALE (m)
0 • a 3 4 5 6 7 8 9 •0
BAND
• • t t J t I I I I I •
BAND •
BAND 3
• BAND 4
Figure10: a) Longitudinal
diffusive
flux. b) Verticaldiffusive
flux. Scalekg m-:s-'.
17.4 Conclusions
An examinationof the application

of acousticscattering
for measuringsuspendedsediment
concentration
in an esmarineenvironmenthas beenconducted.Laboratory,marine,and
theoretical
estimatesfor theacousticparametersfmand• aresimilar,although
differences
in
• havebeennotedanddiscussed. Timeaveraged acoustic concentration
predictions
arein
goodagreement withpumpsampledatatakenin boththeestuaryandthelaboratory. Thehigh
temporalresolutionprofilesshowthe very variablenatureof the transport processwhen
turbulentcurrentsare present.The sedimentanalysisis beginningto showthe behaviourof

P D Thome, R L SoulsbyandP J Hardcastle 349
sedimentdiffusive processesover sandwavesand work is underwayto provide a better

understandingof the connectivitybetweenthe bed, the suspension
and the flow.
17.5 References
AtkinsR., SoulsbyR.L., WatersC.B., andOliver N. (1989) Field measurements

of sedimentsuspension
abovebedformsin a sandyestuary.HydraulicsResearchLtd., Wallingford,Oxford.Report$R 203.
DowningJ.P.,SteinbergR.W., andListerC.B.R. (1981) New instrumentation for the investigation

of
sedimentsuspension
in the shallowmarineenvironment.Mar. Oeol 42: 19-43.
FlammersG.H. (1962) Ultrasonicmeasurement

of suspended
sediments.
Geo. SurveyBull. No. 1141-A
(US GPO, Washington,DC)
HanesD.M., VincentC.E., HuntleyD.A., andClarkeT.L. (1988) Acousticmeasurements

of suspended
sandin theC2S
2experiment
at Stanhope
Lane,PrinceEdwardIsland,Mar.Geol81: 175-183.
NeubauerW.G., Vogt R.H. andDragonettL.R. (1974) Acousticreflectionfromelasticspheres1 Steady-

statesignals.J. Acou. Soc. ,•aner.55 6 1123-1129
ShengJ., and Hay A.E. (1988) An examinationof the sphericalscatterapproximation

in aqueous
suspensions
of sand.J. Acou. Soc. Am. 83 2 598-610
ThorneP.D., HayhurstL., andCampbellS.C. (1990a) Measurements of scatteringfrom a suspension

of sphericalscatters.Proceedingof the Instituteof Acoustics.Acoustics'90.Held at Southampton
University28-30 March 1990.
ThorneP.D., VincentC.E., HardcastleP.J.,RehmanS. andPearsonN. (1990b)Measuringsuspended

usingacousticbackscatterdevices. Mar.Geol. (In Press).

18
Some observations on fluid mud response

to water waves
F Jiangand A J Mehta
Abstract
The behaviour
of softmudunderprogressive
waveshasbeenbrieflyexaminedin the field.
Themainobjectivewasto demonstrate
thatunderwaveconditions
thataretoomildto cause
significant
particulate
erosion,
deeplayersof fluidmudcanundergo
measurable
oscillations
dueto waveloading.Thefieldsitewasin theshallowlittoralmarginof LakeOkeechobee,
Florida. In a test duringwhichwind wave frequencywas on the orderof 0.4Hz and
significant
waveheight
around
10cm,wave-coherent
mudmotion
wasmeasured
20cmbelow
the mud-water interface.
A shallowwaterwavemodel,whichconsiders
thewatercolumnto be inviscidandthemud
layertobea highviscosity
fluid,hasbeenusedtoaidin datainterpretation.
Giventhewave
energyspeca'um,themodelapproximately
simulates
thewatervelocityandmudacceleration
spectra;
deviations
between
simulationandmeasurementarepronouncedin thehighfrequency
rangewhereintheshallowwaterassumptionbreaksdown.A low frequency wavesignalat
about0.04Hz,believedto be forcedby thewindwave,wascharacteristic
of measured
spectra.
Thissignalin themudwasenhanced relativeto theforcingsignaldueto thedependence of
waveattenuation onfrequency, andledto horizontal muddisplacementsontheorderof 2mm.
It is possible
thatthissignalis akinto surfbeatwhichcauses themudto oscillateslowly,
therebycontributingto its mobility.Sucha motionmayhaveimplications for theratesof
nutrientexchange
betweenthebottomandthewatercolumn.
18.1 Introduction
It is generally
wellrecognized thatin shallow,
episodic
coastalor lacustrine
environments
with muddybeds,reworkingof mudby wavescauses thebottomto becomefluidized,with
the fluid statepersistingas longas wavescontinueandthereafter,untilthebottommaterial
dewaters sufficiently
to leadto hardening undercalmconditions.Laboratoryevidencepoints
to fiuidizationundercyclicloadingtobedueto thebreakup of thestructural
matrixof thebed

352 F Jiangand A J Mehta
heldtogether by cohesive, inter-particle

bonds (RossandMehta,1990).Thischange of state
froma bexlto a fluid-supported slurry,andwave-induced oscillations
withinthisslurry,have
obviousimplications forbottomboundary layerrelatedphenomena including,
forexample, the
availability
of fluidizedmudfor transport by current,andpossible changes in theeffective
bottompermeability or resistancetodiffusion,
leadingto correspondingchanges in therateof
exchangeof nutrients or contaminantsbetween thebottomandthewatercolumn.In thisstudy
thespecificmatterof fluidmudmotioninduced by progressive
waveswasexamined fromthe
followingperspective.
18.2 A physicalperspective
A simplephysicalperspective canbechosen to dealwitha rathercomplexproblemwhichis
characterizedby time-dependent changesin mud properties with continuedwave action.
Althoughsuchchanges havebeentrackedto someextentin laboratoryexperiments, field
evidenceis scarcedue to evidentproblemsin deployingrequisitetransducers.Furthermore,
the basisfor any theoreticalexamination
of the time-variability
of suchpropertiesas mud
shearstrengthis presentlyrudimentary.Thereforeit mustbe bornein mindthatin treating
theproblemtheseconsiderations imposecertainoperationallimitations
in datagathering
and
analyticconstraints
in analysis.
In the way of a generaldescriptionof the problemconsiderFig. 1, in which vertical

profilesof sedimentdensity(p) andthehorizontalcomponent of the wave-induced velocity
amplitude(u=) in the watercolumnand bottommud are depictedin a somewhatidealized
manner. In the upperwater column,in which pressureand inertia forcesare typically
dominantin governingwater motionand the flow field is practicallyirrotational,sediment
concentrationusuallytendsto be quitelow, so that the suspension
densityis closeto thatof
water. The lower boundaryof this uppercolumnis characterized by a rather significant
gradientin density,or lutocline,below which the densityof the fluid mud is measurably
higher.
Below the fluidized mud is the cohesivebed having yet higher densities(greaterthan
-1.2g/cm3).
Laboratory
observations
byMaaandMehta(1987),andtheoretical
workbyFoda
(1989) for instanceshow that the wave orbits can penetratethe bed, therebyleadingto
deformations of theupperpartof theotherwisestationary bed. Undercontinued waveloading
suchdeformations cancausefluidizationby eliminationof theeffectivestressassociatedwith
the bed matrix. Due to the generallylow ratesof upwardmassdiffusionabovethe wave
boundarylayerhowever,theproblemof mudmotionby wavescanbeconveniently considered
to be uncomplicated by the effectsof particulateerosiondue to stress-induced
breakupand
entrainment of the cohesivesedimentaggregates at the mud-waterinterface(vanRijn, 1985).
In fact,laboratoryaswell asfield observationsshowthatwaveconditions requiredto generate
measurable bottommotioncanbe quitemoderatecomparedwith conditionsrequirexlto cause
significantparticulateerosion(RossandMehta, 1990;MehtaandJiang,1990). Accordingly,

F Jiangand A J Mehta 353
the followingsystemis considered.
MWL
UB
Mobile
Suspenion
• Lutocline
..... • I•md'•'u c•
Deforming Bed
Figure1' Schematic
of mudbottomresponse
to wavesin termsof verticalprofilesof sedimentdensity
and velocity.
_ (•(x,t)
1P• (712(x,t)
. i mm Water m m m mm mm
z H2 P2' j• Fluid Mud
x ....•;:...•...:.?:.:•.....;.::.•..............:•:.....•?.•?•:..•...•?..•?:•:..
..'.."..':.:'.:.'
Bed
Figure2: Two-layeredwater-fluidmudsystemsubjectto progressive

waveaction.

354 F JiangandA J Mehta
A two-layered,water-fluidmudsystemforcedby a progressive, non-breakingsurfacewave

of periodicityspecifiedby frequency,•, is depictedin Fig. 2. As far as wavedynamicsis
concernedwe will restrictthe problemto one of long waves,which would thereforebe
applicableto veryshallowcoastalor lacustrinewaterbodies,or to themarginsof deeperones
wherewave actionoftenmattersthe most. In the caseof a rigid bottom,the shallowwater
condition
is satisfied
whenHoa/g< 0.1,whereH is thewaterdepthandg is theacceleration
due to gravity. For a given •, this relationshipspecifiesH suchthat, for shallowwater
conditionto hold, the actualdepthmustbe equalto or lessthan that valueof H. When the
bottomis non-rigidthe maximumwaterdepthto which shallowwater conditionis satisfied
will be somewhatlarger,inasmuchasthe wavelengthis greaterthanin therigid bottomcase.
The upperwaterlayerof thickness H• anddensityp• is considered to be inviscid,which

is not unreasonable in comparisonwith the highly viscouslower, homogeneous layer of
fluidizedmud havinga thicknessH2, densityp2 and dynamicviscosity[1. However,the
implicationsof thisassumption of inviscidupperlayerrelativeto the absence of the bottom
boundarylayerin the watercolumnandtheassociated interfacialshearstressarenotedlater.
Beneaththe lower mud layer is the bed, which is assumedto be rigid. The surfaceand
interfacevariationsabouttheirrespectivemeanvaluesareq•(x,0 andq2(x,0. The amplitude
of the simpleharmonicsurfacewave is assumedto be smallenoughto conformto linear
theory,as alsothe responseof the mudlayer. Accordingly,the relevantgoverningequations
of motionand continuitycan be writtenas (Gade, 1958):
Upper layer:
C•U
1 i•q1
•t
+g ax
- 0 (18.1)
i5(q•_qz)+H••u 1 -0
c•t
(18.2)
Lower layer:
•u 2 aTI aZU• (18.3)

aq=
+rg•xx +(1-r)g
/ix• - v az•
&h
dz+c3qz
.0 (18.4)

whereUl(X,t)andu2(x,z,t)are thewave-induced velocities,h = H2+ •1:,r = (p: - Pl)/P: is the

normalizeddensityjump,andv = p/p,•is thekinematicviscosityof mud. Considering thefact
that the fluid domainis boundedbetweenz = 0 and H1 + H,., is infinite in extentin the +x
direction,the lower layer is viscous,and the solutionsoughtis harmonic,the following
boundaryconditionsare imposed:
• • (0,t)= aoCOSOt (18.5a)
tll (eø'0'112(øø'g't)'
Y[1(eø'O
• Y]2
(oo,
t) --,0 (18.5b)
u•(x,O,0- 0 (18.5c)
(x,0 / oz= 0 (18.5d)
wherea0is the surfacewave amplitudeat x = 0. Equation18.5aspecifiesthe surfacewave

form,Equation18.5brepresents thefactthat,dueto viscousdissipation, all motionmustcease
at infinitedistance,Equation18.5cis the no-slipbottomboundarycondition,andEquation
18.5d statesthat becausethe upper layer fluid is inviscid,there can be no stressat the
interface.Note thatby virtueof theassumption of rotationality
in thelowerlayeronly,and
the shallowwatercondition,u,.can vary with z, but not
In orderto generalizethesolutionof theaboveequations andtheboundaryconditions, the

followingconvenient dimensionless quantitiesareintroduced:
= xh/H1,Xh' = xl•q-I•,H2' = Hf-I l, h' = H2' + xh', t' = gt, k' = kH1 (wherek is the wave
number),x'= x•, andz' = z/H1. ThusEquations18.1through18.4become:
Upper layer:
+ 0 (18.6)
iSt' F•
at'
, - v12
/5(vl• ') + c3u•'
ax'
=0 (18.7)

Lower layer:
Ou:' r &qz'+ 1-r&lx' - 1 &Zuz' (18.8)
•' alia'dz'+ aq:'

f0 Ox'
-0 0t'
(18.9)
whereFr= o(H•/g)• is theFroudenumber

andRe = oI-I•/x)is theReynolds
number.Notethat
oH• is thecharacteristic
wave velocity. Note furtherthatthedimensionless surfaceslopeterm
in Equation18.6, as well as the interfacialandsurfaceslopetermsin Equation18.8 are scaled
by 1/F•,whilethedissipation
termin Equation18.8is scaledby I/Re. For typicalwave-
forcedsituationsit canbe easilyshownthatthe dissipationterm is far lesssignificantthanthe
surfaceandinterfacialgradientterms,particularlytheformer. Yet, of course,dissipationplays
a criticalrole in the problemin termsof wave dampingandboundarylayer effect within the
mud. Note alsothatat typicallylow valuesof the waveReynoldsnumber,fluid mudmotion
is wholly laminar(Maa and Mehta, 1987).
The normalizedboundaryconditionsare expressed

as:
q•'(0,t') - A cost' (18.10a)
u•'(•,t'), u:'(•,z,t'), q•'(•.,t') andq:'(•,t') -- 0 (18.10b)
%'(x',0,t')- 0 (18.]0c)
au:'(x,H:',t')/az'
- o (18.10d)
whereA = ao/H•. The solutionsof Equations18.6 through18.9 with theseboundary

conditions
arerelativelystraightforward.Omittingdetailswhichmaybereferenced elsewhere
(Mehta andJiang,1990),we statethe followingsolutionsfor theelevationsandthevelocities:
n•' - A exp[i(k'x' - t')] (18.11)
q2 -A1- exp[i(k'x' - t')] (18.12)

u•'- Ak' [i(k'x' t')] (18.13)

F,2
exp -
[1 - cosh(mz')+ u•h(m•')
(18.14)
sinh(mz')]exp [i(k'x' - t')]
As observedq,.' is damped relativeto Xl,'by themultiprier

1-(k?Fr)
•'. Likewise,u,.'is
dampedrelativeto u,' by thetwomultipliersof Ak'/Fr"(i.e.amplitude
of u/).
Next we requirethe dispersionrelationship for waveceleritybasedon thereal part,lq, of

the wavenumber k, andan expression for the waveattenuation coefficientk•, whichis equal
to the imaginarypartof k. Note thatthe surfacewaveamplitudeis specifiedas ax= aoexp(-
k•x), whereaxis the amplitudeat a distancex. Equations18.11 through18.14 satisfythe
boundaryconditions18.10b,as the amplitudes andvelocitiesvanishat infinitedistance.We
have,by definition,
k'- k,' + ik( (18.15)
where
•'' VI[(¾I + ¾xz)

"t + YR] (18.16)
which is the dispersionrelationship,and
(18.17)
whichis the expression

for the wave attenuation
coefficient.Here
[R[1+R- cos-0
2
(p2+q2)¬]
+
2r(R: +I:)
(18.18a)
I[I-sin•(p
z+q2)•
]}

Yi '
2r(Ra+Ia) [R[l-sin-02
(p2+q2)•
]- (18.18b)
I[ 1*R-c,
os-0
2
(p2.q2)•i]
]
where
p- (1 +R) 2-4rR- I: (18.19a)
q=21(1 +R-2r) (18.19b)
0=tan-I
(•) (18.19c)
exp(4X)
- 1+2sin(2x)
ßexp(2x)
1 (18.19d)
I=H'•[1
-2:1exp(4x)
-+11;-2•(2X)
[exp(nz) 2•-• 'e•2X)
I (18.19e
(18.190
It canbereadilyshownthatk•r depends onH•, r and•. Notethat• is themudlayer

thicknessnormalized
by (2u/c•)
•, whichisrepresentative
of themudboundary
layerthickness.
For the selectedvalueof r = 0.15 (e.g. corresponding
to field condition),this dependence
is
shownin Fig.3 for H• ranging
from0.1 to 1. It is observed
thatfor a givenH•,
decreasesfrom 1 at • = 0, becoming
practicallyconstant
abovea certain•. This trendcan
be easilyexamined,for example,for a given system(H•, H2 fixed) subjectto a wave of a
givenfrequency.
At Z = 0, thebottom
isrigid,hence
k•= Frmeans
c•/k= Co= (gH•)•, which
is theshallowwater,rigidbottomwavecelerity. Increasing
• impliesdecreasing
viscosity
hence
decreasing
p, since
r and,therefore,
p2areheldconstant
in thiscase.Notethatk•
= Co/C,whereC is thecelerityat anyg > 0. Initiallytherefore,as thebottombecomes soft
the wavespeedincreasesover thatdueto rigid bottom. As • -• oo,the lowerlayerbecomes
inviscid,
andthewavespeed
equals
[gHl(1+ H•)]•. Forpractical
purposes,
thelowerlayer
becomes"watery"for valuesof • exceeding2 to 3, and the celeritydoesnot changetoo
rapidlywith furtherincreasein • as seenfrom Fig. 3.

1.0
H:•=0.1
0.2
0.3
0.4
0.5
0.6
0.7
r = 0.15
0.8
0.9
1.0
0.7
0 I 2 3 4
7,=H:•(Re/2)I/2[=H2
((•/2v)
1/2]
Figure3: Dispersion
relationship
basedon theinviscid-viscid
model
In Fig.4 thenormalizedattenuation
coefficient
k;/F,which
alsodependsonH•,r andZ,
isplotted
againstZ forvalues
ofH•ranging
from0.1to1. Wenotethatsince
k•/F•= k•Co/o,
theplotcanbe considered to represent themannerin whichk• changes with increasing Z,
starting
withZ TM 0 atwhichk• -- 0 (rigidbottomcase).An interesting featureof theobserved
variationin k• is theoccurrence of a peakin k• asZ approaches unity. In otherwords,wave
damping is greatest whenthemudlayerthickness (H2)is of theorderof theboundary layer
thickness.As Z increases beyondthisvalue,k• decreases with decreasing thickness of the
boundary layerandapproaches zeroasZ -• o•,i.e.thelowerlayeralsobecomes inviscid.As
Z decreases belowone,although theentiremudthickness constitutestheboundary layer,the
rateof energydissipation reduces relativeto thatat Z = 1, hencek• reducesaswe!!.
In Fig. 5, profilesof thevelocityamplitude,

Urn,areplottedusingtypicaltestparameters,
andg ranging from0.1 to 1.5. Noticetheheavydamping of velocitywithd•reasingg in the
fluidmudlayer. By virtueof themodelassumptions, noboundary layeris foundin thewater
layer,eventhoughthevelocityis dampedwithd•reasingZ, by virtueof theinfluenceof g
on the wave number(Equation18.13). Besidesthe absenceof boundarylayer effect
particularlyjustabovetheinterface, thelackof velocityequalityat theinterfaceis another
manifestation(limitation)of the assumedinviscid-viscidbehaviour. The outcomeis the
suggestion
of steeper
gradients
andhencerotationality
at theinterface
thanin reality,although

laboratorymeasurements (e.g. Maa andMehta, 1987)do indicatea ratherdrasticreductionin

the mud velocityrelativeto that in wateracrossa thin interfaciallayer.
0.16
] I f I • I i I
H•,=1.0
0.12
r= 0.15
0.08
0.04
0.1
0.00
0 I 2 3 4 5
Z=H:•(Re/2)l/2[=H2
(c•/2v)
1/2]
Figure4: Wave attenuationrelationshipbasedon the inviscid-viscidmodel
18.3 Test Site and Experiment

Fig.6 showsthetestsitelocatedaboutlkm offshoreof theeastern
edgeof LakeOkeechobee,
Florida,in theproximityof the OkeechobeeWaterway.A largepart(528 sq.km) of this
shallow,hardbottomlakeis overlainwithmudhavinga thickness of up to 80cm(Kirbyet
al., 1989). Duringthefieldtest,waveconditions
maybecharacterized ashavingbeenmild.
While suchwavesdid generatehorizontalmotionswithin mud, partictdate erosionwas
minimalbecausethe wave-inducedbottomstresses
werebelow0.5 Pa, thecriticalshearstress
for particulate
erosion
of mud(Hwang,1989). Whenstormwavesdo occur,thetop-5cm
thick layer of the mud tendsto dilateto -10cm due to sedimentdiffusion. Above thisdilated
layer,upwardsediment masstransport
tendsto becomparatively
verysmall,butthepresence
of thedilatedlayerdoestendto complicate
thenear-bed
processes
asfar assediment
motion
is concerned(Hwang, 1989).

1.0 I i I • I
•- 1.5=7.
O•'
0.8 ao=8cm
=0.15
• 0.6 WATER
mO
0.4
• _ 0.5
• 0 1.5
=7. FLUID
MUD
• 0.2
o[,• • I • I • I
0 0.1 0.2 0.3 0.4
VELOCITYAMPLITUDE,urn(m/s)
Figure5: Simulatedprofilesof velocityamplitude,umfor differentvaluesof Z usingtypicalfield test

parameters
Two featuresof the bottomenvironmentare noteworthy. Firstly, the mud throughout

includesabout40% (by weight)materialthat is essentiallyorganicand,secondly,the top 5
to 20cmof themudhasnegligible
vaneshearstrength
(lessthan0.39kN/m2),
andpersists
in
the fluid stateeven undercalm conditions(Kirby et al., 1989, Hwang, 1989). This stateis
broughtaboutpartly by the occasionallysignificantwave action,but it is believedthat an
importantfactoris thepresence of theorganicmaterialwhich,presumably by virtueof having
anopenandcomparatively strong,fibrousstructure
of peatyorigin,prevents
rapiddewatering
of the top layer of the deposit.
In consonance with the natureof the problemand the two-layeredflow formulation,the

bulk densityand the dynamicviscosityare the two importantparameters characterizing
the
fluid mud. Verticaldensityprofilesin the mud wereobtainedby a simplebottomcoring
procedure that yieldedapproximate variationof densitywith depth. A mudbulk density
profileis shownin Fig. 7. The substrate underneaththe mud layer may be considered as
"hard";the transitionto hardnessbeingheredefinedas the level at which the instrumented
tower assembly(notedlater) restedon its own account.

27ø1
80ø40'
Test
Site
26ø45._.•
81ø00'
I 80ø40'[
0 5Kin
Depths in Meters
Below Datum
Figure6: Testsitein LakeOkeechobee.

Depthsarerelativeto a datumwhichis 3.81m abovemsl
Mudviscosity
wasmeasured
in a Brookfield
viscometer
(modelLVT) withminiature
vanes,
at different
mudbulkdensities.
Fig.8ashows
thestress-shear
raterelationship
obtained
ata
density
of 1.05g/cm
3. Thebehaviour
of themudisobserved
tobepseudoplastic
atrates
less
thanabout3s'•,andNewtonian
athigher
rates.Sincethewave-induced
shear
ratesin themud
duringthe studyperiodwere on the orderof 0.1-0.2s
'•, it was decidedto calculatea
representative
viscosity,
p, astheslope
of thetangent
tothecurve
passing
through
theorigin
asshownby thedashed line. Thestress-shear
raterelationships
at all otherdensities
were
foundto be qualitatively
similar,andwereusedto obtainthecorresponding valuesof
viscosity.
Theresultingrelationship
between theviscosity
(relative
towater,Pw)anddensity
isshowninFig.8b. Duetolimitations intheapparatus,
muddenser than1.12g/cm 3couldnot
betested.It wasthereforeassumed thatathigher(upto 1.18g/cm
3)densities,theviscosity
couldbeobtainedbyextrapolationof thelog-linear
curveat higherdensities
in Fig.8b,i.e.
assuminga power law.

BULKDENSITY
(g/cm
3)
1.1 1.2 1.3 1.4
01.0 I ii I •
I
-10 - I
\,
ß I
• -20
Mud
_
-30 • Ler_
-40
-50
.....
-6O "Hard"
I ! I
Figure 7: Mud densityprofile at the testsite
Any interpretation
of viscometerdataobtainedusingvanes,in termsof a relationship
between viscosity and the structure of the sediment matrix, must be treated with
circumspection sincevanes,by their very presence,breakup the matrixat the cylindrical
surface definedby thevanedimensions.Furthermore, viscosity
isnotcharacterized
bydensity
in a uniquesense,insofaras viscosityand densityare independent physicalpropertiesof
fluids. Neverthelesstheobserved trendin Fig. 8b doesseemto suggesta rathersignificant
influence of sedimentpacking,asreflectedby thebulkdensity,on viscosity.Apparently,at
densities
lessthanabout1.02g/cm
3,theaggregated
matrixrapidly
became
tightlypacked
with
increasing
density. However,furtherincreasein densitydid not seemto havedrastically
alteredthecompact
arrangement attainedat 1.02g/cm3. Sincethe"granular"density
of the
organics-rich
sediment
was2.14g/cm 3, the sediment volumefractioncorresponding to
1.02g/cm3wouldbe0.018.It isconceivablethatthisvolume fraction
isequaltotheso-called
criticalvolumefractionabovewhichthe rigidityof the matrix(as reflectedby the shear
modulusof elasticity),increases
rapidly(WilliamsandWilliams, 1989).
Duringtheexperiment,time-series
of waterlevel,horizontal
watervelocityandmudmotion
were obtained. Water level was measured with a subsurfacemountedpressuregage

(Transmetrics,Model P21LA). Water motion was measuredby an electromagnetic(EM)

currentmeter (Marsch-McBirney,Model 521), and a biaxial accelerometer (Entran,Model
EGA2-C-5DY) was usedto recordmud motion. A 4Hz low passfilter was usedto remove
high frequencynoisefrom the accelerometer output. Compensative correctionto the output
hadto be madeto accountfor a very slowlineardrift. Finally, a running-meancorrectionwas
madeto theoutputto eliminate
lowfrequency
fluctuations
of theorderof 10'3Hz.
4-
P2
=1.04
g/cm
3
!• =1.90Paos
3--
I
/
I
I
I
/
0 2 4 6 8 10 12 14
SHEAR
RATE
(•1)
Figure8a:Stress-shear
raterelationship
for LakeOkeechobee
mudat a bulkdensity
of 1.04g/cm3
The transducers weremountedappropriately on an 2.45mtall aluminumtowerassembly

with a triangularbasefor bottomstability. The depthof waterat the testsitewas 1.43m,and
mud thickness was 0.55m. The EM current meter was mounted 0.87m below the still water
levelin sucha manner asto allowmeasurementof thetwohorizontal components of velocity,
in orderto enablethedetermination
of thevelocityin thedirectionof wavepropagation. The
accelerometerwasmounted0.20mbelowthemudsurfacefor measuring a horizontalandthe
verticalacceleration
components. The towerwasdesigned to allowpositioning
of thedevice
in sucha way as to direct the horizontalaxis in the dominantdirectionof wave motion.
The samplingrate for all the transducers

was4Hz for 10mineveryhour. Thisratemay
be considered
to be minimallyadequate basedon previousstudies(MehtaandDyer, 1990).

The transducers
were connectedto a dataacquisitionsystem(Tattletale,Model 6) mountedon
a woodenbaseat the top of the tower. The testwas conductedfor a total of sevenhours,
duringwhichthe wavedirectionwasconsistently westerly,corresponding to a fetchof about
50kmo
I I I
103
uJ 102
•W = 1.0087x 10'3Pa ßs -
22oC -
o
O3 10
1
1.00 1.02 1.04 1.06 1.08 1.10 1.12
BULK
DENSITY,
p2(g/cm
3)
Figure8b: Selectedrelationship
betweendynamicviscosityanddensityfor Okeechobee
mud
18.4 Water and mud motions
The time-series
wereanalyzedin termsof theirspectralproperties
andcentraltendencies
via standardprocedures.
Fig. 9 showsthevariationof thesignificant
waveheight,I-I•, over
testduration.Notethateachhourlydatapointis a 10minaverage overtherecordlength.H•
is observedto havebeenrathersmall,peakingto 10cmat 3hr.
In Fig. 10 thevariationof thedominant

(modal)surfacewavefrequency,
fro,definedasthe
frequency at thepeakof thewaveenergydensityspectrum, is shown.An exampleof the

spectrum
itselfis shownin Fig. 11 The modalfrequency
variationis compared
in Fig. 10
with thesamedetermined from thewatervelocityspectra,an exampleof which(corresponding
to Fig. 11) is shownin Fig. 12.
As observed fromFig. 10, thedominantwavefrequency wascomparatively constant,

with
a meanof 0.42Hz (obtainedfromfmvaluesfor thewaveandthevelocityspectra).Thisvalue
canbe shownto correspond to a wind of 20km/hr(moderatebreeze),usingshallowwater
forecastingcurvesfor wave generation over a rigid bottom(CoastalEngineering Research
Center, 1984). This value is believedto be a reasonableestimateof the wind speedat the
timeof study;however,at thisspeedtheforecasted waveheightin therigidbottomcasecan
be shownto be about40cm, whichwould be four timesgreaterthanwhat was measured.It
is surmisedthatthe discrepancy is dueto significantwave dampingover the mudbottom.
o•,10
I I
0 I 2 5 6
TIME (hr)
Figure 9: Variationof significantwave heightduringtest
A mud accelerationspectrum(corresponding to the wave energyspectrumin Fig. 11 and

velocityspectrumin Fig. 12) is shownin Fig. 13. The correspondingmodalfrequency(Fig.
10) is seento be roughlyin agreementwith the wave- and velocity-basedmodalfrequencies
in Fig. 10,thusimplyingthattheaccelerationswerewave-coherent. The slightlyhighervalues

(mean 0.51Hz) in the caseof accelerationin comparisonwith thosefor the waves and the
current are believed to be attributable to anomalies associated with the in situ mount used for
the accelerometer.The accelerationspectrumshowsa markedpeak at a very low frequency
corresponding to a long periodoscillation.Commensurate but comparativelysmallerpeaks
also appearin the wave and watervelocityspectraof Figs. 11 and 12. This low frequency
is distinctfrom the wind wave-inducedfrequency,and is discussedfurther later.
1.0
-1- 0.8
E
Mud Acceleration
0 0.6
0.2
0.0
0 I 2 3 4 5 6 7
TIME (hr)
Figure 10: Variationof modalwave frequencyduringtest
Verticalacceleration measurementsdid notexhibitlongwavesignatures for evidentreasons

but, on the otherhand,the magnitudeof the varianceobtainedfrom the time-seriesof vertical
acceleration
showedthisvarianceto rangefromalmostnil to -0.004cm'/s4;
thelattervalue
beingas highas the varianceof the horizontalacceleration.This behaviourindicatesthat,in
general,thewavesat the sitedid notquiteconformto the shallowwaterassumption.In fact,
it canbe easilyshownthat thisconditionwas valid for waveshavingfrequencies lessthan
around0.2Hz. A subsequent experimentat the samesite, but with the currentmeter set to
measurea horizontalandthe verticalwatervelocities,confirmedthisobservation (Mehta and
Jiang, 1990).

5O
[ I [ I [ I [ I
• 40
30
o
o 0.2 0.4 0.6 0.8 1.0
FREQUENCY (Hz)
Figure 11: Wave energyspectrumat 1 hr

3OO
i I ' I ' I i I
LU 240
i-
180-
•Itl•
Calculated
0 0.2 0.4 0.6
FREQUENCY (Hz)
Figure 12: Measuredandcalculatedwatervelocityspectraat 1 hr

F JiangandA J Mehta 369
0.30 j i
•-,
••. 0.24
•E I
o3z o.18
z•o - Calculated•,,,,•
!•
Measured•
I" i
Q. ,• 0.06
/'
0 0.2 0.4 0.6 0.8 1 .c
FREQUENCY (Hz)
Figure13:Measured
andcalculated
mudacceleration
spectra
at 1 hr
18.5 Response
of LinearizedFluid Mud-WaterSystem
Giventhewaveenergyspectrum, whichessentially
characterizes
themeasuredwaveamplitude
(A) variationwith wavefrequency, Eq. 18.13canbe usedvia the correspondingspectral
transferfunctions(MehtaandJiang,1990)to calculatethewatervelocityspectrum,
andthe
time-derivative
of Eq. 18.14to calculatethe mudacceleration
spectrum.
An evidentdifficultyin adaptingthe modelto measurement involvesthe selectionof a

representative muddensitysince,asseenfromFig.7, thedensitycharacteristically increased
quiterapidlywithdepthbelowtheinterface in thetop-10cm layer. Thematterof selecting
densityof course bearscriticallyontherateof energydissipation, via thedependence of the
viscosity on density(Fig. 8b). In orderto account for thisproblem, Maa andMehta(1987)
andothers(e.g. Shibayama et al., 1989)havedeveloped layeredbedmodelsin whichmud
properties canvary arbitrarilywith depth. It wasfelt thatthe useof suchmodelsfor the
presentcase would amountto an over-specification of the physicalsystem,whose
understanding wasconsmined by theratherlimiteddatacollection effort. On theotherhand,
theassumption of aninvariantmuddensity,especially in viewof itsevidentdepth-dependence
(Fig. 7), is a limitationof modelapplication whichmustbe bornein mindwhileconsidering

model results.
The choiceof appropriate

mudthickness,
H:, wasdifficultin the samesenseas choosing
a representative
density.Withregardto thelatterquantity,
a valueof 1.18g/cm
3wasselected,
correspondingto theposition(AC) of the accelerometer shownin Fig. 7. The computations
were foundto be particularlysensitiveto the selectedvalueof H: by virtueof the effectof
Z on the kinematics,whichin turn signifiesthe stronginter-relationshipbetweenthe mud
boundarylayer and energydissipation.It shouldbe furtherpointedout that,as a rule mud
becomes
ratherimmobile
at densities
exceeding
around1.2g/cm
3. Therefore,
referring
tofig.
7,it wouldbe unreasonable
to selectthe entiremud thickness
of 0.55M for computational
purposes. It was decidedto selectH: = 0.283m. For the casesexamined,this selection
seemedto give thebestresultsby way of agreementwith the data. Unfortunately
however,
consideringthe limitationsassociatedwith model assumption of constantdensity,no
independent
arguments canbe offeredwithregardto anyjustification
for thechoiceof H: on
physicalgrounds.
'•' 0.4
• 0.3
0
•: 0.2
o
Mud Acceleration
•,
0
0.1
- Surface
Wave
/ '--
Water Velocity -
0.0 I I I I I I
I 2 3 4 5 6 7
TIME (hr)
Figure14:Dominantlow frequency
spectral
signature
duringthetest
Usingthesurface waveenergyspectrum in Fig. 11,thecalculated

watervelocityandmud
acceleration
spectraare shownin Figs. 12 and 13. The velocitymeasurementwas0.56m
abovetheinterface, at an elevation
thatwaslikelyto havebeenabovethewaveboundary
layer,sincethe latterwouldbe on the orderof a centimetreonly. Therefore,the model
assumption
of inviscid
upperlayer,whencomparing
calculated
values
withdataessentially
in
theupperportionof thewatercolumn,maynotbeoverlylimiting.Dataandcalculation
show

approximatelysimilartrends.Notethatby virtueof theassumptions

of linearityandinviscid
upperlayer, the calculatedvelocityspectrumis nearlyself-similarto the wave energy
spectrum.
Comparing thecalculated andmeasured mudaccelerations in Fig. 13 indicates thatwhile

modelresultsgenerallyagreewiththedata,thedataexhibitaccelerations at highfrequencies
whichcannotbe accounted for by theshallowwatermodel. Thesehighfrequencywaveswere
proportionatelylessdampedmostprobablydue to thesebeingoutsidethe shallowwater
domain,in intermediatewaterdepth. Note thatthe measured velocityspectrum of Fig. 12
indicates
greaterdampingat highfrequencies relativeto theaccelerationspectrum in Fig. 13,
possiblydueto transducerlimitations.In generalhowever,selective dampingof waterwave
components hasbeenwelldocumented for exampleby WellsandKemp(1986),in theirstudy
on wavepropagation over a largemudflat off Surinam.
18.6 Low FrequencySignature

The dominantfrequencyof the low frequencywave signatureis plottedin fig. 14. It is
observedthatall the spectrayield ratherconsistent
valueswith a meanof 0.043Hz. It is also
noteworthythat by integratingthe accelerationtwice, it can be shownthat a maximum
horizontaldisplacement of mudon theorderof 2mm occurredat -0.03Hz.
Components of
Forcing Wave
Forced
Long
Wave
FREQUENCY
fœ fl f2
---• •'••Af •'[ •Af
Figure 15: Wave energyspectrumshowingtwo forcingwavecomponents

andforcedlongwave

It is seenfrom Figs. 11, 12 and 13 thatthe low frequencywavepeakis accentuated in mud

relativeto that in water, when this peak is comparedwith the corresponding peak of the
forcingwave. This relativeenhancement is due to frequencydependentshallowwaterwave
dampingas canbe gleanedfrom theattenuation coefficientcharacterizedin Fig. 4. Thus,for
example,selecting0.4Hz and 0.04Hz as representative frequencies
for the forcingwaveand
the low frequencywave, respectively,the corresponding dampingcoefficients(assuming
shallow
watercondition)
wouldbe0.0034m
4 and0.0013m
'•, whichindicates
relatively
much
lessdampingat the lower frequency.
Somecomments on thecausative mechanism for thelow frequencysignature

arein order.
In thatcontextit mustbe notedthat the dominantseichingfrequencyin I_akeOkeechobe•is
considerably
smaller,
ontheorderof 10'•Hz.It seems
plausible
thatthelongwaveis in fact
a secondordereffect resultingfrom wind wave forcingleadingto surf beat. This effect,
whichcommonly
occurs
in thefrequency
rangeof 3x10'3to 8x10'3Hz
atopencoasts,
maybe
modifiedby a compliantbottomas in the presentcase,but it wouldservethe purposeto
examinethis mechanismby assuminga rigid bottom,at least for arrivingat a possible
qualitativeexplanationfor the occurrence
of the low frequencywave.
At/--Forcing
Wave
•' I/. , I• t-orcec•
wavem . .t !• I
-5
v v
-1 90 180 270 360

'-F(deg)
Figure16: Forcingwavegroupandlow frequencywavederivedfrom testdataat 1 hr

We considertwo waves,•h(0,t) and•b(0,t), of respectivefrequencies

• and•,. =
where a• is a small difference:
•]1 ' &l•Ot - el) (18.20a)
= azcot - (18.20b)
Here, el and ee are the phaselags. Assumingfurtherthat the amplitudea = a• = a,., the
resultant(forcing)wave, •lr = •l• + •l,., is
n,=2acos(ot
-e)-cos(,aø•:
ae) (18.21)
where• = (• + •z)/2, e = (el + e2)/2,andae = el - e2. It is thusseenthat the wave form,

2acos(•t - e), is modulatedby cos[(a•-t-ae)/2], which causesthe beat effect due to wave
groupiness.
The factthatreal waveshavefinite amplitudesmeansthatsecondordereffectsarisingfrom

changesin water surfaceelevationand associated kinematicscannotalwaysbe ignored. An
effect germaneto the presentcaseis the settingup of a forcedlong wave which modulates
the wave givenby Eq. 18.21. This longwave followsfrom the inclusionof the kinetichead
termin the dynamicfree surfaceboundarycondition,andis obtainedby calculatingthe mean
(relativeto the wave period2•r/c•)water surfaceprofile by carryingout the computations
to
secondorder(Longuet-Higgins and Stewart,1962; Dean and Dalrymple,1984). Omitting
detailswhichare givenelsewhere(Mehta andJiang,1990), and withoutrestrictingthe basic
formulationto shallow water, the secondorder wave can be shown to be (the overbar
representing shortperiod,i.e. 2•r/c•,averaging):
- =-
n a2k [1 * cos(ao-t- e)] (18.22)
wherek = (k• + k2)/2. The first termon therighthandsiderepresents a steadysetdownand

the secondis the forcedlongwave. The energyspectrumthusobtainedis shownin Fig. 20,
whichshowsthatthelongwave,corresponding to two forcingwavesat frequencies
f• -- al/2•r
andf,.-- ad2•randamplitude
a, hasa frequency
f• -- af -- f,.- f• andamplitude
a•,/sinhkH.
It is importantto notethat,comparing Eq. 18.22with Eq. 18.21indicatesthatthe longwave
is •r radians(180ø) out of phasewith the envelopeof the shortperiodwave group.
When forcingis represented

by a continuousspectrum,the treatmentfor determiningthe
forcedlong wave spectrumbecomesinvolved(Sharmaand Dean, 1979). Here a very
approximateapproachis selectedto interpretthe presemdata. Considerfor examplethe
measured,doublepeakedvelocityspectrum correspondingto the forcingwave in Fig. 12.

Assuming
thetwo peakfrequencies
to be theprimarycontributors
to thecorresponding
long
wave peak, we have the following parameters(obtainedfrom the Fourier seriesof the
corresponding surfacewave record):a• = 2.4cm,a: = 2.6cm,long wave amplitudea• =
0.30cm,f• = 0.38Hz, f: = 0.42Hz, f• = 0.049Hz, e• = 191ø,e: = 171ø, and e• = -65ø. Note
thataf -- 0.040Hz, whichis reasonably closeto fe The forcingwave (q•) and the forced
wave (q•) are plottedagainst• = ao.t-ae in Fig. 16. It is evidentthat the forcedwave is
almost;• radiansoutof phasewith the shortperiodwaveenvelopeas theorywouldrequire.
If we assume theshallowwatercondition, theforcedwaveamplitude(fromEq. 18.22)would
bea:/H.Givena -- 0.5(a•+ a•)= 2.50cm
andH -- 143cm,
a:/H= 0.044cm.
Thusthetheory
under-predicts
the amplitudesignificantly,which might be due to inherenttheoretical
limitations
aswell as therathergrossassumptions
madein applyingthetheoryto data.
18.7 ConcludingRemarks
Notwithstanding
thelimitednatureof thedataobtained, thecomplexities
in modelling
mud
motion,and the ratherobviousconstraintsin simulationarisingfrom the shallowwater
assumption
aswell asothers(e.g.inviscidwaterlayer,linearized
response,
homogeneous mud
layer), it is demonstratedthat measurablemud accelerationscan occur tens of centimetre
belowthemud-water interface,
underwaveactionthatis mildenough toprecludemeasurable
particulate
erosion of theinterface.Although bioturbationdoesnotseemto be a significant
factorin LakeOkeechobee (Kirbyet al., 1989),theeffectof persistent
mudoscillation,
even
thoughverysmall,canbe germane to likelychanges in theratesof exchangeof phosphorus
andotherconstituents responsible
for thehighlytrophicstateof thislake. Similarly,there
may be an effect on the formationand upwardtransportof gas bubbleswhich occur
abundantly in the muddyareaof the lake (Kirby et al, 1989).
In thisstudy,weconsidered
muduptoa density
of 1.18g/cm
3tobea highlyviscous
fluid.
In general,however,carefulmeasurements
(e.g. SillsandElder, 1981) indicatethat,at least
underquiescent
conditions,at densities
exceeding
about1.1g/cm 3,clayeymudtypically exists
asa poroussolidratherthana fluid. Vaneshearmeasurements (Hwang,1989)suggest that
theorganics-rich
Lake Okeechobee mudis proneto occurin a fluidizedstateat densities
at
leastuptoabout1.065g/cm
3in quiescent
conditions,
although
viscosity
datapresented
here
seemto suggest
theoccurrence
of a structured
phasefordensities
greater
than1.02g/cm
3.
It is unclearif at 1.18g/cm
3 densitythismudis normally (i.e. in theabsence of wave
action)fluidized,
butit is worthexaminingthisissuebycalculating values of •. Thusgiven
H,.= 0.283m,x>= 1.76x10'" m'/s(selectedvalueformodelcalculations) ando = 5.03rad/s
(correspondingto thedominant forcingfrequency of 0.4Hz) yields• = 2.4, whichwould
mean,asnotedpreviously,thatthemudwouldrespond likea "watery"
material.On theother
hand,at the low frequencyof 0.04Hz, • = 0.76, whichwouldmeanthat the mud would
respond
asa "harder"
material
atthisfrequency.
Hencethechosen
modeldescription
maybe
approximate
in thislatterrespect.Nonetheless,
sincein generalthemodelcalculated
mud

accelerationscomparefavourablywith the measured ones,at leastin orderof magnitude,the

descriptionof mudasa fluid seemsacceptable,
althougha betterdescription,
whichrecognizes
the transitionfrom the fluid phase to the solid phase, would constitutea worthwhile
improvement.
Acknowledgement
Supportprovidedby the SouthFloridaWater ManagementDistrict,West Palm Beach(Lake
Okeechobe• PhosphorusDynamics Study), and the U.S. Army Engineer Waterways
ExperimentStation,Vicksburg,MS (ContractDACW39-89-M-4639), is acknowledged.
18.8 References
CoastalEngineeringResearchCenter(1984). ShoreProtectionManual, Vol. 1, U.S. Army Corpsof

Engineers,Washington,D.C. 608p.
DeanR.G. andDalrympleR.A. (1984). WaterWave Mechanicsfor EngineersandScientists.Prentice-

Hall, EnglewoodCliffs, NJ, 364 p.
FodaM.A. (1989). Sidebanddampingof waterwavesovera softbed. Journalof Fluid Mechanics,201,

189-201.
GadeH.G. (1958). Effectsof non-rigid,impermeable

bottomon planesurfacewavesin shallowwater.
Journalof Marine Research,16(2), 61-82.
HwangK.-N. (1989). Erodibilityof fine sediment

in wave-dominated environments.Rept.UFL/COEL-
89/017,CoastalandOceanographic EngineeringDept.,Univ. of Florida,Gainesville,FL, 77 p.
Kirby R.R., HobbsC.H. andMehtaA.J. (1989). Fine sedimentregimeof Lake Okeechobee, Florida.
Rept.UFL/COEL-89/009,CoastalandOceanographic Engineering Dept.,Univ. of Florida,Gainesville,
FL, 77 p.
Longuet-Higgins
M.S. andStewartR.W. (1962). Radiationstressandmasstransport
in gravitywaves,
with applicationto 'surf beat'. Journalof Fluid Mechanics,13, 481-504.
Maa P.-Y. and Mehta A.J. (1987). Mud erosionby waves:a laboratorystudy. ContinentalShelf
Research,7(11/12), 1269-1284.

MehtaA.J. andDyer K.R. (1990). Cohesivesedimenttransportin estuarineandcoastalwaters,In: The

Sea:OceanEngineeringScience,B. LeMehauteandD.M. Haneseds.,Vol. 9, PartB, Wiley-Interscience,
New York, 815-839.
MehtaA.J. andJiangF. (1990). Somefieldobservations

on bottommudmotiondueto waves.Rept.
UFL/COEL-90/008,CoastalandOceanographic Engineering
Dept.,Univ. of Florida,Gainesville,FL,
85p.
RossM.R. andMehtaA.J. (1990). Fluidizationof softesmarine

mudby waves.In: TheMicrostincture
of Fine-grained
Sediment:
FromMud to Shale.Ch. 19,R.H. Bennetted.,Springer-Verlag, NewYork,
185-191.
Sharma J.N.andDeanR.G.(1979).Developmentandevaluation
of a procedure
forsimulating
arandom
directional
second
orderseasurfaceandassociated
waveforces.OceanEngrg.Rept.No. 20, Dept.of
Civil Engrg.,Univ. of Delaware,Newark,DE, 139 p.
Shibayama
T., Aoki T. andSatoS. (1989). Mud masstransport
dueto waves:a visco-elastic
model.
Proceedings
of the 23rd Congressof I.A.H.,R., Ottawa,Canada,B567-B574.
SillsG.C. andElderD. McG. (1986). The transition

fromsediment
suspensionto settlingbed. In:
EsmafineCohesiveSedimentDynamics,A.J. Mehtaed.,Springer-Verlag,
Berlin,192-205.
vanRijn L.C. (1985). The effectof waveson kaolinite/sand

beds. Rept.M2060,DelftHydraulics,
Defft, The Netherlands.
Wells J.T. and Kemp G.P. (1986). Interactionof surfacewaves and cohesivesediments:field
observations
andgeologicsignificance.In: Estuarine
CohesiveSediment
Dynamics,a.J.Mehtaed.,
Speinger-Verlag,
Berlin, 43-65.
WilliamsD.J.A. andWilliamsP.R. (1989). Rheologyof concentrated

cohesive
sediments.Journalof
CoastalResearch,SpecialIssueNo. 5, 165-173.

19
The reflection of waves off beaches

J Darbyshire
Abstract
The reflectionof wavesoff beacheshasbeen given considerable attentionrecentlyas it is

importantin civil engineeringapplicationsand also in academicproblemssuch as the
productionof microseisms. A numericalmethod,based on Lamb's formula, has been
developed to calculatethereflectioncoefficientfor a givenwavefrequencyoff a beachof any
profile. The methodhas been testedin wave tankswith artificialbeacheshaving a wide
range of topographyand the results have always been found to agree well with the
observations.
It wasdecidedto applythe methodto real beachesandusewasmadeof manyobservations

of beachprofilesand the correspondingwave recordstakenby the authorover severalyears
at Hell's Mouth Bay in the Llyn Peninsula,North Wales.
The waveswere registeredby threerecordersplacedat low tide level and situatedat the
comersof a fight-angledtriangle,suchthatthe secondrecorderwasin mostcases10 m away
from the first,measuredalonga line parallelto thebeachandthe otherrecorderwasalso 10m
from the firstbut alonga line normalto thebeach. FromFourierandcorrelationanalysisof
threewaverecordstakenat or nearhightide, it waspossibleto calculateandplot the phase
differencesagainstthe frequencyfor all threecombinations.It wasalsopossibleto calculate
thewavedirection.If thewaveswereprogressive,
at thisdepth(wherec2 = gh,independent
of frequency),the variationof phasedifferencewith frequencyshouldbe linear. In actual
fact, however,the trendline hasa "wavy"characterdue to the effect of reflection.
It is difficultto calculatethereflectioncoefficientdirectlyfrom thesevariationsas it is an

inverseproblem.It is possible,however,to applythenumericalmethodto theobservedbeach
profileand calculatethe resultingvariationsin phasedifference.Ten suchexampleswere
taken. Reasonableagreementwas obtainedbetweenobservedandcalculatedvariations.The
valuesof thereflectioncoefficients for variousfrequencies so obtainedturnout to be quite
high, oftenbeingof the orderof 10-60%.
The method can be extended to two dimensions.

378 J Darbyshire
19.1 Introduction
The reflectionof waves off beacheshas been given considerableattentionrecentlyas it is

importantin civil engineeringapplicationsand also in academicproblemssuch as the
generationof microseisms.A good deal of theoreticalwork has been doneby Bremmer
(1951), Yoshida (1959), Kujura (1961), Wait (1962), and Meyer (1979). Their work is
summarizedby Mei (1982) and its implicationswill be discussed in the next section. The
presentpaperhasa morepracticalapproachand dealswith measurements in a wave tank and
on an actual beach. It is very difficult to measurethe reflectioncoefficientdirectlyon a
beachas can be done with a tank as it is difficult to get controlledconditions.A possible
solutionis to calculatethe reflectioncoefficientfor eachfrequencyfrom the beachcontours
by a numericalmethodbasedon Lamb's formula. It is thenpossibleto work out the effect
of the reflectionon the wavesand comparethis with what is observed.
19.2 The reflectionprogram

This is basedon the formulagivenby Lamb (1932). Art 176. He showedthat when a wave
of amplitudea• in a channelof widthb• anddepthh• entersa channelof widthb2anddepth
h2,thenthe reflectioncoefficientis givenby
(blC
1 - b2c2)l(blC
1 + b2c2
) (19.1)
and the transmission coefficient is
2blCl/(blC
1 . b2c2
) (19.2)
where
61,2
= (•,hl,2)l/2 (19.3)
Lamb had somereservations aboutthe accuracyof the formulabecauseof the rapid
variationsin depthand width at the pointof discontinuity.
If the width is constant,the reflection formula reducesto:
(C1 - C2)](C
1 + C:•) (19.4)
Thisformula
was derived
forlongwaves(wherec = d'gh),by equating
thefluxonboth
sidesof the boundaryand equatingthe displacement.By this meansthe sameformulacan
be extendedto any gravitywave. However,in the longwave case,the rateof passageof
energyacrossthe boundaryis alsoconserved whereasthisis not so in the generalcaseas the
groupvelocityvarieswith depthin a differentmannerfrom thephasevelocity. Even so,as

J Darbyshire 379
theresultswill show,the formulaappliesover caseswherethe ratioof depthto wavelength

approachesthe deepwatercondition.
In thispaper,the formulais appliedone-dimensionallyovera successionof very small

intervalsovera variableprofile. Thisgetsoverto someextentLamb'sobjections
asthereare
no suddenchanges.
If theoriginisreferred
tothemeasuringposition,
xøreferstothedistance of thewater-line
fromtheorigin,andx• refers toanypointbetween them,thenthereflected waveat x• will
becontributedto byall positions
fromit to thewater-line
xø. Thephase of thecontribution
will dependon thetimetakenby thewaveto travelfromx• to x andback. Expressed
mathematically andusingtheformula,
thenthereflected waveat x• will begivenby:
(19.5)
wherec andk, phasevelocityandwavenumber,are functionsof the depthwhichis in mm

a function of x.
Theintegrationcanbecarriedoutnumericallyforeachvalueof x• between
0 andxøand
thusgivesthereflection
coefficient
atallpoints
between0 andxø. A computerprogramwas
preparedto do this.
In the work describexl

by Mei (1982), the mild slopeconditionis assumed
whichirapies
that:
(ah/Ox)
/ kh<<1 (19.6)
and the bottomprof'deis denotedby:
h=gx wherep dependson x and is <<1 (19.7)
Under theseconditions,by a perturbation

method,the relevantequationscanbe solvedto a
firstorderin g andan expressionis foundfor thereflectioncoefficient.
(19.8)
For the shallowdepthcondition,the integrandis identicalwith thatin equation(19.5), the

only differencebeingthatthe termexp2iS/l•is basedon an averagevalueover the rangeof
integration,whereasin (19.5), thisreplacedby a secondintegrationoverthe samelindts.

380 J Darbyshire
A furtherapproximation is madein assuming that the depthvarieslinearlywith x with a

very small gradientalthoughwith the shallowdepth conditions(19.6) and (19.7) seem
sufficient. They then obtainby usingthe deltafunction,a valueof R for a discontinuous
changein depthwhichturnsout to be thesameas equation(19.4) with c• • c,_.For thecase
of a linearslopeovera distanceL, havinga meandepthh anda changeA betweenthe ends,
they found
R - -as•nk/d
(4hk/.) (19.9)
wherek is the wave numbercorresponding to the depthh. This expression

canbe derived
directlyfrom equation(19.5) for the conditionsassumed.
In derivingtheequations,
conditions
(19.6)and(19.7)havebeenassumed. Equation(19.5)
is derivedfrom moreelementaryconsiderations and no conditionis immediatelyapparent
besidesthe shallowdepth one. However, theseconditionswill be borne in mind when
discussing theresults.As equation(19.8)is correctto a firstorderin g, a value=of g less
than 0.33 shouldgive an accuracyof 10% or more.
19.3 Applicationto tank experiments

The reflectionprogramwastestedwith artificialanalyticalprofilesinsidewavetanksin the
laboratory. The first exampleshownin Figure la refersto data given by Daviesand
Heathershaw (1983)anddealswitha sinusoidalbottomconsisting of 10wavesofwavelength
1 metreand amplitude5cms. The frequencyof the waterwaveswas adjustedby the
wave-maker so thatfor the depthused,the waterwavelength wastwicethe bottomripple
wavelength.b is thebottomrippleamplitude(5cmsin thesecases),
andd is thedepthfor
eachexperiment,andthereflection
coefficient
hasbeenobtainedexperimentally
for theseven
casesshownandare indicatedby thepoints. The curvegivestheresultsobtainedby the
programand showsvery close agreement. This is the case even when the ratio of d/•
approaches
the deepwatercondition. Thusthe two caseswith the lowestvalueof b/d, 0.08
and 0.10, give good agreementalthoughthe wave periodswere 1.17 and 1.20 secs.
respectively
corresponding
to depths
of 62.5cms
and50cmsandgivingvaluesof d/)tof 0.306
and0.245 whichapproachthe deepwaterconditionwhichis oftentakento be reachedwhen
d/•, is 0.25. Thedatausedin Figureslb, c andd, wereobtainedby Drs.M. Belzons
andE.
Guazzelliof the Dept.of Physics,CentreSt. Jerome,Universityof Provence,
Marseilleand
have not beenpublished.

381
J Darbyshire
0.9 VARIATION OF REFLECTION COEFFICIENT WITH b/d
0.8
0.7
0.6
0.5
0.4
0.2
O. 1
b/d
Figure1a: Comparison
of observed
andcalculated
reflection
coefficients
fora sinusoidal
bottom
profile.
Here thebottomtipplesconsistof a mixtureof two sinewavesas shownon the diagrams.

In thesecasesthe measuredreflectioncoefficientsare shownby the pointswhilst the curves
aregivenby thetheory.Theagreeement is particularlygoodforFigurelb. Figurelc shows
notsogoodagreement for the2.5cmdepth.Theagreement is notsogoodfor Figureld but
here,probably,a smallerintegrationintervalwouldhaveimprovedmatters.In Figurelb,
thereis goodagreementfor3.25Hz. and0.04mdepthcorresponding toa wavelength of 0.14m
and h• = 0.286, closeto the deepwater condition.

382 J Darbyshire
REFLECTION OF WAVES FROM A BIFREQUENCY BEACH PROFILE
0.8_
8.7 _
8.5_0 ß
d=2.5 c m•.
0.4_ '
ßß
0.2
0.1
0.8_
0.7
1 2 5
Iß ' ß 4
0.5 •
0.4
0.2
•
mmmm
ß ß
'I d=5.0
cms.
0 1
1 2 • 4 5 -Fr'ect. Hz
0.6
8.5 _ o d= 4.g) c ms.
0.4 - ß ß mm
_ •
0.2 _
0.1
I I ! • I I I I
1 2 5 4 5 •req. Hz
dep!:h=d+sin<&.28518x/12)+sin(&.28•18x/6)• dmx cms.
•' ......... 48 cms. •
Figurelb: Comparison
of observed
andcalculated
reflection
coefficients
forthebifrequency
bottom
profile shown.

J Darbyshire 383
REFLECTION OF WAVES FROM A BIFREQUENCY BEACH PROFILE
0.9
0.8
0.?
0.6
imm
m
0.5
0.4
0.2
0.1
' .
d=2.5cms.
0.8
8.7
0.6
/ ,
0.5
0.4
• • ,•] d=5.0
cms.
0.2
0.1
3 II 4 5 •.e• Hz
0.6
0.5
0.4
0.5
-
0.2
0.1
i 2 5 4 • •re•.
H••
Figure l c: Comparisonof observedand calculatedreflectioncoefficientsfor the bifrequencybottom

profile shown.

384 J Darbyshire
REFLECTION OF WAVES FROM A •IFREQUENCY •EACH PROFILE
0.5 o d=2.5 cm•.
0.5
0.2
t4' mm
m ßßß '
0.1 _
I I I I I
0.8
5 -Freer.•-74
0.7
0.•
i
I
0.5 o d=3.0 cms.
ß
0.4 _
0.5 _ •.'
0.2
0.1
Figure l d: Comparisonof observedandcalculatedreflectioncoefficientsfor the bifrequencybottom

profile shown.

J Darbyshixe 385
Table 1: Gradientandgradienff(kh)
for tankexperiments.
h --mean depth (m) , G -- r.m.s. gradient

f -- frequency
Hz. , k - wavenumber
(m'!)
Table1 showsthevaluesof ther.m.s.gradient and(3h/3x)/khfor theseexperiments.Those

shownin Figurela just obeyconditions (19.6) and(19.7). In thecaseof Figureslb, lc and
ld, the r.m.s.gradientssatisfycondition(19.7) but in no caseis theratiolessthanunityby
an orderof magnitude.The ratio is only lessthanunity for all threediagramsfor 1 Hz.
frequency.It is alsolessthanunity for 2 Hz. in Figureslb and lc. The closeagreement
betweenthe calculatedand observedvalues,particularlythosein Figure lb, suggeststhat
condition(19.6) is too stringent.
Taken as a whole, the resultsare satisfactoryand one can use the programwith some
confidence with real conditions as found on a beach.
19.4 Applicationto beachconditions

With waveson beachesthere is an importantdifferencecomparedto the tank experiments
whichhavebeendiscussed as the depthbecomeszeroat theend. A directapplicationof the
formulaover a region containinga zero depthwould give an infinite answerso the method
has to be alteredto allow for the breakingof wavesin shallowdepths.
In thebreakingzone,it is usuallyassumed thatthewaveheightis proportional

to thedepth.
If we takethepointof measurement asorigin,thexbgivesthepositionof thebreakingpoint,
Xothatof the still waterline andx thatof any other point. The depthd betweenxband x0
will be assumed to be:-

386 J Darbyshire
a- (19.10)
where• is the depthat the breakingpoint.
n canhaveany real valueand determines

the shapeof the profilebetweenXoandXb.
The waveheightin thisregionthenbecomesproportional

to d andsothereflectionformula
for this regionbecomes,ignoringthe phaseterm:
(19.11)
f:•?
(1/2c)
(Xo
-x)"d•/(Xo
-x•,)"
(19.12)
. "=
•ldx - - O.•(gdDua
(Xo-X)"n-•l
(xo-x•,)"n (19.13)
Ol:c)acl& - -n/4(Xo-X) (19.14)
Thus the integralbecomes
(19.15)
-n/4
f,:?
(Xo-X)"-•dxl(Xo-Xt,)
"- 1/4
The value1/4 is thusindependent
of n i.e. of the shapeof the profilein thebreakingzone.
If the zonecoversseveralintegrationsteps,thenthe value 1/4 hasto be dividedbetween

themandmultipliedby the appropriate phasefactorsdetermined in theusualway. In our
case,however, thestandard
program withnostepsistakenandextended tocovernø+1 steps,
the(nø+1)th.stepbeinggivena modulus of 0.25andthisismultiplied
bythephase factorfor
thisdistance.Thevaluex• is takentobethenoth.valueandall thebreakingis assumed to
takeplacebetween thenøth.andnø+1 th.steps.
19.5 Site of wave measurements
The shoreobservations werecarriedoutat Hell's MouthBay in theLlyn Peninsula.Figure

2. Thisbayfacessouth-west andreceives thefull onslaught
of thewintergales.The beach
hasnotbeentouched by manandall thechanges thathaveandaretakingplaceareentirely
due to the effectof the seaon the coast. Thereis extensivebeacherosion,the recession
of

J Darbyshire 387
the coastlineaveragingabout0.3 metresper year.
HELES MOUTH BAY
LLYN
PENINSULA
........ IOfIr
........... 2Of•
i I
Figure2: Map showingsiteof wave andbeachmeasurements.
The nameis foundon theearliestmapsandreflectsthefact thatthe old sailingshipswere

driventhereby seaandwindandwereoftenstranded.The beachhasbeenregularlysurveyed
since1972andfor a greatdealof thattime,fortnightlysurveyswerecarriedoutat low spring
tides. For most of the period 1975-86, the waveswere monitoredby a wave recorderand
during1977-1981therewerethreerecorders whichenabledthedirectionaswell asthe height
and period,to be assessed.The site of the wave recordersis shownby the point C on the
map but the scaleis too small to showthe individualrecorders.

388 J Darbyshire
AUGUST 11 1977
ß 180
"" 160
• 140
• 1:20
• 4•
• 2•
'• -28
0.08
_ 0.0•
_ 0.04
0.02
100
•e.•s
e0.4
o
4 6 8 10 12 14 l& 18 20 22 24
! I I I I I I I I I I I ! I I I I ! I !
1.5
1.8
5 10 15 20 25 :10 55 40
Figure3a: Variationof observedandcalculatedphasedifferences betweenwaveson (2) and(1), and

(2) and(3) andthecalculated
reflectioncoefficients
againstthefi'equency
for August11 1977.Thewave
power spectrumand the beachprofile are shown.

389
J Darbyshire
JULY 9 1979
ß, 8o
•lTI9*
OF
• ½1)
2--3
•&• r18
.0.20
0.12
o.
04ß re<t. x 100
•
o
-:0.2
•2 • •4
4.0
5.0
10 20 50 40 50 60 70 80 90 100 110 120
Figure3b:Variationof observed andcalculated phasedifferencesbetween

waveson (2) and(1), and
(2) and(3) andthe calculated
againstthefrequency
for July 9 1979.The wave
powerspectrum
andthebeachprofileareshown.

39O J Darbyshire
$EPTE•E• 1• 1•?•
2-1
'"' 1,•0
-,= 140
o
• 120
c;•) lem. ½2,)
•- 40
ut
0.10
0.06
0.04
0.0:2
ß •-e<•. x 100
0 '
.,-:0.2
4.0
ii1/111
10 20 :50 40 50 60 70 80 90 lee 110 120
Figure3c: Variationof observedandcalculatedphasedifferencesbetweenwaveson (2) and (1) and

the calculatedreflectioncoefficientsagainstfrequencyfor September13 1978. The wave power
spectrumand the beachprofile are shown.

J Darbyshire 391
OCTO•œ• •0
ß-. 80
0.10
I
•0.08
_ 0.0•
.0.04
, '-'"'"
14 mi• 0m 72i ?4
•e,a.. x 100
•e.6
e04
0 "
le 2e •e 4e 5e 6e ?e ee 9e lee lie lze
Figure3d: Variation
of observed
andcalculatedphasedifferences
between
waves
on(2) and(1), and
reflection
coefficients
againstthefrequency
forOctober
30 1978. The
wave powerspectrumand the beachprofile are shown.

392 J Darbyshire
'"" 1 •0 -
-,=, 140 _
o
a, 120 -
e- 40
.,,..e
0.15
•0.4
o
ø0ø2
Figure3e: Variationof observed andcalculatedphasedifferences

between waveson (2) and(1), and
reflection
coefficients
against
frequencyfor December
7 1978. Thewave
powerspectrum

J Darbyshire 393
OCTOBER 20 1980
140
120
•- 40
_0.12
_ 0.88
_ 0.04
e 32
•r-e<t. x 100
,,: e.6
o • 0
2.0
1o 20 •o 40 50 6o 78 80 9o lOO 11o 12o
Figure3fi Variationof observedandcalculatedphasedifferencesbetweenwaveson (2) and(1), and

(2) and (3) and the calculatedreflectioncoefficientsagainstthe frequencyfor October20 1980. The
wave power spectrumand the beachprofile are shown.

394 J Darbyshire
NOV•M• • 1•80
'"' 1•,0
• 140
o
• 100
q- 40
ß 20
0.10
0.08
0. O&
0.04
0.02
I i4._ , E• I ?2 , i•4
•r-e<l. x 100
'0ø2
4.e .
10 20 :•0 40 50 &O 70 80 90 100 110 120
Figure3g: Variation
of observed
andcalculated
phase
differences
between
waves
on(2) and(1),and
reflection
coefficients
against
thefrequency
forNovember
5 1980.The
wavepowerspectrum

J Darbyshire 395
FEBRUARY ? 1981
•2-t
"" 240
'• 218_
o
u
• 188 -
• 158 -
• 128
.,, :/;"'",,•,,,-,,,/
/-. Zlq•Zll•qC:8 -t t •.
r1.0
, 92
,/•, •;20 , ;2:2,
2-3
;24
0.88
8.84
, • , ,• , ? , ,•e, , , , ?e, •2 i •4
-Freq.. x 180
•84
0 '
ø8ø2
t4
•'•- 6I I I I I I i I I I 8,
4.0 -
10 20 50 48 50 60 78 80 90 180 110 120
Figure3h: Variationof observedandcalculated phasedifferencesbetweenwaveson (2) and(1), and

(2) and(3) andthe calculated
againstthe frequency
for February7 1981. The
wave power spectrumand the beachprofile are shown.

396 J Darbyshire
FEbRUaRY ? 1781
140
• 120
,• 100
ß". 80
!
'" 60
,q- 40
ß 20
19.25
0.20
0.15
_0.10
_ 0.05
I I I I I I I I I •2 I ITM , 0 , •22 , i:24

ß re•. x 1190
a•0.4
o
Figure3i: Variation
of observed
andcalculated
phasedifferences
between
waves on(2) and(1), and
reflection
coefficients
against
thefrequency
for February
9 1981.The
wave powerspectrumand the beachprofile are shown.

J Darbyshire 397
MA•CH 5 1981
240
210
180
150
120
80
&O
_ 0.08
_ 0.06
0.82 x
I I I I6 ! 8
I I I10 I I12 I I14 I I16 I I18 I •0 I •2• •4
4.8
5.0
2.8
1.8
m/m
18 2.8 •0 48 58 68 70 80 •0 100 110 128
Figure 3j: Variationof observedand calculatedphasedifferencesbetweenwaveson (2) and (1), and

(2) and(3) andthecalculatedreflectioncoefficients
againstthefxequency
for March5 1981. The wave
power spectrumand the beachprofile are shown.

398 J Darbyshire
_?
(•)
_6
;24
_7'
<:b) 9.?.79
_6
_5
_4
_2
-I•r' eq.. Hz. x 100

(c) 9.2.01
_5
_4
4 '""*'"
& 14 ' 16 18 20 22 24
I I I I I I I I I I I I I I I I I I I I I
Figure4: Comparison of observed andcalculatedvariationof theratioof powerspectral

densities of
waveson (1) to thoseon (2), againstthefrequency.(a) For the samedataas Figure 3a. (b) For the
samedataasFigure3b. (c) For the samedataasFigure3i.

J Darbyshire 399
19.6 The instrumentsand their lay-out

The threerecorderswere alignedin the form of a right-angledisoscelestriangle,Figures3,
so that the line joining the recorder2 to the outermostrecorder1 was perpendicular to the
beachline. The line joining recorder2 to recorder3 was parallel to the beachline. In all
casesbut one, (Figure 3a), the lengths 1-2 and 2-3 were 10 metres. The recorderswere
situatedabout120 metresseawardof the cliff edgeandwere exposedat low fide. The wave
recorderswere pressuretransducers.The responses were digitizedand sentby a cableto a
shorestationabout200 metresbeyondthe cliff edge. The maintenance of the recorderswas
very difficult as they had to be left unattendedfor long periods. On one or two occasions,
high seascarriedthe wholesystemup to the shoreand the systemhad to be relaid. A more
commonhazard,however,was for the recordersto get buriedunderabout2 metresof sand.
In a way this was the leastof the evils as the unitswere thenquite safebut obviouslytheir
sensitivitywas affected. In analysingthe results,therefore,the amplitudeswere considered
not to be soreliableandthe work wasconcentrated moreon thephasedifferencebetweenthe
recorders.
The digitizedoutputof thethreerecorders

werefed simultaneouslyinto the shorereceiver
and recordedon magnetictape (5 hole papertape in the caseof the earlier records). The
resultswerereadby a micro-computer and thenanalysedby a FastFourierprogramwhich
gave the power spectrafor the three outputsfor a period of 512 or 1024 seconds.The
corresponding co- and quadraturespectrawerealsoobtainedbetweeneachpair of unitsand
thesewere reducedto valuesof coherence ratio and phasedifferencefor eachfrequency
interval. The unit frequencyintervalover whichthe spectralestimateswere calculatedwas
0.02 s'• frequency in thecaseof 512 secsandhalfthisfor 1024secs.With themethodof
analysisused,thiscorresponded to about20 degreesof freedom,allowinga statisticalrange
of about 3 to 1 on the 5% limits. The variation between consecutive values is nevertheless
muchlessthanwould be expectedfor a randomscatter.
Referringto Figures3, if lij is thedistance

between
recorders
i andj andL is thelocal
wavelength, and• is theanglebetweenthewavedirectionandtheperpendicular to thebeach
line thenthephasedifferencebetweenrecorders
i andj is proportional
to theprojectionof line
[j on a lineparallelto thewavedirection
soif i=1 andj=2 then
•2' 3601•2cøs0/L (19.16)
and if i=2 andj=3:
3601:•3sia0/I, (19.17)
In shallow
waterthevelocity
becomes
4'ghwhereh isthedepth.
If f isthefrequency
then

400 J Darbyshire
(19.18)
and the formulaefor the phasedifferencesbecome:
(19.19)
•2' 360112fcøs01
g• aud •23
Thusthephasedifferencebecomes
directlyproportional
to the frequency.The direction
can be found from the two phasedifferences
•231
•)12- (123/112
)t/triO (19.20)
Thusknowing
theIsandthe•)s,canbefoundandthen4'ghobtained
fromboththephase
difference formulae.
In thestraightforwardcasewherethewaveis progressive, theseformulaecompletelysolve

the problemfor any frequency.When thereis reflection,however,we do not get a straight
line but patternsas shown by the full lines on Figures 3, which show a wavy trend
superimposed on a linear one. In the extremecase,wherethereis total reflectionand the
wave is completelystationary,formedpossiblyby reflectionoff someeminencenear the
water-line,therewould be nodesand antinodessetup whosepositionwould changewith the
frequency.Thusat one frequencytwo recorderswouldbe on the samesideof a nodeand
havezerophasedifferencewhereasat anotherfrequencytheywouldbe on oppositesidesand
havea phasedifferenceof 180ø. The phase-frequency diagramwouldthusshowa seriesof
jumpsbetween0ø and 180ø. In the real examplesin Figures3, thereis no perfectreflection
but we have a systemapproaching the stationarywave pattern.
In practice,however, it is easier to deal with the problem by calculatingthe phase

difference
between
tworecorders,
for a progressive
wave,usingthevaluesof 4'ghandO
found from the meantrends. The observedbeachprofile is then adjustedto give the fight
depth at the recorderpositions. The amplitudeand phaseof the reflectedwave at both
recorderpositionsare then calculatedfrom the reflectionprogram. The reflectedwavesare
thencombinedwith the straightprogressivewavesto give a resultantwave at eachpositon.
The phasedifferencebetweenthe two shouldcorrespond to the observedphasechange. As
thereflectionprogramassumes normalincidence,if it is notzerotheintegrationintervalmust
be dividedby cos.O. This assumes someuniformityalongthebeachbut thismaynotalways
19.7 Discussion of results
The resultsare shownin Figures3a.....j. In eachcasethe top part of the diagramgivesthe

observedvariationsin phasedifferenceagainstfrequencyin full lines. The predictedphase

J Darbyshire 401
differencesare givenby the brokenline. The secondpart givesthe powerspectraldensities.

The third part givesthe predictedreflectioncoefficientplottedagainstthe frequency.The
bottompart givesthe observedprofile,referredto the sealevel at the time of recordingand
the positionsof recorders1 and 2.
Figure 3a is the most strikingexample. This examplewas differentfrom the othersin that
the waverecorderswere installedspeciallyonly 25-30 metresfrom a sharprise in the beach.
This was the closestapproachto a laboratorytankexperiment.The resultsshowa classical
caseof a wavyphasedifference frequency variation.In thiscase1•2was5 metresand123 was
10 metres. The reflection coefficientvaries from 25% to 60%. In all the other cases,both
1•2and123are 10 metres. In thesecasesconditions werenot so idealisticas in Figure3a as
the recorders had in manycasesbeeninstalledsometime before. The beachsurveyswere
cardedoutat low springtidesothewaveobservations couldnotbe cardedoutat exactlythe
sametime and therewas usuallya differenceof one or two tidal cycles,so that therecould
be somechangesin the beachprofile. This couldexplainthe constantphaseshift foundin
Figures3d and 3g. Nevertheless in mostof the ninecasesthe resultsare satisfactory.In
Figures3c, 3f, 3g, the wavesare movingnormalto the beach. The agreementis goodfor
Figure3c and Figure3f. In Figure3g, theactualvariationsin thepredictedvaluesagreewith
thoseof the observedvaluesapartfrom the constantphaseshift. Figures3b and 3d showan
angleof approach of nearly45øto theshoreline. The variationsof thepredictedvalues for
Figure3d, like thoseof Figure3g agreewith thoseof the observedonesbut againthereis a
constant phaseshift. The agreement for Figure3d is not unsatisfactory. Figures3i and 3j
showexampleswherethe wavedirectionis about20øoff the normal. Agreementwith both
is satisfactory.A mostremarkablecaseis shownin Figure3e. Here the wave directionis at
70øto theshoreline. In thiscaseno observed profilewasavailablebuta constant depthwas
assumed in thedirectionof 123andthena suddenriseabout40 metresbeyondposition3, (on
the other side to position2). On this basis,someagreementcan be obtainedwith the
observed results.
In mostof thecases,thereflection
coefficients
varyappreciablywith thefrequency,
from
10% to 60%. The 60% valueis reachedin Figures3a, 3e, 3f and3g.
Table2 showsthevaluesof themeangradientand(i)h/i)x)/khfor Figures3a, 3b, 3c, 3d,

3f, 3g, 3h, 3i and 3j. In all thesecasesconditions(6) and (7) are met.
The reflectioneffectshouldalsocausea variationin the wave amplitudeat the three

recorders.Theratioof thespectral powerdensities for eachfrquencyfor thewavesrecorded
at (1) to thoseat (2) areshownin Figure4. The reflectionprogramcancalculatethisratio
andthecalculated resultsarealsoshownfor threeexamples in Figure4. Theseexamples
werechosen because at thesetimes,thesensitivity
of recorders (1) and(2) were nearlyequal.
The first exampleis for 11.8.77wherethereis goodagreement betweenthe observedand
calculated frequenciescorresponding to maximum andminimum ratios.The otherexamples
arenotquitesogoodbutarereasonable. Thecalculated ratiosarehigherthantheobserved
ones.Thiscanbe attributed to someextentto thecoherence between wavesat (1) and(2) not

402 J Darbyshire
being 100% but usually75-95%.
Table 2: Gradientand gradient/(kh)for beachobservations.
h = mean depth (m), G = mean gradient

f --frequency Hz., k = wave number (m
h G
Fig. 3a 1.0 0.047

Fig. 3b 2.5 0.025
Fig. 3c 1.5 0.017
Fig. 3d 3.0 0.033
Fig. 3f 2.5 0.033
Fig. 3g 1.5 0.025
Fig. 3h 1.5 0.017
Fig. 3i 2 5 0.033
Fig.3j 110 0.015
19.8 Applicationto two dimensions

It is notthepurpose
of thispaperto dealwithtwo-dimensional
casesbutsomesuggestions
canbe madehowto extendtheprocess.The mild slopeequation (Williamset al., 1980),
when solved over an area does include the effects of reflection and indeed a one dimensional
form of it hasbeenappliedto thecaseof Figurela, andgivensimilarresults.With a two
dimensional case,however,its solutionpresents somedifficulties.A progressivewaveis
usuallyinputat the openendbut if thereis reflection,thisinputis incorrectandleadsto
errors. A plot of the phasesandto a lesserextentthe amplitudes acrossa longtitudinal
sectionfromthe opento the closedendwouldshow'wavy' variations like thoseshownin
Figures3. If thesewere smoothed out, the resultsfor a purelyprogressivewavewould
remain.If, furthermore,theprogressivewavevaluesfoundby thismethodfor theclosedend
werereinputat thatend,thefinal resultwouldbe reasonably accurate
andshowreflectionat
the openend.
Anotherpossible
methodof mmckis to extendthereflection
program to coverthewidth
b aswellasthevelocitysothatbcratherthanc becomesthevariable
parameter. A separate
waverefraction
programis constructedwhichstoresthe trackof an initiallyparalleland

J Darbyshire 403
equidistant setof raysovera seriesof pointsandgivingandstoringthevelocityat eachpoint

as well asb theratio of the perpendicular
distancebetweentwo adjacentraysand the original
separation.The reflectionprogramcanbe modifiedto readthesevaluesandto carryout the
numericalintegrationalong the ray paths,usingthe appropriatevaluesof b and c. The
programalso usesa similarprocedurewith the transmission coefficientbasedon Lamb's
value. To a firstapproximation, ignoringdiffraction,therecanbe no transferof energyacross
the ray pathswhichcanbe regardedas solidboundaries.As the wavelengthwill usuallybe
comparableto the other lengthsinvolvedand much greaterthan the depth, the problem
becomesidenticalwith thatof theresponse of baysandinletsto longwavesandtheray paths
correspond to the "talweg". The main differenceis that in the responseof bays, there is
assumedto be completereflectiongivingrise to a stationarywave whereasin our casethere
is only partialreflection. The valueof b will vary from quite largevaluesto zero and then
large negativevalues. The effect of very small and very large valuesis largely offset by
breakinganddiffractionandtheunderlyingassumptions regardingtheray pathsno longerhold
so in the programthe signsare ignoredbut the absolutevaluesarekeptbetweenlimits which
are usuallytakento be 0.5 and 2. Apart from theselimitations,the procedureis exactlythe
same as for waves moving up a channelof variable depth and width. When wave ray
channelsintersect,then the wave activity is summedup over the point of intersection,
allowing for amplitude,phaseand direction. On this basis,it is possibleto calculatethe
reflectioncoefficientsat eachpointas well as the resultantamplitudeand phase.
Sucha programhasactuallybeenusedwith somesuccess

but at themomentthereare only
very few setsof simultaneous
beachtopographical
and wave observations
availableto check
the program.
There is an interestingconsequence
of thismethod. The reflectedwaveswouldretracethe
pathof the incidentwavesand so in theorya straightincidentwavefront,althoughgreatly
modifiedby refractionin shallowwater,would still ultimatelygive rise to a reflectedwave
in the openseatravellingin an oppositedirectionto the incidentwave,althoughthis would
be relativelysmall.
Sucha resultmightprovidean explanation for theformationof microseisms.According

to Longuet-Higgins (1950), microseisms are causedby the interaction
of wavestravellingin
oppositedirections. This could happeneither in mid-oceanwhere waves generatedby
opposingwindsin a singlestormor in two stormscan meetor by reflectionoff coasts. An
exhaustivesurveyof microseisms generatedduringthewinterof 1987-88,Darbyshire(1990),
showedthat therewere 20 instances of largemicroseisms beingformed but thesecouldall
be attributedto coastalreflectionandduringthisperiodthemid-oceanconditions werenever
met. One must therefore conclude that the formation of microseisms in mid-ocean is a rather
rare event and that coastalreflectionis the main cause. Most of the coastsinvolved, however,
areveryindentedlike thecoastof Norwayandit is difficultto seeat first sighthowtherecan
be much reflection.

404 J Darbyshire
19.9 Conclusions
A method of evaluating the reflection coefficient for any wave frequency for a
one-dimensional profile is developed.It hasbeenappliedsuccessfully with bottomprofiles
in tanks. The methodcan be appliedto real beachesby usinga triangulararrangement of
wave recorders. In the casesdealt with in this paper,somevery interestingresultswere
obtained,indicatingthat for somefrequencies,thereflectioncoefficientin somecasescanbe
as high as 60%.
The one-dimensional
approachcan be extendedto two dimensions.
Acknowledgments
The author wishes to thank Mrs. Eileen Pritchard and the late Mr. W.R. Davis for theft
invaluableassistance
throughoutthe yearsin obtainingthebeachprofilesandmaintainingthe
pressurerecorders. My thanksare also given to Drs. Belzonsand Guazzelli for the useof
their data.
19.10 References
Belzons,M. and E. Guazzelli,1989: (personalcommunication).
Bremmer,H., 1951: The WKB approximation

as the first termof a geometricopticalseries. Comm.
Pure Appl. Math., 4, 105-115.
Darbyshire,J., 1990: Analysis of twenty microseismstormsduring the winter of 1987-88 and

comparisonwith wave hindcasts.Physicsof the Earth andPlanetaryInteriors,63, 181-195.
Davies, A.G. and A.D. Heathershaw,1984: Surfacewave propagationover sinusoidallyvarying

topography.Instituteof Oceanographic
Sciences,J. Fl. Mech., 144, 419-443.
Kajiura,K., 1961: On thepartialreflectionof waterwavespassingovera bottomof variable depth.

Proc.TsunamiMeetings10thPacificScienceCongress, I.U.G.G. Monograph,24, 206-234.
Lamb,H., 1932:Hydrodynamics.CambridgeUniversityPress,738 pp.
Longuet-Higgins,
M.S., 1950: A theoryof the originof microseisms.
Phil. Trans.Roy. Soc.Lond.,
A243, 1-35.
Mei, C.C., 1982:The AppliedDynamicsof OceanWaves. Wiley Interscience,

740 pp.

J Darbyshire 4O5
Meyer, R.E., 1979: Surface-wavereflectionby underwaterridges. J. Phys.Oceanogr.,9, 150-157.
Ogawa,K. andK. Yoshida,1959: A practicalmethodfor a determination

of longgravitationalwaves.
RecordsOceanographic
WorksJap.,5, 38-50.
Wait, J., 1962: Electromagnetic

Wavesin StratifiedMedia. Pergamon,New York.
Williams,R.G.,J. Darbyshire, andP. Holmes,1980: Waverefraction

anddiffractionin a causticregion.
Proc.Inst. Civ. Engrs.,Part 2, 69, 635-649.

Part V
Modelling (Sediment)
2O
Dispersion in tidally-averaged transport

equation
Ralph T Chengand VincenzoCasulli
Abstract
A generalgoverninginter-tidaltransportequationfor conservative soluteshasbeenderived

withoutinvokingthe weakly nonlinearapproximation.The governinginter-tidaltransport
equationis a convection-dispersion equationin which the convectivevelocityis a mean
Lagrangian residualcurrent,andtheinter-tidaldispersioncoefficientis definedby a dispersion
patch. When the weaklynonlinearconditionis violated,the physicalsignificance of the
Stokes'drift, as used in tidal dynamics,becomesquestionable.For nonlinearproblems,
analyticalsolutionsfor themeanLagrangian residualcurrentandfor the inter-tidaldispersion
coefficientdo not exist,theymustbe determined numerically.A rectangular tidal inlet with
a constrictionis usedin the first example. The solutionsof the residualcurrentsand the
computedpropertiesof the inter-tidaldispersioncoefficientare used to illuminate the
mechanisms of the inter-tidaltransportprocesses.Then,the presentformulationis testedin
a geometricallycomplextidal estuary-- San FranciscoBay, California. The computed
inter-tidal
dispersion
coefficients
arein therangebetween
5x10• and5x10a cm2/sec.,
which
are consistentwith the valuesreportedin the literature.
20.1 Introduction
One of the underpinningmotivationsfor marine scienceresearchis to understandand to

quantifyprocesses thataffectthebalances in marineecosystems.In tidalestuaries,bays,and
shallowcoastalwaters,the transportof dissolvedand suspended substancesis determinedby
theprincipleof conservation. The governing conservation
equationaccounts for thebalances
betweenadvection,localturbulentmixing,andthe sourcesandsinksin a controlvolume. The
relative importanceof these transportfluxes determinesthe mechanismsthat control the
long-termconsequences of the transportprocess.As mostbiochemicalprocesses take place
over a time intervalof severaltidal periodsor longer,the basicissuesconcerning transport
processes mustaddresssolutions of theconservationequationoveran extendedperiodof time.

410 R T ChengandV Casulli
Studiesof conservative solutetransportarefundamental; the sourcesandsinksareconsidered

knownor non-existing.The transport of non-conservativesolutesis morecomplex.Although
the ultimategoal is to understand the transportof both conservativeand non-conservative
substances overan inter-tidaltime span,withoutthe basicandthoroughunderstanding of the
inter-tidaltransportof conservativesolutes,solutionsto the transportof non-conservative
mattersare at best qualitative. Thus, this researchfocuseson improvingthe present
knowledgeof long-termtransportof conservative solutes,suchas salt, in a tidal estuary.
Future researchwill be expandedto include the transportof fine suspendedsediments,
nutrients,andultimatelytheprocesses concerning non-conservative matterssuchasplanktons
or reactive chemicals.
Withoutconsidering the sourcesor sinks,the transportof a dilutesolutein a tidal system

becomesthebalancebetweenthe convectiveflux dueto tidal currentsand the dispersive flux
due to local turbulentmixing. Using the characteristic dimensions of San FranciscoBay,
California,an orderof magnitudeestimatesuggests thatthe turbulentdispersive fluxesare at
leasttwo ordersof magnitudesmallerthanthe tidal currentconvectiveflux. This assessment
is probablyvalid for a largenumberof tidal estuaries, bays,andcoastalseas. The resultsof
thisestimateimply that the intra-tidaltransportprocesses are usuallyconvectiondominated.
Thus,the useof a Lagrangianmethodin analysisof suchprocesses is advantageous. Using
a perturbationtechnique andassuming thetidalsystemis weaklynonlinear, Fenget al. (1986)
haveshownthat the ti_da!!y-avemged transportof solute,at the first orderapproximation, is
a pure convectionprocessin whichthe convectivevelocityis the masstransportvelocity
(Longuet-Higgins, 1969)or thefirstorderLagrangian residualcurrentasusedby Chenget al.
(1986). In the secondordertheory,the Lagrangianresidualcurrentsform an ellipsein the
hodograph plane(Chenget al., 1986). Thesetheoretical developments havepointedout some
propertiesof theLagmngianresidualcurrentanditsrolein thelong-termtransport processes.
Similarinvestigations havebeenreportedin atmospheric literature.Dunkerton(1980) gave
a goodreview of the relatedresearch,and Middletonand Loder (1989) havereachedsimilar
conclusions in weakly nonlinearanalysisfor smallamplitudewaves. Within the limits of
weaklynonlinearapproximation and for shallowwatertidal systems, the weaklynonlinear
analyseshavebeenextendedto three-dimensional and stratifiedflows, and theirresultshave
provided
additional
insights
to residualcurrents
andlong-term
transport
processes
(Fenget al.,
1990; Hamrick, 1986, 1990).
In realistictidalembayments,
thebasingeometry
andwaterdepthdistribution
(bathymetry)
are usuallyquitecomplex. The distributions
of tidal currentspeedanddirectionare rectified
by thebasingeometryandbasinbathymetry resultingin a complexstructureof transient
tidal
flow. Thenonlinearities in theflowareembedd•in thecomplex structureof velocityshears
whichgiveriseto macro-scale mixing(tidalpumpingandtidaltrapping).The variations in
velocityshearcanbe largeandvariable,theweaklynonlinearapproximation maynotbevalid
for partor all of theembayment.As theflowschangefromweaklynonlinearto nonlinearor
stronglynonlinear,the practicaluseof the resultsderivedfrom weaklynonlinearanalysis
becomes increasingly unsatisfactory.An extensionof theproperties of Lagrangianresidual
currentsandlong-termtransport to generalnonlineartidal flowsis highlydesirable.

R T ChengandV Casulli 411
In this paper, a governingequationis presentedfor the inter-tidaltransportprocesses

withoutinvokingthe weaklynonlinearassumption.The governingequationtumsout to be
againa convection-dispersion equation,excepttheconvective velocityis themeanLagrangian
residualcurrent.The tidal-cycleaverageddispersion is shownto haveresultedfromnonlinear
interactionsof tidesandtidal currentswith basingeometryandfrom complexstructureof tidal
currentshear.Neitherthe analyticalsolutionsof themeanLagrangianresidualcurrentnor the
tidally-averaged dispersion
coefficientareknownto eftst,theywill bedetermined numerically.
In the examples,the computedinter-tidaldispersion coefficientis shownto be small and
negligiblein areaswhereweaklynonlinearconditions prevail. In areasof complexbathymetry
andcomplexshearstructures, the computedinter-tidaldispersion coefficientsare consistent
with the valuesreportedin the literature(Eg. Okubu,1971;Fischeret al., 1979;Ridderinkoff
and Zimmermann, 1990).
20.2 Tidally-averagedtransportequation
20.2.1 Weakly nonlinearanalysis
There are two possiblepathwaysthat can lead to solutionsof inter-tidaltransportof a

conservative soluteover a time spanof severaltidal periodsor longer. A straightforward
approachis the directintegrationof the governingintra-tidaltransportequationover a long
periodof time. The tidally-averaged soluteconcentration is derivedfrom time-averagingthe
timedependent solution,or by applyinga low-passfilter to removetidal fluctuations.When
the numericalsolutionfor a long periodof time is sought,a slightnumericalerrorper time
stepof integrationmayrenderthefinal solutionmeaningless.To providean accuratesolution
for long-termtransport,the time integrationmustbe carriedout with sufficientlyfine spatial
and temporalresolutions,and the numericalintegrationmustbe both robustand efficient.
An alternativeapproachis to posethe questionfrom a differentview pointby askingif it

is possibleand advantageous to deriveanothergoverningequationwhichgivesdirectlythe
fidally-averagedsoluteconcentration.This approachhas been attemptedby applyinga
time-averaging operator(a low-passfiler) to the intra-fidalgoverningequation. A form of
the intra-tidaltransportequationfor a conservative soluteis
OC
Ot
+ •,.VC = DVzC (20.1)
whereC(t) denotestheconcentration of solute,t is time,u e is theEuleriantidalvelocity,and

D is the turbulentdispersioncoefficient. Denote both the tidal currentand the solute
concentrationas the sumof an Eulerianmeananda timedependent complementary part. A
tidally-averaged
Eq.(20.1) becomes

O<C>
+ •,,'V<C> + <g,"VC'> = 0 (20.2)
where< > is a tidal cycle averagingoperator,u,, is the Eulerianresidualcurrent,and the

'primed'variablesare therespective time-dependentcomponents (Eg.,Fischeret. al, 1979).
In Eq.(20.2), the turbulentdispersionhas been neglectedwhen comparedto convection.
Strictly,Eq.(20.2)is a consequence of mathematical
formality.Eq.(20.2)is nota well posed
problem without resolving the closure problem associatedwith the correlationterm
u c 'VC'>. A commonapproachis to introduceadditionalassumptions or ad hoc
approximations relatingthe correlation
term<uc 'VC > to themeanvariable<C>.
For a class of weakly nonlineartidal dynamicproblems,using a small perturbation

technique,Fenget al. (1986a,b),Chenget al. (1986) reportedsomeinteresting properties
of
Lagrangianresidualcurrentand inter-tidaltransportprocesses. Subsequently, numerous
extensionsof the weakly nonlinearanalysesof tidal dynamicshave been reported[For
example,seeFengand his co-workers(1990) andalsoseeHamrick(1987, 1990)]. Severe
limitationsexist in usingthe resultsof weakly nonlinearanalysisin realisticapplications.
Beforeextendingthe analysisto generalnonlinearflows, someof the key resultsfrom the
weaklynonlinearanalysesare reviewedand summarized.
Following
thedynamics
ofalabelled
wate_>parcel
inLagrangian
mechanics,
anaccepted
definitionof the Lagrangianresidualcurrent,u t. is given as
Z(T+to)
- Z(to) (20.3)
where
•(T+to)and•(to)aretheending
andbeginning
positions
ofa labelled
water
parcel
releasedat the time to (referringto a tidal phase),and T is the tidal period,(Zimmerman,
1979;ChengandCasulli,1982). ChengandCasulli(1982) pointedout thattheLagrangian
residualcurrentis a functionof the tidal phaseto. Undertheweaklynonlinearapproximation,
Feng et al. (1986a) have shownthat the Lagrangianresidualcurrentcan be givenin a small
parameterexpansionas
u-;,(t
o) - a,, + + ß au(to) + o(,:) (20.4)
in which u •ais the Stokes drift, e is a smallparameteron the orderof the ratio of residual
currentover tidal current. The time dependency of the Lagrangianresidualcurrentappears
in theLagrangian
drift,•d(to),[seeCheng
etal (1986)].TheLagrangian
driftinscribes
an
ellipseon the hodograph planein a tidal cycle. Within the weaklynonlinearanalysis,Feng
et al. (1986b)continuedto showthatthe first ordertidally-averaged transportequationfor a
conservative soluteis a pure convectionequation,

c•<C>
(20.5)
where(u •r + u • is the first orderLagrangianresidualcurrent,Eq.(20.4). Note that (u ½r'•'

U,d is the sameas the masstransportvelocity given in Longuet-Higgins(1969). Similar
conclusions havebeenreportedby MiddletonandLoder(1989). It is interestingto pointout
furtherthat a closurerelationfor Eq.(20.2) hasbeenobtainedwithoutinvokingadditionalad
hoc assumptions.The closureconditionis, of course,
<t•,'.VC'> - g',,t'•'<C> , (20.6)
in which the correlationterm hasbeen shownto be a convectivetransportwith the Stokes'

drift astheconvectivevelocity. Theseresultsareinstructivefrom an analyticalpointof view.
Solutionsof Eq.(20.5) are not necessarilymoreaccuratethanan ad hoc solutionof Eq.(20.2)
in practical applicationsbecausethe weakly nonlinear condition is a fairly stringent
requirementfor mostrealisticembayments.Eq.(20.5) correctlystatesthatthetidally-averaged
transportis principally convectiondue to a mass transportvelocity (or the first order
Lagrangianresidualcurrent),all othermechanisms are of secondary importance.When the
weakly nonlinearconditionbreaksdown,the claimsof Eq.(20.5) becomeinvalid.
Confusionsmay arisein tidal dynamicsof horizontalflows regardingthe definitionof the

Stokes' drift becausethe Stoke's drift is a concept"borrowed"from surfacewaves. To
eliminatethisconfusionin thisand futuretidal dynamicanalyses,the Stokes'drift is defined
to be the tidally-averaged differencebetweenthe Lagrangiantidal velocityand the Eulerian
tidalvelocitytruncatedat thefirstorderof Taylor'sseriesexpansion,(Longuet-Higgins,1969).
The Lagrangianand Euleriantidal velocitydifferenceat (xo, Y0)can be approximated by a
Taylor's seriesof u •(t) and truncatedat the first orderas
Ag-
g•(t,
to)
- g,(t)
-[f•,dt'].V•,+
Higher
Order
Terms, (20.7)
where
•(t, to)andu•(t)aretheLagrangian
andEulerian
tidalvelocities,
respectively.
Specifically,the Stokes'drift is definedas
(20.8)
with all variablesdefinedat (Xo,Y0). Followingthis definition,the sum of the Eulerian

residualcurrentand the Stokes'drift equalsthe mass-transportvelocity,or the first order
Lagrangianresidualcurrent.
In fact Eq.(20.7)canbe usedto setthe limitsfor the weaklynonlinearapproximationto

be considered valid. The expression
in the [ ] in Eq.(20.7)is the displacement
of a water

414 R T ChengandV CasuHi
parcelat time t releasedfrom to. If at any time in a tidal cyclethe valueof the [ ] is of the
sameorderof magnitudeasthecharacteristic lengthof thevelocitygradient,thenthetruncated
Taylor seriesin Eq.(20.7) is insufficientto approximatethe differencebetweenthe Eulerian
and Lagrangian tidal velocities (Zimmerman, 1979). When the weakly nonlinear
approximation becomesinvalid,it furtherimpliesthatthephysicalsignificances of theStokes'
drift (as definedabove)and the masstransportvelocityare alsoquestionable, (Zimmerman,
1979).
20.2.2 GeneralizedLagrangiandynamics
Withoutinvokingthe weakly nonlinearapproximation, a tidally-averagedtransportequation

canbe derivedby followingthe Lagrangianparticledynamics.First,considerthe physical
processof solutetransportin a tidalembayment for whichthe localturbulentmixingcanbe
assumed to be smallandneglected.Underthe statedconditions, thesoluteis transported
by
tidalcurrentconvection alone. The macroscopic mixingis generated by tidalvelocityshears
whichareproduced fromspatialvariations
in tidalcurrentamplitudeandphaseresultingfrom
topographical rectification
of tidalcurrentsin tidalbasinsof complexgeometryandvariable
bathymetry.Thesephenomena of macroscopic mixingarecommonly knownastidalpumping
and tidal trapping.
The mathematical definitionof thetidally-averagedsoluteconcentrationat a fixedposition,

<C(xo,yo)>,is equivalentto measuringsolute(salinity,for example)at a fixed stationand
averagingthe resultsover a tidal cycle. Thus, the tidally-averaged concentration of an
Eulerianvariableis theaveraged valuesof soluteconcentrationsampledfromwaterparcels
thattravelledthroughthe sensorat (Xo,Yo),in sequence, over a periodof a tidal cycle.
Specifically,the soluteconcentrations recordedby thesensorare NOT determined from the
samewater mass. When the local turbulentmixingis neglected,the time-averaged solute
concentration at (Xo,Yo)wouldbe the sameas an ensemble averageof soluteconcentrations
takenfroma subset of waterparcelsthatmovethrough(Xo,Yo)in a tidalcycle. A common
propertythatdefinesthewaterparcelsubset is thateachwaterparcelin thesubset mustmove
through(x0,Yo)at leastoncein a tidal period. Oncethis subsetof waterparcelscan be
determined,an ensemble-averaged valueof soluteconcentration canbe computed from the
soluteconcentrations sampledfrom thesewaterparcels.
In orderto computean ensemble-averaged concentration,

the questionis shiftedto the
identificationandthe determination of thissubsetof waterparcels.Whena labelledwater
parcelis releasedin a tidal systemfrom (Xo,Yo),theLagrangian residualcurrentis definedat
theendof a tidalperiod.In repeating fides,by continuously
releasinglabelledwaterparcels
overa tidalcycle,theendinglocations of thesewaterparcelsforma closedlocusenclosing
a small area in space. An inversequestioncan be posed:What is the distributionof the
startingpositionsof a subsetof waterparcelssatisfying the requirementthateachlabelled
waterparcelmustmovethrough(Xo,Yo)at leastoncein a tidalperiod?Assuming thatthe
startingpositionsof thissubset of waterparcelscanbedefined,theyenclose a smallregion
whichwill be referredto as a dispersion patch. By definitionthen,an equivalence of an

R T Cheng and V Casulli 415
ensemble-averaged soluteconcentration
andthetime-averaged concentration at (Xo,Yo)canbe
established.To helpvisualizethiscomplexsituation,consideragaina weaklynonlineartidal
system. The dispersionpatch in this case can be shownto be a mirror image of the
Lagrangianresidual ellipse (Cheng et al., 1986) becausethe governingequationsare
Hamiltonian.In the generalnonlinearcase,a dispersion
patchexistsbut it is not necessarily
a mirror imageof the regionoccupiedby the Lagrangianresidualcurrents.
Dispersion
Patch
A:[x(t-T), y(t-T)]
O: [Xo, Yo]
CO: liean Lagrangian
residual current
Figure1: A schematic diagramshowingthedispersion patch,meanLagrangianresidualcurrent.Point

A is a generalpoint on the edgeof the dispersion
patch.
Depictedin Figure 1 is a schematicillustrationof a dispersion patch. When a labelled

waterparcelis releasedat time (t-T) from a generalpointA, [x(t-T),y(t-T)],on the edgeof
thedispersion patch,thiswaterparcelwill arriveat point(Xo,Yo)at timet whereT is the tidal
period. For repeatingtides,but notnecessarilyweaklynonlinearsystems, thewaterparcels
defining the dispersionpatch form a closedlocus in space. Thus, the governing
tidally-averaged transportequationis a mathematicalexpressionstatingthe equivalence
betweenthe tidally-averaged soluteconcentration at (Xo,Yo)and an ensemble-averaged
concentrationontheedgeof thedispersion patch.Mathematically, thetime-averaged C at (Xo,
Yo)over a tidal periodis

)>= •1 •feo*l•12
(20.9)
<C(xø
'yø
'tø t•-ra C(xø
'yø
' t)dt
On theotherhand,whentheturbulentdispersion is neglected,
Eq.(20.1)impliesthatthesolute
concentrationis constant alongeachstreakline.In particular,C(xo,Yo,t
) = C[x(t-T),y(t-T),t-T],
whereIx(t-T), y(t-T)] is a generalpointon the edgeof the dispersionpatch. Thus,Eq.(20.9)
can be written as
<C(xo,y•to)> - •1 •,o-r/:
•'to.
rt2C[x(t-T),y(t-T),t-T]dt (20.10)
wherethefight hand-side of Eq.(20.10)is an ensembleaverage.Becauseof theperiodicfide,

the integralon the fight hand sideof Eq.(20.10) canbe writtenin termsof tidally-averaged
concentrations.Thus,the equivalenceof a time-averaged and an ensemble-averaged solute
concentration can be established as
' •1 •'t,.rl2
<C(xø'Yø'tø)> <C[x(t),y(t),to_T]>dt(20.11)
Sincethedispersion patchis definedby [x(t), y(t)], thetimeintegration

on thefight handside
of Eq.(20.11)is an averageof thesoluteconcentrations takenfromthewatermasson theedge
of thedispersion
patch. Denotethecentroid
of thedispersion
patchby (•,•), Figure1, so
that[ and• aredefinedby
1 fyrt2x(t)dt --=-•
, and Y 1 at•-r12
y(t)dt. (20.12)
By subtracting
<C(•,•,to-T)>fromeithersideof Eq.(20.11)anddividingEq.(20.11)by T,
Eq.(20.11) becomes
1 ft•+TI2
<c(xo,•o,to)>
- <c(•,•,t o- • ato-r/2<C[x(t),y(t),to-T)l>dt- <C[E,j',to-TI>
T T
(20.13)
Thetruncated
Taylorseriesexpansion
of <C(xo,Y0,to)>
and<C[x(t),y(0,to-T]>
about(•,•,to-T)
can be written as
c9<C> c9<C>
- <C(•,Y,to-r)> + Tog<C>
<C(xo,Yo,to)> + HOT
+(xo-•)• +(yo-•)oy
(20.14)

417
R T ChengandV Casulli
<C[x(t),y(t),to-T]>
- <C(j,•,to-T)>
+[x(t)-x-]
Ox
+ [y(t)-•]
+_21 [x(t)-œ]
• cgZ<C>
i}x2
+[x(t)-œ]
[y(t)-y-']
O•-<C> + _1[y(t)_y--]
ilxc3y 2
2O •Oy
<C>+ HOT
2
(20.•5)
whereall thederivativesareevaluated
at (•,•,to-T)andHOT standsfor higherorderterms.
Substitution
of theseTaylorseries
expansionsintoEq.(20.13),
afterneglecting
thehigherorder
terms,yieldsthefollowingfidally-averaged
transportequation,
c•<C> 02< C> 02< C> 02< C>
+ <•t,> 'V<C> ' +2 + .
i•t v• Ox
2 v• Oxi•y v•y •y2
(20.16)
In Eq.(20.16),
is themeanLagrangian
residual
currentfollowingthedefinition
givenin Eq.(20.3). The
dispersion
coefficients
are givenby
1 fto+?12 dt (20.17a)
2T2•o-rt2
[x(t)-•]2
1 (20.17b)
2T2•,o-r/2
[x(t)-J]
[y(t)-•]
dt
(20.17c)
¾yym
2T• •eo-rlz
[y(t)-•]=
dt
Thetidally-averagedtransport equation,
Eq.(20.16),is validupto thelimitof theTaylor
seriesexpansions
usedin Eqs.(20.14) and(20.15). Theexactcondition for theconvergence
of theTaylorseriesis difficultto establish,
however,thecharacteristiclengthscalefor the
dispersion
patchisexpected tobeanorderof magnitude smallerthanthelengthscales of tidal
velocitygradients
usedin thedefinition of theStokes' drift,Eq.(20.7).Thus,thevalidityof
thisequationhasbeenextended to generalnonlinear
flowswithoutthelimitation of weakly
nonlinearapproximation. In strongly nonlinear
flows,thedispersion patches canbe highly

contorted.The Taylor seriesexpansiongivenin Eqs.(20.14)and (20.15) may not converge,

and the validity of Eq.(20.16) becomesquestionable.
Nevertheless,the present results, Eq.(20.16), reveal some important properties of

tidally-averaged transportand mixingprocesses.The governingtransportequationremains
in a form of convection-dispersionequation. In the transportequation,the meanLagrangian
residualcurrentis the convectivevelocity. The tidally-averaged dispersionis relatedto the
spatialspreadof theLagrangianresidualcurrentsin a tidal cycle. Theseresultsareplausible
and not at all surprisingfrom a physicalpoint of view. In regionswhere tidal current
distributionsare complexandhighlyvariable,the residualdispersion patchesare expectedto
occupy bigger and irregular areas than in regionsof smoothor weak velocity shears.
Intensifiedmixingandhighvaluesof tidally-averaged dispersion
canbe anticipated in regions
whereintensetidal velocityshearsare expected.In the limiting caseof a weaklynonlinear
approximation,the dispersionpatchbecomesnegligiblysmall, and the mean Lagrangian
residualcurrentis reducedto the first orderLagrangianresidualcurrentor the masstransport
velocity.Theareaof thedispersion
patchis proportional
to e2whichis smallandnegligible
whene - o(1). Obviously,in thislimitingcase,Eq.(20.16)is reducedto Eq.(20.5),a result
thatwas reportedearlierby Fenget al. (1986b).
20.3 A model for the tidally-averageddispersioncoefficient

Strictly,Eq.(20.13)or (20.16) areequivalentformsof an inter-tidaltransportequation,either
onecanbe usedfor solutionof <C>. Eq.(20.13)is a discreteform of the transportequation;
the ensembleaverageof soluteconcentrations representingdispersion canbe evaluatedin the
courseof numericalintegration.In highlynonlinearcases,the localmixing approximation
may not apply,the use of Eq.(20.13) may be preferred. In contrast,when the governing
inter-tidaltransportequationis expressed
in Eq.(20.16),propertiesof Eq.(20.16)canbe related
to the conventionalformulationof convectionand dispersion.
Althoughthe shapesof the dispersion patchare generallyirregular,at this orderof the

approximation thedispersiontensoris a symmetricsecondranktensor,Eqs.(20.17).To better
understand thephysicalprocessesdescribed by theinter-tidaldispersion,
a simplifiedisotropic
dispersion modelis suggested to ascertainthepresentapproachgivesat leasta correctorder
of magnitudeof the inter-tidaldispersioncoefficient. Generaltensorialpropertiesof the
inter-tidaldispersion
coefficients
will be furtherpursuedif thesimpledispersionmodelproves
to beusefulin practicalapplications.A simplifiedmodelis proposed by re-writingEq.(20.16)
as
a<c>
+ <tZt,>.V<C>-v,Va<C> , (20.•8)
at
where <uu> is the mean Lagrangianresidualcurrent,x), is a bulk inter-tidaldispersion

coefficient. Here, •)• is definedto be the meandistancebetweenthe edgeof centroidof the

dispersionpatchsquareddividedby a tidalperiod. In thefollowingexamplesandin practical
applications,both the mean Lagmngianresidual current and the inter-tidal dispersion
coefficientwill be evaluatedby meansof a numericalhydrodynamic model.
20.4 Computationsof the tidally-averageddispersioncoefficient

Assumingthat a robustnumericalhydrodynamic model is available,the transienttides and
tidal currentscanbe adequately describedby themodel[for examplethemodelsdescribed in
Casulli(1990)or in ChengandCasulli(1991)]. Specifically,thepresentinterestis in using
thenumericalhydrodynamic modelto determinethemeanLagrangian residualcurrentandthe
dispersionpatch. In order to definethe dispersionpatch,the governingequationsmustbe
integratedin a reversedtime sequence.Becausethe shallowwaterequationsare nonlinear,
the tidal dynamicsis not a reversibleprocess. An integrationof the governingnonlinear
equationswith a negativeincrementof time does not producethe time-historyof tidal
dynamicsin a reversedtime sequence. Nonetheless,in a discretenumericaldomain,
computations to generatea "reversedtime history"of the tidal dynamicsare possible.
The numericalsolutionsof the shallowwater equationsare solvedby integratingthe

governingequationsforwardin time until the solutionhasreacheda dynamicsteadystate.
The computed tidal velocitydistributions are savedat everyintegration timestepby writing
the numericalsolutionto a time-historyfile for a completetidal period. If the numerical
modelmustbe integrated witha relativelysmallAt dueto reasons of numericalstability,then
thenumericalsolutions canbe savedat appropriate multipletimeintervals. The sequence of
records in this file containsthe complete time-historyof the tidal circulationin the
embayment.A 'reversedtime-history file' is createdby switching theordersof thesequential
recordsfrom the 'forwardtime-history file' to a reversedorderwithoutalteringthe contents
of therecords.The firstrecordin thereversedtime-historyfile is thetidalcurrentdistribution
savedat the last time step. The informationin the 'reversedtime-historyfile' is usedto
performthe reversedtime integration of the Lagrangianmovements of waterparcels. It is
importantto note that in the reversedtime-integrationthe nonlinearitiesof the tidal
characteristics are preserved.Whena labelledwaterparcelis releasedfrom a fixed location
at to,by integrating for onetidalcyclealongthestreakline backward in time,theoriginof the
labelledwater parcel can be determined. During the reversedtime integrationof the
Lagrangian waterparcelmovements, thelocaltidalvelocities at previoustimesarefoundby
interpolation from the numericalsolutionssavedin the reversedtime-historyfile. If the
startingpositionfor an arbitrarywaterparcelcanbe determined, thenthe determination of a
dispersion patchis simplya systematic releaseof labelledwaterparcelsin a reversedtime
sequence overan entiretidalperiod.Because thetideis assumed to be repeatingandcyclic,
the startinglocationsof thissubsetof labelledwaterparcelsform a closedlocusthatdefines
thedispersion patch. Finally,thevaluesof theinter-tidaldispersion coefficientareevaluated
by computing themeandistance between theedgeof centroid of thedispersion patchsquared

420 R T Cheng and V Casulli
and dividedby the tidal period for every point.
20.5 Examples
Resultsfrom two examplesare givento illustratetheproperties andthespatialdistribution
of
tidally-averaged dispersioncoefficient.In bothcases,the numericalsolutionsof the shallow
waterequationsare solvedby a semi-implicit,Eulerian-Lagrangian, finite-difference
method
(Casulli, 1990). The detailsof the Tidal,Residual,andInter-tidalMud-flat model(TRIM) are
reportedin Chengand Casulli (1991), and they are not repeatedhere.
20.5.1. Tidal inlet with a constriction
The firstexampleis designedto gainadditionalinsights.Thisproblemrepresents anidealized

tidal inletwhosepropertiesincludingthetransienttide-induced
eddies,tidalandresidualflows
havebeeninvestigated by Imasato(1983). The testbasinusedin thisinvestigation is similar
to thetestproblemof Imasato(1983) consisting of a rectangular
basin(55 km x 46 km) with
a constriction that divides the inner and outer basins. The inner and outer basins are
connected throughan opening8 km in width (Figure2a). The waterdepthis assumed to be
10 m throughout thebasin,anda uniformM2 tideof 1.0 m amplitudeis specifiedacrossthe
openboundary. The dimensions of this idealizedbasinare chosento representa typical
coastaltidal embayment.
The TRIM modelis usedin the computations of tidal and residualcurrentswith 0.5 km x
0.5 km finite-differencesquaremeshes. The numericalmethodusedin the TRIM model,
underlinearizedanalysis,hasbeenshownto be unconditionally stable(Casulli,1990). The
numerical integrationis carriedoutwith At = 15min.withoutanysignof instability(thegrid
Courantnumberis on the orderof 20). Thirteendroguesare followedfor five days,the
startingpositions
of thedrogues areshownin Figure2a. The Lagrangian
trajectories
of these
drogues,Figure 2b, show a complicated patternin the headlandregions. The drogue
trajectories
exhibitregular
patternsawayfromthecenterof thebasin.Thecomplex trajectory
patternis alsoanindicationthatthespreading
of theLagrangianresidual
currentcanbequite
large. This patternshowsa high degreeof nonlinearinteractions
betweenthe tide and the
currentin regions of highvelocityshear.In thisexample,thespatialvariations
of thevelocity
field (tidalcurrentshears)are inducedby theheadland constriction.A typicaltidalcurrent
distributionis depictedin Figure2c showinga pair of tidal vorticeslocatednearthe inlet
constriction (headland) in theinnerbasinduringflood.A pairof vorticesaredeveloped near
theheadland in theouterbasinduringebb,Figure2d. A slightantisymmetry canbeobserved
in theflow field,whichis caused by theCoriolisacceleration.
In regionswherethevortex
quadripoles developed, thetidalvelocityshearsarehigherthantherestof theregion. The
sizesof thevelocityeddiesareof thesameorderof magnitude asthecharacteristiclengthof
thebasin.Therefore, theweaklynonlinear approximation
is notvalidfortheheadland region
coveredby the vortexquadripoles.Figure 2e and 2f are the computedEulerianand

Lagrangianresidualcurrentdistributions.
Lagrangian Trajectories
Model Tidal Inlet (b) in A Model Tidal Inlet
Open Boundary
Tidal Circulation in Tidal Circulation in

A Model Tidal Inlet (a) A Model Tidal Inlet
Eulerian Residual Current Lagrangian Residual Current

(e) in A Model Tidal Inlet ([) in A Model Tidal Inlet
Figure2: a) A modeltidalinlet,x's arethestarting

positions
of drogues.b) Thetrajectories
of thirteen
droguescovering a periodof fivedays.c) andd) arethecomputed tidalcurrent
distributions
during
floodandebb,respectively. Thetidalvortices onthetwosides of theconstriction
havedeveloped. e)
andf) Thecomputed Eulerian andLagrangian residual
currentdistributions,
respectively.

Followingtheprocedures
described
above,thedistribution
of thetidally-averaged
dispersion
coefficient,
% iscomputedanddepictedin Figure3. Thecomputed x)tranges
between
5x10
•
and5 x 10• cruZ'/sec.
Theregionin whichthevalues of x)tis greater
thanlx105cm•'/s
nearly
coincideswith the residualquadripoles.The meanLagrangianresidualcurrentpatternis
similarto theLagrangianresidualcurrentdistribution at a giventime. The sizesandshapes
of the dispersion patchesare measures of the varianceof the Lagrangianresidualcurrents.
In regionsnearthe constriction headland,the dispersionpatchesare largeandirregularand
the nonlinearinteractionsare significant.In thisregion,the weaklynonlinearapproximation
is violated,andthephysicalmeaningof theEulerianresidualcurrentis questionable, Imasato
(1983) . Likewise,the usageof the Stokes'drift and the masstransportvelocitybecome
questionablein regions of strong nonlinearity(Zimmerman,1979). Away from the
constriction,the nonlinearityis expectedto be weak. The computeddispersion patchesshow
neatlyclosedloci. In somecasesthe dispersion patchescan be approximated by ellipses
consistentwith the resultsof weakly nonlinearanalysis.
Tidally-Averaged
Dispersion
SxlO 6 cm2/sec
5xlO • cm2/sec
5xlO 4 cm2/sec
Figure 3: The distributionof tidally-averageddispersioncoefficientin a rectangulartidal inlet with a

constriction.

20.5.2 San FranciscoBay, California
To evaluatethepracticalvalueof the proposedmodel,thisapproachis testedin SanFrancisco

Bay, California. San FranciscoBay is a complexestuaryconsistingof interconnected
embayments, sloughs,marshesandchannels,Figure4. Amongthe manyfactorsthat affect
theflow properties in theBay, thewaterdepthdistributionis oneof themostimportantfactors
controllingthe spatialvariationsof both the magnitudeand the directionof tidal currents
(ChengandGartner,1985;Walterset at., 1985). The tidesenteringthe Bay througha narrow
openingat theGoldenGatehavea rangeof roughly2 m. composed primarilyof the M 2 and
K1 tides. Within CentralBay the tidesbifurcateandpropagateconcurrentlyinto SouthBay
and the northernreach which includeSan Pablo Bay, CarquinezStrait, and SuisunBay.
CentralBay is not only geometricallycomplex,thevariationsin thedepthdistributionare also
quitelarge. The deepestregionin CentralBay is locatednearGoldenGate, wherea depth
close to 100 m. is found. At the easternboundaryof Central Bay is a broad region of
inter-tidalmud-flat. The geometryof CentralBay is furthercomplicatedby the presenceof
severalislands.Withoutfurtherelaborations,CentralBay is simplya very complextidal basin
in whichlargevariabilityof tidal currentcan be anticipated.
122030 ' 122000 '
HONKER BAY
GRIZZLY
P E TA L UMA
RIVER
SAN PABLO BA
•UINEZ BAY
38 ø00' STRAIT SAN JOA Q UIN
RIVER
GOLDEN
) 10 20KILOMETERS
UTH SAN FRANCISCO BAY
37ø30
CREEK
Figure4: SanFrancisco
Bay, a bathymetrically
complextidalestuary.

Both San Pablo and SouthBays can be characterized by a deep channelwhich is

surrounded bybroadshoals.Tidalcurrent distributionsin thesebasins arelesscomplex.High
intensitytidalcurrents aredistributed withindeepchannels, lowermagnitude tidalvelocities
areobserved overshoalregionsdueto increased friction.The tidalvelocityshears in regions
of shoalsandin themiddleof deepchannels arerelativelyweak. However,in thetransition
zonesbetweenthe channeland the shoals,the tidal velocity shearsbecomelarge, and the
nonlinearitiesare relativelyintense.The computedStokes'drift, usingEq.(20.8),showsan
irregularpatternand a highdegreeof uncertainty in the transitionzoneLrres•tive of the
model grid resolution. This observationfurther ascertainsthat the weakly nonlinear
approximation is invalidin thetransitionzones.Carquinez Straitis a deepchannelconsisting
of relativelysimplegeometry thatconnects SanPabloBay to its westandSuisunBay to its
east. Althoughtheintensities of thetidalcurrents arehigh,thespatialvariabilityof thetidal
currentsis relativelylow. Therefore, theweaklynonlinear approximation mightbemarginally
validin CarquinezStrait. SuisunBay is a shallowandcomplexbasin,it consists of braided
channels andshoals.The complexbathymetrystructure suggeststhatthecharacteristiclengths
of the bathymetfic featuresare relativelyshort. The flowsin SuisunBay is expectedto be
highlynonlinear.Additionaldiscussion of the fidesandtidalcurrents in SanFrancisco Bay
is referred to Walters et al. (1985).
The inter-tidaldispersion coefficients

in SanFranciscoBay are computedusingthe same
numericalhydrodynamic model,ChengandCasulli(1991). The Bay is discretized by a 500
rn x 500 rn rectangularfinite-differencemeshgivingapproximately 18,000totalgridpoints.
Of all the modelgrid points,about30% are active. The computingtime requiredby the
TRIM modelis roughlyproportionalto the activecomputational points. For the purIx)seof
computing theti•lly-avemged dispersion coefficients,onlyan M,•tideis imposedat theopen
boundaries, exceptthe tidal amplitudeis increasedto representrealistictidal rangevalues.
The numericalsolutionsare achievedusingan integrationtime stepAt = 15 min. In this
application,the horizontaltidal velocityshearsare, in a largepart,the directconsequence of
the spatialvariationsof thebasinbathymetry.The computedtidalcurrentdistribution in the
Bay is quitecomplex. ChengandGarmer(1985) haveshownthat the amplitudes of tidal
currentin San FranciscoBay are linearlyproportional to the local waterdepth. It is not
surprisingto identifya fairly strongcorrelationbetweenthe spatialvariabilitiesof the water
depthandthe tidal currents.Withoutdiscussing the complexstructures of the Eulerianand
Lagrangianresidualcurrentsand the Stokes'drift, the computedvaluesof the inter-tidal
dispersioncoefficientis depictedin Figure5. The valuesof thedispersion coefficients
are in
therangeof 3.0x10
nto 3.0x 10• cm2/sec
whichareconsistent
withthepatterns
andthevalues
reportedin the literature(e.g. Ridderinhkofand Zimmerman,1990).
The inter-tidaldispersion
is resultedfrom the time dependency of theLagrangianresidual
currents,whosedistributionis principallya functionof thetidalcurrentshears.While thetidal
velocity shearsare stronglydependentupon the spatial variationsof the basin depth
distribution,the valuesof inter-tidaldispersion
coefficientsare alsocloselycorrelatedto the
variabilityof thebasinbathymelry.High valuesof inter-tidaldispersion coefficientarefound
in regionsof hightidalvelocityshears.Or highvaluesof theinter-tidaldispersion coefficients

R T Chengand V Casulli 425
canbeanticipated in regionswithcomplexgeometryandlargewaterdepthvariations.Central
Bay is just sucha region,thedispersion
coefficientin CentralBay is on the orderof -106
cm"/sec.
Thevelocityshears
in themiddleof deepchannels
in bothSouthandSanPablo
Bays,andin CarquinezStraitarenotnecessarily
high,thecomputed
valuesof thedispersion
coefficient
aresmallto moderate
(- 5x10
• cm'/sec).Similarly,
overtheregions
of broad
shoalswherethe weaklynonlinearapproximation
may be valid, the computeddispersion
coefficients
arelessthan10• cm"/sec.However,
at theedgeof thedeepchannels,
the
dispersioncoefficients
becomelocallylargebecausethe tidalvelocityshearis highin the
transition
zones,Figure5. The computed dispersion
coefficients
in thewesternSuisunBay
aregreater
than106cm2/sec.
Thecomplex
basingeometry
induces
highvelocity
shear,
and
thushightidal dispersion
for thatpartof SuisunBay
Tidally-Averaged
Dispersion
5X10 6 cm2/sec
5x10 s cm2/sec
5x10 4 cm•/sec
Figure5: Thecomputed
tidally-averaged
dispersion
coefficient
distribution
in SanFrancisco
Bay.

20.6 Discussion and Conclusion
Of the two pathwaysto solutionsof long-termtransport problems,thedirectintegrationof the

intra-tidalconservation
equationis applicableonly when an accurate,robust,and efficient
numericalmethodof solutionis available. In practicalapplications,
thisapproachgivesonly
marginally satisfactoryanswers. Alternatively,the solution of tidally-averagedsolute
concentrationcanbe attemptedusingan inter-tidaltransportequation,if sucha solutioncan
be shownto givepracticalanswers.Earlieranalysisundera weaklynonlinearapproximation
has shedsomeinterestinglight in this approach,the resultinginter-tidalWansport equation
givesa lucid explanationof the inter-tidaltransportprocesses.
In realisticapplications,the weaklynonlinearanalysisis oftena poorapproximation in a

tidal estuary. Becausethe Lagrangianresidualcurrentis a functionof time (tidal phase),a
residual dispersionpatch can be introduced. Without invoking a weakly nonlinear
approximation, thedispersionpatchis definedby a subsetof waterparcelsthatmovethrough
a fixed point (x0, Y0) at least once every tidal cycle. An ensemble-averaged solute
concentration is definedas the averagedsoluteconcentration sampledfrom watermasses on
the edgeof the dispersion patch. By examiningthe intra-tidalLagrangianmovements of the
water parcels,an inter-tidaltransportequationhasbeenestablished by the equivalence of
time-averaged and ensembleaveragedsoluteconcentrations.The transportof a dilute
conservativesoluteover inter-tidaltime spansis again governedby the combinationof
convection anddispersion.The convection is dueto a meanLagrangian residualcurrent,and
thedispersion
is represented
by anensemble-average
alongtheedgeof thedispersion
patch.
To gainfurtherinsights, a simplifiedisotropicdispersion modelis proposed andtestedin

two numericalexamples.Throughnumericalsolutions of the shallowwaterequations, the
distributions of inter-tidaldispersioncoefficientsare computed. The distributionof the
inter-tidaldispersion canbe reasonably explained by relatingtheproperties of dispersion
to
thedistributions of tidal velocityshearandnonlinearity in the basin. In regionswherethe
nonlinearity ishigh,thevaluesof thecomputed inter-tidaldispersioncoefficientarealsohigh.
In areaswhereweaklynonlinear approximation is valid,numerical solutionsreproduceresults
of the analyticalsolutions.The presentformulation is furthertestedin a complexrealistic
tidalbasin-- SanFrancisco Bay. The distributionof the computed inter-tidaldispersion
coefficient
is quitecomplex.It is encouraging
thattheproperties
of theinter-tidaldispersion
arewellcorrelated withtheanticipated
distributions
of nonlinearity
in theBay. Thecomputed
valuesof thedispersion coefficient
areconsistent
withvaluesreported in theliterature.
The preliminaryresultsobtainedin this studyof long-term,inter-tidaltransport of

conservativesoluteare quite encouraging. Future investigationsshouldconsiderthe
anisotropic
dispersion
properties.
Theinter-tidaltransport
equationwill be solvedusingthe
computedmeanLagrangian residualcurrentto covera seasonaltime-span.The numerical
solutions
will be compared
with solutions
of theintra-fidal
transport
equation andwithfield
data.

20.7 References
Casulli, V., 1990, Semi-implicitfinite differencemethodsfor the two-dimensional

shallowwater
equations,J. Comp.Phys.,Vol.86,No.1,p.56-74.
Cheng,R. T., and Casulli, V., 1991, A Tidal, Residual,Inter-tidalMud-flat (TRIM) model using
semi-implicit,Eulerian-Lagrangian
Method,to appear.
Cheng,R. T. and V. Casulli,1982, On Lagrangianresidualcurrentswith applications

in SouthSan
Francisco
Bay,California,WaterResour.Res.,Vol. 18, p. 1652-1662.
Cheng,R. T. and J.W. Gartner,1985, Harmonicanalysisof tidesand tidal currentsin SouthSan

Francisco
Bay,California,Estuarine,
CoastalandShelfSci.,Vol. 21, p. 57-74.
Cheng,R. T., S. Feng,andP. Xi, 1986,On Lagrangian

residualellipse,in Physicsof ShallowEstuaries
andBays,Ed. van de Kreeke,LectureNoteson CoastalandEstuarineStudies,Vol. 16, p. 102-113.,
Springer-Verlag.
Dunkerton,T., 1980, A Lagrangianmeantheoryof wave,mean-flowinteractionwith applications

to
nonaccelerationand its breakdown,Reviewsof Geophysicsand SpacePhysics,Vol. 18, No.2, P.
387-400.
Feng,S., R. T. Cheng,andP. Xi, 1986a,On tide-inducedresidualcurrentandresidualtransport,1.

Lagrangianresidualcurrent,WaterResourResearch,Vol. 22., No. 12, p. 1623-1634.
Feng, S., R. T. Cheng,andP. Xi, 1986b,On tide-induced residualcurrentandresidualtransport,2.

Residualtransport with application
in SouthSanFrancisco
Bay,California,WaterResourResearch,Vol.
22., No. 12, p.1635-1646.
Feng,S., R.T. Cheng,SunWenxin,Xi PangenandSongLina, 1990,Lagrangianresidualcurrentand

long-termtransportprocesses
in a weaklynonlinearbaroclinicsystem,in Physicsof ShallowSeas,Eds.,
Wang Huatonget at., ChinaOceanPress,p. 1-20.
Fischer,H.B., E.J. List, R.C.Y. Koh, and N.H. Brooks,1979, Mixing in inlandand coastalwaters,
Academic Press.
Hamrick,J.M., 1986,Long-termdispersion
in unsteady
skewedfreesurfaceflow, Estuarine,Coastaland
Shelf Science,Vol. 23, p. 807-845.
Hamrick,J.M., 1990, The dynamicsof long-termmasstransportin estuaries,in ResidualCurrentsand

Long-TermTransport,Ed., RalphT. Cheng,Spring-Verlag,p. 17-33.
Imasato,N., 1983, What is tide-induced

residualcurrent?,J. of Phys.Ocean.,Vol. 13, p. 1307-1317.

Longuet-Higgins,M. S., 1969, On the transportof massby time-varyingoceancurrents,Deep Sea

Research,Vol.16, p. 431-447.
Middleton,J. F., and J. W. Loder, 1989, Skew fluxesin polarizeAwave fields,J. of PhysicalOceano.,
Vol. 19, No.1, p. 68-76.
Okubu,A., 1971, Oceanicdiffusiondiagram,Deep-SeaResearch,Vol. 18, p.789-802.
Ridderinkhof,
H. andZimmerman,J.T.F., 1990,Mixing processes
in a numericalmodelof the western
DutchWaddenSea,in ResidualCurrents
andLong-TermTransport, Ed.,RalphT. Cheng,Spring-Verlag,
p. 194-209.
Walters,R.A., R.T. Cheng,T.J. Conomos,1985,Time scalesof circulationandmixingprocesses

of San
FranciscoBay waters,Hydrobiologia,Vol. 129, p. 13-36.
Zimmerman, J.T.F.,1979,On theEuler-Lagrangian

transformationandtheStokesdriftin thepresence
of oscillatoryandresidualcurrents,Deep SeaResearch,26A, p.505-520.

21
Effect of bends on dilution rates

R Smith
Abstract
Centrifugal effectsastheflow goesaroundbendscontribute to secondaryflow in theestuary

crosssection. For particleswhichare not neutrallybuoyantthe verticalsamplingof the
secondary flow is non-uniform:buoyantmaterialtendsto be cardedto theoutsideof thebend
anddensematerialto the insideof the curve. Thispaperfocusesattemionon the sensitivity
of the horizontaldilutionratesto the secondaryflow and to the verticalrise (or sinking)
velocityof the particles.
21.1 Introduction
The work describedin this paper was motivatedby my being invited to watch some
spectacular largescaleexperiments conductedby Guymer,Brockie& Allen (1990) in the
SERC flood channelfacility at HydraulicsResearchWallingford. The main concernof the
experimentswas to investigatethe lateral spreadingof neutrallybuoyantparticlesin a
meanderingchannel. However, some flow visualisationexperimentsindicatedthat the
directionof travelandrate of dilutioncouldbe markedlydifferentfor particleswhichare not
neutrallybuoyant. Indeed,it is intendedthat over the next 3 yearsa numberof research
groupswill utilize the SERC flood channelfacility to investigatea varietyof questions
involvingsediments and buoyantparticles. This paperis a theoretician's
attemptto predict
quantitative differences
in the transportanddilutionof bothrisingandsinkingparticles.
21.2 Vertical concentration

profile
For a shallow water flow the vertical distribution•z) of rising (or sinking) particlesis
determinedby a balancebetweenverticalmixing<3andthe upwardsdrift velocityw:

430 R Smith
(21.1)
with
w¾- •az¾=0 on z = -h,O (21.2)
The boundarycondition(21.2) expresses the zeroflux of tracerinto thebedz = -h andout

of the free surfacez = 0. The solutionof theseequations (21.1,2)is givenby
v(z) = exp dz (21.3)
with
(21.4)
wherethereference
levelZois chosen
sothatthedepthaveraged
valueI • I = 1 is correctly
reproduced.
For simplicitywe shallrestrictattentionin this paperto the idealisedcaseof constant

verticaldiffusivity•3 andconstant
driftvelocityw. So,theequilibrium verticalconcentration
•(z) is normalisedexponential.
P extgP) wh z
¾= with P= , • =- (21.5)
1 - exp(-P)' •a h
The verticalP6cletnumberP is positivefor a risingsubstance

andnegativefor a sinking
substance.Figure 1 showsthe profilesfor P = -1,1.
Forairbubbles
of radius5 x 10'• m (.05mm)witha dirtysurface
therisevelocity
is 5 x
10'3ms'• (5 mms'•).Forspherical
particlesof sand
of thesamesizeandwitha density
twice
thatof water,thefall velocitywouldhavethesamevalue. If we specifythewaterdepthand
verticaldiffusivity
h = 5m, % = 0.025m:s-•, (21.6)
then the P6clet numbersfor the bubblesis P = 1, and that for the sandis P = -1. For bubbles
or particlesof othersizestheP6cletnumberscalesas the squareof theradius.

R Smith 431
-0.5 0.0 0.5 1.0 1.5
• ' ß
u/ u
-0.5 0.0 0.5 1.0 1.5
Figure 1: Velocity profilesof the longitudinaland cross-stream flow. The dottedcurvesshow the
verticalconcentrationprofiles•(z) for rising(P = 1) andfor sinking(P = -1) particles.
21.3 Longitudinaland transversevelocityprofilesfor flow around

a bend
Locallytheflow canbe viewedasbeingcircular.Insteadof usingther, 0 polarcoordinate

notation,
weshallinterpretthex coordinateasbeingdirectedalongthecurvingdepth-averaged
flow and they coordinateas beingradiallyoutwards.For shallowwaterflow, the dominant
termsin the horizontalmomentumequationsare
R+y
uv
R+y R+y
r 0•I• I * 0•(,0•u), (21.7)
R uO,,v
R+Y
+vOffR+yu2- -0•IIp
l +a,(va,
v) (21.8)
with

432 R Smith
u-v-O at z--h (21.9)
and
vO•u=vO•v=O at z-O (21.10)
HereR is theradiusof curvature

of theflowliney = O,Ip I is thevertically
averaged
dynamicpressure,and v is the verticaleddy viscosity. At the bed thereis a no-slipzero
velocityboundarycondition(21.9),whiletheuppersurfacez = 0 is asstuned to be stressfree.
(The omittedterms in equations(21.7,8) involve lateralviscosityand lateral gradientsof
velocities).
Again,for simplicitywe shallassumethatthereis onlygradualx, y-dependence of theflow

and we shallmodelthe viscosityas beingconstant.If the transverse
flow v is comparatively
weak, thenthe longitudinalvelocitycanbe approximated (seefigure 1)
3
(21.11)
u=l ull•(1-l•2)
with
_Rh• (21.12)
lul-
3(R +
For a givenpressuregradientthe longitudinalvelocityis fasterat the insideof the bend.

By hylx)thesis
thetransverse
velocityv haszeroverticalvalue.Thecentrifugal
termu:/(R+y)
in equation(21.8) is associatedwith the secondarycirculation(seefigure 1)
v- --f•(5- 33•2 + 35in- 7[ •) (21.13)

5
with
3h2 lu12 5411u

12 (21.14)
•' 56v(R+y)'
0•lpl-35(R+y)
HereI7isthemaximum
outwards
velocity,
atthefreesurface.
Equation
(21.4)quaffties
the slightlyhigherpressureat the outsideof thebendcausedby the curvatureof the channel.

R Smith 433
21.4 Weighted-averagevelocities
The verticalsamplingof the flow by the particlesleadsto a weightedaveragehorizontal
velocity
I •u I - I u lU(P), I •v II- ov(P) (21.15)
The explicit formulae (21.5, 21.11, 21.13) for 'y, u, v, enable us to evaluate the
dimensionless
proportionalityfactorsU,V:
u- 20- exp(-•) p2+2e•(-• + ß
5O4O
• {5 66840
v'=50-exp(-/9) pe+
+exp(-P)
[ 32144
1680 p /,2
+
(21.17)
At P = 0 thereare no singularities:
U- 1, V-0 (21.18)
(i.e. by definitionthe depthaveragedflow is in the x-direction)
Figure2 showsthe velocityratiosU and V asfunctionsof the verticalPec16tnumber. The

curveshavebeenplottedin a way whichgivesemphasis
to thequalitativesimilarityto the
velocity
profries
U/I u • V/Oshown
in figure1.
21.5 Sheardispersion
Fischer(1969)showedexperimentally andtheoretically thatfor neutrallybuoyanttracersthe
effectivecross-stream
mixingratein a curvedchannelcouldbeupto 10 timeswhatonewould
expectin a straight
channel.Thesecondary flowcauses theaugmented cross-streammixing
by thesheardispersion
mechanism firstexplained by G.I. Taylor(1953).Recently, theauthor
(Smith 1991)hasderivedexpressions for the effectsof rise (or sinking)velocitieson the
horizontalsheardispersiontensor

434 R Smith
-0.5 0.0 0.5 1.0 1.5
• 6-
• 4-
-• 2--
• -6
0.0 0.5 1.0 1.5
Figure 2: The dependenceupon the dimensionless

rise velocity P = wh/•:3 of the dimensionless
longitudinaland transversevelocitiesU,V.
(21.19)
Daa=•_
•-• ¾(u-
•rllul)dz
t dz,
•'-•' •1f-,
ø •1 [f_•¾(v-Vv3dz.
tr& (21.20)
•(v- vo)dz
•]dz
(21.21)
Here D12,D21 are mixed terms.
It is comparatively
easyto evaluatethe innerintegrals:
(21.22)
•(u- Vlul)dz•-HI Vl

R Smith 435
(21.23)
where
I (o -
2( 1 1•
2+2p p2 • p2
+
2(• -cxp(-•) • •
(21.24)
{exp(P•)
I5
=33•
2+35•4-
7•
6+ 66• = 140•s+42•s
P
420•2-210•4- 66 840 840- 2520•z

-•(•-•)+
p• ps
5040• 5040 32 144 1680

-c•(-P)
/,5
32 144 1680 5040
5(1-exp(-•) a •
(21.25)
In terms
ofthetwofunctions
I, • wedefine
thebulkflow,curvature
andmixed
coefficients

436 R Smith
as(p
) . 1 - exp(-P)
,p f-ioexp(-p•)12d• (21.26)
au(p
) . I - exp(-P)
P f_• oexp(-P•)/•
d• (21.28)
In principle
theseintegrals
canbeevaluated explicitly.
However,
forourpresempurposes
numericalvaluessuffice
(seefigure3). In theneutrally
buoyant
case(P = 0) theexplicit
values of the coefficients are
a•s - 2, a•t - 37888

, aM 832 (21.29)
105 1126125 34650
Withthedependence uponvertical
driftaccounted
forin theaB,aR,amcoefficients,
the
formulae
(21.19,20,21)fortheshear
dispersion
tensor
takethesimpleform
/•aah2 IIu12'as,D2
• h2•
aa
' o•2 /• h2
I uI 0 (21.30)
Fromfigure3 it isnoteworthy
thatsinking
particles
havegreater
longitudinal
butsmaller
transverse
shear
dispersion
thanrising
particles.
Thiscanbeexplained
intermsofthevelocity
profiles.
Thelongitudinal
velocityhaszeroshear
atthefreesurface
butthetransverse
velocity
also has zero shear near the bed.
21.6Gradual
separation
between
plumes
of neutral
andof rising
particles
In practice
theeffective
longitudinal
velocity
U I u I greatly
exceeds
theeffective
transverse
velocity
V• ofparticles.
Thisleads
togreatly
elongated
particle
distributions.
Consequently,
thestrongest
gradients
andstrongest
sheardispersion
is in thecross-stream
direction.The
corresponding
simplified
version
of thetwo-dimensional
sheardispersion
equation
is

R Smith 437
(21.31)
Here
cisthevertically
averaged
concentration
andI•c:]istheappropriate
weighted
averageof the transverse
turbulenteddydiffusivity.
0.00 0.01 0.02 0.03
I I I
10
4-
0-
-2-
-4-
-6
t
-8
-10
0.00 0.01
I
0.02
I
0.03
I
Figure3' The dependence uponthedimensionless risevelocityP = wh/•:.•of the numericalcoefficients

abe, aR..... , au ..... in the formulae (5.5a,b,c) for the longitudinal,transverseand mixed shear
dispersioncoefficients.
For a steadydischargethe local (constantcoefficient)solutionof equation(21.1) is a

Gaussian
! exp 202 Y- Y*-U I u I•

(2•) a o//u I u I
(21.32)
with

438 R Smith
x-x o
az' 211•. II + D22] I uI (21.33)
u'
To derivethissolution
we regardall theflowproperties
h, U, I u I V 0, I •nc2,
D22as
beingconstant
andwe neglecty/R.
If a neutrally
buoyant
(V = 0) andrising(orsinking)particles
arereleased
together
then
eventuallytheplumeswill become distinct.If we denotethe velocities
anddispersion
coefficients
for thetwospeciesby superscripts{o>,
{•>thenwecanestimate thenecessary
downstream
distancefor separation
otO)
' + oO)
•.
IA1)0
(x-Xo)
o•) l u I I (21.34
i.e. x-xo-2
(21.35)
Thisformulaequates
thesquare
of thecentroid
separation
to thesumof thevariances
of
theindividual
plumes.So,thespreadingandseparation
areof comparablesize.
Theformula
(21.30)for D22allowsusto assess
howthedistance
(21.35)for plume
separation
varieswiththestrength
• of thesecondary
flow. Thefunctional
formis
x- xo- 2[ I h II+• ¾•2

Icø]
ø'cø Ix. (21.36)
U•ø[37888
where X.(P) - 2
iAo•[
1126125 +aa(P)] (21.37)
Thus,
thedistance
forplumeseparation
decreasesrapidlyasthestrengthofthesecondary
flowincreases.
TheworkofFischer
(1969),asupdated byGuymet, Brockie & Allen(1990),
reveals
thatforneutrally
buoyantparticles
D22canbeupto 10times I •c2I in meandering
channels.Forneutral
andrising
particles
releasedtogether,
thedistance (21.36)forplume
separation
canbewithin10%of theasymptote (i.e.retaining
justtheXo.term).

R Smith 439
Figure4 showshow the dimensionless separationdistanceX.ovariesas a functionof the

verticalP6clet numberP. It is in accordwith commonobservationthat stronglybuoyant
flotsamcan rapidlybecomeseparatefrom a neutrallybuoyantplume. Figure4 revealsthat
the separationis just as pronouncedfor sinkingparticles.
oo 0.5 1.0 1.5 2.0 2.5
, I
lo
8-
6-
4-
2-
o-
--2 --
--4 --
--6 --
--8 --
Distance X. for separation
-10
0.0 0.5 1.0 1.5 2.0 2.5
Figure 4: The dependenceupon the dimensionless rise velocity P = wh % of the dimensionless

longitudinaldistanceX.. requiredfor a plumeof rising(or sinking)particlesto becomeseparatefrom
a plume of a neutrallybuoyantparticles.
Acknowledgements
I am mostgratefulto Ian Guymer,Nick Brockie& CathAllen for lettingme seewhatreally
goeson in turbulentflow roundbends.The typescript
waskindlypreparedby theDepartment
of Electronic& ElectricalEngineering.When thiswork was commenced the authorheld a
post-doctoralfellowshipfundedby the Royal Society.
21.7 References
Fischer,H.B. 1969:The effectsof bendson dispersion

in streams.WaterResources
Res.5, 496-506.
Guymet,I . Brockie,N.J.W. & Allen, C.M. 1990:Towardsrandomwalk modelsin a largescale

laboratory
facility. Proc5th Int Confon Mixing andDispersion,
MIT (to appear).

440 R Smith
Smith,R. 1991' Wind Augmentedtransportanddilution.J Fluid Mech.,228, 549-560.
Taylor,G.I. 1953:Dispersionof solublematterin solventflowingslowlythrougha tube.

ProcRoy SocLond, A219, 186-203.

22
Modelling the vertical distribution of

suspendedsedimentin combinedwave-
current flow
A G Davies
Abstract
A numericalmodelof the verticaldistributionof suspended sandis described. This is an

extension
of anexisting"one-equation"
K-modelof therough-turbulent bottom boundarylayer,
whichhasbeenusedto predictthe verticalcurrentstructure in steady,oscillatoryand
combined wave-currentflows. Thishydrodynamical modelis usedto supplytheverticaleddy
viscosity
distribution
(asa function of phaseanglein thecaseof oscillatory flows)for the
sedimentsuspension model.It is assumedthatthesediment diffusivityis equalto theeddy
viscosity
andalsothatthesediment concentration
is sufficiently
lowthatit doesnotaffectthe
flow. Model resultsare presented showingcycle-averaged distributionsof suspended
sedimentconcentrationfor combinedwave-currentflow, andtheseresultsare comparedwith
thosefor boththe wavesandcurrentalone.The influenceof waveson the suspended
sediment
flux is examined,andthesensitivity
of theresultsto theformof the (time-varying)
sediment
referenceconcentrationis assessed.It is concludedthat the "wave-related"contribution to the
flux is likelyto be significant
for grainswhichareconfinedto theoscillatory
boundary
layer.
22.1 Introduction
There have been several recent studies of the combined effects of waves and currents in the
bottomboundarylayerof shelfseas.Thesestudieshavebeenconcerned with the influence
of wave-current
interaction
on therateof energydissipation
at theseabed,on verticalmixing
processes
and,hence,on suchpractically importantmattersasstorm-surgeprediction,scour
aroundoffshorestructures,
and sedimentresuspension.
Oscillatory
boundary
layersarequalitatively
similarwhethertheyareinduced
by wavesor
tides. However,whereasturbulentwaveboundarylayersare thin [O(10-20cm)]andconfined
to thenear-bedlayer,tidalboundarylayersarethick[O(20-80m)]andmayoccupythe entire

442 A O Davies
waterdepthover muchof the continental shell In the combinedwave-current case, the

oscillatoryboundarylayerthickness is significantlygreaterthanthatfor wavesin isolation,
andthe amountof near-bedturbulenceis greatlyincreased.On an erodiblebed, thesefactors
are known to lead to increasedentrainmentand transportof sediment(e.g. Grass,1981).As
a resultof waveaction,substantialloadsmay be carriedby weakcurrentswhich,in isolation,
would be unableto move sediment.Sedimenttransportratesare likely to be greatestduring
stormeventswhen,in additionto bottom-generated turbulence,
turbulentenergyis introduced
into the watercolumnat the free-surfaceby wavebreaking.However,it hasbeenpointedout
by Soulsby(1987)thatthe largestcontributions to longterm transportare dueto fairly large,
frequentlyoccurringwaves,ratherthanto extremeevents(e.g. major stormsduringspring
tides).
Variousmodelshave been developedto predictthe verticalstructureof wave-current

bottomboundarylayerflow, and to assess the nonlinearenhancement of the bed shearstress
and the (oscillatory)boundarylayer thickness.The combinedflow was modelledoriginally
by Bijker(1967)and,subsequently,by Smith(1977),GrantandMadsen(1979),Bakkerandvan
Doom(1979),Fredsoe(1984) andothers.The presentmodelof rough-turbulent boundary-layer
flow, on which the sedimenttransportcalculations are based,was first usedto studythe
boundarylayersbeneathwavesandtidalcurrentsseparately (King, DaviesandSoulsby,1985;
Davies, 1986), and more recentlyto studythe combinedcase(Davies, Soulsbyand King,
1988;Davies,Barberand Soulsby,1990;Davies, 1991).The essentialdifferencebetween the
presentnumericalmodeland previousmodelsis that it utilisesa generalturbulenceclosure
schemewhichis applicableto waves,currentsandany combinedcase(e.g. smallwaveson
a strongcurrentor large waves on a weak current).The turbulenceschemeinvolvesthe
determination of theeddyviscosityas a functionof heightabovethebedandphaseanglein
the wavecycle,via the turbulentenergyequationandvariousturbulencescalinglaws. Thus
the modelsolutionprovidesinformationnot only aboutthe verticaldistributionof velocity,
but also the turbulentenergy.
The hydrodynamical modelis usedhereto providethesedimentdiffusivityrequiredfor the

determination of verticalprofilesof suspended sedimentconcentration.In particular,the
diffusivityis setequalto the derivededdyviscosity. It is assumedalso,for simplicity,that
the suspended loaddoesnot dampthe turbulence and,hence,modifythe eddyviscosity.At
thebed,a "reference concentration"boundaryconditionis adopted,thebottomconcentration
beingprescribexl as functionof the bed shearstress.On thisbasis,calculationsare madeof
the verticaldistributionof suspended sedimentin the casesof wavesand currentsalone,and
also in the case of combined wave-current flow and, hence, estimates are made of the
enhancement
of the suspended
load by wave action.
22.2 Formulation
Only the essentialdetailsof the formulationare describedhere;for a completeaccountthe

A G Davies 443
readeris referredto Davies,SoulsbyandKing(1988).Theprincipalfeatures

of themodelare
shownin outlinein Figure 1.
Linearized momentum equation for horizontal velocity

component (u) in x-direction:
0u
•t _ 1 ap
0 •x + •0 (K•}
•u p
0== pressure
fluid density
boundary
conditions
bed
[ : u = 0 on z = z K = K(z,t) = Eddy Viscosity
0u
surface : K • = 0 on z = h K = K(turbulent energy
I distribution)
Turbulent Energy Equation
[ Generation
of• turbulent
• Diffusion • •k(z,t)Dissipation
energy (c)]
Scaling Laws
Mixing Length l(z,t) < >
K,c : k,1
Sediment Continuity Equation

c = sediment concentration
ac_ az
at a (cs •+
ac WsC) sediment settling
velocity
sediment diffusivity
boundary conditions
bed : c(z 0, t) = bed

function
shear
ofstress
ac
surf ace : cs • + wc = 0
s
Assumption: Cs(z't) = B K(z,t) , B = constant
Figure 1. Model equations

444 A G Davies
Initially, a steadycurrentis generatedin the x-direction,drivenby a (constanO pressure

gradientcorresponding to a chosenvalueof thedepth-averaged currentfor waterof depth(h)
anda bedroughness of lengthscale
(Zo).The boundarylayeris assumed to be rough-turbulent.
It is assumed also that the surface wind stress is zero, and that the effects of thermal
stratificationand of the rotationof the Earth are negligible.
Once the currentis established,wavesare superimposed upon it co-linearly. After the

transientshavedied away,the combinedmotionconsistsof a thin oscillatory(wave+curren0
boundarylayerof thickess O(lm) embedded in a thickersteady(curren0boundarylayerwhich
may occupythe entire water column. The problemis solvedin such a way that the
depth-averaged currentis maintainedat its initial chosenvalue. On accountof the increased
bottomfrictiondue to the presenceof the waves,thisrequiresan increased(steady)pressure
gradient,whichis determinedby themodel. The wavemotioncomprises a uniformhorizontal
oscillation,drivenby a uniformoscillatorypressure gradientin thex4irection.Essentially,the
modelcan be viewedas havinga "rigid lid", suchthat thereis a uniform horizontalflow
betweenthefixedfree-surface (z=h) andtheflat roughbed(Z=Zo).The modelsolutionpredicts
the vertical structure of this flow.
The problemwhich arisesin solvingthe momentumequationis the familiar one of what

to assumefor theeddyviscosityK(z,t). As comparedwith previousmodelsin whichtheeddy
viscosityhas simply been prescribedas a functionof heightabovethe bed and has been
allowed no time-dependence, here we adopta more generalformulationin which K(z,t) is
taken as a functionof the instantaneous turbulentkinetic energy (t.k.e.) distribution. The
turbulenceclosurescheme(seeFigure 1) is basedon (i) the turbulentenergyequation,which
modelsthe local balancebetweenthe generation,diffusionand dissipationof t.k.e. (k); (ii)
a mixing length(1) which dependsuponthe distributionof t.k.e.; and (iii) standardscaling
laws relatingboth the eddy viscosity(K) and the turbulencedissipationrate (e) to k and 1.
The formulationdiffersfrom that adoptedby Davieset al(1988a,b)only by the inclusionof
a "surfacedampingfunction"in theexpression for themixinglength.[Thechoiceof damping
functionhasbeendiscussed by Davies(1990)]. Nezu andRodi(1986) havedemonstrated the
needfor sucha functionin conn<tionwith laboratoryexperiments on steadyflow. Thusthe
eddyviscosity(K) is determinedas a functionof height(z) abovethebedandphaseanglein
thewavecycle. The distribution of K differsfrom the functionalformsadoptedby previous
workers(e.g. GrantandMadsen(1979))and,hence,givesrise to ratherdifferentpredictions
for the peakbed shearstressin the wave cycle andotherquantitiesof interest.
The vertical distributionof suspended sedimenthas been modelledon the basisof the
sedimentcontinuityequation(seeFigure 1). Solutionof thisequationrequiresa knowledge
of the sedimentsettlingvelocity (W.) and physicallyplausibleassumptions for both the
sedimentdiffusivity(e.) and the bottomboundarycondition. The resultsin SectionIll are
basedon threeassumptions.Firsfly, it has beenassumedthat the presenceof suspended
sedimentdoesnot aff•t the distributionof t.k.e in the flow; while thismightbe reasonable
for low sedimentconcentrations, it is not a generallyvalid assumption (Soulsbyand
Wainwright(1987)).Secondly,thesedimentdiffusivity(e.) hasbeenassumed to be equalto

A G Davies 445
the eddy viscosity (K). Again, this is probably an oversimplification(see Dyer and
Soulsby(1988)).Thirdly,andmost importantly,a knowledgeof the "reference concentration"
at the bed hasbeenassumed.This form for the bottomboundaryconditionhasbeenadopted
mainlyfor simplicity. In particular,it hasbeenassumedthat the referenceconcentration at
level Z=Zocan be prescribexl as a functionof the instantaneous bed shear stressin the
wave-current cycle. At the freewatersurface,the conditionof zero sedimentflux is applied.
The systemof equationsin Figure 1 has been solved in two stages,by means of
numericallyconservative,finite differencemethodson a log-linearverticalgrid. Initially, the
required hydrodynamicalsolution has been obtained for wave, current and combined
wave-currentflows. The respectivevelocityand eddyviscositydistributions, determinedas
functionsof heightabovethe bed andphaseanglein the wave cycle,havethenbeenusedin
the solutionof the sedimentcontinuityequation.The resultspresentedbelowrelateto a case
studycarriedout with the followingparametersettings.The bed roughness lengthscale has
beentakenas Zo= 0.5cm,corresponding to sandtippleshavingthe approximatedimensions:
height--5cm, wavelength=50cm. The water depthhasbeentakenas 10m.
Initially,a steadycurrenthasbeengenerated with depth-averaged velocity50cm/s.Waves

havingvelocityamplitude50cm/sand period 10s have thenbeengeneratedwith the same
valuesof Zoandh. On thebasisof linearwavetheory,the surfacewavelength corresponding
to thisperiodis 92m, whichis considerably greaterthanthe waterdepthandjustifiesthe
assumption of horizontaluniformityin theformulation.The flow in theoscillatoryboundary
layeris roughturbulent.Finally,thewaveshavebeensuperimposed co-linearlyon thesteady
current.This superimposition hasbeencarriedout suchthatthe depth-averaged velocityhas
beenmaintainedat its initial valueof 50cm/s. The sedimentcontinuityequationhasbeen
solved for the current, wave and combined wave-currentflows, with various values for the
sedimentsettlingvelocityW,, for exampleW,--0.1,0.25 and 1.0cm/scorresponding to coarse
silt (nominal grain diameter D=0.04mm), very fine sand (D=0.07mm) and fine sand
(D=0.14mm). The useof a constant valueof Zoof 0.5cmimpliestheavailabilityof sediment
froma rippleAbedcomprising a mixtureof grainsizes,thismixturebeingconsistent with the
existenceof tipplesof the sizeindicatedabove.
22.3 Model results
22.3.1 Flow solutions
(i) Steadycurrent
As expected,the verticalprofileof horizontal(mean)velocity(U(z)) is logarithmicin the
lowerpartof the flow (Fig. 2i), andthe (mean)shearstressincreaseslinearlyfrom zeroat
thefree-surface
toitsmaximum
valueof•(Zo)=8.72dyn/cm:
(u.=2.95cm/s)
atthebed(Fig.2ii).

446 A G Davies
L• •vels
1000.0 -
500.0
.•, 100.0
• 50.0
N
• 10.0
•- 5.0
1.0
0.5 ß
o •o 20 •'o •'o 5'0 6'o

(Meen) HorizontelVelocity (cm/s)
Figure 2i: Vertical profiles of horizontalvelocityU(z) for the steadycurrentand <u(z,t)> for the
combinedwave-current(W+C) flow. (Angle brackets<> denotetime-averaging.)The modelgrid is
shown on the inner vertical axis.
Levels
ooo
8oo
600
400
20O
0
0 2 4 6 8 10 12 14
(Me(:]n)SheerStress (dyne/cm2)
Figure 2ii: Vertical profilesof shearstressx(z) for the currentand <x(z,t)> for the combinedwave-
current flow.

A G Davies 447
Levels
1000.0 -
500.0
• 100.0-
E
• 50.0
N c
._o, 1o.o
T 5.0 ß
Current
1.0
0.5
ß' ' 5'0 •60 •50
(Meen)T.K.E.
Figure2iii: Verticalprofilesof (cycle-averaged)

turbulentkineticenergyfor the current,wavesalone
and the combined wave-current flow.
Levels
1 ooo
8OO
W+C
Current
400
200 Waves
0 50 1 O0 150 200 250 300

(Mean) EddyViscosity (cm'/s)
Figure2iv: Verticalprofilesof (cycle-averaged)

eddyviscosityfor the current,wavesaloneand the
combined wave-current flow.

448 A G Davies
The associateddistributionsof t.k.e. (k) (Fig. 2iii) and mixing length(not shown)give rise to
an essentially
parabolicdistributionof eddyviscosity(K) (Fig. 2iv).
(ii) Waves
The verticalprofile of the amplitudeof wave velocity(Fig. 2v) hasthe familiarstructureof
a free-streamregion(z > 20cm), an "overshooting" region(5 < z < 20cm),anda logarithmic
near-bedregion. The associated cycle-averageddistributionsof t.k.e. andeddyviscosityfor
thewavesare shownin Figures2iii and2iv. The solutionconvergedafter60 cycles(10 min
real time).
(iii) Combinedwaves and current

The resultsfor wave-currentflow relate to the co-linearsuperimposition
of the waves (of
Section22.3 1(ii)) andthecurrent(of Section22.3 1(i)). Cycle-averaged profilesof velocity,
t.k.e.,eddyviscosityandshearstressare shownin Figures2i to 2iv. The effectof thewaves
is to greatlyenhancenear-bedt.k.e. comparedwith the currentalone,and henceto decrease
the velocityin the near-bedlayer. Sincethe depth-averaged velocityis requiredto remainat
its value for the currentaloneof 50cm/s,this necessitates an increasein the velocityin the
upperpart of the flow, which is broughtaboutby an increasein the meancomponentof the
pressuregradient. After all the transients
have died away (after about80 cycles)the final
converged,cycle-averaged shearstressprofile is linear with a valueof meanbed shearstress
of 14.0dyn/cm:
(ascompared
with8.72dyn/cm:
for thecurrent
alone).
Levels
1000.0 -
500.0
W(]ves
• 100.0 -
E
o 50.0
o, 10.0
-r 5.0 ,.
1.0
0.5
0 ' 1'0 ' 2'0 ' 3'0 ' 4'0 ' 5'0
VelocityAmplitude (cm/s)
Figure2v: Verticalprofileof the amplitudeof horizontalvelocityfor the wavesalone.

A G Davies 449
Extrapolation
of thestraight-line,
logarithmic segmentof thewave-current profile(15< z
< 200cm)in Figure2i indicates
an"apparent roughness"
forthewave-current boundary layer
of 2.5cm.Thisis 5 timesthephysical roughness lengthscale
(z0). In addition,
theslopeof
thestraight-line
segment suggests
an"apparent meanu."valueof 3.97cm/s. Thisissomewhat
largerthanthetruevalueof 3.74cm/s(i.e.thevaluecalculated
fromthemeanstress atthebed
level).Thisdiscrepancyisexplainedbythenon-logarithmicformof themeanvelocity profile
in the upperpart of the flow.
22.3.2 Solutionsfor suspended

(i) Steadycurrent
Verticalprofries
of suspendedsedimentconcentration
(c) areshownin Figure3i for three
grainsizeshavingthesettling
velocities
W, = 0.1,0.25and1.0cm/s.Eachprofilehasbeen
normalizedby thereference
concentration
c---Co•
atlevelz=z0.No attempt
ismadetoevaluate
Co•in whatfollows;Co•is simplytakenasthevalueof reference
concentration
appropriate
to u. = 2.95cm/sfor therespective
grainsizes.Thustheproblemis reduced
to theprediction
of the verticaldistributionof concentration
relativeto Co•.
Levels
1 000 -
800
600
400
s=O.1 cm/s
200
O I ß I ß I ß I ß I '
0.0 0.2 0.4 0.6 0.8
Concentrotion C/Coo
Figure3i: Verticalprofilesof suspended

sediment
concentration
forthesteadycurrent,
for threevalues
of settling
velocityW.. Eachprofilehasbeennormalized
by theappropriate
reference
concentrationCo•.

450 A G Davies
As expected,the profilesindicatea relativelyhigherconcentration of sediment in the

mid-depths asW, decreases.[Strictly,eachprordeshouldtendtozeroconcentration
at thefree
surfaceon accountof themixinglengthdampingfunction.In fact,dueto therathercoarse
verticalgridintervaladoptednearthesurface,a non-zero
valueisplotted.Thisvaluehasbeen
forced,through
a modified
surfaceboundary
condition,
to beconsistent
in theupperpartof
theflowwiththesedimentdistribution
associated
withthefamiliarRouseprofile.]
Levels
000 ,
800
600
4-00
Ws=O.
1cm/s
200 Ws=0.25
0
o .... .... .... 15' .....
Flux UC/Co (cm/s)
Figure3ii. Verticalprofilesof suspended

sediment
flux for the steady
current,
for threevaluesof
settlingvelocityW,.
Assuming thatthehorizontal
velocitycomponentof thesediment
grains
isequaltothatof
thefluid,thesedimentfluxateachlevel(normalized
bythereferenceconcentration
Co,)is
U(z)C(z)/Co,.Vertical
profiles
of thisquantity
areshown inFigure
3iiforthethreesettling
velocities
W, = 0.1,0.25and1.0cm/s.Thelevelsof maximum
sediment
fluxare445,41 and
1.8cmin therespective
cases.Valuesfor thedepth-averaged
concentration
andsuspended
sedimenttransport
ratefor steadyflow arediscussed
in Section22.4.
(ii) Waves Alone

Fromthe instantaneous
profilesof horizontal
velocityu(z,t)andeddyviscosity
K(z,t)
generated
by themodel,thebedshearstress
x0(O•(zo,Ohasbeendetermined
asa timeseries
through
thewavecycle.Thevalues of x0(Ohavethenbeenusedto generate
a timeseries
of
referenceconcentration
C(Zo,O
suchthat

A G Davies 451
c(zo,O= col%(O/ (22.1)
whereXo•= 8.72dyn/cm
•'is thevalueof bedshearstress
for thesteady
current.It shouldbc
noted,firsfly, that C(Zo,Ohas been taken as a power of the magnitudeof the normalized
instantaneous bed shearstressand, secondly,that no sedimentthresholdmotioncriterionhas
beenincludedin the formulation.Variousexpressions for c(zo,0havebeen suggested by
previousworkers,incorporating differentpowerlaw behaviours,for exampleN = 1/2 (Shi,
1985) andN = 1 (HagatariandEidsvik,1986). Sincehighervaluesthanthesein the range
N = 3/2 to 3 may be inferredfrom resultsfor the cycle-averagedreferenceconcentration(see
Dyer andSoulsby(1988) andNielsen(1988)), thisremainsan areaof considerable uncertainty.
The sensitivityof thepresentmodelresultsto changesin N is discussed in Section22.5. The
omissionof a thresholdshearstressfrom the expression for C(Zo,0takessomeaccountboth
of the persistentnatureof near-beAsedimentmotionin oscillatoryflow and of the effectof
settlinggrains.
Levels
1000-
800
600
400
L.x•Ws=1.0
2OO J Ws=0.25 .
o ' i ' ' '

Mean Concentration <c>/Co
Figure4. Verticalprofilesof cycle-averagedsedimentconcentrationfor the wavesalone,for three

valuesof settlingvelocityW,. Eachprofileis normalized
by the reference concentration
Co•for the
steadycurrent.
On the basisof the aboveformulation for C(Zo,

0 solutions
of the sedimentcontinuity
equationhavebeenobtainedwithN= 1 andB=I (i.e. the sediment
diffusivityis equalto the
eddyviscosity,
seeFigure1). In eachcase,theconverged hydrodynamicalsolutionfor waves

452 A G Davies
alonehasbeenusedrepeatedlyto providethe sedimentdiffusivityas a functionof heightand

phaseanglein thewavecycle. Thusthedetermination of the sedimentconcentration profile
hasbeentackledasan initial valueproblem,with zeroconcentration assumed initiallyat each
level. The cycle-averaged
concentrationprofilesfor the threesedimentsettlingvelocitiesare
shownin Figure4. At the bed,eachprofiletakesthe value<c>/C0,= 4.83 as dictatedby the
cycle-averaged
value<l o(t)/x0,1>. In eachcase,the sediment
remainsconfinedto the
near-bedlayer with the settlingvelocitystronglyinfluencingthe meanprofile. Converged
solutionswere obtainedafter 400, 150 and 25 cycles(67, 25 and 4 min. real time) for the
casesW. = 0.1,0.25 and 1.0cm/s,respectively.
(iii) CombinedWave-CurrentFlow
As for the caseof wavesalone,theeddyviscositygenerated by thehydrodynamical
modelhas
beenusedin the solutionof the sedimentcontinuityequation.The referenceconcentration
has
beencalculatedin the way described earlierandan initial valueproblemhasbeensolvedfor
c(z,t). Convergedsolutionswere obtainedafter 1400, 800 and 200 cycles(233, 133 and 33
minutesreal time) in the casesWs = 0.1, 0.25 and 1.0 cm/s, respectively. Each of the
cycle-averaged profilesin Fig 5i takesthe value <c>/C0,= 5.02 at the bed level z=z0(as
comparedwith the value4.83 for wavesalone). The concentration distributionsfor the case
of combinedwave-current flow maybe compareddirectlywith theresultsfor thecurrentalone
in Figure 3i. An enhancementin the suspended sedimentconcentration is evidentfor each
value of settlingvelocityconsidered.This arisesboth on accountof the increasein mean
bottomconcentration, and also on accountof the enhanceddiffusivity.
The profile of suspended

sedimentflux hasbeencalculatedin two ways. Firsfly,the flux
has beentakenas the productof the cycle-averaged
valuesof velocityand concentration at
eachlevel <c>. Althoughthis is a commonlyusedapproach,it is potentiallyinaccurate
on account of its omission of the "wave-related" contribution to the flux. If the horizontal
velocity(u) and concentration
(c) are written
u = tJ+u,,,+u
t andc = C+c,,,+ct (22.2)
whereU and C are the time-invariantcomponents of u and c, and the subscripts

w and t
denotethe wave(i.e. periodic)andturbulent(i.e. random)components of eachquantity,the
cycle-averaged
flux at a given level is
<uc>= UC+<u,,,c,,,>
+<u,c? (22.3)
whereanglebrackets< > denoteaveragingovera large(integral)numberof waveperiods.

Dyer and Soulsby(1988)
havesuggestedthatthe turbulentcontribution
to thisexpression
is
likely to be small (<1%) and that
<tiC>" fJC +<ttwC•,> (22.4)

A G Davies 453
The first termon the fight handsideof thisexpression

is simplythe meancomponentof flux
<c> since --U and <c>=C.
Levels
1 ooo
800
0.1 cm/s
600
Ws=0.25
400
200
6 ' i ' • ' • ' ; '

Mean Concentration <c
Figure5i: Verticalprofilesof cycle-averaged

suspended sedimentconcentrationfor wave-currentflow,
for threevaluesof settlingvelocityW,. Eachprofileis normalizedby the referenceconcentrationC0•
for the steadycurrent.
Levels
1000 -
800
600
400
200
5'o.... 16o.... 1•)0' ' '

Mean Flux <c>/Coc (cm/s)
Figure5ii: Verticalprofilesof cycle-averaged

sediment
flux <c>for thewave-current
flow, for three
valuesof settlingvelocityW,. Eachprofileis expressed
in termsof thereference
concentrationC0•for
the steadyflow.

454 A G Davies
Levels
6O
5O
'•' 40 Ws=1.0
N ,30
• 20
lO
o ß
o io'4'o''o do
Mean Flux: <c>/Co•; ...... <uc>/Co• (cm/s)
Figure5iii: Vertical profilesof cycle-averagedsedimentflux for the lower part of the wave-current
flow, for threevaluesof settlingvelocityW,. Profilesbasedupon<c> (full line) and<uc> (dotted
line) are shown in each case.
Levels
35-
3O
._.25
E
'•20
N
m15
•10
0
o 2 4 6 8 10 12
Mean Flux'
<c>/Co• ...... <uc>/Co• (cm/s)
Figure5iv: Verticalprofiles
of cycle-averaged
sediment
fluxasin Fig.5iii, butfora largervalueof

A G Davies 455
The profilesof flux calculatedin thesetwo waysare identicalabovethe waveboundary

layer,as shownin Figures5ii-iv. As for the currentalone,eachprof'tleof flux displaysa
maximumat a heightwhichincreases asW• decreases. Figure5ii showsprofilesof UC over
theentiredepth.The levelsof maximumsediment flux are608, 226 and6.4cmfor Ws= 0.1,
0.25 and 1.0 cm/s,respectively.Figure5iii showsprofilesof both<c> and<uc> in the
bottompartof the flow. Abovethe waveboundarylayer(z>35cm)thecurvesfor givenW•
are identical. However, within the wave boundarylayer, substantialdifferencesarise, the
presentformulation(with N=I) suggesting that the wave-relatedcontributionto the flux
increasesthe net Wansport. This is due to the fact that very close to the bed peak
instantaneous concentrations occurat aboutthe time in the wavecycleof maximumforward
velocity,givingriseto a nonlinearenhancement in theflux. Evidently,thiseffectis greatest
for coarsesedimentswhich are confinedto the oscillatoryboundarylayer. Figure 5iv
illustratesthisfor a fourth,largervalueof settlingvelocity(W;--4cm/s).
22.4 Depth-averagedconcentration
and sedimentflux
22.4.1 Notation
In orderto allow a generalassessment

of the results,the profilesof concentration
and flux
have been averagedover the depth. Theseresultsare denotedby an overbarnotationas
follows. For the steadycurrent,
AverageconcentrationC =
1 f••C(z)dz,
h-zo
(22.5)
$uspendeA
scdim•t flux q, = C(z) lJ(z)dz.
FollowingDyer and Soulsby(1988)and Soulsby(1988)we expressthe flux in termsof a

non-dimensional
"shapefunction"I, definedas follows:
q, = Co•(h-z
o)I, (22.6)
whereU is thedepth-averagedvelocity.Heretheflux is dividedintotwoparts:(i) a "stirring
term"givenby the referenceconcentration
Co,,whichis assumed to dependonly on the bed
shearstress
andproperties
of thegrains;
and(fi_.)
a "transporting
term"comprising
theproduct
of the volumetricflow rate per unit width U(h-Zo)and the shapefunctionI½. The later
calculationsconcernonly the transporting
term.

456 A G Davies
For wave-current
flow,depth-averages
aretakenaftercycle-averaging
asfollows:
Average
concentration
• - 1 f•'<c(z,t)>dz(22.7)
Thesuspended
sediment
fluxis calculated
onthebasisof thetwoapproaches
discussed
abovefrom the expressions:
(22.8)
q,•=f•'<c>dz
=f•:U(z)C(z)dz
(22.9)
Following Soulsby(1988) weagainexpressthesefluxesin termsof non-dimensional

"shape
functions"
I,• andI,2, definedfor wave-current
flow in therespectivecasesby:
(22.10)
q.• = co.r(h-zo)L• , (i= •2),
whereCorn
= <C(Zo,t)>is themeanoveran integral
numberof waveperiods of thereference
concentration
at Z=Zo.[It should
benotedthatCorn
andCo,will, in general,
bedifferent.]The
quantity_V
is thecycle-averaged,
depth-averaged
velocity
in wave-current
flow,suchthatthe
product
V(h-zo)is thevolumetric
flowrateperwidth. Sincethewavesaresuperimposexl
on
thecurrent
co-linearly,
•/' hasthesamedirection
asO.
22.4.2 Results
Calculationsof theabovedepth-averagedquantities
havebeenmadefor a rangeof valuesof
settlingvelocityW,. The resultstabulated
belowfor the current,waveandwave-current
casesare baseduponthe flow solutions
discussed earlier.
(i) Steadycurrent
The resultsfor depth-averaged
concentration
C and transportrateper unit widthq, (both
normalizedby thereferenceconcentration
Co,),together
withtheshapefunctionI,, arelisted
in Table 1. As expected,
eachquantitydecreasesrapidlywith increasing
W•.
Further
modelrunshavebeenperformed
withtheroughness
todepth
ratioZo•=10
-4. In
Figure6 theresults
obtained
forI, areplottedagainst
theparameter
S, = Ci)U/3W,,wherethe
dragcoefficient
Ci)= X(Zo)/pU:,
onthegraphof DyerandSoulsby(1988).
Theresults
relate
to a singlesettlingvelocityW,, anda singleroughness lengthscale
Zo(i.e. thesimultaneous
presence of differentbedformtypesis not permitted).Thevarious
curvesplottedin theinset
are for differentfunctionalformsfor theeddyviscosity,andhencesedimentdiffusivity(with

A G Davies 457
15=1). Evidently,Ic is fairly insensitive

to thechoiceof eddyviscosity,andthepresentmodel
resultsconformcloselyto the generaltrend. Althoughall the plottedresultsare derivedfor
low concentrations involvingno dampingof theturbulence by the sedimentgrains,it hasbeen
suggested by Dyer andSoulsby(1988)thattheresultsmayalsobe applicablein casesin which
drag reductionby sedimentin suspension doesoccur.
Table 1' Depth-averaged

resultsfor the steadycurrentobtainedwith I• = 1. Flow parameters:
Zo= 0.5
cm, h = 1030 cm, U = 50 cm/s, u. = 2.95 cm/s.
w I
q½/c Oc
(cm/s) (cm2/s)
-! -!
0.1 5'22x10 2 ß54x 104 5'09x10
-!
0.25 2.06x10 -• 9 ß55x 103 l'91xlO
-2 -2
0-5 5'12x10 2' 14xlO 3 4'29x10
-2 -2
0.75 1'54x10 5' 57x102 l'12xlO
-3 -3
1.0 5'73x10 1' 69x102 3'38x10
-3
1.5 1'47x10 2 ß47x101 4.95x10 -•
2.0 6.77x10 -4 6' 39x10 ø 1' 28x10 -•
-5
3.0 3 ß03xlO -4 1' 26x10 ø 2-52x10
-1 -5
4.0 1.91xlO -4 4.46x10 8'94x10
-2 -6
6.0 1' 07x 10 -4 7' 39x10 1'48x10
(ii) Wavesonly and combinedwave-current flow

As notedearlier,thecycle-averaged concentration
at thebedlevelZ=Zofor thecaseof waves
aloneis <c>/C0,= 4.83, andfor the combinedwave-currentcaseis <c>/C0,= 5.02. Table2
showsresultsfor themean,depth-averaged
concentration
<•'•/C0,calculated
for a rangeof
settlingvelocitiesW,. If theresultsfor wavesalonearecompared with thosefor thecurrent
alonein Table 1, it is seenthatthedepth-averaged concentration
for givenW, is generally
greaterfor the waves. Only for the two smallestvaluesof W,, for whichthe grainsare
stronglydiffusedinto the outerflow, is theopposite
resultpredicted.

458 A G Davies
Sc: CoO/3Ws
001 0.1
1.0
XPresent model
10-I _ P
FC
10-• _
Izo/h
--10
-•'I L
lO-• _ Eddy viscosity
Ic
0'•
$m
F
z/h
ø'6 FHFC L
10-•_ 0•
02
•o-6 _
0. 0 O0
[ I
0.1 02 K/•cu.h
03
I
04 05 0.6
Figure6: Shapefunction
I• versus
Scforsteady
Howwithz0/h=10
-4. Thelabelled
curves
correspond
to the vertical profilesof eddy viscosityshownin the inset, as explainedin detail by Dyer and
Soulsby(1988) andSoulsby(1988) (•: = Von Karman'sconstant).Thepresentmodelresultsareshown
by the crosses.

A G Davies 459
Table2: Cycle-averaged,depth-averaged
concentration
results
obtained
withI• = 1 andN = 1 forboth
wavesonlyandcombined wave-current
(W+C)flow. Flowparameters:z0= 0.Scm,h = 1000cm;mean
velocityV = •'• = 50 cm/s;wavevelocityamplitude
• = 50 cm/s,period= 10s.
<C>/ <C>ratios
Ws Co½
(sin/s)
Waves W + C W + C W + C
Waves Current
O' 1 3' 97x10-1 3' 40x100 8' 6 6' 5

O' 25 1-65x10 -1 1' 76x100 10' 7 8' 5
O' 5 7' 33x10-2 6- 38x10-• 8' 7 12' 5
O-75 4.27x10 -2 2 ß63x10-• 6' 2 17 ß0
1' 0 2.82x10 -2 1' 23x10-• 4 ß4 21' 5
1' 5 1.50x10-2 3' 90x10-2 2' 6 26' 6
2 ß0 9.43x10 -3 1' 68x10-2 1' 8 24 ß8
3' 0 4' 86x10-3 6- 63x10-3 1' 4 21' 8
4'0 3-07x10 -3 3'84x10 -3 1'3 20' 1
6' 0 1-66x10 -3 1' 95x10-3 1' 2 18' 3
8 ß0 1- 10x10-3 1- 26x10-3 1' 2 -
12'0 6-37x10-4 7- 19x10-4 1' 1 -
16'0 4' 65x10-4 5' 18x10-4 1' 1 -
The resultsfor wave-currentflow in Table 2 are comparedwith thosefor boththe current

aloneand the wavesalone. As comparedwith the valuesfor the current,thereis a fairly
uniformorderof magnitudeincreasein <c>/Co•. In contrastto this, comparisonwith the
resultsfor wavesalonesuggests rathersimilarresultsfor thelargervaluesof W,, but anorder
of magnitude differencefor thesmallervalues.Resultsfor thesediment transportrate(q+)per
unit width in combinedwave-currentflow, and alsofor the shapefunction(I+), are shownin
Table 3 for varioussettlingvelocitiesW,. The resultshavebeencalculatedbothexcluding
(subscript1) and including(subscript 2) the wave-related contribution
to the sedimentflux,
as discussed earlier.
Comparison with theresultsin Table1 for thecurrentalonerevealsa substantial

enhancement
in theflux (q.) asa resultof waveaction.For thevaluesof W• considered, theflux increases
by betweenone and two ordersof magnitude. Both q. and I. decreaserapidly as W•
increases.However,I., decreases muchmorerapidly thanI.2, indicatingthe importanceof
the wave-relatedcontributionto the flux for the largergrains.

460 A G Davies
Table3' Meansediment transport

rate(q+per unitwidth)andshapefactor(I+) for combined
wave-
currentflow. Flow parameters
areas for Table2.
Ws q+•/Co
c q+x//qo I +1 q+
2/coc q+
2/ /qo I+2 I +2I+1
(cra/s) (cra
2/ s) (cra
2/ s)
O'1 1' 65x105 6 ß5 6' 59x10-1 1' 65x105 6 ß5 6' 59x10-1 1' O0
0'25 8.15x104 8'5 3'26x10-1 8.16x104 8'5 3'26x10-1 1'00
0'5 2-70'104 12'6 l'08xlO -1 2.70x104 12'6 l'08xlO -1 1'00
0'75 9.91x103 17-8 3'96x10-2 9-94x103 17'8 3'97x10-2 1-00
1' 0 4 ßOlxlO3 23 ß8 1' 60xlO-2 4 ß04xlO3 23 ß9 1' 61xlO-2 1' O1
1'5 9-13x102 36'9 3'65x10-3 9-42x102 38'1 3'76x10-3 1'03
2'0 2'45x102 38'3 9'77x10-4 2.73x102 42'7 1'09'10 -3 1'12
3'0 4.72x10 ! 37'5 1'88x10-4 7.44x10 ! 59'1 2'97x10-4 1'58
4 ß0 1' 74x101 39 ß1 6' 96x10-5 4 ß04xlO1 90' 5 1' 61xlO-4 2 ß32
6'0 4'87x10ø 65'9 1'94x10-5 1.74x101 235 6'94x10-5 3'57
8' 0 2 ßOlxlOø - 8' 03xlO-6 8' 35x10ø - 3' 33x10-5 4' 15
12ß0 5 ß23x10-1 - 2' 09xlO-6 2 ß51xlOø - 1' OOxlO-5 4 ß81
16'0 1.71xlO -! - 6.83x10-7 8.90x10-! - 3-55x10-6 5'20
As for the currentalone, furthercalculationsof shapefunctionhave been made with

zo•=104,andin Figure7 theseresultshavebeenincluded on the_graphof I. versusS.
presented
by Soulsby(1988).
Heretheparameter S. is equalto (CmV/3W,),wherethedrag
coefficient
Cm= I <Xo(t)>
I/PV2. Asexplained
in detailbySoulsby, theresults
shown arefor
the values (i) or.=5 where or. is the ratio of the maximum (instantaneous)
to mean
(cycle-averaged)
valueof u. in the wave-current
cycle;and(ii) B.=100where13.Zo
is the
thickness
of the oscillatoryboundarylayer. The variouscurvesincludedin Figure7 have
been obtained from different assumedfunctional forms (shown in the inset) for the
time-invarianteddyviscosityin wave-current flow. As compared with theresultsin Figure
6, theflux doesnowdependsignificantly upontheverticalprofileof eddyviscosity.As far
asthepresent modelresults areconcerned,a consistent
comparison canbemadeonlybetween
I.• andlheothercurvesplotted,sincethewave-related contribution hasnotbeenincluded in
Soulsby's calculations.
However,valuesfor I. 2arealsoincluded in Figure7. Theseindicate
a markeddifference, by a factorof about6, betweenI.• andI.2 for smallS., for thereasons
discussedearlier.Despitethisdifference,
thepresentvaluesall fall well withinthewiderange
of uncertaintydisplayedby the othermodelresults.

A G Davies 461
zo/h
=10
-4,
(z+=
5,/•+=1001
10-z x Presentmodet
TS2
1-0
Eddy viscosity
10-4
C J2
0'8
I+ I+2
10-6
0-6 /•"GM
z/h
I+1 0-4
10-8
0.0
0.0 0-2 0.4 0.6 0-8 1'0 1-2
K/•<u,>h
100-001 0!01 0-1 1 10

S+: C.+V/3Ws
Figure7: ShapefunctionI+versusS+forwave-current
flowwithz0•=10-4,ct+=5
andI]+=100.The
labelled
curves
correspondtothe(time-invariant)
vertical
profiles
ofeddyviscosity
shown intheinset,
asexplained
in detailbySoulsby(1988)
(K= VonKarman'sconstant).
Thepresentmodel results
for
I+l and1+2areshownby thecrosses.
22.5 Discussion
Theresults
inSections
22.3and22.4wereobtained
withthesediment
diffusivity
setequal
to
theeddyviscosity
(B=1) andwithN=1 in theformulafor thereference
concentration.
These
assumptions
are now discussed.
22.5.1Effectof changing15
Forsteady
flowthesediment
continuity
equation
(seeFigure1) reduces
to

462 A G Davies
de
% dz + WsC' 0 (22.11)
It followsthatif gs=BK,whereB is independent

of z,
(22.12)
dz + - o
For unsteady flow, B cannotbe absorbed

intothe settlingtermin thisway on accountof the
timerateof changeterm8c/i)t. In orderto assess theimportance of thistermin thecontext
of theearliercasestudies,modelrunshavebeenperformedin whichboth15andW, havebeen
variedat a fixed value of W,/B.
For the wavesalone,halvingthe valueof B from 2 to 1 at W,/B = 0.5cm/s,andfrom 1 to

0.5 at W,/8 = 2cm/s, producesin each casean increaseof 5% in the cycle-averaged,
depth-averaged concentration.This increase,which would not occurwith a time-invariant
eddyviscosity, is attributableto thecombined effectof unsteadiness
in theeddyviscosity and
sedimentconcentration distributions.For the wave-currentcase,the resultsarisingfrom the
samechangesin B differ dependinguponthe valueof W,/B. The effect of halvingB from
2 to 1 with W,/8 = 0.5cm/sis to increase<c>/Co,by 28%, and to increasebothq.• and q.2,
andhenceI.• and I.2, by 32%. The effectof halvingB from 1 to 0.5 with W,/B = 2cm/sis
to increaseboth<•'•/Co,and alsoq.• and I.• by 6%, but to leaveq.2 and I.2 virtually
unchanged. Thisindicatesthattheeffectof changingB is ratherlessfor coarsegrainsconfined
to the near-bedregionthanfor finer groinswhichare diffusedinto the outerpartof theflow.
It wouldappearfromtheseresultsthattheeffectof the timerateof changeterm•}c//)tis

significant,and that the valueof 8 shouldnot, in general,be absorbedinto W, in the way
indicatedabove. It shouldbe addedthat 8 may, in reality, be a functionof z, and that our
understanding of its behaviouris presentlyincomplete(Dyer and Soulsby, 1988).
22.5.2 Effect of changingthe power N
Previousinvestigatorshave suggesteddifferentvaluesfor the power N in theft formulae

definingthereferenceconcentration. In orderto assessthesensitivityof thecasestudyresults
to changesin N, calculations for wavesaloneand for the combinedwave-current casehave
beenrepeatedwith N= 1/2 and3/2 for thevaluesW, = 4.0, 1.0 and0.25cm/s.For W, = 4cm/s
the sedimentgrainsareconfinedto thenear-bedlayer,andfor 0.25cm/stheyarewell diffused
into the outerpart of the flow. At the bed level Z=Zothe effect of changingN is to alter
significantlythevalueof <C(Zo,0>.For wavesalone<c>/Co,takesthevalues2.079,4.829 and
11.87 for N=l/2, 1 and 3/2, respectively.For the wave-currentcasethe valuesare 2.098,
5.019 and 12.89,respectively.Thesevariationsat the bed exertthe primarycontrolon the
depth-averaged
values<•'•/Co,. In particular,
the valuesobtained
for <•'•/Co,withN=l/2
and 3/2 fall within 5% of the value obtainedby scalingthe modelresultfor N=I by the

A G Davies 463
appropriatebottomvalue <c(zo,0>. Similar commentscan be madeaboutthe resultsfor

transportrateq+•(i=1,2) listedin Table4, thoughfor the caseof W,--4cm/sthe variationsare
morepronounced thanthoseobtainedby a simplescalingbasedon <c(zo,0>. This is apparent
in the resultsfor I,• (i= 1,2). If changesin I+ with N are expressed
in termsof the valuefor
N= 1 in eachcasethen,for W,=0.25cm/s,bothI+• andI+•.changeby -4% with N= 1/2 and+3%
with N=3/2. For W,= 1.0cm/s,the respectivechangesin bothI.• and I+: are -6% and +5%.
For W,=4.0cm/sthe valuesfor I+• are -10% and +8%, and for I+•.are -25% and +23%,
indicatingagainthe importantrole of the wave-relatedcontribution to the flux.
It may be concludedthat the choiceof powerlaw behaviourhasa stronginfluenceon the

resultsas far as the averagebottomconcentration<C(Zo,0>is concerned,but that its effect
on theshapefactorI. is relativelysmall,exceptfor grainsconfinedto theoscillatoryboundary
layer.
Table 4: Mean sedimenttransportrate (q+per unit width) and shapefactor(I+) for combinedwave-
currentflow. Values for power N in the referenceconcentrationformula are as indicated. Flow
parametersare as for Table 2.
Ws q+ (cm
2/ s) q+2 (cm2
/ s)
(ore/s)
•/co• /c o•
N=l/2 N=i N=3/2 N=l/2 N=i N=3/2
0.25 3 ß28x104 8 ß15x104 2 ß16x10 s 3 ß28x104 8' 16x104 2 ß17x10 s

1'0 1'59x103 4'01x103 1'08x104 1'59x103 4'04x103 1'09x104
4 ß0 6' 56x10 ø 1' 74x10 • 4 ß85x10 • 1' 27x10 • 4' 04x10 • 1' 28x102
I I
+1 +2
N=l/2 N=i N=3/2 N=l/2 N=i N=3/2
-1 -1 - 1 -1 - 1 -1
0' 25 3 ß 13x10 3 ß 26x10 3 ß 37x10 3 ß 13x10 3-26x10 3 ß 37x10
1'0 l'51x10 -2 l'60x10 -2 1'68'10-2 1'52x10 -2 1'61x10-2 l'70x10 -2

4'0 6'26x10 -s 6'96x10 -s 7'54x10-s 1'21x10 -4 1'61x10-4 1'99x10 -4
22.5.3 Phaseof c(z0,t)in relationto the bed shearstressx(z0,t)
The formulafor thereferenceconcentrationwhichhasbeenusedthroughout is basedon the

assumptionof an instantaneousresponseof C(Zo,0to changesin the bed shearstress.This
assumptionmaybe oversimplistic, sincegrainsin closeproximityto thebedmay taketime
to respondto changesin X(Zo,0. In otherwords,a phaselag may be appropriate in the
specificationof the referenceconcentration,
as follows:

464 A G Davies
C(•o,t) Co•]X(•o,t-
At)lxo• (22.13)
Such an approachhas been taken,in effect, by Bailard(1981)who neglectedthe phase

differencebetweenthe bed shearslressand the near-beAfluid velocity and assumedthe
time-varyingsedimentWansport
rateto be in phasewith the flee-stream
velocity.
A numberof runs have been conductedto test the sensitivityof the model resultsto
variationsin At. For a lag At corresponding to 36ø of phase(such that the reference
concentration is roughlyin phasewith the outerfree-streamvelocity),the depth-averaged
concentration for wavesalone increasesby 12% for W,=lcm/s and by 3% for W,--4cm/s.
Theseresultswereobtainedwith N--l; percentage changeswereslightlylesswith N=l/2, and
slightlygreaterwith N=3/2. For the caseof wave-current flow with W•=4cm/sandN=I, the
samephaselag givesrise to a 5% increasein <c>/Co,,a 12% increasein I+• and an 11%
decreasein I+•_.A morecompleteaccountof thesecomparisons will be givenelsewhere.This
is a topicwhichwarrantscarefulinvestigation,not leastin connectionwith vortexformation
andsheddingaboverippledbeds,whichis knownto be importantin the processof sediment
entrainment(Nielsen, 1988).
22.6 Conclusions
A substantialenhancement in the suspended sedimenttransportrate may occurwhenwaves

are superimposed on a current. This arisesfrom the combinedeffect of an increasein the
time-averaged, bottomreferenceconcentration, and an increasein the verticalmixing of
sedimentin the water column. The resultsof a case study indicateenhancements in the
transportrate of betweenone and two ordersof magnitude,dependinguponthe sediment
settlingvelocity.
For sedimentgrainswhich are well diffusedinto the outerpart of a wave-current

flow, a
good approximationto the net sedimentflux may be obtainedfrom the productof the
cycle-averaged profilesof horizontalvelocity andsedimentconcentration
<c>. However,
for largergrainsconfinedto theoscillatoryboundarylayer,the "wave-related"
contributionto
the flux (incorporatedin the term <uc>) may dominatethe net transport.
The modelresultshave been expressedin termsof the "shapefactor"of Dyer and

Soulsby(1988)andSoulsby(1988),whichrepresents a non-dimensionalisation
of thesediment
flux by the productof the volumetricflow rate (per width of the flow) and the bottom
referenceconcentration(cycle-averaged in thecaseof wave-currentflows). For currentsalone,
thepresentmodelis in closeagreement with all previousmodels,confumingthatthe shape
functionis insensitiveto the form of the verticalprofile of eddyviscosity. However,for
wave-current flows,theshapefunctionis stronglydependent upontheeddyviscosity profile,
particularlyfor grainsconfinedto the oscillatoryboundarylayer. The predictionsof the

A G Davies 465
presentmodelfor suchcasesfall well withinthe broadbandof uncertaintyassociated

with
previousmodels.
The present results are dependent upon the assumed relationship between the
(instantaneous)bed shearstressand the (instantaneous) bottomreferenceconcentration.The
powerlaw relationship betweenthesequantifies exertsa significant
influenceon the transport
rate. Thisarisesprimarilythroughchanges in thereferenceconcentration,andonly to a small
extentthroughthe shapefunction. The phaserelationship betweenthe two quantifiesalso
affectsthe results,not leastin respectof the "wave-related"contributionto the flux. This is
a matterwhich warrantsfurtherinvestigation.
Acknowledgements
RichardSoulsby(HydraulicsResearch Ltd., Wallingford)is thankedfor hishelpfulcomments,
and also for allowinguseto be madeof Figures6 and 7. Bob Everitt and AngelaWilliams
are thankedfor their help with the figuresand the text. Financialsupportfor the work was
providedby the U.K. Ministryof Agriculture,FisheriesandFood.
22.7 References
Ballard, J.A., 1981: An energeticstotal load sedimenttransportmodel for a plane sloping beach.
Journalof GeophysicalResearch,86, Cll, 10938-10954.
Bakker,W.T., andT. van Doom, 1979: Near-bottomvelocitiesin waveswith a current.Proceedings

of the 16th InternationalConferenceon CoastalEngineering,1978, American Society of Civil
Engineers,1394-1413.
Bijker,E.W., 1967:The increaseof bedshearin a currentdueto wavemotion. Proceedings

of the 10th
International
Conferenceon CoastalEngineering,1966, AmericanSocietyof Civil Engineers,746-765.
Davies,A.G., 1986:A numericalmodelof the waveboundarylayer.ContinentalShelfResearch,6, 6,

715-739.
Davies,A.G., 1990:A modelof theverticalstructureof the waveandcurrentbottom boundarylayer.

pp.263-297in ModelingMarine Systems,Vol.2, ed A.M.Davies,CRC Press.
Davies,A.G., 1991:Transienteffectsin wave-current

boundarylayerflow. OceanEngineering,18, 1/2,
75-100.
Davies,A.G., Soulsby,R.L., andH.L.King, 1988:A numericalmodelof thecombinedwave andcurrent

466 A G Davies
bottomboundarylayer. Journalof GeophysicalResearch,93, C1, 491-508.
Davies, A.G., Barber,M.E., and R.L.Soulsby,1990: Vertical currentstructureand bedloadsediment

transportin a combinedwave-current boundarylayerflow. pp.52-67in Physicsof ShallowSeas,eds.
Wang H.,Wang J. andDai H., ChinaOceanPress.
Dyer, K•R., andR.L.Soulsby,1988: Sandtransporton the continentalshelf. AnnualReview of Fluid

Mechanics, 20, 295-324.
Fredsoe,J., 1984:Sedimenttransportin currentandwaves.Instituteof Hydrodynamics

andHydraulic
Engineering, SeriesPaper35, TechnicalUniversityof Denmark,Lyngby,37pp.
Grant,W.D., andMadsen,O.S., 1979:Combined

waveandcurrentinteraction
with a rough bottom.
Journalof Geophysical
Research,84, C4, 1797-1808.
Grass,AJ., 1981:Sedimenttransport
by wavesandcurrents.
ReportNo. FL29, SERCLondon Centre
for Marine Technology,26pp.
Hagatun,K., andK.J. Eidsvik,1986:Oscillatingturbulent

boundarylayerwith suspended
sediments.
Journalof GeophysicalResearch,91, Cll, 13045-13055.
King,H.L.,Davies,A.G.,andR.L. Soulsby,1985: A numerical

modelof theturbulent
boundary
layer
beneath
surfacewavesandtides.Instituteof Oceanographic
Sciences
ReportNo. 196,90pp.
Nezu,I., andW.Rodi,1986:Open-channel
flowmeasurements
witha laserDoppleranemometer.
Journal
of HydraulicEngineering,112, 5, 335-355.
Nielsen,P., 1988:Threesimplemodels
of wavesediment
transport.
Coastal
Engineering.
12,43-62.
Shi, N.C., Larsen,L.H., andJ.P.Downing,

1985: Predicting
suspended
sediment
concentration
on
continentalshelves. Marine Geology,62, 255-275.
Smith,J.D., 1977:Modelingof sediment transport

oncontinental shelves,pp.539-577in The Sea,Vol.
6, eds.Goldberg,E.D., McCave,I.N., O'Brien,J.J.,andSteele, J.H., Interscience,
New York, 1977.
Soulsby,
R.L.,1987:Therelative
contributions
of waves
andtidalcurrents
tomarine
sediment
transport.
HydraulicsResearchLtd., Wallingford,ReportSR 125.
Soulsby,R.L., 1988:The structure

of suspendedsedimenttransport
formulaefor unidirectionaland
wave-plus-currentflows.Proceedingsof the Symposium on Mathematical Modellingof Sediment
transport
in theCoastalZone. DanishHydraulicInstitute,May 1988,68-78.
Soulsby,
R.L., andB.L.S.A.Wainwright,1987:A criterionfor theeffectof suspended
sediment
on
near-bottom
velocityprofiles.Journalof HydraulicResearch,25, 3, 341-356.

23
Some considerations on mathematical

modellingof morphological
processes
in
tidal regions
Z B Wang
Abstract
Theconstructionandapplication
of a mathematical
modelformorphologicaldevelopment
in
tidalregions
isconsidered.
Themodel consists
ofa flowmodule,a sediment
transport
module
anda bedlevelchangemodule.Thestructure
of themodelandthechoiceof theflowmodule
andsediment transport
moduleare discussed. The application
aspects
of the modelare
consideredwith a casestudyfor the Yantze estuary.
23.1 Introduction
Themorphologicaldevelopmentof estuarine
andcoastalregions isa longtermprocess.The
timescaleof morphological
processes mayrangefromweeksto decades or evencenturies.
Morphologicaldevelopment
changes thegeometryandbathymetry of a regionandthereby
it
influences
theflowpattern
andallotherphysicalprocessesin theregion.Thereforeknowledge
aboutthemorphological
development is important
for hydraulic engineering aswellasfor
studying
longtermvariability
of otherphysicalprocesses in esmarine andcoastalregions.
Morphologicaldevelopment ofesmarine
andcoastalregions isa verycomplicatedprocess.
Manyfactors
areinvolved in thisprocess:
fiverflow,fide,density flow,windandwaves,
sediment
properties,etc.Although considerable
knowledge existson all theseindividual
elements,
very little is knownhow to includetheminto a mathematical modelfor
morphological
development. Mathematical
modellingof thisprocessisstillinanearlystage.
At DelftHydraulics
themorphological
modelDELMORis underdevelopment.
At this
momentit is not wiseto try to developa generalapplicable
modelby includingall the
significant
factors
onecanimagine.
Thiswill leadto a verycomplicated
model.Littleis
knownaboutthebehaviour
of sucha complicated
model.Thereforethestepby stepstrategy

468 Z B Wang
is followedfor the developmentof DELMOR. In the first stepthe modelwill only include
fiver flow and tidal motionas drivingforcesfor sedimenttransport.Densityflow, wind and
wavesare not takeninto accountfor the time being.Furtherit is restrictedto non-cohesive
sediment.The fundamentalaspectsof sucha simplifiedmodelwerestudiedby Wang (1989),
who developedthe researchmodelESMOR. Someconsiderations on the consauctionand
applicationof sucha relativesimplemorphological modelaremadein thispaper.In thefuture
when the behaviourof such a simplermodel is sufficientlyunderstoodthe model will be
extendedstepby stepby taking more factorsinto account.
In the following sectionthe structureof the model is considered.The model consistsof

threesub-modelsviz. the flow model,the sedimenttransportmodelandthe bed level model.
The three individual sub-models are discussed in section 23.3. Section 23.4 considers the
applicationaspects of the model.Attentionis paidto thedominating conditionandthechoices
of the majorparameters controllingthecomputation procedure. In thissectionthe casestudy
for the Yantzeestuaryis usedas an example.In section23.5 somenumericalaspectsare
considered. Attentionis paidto themodellingof thesourcetermandtheboundaryconditions.
Finally the major conclusions are summarised in section23.6.
23.2 Structure of the model
Despite of the impressiveadvancein computertechnology,a major problem for the

mathematical modellingof morphological
processes is still the computational
cost.Because
of thecomplexityandthelongrunningtimeof themodel,an improperlydesignedmodelwill
costtoo muchcomputationtime makingit uselessin practice.
The morphological
developmentinvolvesthevariations
of flow, sediment
transport
andbed
level.Thesethreesubprocesses
all interactwith eachother.This meansthat the equations
describing
theflowfieldvariables
(h,u,v....), thesediment
transport
variables
(c,s•,Sy...)
and
thebexllevel(zO arecoupledwitheachother.ff thesecoupledequationsaresolvedtogether
thetimestepusedshouldbe muchsmallerthanthesmallestsignificanttimescale,i.e. thetidal
period. Obviouslysucha computationfor a long period (decadesor even centuries)is
impossible.
In otherwords,simplifications
in morphological
modelsarevirtuallyinevitable.
In orderto simplifythe problemtwo importamassumptions are made.The lust oneis the

socalledlow concentration assumption:the influenceof the sedimenttransport
on theflow
fieldis negligible.
Thisassumption isjustifiedin mostestuarineandcoastalregionssincethe
sediment
concentration
by volume
mostly
isof theorderof magnitude
of 10'3.Thisassumption
makesit possibleto separatethe flow moduleand the sedimenttransportmoduleinto
uncoupled sub-models. The second assumption
is thesocalledquasi-steadinessassumption:
the large differencebetweenthe morphological time scaleand the time scaleof the flow
makesit possibleto separatethe flow computation and the bed level computation. This
assumption is justifiedfor smalland moderateFroudenumbers(de Vries, 1959, 1965),

Z B Wang 469
commonin most estuarineand coastalregions.This assumptionmakes it possibleto

distinguishbetween thelongtermvariation processes
(i.e.thebedlevelchange) andtheshort
termvariationprocesses (i.e. theflow andthesedimenttransport).
It is thenpossible
to apply
two differenttime stepsin themodel,onefor thebedlevelchangeandonefor theflow and
sedimenttransport.
The two assumptions

makeit possibleto dividethemorphological
modelintothreesub-
models,viz. the flow model,the sedimenttransportmodeland the bed level model.In one
morphological
time stepthesethreesub-models
haveto be run successively.
It is further
assumedthat the flow field and the sedimenttransportfield remainperiodicduringseveral
tidalperiods.
In otherwordsthemorphological timestepcanbeaslargeasa numberof tidal
periods.Furthermore,thebedlevelchange afteronemorphologicaltimestepisoftensosmall
thatthecontinuitycorrection
methodcanbe appliedfor calculatingtheflow fieldduringthe
nextfew steps(Hauguel,1978,de Vriend,1985).In this methodthe waterlevel and the
dischargeareassumed tobethesameasin theprevious step.Thewaterdepthandthevelocity
field changeonly dueto the bed level change.All theseconsiderations
togetherlead to a
structureof the modelas shownin fig.1.
LINITIAL
STATE
CURRENT
FIELD
ONE
INTIDAL PERIOD
VELOCITY FIELD :•
CONCENTRATION
FIELD •
TRANSPORT
FIELD nuj
o•
UPDATE
TOTAL
TRANSPORT •
I BED
LEVEL
Figure1' Structureof the model
The morphological timestepandthenumberof timesof continuitycorrection

aftereach
completetidalflow computation aretwoof themajorparameters
in themodelandthechoice
of themwill be the subjectof a discussion
in section23.4.

470 Z B Wang
23.3 The three sub-models
As explainedin the previoussection,the morphological modelunderconsideration can be

dividedinto threesub-models, viz. the flow model,the sedimenttransportmodelandthe bed
level model. Much work was done on the individual sub-modelsby many researchers.
Differentcombinations of all typesof the sub-models arepossible.However,theconstruction
of the morphologicalmodelas a wholeis lesselaborated.Thereforethe choiceof eachsub-
model is paid attentionto in this section.Thesechoicesshouldbe basedon the validity
(accuracy)andapplicability(cost)considerations. Thedetailsof theindividualsub-modelsself
are not considered here.
The Flow Model: In the caseof esmarineand coastalproblemsone can choosefrom two
possibilitiesfor the flow model, viz. the two-dimensional(horizontal) and the three-
dimensional model.At present,computercapacityrestrictsaffordablelongtermmorphological
computations basedon a fully 3D flow model.Thereforethe quasi-3Dapproachis chosenin
ESMOR aswell asin DELMOR. The depthaveragedflow is computedwith a 2DH tidalflow
model.A profile modelis appliedto describethe verticaldistributionof the velocityin the
longitudinaldirectionanda secondary flow modelis appliedin orderto takeinto accountthe
velocity componentin the normal direction.In DELMOR, the depth averagedflow is
computedwith TRISULA. The secondary flow modeltakesinto accountthe influenceof the
curvatureof the streamlines of the depthaveragedflow andthe geostrophic acceleration(de
Vriend, 1981, Kalkwijk and Booij, 1986). It shouldbe mentionedthat TRISULA is a multi-
dimensional flow modeldevelopedat Delft Hydraulics.which is able to performfully 3D
computations. This makesit easierto extendDELMOR to a full 3D modelin the futureif
necessary.
TheSedimentTransportModel:The sedimenttransportis dividedintobedloadtransport and

suspended transport.The bed loadtransportis assumed to adjustinstantaneously to the local
transportcapacitywhichis relatedto thelocalflow conditions via a transportformula.For the
suspended transportonecanchoosefrom variouspossibilities varyingfrom a simpletransport
formulato a sophisticated fully 3D model (van Rijn, 1987, Wang, 1989). The available
computercapacityrestrictsthe possibilityto applythe full 3D transportmodelto be applied
in thelongtermmorphological computations.On theotherhandthesimpletransport formula
is usuallyinsufficientto modelthesuspended sedimenttransport in a varyingflow field asin
an estuary.Thereforethe quasi-3Dapproach, originallydeveloped by Galappatti(1983, see
alsoGalappattiandVreugdenhil1985)andlaterextended by Wang(1989), is chosen.
This quasi-3dmodelis derivedby solvingthe full 3D sedimentconcentration

equation
asymptotically.
Fromtheasymptotic solution
of the3D sediment
concentration
equationa 2DH
equationis derivedfor the depthaveragedsedimentconcentration.

471
Z B Wang
in which
c = depthaveraged
sediment
concentration;
u, v = depthaveraged
velocitycomponents;
T. = adaptationtime;
¾ = coefficients;
c. = depthaveraged
equilibrium
sediment
concentration.
The termon therighthandsideof thisequation is a source

term.Whentheactual
concentration
c is smaller
thantheequilibrium
valueccerosion
occurscausing
increase
of c.
Whenc is largerthanccsedimentation
occurscausingdecrease
of c. Thecoefficient
¾
represents
theinfluenceofthenon-uniform
distribution
ofthesediment
concentration
andthe
flow velocityoverthedepthon theadvection.
Aftersolvingthismeanconcentration
equation
the sediment
transport
ratescanbe
calculated.
Detailedinfomarionaboutthismodelis givenin Galappatti(1983) andWang
(1989).
Remark:Equation(23.1)applies
forthecase
of sand.
In caseofcohesive
sediment
thefight
handsideof theequationshouldbereplaced
bythesedimentation
anderosion
rate(Krone,
1962,Mehtaand Partheniades,
1982).
Extensiveanalyses,
theoretically
aswellasexperimentally,
onthevalidityof thismodel
wascarriedout(WangandRibberink, 1986,Wang,1989,1990).Fortunately,
theanalyses
showthatthisquasi-3Dmodel is sufficiently
accurate
toreplace
thefull3D modelin most
practical
problemsin estuarine
andcoastalregions.
TheBedLevelModel:The bedlevelmodelis themostsimpleoneof the threesub-models.
Thebedlevelchange
is described
by themass-balance
equation
for sediment.
(23.2)
Thisequation
isexactandhastobesolved
withthelargetimescale
soit should
be
integrated
overthe tidalperiod.
+ + Y -o (23.3)
In this equation
Az• = bedlevelchangein N tidalperiods;

T = tidal period;
NT = morphological time step;

472 Z B Wang
Tx,Ty=totalsediment
transport
perunitofwidthduring
a tidalperiod.
23.4 The dominatingconditionandmajorparameters

Figure1 shows implicitlythata typicaltideis usedtodescribe thedominating conditionand
thatthemajorparameters characterizingthiscomputation
procedure arethemorphologicaltime
stepandthenumberof continuity corrections aftera newtidalflow computation. The
dominating condition
aswellasthemajormodel parameters arediscussedin thefollowingfor
a practical
casestudyof themorphological processintheYantze EstuaryinChina(fig.2).The
resultspresented
arecomputed withESMOR.Theconclusions drawnfromthiscasecannot
directlybeappliedtoothercases butit mayprovide usefulinsightintothiskindof problems.
TheYantze
riverwitha length
of6380kmanda catchment
areaof 1.8million
km2isthe
largest
fiverin China.It flowsintotheEastChinaSeanorthof Shanghai. Its estuary
covers
anareaof about100x100km2.Unfortunately duringthecasestudyit wasimpossibleto take
thewholeestuary in themodel,becauseof therestricted
computer
capacityaswellaslackof
fielddata.However,themostinteresting
partof theestuary,i.e. thesouthbranch,wherealso
themostfieldmeasurements havetakenplace,is modelled.
Themodelcoversanareaof about
80x20km2 (seefig.2).With a spacestepof 1 km in bothhorizontal directions
themodel
containsabout1000 activecomputational
grid points.
The flow conditions

aswell asthesediment
transport
conditions
in theestuaryarevarying
constantly.
The variationsevencontaincertainstochasticcomponents.Neverthelesstwo
systematic
variation
patterns
canberecognized,viz. thespring-neap
tidalvariation
andthe
seasonal
variation.
Thesetwocomponentsof variationsaredominating
in manyotherestuaries
as well.
Boththespring-neap tidalvariationandtheseasonal variationareconsidered for defining

the dominating condition. In orderto examinethe spring-neap tidalvariation,a oneweek
measuringperiod(from28 Aug.to4 Sept.1984)is simulated. Duringthissurvey thevelocity
andsediment transport in 15 stationsdistributed
in 3 crosssectionsin themodelareawere
measured during3 days(thespringfide,averaged tideandtheneapfide).The timestepfor
theflow computation is 5 minutesandthebedlevelwasadjusted eachday.Thecalculated
totalsediment
transportfieldandthesedimentation-erosion pattern
duringspringfide,averaged
tideandneaptideareshownin fig.3 andrigA. The transport patternsas well asbedlevel
changepatterns are similarduringspringtide,averaged tideandneapfide.Therefore it is
concludedthatit isperfectlypossibleto carryoutmorphologicalcomputations witha properly
chosen standard tide.

473
ZB Wang
Figure2: The YantzeEstuary
The choiceof the standardtide shouldbe basedon the net sedimenttransportduringthe

tidalperiod.Thenettransportin everycrosssection
hastobeabouttheaveraged nettransport
in thespring-neap
tidecycle.In fig.5thecomputed totalebb-andflood-transport
throughthe
three measuredcrosssectionsduringthe surveyare depictedtogetherwith the available
measurements.
The figureshowsthatthe averagedtotalebb-as well as flood- transportare
represented by a tidebetween theaveraged tideandthespringtide.However,concerning
the
nettransport, i.e. thedifference
betweentheebbtransportandthefloodtransport,
theaveraged
valueisrepresented by theaveragedfide.Therefore
theaveraged tideischosen
asthestandard
tide in the longterm morphological computations.
The seasonal variationwouldhaveto be studiedin a similarway but theavailabledatadid

not allow this.Thereforethisvariationis takeninto accountsimplyby adjustingthe upstream
boundary
conditions
andthecoefficient
in thebedboundarycondition in sucha waythatthe
averaged
annualsediment
transport
throughtheestuaryis correctlyreproduced.
The morphological
time stepAttoandthe numberof continuitycorrections M, are model
parameters
only.Theydonothaveanyphysical meaning andtheyhavetobechosen suchthat
the computational
effort requiredis as smallas possiblewhile sufficientaccuracyof the
computations
isensured.
Forthispurpose a sensitivity
analysis
wascarriedout.In thisanalysis
a numberof runshavebeencarriedout with differentvaluesof Attoand M•. The computed

474 Z B Wang
bed level changesafter6 monthsfrom theserunsare comparedwith eachother.From the

comparison it is concluded
thatAt=canbe aslargeas40 timesthetidalperiod(i.e. about20
days),and that • can be as largeas 3.
TRNT?E
ESTUR•T
Figure
3: Netsediment
transport
during
spring
tide(top),averaged
tide(middle)
andneaptide(bottom)
It shouldbe pointedout that if the simpletransport

formulamentioned
in section3 is
appliedinsteadof thequasi-3D modelforthesediment transport,
a timestepmuchsmaller
than40Twould berequired because
oftheCFLstabilitycriterion.
Becauseofthecomplicated
character
oftheproblem thebedlevelequation
(23.3)canonlybesolvedexplicitly.
A Courant
number largerthan1 isonlypossible
if relaxation
effects
ofthesuspendedsedimenttransport
are takeninto account. This meansthat the moresophisticatedquasi-3Dmodelis not
necessarily
moreexpensive thanthe simplerequilibriumtransport
modelin a longterm
morphologicalsimulation.

475
Z B Wang
ii
I (unit = m)
;• • above2
':'" r'---1 i - 2
i -1 - 0
• I -2 - -1
: :• -2
Figure
4: Bedlevelchange
during
spring
tide(top),averaged
tide(middle)
andneaptide(bottom)
An interesting
resultwasthatwithM•=3thecomputational
effortusedfor the flow
computation
andthatfor the sediment
transport
computation
are aboutequal.Thisis
remarkable
asin thesediment
transport
modelonedifferential
equation
(i.e.theconcentration
equation)
hasto besolvedwhilein theflowcomputation 3 differential
equationshavetobe
solved.
Fora totallydifferent
case,(steady flowin a meandering
fiversection
(Waalin the
Netherlands)
withonlybedloadtransport),it alsoappeared
thatthemostefficient
computation
is obtainedwhen the effort for flow computation
and that for the sedimenttransport
computation
areabout equal. It seems
tobethatin a wellconstructed
morphological
model
thecomputational
effortforflowandforthesediment transport
shouldbeof thesameorder
of magnitude.
Afterthedetermination
,ofthemodelparameters
somesensitivity
analyses
onthephysical
parameters,
e.g.thefallvelocity
andthebedboundary
condition
etc.,werecarried
out.Next,
themodelwascalibratedanda 5 yearsimulation
wascarriedout.Detailedinformation
about
thissimulation
is givenin Wang(1989).Hereonlysomeresultsof the5 yearsimulationare
illustratedin fig.6 andfig.7.

476
Z B Wang
million tons million tons

10
Cross section S
10
I J..
-10
I 2 3 4 5 6 7 I 2 3 4 5 6 7
day In measurlno period day In measuring period
JI calc.
ebb
-•
transp
• calc.
flood
transp
meas ebb transp • meas flood transp
FIg. 5 Total ebb- and flood- transport through

the three cross sections
I 2 3 4 5 b 7
day In measuring period
Figure5: Totalebb-andfloodtransport
through
thethreecross
sections
(unit = m)
above 0
-4 - 0
-8 - -4
L_ -12
-16
-20
-
-
-8
-12
-16
-24 - -20
-28 - -24
-28
Figure
6: Theinitialbedlevel(top)andthecalculated
bedlevelafter5 year

Z B Wang 477
23.5 Numerical aspects

It is not the intentionof thispaperto give a detailednumericalanalysisfor the model.Only
two problemsarediscussed. The lust oneis aboutthediscretization
of the sourcetermin the
mean sedimentconcentrationequation(23.1) and the secondone is about the upstream
boundaryconditionfor this equation.
The sourceterm(fighthandside)in themeanconcentration equation(23.1) is a linearterm.

It is stronglyrecommended to discretise
thistermimplicitly,evenwhenan explicitscheme
is appliedto the restof the equation.This is possiblebecausethistermis a linearterm and
it only dependson thelocal sedimentconcentration. In thiscasethetotalschemewill remain
effectivelyexplicit.If an explicitdiscretization
is appliedto the sourceterm,the stabilityof
the computation canbe stronglyaffected.This is especiallyimportantwhenshallowregions
arepresentin the modelarea.In theshallowregionsthe adaptation time T• (~ h/%) is small
andthe factorAt/T• can be extremelylarge.For an explicitschemean absolutelimit for this
parameter is At/T•<2.Especiallywhenanexistingsolverfor theconvection-diffusion equation
is used,one tendsto build the modelsimplyby addingthe sourcetermto it with an explicit
discretization. From the point of view of robustness of the modelthis shouldcertainlybe
avoided.
In the caseof the YantzeEstuary,datawereavailableon the sedimenttransportneitherat

the upstreamnor at the downstreammodel boundary.This is a common rather than
exceptionalproblemof morphological modelling.Thereforesomeattentionis paid to this
problemhere.
During inflow at the upstreamboundarythe sedimentconcentrationor the sediment

transporthas to be prescri .bed.A wronglychosenupstreamboundaryconditiondisturbsthe
computationat the boundaryregion.The disturbancepropagatesinto the modelarea as the
computation proceeds.Its celerityis the smallpropagation
velocityof the bed forms,if the
equilibriumtransportmodelis applied.In morphological modelsfor rivers this problemis
sometimes solvedby applyingtheboundaryconditionfar awayfrom theregionof interestso
thatit is ensuredthatduringthesimulationthedisturbance hasno chanceto entertheregion
of interest.However,in the quasi-3Dmodel the disturbancewill penetrateas far as the
adaptation lengthof thesuspended sediment transport
(--uTa~uh/%--1000h)aftertheveryfirst
morphological time step.In the caseof the Yantze estuarymodelan erroneous boundary
condition will affectthecomputation in thewholemodelareawithina few morphological time
steps.Moreoverit is muchmoredifficultto extenda 2DH modelin the upstreamdirection
of an estuarythan in the caseof a one-dimensional model for a fiver. Thereforeanother
solutionto thisproblemis needed.
In theYantzeEstuarymodelthisproblemis solvedas follows.At the upstream boundary

the dynamicequilibriumconcentration is applied.The dynamicequilibriumconcentration
is
the solutionof the followingequation:

478 Z B Wang
•}C C•-C
(23.4)
This conditionis sufficientfor solvingthe sedimentconcentration

equationbut it is not
sufficientfor the whole morphological model.This becomesclear if the steadycaseis
considered.For the steadycaseequation(23.4) becomes
c - c, (23.5)
which meansin fact that the equilibriumtransportrate at the upstreamboundaryis applied.

In this casethe bexllevel at the upstreamboundaryhasalso to be prescribexlin orderto run
the whole morphologicalmodel.Fortunatelythe bexllevel changefrom 1984 to 1985 in the
areais available,so for the shorttermrunsthe measuredbed level changecan be appliedin
combinationwith equation(23.4). For runslongerthan 1 year the problemhasbeensolved
by recordingthe sedimenttransportrateat theupstreamboundaryduringa wholetidalperiod
in a shorttermsimulationandusingit asboundaryconditionin thelongtermsimulation.This
impliesthat the sedimenttransportthroughthe upstreamboundarywill not changemuch
duringthesimulation.This seemsto be reasonable asthe totalsedimenttransport throughthe
estuaryis mainly determinedby the fiver inflow in the caseof the Yantze estuary.
The solutionpresentedhereis certainlynotideal.Withoutanydoubtit will introducesome

error in the computations. However, this methodensuresthat the bed level changevaries
smoothlyin theboundaryregion.No significantdisturbance dueto theboundaryconditioncan
be observedin the computationalresults.With the availabledata it seemsto be the best
solution,if it is not the only solution.
23.6 Conclusions
Morphological development in estuarine

andcoastalregionsis an importantbutalsoa very
complicatedprocess.Mathematicalmodellingof this processis only possiblewith some
simplifications.In this paper the constructionand the application of mathematical
morphological modelsare considered.
The structureof the specificmodeldiscussed hereis basedon two importantassumptions,

viz. the low concentrationassumptionand the quasi-steadiness assumption.These two
assumptions makeit possibleto distinguish
threesub-models,i.e. theflow model,thesediment
transportmodel,andthebedlevelmodel.The assumptions alsomakeit possibleto applytwo
differenttimestepsin themodel,onefor theflow andthesediment transport
model,theother
for the bed level model.Furtherthe continuitycorrectionmethodhasbeenappliedin order

Z B Wang 479
to savecomputational
costs.
Many possibilitiesexistfor the flow modelandthe sedimenttransportmodel.Basedon the

validity(accuracy)andtheapplicability(cost)considerations
thequasi-3Dapproachhasbeen
chosen for both sub-models.
In orderto apply the model,a dominatingconditionin termsof a standardtide hasto be

defined.For determiningthis conditionthe spring-neaptidal cycle as well as the seasonal
variationare considered. For a practicalcaseit hasbeenshownthatit is possibleto definea
standard tidefor thelongtermmorphological simulation.
In thecaseof theYantzeestuaryit
has been shownthat the averagedtide can be usedas standardfide. However, it shouldbe
pointedoutthatthiscanvaryfromcaseto case.Thisstronglydepends on the upstream fiver
inflow, the tidal force from the seaas well as the formulationusedfor the bed boundary
condition.
The choiceof the morphological time step and the numberof times of continuity
corrections were investigated with a sensitivityanalysis.The conclusionis that a large
morphological time stepis allowedbecauseof usingthe quasi-3Dmodelfor the sediment
transport.
It is alsoconcluded thatthecontinuitycorrection
methodis a usefultool for saving
computationalcost.
Further, the numerical schematizationof the source term in the sediment concentration
equationand the upstreamboundaryconditionhavebeendiscussed. It is recommended to
discretisethe sourceterm implicitly.For the caseof lackinginformationat the upstream
boundarya solutionfor the boundaryconditionis presented.
23.7 Acknowledgement
The mainpartof the workdescribed in thispaperis carriedoutduringthepreparation of the
doctoralthesisat the Delft Universityof Technology.The writer wishesto expresshis
appreciation
of the adviceandencouragement givenby the supervisorsprof. dr. M. de Vries
andprof.dr.C.B. Vreugdenhil.Thanksaredueto Mr. J.C.Winterwerp for hisencouragement
andconstructivecomments duringthe preparationof thispaper.
23.8 References
Galappatti,R. (1983), A depth-integrated modelfor suspendedtransport,Reportno. 83-7, Delft

Universityof Technology,Dept. of Civil Eng.

480 Z B Wang
Galappatti,
R.andVreugdenhil,
C.B.(1985),
A depth-integrated
modelforsuspended
sediment
transport,
J. of Hydr. Res., Vol. 23, no. 4.
Hauguel,
A. (1978),Utilization
desmodeles
mathematiques
pourL' etude
dutransport
solide
sous
1'
actiondestourantsde Maree,ReportFA2/'78.41,
EDF, DirectiondesEtudeset Recherches.
Kalkwijk,
J.P.Th.
andBooij,R. (1986),Adaptation
of secondary
flowin nearlyhorizontal
flow,J. of
Hydr. Res., Vol. 24, no. 1.
Krone,
R.B.(1962),
Flume studies
of transport
of sediment
in esmarial
environments,
Finalreport,
University
ofCalifornia,
Hydraulic
Engineering
andSanitary
Engineering
Laboratory,
Berkeley.
Mehta,A.J.and
Partheniades,
E.(1982),
Resuspension
ofdeposited
cohesive
sediment
beds,Proceedings
oftheXVIII Coastal
Engineering
Conference,
ASCE,
Vol.2,Cape
Town,
SouthAfrica,
pp1569_1588.
Rijn,L.C.van(1987),Mathematical
modelling
of morphological
processes
in thecaseof suspended
sediment
transport,
Doctoral
Thesis,
DelftUniversity
of Technology
Vriend,
H.J.de(1981),
Steady
flowin shallow
channel
bends,
Doctoral
thesis,
DelftUniversity
of
Technology,Delft, The Netherlands.
Vriend,
H.J.de(1985),
Flowformulation
in mathematical
modelsfor2Dhmorphological
Changes,
Reportno.R1747-5,Delft Hydr. Lab.,Delft, The Netherlands.
Vries,
M.de(1959),
Transients
inbed-load
transport
(Basic
considerations),
Report
no.R3,DelftHydr.
Lab., Delft, The Netherlands.
Vries,
M. de(1965),
Considerations
about
non-steady
bedloadtransport
inopen
channels,
Publ.no.36,
Delft Hydr. Lab., Delft, The Netherlands.
Wang,Z.B.(1984),
Thevalidity ofadepth-integrated
modelforsuspended
sediment
transport
andthe
extension
ofthismodel
totidalrivers,
Report
no.10-84,
DelftUniversity
ofTechnology,
Dept.
ofCivil
Eng.
Wang,
Z.B.andRibberink,
J.S.(1986),
Thevalidity
ofadepth
integrated
model
forsuspended
sediment
transport,J. of Hydr. Res.,Vol. 25, no.1.
Wang,
Z.B.(1989),
Mathematical
modelling
ofmorphological
processes
in estuaries,
Doctoral
thesis,
Delft Universityof Technology.
Wang,
Z.B.(1990),
Theoretical
analysis
ondepth-integrated
modelling
ofsuspended
sediment
transport,
papersubmitted
to J. of Hydr. Res.

24
A three-dimensional transport model for

dissolvedand suspendedmatter in
estuaries and coastal seas
G C van Dam and R A Louwersheimer
Abstract
This paperdescribesa modelwhich simulatestransportof dissolvedor suspended matterin

threedimensionsby setsof discreteparticlesbeingdisplacedin threedimensionsat eachtime
step. The transportmodel is uncoupledfrom the models providing the required three-
dimensionaldeterministicvelocityfields. The deterministicfields are supplemented with
movementsthat accountfor vertical exchangeand for sub-gridstructuresin the fields of
horizontal velocities. These movements have a random character but in the horizontal
directionsthey are organized(scaled)as to accountfor the spectralstructureof the field in
nature.
At present,the 3D deterministic fieldsare generated by an analyticalextensionin vertical

directionof the velocityfield computedby a horizontallytwo-dimensional numericalflow
model.Distortionof verticalvelocitydistributionsby wind or densitygradientscan alsobe
takeninto account.In all casesthe analyticalfunctionsare giventhe simplestpossibleshape
while care is taken that the verticallyaveragedvelocitiesare not affected.The randomwalk
in verticaldirectioncorresponds with an exchangecoefficientI• that doesnot dependon
verticalposition.In the caseof sediments, the particlesare givena fall velocityand criteria
for deposition anderosionat thebottom.Thehorizontal displacementsby thedepthdependent
velocitiesare supplemented with scaledependent randomdisplacements thattake careof the
"sub-grid" velocitiesnotmodelledby thenumericalflow model.Computational resultsbased
upontheaboveprinciplesare comparedwith datafrom tracerexperimentsfor calibration.For
variouscases, computed concentration distributions
at smallandlargescalesarepresented and
discussed.

482 G C van Dam and R A Louwersheimer
24.1 Introduction
The transportand dispersionof matterin naturalbodiesof water is governedby physical

m•hanisms in threedimensions.The horizontalspreadingof materialcan only be properly
simulatedwhenprocesses in the verticalare explicitlytakeninto account.
In thepresentthree-dimensionaltransportmodel,thenecessary 3D deterministic
flow field
is derivedby extendinga computedverticallyintegratedvelocityfield into threedimensions
by meansof simpleanalyticalfunctions.The analyticalvelocityprofriesare steadyandapart
from additionaldistortions
by wind or densitygradientstheyare unidirectional,sotheydo not
accountfor phasevariationsand skewedsurfaceandbottomEkmanlayers(Hamrick1986).
Thisprovisionalsimplificationseemsto bejustifiedby certainresultssuchasthoseof section
23.3.1 and it alsomakespossibleto investigatethe sensitivityof the dispersion
processes to
variousparameters. Neverthelessit is intendedto evaluatethe improvements to be obtained
by quasi-3Dor "2•/D" approaches (Prandle1982, Van der Giessenen Jansen1988, Davies
1990)for obtainingtherequired3D velocityfields.Finally,theidealway to derivethesefields
would be by completethree-dimensional hydrodynamicnumericalmodelling which has
becomefeasiblein recentyears.Thesevariousalternativesfor obtainingbetter 3D velocity
structureswould leave the structureof the transportmodule unchanged:discreteparticle
approach,displacementin three directionsat each time step including supplementary
displacements which accountfor vertical exchangeand for spectrallystructuredhorizontal
"sub-grid"velocities.
The particleapproachandthe scaledor spectrallystructured sub-gridmodellingdescribed

below hasessentiallyno lower boundto spatialresolution.Resolvingfiner structures is just
a matter of particle numbersand sufficientlysmall timesteps(and thereforea matter of
economyas well). This sub-gridmodellingdiffersessentially from sub-gridparameterization
in the senseof accountingfor the effectof sub-gridprocesses
on phenomena at largerscales.
24.2 The model
24.2.1. The flow model and the use of the velocityfields
The two-dimensional hydrodynamicnumericalmodel usedfor the computationof the 2D

horizontalvelocityfield, in all applications
sofar, is a recentversionof the shallowwaterflow
simulationmodel WAQUA, which originatesfrom the well-knownconceptof Leendertse
(1967). The presentversionincludesthe improvements in the numericalschemes developed
by Stelling(1983). All applications with the presenttransportmodelhaveso far beenmade
with implementations of the flow modelon a rectangulargrid. The applicationswere made
with the North Sea model GENO (meshAx= Ay= 8000 m), the southernNorth Sea model
ZUNOWAQ (mesh= 3200 m) and an Ems estuarymodel with a meshof 300 m. Two-

G C van Dam and R A Louwersheimer 483
dimensional predecessorsof theparticletransportmodel(Van Dam 1985') havebeenused(or

are still in use) with WAQUA-models of WesternScheldt(mesh= 400 m) (Van Dam 1985',
1986b) andIJsselmeer
(mesh= 2000 m). The latteris a non-tidalfreshwaterlake.In the
presentarticlethe specialcharacteristics
of non-tidalapplications
will not be discussed.
In thepresentapplications, flow computation andtransport computation areuncoupled.The

tidalflow computations weremadewith cyclicboundaryconditions (containingseveralFourier
modes).After the flow in the computation hasbecomecyclic,all velocitiesand water levels
for the last cycle at a certainnumberof time levels(between10 and 30 minutesapart)are
stored.In the transportcomputation, this set of velocityfields is usedrepeatedly,as many
timesasrequiredfor thelengthof thesimulationperiod.Thismay amountto severalhundred
cycles.The time stepsof the transportcomputationare usuallysomewhatsmallerthan the
interval betweentwo successivetime levels of the storedvelocities.The velocities used, are
obtainedby linear interpolation
betweentime levels;furtherthereis a bilinearinterpolation
in spaceamonggridpoints.In someapplications thevelocityfieldswereFourieranalyzed(in
time) in orderto saveon computermemory.
In general,the amountof velocitydatafor one cycleis too largefor the workingstoreof

the computerso that only two successive levels are presentat a given time. Updatingthe
velocityfieldsfor successiveintervalstakesa considerable amountof the total processing
time.Dependinguponthetypeof computerit is of the sameorderasthenecessary computing
time (CPU time) or larger.
The separation of flow computation andtransportcomputation makesit possibleto perform

great numbersof transportcomputations (variouslocationsand typesof sources,various
durations,transportparametervaluesand verticalvelocitydistributions) on the basisof just
oneflow computation. In practicetheuncoupled approachcanonlybe usedin caseof strictly
periodicflow, shortnon-periodic flow historiesandstationary
fields.Non-periodictransport
computations for long periodshave to be performedin a coupledmodeof flow model and
transportmodel.
24.2.2. Supplementalhorizontalvelocities
Becauseof its finite mesh,the numericalflow modelcannotreproducedetailsof the 2D

horizontalvelocityfield belowa certainlengthscale.This lengthis usuallylargerthanthe
meshof thegrid (Van Dam, 1988,1989)butfor convenience thetermssub-gridscalesand
sub-gridvelocitieswill be usedin this connection.
The particletechniquemakesit possibleto resolvedetailsof concentration

distributions
on
muchsmallerscalesthanthe gridmesh.Thiscanonlybe donerealistically if thecomputed
flow field is supplementedin oneway or another.The supplementary mechanism mustbe of
a stochasticcharacteras long as the given flow model is not refined. If a deterministic
representation of smallerdetailsis for somereasonrequired,
theusualapproach is thenesting
of a finer grid flow modelwithin the givenone.

Two methodshavebeenappliedfor supplementing the horizontalvelocityfields:1. The

additionof anartificialeddyvelocityfield,composedof harmonicfunctions of space,coveting
thespectralrangemissingin the givenflow field,downto somelowerboundof therangeof
wavelengths,corresponding with therequiredresolution.
2. The additionof a scaledrandom
walk process,i.e. usinga randomstepdepending uponthe size of the particle(sub)cloud,
which is usuallyrelatedto its age.
Both methodshavea (gradual)cut-offat the upperend of the spectralgapthathasto be

filled.Froma physicalpointof view,themethodof theadditionalflow fieldis preferable,but
it is more time consumingand gives sometechnicalproblemsat boundaries.The second
methodis moreeconomicbut hasphysicaldrawbacks.It assumes thatparticleshavea memory
or at least"know"theaveragedistanceto theotherparticlesof theirgroup.Furtherit produces
rathersmoothdistributionswhile thosefrom the other methodhavea morerealisticaspect.
In certain casesthe unsatisfactoryphysical basis may cause unacceptabledeviationsin
concentration
values(VanDam 1988
b)butformostpractical
purposes
it is goodenough.
The
ideaof a "scaling"of a randomwalk wasfirstproposed by Maier-Reimer(1973);theprinciple
of a gradualcut-offat the end of the spectralgapwasaddedby Van Dam (1982, 1985a).
The numericalexperiments with the methodof the additionalflow fieldshaveplayedan

importantpart in gaining insight and in the interpretationof field data and resultsof
computations.
Somedetailswill be given in section24.3.
24.2.3. Extension in vertical direction
The 3D extension of the model exists of
- a decomposition of verticallyintegratedvelocitiesinto verticalvelocitydistributions

in
analyticform;
- the attributionof a vertical position(z) to each particle (z has been chosenpositive
upwards,zero at the bottom);
- the introductionof a verticalexchangemechanismby meansof an ordinaryrandomwalk
in verticaldirectioncorresponding with a depthindependent exchangecoefficientKz.
The particlesare of coursedisplacedin horizontaldirectionsby the horizontalvelocity

correspondingwith their verticalposition.
The analyticalexpressions
for the verticalvelocitydistributions
werechosenas simpleas
possiblei.e. Van Veen distributions(Van Veen 1936, 1938) for undistortedverticals
(neglectingphasedifferences):
(24.1)
(V} = givenvertically
averaged
velocity;
H = waterdepth;
ot= shape
parameter)
andthesum

of two expressions
of the sameform to accountfor possibledistortions
vd (z) by wind or
densitycurrents:
(24.2)
(•do= value
atthesurface;
[•and7shape
parameters).
Bothdistributions do not effectthe given2D horizontalvelocityfield. The verticalaverage
of (24.1) is equal to the given velocity v; the verticalaverageof (24.2) is zero. Both
distributionslie in verticalplanes,but (24.2) canmakeany anglewith (24.1) and a sumof
(24.1)
and(24.2)
isgenerally
notina plane.
Thedistortion
vd (z)istaken
neutral
(vertical
averageis zero);the effectof the causeof the distortionuponthe verticallyaveragedflow
field is takeninto accountin the numericalflow computation.
v (z) -v (a + 1) g
Figure 1: Construction
of "distortion"
vd(z) of verticalvelocitydistributionby wind or densitycurrent.
The simplepowerexpressions havebeenchosenfor reasonsof simplicity.The numerical

differencebetweena logarithmicdistributionand a powerdistribution(with properexponent

or)is extremely
small.Theeffectof thisdifference
upondispersion
patterns
wouldnotbe
noticeable.
Thevalueof otfor neutralbuoyancy is << 1; a common choicefor shallowwatersis ot=

1/6.It will tendto be evensmallerfor deeperwater(upperpartsof thevelocitydistributions
become moreuniform).To approximate verticalvelocitydistributions
in stratified
situations,
muchlargervalues
of otcanbeappropriate,
evenot= 1 orevenlarger.Of course,
in thiscase
theapproximation
remains alwayspoorbecause
of theusualjumpsthatcorrespond with
density
discontinuities.
Thiscanonlybeimprovedby usingexplicit3D flowcomputations.
Z0
0,9-
0.8-
0,7-
0.6-
0,5-
0,4-
0,3-
0.2-
0, 1-
0.01 0.1 1 10 100
Figure2: zo/H(seefig. 1) asa functionof •, for T = 1/9, 1/6 and1/3.
If the verticalmodellingis to be refined,theassumption of heightindependent

K• would
be the first one to be modifiedsinceit is in fact inconsistent
with almostany non-uniform

Ca C van Dam and R A Louwersheimer 487
verticalvelocitydistribution.
A possiblerefinementshouldbe preceded by sensitivitytestsas
to theeffectsupondispersion. The matteris probablyratherinsignificant
for dissolvedmatter
dispersionbutit couldbe of interestin thecaseof sediments.Anotherref'mement couldbe to
accountfor the time variationof I• in a tidal flow.
In (24.2), oneof the exponents,at leastin practice,hasto be > 1, say {3> 1, while the
otheris < 1. The termwith ]3> 1 canbe considered as theprimaryinfluenceof the wind,
downfrom the surface(fig. lb). A highervalueof ]3meansa smallerrelativepenetration
depth,so that• couldbe higherfor deeperwater.The termwith ), (normally),< 1) canbe
seenas the compensation currentthatneutralizesthe effectof the otherterm uponthe depth
averagedvelocity.Thereforeit seemsappropriate to take), roughlyequalto ix. The heightz
= zowhere
vol(z) = 0 liescloser
tothebottom
as• issmaller
(relative
penetration
depth
greater).Figrife2 giveszo/Has a functionof {3for fixed valuesof ¾.
24.2.4 Two dispersionmechanisms
The model containstwo (horizontal)dispersionmechanisms. When the resultsof the

simulations
are comparedwith field data,as will be shownbelow,it may be concludedthat
the main mechanisms in nature are indeed contained in the mechanisms of the model.
The two mechanisms are (i) the horizontaldispersionby verticalcurrentshearand the

accompanying verticalmixing and (ii) the spreadingcausedby the inhomogeneity and time
dependence of thefield of horizontalvelocities.In the secondmechanism horizontalsheafing
of velocitiesplaysa similarrole asthe verticalshearin the first mechanism,
but in the second
casethe effect occursat numerousdifferentlengthscalesand in all directions.
In the model,the (vertical)sheardispersion (i) is dealtwith by the verticalextensionof

velocitydistributionand accompanying verticalmixing. The spreading(ii) causedby the
structureof the field of horizontalvelocitiescan be modelledby an additionalfield or a
(scaled)randomwalk in the "sub-grid" rangeof the flow model(section24.2.2); the larger
scales are takencare of by the structureof the computedflow field itself.
Thetwoscalerangesdonotactentirelyindependent. Thelargescalestructure
of a velocity
field alonedoesnot inducethe samepatternsas its combinationwith smallerscalestructures.
Smallerscalevelocitystructures havea smoothing effectupondispersionpatternsas would
be causedby the large scalestructurealone.
Much insightin the effectsof the horizontalstructures

hasbeenobtainedfrom numerical
experimentswith syntheticvelocityfieldsin two dimensions(Van Dam 1987, 1990a,c).
It was
foundthatthespatialstructureis in factdeterminingthespreading.Growthrateof patchesis
virtuallyindependent of the life time of the eddiesof varioussizes.Only if all spectral
components arecompletely stationary,
specialeffectssuchastrappingof particlesmayoccur.
A minordependence on timeis sufficientto avoidthis.This meansthatfrom theviewpoint
of growthrate or dilutionof patchesof dissolvedmatterthe effectof persistentstructures

relatedto topography is the sameas the effectof stochasticcontributions

of the samekinetic
energyat the samelength scales.Therefore,from the viewpointof growthrate of single
patchesthe 2D horizontalvelocitystructureof an entireregionsuchas the North Seacanbe
fully describedby a spectrumof kineticenergyof spatialvelocitydifferences.Fromnumerical
experimentscombiningsuch spectrawith the effects of (vertical) shear dispersionand
comparingthesewith field data,it wasconcludedthatfor theNorth Seathe presentempirical
data agreefairly well (figs. 3 and 4) with a kineticenergyspectrumof horizontalvelocities
of the simpleform
E• = cnk-•' (24.3)
with F_•= energyper unit of massand unit of wave numberk; ckand m positiveconstants.
On thebasisof comparisonbetweennumericalsimulationandexperimentsthe valuesusedfor
theNorthSeaat present
areck= 2.87x 10'4m2 S'2,m = 1 (VanDam 1987).A spectrum
of
the simpleform (24.3) in two dimensionsand in absenceof othereffects(in particular3D
mechanisms) corresponds with a behaviourof the maximumconcentrationof a singlepatch
(instantaneous
point source)
(•) = cot-• with 0 = 4/O-m) (24.4)
Thesuperscript (1) refersto thereleaseof oneunit;the2D characterimpliesa dimension[L]'2

(unitsper unitof area);Coand[3arepositiveconstants. For m = 1, 13= 2 thevalueCk= 2.87
x 10'4m2 S'2(in combination withm = 1, 13= 2) correspondswithCo= 7.09x 103m'2S2.
If we choosea scaledrandomwalk techniquefor simulationof theeffectsof thehorizontal

velocitystructure,the (average)size of the randomstepin horizontaldirectionfor the case
Cmax-r2willgrowas tø'5.
In general,
anincreasing
random step
- tø•o
corresponds
toCmax-
t-(2o•o+
1)(VanDam1985b).It should
berememberedthatthecorrespondence
between
(24.4)
and (24.3) and the relationship
I• = 4/(3-m) only hold as long as the kineticenergyof
horizontalvelocitiesaccordingto (24.3) is the only mechanismof dispersion.Somewhat
similar,
a scaled
random
walkwitha random
step- tøtowillonlycause
a Cmax-t-13
(with
[3= 2Oto+ 1) if therandomwalkis theonlymechanism of horizontaldispersion
in thesimu-
lation.Of course,thephysicalcauseof thescaleeffectis notthetimebuttheincreasing size
of thepatch,allowinglargereddiesto affectit. Therefore,if othermechanismsarepresentin
thesimulation, suchas(vertical)sheardispersion,
a correct
scaled randomwalk,corresponding
with the assumed spectrumof horizontalvelocities,hasto be establishedby relatingthe
randomstepsizeto the sizeof thepatch,in sucha way thatthestepsizeusedwith a certain
patchsizehasto bethesameasit wouldbefor thesamepatchsizeif thehorizontal velocity
structurewere the only mechanism.
If thereis a structuredgiven(computed)velocityfieldpresent,
therandomstephasto be
levelledoff (gradually)at thescalewherethegivenvelocityfieldbeginstoprovidethecorrect
structures.

24.3 Model studies for further investigation of dispersion

mechanisms
24.3.1 Relative importanceof the two mechanismsof horizontaldispersion
For investigatingthe relativeimportanceand the combinedeffectsof the two mechanisms, it

is sufficientand convenientto use a simplifiedmodel with a uniform velocity field and a
uniformwaterdepth.For a ratherwide rangeof scales,up to the resolutionof the computed
velocityfield concerned,the resultswith the uniformfield and with the computedfield (in a
region of uniformdepth) will be identical.For the larger scales,the non-uniformityof the
computedfield is takeninto accountin the simplermodelby continuingthe scaledrandom
walk beyondthe scale at which otherwisethe computedfield would provide the proper
velocitystructuresand their dispersiveeffect.
On the basisof theseconsiderations,computations have been made in a model with a

uniformvelocityfield of a simpleharmonictimedependence (tidal ellipseshownin the inset
of figures3a and b) and a period of 44700 s. The uniform depth H implies that vertical
exchangecoefficientI• and verticalmixing coefficientTz, relatedby
or T - (24.5)
K'' 2r= = 2Kz
becomeequivalentparameters. By workingwith Tz, thevalueof H canremainundefined.The

factor1/2 in (24.5) correspondsto a definitionof T• asthetimeneededto reacha • = H after
an instantaneous point releasein absenceof boundaries(water surfaceand bottom).
Sincein the computationin questionboth mechanisms act simultaneously,

it is necessary
to relate the randomstepsize to patchsize and not to time as explainedin the preceding
section.The shapeof the patchesis not isotropic;patchsize is definedby
r = }/2o•on (24.6)
if • and q are the directions
of the principalaxesof the patch.The useof r insteadof t
impliesthat• and{Jnhaveto be computed fromtheparticledistribution
quitefrequently.
In figure 3, the resultof a numberof simulationsis presentedin the form of a 2D

maximumconcentration asa functionof timeafteran instantaneous pointreleaseof oneunit
at thesurface,for variousvaluesof T• anda fixedkineticenergydislributionfor thehorizontal
velocityvariations.The chosenenergydistribution is the spectrumassumed for theNorth Sea
as described in the precedingsection.This choiceis particularlyinterestingbecauseof the

490 van Dam and R A Louwersheimer
largenumber of available fielddalafortheNorthSea(Joseph etal 1964;Barrett et al 1969;

TalbotandTalbot1974;VanDam1982;Suijlenet al 1990"•; Weidemann 1973)(fig.4). The
assumed tidalellipsesarepicturedin theinset.Of course thereis a physicalrelation
between
tidalenergy andspectral energy,thefidesbeingthemainsource of kineticenergy.Therefore
it is notmeaningful to takejustanyamplitude for thetidalvelocitiesin thesesimulations.
10-2
io--4 io-4
TZ = 0
/
Z = 3600
= 7200
io-6 lO-6
Z = 14400
io-8 i0-8
Tz = 28800j
T = 8640
Z
iø-iø & io-iOi b

io i hours io-i io i h0u
$
Figure3: Maximumverticallyintegrated concentrationr,(_l_)

(2D) after
•'•man an instantaneous
point release
of one unit of dissolvedmatterat the surface,computed with uniformdepthanda simpletidalellipse
(inset).Kineticenergyof tidein caseb is half of thatin a.
It is interesting
to compare figure3 withthecompiled datafromtracerexperiments in the
NorthSea(fig.4). Thecomparison supportsthechosen approximationof theenergyspectrum
andit strongly suggests thatthenaturalvariations
of T,_(thehighervaluesarerealisticin stra-
titled areas)are the main causeof the spreadamongthe variousexperiments, ratherthan
regionaldifferences in kineticenergycontentor its spectralstructure.
The correspondence
betweenthe bandof observations andthe bandof simulations is quitesatisfactory.
Note that
the bandof simulationshasa lowerbounddue to the fact that from a certainvalueof T,_the
increase
of dispersion
rateswith increasing
T,_comesto an endandevenchanges
into a
decreaseat still largerT,_values.

10-6
zo-8
10-10
].o hours
Figure4: Compiled
NorthSeadataof instantaneous
pointreleases
of tracer(seerefs.).
Thelattereffecthasbeenstudied separately
in a vertically
two-dimensional (2DV) model
(essentiallythesamemodelwaspresented in 1982by Allen).A resultof thesestudies is
presented in figure5. The dispersion coefficientsplottedin the verticaldirectionare the
constantvalues (nofide,netcurrent only)or theeffectivetidalaverages,bothreached after
a certaininitialperiodrelatedto theverticalmixingtime.

Deff
300 m2s'1
200 6
81% osc.
•n.
100
3O 98
•0
Tz: T/6 T 6T
Figure5. Effective(longitudinal)dispersioncoefficient(netvaluesper period)as a functionof vertical

mixing time for variousmixturesof tidal and net current,computedin a 2DV particlemodel.For all
curvesthetotaltidallyaveraged
kineticenergyis thesame('• = « m2 s-2).
Figure 6 assistsin explainingthe role of T,_in (vertical)sheardispersion.It depictsthe

behaviourof a verticalline sourceduringa tidal period,with verticalvelocitydistributionand
verticalexchangeas only mechanisms, for a T,_<< tidal periodT, a T,_-- T anda T,_>> T.
Figure3 showsthatwith therelativelylow energylevel of the horizontalvelocityvariations

foundfor the North Sea, thereis a fairly long periodin which the mechanismof (vertical)
sheardispersion playsan importantor evendominantrole, the moresowhenT,_is larger(in
therangeT,_ < T); the field datashowthatall T,_valuesof thisrangedo occur(andpossibly
beyond;this is certainlyso in stronglystratifiedregions).After the periodthat the relative
importance of themechanism itselfis significant,thereis a periodin whichtheobtained"lead"
in dispersionstill hasan impacton concentrations ("memory"effecO.
In a well mixedestuary,the situationis different.Verticalmixingtimesare short,so that

the (vertical)sheardispersionbecomeslessimportant.On theotherhand,becauseof themuch
morepronouncedtopographyand the usuallyhighertidal energies,the energylevel of the
velocityvariationson lengthscalesup to the tidal excursionlengthis considerably
higher.

!!
1/3
1/2
2/3
516
....•,':'
.. • ..•...• . :)
ß:%•2.',½.: x T
Tz = T Tz >> T.
Figure6. Illustration
of (vertical)sheardispersion
of aninstantaneous
verticallinesource
for 100%tidal
current(comparefig. 5), showinghowthelongitudinal dispersion
decreases
comparedwith thecaseTz
= T, whenTz becomes(much)smalleror (much)larger.
24.3.2 Chaotic advection
If a fine meshnumericalflow modelof an estuaryis available,from a certainscaleonwards

the dispersiveeffect of the spatialvelocity structurecan be demonstrated by using the
computed velocityfield withoutanysupplement.An exampleis givenin figure7, basedupon
a 300 m meshrectangulargrid (WAQUA-) modelof theEms estuary.It showsthe evolution

during
onetidalperiodof a contour
which
isinitially
a circleof 1500m radius,using the
computedflowvelocities
only.Onthisscale,
therelative partof thehorizontal velocity
structure
in thedispersion
is dominant
andthemainpartof itsenergy is reproduced in the
flowcomputation.
Similar
results
havebeenpublished
byRidderinkhof (1990).In figure 7c
thedistribution
of a dense
setof particles,
fillingthecircle,is givenat theendof thesame
tidalperiod,
including
theeffects
of "sub-grid"
scale
velocity
variations
andvertical
shear.
0 • 20000
s
_ 5000
s 2•0•00•
•50•
0•a
45000
-period)
• • • 15000s
b
4500
C (eM2
peri
d) _
Figure
7. Deformation
ofa circular
contourduring
onetidalperiodinEmsEstuarymodel (mesh= 300
m)bycomputed velocity
fieldonly.Difference
in initialphase
betweena andb is onehalfof tidal
period
(22500s).Extradispersion
by"sub-grid"effects
(vertical
shear
withvertical
mixing;"sub-grid"
horizontal
velocity
variations)
forcasea shown
in inset(c) (situation
atendof period
only).
Theadvectionbythecomplex timedependentvelocity fieldistypically

a chaoticprocess.
Thepattern afteronetidalperiodstrongly
depends uponthetidalphase at thebeginning
(figures
7aand7b).In otherwords, thedisplacement of a particle
duringa tidalperiod
and
therefore
itstidallyaveraged
Lagrangianvelocityis verymuchdependent upontidalphase
(VanDam,1984,1986•; Ridderinkhof1990).Thislimitsthephysical meaning andpractical
utilityof "Lagrangian
residual
velocities".

a
b
5oo j
t = 4000
Figure
8: Effects
similar
tofigure
7,obtained
bysimulation
with2Dsynthetic
eddyfields
(8pairs
of
harmonic
components)
in anunbounded
region.
a. Stationaryfield
b. Eddy life time 1.5 revolutions.
In case
b,onlythecontour
ofthefluidblobhasbeendrawn;
graphs
(candd) show
evolution
ofoverall
sizeandcontourlengthin time.
Similar
patterns
wereobtained
intheseparateinvestigations
of synthetic
velocity
fields.
Figures
8aand8bwereobtained
with8 "eddycomponents"(8pairsofharmonic
functions),
figure
8ainastationary
mode,fig.8bwithfinite
lifetimes
(1.5revolutions
pereddy).
Units

areessentially arbitrary,butfig. 8b canbe interpreted

asa patchof some250 m diameterand
anageof 4000s. If theexponential behaviourof thecontourlengthis extrapolated,
whichis
justifiedfor reasonsof similarity,onearrivesat a lengthof some1000000 000 000 m when
the diameterhasbecome 1 km (Van Dam 1989•).
Thenumerical
experiments
withsynthetic
fieldsarea quickmeans
to interpret
dispersion
patternsin termsof energyspectra(section24.2.4).
24.4 Practicalapplications
24.4.1 Dissolved matter
Applicationsmay vary from time scalesof a few hours(or less;thereis no lowerboundto

spatial
resolution)
totimescales of theorderof a year(orlonger).
In practice,
atthismoment,
simulation
timesof theorderof a yearareonlypossible withmerelyperiodicvelocity
fields.
ß .
..
Tz - 3600 s Tz - 7200 s Tz - 14400 s
Figure9: Bird's-eye
viewof particlesetsin a tidalcurrent
(Northgoingphase)
withmoderate
wind
fromtheWest,computed for threedifferent
verticalmixingtimesTz.
Pointreleaseat thesurfaceat t = 0; distributions
drawnafter40, 80, 120,200 and360 minutes.
Patchwithoutwinddrawnfor t = 360 minutesonly.

The simulation
of periodsup to a few dayshasbeenoperational with pr•eding 2D
versions
of themodelfor accidental
spillsat sea(DelftHydraulics
Laboratory
1989)or in the
Western
Scheldt
(VanDam1985%
1986b;
Lefevre
andVanderMale1988).Forthesame
type
of applications
a 3D versionof the modelis now beingimplemented
for the North Sea
(Diependaal
1990;Hendriks
1990
a'b)
andwillbeadopted
forotherregions
in thenearfuture.
An illustration
of an applicationon a shorttimescaleis givenin figure9, presentinga
bird's-eyeviewonly(particles at all depthlevels).Thechosennumberof particles(300) is
smallerthanin cases whereiso-concentration finesare required.The picturegivesan
impression of windinfluence(moderate crosswind)anditsdependence uponT•.
I I
•ookm
Figure 10a Figure lob
Figure10. Simulation
with 3D modeof transport
modelof a continuous release(nearthemouthof
riversRhineand Meuse)in the North Sea. Periodicflow; (periodic)steadystateconcentration
distribution.
a. Computed
withnon-uniform
account
of density
current
perpendicular
to thecoastandsuppressed
verticalmixing,decreasing
with distancefrom river mouth.
b. No densitycurrentandmixinguniformover the wholearea.
White= land(figures
10,11,12).Shades
in figures
10through
15represent
concentration
ranges
of a
factor 4.
Figures10,11and12givesometypicallongtermapplications, whichwereonlyfeasible
bytheuseofa strictlyperiodic velocity
field.Of course,theperiodicity
oversucha longtime
(orderof oneyear)limitstherealisticmeaning of theresult.Theperiodicitydoesnotonly
enable theseparation
of the(relatively
short) flowcomputation andthetransportcomputation
butit further
enablestheuseof superposition (VanDam1985a),whichsaves a greatamount
of computer time.Theresults havebeenobtained by following only300particles fromthe
source, amounting
toa totalof some100000particles in thefinaldistribution
bysuperposition

of almost400 successivestates(for largertimes,partsof the groupsof 300 particleshaveleft

the modelareaat the openboundary),eachrepresenting a periodof 9(K}• s (= thecycletime
of the flow computation).
The figurerepresents somegeneralresultsobtainedby simulations
overa periodof 405 days(in orderto obtaina (periodic)steadystateconcentrationdistribution
in theentireregion).Shadesrepresent verticallyaveraged(3D) concentrations.The underlying
cyclic 2D velocity field was obtainedby one particularrun of the WAQUA-model
"ZUNOWAQ" (mesh 3200 m). One shouldrealize that different resultsare obtainedwhen
otherboundaryconditionsare applied.
.............................................................
............................................ 2..•_;,.-.•:.•
•"•' :::::::::::::::::::::::::::::
::::::::::::::::::::::::
,•;•.z•.•:••5=• .•:=; ß,
================================ .• ....:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
...... ==j , .:'.,.;........:, ,:.:::;:,::, .,,..,•.. -• •, , '.."

•=• ..'.:::,:.,:':• :: •,•,•.•:' < .. -
5::5::5•55••5•[•, ::•'::
:::7'
• ••.• " ""
......
=:==.==.= :,2•.-
7•. ._'•-
Figure 11a Figure llb
Figure 11: a. Simulationwith 2D modeof transportmodelof a continuousreleasefrom samesource

as figure 10. No extradispersionaddedto parameterizeabsenceof dispersionby verticalshear.
b. Same as a, but tidally averagedvelocities(Eulerianresidualtransportvelocities)usedinsteadof
completetime dependentvelocity field.
Figures1la and 12awerecomputed in a 2D modeof theparticlemodel.In figure10bthe

verticalsheareffectsare takeninto accountin a uniformway (corresponding to just one of
the lower curvesof figure 3a). Figure 10a accountsalso for the local stratificationnear the
mainfiver mouth,vanishingat largerdistances, andfor the densitycurrentsperpendicular to
the coast, decreasingat largerdistances from the mainsourceof freshwater(compareDe
Ruijteret al 1990andVisser1990).Onlytheapproach leadingto figure10aproduces realistic
fiver water concentrations near the fiver mouth if the source is identified with the fresh water
outflowof the local fiver mouth.At largedistancesfrom the sourcethe differencesbetween
the resultsof figures10a, b and 1la becomeinsignificant.
Figure l lb and 12b as comparedto 11a and 12a are to illustratethe effects due to
differencesin advection(Van Dam 1984,1986•, 19889thatariseif Eulerianaverages of tidal
velocitiesare used in order to save computertime. It is seen that the effects are more

pronouncedin one area than in another.However, thereis little reasonto use the fasterbut
lessaccurateapproach(figures1lb and 12b)nowthecompletetimedependent simulationhas
become feasible.
'--.•.E•-.- ,•..--:•-'
7•2=g•.• •i-':•.•. •.,•;-nñ .............................................
.
;.•?.:..•.:/•=g•:=----====:_z.:_z.=E-!-!'-"!-•..•--=_=
•'....,,--l•- r-"":-7.:-
::.•.--.7.:•.:.7.:•:--
•
...................................................
...............................................
Figure 12a Figure !2b
Figure 12: As figure 11. exceptfor locationof source.
It shouldbe pointedout that the (very)longtermsimulations haveto be regarded as

preliminary. Somerefinements of themodelforthistypeof applications arepossibleandmay
be desirable. Oneaspectis theuniformity of variousparameters in thepresent programme,
especially K, (oralternatively
T,). Forevents in restricted
areasthisis usuallynota problem.
Ontheotherhandit is noteasyto indicate howK, will varyoverthewholearea.Theglobal
dependency onlocationin figure10a,hasbeenacquired by a "trick";thecomputation of the
fateof thebasicgroupof 300 particleswasinterrupted at certainstagesandrestatted with
othervaluesof someparameters (T,; shapeparameter of VanVeenvertical;densitycurrent
intensity)dependent onthechanging positionof theparticlecloud.In recentlongtermsimula-
tions,particularly thosewith sediments (nextsections) a uniformK, waspreferred to a
uniform T,_in ordertoavoidunrealisticslowmixingin shallow areas. In thenearfutureit may
become possible to makeK, heightdependent in a meaningful way.Oneof theagents forK,
is theenergy of thecurrent.Thisalsomeans thatin factK, variesduringa tidalperiod.Also
thiseffectmaymainlybe of interestfor sediment applications.
24.4.2 Extensionto suspended

matter
Theextensionof themodelintoa version

forsimulating
thetransport
of (settling)
suspended
materials
is quiterecentandtheprovisions
madeforthispurposewill mostlikelyneedsome
refinements
in orderto obtainmeaningful
results,
hopefullyof somepredictive value.

A fall velocitywas introducedas an input variable.Fractionswith differentfall velocities

haveto be dealtwith separately.For eachfractiontwo criticalvelocitiescanbe chosen(which
can be relatedto shearstresses
at the bottom)on a heightthat canbe specified.The smaller
velocitycorresponds with sedimentation,thelargerwith erosion.No timelagsareusedsofar.
Thismeansthata particleis immediately
reactivated
whentheerosionvelocity(ona specified
heigh0 is exceeded.
An importantimprovementcomparedto modelsin two dimensions,is the fact that a

particlewill only settleon the bottom,when it hits the bottomand at the sametime the
velocity(on theprescribed height)is belowthecriticalvalue.Reversely,whentheparticleis
re-activatedat a later time, its initial positionis very closeto the bottom.
The aboveprovisionsincludethe casethat all materialremainsin suspension, evenwith

highfall velocities;
oneonlyneedsto specifya threshold velocityequalto zero.In a uniform
regionof depthH, verticalexchangecoefficientK, andM kg suspended materialin a water
column(unit area), the verticaldistributionwill thenbecome(Van Dam 1982):
_ Iw•l
z (24.7)
c (z)- c
where
_ttl,•el• (24.8)
(l%l= absolute
value
of fallvelocity;
withz positive
upwards,
we< 0) or,with(24.5)
formixingtimeTzandTv= H/l% I forfallingtimeandz* = z• forrelative
vertical
position:
(24.9)
c (z,) = c (o) e
ForTz<< Tv, theverticaldistribution

is practically
uniform(veryfinesediments).
For T• >>
T•, mostof thematerialis concentrated
in a thinlayernearthebottom.Maybethelattercase
canbe usedas a furstapproximationof a "fluidmud"layer.
Figures13, 14 and 15 give someexamplesof typicalresultsthatcanbe obtainedwith the

present provisions.Theypertainto specificfractions
witha fixedfall velocity.Thebehaviour
of mixturescanbe studiedby separate computations for differentfractions.At thismoment
it is notyetpossibleto evaluatetheresults.
It mayberatherdifficultto findproperfielddata
for verificationwhenthe simulations acquiresufficientrealisticvalue.

iii!i!ii!!i:i::::::::::!i::!::!::::::::ii:ii!:i:iiiii!!!!iiii
!!!!i i i i•.iiii:::•::i::!:.:•:.•:•....`.•.:•;i:•i:..!i•:•.:•i:.::•i!ii•ii•.:i!!..?.`•::
•i!;•;:'.':,'•i:..'.:::!i•iiiii::":::
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
.;.;::.,,..:,•.:.•..•
.',.
•••'-'•,•.
;.:.;•,:
;:...,.::•:::::
•.:(...!.
:.:...•:.:'::..'.:
................
!i::i•::::::i::!:iii::::i!iii::iii::ii::!::!!::::iiiii:::i:!•i•ii::i:::.:.•i
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
::::::::::::::::::::::::::::::::::::
v..•
::::::::::::::::::::::::::::::::::::::::::::::::::::::
.--.-'.'
::::::::::::::::::::::::::::::
:,_:..,-.-.
::::::::::::::::::::::::::::::::
•...
......................................
........................................
..............................
..............................
::::::::::::::::::::::::::::....
........................
::::::::::::::::::::::::::::::::::
!•iiiii•;•iiiii
..............
ii::.... IOO m ........................................
i!•i::::::::iii::"i
........ iii:.":
........
Figure 13: Computeddistributionalongthe Netherlands'coastof suspended part of a certainsediment

fractionfrom a continuoussource(samelocationasfig. 10 & 11), for two differenttidal phases(•AM2-
periodapart).
Black = land (figures13, 14, 15)
..............................................................
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
::::::::::::::::::::::::::::::::::::
:::::::::::::::::::::::::::::::::::::::::::::::::::
:::::::::::::::::::::::::::::::::::::::::::::::::
::::::::::::::::::::::::::::::::::::::::
=================================
....................................... ::::::::::::::::::::::::::::::::::::::::
::::::::::::::::::::::::::::::::
............................
:::::::::::::::::::::::: ....
........................................ ....................
::::::::::::::.. ..::::::..::
Figure 14: Settledpart of the sedimentof figure 13 at sametidal phases.Sedimentation

zonemoves
alongthe coastwith the tidal waveandcorresponds with zonesof loweredconcentration
in figure 13.
The nonchangingspotsrepresentpermanentlysettledmaterial(undertherigidly constantconditionsof
the simulation).

..................................................................................................
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
..::::::::::::::"
..................................................................
..................................................................
..................................................................
..................................................................
..................................................................
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
::::::::::::::::::::::::::::::::::::::::::::::::::::
:::::::::::::::::::::::::::::::::::::::::
•:' ;.•:.::.::•
:::::::::::::::::::::::::::::::::::::::: ::::::::::::::::::::::::::::::::
:::::::::::::::::::::::::::::::::::::::: ............................
:::::::::::::::::::::::: ....
::::::::::::::::::::::::::::
....
........................
........................................
......................
Figure15: Simulationwith strongdensitycurrent(or similarcirculationin verticalplane)perpendicular

to the coast.Sourceof sedimentas in precedingfigures.
a. Fractionwith samecharacteristicsas precedingfigures.
b. Criticalvelocitiesfor sedimentation
anderosionreducedby a factor% (fall velocityunchanged,
0.0003m s'l).Notegreatimpactof thisrelativelysmallalteration.
Anotherdifficulty will be the behaviourof the flow modelnearthe closedboundaries.In

figures14 and 15 it is visiblethatthe presentflow field of ZUNOWAQ createsnarrowzones
of low velocitiesalong the Netherlands'coast, which act as seAimentation zones in the
simulation.This effect will be very sensitiveto the rules appliedin the flow modelat the
closedboundaries: "slip","no-slip"or someintermediate prescription.
In the list of possiblerefinements,

the introduction
of a scourlag is probablythe most
urgent.Depthandflow dependence of K, maycomenext.Flocculation anddisintegrationare
importantbut for thenearfutureseemto be beyondthereachof meaningful modelling.
24.5 Conclusions
The described3D transportmodeltakesexplicitlyinto accountthe mechanism of horizontal

dispersion due to verticalvelocityshearand the mechanism of horizontaldispersion by
horizontaleddiesandit reproduces very well theshapeandthewidthof thebundleof curves
obtainedfrom the numerousdispersion experiments with dissolvable
tracersin theNorth Sea

(figures3 and4). In the simulation,the upperboundof the setof curvescanbe adjustedby

calibrationof the energyspectrumassociated with the secondof the two mechanisms.
From thisresultit canbe concludedthatall substantialdispersionmechanismsare modelled

and this thusforms the basisfor a wider application(non-dissolvable matter,dispersionin
estuariesetc.).The rathersimpleway in whichthehydrodynamically computed2D (vertically
averaged)velocity field is extendedinto three dimensionsis apparentlygood enoughfor
severalapplications. Nevertheless
it is intendedto evaluatethe improvementsto be obtained
by a moreexplicitcalculationof the 3D velocityfield, eitherby a "2•6D" approachor by a
fully three-dimensionalhydrodynamic computation of the flow.
24.6 References
Allen, C.M., 1982. Numericalsimulationof contaminant

dispersionin estuaryflows. Proc. Roy. Soc.
Lond. A 381, 179-194.
Barrett,N.J., D. Munro and A.R. Agg, 1969. Radiotracerdispersionstudiesin the vicinity of a sea
outfall.Proc.4th Int. Conf. on Water PollutionResearch,PergamonPress,New York, 863-882.
Davies, A.M., 1990. On extractingtidal currentprofilesfrom vertically integratedtwo-dimensional

hydrodynamical models.Jour.of Geophysical Res.95, No.C10, 18317-18342.
Delft HydraulicsLaboratory,1989.Gebmikershandleiding
MARS (MarineSpills).Rept.T 470, 47p.
Diependaal,R.J., 1990. MARS 2.0, Processen

en Transport.Delft HydraulicsLaboratory,T 607.
Hamrick,J.M., 1986. Long-termdispersion

in unsteadyskewedfree surfaceflow. Est., Coast.andShelf
Sc. 23, 807-845.
Hendriks,A., 1990' MARS 2.0, Gebmikershandleiding.

Delft HydraulicsLaboratory,
T 607.
Hendriks,
A., 1990
b.MARS2.0,Deconstructie
vaneenhydrodynamische
database.
DelftHydraul.
Lab.,
T 607.
Joseph,J., H. SendnetandH. Weidemann,1964.Untersuchungen

iiberdie horizontaleDiffusionin der
Nordsee.Dtsch.Hydrogr.Z., 2, 57-75.
Leendertse,
J.J.,1967.Aspectsof a computational modelfor long-period
waterwave
propagation. The
RandCorporation,SantaMonica,Calif., rpt. RM-5294-PR/Ph.D. Thesis,TechnicalUniversityDelft
1967, 165p.

Lef•vre, F.O.B. and C. van der Male, 1988. Gebmikersdokumentatie programmapakket CALM,
Rijkswaterstaat
(Neth.),Tidal WatersDivision,rpt. GWWS-88.407,62p.
Maier-Reimer, E., 1973. Hydrodynamisch-numerischeUntersuchungen zu horizontalen

Ausbreimngsvorg•ingen
in derNordsee.Dissertation
Univ.Hamburg/Mitteilungen
Inst.filr Meereskunde
der Univ. Hamburg,XXXI, 1-56.
Prandle,D. 1982. The vertical structureof tidal currentsand otheroscillatoryflows. Cont. Shelf Res.
1, No. 2, 191-207.
Ridderinkhof,H., 1990. Residualcurrentsand mixing in the WaddenSea. Ph.D. Thesis,University

Utrecht,Neth., 91p.
De Ruijter,W.P.M., and A. van der Giessen,1990. Currentand densitystructurein the Netherlands'

coastalzone.Proc.5th Int. Bienn.Conf.Physicsof Estuariesand CoastalSeas,Gregynog,Wales.
Suijlen,J.M., J.S. Sydow,C. Heins andP.C. Beukenkamp,1990'. Measurement of turbulentdiffusion

andresidualdisplacements by dye experiments in thesouthern
NorthSeain 1971, 1973, 1978and1979.
Data Report.Rijkswaterstaat (Neth.), Tidal WatersDiv. GWAO-90.022.
Suijlen,
J.M.,1990•. Measurement
of turbulent
diffusion
andresidual
displacements
bydyeexperiments
in the southernNorth Sea in 1982. Data Report.Rijkswaterstaat
(Neth.), Tidal WatersDiv. GWAO-
90.023.
Stelling,G.S., 1983. On the construction

of computational
methodsfor shallowwaterflow problems,
Ph.D. Thesis,TechnicalUniversityDefft, Neth.,232p.
Talbot, J.W. and G.H. Talbot, 1974. Diffusion in shallowseasand in Englishcoastaland esmarine
waters.Rapp.,P.V. Cons.Int. Explor.Met., 167, 93-110.
Van Dam, G.C., 1982. Models of Dispersion.In: Pollutanttransferand transportin the Sea, G.
Kullenberged. CRC PressInc., BocaRaton,Florida,91-160.
Van Dam, G.C., 1984. Reststromen en resttransportenin modelberekeningen.

Rijkswaterstaat
(Neth.),
Dir. W & W, PhysicsDiv., rpt. FA 8402, 33p. (Englishversion:Van Dam, 1988').
Van Dam, G.C.,1985'.Deeltjesmodellen,2DH (Particlemodels2DH). Rijkswaterstaat

(Neth.),Dir. W
& W, PhysicsDiv., rpt. FA 8507, 44p.
Van Dam,G.C., 1985•. Konstante

en aangroeiende
randomstap
in deeltjesmodellen.
Rijkswaterstaat
(Neth.),Dir. W & W, PhysicsDiv., note85-FA-219,5p.

G C van Dam and R A Louwersheimer 5O5
Van Dam, G.C., 1985c. A particlemodelappliedto the WesternScheldt.LiverpoolConferenceon

mixinganddispersionin estuaries,
4-25 Sept.1985/Rijkswaterstaat
(Neth.),Dir. W&W, PhysicsDiv.,
85-FA-226
Van Dam, G.C., 1986'.Residualcurrentsandtransport

in two-dimensional
modelcomputations.
Neth.
J. Sea Res. Publ. Series, 13, 66-67.
VanDam,G.C.,1986
• Particle
modelling
of2DHtransport
phemomena
intheWestern
Scheldt
Estuary.
Abstracts
International
Symposium
onPhysical
Processes
in Estuaries,
Noordwijkerhout,
TheNeth.:1.1.2.
Van Dam, G.C., 1987.Eddysimulation

of two-dimensional
dispersion.
Rijkswaterstaat
(Neth.),Tidal
WatersDivision,rept. GWAO-87.022, 26p.
Van Dam, G.C., 1988'. Residualcurrentsand transportin connectionwith two-dimensional

model
computations.
Rijkswaterstaat
(Neth.),TidalWatersDivision,GWAO-88.042,20p.
VanDam,G.C.,1988
b.Themathematical
modelling
of transport
phenomena
intheNorthSeainrelation
to experiments.
Rijkswaterstaat
(Neth.),TidalWatersDivision,rept.GWAO-88.043,27p.
VanDam,G.C.,1989'.Eddysimulationof dispersion.
Workshop on applying
non-linear
dynamics and
chaosin fluidmechanics,
June1989,Woudschoten,
Zeist,TheNetherlands/Rijkswaterstaat
(Neth.)Tidal
Waters Division, GWAO-89.313.
VanDam,G.C.,1989
b.Dispersion
ofdissolved
matter
intheNorthSea.ICES,Hydrogr.
Committee,
C.M.
1989/C:13,21p.
VanDam,G.C.,1990'.Two- andthree-dimensionalmodellingof dispersionand mixingin shallow

waters.IUTAM Symp.on FluidMechanics
of StirringandMixing,La Jolla,1990.
VanDam,G.C.,1990
•. Two-andthree-dimensional
models
of dispersion
based
uponmeasurements
in
theNorthSea.Proc.IAWPRC/EWPCAInt. Conf.on NorthSeaPollution,Amsterdam,
14p.
VanDam,G.C.,1990'.Numerical
simulation
of dispersion
by eddyvelocityfieldsin twodimensions.
Submittedfor publication.
VanderGiessen,
A. andR.R.Jansen,
1988.Het2,5D-model.
Eennieuwe
methode
terbepaling
vande
vertikalestromingssmfictuur
in deNederlandse
kustzgne.
Rijkswaterstaat
(Neth.)TidalWatersDivision,
GWAO-88.013.
VanVeen,J,, 1936.Onderzoekingen
in deHoofdenin verband
metdegesteldheid
derNederlandse
kust.
Alg. Lands•erij, Den Haag,Neth.,258p.
Van Veen,J., 1938.Watermovements

in theStraitsof Dover.,J. Cons.Int. Explor.Mer. 13,7-36.

Visser,D., 1990.A non-lineardiffusionmodelfor spreading

of river discharge
in a shallowsea.5th InC
Bienn.Conf. Physicsof EstuariesandCoastalSeas,Gregynog,Wales.
Weidemann,H. (editor), 1973. The ICES diffusionexperimentRFIENO 1965. Rapp.P.V. R6un.Cons.

Int. Explor. Mer. 163, 11lp.

25
An estuarine and coastal sand transport

model
B A O'Connor and J Nicholson
Abstract
This papercontainsa descriptionof an estuarineand coastalsandtransportmodel which is

currentlyunder development. Also describedare the resultsof teststo choosea suitable
mathematicalanaloguefor thevariousprocesses whichcontrolthe transport
of sediment.The
resultingmodeland its manner of operationwere thencheckedagainsta field situationand
shownto give promisingresults.
25.1 Introduction
Engineeringworksin estuariesand on coastlinesare increasingin both numberand scope.

As theseworkscan causemarkedchangesin the local morphology,whetherintentionalor
unintentional,
thereis a needfor sedimenttransportmodelswhichcanpredictsuchchanges.
A depth-averaged model of sandtransport,which can be appliedto both estuariesand

coasts,is currentlyunder development. The proposedschemeis modularin form, each
modulerepresenting one or more of the processes governingthe transportof sediment.
Describedherein are the model itself, including the results of tests which have been
undertakenin order to establisha suitablemethodof simulatingthe varioustransport
processes,
andthe applicationof the modelto a field situation.
25.2 Model description
25.2.1 General
Fine sandis commonlyfoundin estuarine

andcoastalareas,the suspended
component
of

5O8 B A O'Connor and J Nicholson
which, unlike the bed component,is unlikely to attain equilibriumwith the unsteadyand
non-uniform ambient flow conditions. For this reason, therefore, the model treats the
suspended andbed components of the sedimentload separately.In addition,a cross-shore
transportmoduleis includedin orderto reproducethe development of beachprofiles.
The model computesthe bed level changesin two stages. Firsfly, the depth-averaged
versionof theunsteady-state
diffusion-advection
equationis solvedfor thesuspended
sediment
concentrations.This equationtakesthe form:
DISPERSION
(25.1)
+--••&(ht,
UC)
+•(hrVC)
+Ro
- R•. 0
ADYECTIOIg SOURCE
/SINK
where:
(2 = depth-averaged
suspended
(kgm'3);
Dx = longitudinal
dispersion
coefficient
(m
Dy = lateral
dispersion
coefficient
(m2
hv = flow depth(m);
Rv = sediment
deposition
rate(kgm'2s4);
R•. = sediment
entrainment
rate(kgm'2s4);
t = time (s);
0 = depth-averaged
longitudinal
velocity(m s4);
•/ = depth-averaged
lateralvelocity(m s4);
x = longitudinalco-ordinate(m);
y = lateralco-ordinate(m).
In Eq. 25.1, detailsof theflow field (hv,U,V)aresuppliedby a separate

model,for example
that of Yoo and O'Connor(1988). The depositionrate (R•), which is a functionof the
unknown
variable
((2),is described
belowalongwiththeentrainment
rate(Rv),whichis
computedindependently.An alternatingdirectionimplicit (ADI) finite differenceschemeis
usedto solvethe overallequation.The resultsof two testswhichsimulatedthe progressof
a slugof suspendedsedimentsubjectto advectionand advectioncombinedwith dispersion,
are shownin Figs.la and lb respectively.
The flow velocitywas 1.4 ms4, thedispersion
coefficient
was0.1 m2s 4 andthetimeintervalwas0.5 s. Asis indicatedby themaintenance
of the shapeof the sedimentslugin Fig. la and the lack of any disturbance
in the wake of
the slug in both Figs. la and lb, the effectsof numericaldiffusionand dispersonwere

B A O'Connor and J Nicholson 509
minimal.In addition,
a checkonthequantity
of sediment
present
duringtheprogress
of the
slug showedthat masswas conserved.
START TIME = 0'5s

_1
"l
i
m 80
z
_o 6O
< 40
z
,,, ø
z
o
o 1.2m
• 1.0 1.0
1 '2m
1.o
0.8
0.•6
0.8
0.4 0'4
0'8
0.2
(o) Advection
START TIME = 0.5s

I
I
•" 100
'E I
• 80 I
z
o_ 6O I
4O
z
z
o
.2m
lo2m'•1.0 ' 1'0
1.00.'••6 0••.6
0-8 ,", 0.
0-8
81'0 0'4 0.4

, 0'8
0.2
(b) Advection and dispersion
Figure1: Modeltest- transport

of suspended
sediment
Thesecond stageof thecomputation

of thebedlevelchanges involved
solvingthebed
masscontinuity
equation.Bedmasscontinuity
is described
by theexpression:
(25.2)
(1-p)p,at
where:

510 B A O'Connor and J Nicholson
p = porosityof thebedmaterial(= 0.4);

Q•x = longitudinal
bedloadtransport
ratedueto tidal
andwave-inducedlittoralcurrents
0cgm4 s4);
Q•v= lateralequivalent
of Q•x0cgm4 s4);
Qcx= longitudinal
cross-shore
transport
ratedueto
undertow,waveasymmetry andlocalbedslope0cgm4 s4);
Qc•= lateralequivalent
of Qcx0cgm4 s4);
% = elevationof thebedrelativeto an arbitrarydatum(m);
p, = density
of thebedmaterial
(= 2650kgm4).
The only unknownin Eq. 25.2 is theelevationof thebed (%), detailsof the computation
of the independent
variables(Q•x, Q•v, Qcx, Qc•, Rx•,R•) being set out below. As the
deposition
rate(Rx))is a function
of thedepth-averaged
concentration
((2),thisvariable
is
dependent
on the solutionof Eq. 25.1. Eq. 25.2 is solvedby meansof an explicitfinite
difference scheme.
The structureof the model setup to solveEqs. 25.1 and 25.2 is containedin Fig. 2. A
description
of the constituent
partsof the modelfollows.
I INITIAL
BATHYMETRY
I
i HYDRODYNAMIC
DATA
I
BED LOAD DUE

TO CURRENTS
CROSS-SHORE
TRANSPORT
i ENTRAINMENT
I
I
I BATHYMETRIC
CHANGES
[ =,NAL
BATHYMETRY
I
Figure2: Model structure

25.2.2 Bedforms
The heightof the bedformsat any pointin plan is neededto fix the thickness of the mobile
bedlayer. Bothcurrentandwaveactionmaybe presentat anygivenlocationbuttheauthors
are unawareof any quantitative informationin theliteratureaboutbedformsgenerated under
suchconditions.This problemwasovercome,therefore,by determiningthe bedformheight
producedby currentsand wavesseparately and thentakingthe largerof the two to be the
representative
value. The determination of the characteristics
of the current-generated tipples
is basedon the work of Yalin (1977) and thatof wave-generated tipplesis basedon the work
of Nielsen(1985) or Swart(1976). The presentversionof the modelmakesno provisionfor
the occurrence of marine dunes.
The results of a comparisonof the Nielsen and Swart methodsof deriving the
wave-generatedtipple heightsis shownin Fig. 3. Clearly,thesetwo approaches
yield values
whichdiffer markedlyin trend. However,recentwork by RibberinkandA1-Salem(1990) has
confumedthat the Nielsen methodgivesthe more qualitativelycorrectresults,a sandbed
tendingto becomeplaneunderintensewave activity.
0'012
q(m)
0.010 _
0 038 -
GRAIN SIZE = 250 10'6 m
0.006 _ FLOW DEPTH= 2-5m
WAVE PERIOD = 10.0 s
NIELSEN (1985)
o-00z. _
--0 SWART (1976)
0-002 --
H(m)
0
0 0-2 0-4 0-6 0 8 1-0 1-2 1.4 1-6
Figure3. Bedformheight(q) versuswave height(H)

25.2.3 BedloadTransport
Bedloadtransport dueto currents,with or withoutthepresence of waves,is computed in three

stages.Firsfly,the steady-state
suspended loadtransport
rateis calculatedusingan empirical
expressionbasedon laboratoryand field data (unpublished).In its simplestform, this
expression appliesto unidirectional
flowsbut allowanceis alsomadefor the effectsof wave
actionandreversing(tidal) flows. Wave actionincreases the transportrate but theeffectsof
tidal actionare more complex. The finite durationof a tidal cycle may causethe water
columnto be undersaturated and inertiamay have the sameeffectduringaccelerating flows
but the oppositeeffect during deceleratingflows. The presentform of the empirical
expression doesnot differentiatebetweenaccelerating anddeceleratingflows.
Secondly,the ratio of the steady-state suspended load transportrate to the bed load
transportrateis determinedusingthe Bijker (1968) versionof the methodoriginallyproposed
by Einstein(1950). The Bijker-Einsteinmilo, however,hasbeenslightlyalteredfor usein
thepresentmodelin thatthe thickness of themobilebedlayeris setequalto half thebedform
heightratherthanthe totalbedformheightadoptedby Bijker. This resultsin an expression
of the form:
Os. 2.47 I: +I:
where:
I• = first Einsteinintegral;
I: = secondEinsteinintegral;
QB= bedload
transport
rate(kg m4
Qs= steady-state
suspendedloadtransport
rate(kg m
rB= bed roughness
height(m).
The lattervariable,in turn,is derivedusingthefollowingrelationship,

whichwasproposed
by val• Rijn (1982):
(25.4)
where:
D9o= ninetiethpercentile-by-weight
groinsize (m);
q = bedformheight(m);
X = bedformlength(m).
Finally, the known steady-state

suspended load transportrate (Qs) is combinedwith Eq.
25.3 to yield thebedloadtransport
rate(Qa). This methodof determining thelatterquantity
has beenadoptedbecauseit ensuresthat the bed load is compatiblewith the steady-state

suspended load and the latter is computedusingan expression

basedon a wide rangeof
laboratoryandfield conditions.Typicalvaluesproducedfor Q•QB andQBare givenin Fig.
4.
50 I i I 1 I I I I O-5
Qs/% Oe(kgm-1s-1)
40 O-4
GRAIN SIZE = 250 10'6 m
FLOW DEPTH = 2'5m
30
-- FLOW
VELOCITY
=1.0
m$-1 • 0.5
WAVE
PERIOD
10-0s •sI••
2O 0.2.
0.1
H(ml
0 I I I I I I I 1 o
0 0 2 0 z. 0-6 0.8 1 0 1 2 1./.. 1-6 18
Figure4: Qs/Q• and Qa versuswave height(H)
25.2.4 Cross-ShoreTransport
A comparisonwas madeof the followingcross-shore

transportsub-modelsin orderto select
a suitable scheme for inclusion in the main model:
Bailard (1982)
Nishimuraand Sunamura(1986)
Watanabe(1985)
Shibayamaand Horikawa(1985).
This comparisonentailed running the four schemesfor various deep-waterwave

characteristics
(Ho, Lo), grainsizes(D5o)and initial planebeachslopes(io); hydrodynamic
data,neededto drivethe sedimentmodels,weregenerated by meansof theYoo andO'Connor
(1988) wavemodel. The results,expressed in the form of the occurrence
of accretion,erosion
or no change(stability)at the intersectionpointbetweenthe meanwaterlevel and the beach
profile,werethencheckedagainstfield datacollatedby SunamuraandHorikawa(1974). The
latterhaveshownthatthe overallshapeof a beachprofileis characterised by the movement
of themeanwaterlevelintersection point. Basedonthisapproach, theWatanabemethodgave
thebestresults,a summaryof thelatterbeinggivenin Fig. 5. Even so,theWatanabemethod

producederodingbeacheswhentheyshouldhavebeenstableandstablebeacheswhenthey
shouldhavebeenaccreting.Clearly,this is a subjectrequiringfutureimprovement.
Ho/L
ø I •
FIELD SUNAMURA
&HORIKAWA
(197/.) • J
10'2
10'3 ,J I
10-•
{iø)-ø
:•?(Dsø/Lø)ø'•?
10-•;
Figure5: Beachtype as functionof wave andsedimentcharacteristics
25.2.5 Sediment Entrainment
Nine possiblemethodsof derivingthe sedimententrainmentrate were considered.These

variousmethodsfall into two categories,
namelythosewhichgivetheentrainment ratedirectly
and thosein whichthe entrainment rate is derivedfxomthe steady-state
near-bedsuspended
sedimentconcentration.In the lattercase,theentrainment rateis setequalto the steady-state
depositionrate so that:
- % c; (25.5)
where
C•ø = steady-state
near-bed
suspended
sediment
concentration
(kg m'3);
Ws = sediment
fall velocity(m s'•).
The relevantentrainmenttheories,and the conditionsfor whichtheywerederived,are as

follows:

Direct approach:
Van Rijn and Tan (1985) - unidirectionalflow and waves;

Van Rijn (1985) - unidirectionalflow;
Nakagawaand Tsujimoto(1980) - unidirectionalflow;
FernandezLuque (1974) - unidirectionalflow;
De Ruiter (1982) - unidirecfiona•flow.
Indirectapproach:
Nielsen (1986) - waves;

Engelundand Fredsoe(1976) - unidirectional
flow;
FernandezLuque and van Beek (1976) - unidirectionalflow;
O'Connor (O'Connor, Nicholson and Yoo, 1989) - unidirectionalflow and waves.
30
i I I ! I ! i I
RE (kgm-2s-•) VAN RIJN AND TAN [1985)
VAN RIJN [1985}

NAKAGAWA & TSUJIMOTO[1980)
........ F. LUQUE {197/,}
../ ......... DE RUITER [1982)

..... NIELSEN [1986)
20 ............ ENGELUND & FRED$OE [1976)
....... F. LUQUE & VAN BEEK [1976)
............... O'CONNOR
..J
.J
Figure6: Entrainment
rate(R•.)versuswaveheight(H)
An exampleof theresultsproduced by the varioustheories

is shownin Fig. 6. When
producingthe data containedin this figure, it was assumedthat theoriesderived for
unidirectional
flow conditions
couldalsobe appliedto situations
involvingwaveaction,
providedthat the appropriatevalue was usedfor the bed shearstress; the latter was

calculatedusingthe O'Connorand Yoo (1988) bed frictionmodel. The contentsof Fig. 6

indicatethat all the resultsare in generalquantitativeagreement,exceptfor the very large
entrainmentratesgivenby the EngelundandFredsoeandthe FernandezLuqueand van Beek
methods. This discrepancycan be explainedby the fact that a value of one or two grain
diameterswas adoptedfor the thicknessof the mobile bed layer in the latter two methods,
regardlessof the presenceof bedforms. However, if it is assumedthat the mobile bed
thicknessis directlyrelatedto thebexlformheight,thenthe near-bedconcentrations, andalso
the entrainmentrates,are reducedaccordingly.
25.2.6 SedimentDeposition
The sedimentdepositionrate is def'medby the relationship:
- (25.6)
where:
Cs = instantaneous
near-bedsuspended
(kg m'3). However,the
solutionof thedepth-averaged
diffusion-advection
equation(F-xl.25.1) yieldstheinstantaneous
depth-averaged
concentration
(2. Hence,if theratiobetween
thesetwoquantities
is written
as:
(25.7)
combiningEqs. 25.6 and 25.7 yields:
- ca
The onlyunknown quantityin Eq. 25.8 is I•. A valueis establishedfor thisvariableby

assuming thattheform of the verticalconcentration profilecanbe represented by theRouse
(1937)distribution.
Thereference levelcontained in theRousedistribution is equatedto the
mobilebedthickness; thelatter,in turn,is setequalto halftheroughness heightgivenby Eq.
25.4. This approachyieldsthe followingexpression for the concentration
ratio:
I• "0.22 (25.9)

where:
mobilebed layer thickness(m).
Valuesof [5,as a functionof flow velocity(U) for a fixedmediangrainsize(D•0),flow

depth(hv), wave height(H) and wave period(T), are shownin Fig. 7; both the Nielsen and
the Swart methodshavebeen usedto computethe wave-generated bedformcharacteristics.
Theunrealistic
change
in thevalueof [5at a flowvelocityof approximately
0.4 m s'1reflects
the separationof unidirectionalflow effectsfrom wave effectsin the determinationof the
bedform characteristics(see Sect. 25.2.2).
12 I I
r•(lO3)
I
WAVE GENERATED RIPPLES:
NIELSEN (1985}
1.0
SWART (1976}
GRAINSIZE = 25010'Gm
FLOW DEPTH = 2-5m
0.8
WAVE HEIGHT = 0-Sm
WAVE PERIOD = 10-0s
0.6
0'4 -WAVE-GENERATED RIPPLE.• CURRENT-GENERATED RIPPLES WAVE-GENERATED RIPPLES -
0.2
I I I ! I I • (ms-l)
0 0.2 0-4 0-6 0-8 1.0 1-2 1-4 1-6
Figure7: [•-function
versus
flowvelocity(0)
25.3 Model application
25.3.1 Field Dam
The modelwastestedby applyingit to a field situation,detailsof whichhadbeenmonitored

by the Danish Hydraulic Institute (1984). The field situation consistedof a trench

approximately
40m wide by 6m deepwhichhadbeendredgedacrossthe inshorezoneof the
DanishNorth SeaCoastin orderto bringashoreoil andgaspipelines.The trench,the layout
of which is given in Fig. 8, had been surveyedat regular intervals and continuous
measurementsmade of the wave climate. Details were known, therefore, of the siltation in
the trench and the associated wave conditions.
ß ß
....
ß ß ßßII 'ßOm /
ß Ii
ß ß ß
ß
ß e
e e
ßß,
ß ß ß ß ß ß
COASTLINE
DEPTH [m)
6
5
5
6
TRENCH
1600m
Figure 8: Layout of the dredgedtrench
A severestormoccurredbetweentwo successive surveysand,asthewaverecordsindicated

that virtuallyall the sedimentmovementassociated with the monitoringperiodcouldbe
ascribexlto this one storm,only the latter was reproducedin the model. In addition,it was
assumed that a singlewaveheightcouldbe usedto represent the stormdueto the abrupt
growthand decayof the wave intensity. The resultingrepresentative conditionswere as
follows:
Wave period = 5.5 s

Breakerdepth = 4.8 rn
Breakerheight = 3.0 rn
Breakerapproachangle = 7.5 o
Stormduration = 4.3 x 10• s (5 days)
Mediangrainsize = 200 lam.
The hydrodynamicdata neededto drive the sedimentmodel consistedof a cross-shore

profileof the longshore
current,whichhadbeenproduced by Deigaardet al (1986),anda
wavefieldbasedona mixtureof availableinformation
andinformation
derivedfroma plane

beachconfiguration.
Tidaleffectswereconsidered
to benegligible,
themeantidalrangebeing
less than one metre.
25.3.2 Model Results
Beforethe startof a modelrun, the systemwas operatedwith fixed bathymetryuntil the

suspendedsedimentconcentrations had attaineda state of equilibrium. The resulting
concentrations werethenusedastheinitialvaluesfor themodelrun. The lattertookplace
in a seriesof time steps,at theendof eachof whichtheflow field wasadjustedto allow for
theeffectsof depthchanges.Thisdirectcouplingbetweenthehydrodynamic "model"andthe
sedimentmodelcausedthe outputfromthe latterto be dependent on the time step.
10s
• 10
,
103
103 10•, 10s 10•
DURATION is)
Figure9. Time stepeffects
The effectof thetimestepon themodeloutputis illustrated in Fig. 9, whichshowsthe

volumeof siltationin thetrenchasa functionof time. Theparticular versionof themodel
associated
withFig.9 incorporated theSwartbedform sub-model andthevanRijnentrainment
sub-model.Fromthecontents of thefigure,it canbe seenthattheamountof siltationvaried
inverselywiththetimestepbutthatthistrendbecamenegligible
if thetimestepwaslessthan
100s. It wasconcluded,therefore,thata timestepof 100s yieldedthebestcompromise
betweenoperating
requirements
andaccuracy.

The contentsof Fig. 9 alsoindicatethattherateof infillingdecreaseswith thepassage of

time. This factor can causethe model outputto be misinterpreted ff short-termresultsare
extrapolated
to givelong-termvalues.Forexample, running themodelfor 10• s realtime
witha timestepof 100s leadsto anextrapolated
infillingvolumeof 184000m3 after4.3
10•s.Thiscompares witha fullycomputedvalueof only86 000m3.
The totalmeasuredsiltationin thedredgedtrenchwasthencomparedwith theoutputgiven

by versionsof the sedimentmodelwhichincorporated variouscombinations of the bedform
and entrainmentsub-models. Only a limited numberof thesetestscould be completed,
however,due to instabilityproblems.The latterwere causedby very largedeposition rates
which so alteredthe bathymetryduringthe courseof a time stepthat the waterdepthswere
no longercompatiblewith the flow field. Althoughit may havebeenpossibleto completea
failed testby employinga reducedtime step,the run timesundersuchcircumstances would
havebeenunacceptably large. Nevertheless,the fact thata testfailedwasa strongindication
thattheparticularversionof themodelbeingexaminedyieldedunrealistic values.The results
of the successfultestsare containedin Fig. 10, which showsthat the versionof the model
containingthe Swartbedformsub-modeland the de Ruiterentrainmentsub-modelyieldeda
totalsiltation
volumeof 114000m3. Thisquantity
compares favourably
withthemeasured
valueof 150 000 m3, an errorof 24 per centbeingwell withinthe accepted
level of
engineeringaccuracy.
105
/.."'
•z 10•
/
•/B E___DF
ORM$ .ENTRA_I__
NIELSEN VANRIJN& TAN
-- SWART DE RUITER
SWART VAN RIJN & TAN
SWART F. LUOUE
I
103
103 10s 10•
DURATION
Figure 10: Comparisonof computedinfilling rates

A moredetailedcomparison betweenthe modeloutputand the field data,basedon the

distributionof infilling alongthe lengthof the trench,is shownin Fig. 11. There are three
distributionsin thefigure,namelytheoutputgivenby thepresentmodel(incorporating Swart
bedformsand de Ruiter entrainmen0,the output given by a longshoretransportmodel
developed by Deigaardet al (1986) andthe measured values. Again,thereis a reasonable
measure of agreement betweentheoutputfromthepresentmodelandthefielddata,especially
whenallowanceis madefor theapproximate natureof the hydrodynamic datausedto drive
the former.
Finally,thecomputed andmeasured distributions

of infillingin two trenchcross-sections
arecompared in Fig. 12. The agreement betweenthecomputed andmeasured valuesis very
poorin theinshorecross-section (chainage700), asis to be expectedfrom thecomparison of
the bulk siltationquantitiesshownin Fig. 11. A probablereasonfor thisdiscrepancy is the
omission of anyallowancefor theeffectof breakingwaves on sediment entrainmentinside
the surf zone. On the other hand, the agreementbetweenthe computedand measured
distributionsof infilling in the offshorecross-section(chainage1350) is good. Again, this
reflectsthe comparison containedin Fig. 11.
1000
NEASURED
I
800 ------- CONPUTED
'7E ....... DEIGAARD ET AL (1986)
• 600
• 400
200
0
5OO 700 900 1100 1300 1500
CHAINAGE {m)
betweenmeasured
andcomputedlongitudinalinfilling

25.4 Conclusions
A depth-averaged estuarineand coastalsedimenttransportmodelhasbeenestablished which

incorporatesthe effectsof wavesand currents.The modelis of modularconsauctionso that
it can be updatedin the future as understanding
of sedimentprocessesis improvedthrough
furtherresearch.Testson the varioussedimentprocessmoduleshaveindicatedthatbexlload
transportand sedimententrainmentare adequatelyrepresented, bedformsand cross-shore
transportarerathercrudelyrepresented andmayneedimprovement while sedimentdeposition
appearsto be unrealisticundercertainflow conditionsandmay also needimprovement.
!CHA,NAGE
700i
• ORIGINAL
..... MEASURED -
COHPUTED
T
ee e * ß ß ß ß ß ßßß ß ß
W ß
C• ß
ß
ß
ß
12
20 0 20 40
OFFSET Iml
[CHAINAGE
1350J
•
ORIGINAL i
..... MEASURED
COMPUTED
• ".2
12
40 20 0 20 40
OFFSET Im}
betweenmeasured
andcomputed
lateralinfilling

Application
of themodeltoa simplified
fieldsituation
involving
thesiltation
of a dredged
trenchin a coastalenvironment
yieldedrealisticanswers, whichwerewell withinengineering
accuracy. Thefieldapplication
alsoshowed theimportance of including "feedback"between
morphological changes andthehydrodynamic flow field,theneedto choose a suitable
model
timestepandtheneedto operate themodellongenough to ensurethatany"scaling-up" of
short-termruns,whichreduces modeloperation costs,is representative
of thefinalconditions.
Finally,a detailedexaminationof the longitudinalsiltationpatternin the dredgedtrench
showed that the model underestimatedthe siltation in the inshore sectionof the trench, which
suggests
thatadditionalfield testingis requiredin orderto providefurthercheckson the
accuracyof the scheme.
Acknowledgement
This work was undertakenas part of the MAST G6 CoastalMorphodynamics research
programme.It wasfundedby theCommission of theEuropean
Communities, Directorate
Generalfor Science,Researchand Development,
underMAST contractno. 0035.
Symbols
(2 = depth-averaged
suspended
sediment
concentration
(kgm'3);
Ca = instantaneous
near-bexl
suspended
sediment
concentration
(kgm'3);
C•ø= steady-state
near-bed
suspended
sexliment
concentration
(kgm'3);
D•o = mediangroinsize (m);
D•o = ninetyper centfiner-by-weight
grainsize(m);
Ds = longitudinal
dispersion
coefficient
(m"
Dy = lateral
dispersion
coefficient
(m"s'•);
ha = mobilebed layer thickness(m);
hv = flow depth(m);
H = wave height(m);
Ho = deepwater wave height(m);
io = initial beachslope;
I• = first Einsteinintegral;
I,. = secondEinsteinintegral;
Lo = deepwater wave length(m);
p = bed layer porosity;
Qax= longitudinalbed load transportrate due to tidal and wave-inducedlittoral currents
(kg m'• s'l);
QB¾
= lateralequivalent
of QBx(kgm'1

Qcx= longitudinal
cross-shore
transport
ratedueto undertow,
waveasymmetry
andlocalbed
slope(kg m4 s4);
Qc•= lateralequivalent
of Qcx(kgm'l s'l);
Qs = suspended loadtransport
rate(kgm'l s'l);
rB = bedroughness
height(m);
RD = sediment
deposition
rate(kgm'2s'l);
RE = sediment
entrainment
rate(kgm'2s4);
t = time(s);
T = wave period (s);
0 = depth-averaged
longitudinal
velocity
(m s'l);
9 = depth-averaged
lateralvelocity(m
x = longitudinalco-ordinate(m);
y = lateralco-ordinate(m);
Ws = sediment
fall velocity(m s'1);
za = elevationof the bedrelativeto an arbitrarydatum(m);
= cde;
,,t = modeltime step(s);
q = tipple height(m);
•, = tipple length(m);
p, = sediment
density
(kgm'•).
25.5 References
Bailard,J.A., 1982: Modelingon-offshoresedimenttransportin the surfzone. Proc.EighteenthCoastal

EngineeringConf., ASCE, 1419-1438.
Bijker,E.W., 1968: Littoraldrift asfunctionof wavesandcurrent.Proc.EleventhCoastalEngineering

Conf., ASCE, 415-435.
DanishHydraulicInstitute,1984: Monitoringof sedimentation

in a dredgedtrench.Report,Horsholm,
Denmark.
Deigaard,R., Fredsoe,J. and Hedegaard,I.B., 1986: Mathematicalmodel for littoral drift. Jour.
Waterway,Port, CoastalandOcean Engineering,ASCE, Vol. 112, No. 3, 351-369.
De Ruiter,J.C.C., 1982: The mechanismof sedimenttransporton bed forms. Mechanicsof Sediment

Transport,Euromech156, 137-142.
Einstein,H.A., 1950: The bed-loadfunctionfor sedimenttransportation in open channelflows.

TechnicalBulletin No. 1026, United StatesDepartmentof Agriculture,Soil ConservationService,
WashingtonD.C.

Engelund,F. andFredsoe,J., 1976: A sedimenttransportmodelfor straightalluvialchannels.Nordic

Hydrology,Vol. 7, 293-307.
Fernandez
Luque,R., 1974: Erosionandtransport
of bed-loadsediment.
Dissertation,
KRIPS ReproBV,
Meppel,Netherlands.
FernandezLuque,R. andvan Beelc,R., 1976: Erosionandtransportof bed-loadsediment.Jour.Hyd.

Res., Vol. 14, No. 2, 127-144.
Nakagawa,H. andTsujimoto,T., 1980: Sandbedinstabilitydueto bedloadmotion. Proc.ASCE, Jour.

Hyd. Division,Vol. 106, No. HY12, 2029-2051.
Nielsen,P., 1985: A shortmanualof coastalbottomboundarylayersandsedimenttransport.Technical

Memo. 85/1, Public Works Dept., Sydney,Australia.
Nielsen,P., 1986: Suspendedsedimentconcentrations

under waves.CoastalEngineering,Vol. 10,
23-31.
Nishimura,H. andSunamura,T., 1986: Numericalsimulationof beachprofilechanges.Proc.Twentieth

CoastalEngineeringConf., ASCE, 1•.•.•.-1455.
O'Connor,B.A., Nicholson,J. andYoo, D.H., 1989: CoastalImpactModelling. Volume3. Report,

Universityof Liverpool.
O'Connor, B.A. and Yoo, D.H., 1988: Mean bed friction of combined wave-current flow. Coastal
Engineering,Vol. 12, 1-21.
Ribberink,J.S.andAl-Salem,A., 1990: Bedforms,sediment

concentrations
andsedimenttransportin
simulatedwaveconditions.Abstrs.Twenty-Second
CoastalEngineering
Conf., ASCE, 501-502.
Rouse,H., 1937: Modernconceptions

of the mechanicsof fluid turbulence.Trans.ASCE, Vol. 102,
463-543.
Shibayama,
T. andHorikawa,K., 1985: A numericalmodelfor two-dimensional
beachtransformation.
Proc. JSCE, No. 357/II-3, 167-176.
Sunamura, T. and Horikawa, K., 1974: Two-dimensional beach transformation due to waves. Proc.
FourteenthCoastal EngineeringConf., ASCE, 920-938.
$wart,D.H., 1976: Coastalsediment

transport.Computation
of longshore
transport.Report968-I, Delft
HydraulicsLab., Netherlands.
Van Rijn, L.C., 1982: Equivalentroughness

of alluvialbed. Proc. ASCE, Vol. 108, No. HY10,
1215-1218.

Van Rijn, L.C., 1985: Mathematical modelsfor sediment

concentration
profilesin steadyflow. Publ.
No. 354, Delft HydraulicsLab., Netherlands.
Van Rijn, L.C. andTan, G.L., 1985: Two-dimensional

verticalmathematical
modelfor sedimentation
in dredgedchannelsandtrenchesby currentsandwaves.Rijkswaterstaat
Communications,No. 41.
Watanabe,A., 1985: Three-dimensional

predictivemodelof beachevolutionarounda structure.Proc.
Water Wave ResearchConf., Hannover, 123-141.
Yalin, M.S., 1977: Mechanicsof sedimenttransport.PergammonPress,Oxford,England.
Yoo, D.H. andO'Connor,B.A., 1988: Numericalmodellingof wavesand wave-inducedcurrentson

groynedbeach. Proc.IAHR Symp.on MathematicalModellingof SedimentTransportin the Coastal
Zone, Copenhagen,127-136.

Part VI
Applied Studies
26
Current and density structure in the

Netherlands coastal zone
W P M de Ruijter, A van der Giessenand F C Groenendijk
Abstract
Resultsare describedof a measurementprogrammein the outflow zoneof the Rhine river that
hasbeengoingon over the pastyearswith varyingintensity. Its aim is to describethe three-
dimensionalhydrographicstructurein the area at differenttime and spacescales.
The densitydistributionin the Dutchcoastalzoneis controlledby the salinityfield. Here

thelargestsalinitygradientsof theNorth Seaare observed.This is dueto therelativelyhigh
fiver discharge
and the fact that its northwardspreading
is mostlyconfinedto a coastalband
within 20kin from the shore.
Varyingconditions of bothwindandfreshwaterdischarge leadto a highlyvariabledensity

and currentstructure. Especiallyin the upperlayer and up to 80kin from the fiver outlet
almostall individualsalinitypatternsdeviatesignificantly
fromthe mean. In thisregionthe
water columnalso often is stronglystratified.
Currentswere measuredcontinuouslyover a one year period in a cross-shore transect

locatedapproximately40kin north of the main outflow port. As expected,the mean
southwesterlywindsmainlydrive a longshore northwardcurrent,with averagevaluesin the
upperlayerbetween7 and 1lcm/s. The largecross-shore densitygradientis relatedbothto
a longshoregeostrophicflow componentwith large vertical shearand to a cross-shore
estuarinecirculationcomponent.Thus,thislongshoreflow is enhancednearthe surfaceand
reducednear the bottom. Almostpermanentlythe observednearbed residualcurrentwas
directedon-shore,with a one year meanvalueof 3cm/s.
Underextremedischarge conditions
combinedwithstrongwindson-shore nearbedcurrents
of ca. 10cm/swereobserved.Duringthiseventstratification
wasveryslxong throughout
the
coastalzone. In the upperlayerlongshore
currentspeedsappearedof up to 50cm/s,whereas
in the lower layer thesewere only a few cm/s.

53O W P M de Ruijter, A van der GiessenandF C Groenendijlc
Vertical currentprofilesare very sensitiveto wind direction. Wind- and densitydriven

components may counteractor reinforceeachother. Consequently, duringeasterliesa large
veeringof thecurrentdirectionwith depthwasobserved.Duringsouthwesterlies thisis only
minor,but the amplitudedecreases stronglywith depth. In caseof northerliesonly a small
vertical variationof residualcurrentspeedwas observed.
26.1 Introduction
The input of nutrientsand contaminants into the Netherlandscoastalzone (Fig. 1) by the

Rhine fiver outflow has a strongimpacton its water quality. In the SouthernBight of the
North Sea the dischargeof the Rhine/Meuseestuaryconstitutesthe major sourceof fresh
water. Consequently, stratificationoccursin a large part of the Dutch coastalregion(Van
Alphenet al, 1988). The three-dimensional characterof the densityfield is certainlynot
limitedto the immediatevicinityof the mainfiver wateroutlets,theRotterdamWaterwayand
the Haringvlietsluicecomplex(seeFig. 2 for locations).
[ I I
0 50 100 km
SOUTHERN 8/GHT
Figure1: The SouthernBightof theNorthSeaandits bathymetry(depthcontours

in meters).The area
is indicatedwheremostof the observations
weretakenreportedin thispaper.

W P M de Ruijter, A van der GiessenandF C Groenendijk 531
Overa largedistanceafterits debouchment intotheNorthSeafiver waterspreading causes

spatialinhomogeneities in the salinityfield. Associatedwith it is a markedthree-dimensional
structureboth of currentsand suspended matter (Van der Giessenet al, 1990, Visser et al,
1991). Informationaboutthisverticaldistribution is essentialto developappropriatemodels
to supportthe managementof the coastalenvironment.The extentof the stratifiedarea is
largelycontrolledby windandfiverdischarge conditions,
bothlargelyvariable.Consequently,
the salinitypatternis highlyvariableas well.
0•• 5•70
ß
' 5850
Callan
5•20
IJ•iden
•oo
o
y78o
y/7o
Tert•lde
o
5750
Y7•O
Figure2: Locationof cross-shore

transects
whererepeated
CTD measurements
weretaken. Coordinates
are in km (UTM-system)

532 W P M de Ruijter, A van der GiessenandF C Groenendijk
In thispaperthe meandensityfield andits variabilityarepresented andanalyzed(section

26.3), as determinedfrom repeatedcruisesin 1986 and 1987 in the areabetweenthe Dutch
coastand 30km off-shore(Fig. 2). Beforethat,in section26.2, a generaldescription of the
hydrography of theareais given. Section26.4 reviewstheresidualcurrentstructure
measured
for a one-yearperiodfrom a seriesof currentmetermooringsin a cross-shore transectnear
Noordwijk,located40km northof the main fiver outlet. Finally, in section26.5 the results
are shortlysummarizedand discussed.
26.2 General Characteristics of the Dutch Coastal Zone
The southern partof theNorthSeais characterized by its shallowness

(Fig. 1) andstrongtidal
currents,with maximumcurrentspeeds rangingbetween0.7 and 1.1m/sanddirectedmainly
parallel to the coast. Consequently, over mostof the area the water columnis well mixed
throughout the year. Major exceptions are the outflowregionsof the riverswherestrong
halinestratification existsthroughoutthe year. Currentvariationsoverthe spring-neap tidal
cycleare substantial (Van der Giessenet al., 1990. The dischargeof the Rhinefiver varies
strongly.It ranges
between
600and13000m3/s
witha longtermaverage
of 2200m3/s
(Fig.
3). Generally,discharges
are low in late summerandfall, higherdischarges
are observedin
winterandspring.Shorterindividualpeaksfrequentlyoccur,mainlydueto heavyrainevents
over the catchment area.
6000.
'1500.
E
_c•ooo.
o 1500.
--
I C •
o
•,
o oo
Figure3: Discharge
andwindconditions
duringtheobservation
periods.A: Rhinedischarge.
B: wind
vector. C: periodnumbersindicatingindividualcruisesin the coastalzone

W P M de Ruijter,A van der GiessenandF C Groenendijk 533
Evidently,the otherkey factorcontrollingthe densityand sub-tidalflow fieldsis the wind.

It is also highly variable,with southwesterliescharacterizingthe climatologicalmean (Fig.
3).
Thebulkof theRhinewatercombines
withtheMeusefiverdischarge
(averaging
300m3/s)
anddebouches
into theNorth Seathroughboththe RotterdamWaterwayandthe Haringvliet
sluices
(Fig.2). Thelatteraremanipulated
withtheaimto guarantee
a discharge
of 1500m3/s
throughthe Waterwayas long as possibleto preventsalt wedgeintrusion. Above a fresh
waterdischarge
of 1500m3/stheremainder
entersthecoastal
zoneviatheHaringvliet.Only
at discharges
over6000m3/s
anextratransport
throughtheRotterdamWaterwaytakesplace.
In front of the Dutch coastthe fiver water mixes with open sea water that has enteredthe
southern NorthSeathroughtheDoverStrait. Thisinfluxof Channelwatervariesconsiderably
with wind speedand direction(Prandle,1978).
Importantlengthscalesthat characterize the nearfield areaof the outflowingriver water

are the internalRossbyradiusof deformationandthe inertialdeformationradius. The former
definesthe horizontalscaleover which the Coriolis force and the pressuregradientforce
relatedto the horizontalsalinitygradientbecomeof equalorderof magnitude. The latter
estimatesthe scaleafter whichinertialand Coriolisforcesmay balance. Both lengthscales
are closeto 10kin in the Rhineoutflow(Van Alphenet al, 1988) whichis the sameorderof
magnitudeas the tidal excursionlength. At largerdistancesfrom the mouththe influenceof
earth rotation becomesmore significant. Observationsnear the mouth of the Rotterdam
Waterwayrevealformationof a sharpsurfacefrontwhentheconvergence is strongest
between
buoyantfiver outflow and maximumsouthwardebb-tidalcurrenm.At thisphaseof the tide
fiver outflow is maximal and driven southwardby the longshoretidal current. The front is
clearly visible by an abruptchangein colourand collectionof foam and floating materialat
the convergence line. The very strongsurfaceconvergence is accompanied by significant
downwellingandinternalfrontformationat the northernedgeof theoutflow(Van Alphenet
al, 1988). Near field observations have shownthat plume dimensions vary considerably
dependingon dischargeand wind field.
26.3 The ObservedDensity Field

Over the years 1986 and 1987 repeatedcross-shore CTD-sectionswere taken in a zone
betweenthecoastand30kmoffshore(Fig. 2). Measurements wereperformedby theresearch
vessel'Holland'usinga GuildlineCTD probe. The observational periodcomprises a wide
varietyof meteorologicalandfiver discharge
conditions(Fig. 3). At everystationthevertical
temperatureandsalinitystructurewasobserved at depthintervalsof ca. lm. Underway,near
surfacesalinitieswere takencontinuously.

534 W P M de Ruijter,A van der GiessenandF C Groenendijk
NoorthftJk 890926
•um
Temper•t,Umre
, I Tempermture
-B
Figure4: Temperature,Salinity andsigma-t-fieldscross-shorenearNoordwijk(40km northof the river

mouth). The observations were takenon two consecutive late Septemberdayswhen the wind veered
from southwestto north-northeast (wind shownon the bottomof the figure). It showsthat in this area
the salinityfield determinesthe densitystructure.The temperaturethenfollowsand adjustlocallyto
atmospheric conditions.This responseappearsto be stronglyaffectedby the local depthof the wind-
mixedlayer.
In thisregionthestructure of thedensityfield is determinedby thesalinityfield dueto the

dominantinfluenceof the fiver outflow. As an example,Fig. 4 showstemperature, salinity
anddensityfieldsin theNoordwijktransect measured on twoconsecutive lateSeptember days
whenthe wind veeredfrom southwest to north-northeast.The responseto the changein wind
directionis remarkable,due to the associated Ekmantransportin the upperlayer changing
fromonshoreto offshore.Most strikingis thechangeof the temperature field. Temperature
is clearlynon-conservative, in particularin the upperlayer whereit adjustslocallyto the
atmospheric conditions.In thisexample(Fig. 4) therelativelycoolnortherlies leadto lower
temperatures in theupperlayer,theeffectbeinglargestaround4 and20kmoffshore,because
therethe mixedlayer depthis minimaland thuscoolingmaximal. The temperatures in the
lowerlayerstill reflectthe conditionsof the priorday.
As a consequence of theabove,temperature is nota usefultracerto derivee.g.advective

properties
of the flow field, nor is an analysisusingTS-diagrams veryusefulin thisregion.

In the sequelthe analysiswill thereforebe limited.
Expectedly,the averagesalinityfield (Fig. 5,6) showsa gradualweakeningof vertical

stratificationwith increasingdistancefrom the main fiver wateroutlet. Averageswere taken
over 8 cruises(see Fig. 3). In the Ter Heijde transect(Fig. 5a) 10km northof the mouth,
stratificationis strongest,The relativelyfreshwaterlayer is conf'med
mainly to the upper5-
8m. Variationsin wind and dischargeare reflectedmainly in largevariabilityof the upper
layer characteristics.This closeto the fiver mouththe large horizontalsalinitygradientis
associated with a persistentestuarine-type crossshorecirculationcomponent.It showsin a
largeslopeof the isopycnals, far inshorepenetrationof dense,highlysaline,openseawater
and relativelysmall variabilityin the lower layer (also due to tidal mixing). A relative
minimumof variability(in termsof standarddeviation)around4km off-shoremay indicate
a preferredupwellingarea. Northerliesand offshorewinds may furtherthe occurrenceof
upwelling,related to offshoreEkman transport. A similar phenomenonis observedoff
Noordwijk (Fig. 5b), 30km furthernorthward.
However,the structureof averageand standarddeviationof the salinityfield in the Ter

Heijde transectis also stronglyaffectedby near field processes as it is only one internal
Rossbyand inertial radiusaway from the source. Both analytical(Garvine, 1987) and
laboratorymodels(McClimans,1988) of fiver plumesandfrontsindicatethat nearthe mouth
thereis an inertialturningregion. Streamlines
of theoutflowseparate from thecoaston both
sidesof the sourceandstartturningto the fight underthe influenceor earthrotation. Within
severalradii of deformationthe plumereattaches to the coastandformsa coastalcurrent.
Althoughthesemodelsignorewind and tidal effectsthe structureof mean salinitiesand

theft standarddeviationat Ter Heijde and Noordwijk seem to reflect the effect of the
dynamicalprocesses incorporated.The mean positionof the turningplume is situated
approximately 10kmoffshorefromTer Heijde(Fig.5a);reattachment takesplacebetweenTer
Heijde and Noordwijk and the positionof a coastalfront around7km from the shoreis
suggested by the crowdingisolinesof the standard
deviation(Fig. 5b) at Noordwijk.
Spreadingand mixingprocesses, involvingconsiderable sheardispersion, act to weaken

stratificationandcross-shore gradient.Consequently, movingnorthward salinityvariability
decreases, in particularin theupperlayer,but the widthof the areaaffectedby fiver water
becomes larger.Thisis a resultbothof verticalexchange in combinationwith highlysheared
currentsandreducedon-shorepenetration of the salinewedge,relatedto a weakeningof the
cross-shore estuarineflow component.On movingfurtherawayfrom thefiver watersource
the watercolumnbecomesmoreandmoreverticallyhomogeneous (Fig. 5c, d), reflectedin
steeper isopycnals andisolinesof variability.Nevertheless,
in theCallantsoogtransect, located
ca. 100kmnorthof thefiver mouth,eventhemeansalinityfield is still weaklystratified.

TerHeyde
11• • T• lH15 lHlO 1807 11• lHI.5
.' 3•.5' t.5 ' 5 ' ' ' •.5

-5
-10
-20
•5 20 t0
-25
t)Jstance to coast
11•
-5 _
-10
.4
• -•0
0 • • 15 10 5
-•5
t)Jstance to coast
Noord•jk
• • • 115 IlO I7 I4
3. 5 ' 3t.5 30.5 .5 8.
33 5 5 .5
33. ß
5 2. t.5 t. 30.5
• 33.
Bistance to coast {km)
-5
-t0
-•5 '6
6
-2{)
20 t5 t0 5
-25
O•stance to coast
Figure5a,b: Averagesalinitydistribution(in 0KI0) andstandarddeviationover 8 cruisesin

1986 and 1987; a: Ter Heijde, b: Noordwijk

W P M de Ruijter, A van der Giessenand F C Groenendijk 537
Egmond
• E•5 E•O •l• •lO ,EN07 ,Era ,ENI
-iO . .5 . i.5 3t .5
ß
-20 33.5 ß
-25F 25 20 15 t0 5
• E•5 E•O •l• •lO EN07 ,Era ,ENI
-20
5 20 t5 t0 5
-25
Distance to coast
Callantsoog
cuo u.zs CL2O cu5 ,cLio ,cLo7 ,cm ,cu.5

33. 5
-5
-t0
32.5
-20
20 t5 t0 5
-25
(130 II.Z5 (1..20 CLI5 CLIO CL07 CL04 CLI.5
-5
-t0
-is
-20
-25
Distance to coast (Ice}
Figure5c,d:Averagesalinitydistribution(in 0/00) andstandarddeviationover 8 cruisesin 1986 and

1987;c: Egmond,d: Callantsoog,e: Goeree(seeFigure2 for locationsof transects

538
W P M deRuijter,A vanderGiessen
andF C Groenendijk
It is interesting
to notetheincreasedgradient of salinityvariabilitybetween10 and15km
offshore fromCallantsoog (fig.5d). Thisis alsotheareaof relatively maximum stratification
and crossshoredensitygradient,suggesting that herethe positionof the frontalzone
separating openseawaters fromtheregionof majorfiverinfluence is relatively
stable.This
maybe a consequence of therathersteepbottomslopeinhibiting (byrotational constraints)
cross-shore displacementof watercolumns andsteering theflow alongit. However,sofar
the numberof cross-shoresectionstakenat this locationis too small to draw firm conclusions.
In thisareaalsoa minimum
of suspended
matterhasoftenbeenobserved,
mostlyduring
wintercruises(Visseret al, 1991).
1 METER BELOW SEASURFACE
..
10 METER BELOW SEASURFACE
Figure
6: Average
salinity
fields(over8 cruises)
andtheirstandard
deviation
(ino/oo)at lm and10m
belowtheseasurface.
Clearly,variability
intheupperlayerismuchlarger
thanin thelowerparts
of
the water column.

The above three-dimensional picture of the salinity field is further confuted by its
horizontaldistributionat a few selecteddepths(fig. 6). However,thesemeanfields don't
appearexcitingas manyof the interestingfeatureshavebeenaveragedout (thoughthe fields
of standarddeviationindicatethat there'smore goingon).
A muchmore revealingpictureappearsif the averagesalinityfieldsof upperand lower

layers are subtracted. What emergesis a figure showingthe horizontaldistributionof
stratificationin the coastalzone(Fig. 7). The buoyantplumespreadingfrom thefiver mouth
standsout clearly. Northof Noordwijk(i.e. 40km northof the source)theplumeaxisseems
to splitinto two branches.This may indicatetwo favouritespreadingmodes.In the first one
theplumereattaches andflowsnorthwardasa narrowcoastaljet within 10kmfrom the shore.
In the othermodeit eitherdoesn'treattachor, while on the way northward,it separates
from
the coast,due to windsfavouringoffshoreupperlayer transport.
It is temptingto try to interpretthe whole figure in termsof horizontaladvectionand

spreading.However,as described earlier,cross-shore circulationin the verticalplanemay
play an equallyimportantrole. The bandof relativelylargestratification around15kmfrom
the shoremay alsoindicatethe mainpositionof a front,separating the regionof majorfiver
influencefrom the opensea.
Figure7: Horizontaldistributionof meansalinitystratification,

definedas the differencebetweenthe
averagesalinitiesat lm and10mbelowtheseasurface.It suggests thatriverspreading takesplaceboth
alongthe coastand offshorein a bandcentredbetween10 and 15km from the coast. In the northern
partof theregionthisbandof relativelylargestratification
mayalsoberelatedto thefrequentexistence
of an offshorefront (compareFig. 5d).

54O W P M de Ruijter, A van der GiessenandF C Groenendijk
After its debouchmentinto the North Sea some fiver water is transportedsouthward,
particularlyin caseof northerlies.Consequently, in the Goeree-section
some20km southof
the RotterdamWaterway(Fig. 2) the influenceof the outflowingfiver is still significantup
to 15kmoffshore.Variabilityis of thesameorderof magnitude asin theNoordwijk-transect.
Verticaldensityinhomogeneities are smallerdueto the shallowness
of thecoastalzonenear
Goeree.Underpersistent windsfrom thenorthRhinewatermayevenreachthemouthof the
ecologicallyvulnerableEasternScheldt.In suchan extremesituationcombinedwith a high
fiver dischargea fiver water fractionof about25% wasobservedat this location,which is
situatednearly40km southof the Haringvliet(De Ruijteret al., 1991).
In view of the large variability of the densitystructureit is useful to considera few

characteristic
individualsituations.Somedistinctpatterns,representingperiodsof high,mean
and low fiver discharge,respectively,are presentedfor the transectsnear Ter Heijde and
Noordwijk. Figs.8-10 showthe salinityfieldsduringsuchsituations.
• •0 Thi5 Thio lhO? • Thi.5
-5 _
.5
-10 _
33.5
-15 _
33.5
-20 _
TerHeyde 870213
0 15 lO 5
-25
Distance to coast
• 11115 III10 1•07 11104 Illi.5
-5
2.5
-t0
-15
32.5
-20
Noordwi jk 870213
5 20 t5 1o 5
-25
Distance to coast
Figure 8: Salinitydistributions
in the Ter Heijde (a) and NoordwijkCo)transectsat averageriver
discharge
(2300m3/s)andwindswithaneasterly
component.
Dateof survey:
February
13,1987

541
andF C Groenendijk
In the situation
with dischargecloseto averagein theTer Heijde-transect
up to 20km
offshorestratification
isstrong.Mixingprocesseshavesubstantially
reducedthestratification
in theNoordwijk-sextion,whereit appearsmainlyin a zonebetween 5 and15kmoffshore.
Fig. 8 alsoshowssomeearliermentioned phenomena observedin thesetmnsects.For
instance,nearTerHeijdethepattern of isohalines
indicatesanestuarine-like
circulation,
with
lesssalinewaterdrivenoff-shoreandopenseawatertransported on-shore.Ca. 4km from
shorea frontnearthebottom(TerHeijde)canbeobserved. Northerlies
andoff-shorewinds
maystimulate
the occurrence
of thesephenomena.
ß . . ß
• TerHeyde 860i30
5 20 ',5 f.O 5
-25
OJstance to coast
NI• 14M20 Nllt5 NlltO NM07 NM04 Nllt.5
-2O
Noordwijk 860130
20 t5 •0 5
-25
OJstance to coast [km)
Figure
9:AsinFig.8,nowatabove
normal
discharge
conditions
(6000m3/s).
Dateofsurvey
January
30, 1986
Extreme
fiverdischarge
conditions
leadto substantial
deviations
fromtheaverage
salinity
pattern.A salinity
distribution
during
a peakdischarge
(:!:6(X}0m3/s)
is depicted
in Fig.9.
Stratification
is extremenow, throughout
the coastalzone. Comparedto the Ter Heijde
transect
onemaynoticethedeepening of theupperwindmixedlayerin thestratified
system
nearNoordwijk,30km furthernorthward.It showsthe impactof strongwinds(5 Bf)
occurring
duringthisspecific
period.However,thereareindications
thatalsoataverage wind
conditions
thebuoyantspreadingandmixingof thefiverwaterplumeis affectedstronglyby

542
W P M deRuijter,
A vanderGiessen
andF C Groenendijk
wind-induced
stirringof theupperlayer.
lb20 lh15 fido lh07 Th04 lhl.5
-5
•.5 •. 3t. 29.
•
-tO
-t5
.
•.5i• 3t
•.5 5t5 t5 .5 29 . . 3. .5 ß
-20
ß
TerHeyde 871110
20 t5 tO 5
-25
Distance to coast
tk25 • •i5 •iO Ik07 III104 •i.5
-5 __
-10 __
-t5 __
-20 __
Noordwiik 871110
.'5 20 t5 tO 5
-25 ,
Distance to coast
Figure
10:AsinFig.8,nowatrelatively
lowRhine
river
discharge
(1400m3/s).
Date ofsurvey
November 10, 1987
During
periods
oflowfresh
water
discharge,
stratification
isalmost
absent
(Fig.10)except
in a small
areaof theTerHeijde-section,
closetothemajor fiveroutlet.Nevertheless,
significant
horizontal
densitygradients
stillexist.
Inthelong-term
average
situation
these
lateral
density
gradients
decrease
withincreasing
distance
fromshore.
However,
ina specific
casetheactual
meteorological
andhydraulic
conditions
may
disturb
this
pattern
considerably.
Therefore
anyindividual
density
pattern
may
deviatesignificantly
fromtheaverage
situation.
26.4 The ObservedResidual Current Field
Currents
weremeasured
continuously
in thetransect
nearNoordwijk
fromAugust
1985
through
September
1986.(see
Fig.11forlocation
ofmoorings
and'positioning
ofcurrent

543
andF C Groenendijk
meters
in thewatercolumn.)Someremarks
aboutthetidalcurrents
in thisregionweregiven
in section 26.2. Here we will focus on the residual currents.
NOORDWlJ K-transect
30km 20 10 5 3
I
(
5m-
10-
15-
20-
¸ Current
meter
::: if::
• i I--IPressure
recorder
Figure11:Location
ofcurrent
metersduring
theone-yearobservational
period
inthecross-shore
transect
nearNoordwijk.
Currentmeterswererobust
self-recording
Flachsees.
Every10minutes
theyregistered
10 minute-averaged
velocities
andinstantaneous
currentdirections
Residualcurrentsareinducedby several
physical mechanisms.
A basiccontribution
is due
to nonlinearfide-topographyinteractions
(e.g. Zimmerman,1978). Anotherdominant
componentis wind-induced.In caseof windswitha predominant
coast-parallel
component,
within severaltensof kilometresfrom the shore,the flow mainlyfollowsthe coastlineas a
geostrophic
jet in downwind
direction
ateverydepth(decreasing
in magnitude
withdepthdue
to friction).Besides
alongshore
flow,windsalsoinducea cross-shore
circulation
component,
eitherbya combination
of fictionandCoriolisforce(asanEkmanspiral)orasa driftrelated
towindsetup.In coastal
waters density
drivenresidualcross-shore
currentscanbeimportant.
For instance,
for the Irish SeaHeaps(1972) hasshownthatthe observed averageon-shore
flow neartheseabed mayresultfromthebuoyancy effectdueto thelateralvariation
in the
densityfieldwith relativelyfreshwaterinshore
andhighersalinities
offshore.The same
applies
to theDutchcoastal zonewherethehorizontal
densitycontrastis largerdueto the
highermeanfiver discharge comingfromMeuseandRhine.
Suchan estuarine-typecirculation
component is constant
in directionandthereforemay
contribute
significantly
to thelong-termdistribution
andtransportof waterproperties.
Thelongtermaverage currents
nearthebottomvaryfrom2.5 to 3.5cm/s(Fig. 12). The
maximumof i}p/3x(=2.10'4kg/m
4,Fig.4b)between5 and10kinoff-shorefromNoordwijk
corresponds
with a maximumin theon-shore
nearbexlflow anda relativelylargelongshore

544 W P M de Ruijter, A van der Giessenand F C Groenendijk
mean current near the surface,indicatingthat both the estuarineand geostrophiccurrent

componentsare importanthere. This may be the heart of the coastalcurrent(see section
26.3).
30km
•'•
•, /....
......>.-
Mid-column
.................• Bottom
.....
ili!
•:
Figu;'•:12: Mean currents(closeto the surface,mid-columnand near the bottom)in the Noordwijk
transect. The averagewas taken over the full periodof observation,
August1985-September 1986.
Relativelylargelongshoremeancurrentsat 5 and3km offshoreare indicativeof frequencyoccurrence
of a narrowcoastaljet.
Large variabilitywasobservedof the three-dimensional currentstructureon differenttime

scales,mainly inducedby variationsin wind conditions(Van der Geissenet al., 1990). As
an illustrationperiodscharacterizingfour differentpersistentwind situationswere selected
fromtheone-yeardataset,otherwise withcomparable hydrologicalconditionssuchasaverage
Rhine dischargeand equalmagnitudeof wind speed(of 4 h 5 Bf, Fig. 13).
To a largedegree,the three-dimensional currentstructurecanbe qualitatively understood

by combiningthe earliermentionedphysicalmechanisms: a persistent
cross-shore estuarine
circulationcomponent,longshoregeostrophic flow (relatedto the fiver inducedsalinity
gradient)northwardin the upperlayer with significantverticalshear,and varyingwind-
inducedcurrents (e.g.Csanady,1982). Southwesterlies reinforcethelongshore jet by adding
a barotropiccomponentto it (averagesurfacevelocityof +14cm/s),northerliesdo thereverse
to the longshoreflow and reinforcethe cross-shore
circulationcell. The effect of the Ekman

W P M de Ruijter, A van der Giessenand F C Groenendijk 545
spiralsdrivenby cross-shore
windsis alsoclear from the observations
(Fig. 13).
Consequently, duringeasterliesthe currentdirectionveerssignificantlywith depth;for

southwesterliesthisis onlyminorbutthentheamplitudedecreases stronglywith depth. When
the wind hasa componentblowingfrom the northa smallverticalvariationof residualcurrent
speedwas observed.
• DWIJK
..'......DWIJK
DWIJK
5cm/s
:,, Surface
........ ,..- Mid-column
........................
-• Bottom
Figure13:Residual
currents
20kmoffshore
fromNoordwijkunderpersistent
windsblowingfromfour
distinct sectors

546 W P M de Ruijter, A van der Giessenand F C Groenendijk
Thesequalitativecurrentprofilescan be rathereasilyreproducedusinga diagnostic(i.e.

giventheobserveddensityfield) linearmodelinvolvingtheabovegivenprocesses. However,
correctcurrentvelocitiescan only be obtainedby imposinga longshoreseasurfaceslopeof
a few cm per 100km,(Groenendijk,1988). Apparently,due to the wedgeshapedgeometry
of theSouthern NorthSea,with onlya smallgapat theDoverStrait,themomentum inputby
the wind can be partly balancedby wind set-upagainsta southernboundary(seealso De
Ruijter et al., 1991).
• I
o 2Okm
3o
E-bottom
26
27 28 29 30
27 D-surface
26
JANUARY 1986
winds
Figure14: Progressive
velocityvectordiagrams,nearthe surfaceat 10km(D) andnearthe bottom(E)
at 20km offshorefrom Noordwijk. Periodof observation:January25 - January30, whenthe Rhine
discharge
waslarge(+6000m3/s)
andwindsstrong
andvariable
(shown
inthelowerpartoftheFigure).

W P M de Ruijter,A vanderGiessenandF C Groenendijk 547
Southwesterlywindsare mostcommonoverthe Dutchcoastalzone. Theseconditions

prohibitcross-shoremixingof fiverwithseawaterasthewind-anddensityinduced cross-
shoreflowcomponents counteract.
TheRhineoutflowzoneis therefore rathernarrow,with
themajortransport of fiverwatermostlytakingplacewithin15-20kinoffshore.The sl•ong
currentfluctuations
inducedbothby varyingfiver discharge andwindare illustratedclearly
fromthehighdischarge eventof January
1986. Whenthewindturnedfromnorthwesterly to
southwesterly,
longshore wind-anddensitydrivencurrentspeeds
closeto 0.Sm/sresulted(Fig.
14). Furthermore duringthisperiodandthemonthfollowingit, whentherewasstill much
fresh water in the coastalzone, all five near bottomcurrentmetersregisteredan on-shore
residualcurrentaround10cm/s(Fig. 15).
•o
CM/S'
3 KM
- 10 near bottomcross-shore
currentcomponents
-
t0
l0
10
•0
1o
'0 30
- t0
B000
6.000
• 000 Rhine discharge
2 000
t t ' APRIL
JANUARY FEBRUARY MARCH
1.o,86
Figure15:Low-pass
filteredcross-shore
currentcomponents atfivenearbottomcurrentmetersoff-shore
fromNoordwijk.Persistenton-shore
currents
existaftera periodof highRhinedischarge.Fluctuations
on time scalesof severaldaysare mainly wind-induced.

26.5 Summaryand Discussion

The averagesalinitydistributionin the Netherlandscoastalzoneshowsthe outflowof Rhine
fiver waterto takeplacemainlynorthward withinan areawithoffshorewidthshrinkingfrom
20-25kmnearthe mouthto 10-15kinsome100kinnorthward.Mostlythelesssalinelayeris
confinedto the upper4-8m of the water column. Up to 100kin north of the Rotterdam
Waterwaystratificationis observedfrequentlyin the outflowzone.
The horizontaldistribution
of meanstratification (Fig. 7) suggests
two favouritemodesof
fiverplumespreading.In bothcasestheoutflowjet separates fromthefivermouthandturns
to thefight influencedby earthrotation.In thisfirst modetheplumereattaches to the coast
(probablybetweenTer HeijdeandNoordwijk),to forma narrow(ñ10km)coastal jet. In the
othermodereattachment doesn'ttakeplace,or whilemovingnorthward, separationfromthe
coastis inducedby windsfavouringoffshoreupperlayertransport (particularly
windsbetween
North and Eas0. In the mean field a bandof relativelystrongstratificationis observed
between10 and 15kmoffshore.Thisbandmay alsobe relatedto the frequentexistence of
an offshore front.
The salinityfieldshowsa largevariability,especially

in theupperpartof thewatercolumn
in anareabetweenGoereeandNoordwijk.Mostactualdensitydistributions occurring
in this
regiondeviatesigndicanflyfromthemean.Windstrongly aff•ts thespreading
characteristics
of the fiver plumeboth by the wind-drivencurrentfield and verticalturbulentmixing.
Obviously,the time-varyingdischarge is anothercrucialparameter.
The largeobservedvariabilityof thecurrentfieldis mainlywind-induced.If strongwinds

anda highfiver dischargecoincidethena distincttwo-layersystemcandevelop.Non-tidal
currentspeedsof over40cm/sweremeasured duringsuchperiods.Averaged overa full year
of observations
theresidualcurrents in a cross-shore
transect
nearNoordwijkranged between
7 and1l cm/sin theupperlayer. An estuarine circulation
component showed dominantly near
thebottom, wherethesub-tidal
current wasalmost permanentlydir•ted on-shore,withspeeds
up to 10cm/sunderfavourable densityandwind conditions.One yearaveragevaluesare
between 2.5 and 3.5cm/s.
Thelargevariabilitynecessitates
anelaborate
analysis
of individual
density
patterns
(fronts,
stratification),
characterizing
a widevarietyof windandfiverdischarges
conditions.
Extensive
fieldstudies,
including
repeated
hydrographic
sections
andlongtermobservations
from mooredarrays,are indispensable for the development and validationof 3-D
hydrodynamic anddispersion
models.Theresultsreported
hereindicate
thatemphasisshould
beona 25kmwidezonebetweenGoereeandNoordwijk(i.e. some40kmnorthof theoutflow
port). Longshoreresolutionof the observational
networkhasto be muchfiner thanwas used
forthisstudy.(Recently,
aspartof thePROFILE-project
(Processesin Regionsof Freshwater
Influence)
withintheEuropean
Community MAST-program a collaborafive
efforthasstarted
to fill in this gap).

All of the above showsthat for mostapplicationsa treatmentof the Dutch coastalzone
usinglong-termmeanhydrographic parameterscanleadto totallyerroneous
results.The same
holdsif verticallyintegratedinsteadof three-dimensional
currents(and models)are applied.
Acknowledgement
We gratefullyacknowledge themanypeoplewhocontributed to thecompletionof thisstudy,
in particularthe crew from the RV "Holland",the personnelfrom the North SeaDirectorate,
The Tidal Watersdivisionandfrom the formerdivisionHellevoetsluisof Rijkswaterstaatwho
organisedand assistedin the data collectionand analysis.
26.6 References
Csanady,G.T., 1984: Circulationin the coastalocean. D. Reidel PublishingCompany,Dordrecht/

Boston/Lancaster,
279pp.
De Ruijter,W.P.M., J.C. BorstandJ.S.L.J.van Alphen,1991: Spreadingof Rhineriver water at high

dischargeand strongwinds. Subumitted.
Gatvine,R.W., 1987: Estuaryplumesand frontsin shelfwaters. J. Phys.Oceanog.,17, 1877-1896.
Groenendijk,F.C., 1988:Residualcurrentstructurein the Netherlandscoastalzone. MastersThesis(in

Dutch). Univ. of Utrecht, Inst. of Meteorologyand Oceanography.ReportIMOU V-88-6, 94pp.
Heaps,N.S., 1972:Estimationof densitycurrentsin theLiverpoolBay areaof the Irish Sea. Geophys.

J.R. Astron. Sot., 30, 415-432.
McClimans,T.A., 1988:Estuarinefrontsandriverplumes.In: Physicalprocesses,

In: Physicalprocesses
in estuaries,
J. DronkersandW. van Leusden(eds.),SpringerVerlag,55-70.
Prandle, D. 1978: Monthly-mean residual flows through the Dover Strait, 1949-1972.
J.Mar.Biol.Ass.UK, 58, 956-973.
Van Alphen,J.S.L.J.,W.P.M. de RuijterandJ.C. Borst,1988:Outflowandthre-dimensional spreading

of Rhineriver water in the Netherlands
coastalzone. In: Physicalprocesses
in estuaries,J. Dronkers
andW. van Leussen(eds.),SpringerVerlag, 70-92.
Van der Giessen,A., W.P.M. de RuijterandJ.C. Borst,1990:Three-dimensional

currentstructurein the
Dutch coastalzone. Neth.J.SeaRes., 25, (1.2), 45-55.

55O W P M de Ruijter, A van der GiessenandF C Groenendijk
Visser,M., W.P.M. de Ruijter andL. Postma,1991' The distribution

of suspended
matterin theDutch
coastal zone. Submitted.
Zimmerman,J.TJ•., Topographicgenerationof residualcirculationby oscillatory(tidal) currents.

Geophys.Astrophys.Fluid Dyn., 11, 35-47.

27
On the distribution of suspendedmatter

and the density driven circulation in the
Dutch coastal area
M Visser
Abstract
Especiallyin winter, observationsof the near-surface

suspended matterdistributionin the
Dutchcoastalzoneshowa suspended matterminimumlocatedbetweensalinitymaximumand
the coast. This minimum is often presentin an area with relatively strong(cross-shore)
densitychanges,relatedto Rhine fiver outflow.
A simplemodel of cross-shore densitydrivencirculationshowsthe possibleinfluenceof

a densityfront on the transportand sedimentation of suspended matter. Particleswith a
settlingrate comparableto the verticalvelocityscaleof the circulationare forcedto move
off-shoreuntil theyreachan areawheretheycansinkto thebottomlayer.In thebottomlayer
theyare transported onshoreagain. Wind conditionsare importantin modifyingthe strength
andpositionof a suspended matterminimum.
27.1 Introduction
The SouthernBight of the North Seais exposedto high inputsof contaminants, especially
through rivers and dumping activities. Part of those contaminantsare dissolvedand
transportedwith the (residual)circulation,othersinteractwith suspended
particles(silt,clay).
Thereforethe distributionand transportmechanisms of suspendedmattermay give more
insightinto the fate of pollutants.
From1975to 1983a waterqualitymonitoring programhastakenplacein theDutchcoastal

area(Fig. 1). Duringthisprogramsalinity,temperature, suspendedmatteranda rangeof
otherwaterqualityparameters weremeasured.The suspended matterpatternsshowa clear
seasonalvariability(Visseret al., 1991). In wintera localconcentration
minimumis present
betweensalinitymaximumand the Dutchcoast. A similarminimumwas foundby Lee &

552 M Visser
Folkard(1969)andPostma
(1981).Postma
(1981)assumed
itsposition
to coincide
withthe
seawardendof a densitydrivencirculation.
1
0 tOO &, m
SEA
,!!iiii!iii!!!i
B E LG I U
Figure
1:SouthernBight
of theNorth
Seawithisobaths
andinvestigated
area.The76measuring
locations
aredistributed
overtencross-transects
along
theDutch
coast.
Thelocations
areindicated
by
thefirstletter(s)
of thetransect
andtheirdistance
fromtheshore(km). Fromsouth
to norththe
transects
are:Appelzak
(A),Walcheren
(W),Schouwen
(S),Ooeree
(O),TerHeijde
CTH),
Noordwijk
(N),Egmond
(E),Callantsoog
(C),Terschelling
(Ts)andRotmmerplaat
(R).
In thispaper
westudy
thepossibleinfluence
ofthe(cross-shore)
density
drivencirculation
on thesuspendedmatterdistribution
usinga simpleanalytical
model.In section
27.2a
description
of theDutch
coastal
zoneisgiven.Section
27.3shows
observed
salinity
and

M Visser 553
suspended matterdistributions.A simplemodelof the densitydrivencirculationand the

associated
(advective)transportof suspended
matteris presented
in section27.4. In section
27.5 we summarizeanddiscusstheresultsof thisstudy.
29
Figure2: Averagesalinitydistribution
(O/DO)
overthefull period(1975- 1983).The freshfiverwater
is spreadnorth-eastward
alongthe Dutchcoast.

554 M Visser
27.2 Dutch Coastal Zone
The Dutch coastalzone is locatedin a shallowpart of the southernNorth Sea with water
depthslessthan30 m (Fig. 1). The greaterpartof thewatermassoriginatesfromtheEnglish
Channel.A smallpartentersthecoastalzonethroughthecontinental rivers,in particularthe
Rhine.
Spatialdensityvariations aremainlydueto variationsin salinity;temperature

variations
are
of lessimportance.Over a largeportionof the areathe watercolumnis stratified.Strong
stratificationoccursin the immediatevicinity of the outtlowport. The three-dimensional
densitystructurein the Ter Heijde (TH) transect(ca. 10 km northof the main outletof the
Rhine-Meuseestuary)is still to a greatextentdetermined
by the initial mixingandoutflow
processesof fiver water. Going to the north stratificationdiminishes,but even in the
Callantsoog(C) transect,locatedsome100 km northof the fiver mouth,weak stratification
is oftenobserved(De Ruijteret al., this volume). Cross-shorethe densitygradientis still
very strong.
Tidalcurrents arestrongandnearlyparallelto theshore.On timescales

of daysto weeks
thesub-tidalcirculation
is predominantly
wind-driven.Thedominantsouth-westerlies
(Visser,
1987)drivelong-shore currents
in northeastward
direction
withmagnitudesof theorderof 104
ms4 in theDutchcoastal
zone.Changing
windconditions
cause
considerable
deviations
from
the averagecirculation(Van der Giessenet al., 1990).
TheaverageRhine-Meuse dischargeis + 2500m3s4. It variesstrongly

withanassociated
timevaryingdensitydistribution
in thecoastalzone.Remarkably, theposition
of thesalinity
maximum is almost constant at the axis of Channel water inflow some 75 km off-shore
(Visser,personal
communication).
Theeightyearaveragesurfacesalinitydistribution
shows
thefiver waterspreading
northeastward
alongtheDutchcoast(Fig 2).
27.3 Observations
FromMay 1975to February1983everytwoweeksa surveywasmadeby theR.V. "Holland".

Duringeachcruisenear-surface watersamplesweretakenat 76 locations, distributed overten
cross-transects (70 km wide)alongtheDutchcoast(Fig. 1). Thelocations areindicated by
thefirst letter(s)of the transectandtheirdistancefromthe shore0an). Fromsouthto north
thetransects are:Appelzak(A: 1, 2, 4, 10, 20, 30, 50 and70 km off-shore), Walcheren (W:
1, 2, 4, 10, 20, 30, 50 and70 km off-shore),$chouwen(S: 1, 4, 10, 20, 30, 50 and 70 km
off-shore), Goeree(G: 6, 10, 20, 30, 50 and70 km off-shore), Ter Heijde(TH: 1, 2, 4, 10,
20, 30, 50 and70 km off-shore), Noordwijk(N: 1, 2, 4, 10,20, 30, 50 and70 km off-shore),
Egmond(E: 1, 2, 4, 10, 20, 30, 50 and70 km off-shore), Callantsoog(C: 1, 2, 4, 10,20, 30,
50 and 70 km off-shore),Terschelling (Ts: 4, 10, 20, 30, 50 and 70 km off-shore)and
Rottumerplaat (R: 3, 5, 10, 15, 20, 30, 50, 70 and 100 km off-shore).At everylocation

M Visser 555
salinity,
temperature,
suspended
matter
concentration
andarange
ofwater
quality
parameters
weremeasured.
Thesuspended
matter
concentration
wasdetermined
froma water
sample
of
1 lilxethatwasf'fitered
witha 0.45pmcellulose-acetate
filter. Theresidue
waswashed
with
demineralized
watertoremove
thesalt.Afterit wasdriedata temperature
of 105øC,
the
residue
wascooledandweighed.
Nodistinction wasmade betweenorganic
andinorganic
suspendedmatter.
lO %
10
0 •0gm
Figure
3: Average
suspended
matter
distribution
(mgl
-•)over
thefullperiod
(1975
- 1983).High
concentrations
aremeasured
infront
ofthecoast.
Furtheroff-shore
theturbidity
is muchsmaller
and
morehorizontally
homogeneous.
Howevera weaklocalminimumispresent
atsome 25kmoff-shore
northof Noordwijk.

556 M Visser
Theaverage(overthefullperiod) suspended matter

distribution
(Fig.3) shows
a verysteep
concentration
gradient in a narrowstripalongthecoast,
saywithin5 - 10km. In thisstudy
ouremphasiswill be ontheareaoff-shore fromthiszone,wherethewateris moresaline.
Theretheturbidity
is muchsmaller andmorehorizontally homogeneous. However,eventhe
averagesuspendedmatter pattern
shows a weakminimum northofNoordwijk between20and
30 km off-shore.
5 ß
..4
3 ,50
,
,/50
10'
6 *
20;/
5O
IO0
0 50 km
Figure
4: Average
suspended
matter
distribution
(mgl
4) over12January
cruises
(1976- 1983).A clear
minimumis presentat some25 km off-shore.

M Visser 557
Especially
in wintera pronounced
suspended
matterminimumis present.The average
Januarypattern (Fig. 4) shows a clear minimum at some 25 km off-shore. Individual
measurements oftenshowa localminimumin whichabruptconcentration
changes
occur.
Thesefrontalchangesoftencoincidewith a largedensitygradient.
/
/
/
/
,/
/
-".31
/
/
34 /
,/
25
35
25
33
o 50kin
Figure
5a: Salinity
in o/oo(a)andsuspended
matter
in mg1-1
Co)measured
during
thesurvey
of 5 - 8
March1979. A salinityfrontis coinciding
witha suspended
matterminimum.

558 M Visser
2x, , ./ .--" ,5 7
' •" ..................
•.10<?
10,',\ ; / ••:• /' I
1
34' '
/
2
10
ß
lOO
o 5okm
Figure5b: Salinityin o/oo(a) andsuspended

matterin mg1-1
(b) measured
duringthesurvey
of 5 - 8
March 1979. A salinityfront is coincidingwith a suspended
matterminimum.
Fig. 5 showsthe salinityand suspended matterdistribution,observedduringa cruisefrom

5 to 8 March 1979. In this characteristic winter pattem a narrow strip of low suspended
matterconcentration is presentparallelto the Dutchcoastsome25 km off-shore(minimum
concentration
lessthan1.0mgl4 at 30km off-shore
in theNoordwijktransect,
Fig.5b).High
suspendedmatterconcentrations
are foundin the southernWansects, indicatingthe influence
of the Hemish Banks. The salinitydistribution(Fig. 5a) showsa densityfront near the
suspendedmatterminimum(Noordwijktransect,salinityat 20 km off-shore:28.3 o/oo,at 30

M Visser 559
km: 34.1 o/oo).
In the absenceof large sourcesin the Dutchcoastalzone the globaldistribution of

suspended matter(highconcentrations in winter;lowconcentrations in summer) is expected
to be largelydeterminedby theboundary.Particularly seasonally varyingfluxesthrough the
EnglishChannel areimportant
(Visseret al., 1991). Ltx:allyfor theexistenceof a minimum
in suspended matterconcentration(between salinitymaximum andshore)a largeenough fiver
discharge is necessary.In combination with a favourablewind (long-shore, southwest)a
salinityfrontcandevelop parallelto thecoast.If theassociateddensitydrivencirculationis
strongenough, thecross-shore
component maycausea depletion of suspendedmatterin the
neighbourhood
of this front.
(0, y, •(0))
(0, y, 0)
/I
/
! I
! I
! I
! I
! I
!
(0, y,-h)
! !
! !
! !
! !
! !
!
! ........:.........
! !
! !
! !
! !
! .... ::::::::::::::::::::::::::
! .................
! !
! !
! ! .... :.:.:.:.:..........................
! ..:.:.:.::.:.:.:.:.:.:.:.:.:.:.:.:
ß
!
/ !
z = ((x)
• • z y
I !
I !
I !
I !
z=-h I• (o, o,-h)
(x0,0,-h) x
Figure 6: Model set up: x-axiscross-shore positiveeastward;y-axis long-shorepositivenorthward;

z-axisverticalpositiveupward;surfaceat z = •(x); bottomat z = -h (uniformwaterdepth);coastat x
--0.

560 M Visser
1028.0 i I I I I I I 3.5e-04
-3.0e-04
1027.0
- 2.5e-04
- 2.0e-04
1026.0
- 1.5e-04
1025.0
1024.0
1023.0
0
1028.0 3.5e-04
-
- 3.0e-04
.
1027.0
- 2.5e-04
.
- 2.0e-04
1026.0
- 1.5e-04
1025.0 - 1.Oe-04
p-gradient
- 5.0e-05
1024.0
- O.Oe+00
.
1023.0 I I I I I I I , -5.0e-05
0 5 10 15 20 25 30 35 40
Figure7: (a):Average densityin kgm-3(.... ) anddensitygradient

(- - -) over12January
cruises
in the
Callantsoogtransect in kgm-3(alonghorizontal axisdistancefromtheshorein km).(b):Prescribed
density
field(..... ), p(x)= 1028.0- 4.0exp{ -(x/-xo) n }, withXo= 2 104m anddensitygradient
(- - -).

M Visser 561
27.4 Modelling
To illustratethe possibleeffect of a densityfront we usea linear, stationarymodel,with a
coordinatesystemas shownin Fig. 6. In the modelwe describethe advectivetransportdue
to the densitydrivencirculation. In the northernpart of the Dutchcoastalzone (northof
Noordwijk) long-shoreand vertical density changesare inferior to the still very large
cross-shore gradients. In this part of the area a local suspended matter minimum often
occurs. Thereforelong-shoredensitychangesare not takeninto accountin the model.
Furtherwe assumethe densitydistributionto be stationary.The freshwaterinflow from

theRhine-Meuseestuaryis supposed to maintaina constantcross-shore
densitygradient.This
impliesthatthe velocityfield is only dependingon the cross-shore
x-coordinateand on the
verticalz-coordinate.To simulatea relativelystrongdensitygradientat x = -Xowe choose:
(27.1)
in whichPo= 1028kgm'3thestandard
density
of seawater;Pdthedensity
difference
between
coastaland sea water; and Xoa cross-shore lengthscale,approximatelythe positionof the
largestdensitygradient. This densityfield is in reasonable
agreement with measurements of
the densityprofile in the CallantsoogWansectduringJanuary(Fig. 7).
The pressurefield p(x,z) simplyfollowsfrom the hydrostaticbalance:
Oz
g (27.2)
The horizontalvelocitiesu(x,z) and v(x,z) satisfythe momentumequations:
10p f v- ka2u (27.3)

Po ax azz
f u=kaZv (27.4)
Oz2
wheref = 10'• s'• is theCoriolis

parameter;
g = 9.8 ms'2thegravitational
acceleration
andk
an eddyviscosity.The verticalvelocityw(x,z) followsfrom the continuityequation,which
reduces to:
• + ' 0 (27.5)
ax az
due to the fact that the long-shorevelocitycomponentv is independent

of the y-coordinate.

562 M Visser
This providedthat we canintroducea 'stream'function•(x,z) definedby:
u ' -• • w ' • • (27.6)

az ax
The 'stream'functiondescribes thecirculationin thex-z plane. The boundaryconditionswe

have to use are (partly) determinedby the area to which we want to apply the model. For
(27.2) we haveto usecontinuityof pressure at the seasurfacez = •:
p(x,•) - 0 (27.7)
where• = •(x) is the surfaceelevation.Furtherwe havethe kinematicconditionat the

surface:
u a(
• I
w (27.8)
•x
At the bottom z = -h the kinematic condition is:
w(x,-h) - 0 (27.9)
At the coastalboundaryx = 0 we havethe conditionof no net transportthroughthe coast:
f_:
su(x,z)dz' 0 (27.10)
Properlywe haveto requirethatthevelocitycomponent normalto thecoastvanishes.But
in realitythebottomz = - h(x) depends
on thecross-shorecoordinate.Approaching thecoast
the depthdiminishes and is zeroat the coast. In thatcasewe haveto usethe boundary
conditionthatthevelocitycomponent normalto thebottomvanishes andthatu(0) = 0. In this
simplemodel,in whichwe usean uniformdepthh, we haveto usea transport conditionon
the coastlinex = 0. This impliesthat we allow that sUeamlines intersectthe coast. We
assumethatthereexistsa 'small'boundary layerin whichthedepthdiminishes to zero. The
coastalboundary conditionprovidesthatthereis nonettransport
intoor outof thisboundary
layer.
Depthintegrationof thecontinuityequation(27.5)andapplyingboundary conditions

(27.8),
(27.9) and(27.10) givesthat thereis no net transportin cross-shore
directionoverthe whole
area. Depthintegrating(27.4) showsthatbottomandsurfaceslresses in long-shore
direction
have to be equal:

M Visser 563
z- ½ z- - k
This is a directconsequenceof the fact that we neglectthe (small) densityvariationsin

long-shore direction.Cross-shore
bottomandsurfacestresses do not haveto be equal.
At the bottom we use a stress condition:
[/1_••]
•--•,. Po
x•
k ' [0•]
•--t,. Po
x•
k (27.12
with thebottomstressproportional
to thebottomvelocity(Csanady,1982):
x• - poL u(x,-h), 't•- poL v(x,-h) (27.13)
Furtherwe apply a wind stressconditionat the surface:
(27.14)
ß •-½ Po
k' •.½ Po
k
The wind stressis givenby
x W• :• . Wy (27.15)
with W = (W •, W y) the wind vector.
For W we choose:
•(x,-h) - 0 (27.16)
Representative
parameter
values
arek = 0.01m:s'• (Bowden,
1983),Xo= 2 10• m,h = 20 m,
L = 10'3ms'• (Csanady,
1982)andp(x)prescribed
by(1) withPoTM
4.0kgm
'3. Withthese
boundaryconditionswe candeterminethehorizontalvelocitiesu andv, the 'stream'function
• andthesurface
slope• (theformulae
aregivenin Appendix
A).

564 M Visser
I I , I
-2
-/4_
-6_
-8_
-1/4_
-16_
-18
-2O
-35 -30 -25 -20 -15 -10
X (' 1000 m) •
Figure8: StreamfunctionW(x,z) in absence

of wind: off-shoremotionin the surfacelayer,onshore
motionnear the bottom. The circulationis restrictedto the areawherethe densitygradientdiffersfrom
zero. This is a projectionon the x-z plane. The long-shore velocity componentv(x,z) is not included
in this picture. The axesare definedin Fig. 6.
In absenceof wind the associated'stream'functionis depictedin Fig. 8. As expectedthere

is off-shoremotion in the upper layer and onshoremotionnear the bottom. The circulation
is restrictedto theareawherethedensitygradientdiffersfromzero. To illustratethepossible
effect on the sedimentationof suspendedmatteran equationfor the sedimentconcentration
is added:
u(x,z)
OCr
ax +v(x,z)
i]C+[w(x,z)
3-Y'y +W,]i]C
az . 0 (27.17)
describingadvectivetransportof suspended matterwith constantsettlingrateW,. Becauseof
the fact that we only look at advectivetransportwe can determinethe characteristic curves
alongwhichthe suspended matteris transported.The setof characteristiccurvesis givenby
the intersectionof two familiesof surfaces(seee.g. Kevorkian,1990):

M Visser 565
,t,(x ,z) + W,x - cnstant (27.18a)
(27.18b)
h v(x,•)d• - y [w(x,z)+ W,]" cxmstaat
In absenceof wind, the projectionof the trajectorieson the x-z plane is given in Fig. 9.
For differentsettlingvelocitiesthe patternsshowconsiderable differences.Obviously,if the
verticalvelocitycomponent w is muchsmallerthanthe settlingvelocityall theparticleswill
immediatelysink to the bottom(Fig. 9a). If the verticalvelocitycomponentis muchlarger
thanthe settlingrate the particleswill follow the residualcirculationand hardlysettledown
(Fig. 9c).
If settlingvelocity and vertical velocity componentare of the same order a complex

suspended mattercirculationpatternexists(Fig. 9b). Thereis an areain whichthe trajectories
do not intersectthe bottom,in this area suspended matterparticleswill stay in suspension.
This areais comparable to the 'regionof retention'asdescribedby Stommel(1949) in a study
aboutsmallparticlessinkingthroughconvection cells. Suspendedmatterparticlespresentin
the surfacelayer (outsidethe 'regionof retention')are forcedto movein off-shoredirection
until they reach the area where the vertical velocity can not compensatethe settlingrate
anymore. This happensat the seawardsideof the densityfront, wherethe particlessink to
the bottomlayer in whichtheyare transported in the onshoredirection. The settlingrate of
a particledependson its size amongotherthings. Most suspended particlesare finer than
30gm (Eismaet al., 1980). Settlingvelocitiesof suchfine grainedparticlesare of the order
of 10'5 ms'• (Meade,1972). Murray(1970) has shownthat the averagefall velocity
determinedby experimentin various turbulent fields is reduced by 30% below the
corresponding
still-waterterminalfall velocity.
Only particleswith a settlingrate comparableto the magnitudeof the verticalvelocity

contributeto the formingof a suspendedmatterminimum. To investigatethe influenceof
winddirection
andvelocitywe choose
in thefollowingW, = - 10'5ms'•. Thepatterns
for
different wind directionsare given in Fig. 10. With a northerly(long-shore)wind the
associated Ekman transportis off-shoredirected. The suspended mattertransportpathsare
stretchedin the off-shoredirection. Particlespresentin the surfacelayer are forcedto move
furtheroff-shorebefore they can sink, than in the caseof no wind (Fig. 9). Southerly
(long-shore)windsleadto an onshoredirectedEkmantransport.Materialpresentfar off-shore
in the surfacelayer is forcedto movein the onshoredirectiondue to the wind, while material
presentin thesurfacelayerof thecoastalareais forcedto moveoff-shoredueto thedensity
gradient.

566 M Visser
Fig.9a
-35 -30 -25 -20 -15 -10 -$
X (' 1000 m)
0 I .... I , , • • I • • • • I , • • • I , , , • I • • I • , • , I , , ,
Fig.9b
_
-i
-1
-35 -30 -25 -20 -15 -10 -5 0
X (* 1000 m) ;
Fig.9c
-•6•
-20
-35 -30 -25 -20 -15 -10 -5 0
X(' 1000 m) •.
Figure9: Characteristic
curves
of suspendedmatter(projection
onthex-zplane)withdifferent
settling
velocities
(a) W, =- 10.4ms-';(b)W, =- 104 ms-';(c)W, =- 10-• ms-'.

M Visser 567
Fig. 10 a
X (' 1000m)
-3% -30 -25 -20 -15 -10 -5 0
X (' 1000 m) •.
curvesof suspended
matterwithdifferentwinddirections
anda windspeed
of 5 ms4 (projection
onthex-z plane).a:northerly
(long-shore)
wind,b: easterly
(off-shore)
wind,
c: southerly(long-shore)wind, d: westerly(onshore)wind.

568 M Visser
-4_-
Fig.10c
•2
14
•6
•8
20
-35 -30 -25 -20 -IS -•0 -5
X (* 1000 m) --•
3b 30 -25 20 -1S -!0 -5
X (' 1000 m) --•
curves
of suspended
matterwithdifferent
winddirections
anda windspeed
of 5 ms-1(projection
onthex-z plane).a:northerly
(long-shore)
wind,b:easterly
(off-shore)
wind,
c: southerly(long-shore)
wind, d: westerly(onshore)wind.

M Visser 569
The verticalstructureof the wind-drivencirculationshowsan Ekman spiralwith surface

currentsin a 45ø directionto the fight of the wind. Thereforeeasterly(off-shore)windsgive
rise to an off-shoresurfacevelocity componentand in this way the transportpattern is
comparableto the patternwith a northerlywind. In the otherway, westerly(onshore)winds
give rise to an onshoresurfacevelocitycomponentand the transportpatternof suspended
matteris comparableto the caseof southerlywind. The wind-drivencirculationis addedto
the densitydrivencirculation,thereforethe suspended matterminimumis morepronounced
in circumstances with westerlyor southerlywinds.
Not only the wind directionis importantbut alsoits magnitude.Fig. 11 showsthe pattern
witha southerly
windwitha magnitude
of 10 ms']. In thecaseof a windvelocityof 5 ms']
(Fig. 10c) the suspended
matterminimumis locatedat some28 km off-shore,while in the
caseof a magnitude
of 10 ms'] theminimum
is located
closerby thecoast(at some24 km
off-shore). This is due to the fact that the circulationin the surfacelayer is predominantly
forcedby the wind. The influenceof the densitydrivencirculationis smallerin the caseof
a strongerwind. Thereforethe particlespresentfar off-shorein the surfacelayer can move
furtherto the coastbeforethey sink out of the surfacelayer.
-2
-4
-8_
-12_
-16_
-18_
-2O
-25 -•0 -15 -10 -5 0
X (* 1000 m) g
curvesof suspended
matterwith a southerly
(long-shore)
wind of 10 ms-*
(projection
on the x-z plane). Comparing
to Fig. 10c(southerly
wind5 ms-*)thepositionof the
suspended
matterminimumis switched
to thecoastat some24 km off-shore(against28 km in Fig. 10c).

570 M Visser
27.5 Discussion and Conclusions
Locally for the existenceof a minimumin suspended matterconcentration

betweensalinity
maximumand the shorea largeenoughfiver dischargeis necessary.In combinationwith a
favourablelong-shoresoutherlywinda salinityfrontcandevelopparallelto the coast. ff the
associateddensitydrivencirculationis strongenough,the cross-shore
component may cause
a depletionof suspended matterin the surfacelayer.
For particleswith a settlingrate comparableto the verticalvelocityscalethereexistsan

area in which the circulationis strongenoughto keep the particlesin suspension.Particles
presentin the surfacelayerare forcedto moveoff-shorebeforetheycan sinkto the bottom
layer. The depletionof suspended matterin the surfacelayertakesplaceat the seawardside
of the densityfront. During winter observations of the near-surfacesuspended matter
disU'ibufiona suspended matterminimumis often locatedin an area with strongdensity
changes.Southerlyandwesterlywindscanstrengthen a minimum. Duringthecruiseof 5-8
March 1979 therewas a long-shoresoutherlywind. In that casethe minimumwas located
some 25 km off-shore.
In the model we describethe advectivetransportof suspended particlesonly, turbulent

mixing is not taken into accountfor reasonsof simplicity. Partly the effect of turbulent
mixing can be simulatedby taking an "effective"settlingvelocitysmallerthan the actual
settlingrate for particleswith that size (Murray, 1970). Stommel(1949) arguesthat the
'regionof retention'mustgraduallyloseits particlesdue to turbulentexchangethroughthe
boundaries.In our modelthe regionof retentionis embeddedin an asymmetriccirculation
pattern.Thereforeit is possiblethatthe exchangeof materialtakesplaceat a favourableside.
The regionof retentionis separated from theareaof trajectories
intersectingthebottomby a
separatrix,which containsa saddlenode. In a study on diffusion acrosscharacteristic
boundarieswith critical pointsMatkowskyet al. (1983) have shownthat the distributionof
exit pointsis peakedat thecriticalpoints.In ourcasethiswouldimply thattheexchange of
materialbetweentheregionof retentionandthesurrounding areatakesplacemostprobably
at thecoastalside. In thisway thedensityfrontmayactlike a kindof barrierthroughwhich
exchangebetweencoastaland off-shorewatersis hardlypossible(as proposedby Postma,
1981).
Variationof the eddyviscosityk hasits influenceon theverticalstructure of the velocity

field. The suspended mattercirculationpatternswill look slightlydifferentfor othervalues
of k. Anotheruncertainty is the magnitudeof the settlingvelocityWs,althoughthe settling
rate of the particlesin suspension is determinedto a greatextentby the magnitudeof the
verticalvelocitycomponent.In a numericalstudyon turbiditymaximain partiallymixed
estuariesFesta& Hansen(1978) concludedthat individualestuarieswill selectfrom the full
rangeof particlessuppliedto them,a smallrangefor storagein the watercolumn. This is
due to the estuarinedynamics.The samesituationoccursin our model:onlyparticleswith
a settlingratecomparable to themagnitude of theverticalvelocitycontribute to theformation
of a suspended matterminimum;largerparticleswill settledownon thebottomwhile smaller

M Visser 571
particlesremainin suspension
overthewholearea. The rangesizedistribution of suspended
mattervariesasthedynamicsof the systemchange.Storereel(1949) describes a comparable
sortingprocess.Flocculation
is notan important
process in thisarea:salinitydifferences
are
"small"andsuspended matterconcentrations
are relativelylow.
The modelpresented in thispaperis a strongsimplification of reality. The fact thatthe

pressure(and velocity) field is independent of the long-shorey-coordinateis a strong
assumption.It providesthatbottomand surfacestresses in long-shore directionare equal.
Furtherit providesthatwe canusea streamfunctionto describe thecirculation.Includingan
y dependence will makethe modelmorerealistic,but alsomorecomplicated; an analytical
solutioncan not be foundeasily. Suspended mattersources(andsinks)are alsonot included
in the model. This is donebecausethe majorsources(EnglishChannelandFlemishBanks)
are locatedoutsidethe investigated area. Becauseof the y dependence of the concentration
in the sedimentequation(27.17) it is possibleto includean inflow of suspended matterover
a southernboundary(e.g. at y = 0). Then a concentration field C(x,y,z) canbe determined
with the characteristiccurves. The wind-drivencirculationhas an importanteffect on the
suspended matterdistribution,not only by affectingthe positionand strengthof a local
minimum,butalsoby affectingthe totalsuspended matteramount. It affectsthe activityof
sourcesand the large scale (wind-driven)circulation,which are importantfactors in
determiningthe seasonalvariationsin the suspended matterdistribution(Visseret al., 1991),
althoughtheyare not includedin the modelpresentedhere. Thereforein thisform the model
is not directlyapplicableto the Dutchcoastalzone. But it showsclearlythatstrongdensity
changescan havea largeinfluenceon the formationandpositionof a local suspended matter
minimum and therefore it is useful.
Acknowledgements
I would like to thank W.P.M. de Ruijter, H.E. de Swart and D. Visser for discussionsand
information.FurtherI acknowledge the North Sea Directorateof Rijkswaterstaat (Rijswijk,
NL) for puttingthe dataat my disposal.The investigations were supported by the Working
Group on Meteorology and Physical Oceanography(MFO) with financial aid of the
NetherlandsOrganizationfor the Advancement of Research(NWO).
27.6 References
Bowden,K.F., 1983, PhysicalOceanography

of CoastalWaters,Ellis HorwoodLtd., Chichester,
pp.302
Csanady,G.T., 1982, Circulationin theCoastalOcean,D. ReidelPublishing

Company,Dordrecht,pp.
279

572 M Visser
De Ruijter,W.P.M., Van der Giessen,A., & Groenendijk,

F.C., 1991, Currentanddensitystructure
in the Netherlands coastal zone, this volume.
Eisma,D., Kalf, J., & Veenhuis,M., 1980, The formationof smallparticlesandaggregates

in the
Rhineestuary,Neth.J. SeaRes., 14, 172-191
Festa,J.F.& Hansen,D.V., 1978, Turbiditymaximain partiallymixedestuaries:

A two-dimensional
numerical model, Estuarine and Coastal Marine Science, 7, 347-359
Kevorkian,J., 1990, PartialDifferentialEquations

AnalyticalSolutionTechniques,
Wadsworth
&
Brooks/Cole,PacificGrove,California,pp. 547
Lee, AJ. & Folkard,A.R., 1969, Factorsaffectingturbidityin the southerly

NorthSea,J. Cons.int.
Explor. Mer, 32, 291-302
Matkowsky,B.J.,Schuss,Z., & Trier, C., 1983, Diffusionacrosscharacteristic

boundaries
with critical
points,SIAM J. Appl. Math., 43, 673-695
Meade,R.H., 1972, Transportanddeposition

of sediments
in estuaries,
In: Nelson,B.W., (ed.),
Environmentalframeworkof coastalplaneestuaries,Geol. Soc.Am. Memoir., 133, Boulder,91-120
Murray,S.P.,1970, Settling
velocities
andverticaldiffusion
of particles
in turbulent
water,J.Geophys.
Res., 75, 1647-1654
Postma,H., 1981, Exchangeof materialsbetweenthe North Seaandthe WaddenSea,Mar. Geol.,40,

199-213
Stommel,H., 1949, Trajectories

of smallbodiessinkingslowlythroughconvection
cells,J. Mar. Res.,
8, 24-29
Van der Giessen,A., De Ruijter,W.P.M., & Borst,J.C., 1990, Threedimensionalcurrentstructurein

the Dutch coastal zone, Neth. J. Sea Res., 25, 45-55
Visser,D., 1987, Analysisof wind measurements

at lightvessels
Noord-Hinder
andTexelover the
period1951-1981,
Report87-3,Inst.Meteorology
andOceanography,
University
of Utrecht(NL),pp.
31
Visser,M., De Ruijter,W.P.M.,& Postma,

L., 1991,Thedistribution
of suspended
matterin theDutch
coastal zone, Neth. J. Sea Res., 27, 127-143

M Visser 573
Appendix A
In this appendixthe formulaefor the pressurefield, the horizontalvelocities,the stream
functionand the surfaceelevationare given.
Integrating(27.2) andapplyingboundarycondition(27.7) givesthe pressurefield:
p(x,z.) -g p (x) [½(x) - z] (27.A0)
Substitutionof (27.3) and (27.A0) in (27.4) and applying(27.12) - (27.14) leads to the
horizontalvelocitycomponents u and v:
U •
2pofC c•x
2Poke
+
2PokC
•b
• (•-2a')
sin (•cosh
(q)
- co•
(•-2d)
anh
(• ,]
cosh(•2a'). ,x..(q sinh (•2a')
y sin(• +d)cosh
(•-d) - cos
(• +d)si• (•-a')-]
2pokC - an (•-a') cobh(• +d) + • (•-a') sinh(• +a')]
2PokC - aia (n) cosh(n-2d) + c• (n) shh (n-
(27.A1)

574 M Visser
2p•fc ax
- sin
(n+eO
co•
(n-eO
+co•n
+a')
• (•-a')
.]
z
+
2PokC
•b
- sia
(n-2a')
• cosh
(n)+• (n-2a')
siah
(n)+]
(n) ½,mh(n-2a') - ex• (n) •.h (n-2a')
2PokC ,[-sia
(n+a')
cosn
sin (n-a')
- ex,s
(n+a')
siah
(n-a)
+]
2PokC sin
(q-2d)
cosh
(q)
- cos
(q-2d)
siah
(q)
+]
sin (n) ½osh(q-Za) + cos(q) •.h (n- •d)
+ [o(d- n)]
Po/ Ox
(27.A2)
Integrating(27.6) and applyingboundarycondition(27.16) leadto the 'slxeam'function
20o/½ ,co• (n-2d) cosh(n) - cosh(q- 2d) co• (n) + C +

toe (n +d) cobh(n-d) - toe (n-d) coeh (n +d)]
•2
2poke
•,•[sin
(q-d)
sinh
(q+tO
- sin
(q+d)
sinh
(q-d)]
•2 x
200/:<2
%[sin
(q)sinh
(n-2d)
- sin
(q-2d)
sinh
(q)]
82
[cos(n +d) cobh(n-d) - cos(n-d) cosb(n +d)]
2PokC
82
2potcC
•[[co•
(n- 2d)
cosh(n)
- cos
(n)cosh
(n-2d)
+C]
(27.A3)
where:

M Visser 575
z+h •+h
q- ; d-
C- cosh (2d) - cos (2d)
(27.A4)
From (27.11) - (27.14) it follows that:

y (27.A5)
x,,,- Po/'Vl
Substituting(27.A2) in (27.A5) givesan expressionfor the surfacegradient:
P a-• =- ax
,{, f [sin(d)
cos(d)
-sinh(d)
sic [- sia(d)•
(d) c,
osh(d)
+ c• (a') ]}
si• (a')J
f
% [sin (d) cosh(d) + cos (d) sinh(d)]
8kC
f
[sin(a')cos(d) + cosh(• sinh(a')]
8kC
(27.A6)
From (27.13) and (27.A1) an expression

for the cross-shore
bottomstressis derived:
+ L-sinh(d)cosh
(d)- cos(d)sin(d)
x•, [•nh (d) cos(d) - cosh(d) sin(a')]

tQc
(27.A7)
where:

576 M Visser
õL [sinh
(2d)- sin(2d)] (27.A8)
Nowoursystem
is completely
determined
if weprescribe
windanddensity
field.

28
Coastal dynamics along a rugged coastline

B King and E Wolanski
Abstract
A two-dimensional(depth-averaged)numerical model is proposed to describe the

hydrodynamics of the CentralGreatBarrierReef andis usedto investigate the dynamicsof
its coastalwaters. The model incorporates the dominanthydrologicaland meteorological
forcesand possesses sufficientresolutionto examinethe small-scaledynamicswhich are
influencedby both the reefs and the complexcoastaltopography.Field measurements
collectedover the shelfconfirm the broad-scalefeaturesof the model,thusyielding support
to predictions
of fine-scaleprocesses. Model simulations suggest thatthepresence of thereef
matrixenhancesthe velocityshearwithin the coastalzone,in boththe long-shoreandcross-
shoredirections.Trackingsimulations suggestthatthe shearactsasa steeringmechanism for
particulates
withinthewatercolumn.This shearwill hindertheexchange of wateracrossthe
shelf, thuscoastaltrappingcanoccur. Furtherimplications for effectiveregionalecological
management are that little terrestrially-derived
matterwill reachthe reef system.
28.1 Introduction
The CentralZone of the Great Barrier Reef Marine Park (CGBR) encompasses
mostof the
coastal and continental shelf waters off Townsville in North-east Australia. The shelf in this
regionis about120 km wide and its bathymetrygentlyslopesfrom the coastto a depthof
about 100m at the shelf-break(figure 1) with the isobathsapproximatelyrunningparallel to
the shoreline.Thusthe along-shore directionof thispartof the shelfis definedto be roughly
alignedwith the north-westward, south-eastward axis.
The CGBR is characterizedby: a complex seriesof headlands,capes and shallow

embayments at its westernboundary;an extensivereef matrixtowardsits easternboundary;
anda relativelytopographic-free
regionin between,knownas the 'lagoon'. The cross-shelf
widthof thelagoonis comparable to thatof theoffshorereef systemandthe two domainsare
generallyseparatedby the 40m isobath.

578 B King andE Wolanski
95 147
19S
- g
North
• •' • Site
4
• Site 3
• x
Townsville x
Site 2
Site I
Bowling
Green
Bay
Upstart Bay
Site 5 •
0 20 ß
I x 65
Figure 1' Grid boundariesof the studyareawithin the cenlxalGreat BarrierReef. The currentdata
collectedfrom within the regionare from mooringslocatedat sites1-6.

B King and E Wolanski 579
The shelfwatersof thisareaare usuallyverticallywell-mixedin temperature and salinity

duringtheaustralwinter(April - September)
whilesomestratification occursfollowingperiods
of calm weatherin the australsummer(WolanskiandBennett,1983). The innershelfwaters
alsoreceivefreshwaterpulsesfrom fiver runoff;but runoffoccursin events,whichusually
lastsonly a few daysat a time and occursprincipallyin the australsummer.
The currentsin the CGBR are predominantlybarotropicand are influencedby three

phenomena (WolanskiandPickard,1985):diurnalandsemi-diurnaltidal forcing,meso-scale
wind eventsand a polewardoceancurrentknownas the East AustralianCurrent(EAC).
28.1.1 Tidal currents
Analyticalmodelsof the currentsin the CGBR predicttidal ellipseswith the major axis
predominantlyin the across-shelfdirection (Church et al. 1985). Further, numerical
simulationsby Andrewsand Bode (1988) showthat the tidally-inducedcomponentof the
currentsin this regionis generallyweak (peaktidal currents< 0.2 ms'l) and field
measurements indicatethatthetidalcurrentsarehighlymodulated anddominatedby thesub-
tidal flows(WolanskiandPickard,1985). Work undertaken by AndrewsandFurnas(1986)
suggestthat the tidal flows would not contributesignificantlyto the large-scaleadvective
processes of this region.
28.1.2 Wind forcing
Duringtheaustralwinter,thewindfield is dominated
by thesouth-east
tradesandis highly
coherent
overdistancesgreaterthan1000km (Wolanski,1982). The windspeedgenerally
ranges
from0 to 15 ms4 andblowspredominantly
north-westward
(i.e.along-shelf)
in the
CGBR with typicalwindeventslastingfor a periodof 1-2 weeks. Non-tidal,along-shore
currentshavebeenfoundto be highlycoherentwith thelocalwindforcingduringthistime
(Wolanski andBennett,1983;Bunageet al. 1991a).
Due to the narrowingof the continental

shelfnorthof theCGBR and thepresence of the
densereef matrixin this northernregion,thereexistsa naturalbarrierto the equatorward,
along-shorewind-drivencurrents. This resultsin a pile-up of water and establishes a
longshore pressuregradienton theshelf,setup in responseto theuniformalong-shorewind
stress.
28.1.3 East Australian Current
TheEAC flowspolewardalongtheoutershelfandshelfedgein thisregionandis partof the

south-westPacificOceancirculation(Church,1987). It is a boundarycurrentwhich is
associated
with an offshorealong-shelfslope. Measurements of the currentsoffshoreof the
CGBRshowspeeds typicallyof theorderof 0.3 ms4 in thetop200mof theoutershelfand
shelfbreakwaters(Church andBoland,1983).In theabsence of localwindforcing,theeffect

58O B King and E Wolansld
of theEAC propagates acrosstheentireshell through

thegenerationof a longshorepressure
gradienton the shelf. This gradientdrivesa quasi-steady
southward flow in the shallow
coastalwatersaswell asthroughthereefmatrix(Burrageet al. 1991a;Wolanskiet al. 1989).
In summary, Godfrey(1973a,b)showsthatthetwopressure gradients setup on theshelf

as a resultof the wind forcingandEAC combineandthe resultingpressuregradienton the
shelfdrive the low frequencycurrenmwhichhavebeenshownto play the majoradvective
exchange role in thisregion. WolanskiandPickard(1985)andBurrageet al. (1991b)also
showthatat low frequencies, thealong-shore
currents
aresignificantly
largerthanthecross-
shelf currents. Thus given the aboveassumptions, we can simplify our work by only
considering the role that the dominant,non-tidalphenomena contributesto the coastal
dynamics of theCGBR. This wasalsotheprocedure usedin the studyof oil dispersion
in
the CGBR by Andrewet al. (1983).
Finally,giventheaboveknowledge of theCGBR,we implemented a numerical modelof

thisregionas an extension of the workof WolanskiandBennett(1983) andBurrageet al.
(1991 a,b). Theseauthorshaveshownthatspatialvariabilityof the sub-tidalcurrentsexists,
in boththe cross-shelf and the along-shelfdirection;and it is the aim of this paperto use
numericaltechniques to: (a) resolvethenatureandstructureof thisspatialvariabilityin much
finerdetailand(b) investigate theinfluenceof theunusual
topographical featuresof theCGBR
on the advectivecharacteristics
of its coastalwaters. This is importantin orderto determine
the fate of land-derivedorganicand inorganiccompoundsand variouspollutantsand to
understand what their influencewouldbe on the reef systemand the coastalzone.
The modelutilizeshydrological andmeteorological

dataof theregion.Currently,suchdata
setsare availableonly for the australwinter.
28.2 Available data
During1985,current,windandwaterelevationdatawerecollectedin theCGBR (seeBurrage

et al. 1991bfor details). In particular,four currentmetermooringswere locatedacrossthe
shelf (shownas sites 1-4 in figure 1). Each mooringcomprisedof multiplecurrentmeters
situatedat variousdepthsthroughout thewatercolumn.SinceBurrageet al. (1991b)reported
onlyveryweakverticalshearin thecurrentsmeasured on theshelf,onlythemid-depthcurrent
meterdatawere usedfor modelverification,as thiswasassumedto be representative of the
depth-averaged currentvelocity. Additionalcurrentmeterdatawerealsocollectedin 1980 at
sites5 and 6 (seeWolanski and Bennett, 1983).
The mid-depthcurrentmeterdatathat were collectedat the four locationsacrossthe shelf

were analysedto examineonly the sub-tidalcurrentsand their responseto wind forcing.
Duringthe SE tradewind season,highcorrelationbetweenthe sub-tidalalong-shelfcurrents
(V), lines(a) - (d) in figure2 andthealong-shelf
windvelocity(Wy),line(e) in figure2, was

B King and E Wolansld 581
observed at all inshore sites.
0.5
0.0
--0.5
0,5
0.0
• o.o
-0.,5
0.5-]
0.0
-0.5
12.0
4.0 e
-
-4.0 • • •
120 150 180 210
Time (d•ys)
Figure2: Time-series
plotsof thehydrological
andmeteorologicaldatausedfor modelcalibration
and
forcing.The sub-tidalalong-shelfcurrents
(V), fxomsites1 - 4, aredepicted
in graphs(a) - (d)
respectively.
Thelow-pass
filtered,
along-shelf
windforcing
(Wy),asmeasured
in Mackay,
is shown
in graph(e). The timeis givenas thedaynumberin 1985andcoverstheaustralwinterseason.

Regressionanalysis,using the Methodsof Least Squares,was employedto describea

general
linearrelationship
between
V andWyat eachsite,whichis independent
of time. The
resultingregression equationsare givenin table 1. The confidencelevelsof the correlations
betweenlongshorecurrentsand winds are particularlyhigh in the shelf watersbut lost
significanceat the shelf-break(site 4). Thesesimplefunctionsgive a measureof the mid-
depthcurrentresponseto local wind forcingandare usedfor modelverification.
Table 1: Linear regressionequationsgiving the along-shelf,mid-depthcurrent(V) responseto local

along-shelf
windvelocity(Wy)for eachmooring
siteduringtheaustral
winter;withthecorresponding
R: andthetime(inhours)
thatV lagsWy,calculated
fromtheanalysis.
* denotes
R:values
which
are
significantat the95% confidence
interval.
Site Regression
Eqm. R" Lag(HR)
4 V4= 4.92x 10'•'Wy- 0.312 0.463 24
3 V• = 4.00x 10'•'Wy- 0.182 0.737* 15
2 V•.= 3.70x 10'2Wy- 0.0973 0.738* 15

1 V• = 2.94x 10'•'Wy- 0.0619 0.740* 9
28.3 A numerical model
To understand
and studythe sub-tidalhydrodynamiccharacteristicsof the coastalflows in the
CGBR, a detailednumericalmodelof the entireshelfwas set-upin orderto encompass the
small-scale
andshelf-scale
influencesin thisregion. The numericalmodelusedto achievethis
is basedon the work of Leendertse(1970) andFalconeret al. (1986) and solves,usingan
implicit scheme,the followingdepth-averaged
two-dimensional hydrodynamic equations:
c• & C2 - W
-% p
(28.2)

B King andE Wolanski 583
8 LrH il VH
/}q * * - 0 (28.3)
where t is the time; x and y the horizontalaxes, orientatedacross-shelf

and along-shelf
respectively;
H the totaldepth;U and V the depth-averaged
horizontal
velocities;
• is a
correctionfactorfor the non-uniformityof the verticalvelocityprofile(=1.016);f the Coriolis
parameter; g thegravitational
acceleration (=9.81);q thefreesurfaceelevation;p is thewater
density;
T is thedepth-integrated
Reynolds
stress;
C is theChezycoefficient
(=H1/•/n,
where
n is the ManningCoefficien0;andW is the wind forcing.
The depth-averaged
equationsare valid for this applicationdue to the predominantly
horizontal flow which occurs in shallow shelf waters. Three dimensional currents have been
observedaroundreefsand in wakesbehindislandsand headlandsbut thesepresumablywill
contributelittle to the large-scaleflows and are mostlikely to be sub-gridscalephenomena
in thismodel(WolanskiandHamnet,1989;Deleersnijder, et al. 1989). Barocliniceffectsand
atmospheric
pressure
differences
duringtheaustralwinterare negligible.
Further,this numericalschemehas been usedextensivelyin other hydrodynamicstudies,

includingregionswithintheCGBR. Thispresents uswiththeadvantage thattheformulations
andparameters usedin equations
(28.1) - (28.3) havealreadybeenpreviouslytested,calibrated
and verifiedwith numerousfield datasets(e.g. Wolanskiet al. 1989;Falconeret al. 1986;
WolanskiandKing, 1990)thusallowingus to utilizeour limiteddatafor verificationof the
presentscheme. The work of Burrageet al. (1991b)alsoprovidesrealisticestimatesof
boundaryconditionsand model parameterswhich we have employedin establishinga
representative
shelf-modelof the CGBR.
The studydomainwasrepresented by a meshwith 65 pointsin the across-shelf direction

(X axis, positive eastward)and 95 points in the along-shelfdirection(Y axis, positive
northward)as shownin figure 1. The resultingunit meshsize was 2031.2 meterssquare,
whichwasa suitablesize in orderto encompass the influencesof boththe large-scaleshelf
dynamicsand thatof the topographically-complex,small-scalereefsandcoastaltopography.
The coastlineclosedthe westernboundaryof the modeldomainwhile theremainingthree

boundaries
were openand were forcedby sea-level(q) fluctuations.
The easternboundaryoccurredat the shelf-breakand approximatelycoincidedwith the

100misobath.This boundaryis usedto controlthealong-shelfgradientin the waterelevation
(Godfrey,1973b),whichis dueto the combinedpresenceof large-scalewind forcingand the
EAC as detailedearlier in this text. At this stage,the absolutevalue for this along-shelf
gradientis unknown,thereforeit is simplydeterminedby trial-and-error.That is, the slope
is calibratedby adjustingthe water elevation along this boundary;such that the model
predictedcurrentat site4, matchthosemeasuredat site4, for any givenwind forcing. Thus,
havingcalibratedthisopenboundary,the remaininginshoresites(1-3) are usedto verify the

584 B King andE Wolansld
model'sresponse
to thisforcingacrossthe shelf. Thesecomparisons
are listedin table2.
Table 2: Comparison
of the numericallypredictedalong-sheff
component
of the currentvelocity(Vm;
ms'l),withtheobserved
(V•,;ms-l)values
extrapolated
fromthecorresponding
regression
equation,
listed
in table 1, for the 6 givenwind conditions.
Run # 1 2 3 4 5 6 7
sit#
-0.31 -0.31 -0.21 -0.21 -0.12 -0.12 -0.02 -0.02 0.08 0.08 0.18 0.20 -0.31 -0.25
-0.18 -0.18 -0.10 -0.09 -0.02 -0.01 0.06 0.09 0.14 0.12 0.22 0.17 -0.18 -0.18
-0.10 -0.10 -0.02 -0.03 0.05 0.06 0.13 0.11 0.19 0.20 0.27 0.26 -0.10 -0.08
-0.06 -0.06 0.00 -0.01 0.06 0.05 0.12 0.08 0.17 0.20 0.23 0.22 -0.06 -0.03
For thenorthernandsouthern across-shelfopenboundaries,WolanskiandBennett(1983)

andBurrageet al. (1991a)suggestthe across-shelf
slopeis in geostrophicbalancewith the
sub-tidalalong-shelf
flows. Therefore,boththeseboundaries are controlledby:
gH8q
Ox
=fVH (28.4)
whereV andH werelocallydeterminedfrom theprevioustime-step.
Simulations
wereperformedfor sixdifferentwindconditions
whichdominatethewindfield
duringtheaustralwinter(Wolanski,1982).Theseconditions
assume south-easterly tradewind
withvelocities
rangingfrom0 to 10ms'1at 2 ms'1increments,andareidentified asrun# 1 to
#6 respectively.A simulationwasalsoperformedidenticallyto run # 1, but withoutthereef
matrix (run #7) in order to assessthe influenceof the reef on the near inshorewaters.
The model was started from rest and allowed to run for 120 hours in order to reach a
steadystate. Testswere thenperformed,utilizingthe along-shelf
slopealongthe outer
boundaryunderno wind conditions,usingvarioustimesteps,to examinethe effectsof
numericaldiffusionon the simulation. The predictedflows for each test revealedno
significant
dependency
on timestep.The timestepwassetat 600 seconds.
Finally,thenumerically-predicted
currentsof thefirst6 runswereusedto drivea simple
advection
model.Theadvectionmodelutilised
thesheif-model velocities
todetermine
drogue

trajectories.For eachran, thesesimulateddrogues,representing any passiveparticle,were

released;oneat eachgridcell, in an across-shelf
transectwithinthecoastalandlagoonregion.
This techniquebetterenabledus to examinethe effectsof the across-shelf shearby taking
advantageof the greaterresolutionof the shelf-model. Thus we were able to assessthe
transportdynamicsof the coastalregionfor a varietyof weatherscenarioswhich are likely
to occurduringthe australwinter.
28.4 Results
The resultsfrom eachof the first 6 simulations

werecomparedwith datafrom sites1 to 3 for
model verificationafter the easternboundarysea-levelgradientwas adjustedto match the
flows observedat site 4. They were generallycloseto thosepredictedby the regression
equations(see table 2). This is not surprisingconsideringthe previouswork with this
numericalschemewithin this region. Howeverit is encouraging to see that the sub-tidal
currents,as reportedby Burrageet al. (1991a) anddescribedby the regressionequationsof
table1, arereproduced.Thuswe werenowconfidentthatthe finerdetailgivenby the shelf-
modelof the shearin the sub-tidalcurrents,is a reasonable
representation
of reality.
Analysisof the momentumbalancefor mn #1 revealsthat, underno wind conditions,the

coastalwatersare predominantlyinfluencedby the offshoresea level gradientand flow
southwardalong-shelf. Preliminaryrestfitsfor run #1 (reef, no wind) and #7 (no reef, no
wind) werecomparedwith datafrom sites1 to 6. Thisexerciseshowedthe inshorecurrents
predictedby mn #7 under-estimatexl thoseobservedin the field. However,mn #1 yielded
predictedcurrentswhichcomparedwell to the observedones. Fromtable2, we can seethat
velocityshearexistswithinthelagoondueto across-shelf variationsin the frictionalinfluence
of this slopingshelfin boththesecases. However,while shearwill existacrossany shelf
undertheseconditions,it canbe notedfrom the datain table2, thatthe presenceof the reef
matrixstrengthens the inshorewatervelocity(comparing mn #1 with mn #7). This suggests
that the reef matrix plays an importantrole in the hydrodynamics of the coastalwaters,
probablyby channellingwatersfrom thereef complexto the innershell Thischannellingis
enhancedby the pre•nce of the large southernreef systemsin the domainand this can be
seenin figure3a.
DatafromWolanskiandBennett(1983) suggested thatthealong-shelfcurrentsat site5 and

6 wouldbe approximately 0.2 m/s and0.4 m/s respectively,to the southin the absenceof
wind forcing. This agreeswith the predictedvelocitiesgivenby mn #1. Given that the
frictionalinfluenceat site 1 is similarto 5 andsite2 is similarto 6, thenthe along-shore
variationin velocitystrengthcan alsobe attributedto the influenceof the reef matrix and
coastaltopography. Topographic steeringby thecoastalheadlands andreefclusters promotes
the channellingof water throughthe lagoonand this will contributeto the along-shore
variationin theadvectionmechanisms. Thiscanbe seenin detailin figures3a and3b.

Flows within embayments alsodemonstrate somecross-shelf shear;mostlydue to the

increasedfrictionalinfluenceof the shallowwaters. Stagnationzonesalsooccurwithin the
embayments asa resultof upstream blockingby headlandsandcapes.
The influenceof this across-shelfshearon the transportof neutrally-buoyant

matteris
demonstratedby simulatingdroguetrajectories
with the advection
model,giventhe flow
patternshownin figure 3a. Snapshotsof the droguepositionscan be seenin figure4.
Velocityshearin theembayments
andnearthecoastpreventtheinshore drogues
frommoving
offshore.Thus,inshorewatersappearto remaincoastally-trappedfor longperiodsandarenot
readilyflushedundertheseconditions,
thatis, by theadvectiveexchange processespresentin
these coastal waters.
The drogues releasedin deeperwatersaresubjected to stronger

andlessobstructedflows
andareadvectedfurtherdownstream thanthoseinshore.The velocityshearacrossthelagoon
tendsto alignthesedroguesinto a slickformation(seefigure4f). Thustheadvectiveprocess
here, underthe influenceof velocity shear,aggregatethe drogueswhich may opposeany
diffusiveprocesses operating.
In the presenceof light SE winds(1-3 m/s; mn//2), decouplingof the shallowcoastal

currentsfrom the offshoreshelf currentsoccursin the lagoon(see figure 5). The inshore
currentsare dominatedby the wind forcingand movenorthward,while the offshorecurrents
remainunderthe influenceof the along-shelfslopeof the EAC and flow southward.There
is little currentin the zonebetween. This decouplingresultsin a separationof coastaland
offshoreshelfwatersand canbe expectedto reducethe exchangeof watersfrom the coastal
zoneto the reef matrix and vice versa. The offshoreextentof theresulting'separationfront'
is foundto be dependenton the wind strengthand water depth.
The 'separationfront' runscontinuously throughthelagoonof theCGBR undertheselight

wind conditions,exceptfor one break offshorefrom Cape Bowling Green (see figure 5).
Local coastaltopographic steeringof the inshorecurrents,by the headlandin thisparticular
case, have causedthe front to become segmented. This may reduce the along-shore
progressionof shallowwater,but wouldlocallyenhancethe cross-shelf exchangeof water.
Figure6 showsthe synopticviewsof the droguelocationsfromrun//2. The embayment

watersare very slow to flush due to the weak velocitiesin the coastalzone. Again the
droguesdeployedin the deeperregionof thelagoonalignthemselves into a slick. Two slicks
are presentin the later stagesof the simulation,emphasising
therole of the separationfront
No evidenceof water exchangefrom the coastalzone to the reefal systemis presentfrom
thesetrajectories.

Figure3a: The predictedcurrentprofriesof the modeldomain,forcedby an along-shelfslopeonly (ran

#1). Thelength
of thearrows
represents
therelative
current
velocity
strength
anditsorientation
indicatingthecorresponding
directionof flow. Part(a) showsthepredictedshelf-scale
flowswhile (b)
gives a subsetof the domainfocusingon the predictedflows within the coastalzone of the CGBR
lagoonand its embayments.Note the velocityshearacrossthe lagoonand the variationin current
strengthalongthe lagoon.

IIIIIi111
IIIIIIii1111
•111111111
slllllillll
II•111111111
s%11•11111111
• • i I I I I I I I I I I I
] .... •11111111111111
/ --''•s•11111111111
Length
scale
,,,,
z/ /I//
/ ...... .......
, • •;;txilXl //
VelocityScale • 0.25
Figure
3b:Thepredicted
current
profiles
ofthemodeldomain,
forced
byanalong-shelf
slopeonly
(ran
#1). Thelengthof thearrowsrepresents
therelative
current
velocity
strength
anditsorientation
indicating
thecorresponding
direction
offlow.Part(a)shows
thepredicted
shelf-scale
flowswhile
(b)
givesa subset
ofthedomainfocusing
onthepredictedflows
within
thecoastal
zoneoftheCGBR
lagoonandits embayments.Notethevelocity
shearacross
thelagoon
andthevariation
incurrent
strengthalongthe lagoon.

Figure4: Snapshots
of drogue distribution
drivenby anadvectionmodelusingtheoutput of run#1.
Plot(a) shows
theinitialdistribution
of 26 drogues
alongthetransectY=55 for 10 < x _<35 at time=
0 days.Theremakning plotsgivethedistribution
after(b)2 days,
(c)4 days,
(d)6 days, (e)8 daysand
(f) 10 days.

59O B King andE Wolanski
Figure
5:Theflowpatterns
whichexistinthecoastal
region
fromsimulation
#2. Thedecoupling
of the
currents
across
theshelfthatresults
is highlighted
by a separation
zoneandstrong
shearpatterns.The
wind forcingis directedup the page.

a a b a
c d
e f a
: '•.
Figure6: Snapshotsof droguepositiondrivenbyanadvection

modelusingtheoutputof mn#2,given
thesameinitialdistribution
shown in figure4a. These
plotsgivethedistribution
after(a) 2 days,(b)
4 days,(c) 6 days,(d) 10 days,(e) 15 daysand(f) 20 days.
Asthewindvelocitystrengthensfromthesouth-east,
thelocalwindforcingdominates and
flowreversaloccursacross
theentireshelf.Undertheseconditions,
velocity
shearstilloccurs
nearthecoast andisparticularly
evidentin theembayments.Seefigure7 forexample.The
existence
of aneddyin BowlingGreenBay,in theleeof thecape,canalsobeobserved. This
will resultin sometrappingof the waterin thisarea.

•, / \.xx•tttttttt
, I • ltltttttlttl
• • • f ffttt tttt tt
..... ///flftttttltt lilt
..... ,•ftttttttttt 1111
..... •tttttttttttt
..... ,•T t tit t ttt t 1
...... ,\•111t tittit
....... \•111 tt ttttt
.. .ITtttt t flirt
- •z//lfff f ill I!
•,//II ftt f I It I f
_....• • / / I I f t f f t t I t 1
/,/zz//tttfftlttttt
/ •ttttttf fttttttttt
/..,•,,,,,x\\•11
\ ::
\ '"'-'/.//,///,;III
! / / / • /II
.. • /////,'/II
" ///IZll, II!
• /• //////,'//!
• ... /,/////////
/,////II/I////
/•////////////
• •/////////Z//
/,///////////ZZ
•////////////Z•
-_)s////////////Z•
_ ¾, /////////////Z•
• !, ////III//////Z•
• i %//////I/I/I//ZZœ•
/ .... ,/••////II/ZZœ•
,,,,,•, •//III///////
Velocity Scale • 0.25 m/s
Figure7: Increasing
windstrength
(run#3)causesflowreversal
across
theentirelagoon
andshelf.This
leadsto strongshearzonesacross
themouthsof theembaymentsandaneddydevelops in theleeof
CapeBowlingGreen. The wind forcingis directedup thepage.

a a b a
Figure8: Snapshotsof droguepositiondrivenby an advectionmodelusingtheoutputof run #3 given

the sameinitial distributionshownin figure4a. Theseplotsgive the distributionafter (a) 2 days,(b)
4 days,(c) 6 days,(d) 10 days,(e) 15 daysand (f) 20 days.

594 B King and E Wolanski
Length Scale
5000 m
Velocity Scale • 0.25 m/s
Figure9: Strongwind (run#6) drivesstronginshorecurrents.Thisleadsto strongshearzonesacross

the mouthsof the embayments andan eddydevelops
in the lee of CapeBowlingGreen. The wind
forcingis directedup the page.

The inshoreflow increases correspondinglywith wind strengthasshownin table2. Under

the milderconditionsof run #3, the droguesremainresidentwithin BowlingGreenBay for
up to 10 days(figure8) whilerung6 predictsrapidflushingof theembayment by the second
day (not shown).
However the observedtrajectoriesof the droguesfor runs #3 to #6 all show a similar

distributionto that depictedin figure 9. This is due to the similarvelocityshearpatterns
whichexistundernorthward-moving coastalflows. All simulationsshowslick aggregation
of the off-shoredrogues.
28.5 Discussion
The behaviourof inshorewateradvectiveprocesses for the CGBR havebeendescribedfor a

varietyof hydrologicaland meteorological scenarios,commonlyoccurringduringthe austral
winterperiod. This wasachievedby utilizingexistingfield dataanda numericalmodelof the
shelfof theCGBR to resolvethe detailedflow regimeof theregion. This analysishighlighted
the role that the reef matrix plays in modifyingthe dynamicsof the coastalwaters,by
topographically-steeringand channellingwhichenhancesthe inshoreflow. This processalso
leadsto increasedvelocityshearwithin the lagoonregionof the shell
The channellingeffectsproducedby the irregulartopographyof the coastandreef matrix

in the CGBR also explainthe along-shelfvariationin currentvelocitieswithin the lagoon
(comparingsites1 with 5 and2 with 6). This demonstrates the importanceof our approach
herein designinga fine-scaleshelfmodelin orderto studythecoastaldynamicsof theregion.
Converselyonemustproceedwith cautionwhenusingsimplermodelsfor predictingcoastal
and shelfflows in sucha complexenvironment.
From our understandingof the dominant processescontrolling long-term water

transportationin the CGBR; numericalsimulations basedon theseprocesses were analysed,
in orderto determinethelikely coastaldynamicsandcross-shelf exchange processespresent.
Theseare madeapparentby calculationsof the fate of neutrally-buoyant particles. These
simulations revealthatunderall windconditions, strongvelocityshearexistacrossthelagoon
whichinhibitwaterexchangefrom thecoastalzoneto thereef system.Thusinshorewaters
remaincoastallytrapped. Flushingis apparentlyreducedby the velocityshearacrossthe
openingsof embayments and the generationof eitherstagnationzonesupstreamand eddies
in theleeof headlands whichpromotelocaltrapping.In theembayments, flushingof coastal
watersis furtherinhibitedby trappingin thefringingmangroveswamps(WolanskiandRidd,
1990).
Thisstudyhasimportantimplications
for thesoundecological
management of thecoastal
andreefsystems.Thatis,anyland-derived
matter(suchasdissolved
nutrients,
pesticides
and

otherpollutantsand suspended sedimen0enteringthe coastalsystemare not readilyflushed

andremaincoastallytrappedundermostweathers conditions,
in theaustralwinter. Depending
on thenatureof theinjectedmatter,accumulation
maypollutethecoastalwatersin thisregion.
However,sincelittle across-shelf exchangeprocessesare evidem,the reef systemwould

appearto be isolatedfrom this threat. Indeed,observations of low levelsof agricultural
pesticidesfoundin reeforganisms in theregionby Olafson(1978) supportourclaim. Olafson
postulatedthat the CGBR was not underthreatfrom land-derivedpesticidesfrom adjacent
fiver runoff due to the natureof the oceanographic featuresof the regionas observedfrom
terrestrial sediment distributions on the seabed.
Finally,whiletheacross-shelf
tidalexcursiondoesnotsignificantly
contributeto theacross-
shelf exchangeof water, the role of tidal diffusionin mixing the watersis yet-to-be
determined.Although,from an ecologicalperspective, this processwouldhelpdiluteany
injectedcontaminants,furtherresearchinto its effectsis warranted,in orderto give some
measureof the mixing ratesin the coastalzone.
28.6 References
Andrews, J.C. and Bode, L., 1988, "The tides of the central Great Barrier Reef'. Continental Shelf
Research.,Vol. 8, 9, 1057-1085.
Andrews, J.C. and Furnas, M.J., 1986, "Subsurfaceintrusionsof Coral Sea water into the central Great
BarrierReef - I. Structuresand shelf-scaledynamics".ContinentalShelf Research.,Vol. 6, 4, 491-514.
Andrews,J.C., Mitchell, A.W. andBellamy,N.F., 1983, "Field studiesof currentsandsimulationsof

oil dispersionin the CentralGreat BarrierReef'. Sixth AustralianConferenceon Coastaland Ocean
Engineering.Gold Coast,Australia,13-15thJuly, 1983.
Burrage,D.M., Steinberg,C.R. and Church,J.A., 1991a, "Currentvariability and shearacrossthe

continentalshelf and slopeof the CentralGreat BarrierReef'. (submittedto Journalof Physical
Oceanography).
Burrage,D.M., Church,J.A. and Steinberg,C.R., 1991b,"Linearsystemsanalysisof momentumon the

continentalshelf and slopeof the CentralGreat BarrierReef'. (submittedto Journalof Geophysical
Research).
Church,J.A., 1987, "EastAustraliancurrentadjacentto the GreatBarrierReef'. AustralianJournalof

Marine and FreshwaterResearch,38, 671-683.
Church, J.A., Andrews, J.C. and Boland, F.M., 1985, "Tidal currentsin the central Great Barrier Reef'.
Continental Shelf Research,5, 515-531.

Church,J.A. andBoland,F.M., 1983,"A permanent undercurrentadjacentto theGreatBarrierReef".

Journalof PhysicalOceanography,13, 1747-1749.
Deleersnijder,
E., Wolanski,E. andNorro,A., 1989,"Numericalsimulation
of thethree-dimensional
tidal
circulation
in an island'swake".Proceedings
of the 4th International
Conference
on Computational
MethodsandExperimentalMeasurements,
Capri,Italy, 23-26 May, 1989.
Falconer,R.A., Wolanski,E. and Mardapitta-Hadjipandeli,

L., 1986,"Modellingtidalcirculation
in an
island'swake".Journalof Waterway,Port,CoastalandOceanEngineering,
AmericanSocietyof Civil
Engineering,112, 234-254.
Godfrey,J.S., 1973a,"On the dynamicsof the westernboundarycurrentin BryanandCox's (1968)

numericalmodelocean".Deep-seaResearch,20, 1043-1058.
Godfrey,J.S., 1973b,"Comparison
of theEastAustralian
Currentwith the westernboudaryflow in
BryanandCox's (1968) numericalmodelocean".Deep-seaResearch,20, 1059-1076.
Leendertse,J.J., 1970, "A water-qualitysimulationmodel for well-mixedestuariesand coastalseas:

Volume1, Principles of computation".Report No. RM-6230-RC.The RandCorp.,77.
Olafson,
R.W., 1978,"Effectof agricultural
activityonlevelsof organochlorine
pesticides
in hardcorals,
fish and molluscsfrom the Great BarrierReef". Marine EnvironmentalResearch,Vol. 1, 2, 87-107.
Wolanski,E., 1982, "Low-leveltradewindsover the westernCoral Sea".Journalof Applied

Meteorology,21, 881-882.
Wolanski, E. and Bennett, A.F., 1983, "Continentalshelf waves and their influence on the circulation
aroundthe Great BarrierReef'. AustralianJournalof Marine andFreshwaterResearch,34, 23-47.
Wolanski,
E., Burrage,
D.M. andKing,B., 1989,"Trapping
anddispersion
of coraleggsaroundBowden
Reel GreatBarrierReel followingmasscoralspawning".
Continental
ShelfResearch, Vol. 9, 5, 479-
496.
Wolanski,E. andHamner,W.M., 1988, "Topographically

controlledfrontsin the oceanand their
biologicalinfluence".Science,241, 177-181.
Wolanski,E. andJones,M., 1981,"Physical

properties
of GreatBarrierReef lagoonwatersnear
Townsville.1. Effectsof Burdekin
Riverfloods".Australian
Journal
of MarineandFreshwater
Research,
32, 305-319.
Wolanski,
E., Jones,
M. andWilliams,
W.T., 1981,"Physical
properties
of GreatBarrierReeflagoon
watersnearTownsville.2. Seasonal
variation".Australian
Journalof MarineandFreshwater
Research,
32, 321-334.

Wolanski,E. and King, B., 1990, "Flushingof BowdenReef lagoon,Great BarrierReef'. Esmarine,
Coastal and Shelf Science, 31, 789-804.
Wolanski,E. and Pickard,G.L., 1983, "Currentsand flushingof BritomartReef lagoon,GreatBarrier

Reef'. Coral Reefs, 2, 1-8.
Wolanski,E. andPickard,G.L., 1985, "Long-termobservations

of currentson the centralGreatBarrier
Reef continental shelf". Coral Reefs, 4, 47-57.
Wolanski, E. and Ridd, P., 1990, "Mixing and trapping in Australiantropical coastalwaters".In
"ResidualCurrentsand Long-termTransport",R.T. Cheng(Ed). Springer-Verlag,New York.

29
Transport of hypoxic waters: an estuary-

subestuary exchange
A Y Kuo and K Park
Abstract
Hypoxic or anoxic conditionsin the subpycnoclinewater of ChesapeakeBay persist

throughout the summer.The effecton the dissolvedoxygenconcentration in the deepbasin
of the lower RappahannockRiver, a subestuaryon the westernside of the bay, was studied
with an observationalprogram. The data indicatethat in the lower portionof the water
columnthe subtidal(or residual)currentwas directedinto the subestuarymostof the time.
The massfluxesof salt and dissolvedoxygeninto the subestuary througha point near the
estuadnebottom at the fiver mouth were calculatedfor tidal and subtidalcomponents
respectively.From the analyses,we concludethat the massexchangeowing to the tidal
componentis at leastan orderof magnitudesmallerthan that resultingfrom the subtidal
component. Characteristic differencesin the propertiesof water transportedinto the
subestuary wereobserved.On five occasions, eachlastingabouttwo days,duringthe one
monthperiodof field measurements, the importedwaterwascharacterized by low dissolved
oxygen
andhighsalinity,
typically
3 to4 mg1-1lowerdissolved
oxygen
and2 to3 psuhigher
salinitythanat othertimes. The low dissolvedoxygen,high salinitywatermasseswere all
accompanied by strongsubtidalcurrentandsouthwest wind. Duringperiodsof strongwind
fromthesouthwest quadrant,thedensity-drivencurrentnearthebottomwasenhanced by the
wind-drivencirculation.Furthermore,thesurfaceset-upfavorsthetransport of thewaterfrom
the deepportionof the bay resultingfrom (1) tiltingof the pycnoclinein the bay and (2)
shoreline andbathymetricconfigurationsaroundthe estuary-subestuaryjunction.
29.1 Introduction
Problemsof hypoxia and anoxia in estuarineand coastalwaters have receivedincreased

attentionin recentyears(Schroeder,1985).Thedepletion of dissolvedoxygen(DO) generally
resultsfroma combination of biologicalandphysicalfactors.It is mostfrequently
observed
in subpycnoclinewatersduringsummer whenverticalmixingis weakandwatertemperature
is high. In Chesapeake Bay,for example, summeranoxiahasbeenobserved in thedeep

600 A Y Kuo and K Park
bottom waters since the 1930s (Newcombe and Home, 1938), and has become more
widespreadand of longerdurationduringrecenttimes(Flemeret al., 1983).
All majorsubestuarieson thewesternsideof Chesapeake Bay havea deepbasinneartheir

mouths(Fig. 1). Hypoxia has been observedfrequentlyin the deep basinsin all these
subestuariesexceptin theJamesRiver, thesouthernmost subestuary.Kuo andNeilson(1987)
madea comparative investigation of thebottomDO in the threeVirginiasubestuaries.
They
reportedthathypoxiaoccurredmostfrequentlyin thedeepwaterof theRappahannock River,
lessfrequentlyin the York River, andrarelyin theJamesRiver,eventhoughit receivedthe
heaviestwastewaterloadingsamong the three subestuaries.This was attributedto the
relativelystronggravitationalcirculationin the JamesRiver.
•7o
00'
) KILOMETERS
?7o00' 76000 '
Figure1'. LowerChesapeake
Baywithitsfourmajorsubestuaries
(Theshaded
areasaredeeperthan20
m, the depthcontoursare 13 m; X marksthe stationwith time seriesmeasurements
and O marksthe
slackwatersurveystationin the bay).

A Y Kuo and K Park 601
Sincethe variabilityin hypoxicconditionsamongthe threeVirginia subestuaries follows

a similartrendof decreasing bottomDO from southto northin the main stemof Chesapeake
Bay, it is possiblethat the variabilitymay be due,at leastin part, to the qualityof the waters
importedinto the mouthsof the subestuaries from the bay. In this paper,the effect of the
hypoxicor anoxicconditionsin thebay on waterqualityin the lowerRappahannock River is
investigated by examiningthe masstransportnearthebottomat thefiver mouth. Time series
datafrom simultaneous measurements of DO, temperature, salinityandvelocityat the bottom
of the fiver mouthwereanalyzedto demonstrate the primaryfactorsdominatingthe estuary-
subestuaryexchange.
29.2 Field observation
Field datafor thisstudywerecollectedduringsummer1987. Data collectioncanbe divided

into two major groups.
29.2.1 Slackwatersurveys
A totalof 13 slackwatersurveyswereconducted fromJuneto Septemberat slackwaterbefore

ebb. During each survey,temperature,conductivityand DO were measuredat 11 stations
alongthefiver axisandonestationoff thefiver mouthin themainstemof thebay. The fiver
stationsextendedfromthefiver mouthto 57.8 km upriverbeyondthedeepbasinin thelower
subestuary.
Temperature andconductivity weremeasured withan AppliedMicro Systemconductivity-

temperature-depthprobe (CTD). Continuousverticalprofiles,top to bottom,for these
variableswereobtainedat eachstation.DO wasmeasured usingan oxygenprobe(Yellow
SpringsInstruments)everymeterfromthe surfaceto 15 m depth,thenevery2 m until the
bottomwasreached.Conductivity andtemperature measurements wereconverted
to salinity
as psu (UNESCO, 1983). Salinity, temperatureand DO data for the fiver stationswere
displayedas isoplethsin the vertical-longitudinal
planefor eachsurvey.
Two examplesof spatialdistributions

of DO and salinityare presentedin Fig. 2. In
general,the surfaceDO concentrations
werealwayshigh,with little variabilityalongthis
stretch
of thesubestuary.
TheDO concentrationsin thebottomwaterwerelow in all surveys,
with a characteristic
longitudinalpattern. The bottomDO concentrations
decreasedin the
upriverdirection
fromthefivermouth,reached
a minimumandthenincreased
aswaterdepth
decreased.Therewasa strikingdifferencein verticalsalinityandDO distributions
at thefiver
mouthbetweenthetwodates.The watercolumnat thefivermouthwasmoderately stratified
on August10 andbecamenearlyhomogeneous on August17 (Fig. 2a and2b). Fig. 2b also
showsan apparent intrusionof highsalinitywaterat mid-depth, thusimplyingprobability of
reversedthreelayeredcirculation
(Elliott,1978). Thiscreatedanunstabledensitystratification
in thelowerportionof watercolumn,andenhanced theverticalmixing.

6O2 A Y Kuo and K Park
DISTANCE UPSTREAM FROM MOUTH (Kilometers) DISTANCE UPSTREAM FROM MOUTH (Kilometers)
Figure2: Distribution
of dissolved
oxygen
(mg14)andsalinity(psu)in thelowerRappahannock
River.
a) August10, 1987 andb)August 17, 1987 (-marks the measuringlocation).
US EPA (Environmental ProtectionAgency)hasa Chesapeake Bay monitoringprogram

measuringwater quality parametersthroughoutthe bay at semi-monthlyintervals. As
indicatedby the August3-4, 1987 sampling,the bay was highly stratifiedand DO in the
bottomwaterwasseverelydepressed (Fig. 3). To demonstrate
the time historyof the vertical
distributionof DO and salinityin the bay, the EPA programdataand our slackwatersurvey
data, both at the stationoff the Rappahannock River mouth (Fig. 1), were combinedto
construct constantvaluecontours in thedepth-timeplane(Fig. 4). It showsthatthebay was
stratifiedthroughoutthe period. SurfaceDO was alwaysat or near saturationlevel while
hyIx•xicconditions in thebottomwaterstartedto developin May andpersisted throughout the
summer.

--1 1
--4
--7
--10
--1.3
/---
-• -
.
t
--19
/
--22
/
--28
¸
z
,
z
-34- -F
/
o) DO
-57-
-4.0 -
--430 •i0 2•0 • i 41

ø t 5'o
,,,/
o
0
z
z
--
--430 •0
I I I
20
I
50
I ,
40
I ,
50
DISTANCE NORTH TO SOUTH (kin)
Figure3: Dissolved
oxygen(mg1-1)andsalinity(psu)in themainstemof Chesapeake
BayonAugust
34, 1987 (The largertick markson abscissa
indicatemeasuringstations).

60• A Y Kuo and K Park
o I I .I . I , ,11 . I I II_,_I, ., I t ,•l I• I
-3
-5
/
-6
-7
-8
(M)
-9
-lO
-11
-12
-13
91 121 51 181 211
-1
-2
-3
.'I••I r,
(I I /1I I11
Ii )ib)
iSalinity
ii i 1
-4
-5
D
E
-6
P
T -7
H
-8
(M)
-9
-lO
-11
-12
-13 I I
12_1 151 181 2_11 2_41
dulian Days
Figure4: Dissolvedoxygenandsalinityisopleths
in depth-timeplane(The tick marksat thetopof the
figure indicatesamplingdates).

29.2.2 Time series measurements
Time seriesmeasurements of temperature,conductivityand dissolvedoxygenwere taken at

two stations,oneat the river mouthand the otherat 16.6 km upriverin the deepbasin(Fig.
1). The datawere measuredandrecordedeveryhalf hourwith HydrolabDataSounde1. One
sensorpackagewas deployedat thefiver mouthnearthebottomat 11.8 m depth,while at the
upriverstationtwo were deployed,one at mid-depth(7.5 m) and one at the bottom(16.2 m).
The meterat thefiver mouthstationwasdeployedfrom 5 Augustto 2 September.Duringthis
period,it wasrepeatedlyserviced,calibratedandredeployedfor periodsrangingfrom 7 to 10
days. Batterylife andfoulingrestrictedthe deploymentperiod,thuscreatinggapsin the time
seriesdata. The metersat the upriverstationwere not deployeduntil 25 Augustbecauseof
availability.
Currentvelocitiesweremeasuredwith in-situ,selfrecordingmetersthatweredeployedwith
taut wire mooringsat the river mouthstationand the station16.6 km upriver. Two typesof
currentmeterswere used;Inter-OceanS4 metersand modifiedBrainconHistogrammeters.
The S4 meteris an electromagnetic typecurrentmeterwith solidstatememory,someof which
are also equippedwith temperature,conductivityand pressuresensors.The meterswere set
to recordthe averagevaluesof all variablesevery 30 min. The Brainconmeter measures
currentmagnitudeby a savoniusrotor, with a vane attachedfor directionmeasurement.All
threeBrainconmetersdeployedfunctionedproperlyfor a periodof only a few daysand then
succumbed to electronicproblemsor fouling. This leavesavailabledatafor two depthsat the
fiver mouth station,and for threedepthsat the upriver station.
The velocitydatafrom eachcurrentmeterwere analyzedto determinethe principalaxis

of the flow at themeterlocation.The directionof theprincipalaxisis definedasthe direction
alongwhichthe sumof the absolutevaluesof velocitycomponents from all datapointsis
maximum.Thisdirectionwastakenasthelongitudinalaxisof a coordinate systemfor further
currentanalyses.
Threetide gaugeswereinstalledduringthisstudyat locationsspanning

the lengthof the
subestuary.Water surfaceelevationswere recordedevery 6 min. Thesedata were then
convertedto hourly surfaceelevations.
Wind datafrom Norfolk andRichmondairportswereexamined.Norfolk is located65 km

to the southandRichmondis located80 km to the westof theRappahannock River mouth.
Bothsetsof winddataexhibitsimilarwinddirection andmagnitude exceptduringperiodsof
weakwindwith variabledirection.The winddatain Norfolkis presented in Fig. 5a andused
for later analyses. Fig. 5 also displaysthe time seriesdata of water surfaceelevation,
longitudinalvelocitycomponents, salinityand DO at the fiver mouthstation.

06
oo
-06
06
O0
-06
22
"• 20
._
215 225 235 245
Julian Days
Figure5: Timeseriesdata. a) WinddataatNorfolkAirport(y-axispointsto thenorth),b) Filteredand

unfilteredsurfaceelevationat river mouthand c-e) Filteredand unfilteredtime seriesdata near the
bottomat theriver mouth(positivevelocityis in ebbdirection).The verticallinesindicatetheevents
of high salinity,low DO water intrusion.

29.3 Data analysesand massflux calculation

All timeseriesdataweresubjected
to a low-pass
filterwitha cut-offfrequency
of (36 hr)'1.
The filter wasa modificationof the low-passfilter designedby Godin(1972). The filtering
processessentiallyeliminatesthe diurnaland semi-diurnaltidal constituents
and fluctuations
of higherfrequencies.The resultingfilteredserieswereconsidered as subtidalcomponents.
The differencebetweenthe measuredtime seriesand a subtidalcomponentis consideredas
the tidal componentfor the followingdiscussion.That is
q(t) = <q>(O + Q(O
where
q = total component,or the measuredtime seriesdata,

<q> = subtidalcomponent,or the filtered time seriesdata,
Q = tidal component,
t = time.
Fig. 5 alsopresentsthe subtidalcomponents of surfaceelevation,longitudinalcurrent,

salinityand DO. The subtidalcurrent(or Eulerianresidualcurten0 was directedinto the
subestuaryalongthe bottomat all times (Fig. 5c). The averagevalue over the periodof
measurement is -6.3 cms4. Theaverage valuesof thelongitudinal
velocitycomponents at
the upriverstationwere 2.9, -3.0 and -3.7 cm s'• at depthsof 1.2, 10.0 and 18.7 m,
respectively.This agreeswith theclassicalestuafinecirculationof seawardflow in the upper
layer and landwardflow in the lower layer.
Sincethe hypoxicconditionsexistonly at the lower portionof the water column,both in

the bay and subestuary,the influenceof the bay on the subestuarymay be investigatedby
consideringthemasstransport throughthelowerportionof thewatervolumeat thesubestuary
mouth. Quantitativeestimates of masstransportshouldbe obtainedby spatialintegrationof
theproductof velocityandconcentration overtheportionof thecross-section throughwhich
the transportis landward.
M(t)= -ff• uty,z,0

cry,z,0
ayaz
where
M = masstransportrate into the subestuary,

u = longitudinal
velocitycomponent (positiveis out of subestuary),
c = concentrationof dissolvedsubstance,
A = theportionof the cross-section in whichthe transportis
into the subestuary,
y, z = the spatialcoordinatesin the cross-section.

In thisstudy,therearemeasurements of velocity,salinityandDO at onlyonepointin the

lowerlayerof the transectat thefiver mouth. No information on the spatialvariabilityof
masstransportmay be extracted.However,the temporalvariabilitymaybe investigated, at
leastqualitatively.The massflux, or the masstransport per unit area,at the currentmeter
locationmay be calculatedas:
m(t) = - u(t)c(t)
where the concentration c may be salinityor DO (or dissolvedoxygendeficit). If it is

assumed thatspatialdistributions
of u and c overthe areaA do not vary significantly
with
time, temporalvariabilityof m shouldbe the sameas that of M.
The net flux of massmay be decomposed

into two components,
i.e.,
• - - <c> - UU
wherethe overbarsdesignatethe temporalaverageover the studyperiod. The first termon

the fight handsideis the net transport
by the subtidalcomponent andthe secondtermis the
net transportby the tidal component.The datacollectednearthe bottomof the fiver mouth
were usedto calculatethe net massfluxes of salt and DO, contributedby total, tidal and
subtidalcomponents respectively.The resultsarelistedin Table 1, togetherwith theaverage
valuesof eachcomponentof longitudinalvelocity,salinityand DO. The root meansquare
valuesof tidalcurrent,andtidalcomponents of salinityandDO fluctuationarealsopresented
in Table 1. It showsthat, even thoughthe tidal currentis muchstrongerthan the average
current,the salinity and DO fluctuationsat tidal time scalesare much smallerthan theft
respective averagevalues. Furthermore, Fig. 5 indicatesthatthe salinityandDO fluctuations
at time scaleslongerthan tidal are muchlargerthanthoseat tidal or shortertime scales.
Table 1. Averagevaluesof measuredquantifiesand massfluxes
Total Subtidal Tidal Rc.•, Mean Square

Component Component Component ofTidal Component
(a) Average over 596 hours
u (cm s-x) -6.31 -6.28 -0.03 16.6
s (psu) 18.5 18.5 0.00 0.26

- u x s 119 118 1
(b) Average over 333 hours
u (cm s'x) -7.93 -7.90 0.03 17.4
0.02 0.72
DO (mg1-•) 5.63 5.61
- u x DO 37.9 41.1 -3.19

The net massfluxesof salt and DO by the tidal componentamountto only 1% and 8%
respectivelyof thoseby the subtidalcomponents (Table 1). Therefore,we needto consider
only the subtidaltransportin studyingthecharacteristics
of the estuary-subestuary
exchanges
of salt and DO in this case.
29.4 Temporalvariability of massexchanges

The subtidalcurrentwas variablewith a dominanttime scaleof severaldays(Fig. 5c). Since
thefreshwaterdischargeinto theRappahannock River wasfairly constantoverthetwo months
priorto andduringthe field measurements, thevariabilitymay be attributedto meteorological
forcing,primarilythewind. The windeventsnotonlyexertvariationsin wind stressoverthe
water surface,but also causevariationsin salinitystructurein the bay nearthe fiver mouth.
The salinitystmcturein the bay may also vary in responseto changesin dischargefrom the
Susquehanna River, the majorfreshwatercontributorof the bay. However,the time scaleof
variationsresulting from this change should be much longer than several days. The
magnitude,andits amplitudeof variation,of theobservedsubtidalcurrentare consistent with
the valuesderived from the theoreticalmodel of Hansenand Rattray (1965). They have
derivedan equationdescribingthe verticaldistributionof steady-state estuarinecirculation.
In the equation,the circulationconsistsof three modes:barotropiccirculationforced by
freshwaterdischarge,barocliniccirculationdriven by densitygradientand the wind-driven
circulation. Basedon the observedlongitudinalsalinity gradient,the model estimatesthat,
underthecondition
of 5 m s'• windspeed,
ug/u• = 0.32
whereugandu• are,respectively,
themaximum
density
drivenandwind-driven
velocities
in
the lower layer of the water column. Applicationof the formulaproposedby Hansenand
Rattray(1966) gives
u• = -3.3cms'•
Uw= - 10.2cm s'•
Thesevaluesare consistentwith the subtidalcurrentpresentedin Fig. 5c.
Furtherevidenceof wind-induced variabilitymaybe discerned

by considering Figs.5a, 5b
and 5c. Increasedlandward(or westward)currentnear the fiver bottom(Fig. 5c) always
occurredwhen a strongwind was blowingfrom the southwestquadrant. The wind drove
surfacewaterout of the fiver, and thusloweredthe surfaceelevation(Fig. 5b) and causeda
setupin the bay that drove the bottomwater into the fiver.
Therewerecharacteristic differencesin the properties

of watermasses
transported
into the
fiver overdifferentperiodsof time. On five occasions
duringthemonthof measurements, the

incoming
watermasses
werecharacterized
by lowDO andhighsalinity,
typically
3 to4 mg
1'• lowerDOand2 to 3 psuhigher
salinity
thanthose
of otherperiods.
Theimport
of low
DO, highsalinitywatermasses
wasalwaysaccompaniedby increased landwardcurrentand
southwest
wind. ComparingFigs.5d and5e with Figs.3 and4, it is evidentthatthe low
DO, highsalinitywatermasses
werederivedfromthelowerportionof thewatercolumnin
the bay. In the absenceof strongsouthwest
wind, the shallowsill off the fiver mouthblocks
the bottomwaterin the bay from enteringthe fiver, andtherefore,
the importedwateris
derivedfromthesurface layerin thebay,withhigherDO andlowersalinity.
29.5 Mechanisms
of windinducedmassexchange
Asindicated
inFig.5, theintrusion
of thelowDO,highsalinity
waterintotheRappahannock
Riverhasa timescaleof 2 to 3 days. Thisagreeswiththetimescaleof theeffectof local
windforcingin ChesapeakeBay. WangandElliott(1978)observedthatthesubtidalsealevel
fluctuations
in Chesapeake Bayhadspectralpeaksat periods of 20, 5 and2.5 days.Their
analysesassociatedthe 2.5 day fluctuations
with the localwindforcing,whilethe lower
frequencyfluctuations
wererelatedto non-local windforcingthrough theinteraction
with
coastalsea. With localwindas forcingfunction,therearetwo possible mechanisms that
contribute
tothetransportof thesubpycnocline
waterof thebayintotheRappahannock River.
For simplicity,we may look at the wind-drivencirculationaroundthe fiver mouthin two-
dimensional
space,
onein thehorizontal
planeandtheotherin thevertical
plane.
29.5.1 Horizontalplane
Fig.6 depicts
theforcing
of bottom
current
around thefivermouth.Theleftpanelis for
thewindfromthesouthwest. Surface
set-up
in twotransectsacross
thebayis considered
(Fig. 6b). The surfacecurrentin bothtransects
is towardtheeastbecauseof directwind stress
and Corioliseffect. The boundaryon the eastcausesthe watersurfaceto rise and a return
westward
flowalongthebottom.Thereis a difference
in thedegree
of surfaceset-downon
thewestern
boundary
betweensections
AA andBB. Owingto thesurface out-flowfromthe
fiver, the set-downat sectionAA will be lessthan that at sectionBB. This resultsin a
barotropic
flowawayfromthefivermouth
alongthewestern
boundary
of thebay(Fig.6c).
Therefore, thebottomwatertransported
intothefiveris mainlyderivedfromthemainstem
of thebay wherehypoxicconditions existin thesummer.On theotherhand,if windis
blowing fromthenortheast
quadrantthesurfaceset-upatthefivermouthwill belessthanthat
at sectionBB. Thisresults
in analongshorebarotropicflowtowardthefivermouth.If the
density-driven
gravitational
circulation
at thebottom of thefivermouth is strongerthanthe
wind-drivencurrent,
thebaybottomwaterwill stillbetransportedintothefiver. However,
thisbottomwaterisderived
fromnearshore regionstothenorthandsouth of thefivermouth.
Because oftheshallowdepths
inthese
regions,theimported watermass hashighDOandlow
salinity.

a). TOP VIEW

Southwest Wind Northeast Wind
BAY
BAY
A•.•RIVER
1
b). SIDE VIEW
SECTION
AA ß ,
T •
SECTION BB
T
c). TOP VIEW

of bottom current
ha > hb ha < hb
Ug
• • Uw
•
Figure6: Directionsof wind-drivencurrents

aroundan estuary-subestuary
junction.
29.5.2 Vertical plane
It wasobserved thatthebay wateris highlystratified, bothin salinityandDO in summer

months.Fig.7 depicts thewatersurface andpycnocline in a transect
across thebay(Section
AA of Fig. 6). In theabsence of wind,thepycnocline is nearlyhorizontal andis locatedat
aboutthesameheightasthesill nearthefiverentrance.Mostof thewatertransported into
thefiverby gravitational circulation
is derivedfromthebayat a heightabovethepycnocline.
In thetimeof strongsouthwest wind(Fig. 7b), thesurfaceset-upandwestward returnflow
nearthebottomforcethepycnocline to tilt upwardonthewestern sideof thebay. If thetilt
is largeenough to raisethepycnocline abovethesill, it notonlyforcesthesubpycnocline
waterintothefiver,butalsostrengthens thebottomflow intothefiver. Therefore, strong
inflowof low DO, highsalinitywaterwasobserved (Fig. 5). Similarargument hasbeen

proposedtoexplain
theobserved intrusion
of hypoxicwaterontotheshallowflanksofthebay
(Tyler,1984;Maloneet al., 1986). On theotherhand,if strongwindis blowingfromthe
northeast,
thepycnocline
will bedepressed belowthesillonthewesternsideof thebay(Fig.
7c). Thewatertransported intothefiverwill be thatabovethepycnocline,thuswith high
DO, low salinity.
rivermouth
t v no
wind /
a) I g pycnocline
/•
,• '
Figure7: The tiltingof pycnocline

in a cross-bay
sectionin response
to wind.
29.6 Discussion
In thisstudy,we havefocusedon the importof watersinto the subestuary

throughthe lower
layerof the watercolumn. Analysesof observed dataindicatethattransportsof water,DO
and salinityby subtidalcomponents are an orderof magnitudelarger than thoseby tidal

componentsof watermovements.Sincewe havemeasurements at onlyonepointin thewater

columnno quantitativeestimateof theamountof masstransportmaybe made. However,the
temporalvariability of mass transportmay be assesseA assumingsimilarity in spatial
distribution over time.
Two distinct characteristicwater masseswere observedbeing transportedinto the

subestuary, the low DO, high salinitysubpycnocline water and the high DO, low salinity
surfacewater. Throughsimultaneous comparison of winddatawith surfaceelevation,subtidal
current,salinityand DO, we can concludethat wind is the dominantfactorinfluencingthe
propertiesof importedwaters. The hypoxicwater was transportedinto the subestuaryas
episodiceventswhenstrongwindwasblowingfromthesouthwest.Duringtheseperiods,the
pycnoclinein the westernside of the bay rose above the sill at the fiver mouth and
subpycnocline waterin thebay 'rushed' into thefiver alongthefiver bottom. The landward
flow of gravitationalcirculationwasfurtherenhancedby thewind-drivencurrentandincreased
longitudinal,salinitygradient.
Otherstudiesalsodemonstrated the temporalvariabilitiesof bottom-waterintrusionsover

the entrancesills of estuariesor subestuaries.Cannonet al. (1990) reportedthat the onsetof
the intrusionsinto PugetSoundwere associated with the combinedeffectsof neapfidesand
the horizontaldensitygradientacrossthe sill. The regularspring-neap cycle was the result
of mixingeffectsover the sill; however,the episodicoccurrence of bottom-waterintrusionin
winter was attributedto wind eventson the PacificCoast. Other causeshavebeenproposed
for the variabilityof the densitygradientacrossthe sills of fjords (e.g., Gade and Edward,
1980). Theseincludeseasonalvariationin densityandupwellingoutsideof the entrancesill.
For a fjord estuary,bottom-waterintrusionis consideredto play a dominantrole in the
replacement of water below sill depth,and the flushingof contaminants. On the otherhand,
the intrusionsof bottom water from ChesapeakeBay into its subestuaries could mean the
intrusionof hypoxicor anoxic water in the summerseason,adverselyaffectingthe water
qualityin subestuaries.Sanford(1988) alsoreportedthatintrusionof lower-layerwater from
thebay into theChoptankRiver occurredasan episodic,wind-driveninternalsurgeratherthan
as a slow,steadyflow. Sincethe ChoptankRiver is on the easternsideof Chesapeake Bay,
the intrusioneventsoccurredwhen strongwind was blowingfrom the northeast.
The episodicnatureof subpycnocline-water intrusionsoften resultsin the reversalof the

longitudinalsalinity gradientin the lowestreachof the subestuary.This would posesome
difficultiesin specifyingboundaryconditionsfor numericalmodellingof subestuaries.A
priori knowledgeof conditionsin the main estuaryis required,or the modeldomainwould
needto be extendedto includeportionsof the main estuary.

Acknowledgements
This studywas fundedby the Commonwealth of Virginia as part of Chesapeake Bay
Initiatives.The authorswishto expresstheirappreciation
to Mr. K. Curlingfor providingthe
dataof the main stemof the bay, collectedunderthe US EPA Chesapeake Bay Program.
29.7 References
Cannon, G. A., J. R. Holbrook, and D. J. Pashinski,1990: Variations in the onsetof bottom-water

intrusionsover the entrancesill of a fjord. Estuaries,13(1), 31-42.
Elliott, A. J., 1978: Observationof the meteorologically

inducedcirculationin the Potomacestuary.
Estuarine Coastal Mar. Sci., 6, 285-299.
Flemer,D. A., G. B. Mackierman,W. Nehlsen,andV. K. Tippie, 1983: ChesapeakeBay: a profile of

environmental change.U.S. EnvironmentalProtection
Agency,Washington, D.C., 199pp.
Gade,H. G., andA. Edwards,1980: Deep-waterrenewalin fjords. In H. Freeland,D. Farmer,andC.

Levings(eds.),FjordOceanography,PlenumPress,New York,453-489.
Godin,G., 1972: The analysisof tides. Universityof TorontoPress,Canada,264 pp.
Hansen,D. V., andM. Rattray,Jr., 1965:Gravitational

circulationin estuaries.J. Mar. Res.,23, 104-
122.
Hansen,D. V., andM. Rattray,Jr., 1966: New dimensions

in estuaryclassification.Limnol.Oceanogr.,
11, 319-326.
Kuo, A. Y., andB. J. Neilson, 1987: Hypoxia andsalinityin Virginia estuaries.Estuaries,10(4), 277-
283.
Malone, T. C., W. M. Kemp, H. W. Ducklow,W. R. Boynton,J. H. Tuttle, and R. B. Jonas,1986:

Lateralvariationin the productionandfateof phytoplankton
in a partiallystratifiedestuary.Mar. Ecol.
Prog. Ser., 32, 149-160.
Newcombe,C. L., and W. A. Home, 1938: Oxygenpoor watersof the Chesapeake

Bay. Science,88,
80-81.
Sanford,L. P., 1988: Bathymetriccontrolof tributary-mainstem

interaction:the caseof the Choptank
River. EOS, Trans.Am. Geophys.Union, 69, 1256 (abstract).

Schroeder,W. W. (convener),1985: The occurrence of hypoxicand anoxicconditionsin estuariesand

coastalenvironments.Abstractfor theEighthBiennialEstuarineResearchConference,
Estuaries,8(2B),
43A-47A.
Tyler, M. A., 1984: Dye tracingof a subsurfacechlorophyllmaximumof a red-tidedinoflagellateto

surfacefrontalregions. Mar. Biology,78, 285-300.
United NationsEducational,ScientificandCulturalOrganization,1983: Algorithmsfor computationof

fundamentalpropertiesof seawater.UNESCO TechnicalPapersin Marine Sci., 44, Pads, France,53
PP.
Wang, D. P., A. J. Elliott, 1978: Non-tidal variabilityin the ChesapeakeBay and PotomacRiver:
evidencefor non-localforcing. J. Phys.Oceanogr.,8, 225-232.

30
Interdisciplinary study on the tidal front in

the Bungo Channel, Japan
T Yanagi, O Matsuda,S Tanabeand S Uye
Abstract
An interdisciplinary
studywascarriedoutaroundthetidalfromin theBungoChannel,Japan.
Temperatureand salinitydifferencesacrossthe tidal front are 3'C and 0.3, respectively.
Nutrientsare richer in the well-mixedregionthanin the surfacewater of stratifiedregion
wherethey havebeenconsumed by phytoplankton.Man-madeorganicchemicalssuchas
PCBs are concentrated in the surfacewaterjust on the front andthey are accumulated
in the
planktonwhich are gatheredthere. Neustonsuchas marine skatersand Vellela lata are
accumulatedon the front. The tidal front plays as a barrier for some oceanic and neritic
pontellidcopepods.
30.1 Introduction
The remarkabletidal frontis developedfrom May to Augustin thenorthempartof the Bungo

Channel,theSetoInlandSea,Japan(YanagiandKoike, 1987). The detailedphysicalstructure
of this tidal frontis akeadyrevealedandits temporalandspatialvariationsare well explained
by the variableefficiencymodel(Yanagiand Tamara, 1990).
However, the chemicaland biologicalcharacteristics

of this tidal front have not been
clarifiedyet. Hencean interdisciplinary
studywascarriedout in collaboration with physical,
chemicaland biologicaloceanographers.
30.2 Observations
The intensiveand comprehensive field observations

of the tidal front at the northernpart of
the Bungo Channel were carded out on board T & R/V Toyoshio-Maruof Hiroshima
Universityfrom 16 to 19 June 1986, from 22 to 24 May 1987 and on 21 May 1988. The

618 T Yanagi, O Matsuda,S TanabeandS Uye
bottomtopography
andthemapof theobservation
siteareshownin Fig.1.
131ø20'E 132ø00'E
33ø40'N
THE SEA OF IYO
KUNISAKI
60
PENINSULA
SHIKOKU
..
60
:!: HAYASUI
BEPPU STRAIT
¾• AJIMA
KYUSHU •v
)1) o+5
JAPAN• c;Pti +6
•
•oo• 1
BUNGO CHA'NNEL
32ø50'N
(Km)
Figure1: Observation
stations
in theBungoChannel.Numbered
contours
showthedepthin metres.
Verticalprofilesof temperature
and salinityweremeasured at 11 stations
alongthreelines
(W1-W3, C1-C5 andEl-E3) from 16 to 19 June1986,at 4 stationsalongone line (P1-P4)
from22 to 24 May 1987andat 6 stationsalonganotherline (1-6) on 21 May 1988by using
a CTD (ALEC Model AS-100). Water andplanktonsampleswere collectedat thesestations
with use of a pair of Van Dom water bottlesand planktonnet, respectively. Water
temperature,salinityandchlorophylla weremeasuredcontinuously at a depthof ca. 3 m
alongtheaboveobservation lines(waterwaspumpedup from 3 m depth)usinga multiple
environmentalsenser(Martec Mark VI) anda fluorometer(TurnerDesigns)in orderto detect
the exactpositionand sharpness
of the front.
30.3 Physicalaspects
Water temperatureand salinityabruptlychangedfrom 18C and 33.8 to 22'C and 34.3,
respectively,
acrossthe frontalregionwith a widthof ca. 500 rn on 18 June1986. Density

T Yanagi,O Matsuda,S TanabeandS Uye 619
alsochangedfrom 24.5 to 24.0 in sigmat acrossthe tidal front (Fig.2).
('C)
22.0
c-1 c-2
....
c-3 c-4 c-5
T('C)
t34.4
20.
o[
18.0[ t34.2
34.0
33.8
ira)
ø1 ' X-•' '24.o
18.0
II •..• :____'----.'--
:Z-_._2.4.2
. 24.4
....:•' 2 .0
:(•,'/. '....
lOO
[ :::' June 18.1986 .... ::"'""

......""
(Km)
Figure2: Horizontaldistributionof watertemperatureca. 3 rn below the seasurfaceon 18 June1986.

Thin line showsthe shipcruisetrack (left). Horizontalvariationsof water temperature,salinityand
density3 rn belowthe seasurfaceandverticaldistributionof densityalongthecentralobservation line
on 18 June1986 (right).
The waterin the northernpartof thisfrontwaswell mixedverticallydueto the strongtidal

currentnear Hayasui Strait while that in the southernpart was stratifieddue to weak tidal
current. The transparency (secchidiscdepth)was 14 rn in the northernpart of the front and
it was 16 rn in the southernpart.
The positionof the front was noticedeasily from the researchvessel,as the front was
accompanied by a narrowband of foam and variouskinds of drifting materialon the sea
surface.Suchaccumulation of driftingmaterialssuggests theexistence
of surfaceconvergence
at the tidal front. This basicphysicalstructurewas maintainedconsistently throughoutour
observation periodswith only minor spatialchangesaccordingto tidal phase. The detailed
physicalcharacteristics of this tidal front are describedby Yanagi and Tamaru(1990).
30.4 Chemicalaspects
Representativevertical distributionsof DIP (DissolvedInorganicPhosphorus)and PP
(ParticulatePhosphorus)on 24 May 1987 are shownin Fig.3. The front existedexactlyat
Sta.P3, while StasP1 and P2 were situatedin the well-mixedregionand Sta. P4 in the

620 T Yanagi,O Matsuda,

S TanabeandS Uye
stratified
region(Matsuda,1990).Theconcentration
of DIP ismoreimportant forthegrowth
of phytoplanktonthanthatof DIN because
phosphate
is a limitingfactorin the$etoInland
Sea(Joh,1986). DIP wasabout0.3 pg-at/1in the well-mixed
regionandit abruptly
decreased acrossthe tidal front to about0.15 pg-at/lin the surfacelayerof the stratified
region.On theotherhand,PP waslow (about0.1 pg-at/1)in bothsidesof thetidalfrontbut
high(about1.15pg-at/1)at the tidalfront. This is associated with higherphytoplankton
biomass asevidenced by higherconcentrationsof chlorophyll
a andpheo-pigments justat the
tidalfront(Fig.4). Thedissolved oxygenconcentration washighin thesurface layerat the
tidalfront(Fig.5). Thesefactsindicatesthattheactivityof phytoplanktonis highat thetidal
front.
P2 P3 P4
W ,
L
2O
. .X
0.2
4O
• 0.25'""
6O
8O
100
2• •ay • 987
(Ill)
Stn. P1 P2 P3 P4
2O
ß 0.10/•1 15 1.I0
0.10
4O
ß .
lOO
0
""";
'•t/• 5
) 10k•
(m)
Figure3: Vertical
distributions
ofDIP andPPfromSta.P1
toSta.P4
on24May1987.Tidalfrontexists
on Sta. P3.

T Yanagi, O Matsuda,S Tanabeand S Uye 621
Continuousobservationsof water temperature,salinity, chlorophylla and beam

transmittanceof thesurfacewaterwerecarriedoutbetweenStations1 and6 on 21 May 1987
(Fig.6). On thisoccasion,
chlorophyll
a concentration
washighandbeamtransmittance was
low at the tidal front.
Stn, P1 P2 P3 P4
20
40
80
lOO
(m) 24 May 1987

chlorophyll
a__
(/•
g/11
)
stn, P1 P2 P3 P4
L •.B• •3
/
20
40
6O
2O
lOO
(m)
Figure4: Verticaldistribution
of chlorophyll
a andpheo-pigments
fromSta.P1to Sta.P4on24 May
1987. Tidal front exists on Sta.P3.
Thereasonsfor thehighconcentration
of chlorophyll
a at thetidalfrontmaybe explained
asfollows.Phytoplanktoncannotgrowsuccessfully
in bothsides of thefront.Because
they
can not stayin the shalloweuphoticzonedueto strongverticalmixingin the well mixed
region,andnutrients arelackingin thedeepeuphotic
zoneof stratified
region. Ontheother
hand,phytoplankton
cangrowsuccessfully
nearthefrontwheretheextentof verticalmixing

622 T Yanagi,O Matsuda,S TanabeandS Uye
and the supplyof nutrientsare moderate.
Table 1' Concentrationof DDE and PCBs in planktoncollectedfrom the surfacesea water in the
northernpart of the BungoChannel(Tanabeet al., 1990).
Concentrations of p,p'-DDE and PCBs in ptank•on corrected

from surface seawater of Bungo Channel
SampLingstation p,p'-DDE concentration PCBs concentration
(ng/g on dry wt. basis) (ng/g on dry wt. basis)
C- 1 < 1.2 9.9

E-2 5.5 26.6
C-5 < 1.2 5.4
Stn, P1 P2 P3 P4
or ß .-- • •___---. II ' i
.../'"' •'\ 11)2 .

2o
' ß 9.o •9 • '
• ! \ :•oo
•
4o
6o
8O
DO (%)
.1.oo ß
(m)
of dissolved
oxygen
fromSta.P1 to Sta.P4on24 May 1987.Tidalfront
exists on Sta.P3.

623
T Yanagi,0 Matsuda,S TanabeandS Uye
Stn.1 2 3 4 5 6
I I I I i I
19.0
T e m p . (øC)
18 0
16 0
t -
I;,, ,; ':T: fi_k•, • f I• I
342 S a 1. (%) ;I ill, •: i ;i' ';'ii,i
33.8 '- - -"

21 Ma'! 1988
Ch1. a(l•g/t2)
i ---- ___2_-
5.0- t --
---- i__'•-
4.0
3.0 ---i•
2_
z.oT
1.0 I___
0.0 . ._-----:
.....
60
70 - _
80
100
90 !_4_•
0 5 10km
Figure6: Horizontalvariations
of watertemperature,
salinity,chlorophyll
a andturbidity(reciprocal
of
beam transmittance)ca. 3 rn below the seasurfacefrom Sta.1 to $ta.6 on 21 May 1988.

The concentrations of HCHs (Hexachlorocyclohexanes) were high in the surfacelayer of

the tidal front as shownin Fig.7 (Tanabeet al., 1990). The differenceof 0.2 ng/l in Fig. 6
is significantbecause variationin reproducibility
is lessthan 10%. Our resultsindicatethat
the tidal front accumulatesoil and particlesin the surfacelayer due to the surface
convergence.The surfaceconvergence at the tidal front wasnot observeddirectlybut the
accumulationof foams and many kinds of drifting matter suggestedits existence. The
physicochemical properties
of HCHs suchas lipophilicandhighparticleaffmityenablesthe
enrichment of this contaminant at the tidal front. The concentrations of DDE and PCB in the
planktonof the surfacelayeraroundthe tidalfrontwereexaminedandthe resultsare shown
in Table 1. The concentrations of DDE andPCB in the planktonwere low in both sidesof
the front,StasC-1 andC-5, but highat Sta.E-2 on the tidal front. Theseresultsalsosuggest
thatthe man-madeorganicssuchasPCB andDDE whichare enrichedin the surfacelayerof
the tidal front are bioaccumulated
in the plankton.
132ø00'E
0 m
Stn.
C-1 C-2 C-3 C-4 C-5
! I I ! I
50
lOO
(ng/[)
Figure7. Horizontaldistributionof HCH concentrations in the surfacelayer on 18 June1986 (upper).

Vertical distributionof HCH alongthe centralobservationline on 18 June1986 (lower).

T Yanagi,O Matsuda,
S TanabeandS Uye 625
.• b•nd
/ 1000
m
2
Marine skaters 0
o ind./1000
m2
i:!iii!i!iiiiii!ii!i:i:i:i:!:i:i:i:!:i:i:i:i:!:!:i:!:i:
......... oC oEO///
'"'"'•'"•:•.•••"..••"...••ii:•i',
'""•:•'."•
...... •c •t••••_•
W30 •
0C4
0C5
Velella lata 0 100ind./I000m2

0 0 ind./ 1000 m2
33.8 •o
870o
• • 2 18øC
0C5
Figure8' Horizontaldistribution
of marineskatersandVelellalata in the surfacelayeron 18 June1986.

626
T Yanagi,O Matsuda,
S TanabeandS Uye
o Labidocera rotunda
NERITIC PONTELLIDS ,,, Pontella

rostraticauda
[] Pontellopststenuicauda
E1 o o tnd./100
m3
-- t-10
c• ©"•/ -•-•oo
• C2 18øC
• C3 o
C4
-- O Labidoceradel:runcata _
OCEANIC
PONTELLIDS o L. acu•rrons
,• Pontella securtfer
[] Pontellopsts vttlosa
....:...• '•'
• Pontetllna
P mortl
olumata
E1
'""
• 3.8 V•,•E/•
•.•Xo
i•"'"""'"'•:':':•""•••
•o
• 2 __ '•"'•øø•
Jl
--11--100
C1 V -- 1-10
I!-''"'"'"--'•""""•"•:':'.'•:•••U:i::.,•
'18 øC _ • " •,.x,,,
•
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
,•. /•
"h •
E318•C
/
=======================
.......
18
•c •""
::i:i:i:i:i:!:!:i:i:!:i:!:i:i:i:i:!:i:!:!!!:i:!:!:!::"'• c3
:!:::':'"'
"':':':':'•:•i:i::•/ v o4
?:?!:i:i:i:i:i:i..........
:i:[:[:?i:!:i:?i:[:i:i:i:i:i:i:55i:[:i:!:55i:i:55i::':':"' v
Figure
9: Horizontal
distribution
ofneritic
andoceanic
Pontellid
copepods
inthesurface
layer
on18
June 1986.

627
T Yanagi,0 Matsuda,S TanabeandS Uye
30.5 Biologicalaspects
Epineuston
(whichliveson theseasurface)
suchas marineskaters
whosespecies
is not
identifiedandpleuston
(whosebodyexistsacross
theseasurface)suchas Velellalata are
accumulated
aroundthe tidal front as shownin Fig.8 (Uye et al., 1990).
Both marineskatersand Velella lata are warm waterspeciesandthey are considered

to
be advectedfrom the Pacific Ocean and accumulatedaround the tidal front in the Bungo
Channeldueto thesurface convergence. Althoughhyponeuston (whichlivesjustbelowthe
seasurface)suchaspontellidcopepods werenotaccumulated aroundthetidalfront,nedtic
pontellids
wereobserved mainlyin thenorthernpartof thefrontandoceanic pontellids
in the
southern
partof thefrontasshown in Fig.9. Suchresultsmeanthatthetidalfrontplaysa role
of barrier of the distributionof hyponeuston.
P1 P2 P3 p•, P1 P2 P3 P•
0 ! I I 0 . . .I,
20 •s
3
2C ß
_
2O
40 2o 40 0
60
i Zooplankton
i. •0
( Number/I )
. .
8 .. ß :: : .....
CopepodNauplii ( Number/I )
pl P2 P3 p•,
I I • I
60
$0
,..........
.,.6:
. . .....;::...
Copepod Adult ( Number/I )
ß .
of zooplankton,
copepod
naupliiandcopepod
adultsfrom Sta.P1to
$ta.P4 on 24 May 1987. Tidal fxontexistson $ta.P3 whichis shownby an arrow.
Figure10 showstheverticaldistributions
of totalzooplankton,
copepodnaupliiandadult
copepodsacrossthetidalfronton 24 May 1987. Theirpopulation
densities
werehighin the
surfacelayer of the stratifiedregion.

Egg productionexperimentof a copepod,Palacalanussp. was carriedout with use of

sampledsurfacewater around the frontal region and its result is shownin Fig.11. Egg
productionrate is positivelyrelatedto chlorophylla concentration. The sameresult is
obtainedat the frontal zone in the NorthwestNorth Sea (Kiorboe et al., 1988). Egg
productionrate of a copepod,Palacalanussp., is consideredto be high on the tidal front
wherechlorophylla concentration is high as shownin Fig.4. However,the concentration of
copepods (naupliiandadults)wasnotnecessarily highon thefront. Further,Uye et al. (1990)
demonstrates thatthe densityof copepodnaupliiis lowerat the frontthanin bothsidesof the
front. Suchdiscrepancy suggeststhatcopepodnaupliiwhichmightbe producedat highrate
at the front are intensivelypredatedby carnivoressuchas Velle lata andjelly fish whichare
accumulated in the surfacelayer of the front.
Similarobservational resultsof high concentrationof chlorophylla and low population

densityof zooplankton at thetidalfrontwereobtainedin theEnglishChannel(Holliganet al.,
1984) and in the NorthwestNorth Sea (Kiorboeet al., 1988).
50
•-•
O•
'-•::n
30
o
oo
o• 20
o ---
3 6X-16 9
,
o o
LU 10
i i i
1.0 2.0 3.0
Chlorophylla Concentration(pg/i)
Figure11. Relationship
betweeneggproduction
rateof thecopepod
Paracalanus
sp.andchlorophyll
a concentration.

T Yanagi, O Matsuda,S Tanabeand S Uye 629
30.6 Discussion
The tidal front is generatedin a transitionzone from the well-mixedregionto the stratified
one.Nutrientsare rich in the well-mixedregionandpoorin the surfacewaterof the stratified
region.On the otherhand,the biomassof phytoplankton is high at the edgeof the stratified
regionbecause the verticalmixingis too strongin the well-mixedregionfor phytoplankton
growth. The oleophilicchemicalsare concentrated alongthe tidal front and bioaccumulated
in theplanktonwhicharealsogathered alongthetidalfrontdueto thestufface convergence.
Though the tidal front is a physicalstructure,it also affectsthe distributionof chemical

compounds and planktonicorganisms as schematically shownin Fig.12. It actsas a barrier
for somezooplankton, i.e. oceanicspeciesare foundonly in the southern partof the frontand
neritic speciesin the northernpart. At the same time, the tidal front concentratesthe
particulatematerialsincludingliving planktonalongthe frontalregionin the surfacelayer
wherethechemicalmaterialareaccumulated in theorganisms of highertrophiclevelthrough
the food chains.
Coastal
area/T•dalfront
low
water temperature
/
_ /-'
/
r•
Offshore
area
low
salinity
// / hwater
iwgah
richnutrient.1•/h?_h /..
temperature
•/ PCBs.///' highsailnit
lowChl.a ,"'/high ,'/"/- '"-'7' Y
,,/' '•.•'h'l
• ,'/•'/ poornutrient
neriticzooplankton
//'
_h•gh neuston,/
Chl.a..// low
I --Chl.a-'-'
//abundances/'
•'• /_l,o_w
n,
aupli!/
oceanic
zooplankton
well-_.m
..... / ", ..
euphotic layer
Figure12. Schematic
representation
of thephysical,
chemical
andbiologicalstructures
of thetidalfront.

63O T Yanagi, O Matsuda,S TanabeandS Uye
The moreintensivestudyon thehighertrophicorganisms suchasjelly fishandfisharound

the tidal frontwill be n<essaryin the future. Thoughthe physicalstructureof tidal frontis
consistent fromlatespringtill latesummer,its biologicalrole seasonally changes.Therefore,
the seasonal variationin the biologicalrole of the tidal front needsto be revealedin orderto
clarifythe <ological meaningof the tidal front in the coastalsea.
Acknowledgement
The authorsexpresstheir sincerethanksto Prof.T.Onbe and Mr.S.Ohtsukaof Hiroshima
University for their commentson zooplanktondistributionand the crew of T & R/V
Toyoshio-Marufor theirkind cooperation
duringthe field observations.
30.7 References
Holligan,P.M., R.P. Harris,R.C. Newell,D.S. Harbour,R.N. Head,E.A.S. Linley, M.I. Lucas,P.R.G.

Tranter and C.M. Weekley (1984): Vertical distributionand partitioningof organiccarbonin mixed,
frontal and stratifiedwatersof the EnglishChannel.Mar.Ecol.Prog.Ser., 14, 111-127.
Joh,H. (1986): Studieson the mechanismof eutrophication

andthe effectof it on fisheriesproduction
in OsakaBay. Bull. OsakaPrefecturalFisheriesExperimentalStation,7, 1-174.
Kiorboe, T., P. Munk, K. Richardson,V. ChristensenandH. Paulsen(1988): Planktondynamicsand

larval herringgrowth,drift and survivalin a frontalarea.Mar.Ecol.Progr.Ser.,44, 205-219.
Matsuda,O. (1990): Chemicalprocessaroundthe tidal front.In "The Scienceof Siome"ed.byT.Yanagi,

Koseisha-Koseikaku, Tokyo, 37-50.
Tanabe,S., S. Hanaoka,T. Yanagi, H. Takeokaand R. Tatsukawa(1990): Persistentorganochlorines

in the coastalfront. Mar. Pollut. Bull., (in press)
Uye, S., T. OhtsukaandT. Onbe (1990): Biologicalprocessaroundthe tidal front. In "The Scienceof
Siome"ed. by T. Yanagi,Koseisha-Koseikaku, Tokyo, 78-100.
Yanagi, T. and T. Koike (1987): Seasonalvariationin thermohalineand tidal fronts,SetoInland Sea,

Japan.ContinentalShelf Research,7, 149-160.
Yanagi, T. and H. Tamaru (1990): Temporaland spatialvariationsof a tidal front. ContinentalShelf

Research, 10, 615-621.

31
Hydrodynamic modelling for a tidal power

project
T L Shaw
Abstract
This paperreviewsthe programmeof hydrodynamicstudiescarriedout to datein connection

with the developmentof theproposedSevernBarragetidal powerscheme.It alsooutlinesthe
furtherwork consideredto be necessary
beforea decisionto constructthe barragecouldbe
taken.
The fidesandhencetheprocesses whichrelateto thesearefundamental

to theperformance
of the project, both as an energy producerand in respectof the acceptabilityof its
environmentaleffects. For example,salinityvariationsand the consequencesof sediment
transportdeterminemuch aboutthe regimewhich the barragewould affect and create. The
work done to date on thesesubjectsis summarised.
The accuracyrequiredof predictionsand the availabilityof methodsto achievethis are

considered.
31.1 Introduction
The recentdecisions of theUK Govemmentto curtail,at leasttemporarily,thedevelopment

of nuclearpowerstationsand to expandsupportfor renewableenergyschemeshavegiven
addedimpetusto the possibilityof exploitingwind, wave and tidal sourcesfor commercial
purposes.Each of thesesourcesis capableof supplyingmore than 10% of presentUK
electricitydemand,henceis regardedas potentiallysignificantin nationalterms. However,
only tidal powerhasa long trackrecordand is capableof development in large units; the
'smallis beautiful'conceptis bettersuitedto powerproductionat a localratherthannational
scale,whenit is organisationallycumbersome andeconomically dubiousto promote.
The fidesof the SevernEstuaryare thehighestin Europeandthe cross-sectional

profile
of thisestuaryis generallywell suitedfor powergenerationfrom thissourcein termsof the

632 T L Shaw
depth and width required to accommodatethe number and size of turbinesand other
equipmentrequiredto harnesstheavailableenergy.Eitherexcessive or insufficient
depth
and/orwidthcanintroduceseriouscapitalcostpenalties
to thistypeof project.Fig. 1 shows
the locationproposed
for the SevernBarrage(D.En. et al, 1989); thisis towardstheseaward
endof whatis knownas the SevernEstuarybut which,as we shallsee,hasfeatureswhich
makemuchof it moreakinto an opencoastline thanto a fiverestuary.
•nmOD
Severn Bridge
NEWPORT
;Avonmouth
• ///iI ••// BRISTOL
i __
L_ CARDIFF
Laverhock
Point
///
if Flat
[---:::-;i
• ,,' •' Holm
, /
/
to r.• Brean
Down
Bridgwater
_
ß_ -10
0 5 1Okra
i I I
Figure1' Barragelocation
A barrage
in thislocation
wouldproduce some7-8%of thepresentelectricity
demandof
England
andWales,equivalent totheoutputof 2-3 majorthermal
powerstations.
Thiswould
makeit sometentimeslargerin sizethantheproposed
Mersey
Barrage,
thirtytimeslarger
thanthe RanceBarragein Brittanycommissioned
in 1966,andmorethana hundredtimes
largerthanmostpossible
UK tidalpowerschemes
(BinnieandPmrs,1989).Allhough
tide
rangeis a factorin thesecomparisons,
thedifferences
from siteto sitebetweenthesearesmall
comparedwith the basinareaswhicheachschemewouldcreate,ie., with the live water
volumes available
to eachsetof turbines
(volumes
whichlargely
determine
theappropriate
installed
generating
capacities).
Thebasin formed
bytheSevernBarrage
wouldhavea high
water surface area of over 500km 2.

T L Shaw 633
Thispaperoutlines
theworkdoneto determine
thehydrodynamic conditions
in whichthis
barragescheme wouldhaveto be constructed
andin whichit wouldoperate.Thelargesize
of theestuary
meansthatwaveclimatebecomesanimportantconsideration,
in additiontotide
rangeand associated
currents.
31.2 Tide regime

Earlierstudiesof theproposed
barrage(SevernBarrageCommittee,
1981)madeuseof 2-D
numericalmodelsof the SevernEstuaryand BristolChannel,one of thesemodelsextending
on to the ContinentalShelf (IOS, 1980/1). Thesemodelsconfirmedthe importanceof good
representation
of coastalbathymetryandestuarine
detail,alsotheseaward
tideregimesothat
thefull implications
of thebarrageandits methodof operationwereproperlysimulated.
I i i • i, i I . i
Figure2: Representative
fide andestuarywaterlevel curves- RanceEstuary
This informationalloweda more targetted2-D plan modelto be developedfor the recent

studyof thisbarrage(WRc, 1989a). The aim of thisworkrelatedcloselyto theinvestigation
of a rangeof issuesnot hithertostudiedin detail,for exampletidal flow patternsand their
consequences duringbarrageconstruction,variationsof salinityandthedispersion
of effluents.

634 T L Shaw
Thestudywaslessconcerned withrefiningthedetailsof theworkdonepreviously. The

decision
to proceedin thiswaywastakenbecause of theneedto improveawareness of a
number of issues
of particularenvironmental importance, notleastbecause of thepublic
concernnowattached to thisaspect
of all newprojects.It wasalsorecognisedthatattempts
toimprove theaccuracyof tidemodelling couldnotsensiblyprogress
untilbarrage operating
procedures
hadbeenspecified, in particular
useof theturbineequipmentin reverseaspumps
aroundthetimeof hightidetoraisethelevelof thebasinabovethatwhichgravity refilling
alonewouldachieve.Fig.2 shows thisoperation asit ispractised
at theRanceBarrage in
Brittany;
theamount ofpumping doneclearly relates
totiderange,beingleastonspring tides
andmorewhentherangeis lessthanaverage.
Fig 3 showsthefourlevelsof modellingder'tuition

used,thesebeingselected
to suitthe
requirements
forinformation
in eachpartof theestuary.Theperformanceof themodelwas
checkedagainst
simultaneous
tidegaugedatarecorded at24 stations.
Fig4 shows a selection
ofthecomparisions
achieved
bythemodel
whenit hadbeendeveloped
totheformaccepted
forusethroughout
theinvestigation.
Thevalues
of thebedroughness
parameter
variedover
thewholeestuary
andwithfiderangebyfactors
of 3-4,aswouldbeexpected
(WRc,1989a).
• l/-• 14
Scl :5963,000
6 7 8 $ I0 I 1213141516I118192021222324252•2?2823303132333435• 3?
32
31
30
29
28
Stackpol 27
'l II
25
24
23
II 22
II 21
II 20
I I
I I 8
II 7
II
II 5
4
3
2
1
0
8
7
5
4
3
2
1
I 2 3 4 5 t• 7 8 $ I0 11 12 13 14 15 It• 17 18 I$ 20 21 22 23 24 25 2S 27 28 2g 30 31 32 33 34 5 3• 3?
Figure3a: 2-D tide modelgrids4500m + 1500m

T L Shaw 635
MODEL GRIDS - 1500 m. + 500m.
',',I I I II .......
..........................
Figure3b: 2-D tide modelgrids1500m+ 500m + 167m
It wasrecognised thatthemodelassumed unchanged andunchanging bathymetry,notonly

overtheperiodsincethe mostrecentsurveysof eachpartof theestuaryweredonebut also
in respectof the consequencesof the barragefor the distributionof sedimentdeposits.The
modelwasthereforeusedto determine theextentto whichlocalalerationsin depths,whether
by accretionor erosion,wouldsignificantlyaffectanyaspectof thetiderange.No discernible
changes werenotedwithintheconsiderable rangeof depthchanges postulated
to occurwithin
the mainSevernEstuary,thoughthe effectsof expectedaccretionin thefiver estuaries could
be significanL
Anotheruncertainhydrodynamic factornecessarilyspecifiedin the model is the rate of

dispersion
of momentum.Froma rangeof typicallyaccepted values,a suitablenumberwas
chosenfor use in all conditions,includingthosesimulatingthe regimeduringbarrage
construction
and operation.The sensitivityof the chosenvalueto the conditions simulated
was acceptablysmall.

636 T L Shaw
./
HRS AFTER H.W.
AT PORT ISAAC
BARRY HINKLEY POINT
MODEL + ---.
OBSERVED E)
NEWPORT PORTBURY/AVON
eeeee
coooeec e
;HARPNESS
EPNEY
Figure4: Springtides- observedandmodelwater level data

T L Shaw 637
Tide modellingis usuallyemployedto showfeatureswhichoccurduringa singletide or

asa resultof a relativelyshortsequence of cycles. Many aspectsof theeffectsof the barrage
havebeenresolvedin thisway. However,theaccountof theevolutionof the SevernEstuary
(D.En. et al, 1989) whichprincipallyoccurredin responseto risingmeansealevel sincethe
last ice age,confirmsthe importanceof alsoconsidering the longertimescalebecauseof the
changewhich the barragewould make to meanwaterlevel in the basinarea and the effects
which,by inference,thisis likely to have. Fig 5 showsthischangeat representativelocations
aboutthe Estuary.
HRS AFTER H W. o
AT PORT ISAAC
-2
-a
-,
HINKLEY POINT IlARRAGE - •Irlt PUMPING X WESTON SUPER MARE

WITHOUT BARRAGE 3t-
OLDBURY EPNEY
Figure5: Springtidecurves- withoutandwith barrage- with pumping
The fact that transformationfrom the full range(whichmodellingshowswouldbe little

changed duringbarrageconstruction) to a muchreduced rangewouldprobablybe madeover
a periodof notmorethana weekor so meansthatmanyfeaturesof theestuarywouldquite
suddenly be outof regimewith thecontrollingtidalinfluence.A periodof adjustment would
follow,extending to decadesandperhaps to centuries,
andwhichwouldto someextentdepend
on how thebarragewasoperated (eg Fig 2). Possiblemethods of predictingthesechanges
basedon the use of regimeconceptsand field surveydatahavebeenconsidered (D. En.,
1989). They differ significantly from conventional modellingmethodsand are still at the
formativestage,mainlybecause theyhavenothithertogenerally beenneededfor predictive
purposesin connection withproposed engineering works. The widelyanticipatedbutstill to

638 T L Shaw
be provenworldrise in meansealevel forecastto occuroverthe next50-100yearsis adding

momentumto work in this areabecauseof the effectsit will haveon coastalprocesses
and
seadefenceprovisions.
Earlier work on the proposedSevernBarrage(eg SevernBarrageCommittee,1981)gave

more attentionto tide conditionsin the main SevernEstuarythanin the fiver estuaries(Fig
1). However,manyof the reasonsfor wishingto know moreaboutthe implicationsof the
barragenow relate to its likely effectsin the fiver estuaries,eg, changesin salinity,water
quality,turbidity,sedimentmovementand flood risk. The presentgeometryof eachfiver
estuaryis largelydeterminedby the hydrodynamic factorsof tideregimeandfiver flow. Fine
sediments arereadilyavailableandin somefiver estuaries therearegeologicalcontrolson the
otherwise unconstrained evolution of these interfaces between fivefine conditions landward of
thetidallimitsandquasi-maritimeconditions seawardin themainSevernEstuary.However,
because thesignificance
of thesecontrolsis onlylocal,thereis considerable
similarityof form
betweenthesefiver estuaries,
a consequence essentially
of thecommoninfluenceof hightide
range.
For thisinitial studyof the effectsof the barrageon conditionsin thefiver estuaries(D.En.
et al, 1989),it wasthereforeappropriate to focusdetailedworkon oneestuaryandto seekto
applythe lessonsso learnedto the others. In this way it provedpossibleto makegeneral
progress andto identifythe moredetailedeffortrequiredin duecourse.The geometryof the
fiver estuariesalloweda 1-D modelto be used,and this was appliedto the simulationof
turbidityand erosion/deposition (Wimpol, 1989). However,as previouslyexplained,the
consequences of the rise in meanwaterlevel in the basinareawouldencourage permanent
accretion on a muchlongertimescale thancouldsensibly be simulatedby a modelof fide-by-
tide processes.The techniques considered for this purposeand that appliedin detailare
described by theUniversityof Liverpool(1989). A timescale of about100yearsto reach95%
of theaccretionneededto give theultimatelystablebedformaccording to thechanged tide
regime and mean water level was deduced. The tide model was then run with this revised
bathymetry to showhowhydrodynamic conditions in thefiver estuaries
wouldthencompare
with thosewhichnowexist. Thosewhichwouldoccurin theshorttermfollowingbarrage
construction were also assessed.
The suitabilityof a I-D modelto simulateflows in the fiver estuaries,and of a 2-D model
for the main SevernEstuary,respectivelybecomeincreasingly questionable as the fiver
estuariesbroadentowardstheirmouthsandas the mainestuarydeepens furtherseaward.In
theseouterregions,bothgeometric anddensityeffectsprogressively influencethe simpler
regimeassumed with justification
to applyin eachcasefurtherlandward.Entending this
simplicityto areasin whichit is knownnot to be as strictlyapplicable givesrise to errors.
Thesemustbe quantifiedin orderto establish whethermorecomplexmodelswill be needed
in futureand,if so,whattheirspecificationsarelikely to be, thefield dataneededto validate
themandthecostsof collecting thosedataandof developing andrunningthemodels,etc.

T L Shaw 639
In the recent study of the Severn Barrage, an initial assessmentwas made of the
specifications for andfindingsof 3-D modelsappliedto thetideregimeof the SevernEstuary
and BristolChannel(McLarenand Collins, 1989). Their resultswere comparedwith those
of severalother2-D plan models.The resultssuggestthatthe overallstructureof the currents
in someshallowareasaway from the main flows and in deeperareascouldincludevertical
components.However, theseappearto be generallyweak comparedwith the main flows in
plan, and it remainsto be shownwhetheror not theyjustify furtherseriousconsiderationin
the contextof the barrage. Presentindicationsare that any effectswill only be of secondary
importanceand are more likely to be significantin some specific situationsrather than
generally.If this is confirmedit may still meanthatmuchof the SevernEstuarywill at some
stagehave to be modelledin threedimensions.
31.3 Salinity regime

In the transitionfrom freshwaterto fully marineconditions,salinityconcentrationsof 5 and
18partsper thousand(ppt) are commonlyusedto describethresholds of tolerancefor various
speciesof plant and animal. A near-marinethresholdof about33 ppt is alsoidentifiedfor
somespecies. However,the salinityat any locationin an estuaryvariesthroughouta tide
cycle,the amplitudeof changedependingon tide range. Also, duringperiodsof higheror
lower than averageincomingriver flow, the mean salinity will be reducedor increased
respectively,andin thenearshore zonewill be affectedby localranoff,eg.,duringprolonged
heavyrainfall.
The abovereferenceto thesignificanceof specificvaluesof meansalinityshouldtherefore

be readwith referenceto the extentto whichtheychangeon the shortertimescaleof a tide
cycleandon thelongertimescaleof varyingriver flows. However,thereappearsto be a lack
of evidenceon which to basethisjudgement. Indeed,in the caseof highertide range
conditionssuchasthosein theSevernEstuary,it mustbequestionable whetherthebiological
regimeis significantly
determined by salinityor whethertidal exposureand submergence,
strongcurrents,storm waves,high turbidity,mobile substrates, etc., either singly or
collectively
exertmoreinfluence.Clearlyit wouldbe wastefulto definethesalinity(or any)
regimein moredetailthanis necessary, whichmeansthata specification for thishasto be
made. This remains to be done.
A significantstepin thisgeneraldirectionwastakenduringtherecentstudyof theSevern

Barrage.It wasarguedthatit wouldbe instructive to knowhowthebarragewouldchange
thesalinityregimethroughout theestuary,
bothin termsof fide-averagedvaluesaccordingto
tiderangeandfiver flow, andtheamplitudes of variationper tidecycle. Earlierworkdone
using a 1-D tide model (SevernBarrageCommittee,1981) suggested that average
concentrationslandward of thebanagewouldfall byat least2-3 ppt,andby morein thefiver
estuaries.

64O T L Shaw
Salinity (g/kg)
2..9
25
I0
1.5cms = 20km
Figure6: a) Sevem- salinitycontoursfor Springtide, 10%flow, no barrage

b) Severn- salinitycontours
for Springtide, 10%flow, barrage

T L Shaw 641
The 2-D modeldeveloped for therecentstudy(WRc, 1989a)wasusedasthehydrodynamic

basisfor simulatingsalinity(WRc, 1989b). Typicalexamplesof the outputof this work are
shownin Fig 6, in this caseshowingthe concentration contoursfor the without and with-
barrageconditionsat low waterof a springfide. The full resultssuggestthat the average
concentration at any locationin the main SevernEstuaryon any tide would be increased
slightly(up to aboutl ppt), thatthe excursionper tide cyclewouldbe reducedandthat there
wouldbe a seawardmovementof salinity(establishing freshetconditions)
in theupperreaches
of the fiver estuaries.
The reasonswhy the 1-D and 2-D modelsdifferedin their forecastsof the directionof
salinitychangein the basindeserveto be resolved,thoughthe absolutemagnitudeof the
differencebetweenthe forecastsmay not be too significant. The fundamentaldifference
betweenthe modelsis, of course,the way in whicheachrepresents lateralvariationsin water
levels,currents,etc.,acrosstheestuary.Early modelsof tidepropagation throughtheBristol
Channel and Severn Estuary (Fig 1) assumedlateral homogeneityand were able to
demonstrate the principaldetailsof the fide regime. However,lateral gradientsof levels,
currents,salinity,etc.,are knownto exist,hencethereis a casefor using2-D representation
for the presentregime.
RESIDUAL VELOCITIES - SPRING. NO BARRAGE.
RESIDUAL
VELOCITIES
-SPRING.
BARRAGE WITH PUMPS.
Figure7: a) Residualvelocities
- Spring.No barrage.
b) Residualvelocities- Spring.Barragewithpumps.

642 T L Shaw
Furthermore,the 2-D model confirmsthat the layout of turbinesand sluicesalong the

barrageand the sequencein which theseare usedthrougha tide cycle would substantially
increasethe magnitudeof thesedifferentialflows, particularlyin the regionof the barrage.
(Fig 7 showsthe residualcurrentsfor a springtide without and with the barrage). The
importanceof the strongcontra-rotatingresidualsset up by the barrageextendsbeyond
salinity,for which it would ensurethat a more rapid exchangeof more salinewater from
seawardandfresherwaterfor landwardwouldoccur. Accordingto theresultsgivenby WRc
(1989b),thenetresultof thismotionis to alterthepresentsalinityregimeas indicatedabove.
Verticalgradients
of salinityin theSevernEstuaryhavebeenshownto be generallysmall,
lessthanlppt. Only in the deeperwatersof the BristolChanneldo greaterdifferencesoccur
but eventheretheyarebelievedto be of little consequence
to thewaterqualityandecological
regimes. Moreover,it is unlikelythattheywouldbe muchalteredby the barrage,thoughthe
possibleoccurrenceof locallysignificantchangesfor someplantandanimalcommunities, and
the needto designthe barrageto influencethis, shouldnot be ruledout.
31.4 Effluent dispersion

The effectsof the differentialcurrentssetup by the turbinesandsluicesare mademoreclear
by the way in which the patternsof dispersionof flows from individualouffallswould be
affected. In orderto quantifythe change,two hypotheticaloutfall locationswere selected
(positionsA andB on Fig 1). The discharge of bothdegradable (eg,bacteria)andconservative
(eg, metals) effluentswere considered. Various bacterialifetimesto 90% decay were
consideredin orderto representthesomewhat uncertain
effectsof highturbidityin reducing
typicaldecayrates. The 2-D tide modelwas alsousedfor this study(WRc, 1989c).
Fig 8 comparesthe plumesof degradable effluentfrom the two outfallswithoutandwith

the barragein place. The changesto the regimedue to the barrageare particularlymarked
for outfall B, seawardof the barrage. In this case, the combinedeffect of the sluices
permittingdischargelandwardthroughthebarrageduringthefloodtidephaseandtheturbines
furtheroffshorereleasingit seawardis to accelerate dispersion
many-foldcompared with that
of the comparativelyrectilinearflows of the tide alone.
The effectof thebarrageon flowsfromoutfallA, becauseof its greaterdistancefrom the

barrage,wouldbe lessmarked.Nevertheless, datafromthemodelsuggest thatevencloseto
thisouffall,contaminant levelswouldbe reduced.Theeffectof accelerated decayshouldthe
fall in turbiditylevelsbe sufficientto increaselight penetration
significantly
wouldbe an
addedconsequence of the barrage;as indicatedabove,thisremainsto be quantified.

T L Shaw 6•3
E.coli/100ml
2OOO
50O
I00
2.6cms = 20km
Without barrage: low water
With barrage: low water
Figure8: a) No barrage,Spring,T90 = 20 hrs,No wind,D. Coeff. 10.0M*M/s

b) Barrage,Spring,T90 = 20 hrs, No wind, D. Coeff. 10.0 M*M/s

644 T L Shaw
Basedon the evidencegiven by WRc (1989c) for the SevernBarrage,the instinctive

supposition thatthistypeof projectwouldlowerthequalityof thewaterin thebasinis not
correct. Hydrodynamicmodellingof dispersionacrossthe main estuaryconfirmsthat
conditionsmaybe expected to improvefollowingbarrageconstruction.
However,in thefiver
estuariestheremay be specificdischargelocationsandphases,duringlower rangefidesin
particular,whenthis improvement may not occur. Theseconditionsremainto be studiedin
detail; the combinedeffectsof reducedcurrents,greatermeanwaterdepths,restrictedlateral
estuarywidthsand reducedpeak turbidityneedto be assessed beforecommentsof eithera
generalor site specificnaturecanbe made. (The effectof long-termaccretionwouldbe to
reducedepthsandincreasecurrentsandhenceturbidity; the neteffectof thesechangesmust
also be considered).
31.5 Wave climate
The orientationand depthof the Bristol Channelpermit waveshavingperiodsof up to 10

secondsto penetrateas far as the barrage(Fig 1). Sedimenttransportin this part of the
estuarymay thereforebe expectedto be influencedby wave action,thoughstrongtidal
currentsare the major force at work. However,in the shallowcoastalzone, this role is
reversed;the risk of stormdamagein somelocationsis high,as experiencesoftenconfinn.
The wave climatein the Bristol Channeldoesnot penetratefar into the SevernEstuary.
The coastfinesseawardof Clevedonand from Cardiff to Newport are exposedto conditions
whicharein partof thisoriginbutarealsodueto morelocallyraisedseas.Furtherup-estuary
and in the fiver estuariesit hasbeenconcludedthat apartfrom the influenceswhichreduced
currentsmay haveon refractionandits effects,the waveclimatewouldbe little affectedby
thebarragebecausethesewaveregimesare generated by localwindswhosefetchwouldnot
be affectedsignificantly,if at all (HRL, 1984).
The most uncertain influence of wave climate is its effects in shallow water and on
coastlinesas far as ecosystemproductivityis concerned.Referencewas madeearlierto the
informationnow availableon the historicevolutionof the SevernEstuary.Wavesarebelieved
to haveplayedan importantpart in this, in conjunctionwith currents. Over the pastfew
hundredyears,the processhas been alteredby the construction of sea walls, and further
changesmay be expectedto occurdue to the presenceof the barrageand how it wouldraise
mean water level in the basin and reducethe currents(more so on the ebb phasethan the
flood).
The resultanteffect on the stabilityof coastalsedimentsremainsto be resolvedand hasa

numberof importantimplications. The wave climate further offshoreappearsto be less
environmentallysignificant.

T L Shaw 645
31.6 Conclusions
Sufficientbasicwork hasbeendone to illustratethe principal hydrodynamiceffectsof the

proposedSevernBarrage. By takingan objectiveapproachto requirements, the recently
completedstudy of the barragehas identified many significanteffects which were not
previouslyappreciated.The principaltasknow is to improvethe level of scientificawareness
in a numberof hydrodynamically-relatedareas,including:
1 The significancefor the effectsof the barrageof thr•-dimensionality in the flow structure
of the BristolChannel,thisbeingdeterminedby topographyand salinity.
2 The extentto whichthe flow structuresetup by the barragemay needto be and may be
modifiedby the designof the barrage,includingthe distributionand alignmentof the
turbinesandsluices,the sequenceof useof thisequipment,the extentandregularityof its
use includingthat for pumping,and the conditionswhich would occurduringbarrage
construction
accordingto the constructionmethodsandprogrammeadopted.
3 Thesignificance, if any,of thedurationof theperiodoverwhichthetideregimeis changed

from that existingat the end of the barrageconstruction
process,but whenall sluicesand
turbinesarestillpermanently open,to thatrequiredforpowergeneration,includingthetime
of year and any other factorsrelevantto the estuary'secosystems.
4 How the rate of decayof degradablecontaminants is determinedby the extentto which

sunlightpenetratesinto the watercolumn(ie by turbidity),togetherwith the influenceof
salinity.
5 How the stabilityof coastalsedimentsand hencethe compositionof this zone are

influencedby the strengthof currents,
the availabilityof material,andthe frequencyand
severityof wave action(informationwhichmustin practicebe extendedby dataon the
plantandanimalcolonisation of specificareas,theseinfluencingstability).
6 How thebarragewould,in the longterm(decades),

alterthedistributionof sediments
and
how thiswould affect the overallregime.
This paperhas not dealtspecificallywith sediments

thoughtheir importanceto the total
regimeof the estuarymustbe stressed.They have been mentionedin the contextof the
effectsof turbidityandof the compositionof thecoastalzone. In theirownway theyare
oftenregarded as thekey to determiningtheproductivityandhencetheecological
character
of anestuary,andit is clearfromwhatis nowknownaboutthebarragethattheprojectwould
alterthischaracterof theestuaryin a numberof significant
ways.
However,it will notbe possibleto determinethesechangeswith confidenceuntil someof

the hydrodynamic implications
of the schemeare morefirmly specified.Whereasthistask
is notregardedasexcessively demanding in scientific
terms,it is in parturgentforresolution

646 T L Shaw
in orderthat attentionmay be givento certainimplications.
The fact that the proposedbarragewouldcreatea tide regimewhichin manyrespects

wouldbe typicalof thatwhichoccursin manyUK (andother)estuaries makesit important
to look elsewherefor guidanceas partof the processof forecasting the changeswhichthe
barragewouldproduce.Basedon thefindingsof therecentstudyof theproject(D.En.et al,
1989), thesechangeswill be lessextensivethan has previouslybeensuggested, and there
appears to be considerable opportunity to influencethem. A furtherconcertedprogramme of
moregeneralstudyas outlinedin thisPaperis neededbeforeestuary-specific conclusions
abouttheenvironmental implicationsof theSevern(or anytidalpower)barragemaybe made.
However,the resultsof the recentstudyhaveaddedconsiderable confidenceto the viability
andacceptability of the projectin a numberof hydrodynamic areas.
31.7 References
Binnie andPartners,1989, "The UK Potentialfor Tidal Energyfrom Tidal Estuaries",Rpt. to ETSU,

No. ETSU/TID 4048 Pt. 1.
Department
of Energy,CEGB and STPG, 1989, "TheSevernBarrageProject",EnergyPaperNo. 57,
HMSO.
HydraulicsResearchLtd., 1984, "WavesUpstreamof the SevernBarrage",Rpt. to the SevernTidal

PowerGroup,No. SD 2.4.
Instituteof Oceanographical
Sciences,1980/81,"Tidal BarrageCalculationsfor the BristolChannel",
Rpts.to the SevemBarrageCommittee,Nos. STP 15-17.
McLaren,P. andCollins,M.B., 1989, "SedimentTransportPathwaysin the SevemEstuaryandBristol

Channel",Rpt. to the SevernTidal PowerGroup,No. SBP 18.
SevernBarrageCommittee,1981,"TidalPowerfromtheSevernEstuary",EnergyPaperNo. 46, 2 Vols,

HMSO.
Universityof Liverpool,1989,"Assessment
of RiverEstuaryGeometry",Rpt. to theSevernTidal Power
Group,N. SBP 10.
Water ResearchCentre, 1989a,"2-D Plan HydrodynamicModellingof the BristolChanneland Severn

EstuaryWithoutandWith Barrage",Rpt. to the SevemTidal PowerGroup,No. SBP3.
Water ResearchCentre,1989b, "SalinityModellingof the SevemEstaury",Rpt. to the SevemTidal

PowerGroup,No. SBP 9.

T L Shaw 647
WaterResearch
Centre,1989e,"TheEffectsof a Barrageon theDispersion
of Pollutants
in theBristol
Channel
andSevernEstuary",
Rpt.to theSevern TidalPowerGroup,No. SBP34.
Wimpol
Ltd.,1989,"I-D Mathematical
Modelling
Study
of theTidalReaches
of theRiverUsk",Rpt.
to theSevernTidal PowerGroup,No. SBP5/5A.

Coastal and Estuarine Sciences Vol. 18: F. Bo Pedersen,
Environmental
Hydraulics:
Strat-
ifiedFlows.278 pages.1986.
1. Coastal Upwelling,FrancisA. Richards(Ed.) 529 Vol. 19: K.N. Fedorov,The PhysicalNatureand Structure
pages. 1981. of OceanicFronts.333 pages.1986.
2. Oceanographyof the SouthwesternU.S. Continen-
tal Shelf, L. P. Atkinson, D. W. Menzel, and K. A. Vol. 20: A. Rieser,J. Spiller,D. VanderZwaag(Eds.),
Bush (Eds.) 156 pages. 1985. Environmental Decisionmaking in a Transboundary
Re-
3. Baroclinic Processes on Continental Shelves, gion.209 pages. 1986.
ChristopherN. K. Mooers (Ed.) 130 pages. 1986. Vol. 21: Th. Stocker,K. Hutter,Topographic
Waves in
4. Three-Dimensional Coastal Ocean Models, Nor- Channels
andLakesonthef-Plane.176pages.1987.
man S. Heaps (Ed.) 208 pages. 1987.
Vol.22: B.-O.Jansson(Ed.),CoastalOffshoreEcosystem
Interactions.
367 pages. 1988.
Coastal and Estuarine Series
Vol.23: K. Heck,Jr.(Ed.),Ecological
Studiesinthe Middle
(formerlyLectureNotes on Reachof ChesapeakeBay.287 pages.1987.
Coastal and EstuarineStudies)
Vol.24: D.G. Shaw,M.J. Hameedi(Eds.),Environmental
Vol. l:J. S(•ndermann,K.-P. Holz (Eds.), Mathematical Studiesin PortValdez,Alaska.423 pages.1988.
Modellingof EstuarinePhysics.Proceedings,1978. 265
pages. 1980. Vol. 25: C.M. Yentsch,F.C. Mague,P.K. Horan(Eds.),
Immunochemical
Approachesto Coastal, Estuarineand
Vol. 2: D.P. Finn,Managingthe Ocean Resourcesof the
Oceanographic
Questions.
399 pages.1988.
United States: The Role of the Federal Marine Sanctu-
aries Program. 193 pages. 1982. Vol.26: E.H. Schumann(Ed.),CoastalOceanStudiesoff
Vol. 3: M. TomczakJr., W. Cuff (Eds.), Synthesisand Natal,SouthAfrica.271 pages.1988.
Modellingof IntermittentEstuaries.302 pages. 1983. Vol.27: E. Gold(Ed.),A Lawof the Sea forthe Caribbean:
Vol. 4: H.R. Gordon,A.Y. Morel, Remote Assessmentof An Examination of MarineLaw and PolicyIssuesin the
OceanColorfor Interpretation
of SatelliteVisibleImag- LesserAntilles.507 pages.1988.
ery. 114 pages. 1983. Vol. 28: W.S. Wooster(Ed.), FisheryScienceand Man-
Vol. 5: D.C.L. Lam,C.R. Murthy,R.B. Simpson,Effluent agement.339 pages. 1988.
Transportand DiffusionModels for the Coastal Zone.
Vol.29: D.G. Aubrey,L. Weishar(Eds.),Hydrodynamics
168 pages. 1984.
andSedimentDynamics ofTidalInlets.456 pages.1988.
Vol. 6: M.J. Kennish, R.A. Lutz (Eds.), Ecologyof
BarnegatBay, New Jersey.396 pages. 1984. Vol.30: P.B.Crean,T.S. Murty,J.A. Stronach,Mathemat-
ical Modellingof Tides and EstuarineCirculation.471
Vol. 7: W.R. Edeson,J.-F. Pulvenis,The LegalRegime
pages. 1988.
of Fisheriesin the CaribbeanRegion.204 pages.1983.
Vol.8: O. Holm-Hansen,
L. Bolis,R. Gilles(Eds.),Marine Vol.31: G. Lopez,G. Taghon,J. Levinton
(Eds.),Ecology
Phytoplanktonand Productivity.
175 pages.1984. of MarineDepositFeeders.322 pages.1989.
Vol. 9: A. Pequeux,R. Gilles,L. Bolis(Eds.),Osmoregu- Vol. 32: F. Wulff,J.G. Field,K.H. Mann (Eds.),Network
and MarineAnimals.221 pages.1984. Analysisin MarineEcology.284 pages.1989.
lationin Estuarine
Vol. 10: J.L. McHugh,FisheryManagement.
207 pages. Vol. 33: M.L. Khandekar,OperationalAnalysisand Pre-
1984. dictionof OceanWindWaves.214 pages.1989.
Vol. 11: J.D. Davis,D. Merriman(Eds.),Observations
on Vol. 34: S.J. Neshyba,Ch.N.K. Mooers,R.L. Smith,R.T.
the Ecologyand Biologyof Western Cape Cod Bay, Barber (Eds.), PolewardFlows Along Eastern Ocean
Massachusetts. 289 pages. 1984. Boundaries. 374 pages.1989.
Vol. 12: P.P.G. Dyke, A.O. Moscardini,E.H. Robson
Vol.35: E.M. Cosper,V.M. Bricelj,E.J. Carpenter(Eds.),
(Eds.),Offshore
andCoastalModelling.
399 pages.1985.
NovelPhytoplanktonBlooms.799 pages.1989.
Vol. 13: J. Rumohr,E. Walger, B. Zeitzschel(Eds.),
Seawater-Sediment Interactions in Coastal Waters. An Vol.36: W. Michaelis(Ed.),EstuarineWaterQualityMan-
Interdisciplinary
Approach.338 pages.1987. agement.478 pages. 1990.
Vol. 14: A.J. Mehta (Ed.), EstuarineCohesiveSediment Vol. 37: C.M. Lalli(Ed.), EnclosedExperimental
Marine
Dynamics.473 pages.1986. Ecosystems:A Review and Recommendations.X, 218
pages. 1990.
Vol. 15: R.W. Eppley(Ed.), PlanktonDynamicsof the
SouthernCalifornia Bight.373 pages.1986. Vol.38: R.T. Cheng(Ed.),ResidualCurrentsand Long-
Vol. 16: J. van de Kreeke (Ed.), Physicsof Shallow termTransport.XI, 544 pages.1990.
Estuariesand Bays.280 pages.1986. Vol.39: M.I. EI-Sabh,N. Silverberg
(Eds.),Oceanography
Vol. 17: M.J. Bowman, C.M. Yentsch, W.T. Peterson of a Large-Scale Estuarine System.X, 434 pages.1990.
(Eds.),TidalMixingand PlanktonDynamics.502 pages. Vol. 40: D. Prandle(Ed.), Dynamicsand Exchangesin
1986. Copyright American Geophysical Union
Estuariesand the CoastalZone.648 pages.1992.

Coastal and Estuarine PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Coastal and Estuarine PDF

Uploaded by

Copyright:

Available Formats

Coastal and Estuarine Studies

David Prandle, Editor

Dynamics and Exchanges

American Geophysical Union

Library of Congress Cataloging-in-PublicationData

Dynamicsand exchangesin estuariesandthe coastalzone/ David

Copyright1992 by the AmericanGeophysicalUnion2000 FloridaAvenue,NW, Washington,DC

Figures,tables,and shortexcerptsmay be reprintedin scientificbooksand journalsif the source

Printed in the United States of

Part II. BaroclinicDynamics

Tidal mixing near the sill of a Scottishsealoch

Part III. Circulation

Residualflow in NaplesBay andits effecton constituent

Water level fluctuationsin the AtchafalayaDelta, Louisiana:

Modellingof low-frequencysalinityvariationsin the St. LawrenceEstuary

On the estuafinecirculationwithin the Kattegat

Part IV. Sedimemation

Velocity asymmetries in frictionally-dominated

Effectsof sea-levelrise on muddycoastalmargins

Someobservations on fluid mudresponse to waterwaves

The reflection of waves off beaches

Part V. Modelling (Sediment)

Effect of bends on dilution rates

Modellingthe verticaldistributionof suspended

Part VI. Applied Studies

Transportof hypoxicwaters:an estuary-subestuary

R.T. Cheng,U.S.G.S., SanFrancisco,USA

As organiserof the precedingconference, Ralph Cheng'sadviceand encouragement was

In thepreparationof thisvolume,thepainstakingeffortof therefereesis duly recognised,

Finally, neitherthe conferencenor this volumewould have happenedwithoutthe efforts

der UniversitaetHamburg PO Box 20907

Troplowitzstr.7 25(}0EX The Hague

MathematicsDepartment Marine Science Laboratories

Universityof Trento Menai Bridge

PlymouthPL4 8AA MerseysideL43 7RA

Miami, Florida 33149-1098 Florida 32611

ScrippsInstitutionof Oceanography Departmentof Civil Engineering

2500 EX The Hague School of Marine Science

BedfordInstituteof Oceanography USA

BatonRouge,LA 70803 United Kingdom

Division for Earth, Marine and Atmospheric

GwyneddLL59 5EY LouisianaStateUniversity

Schoolof Civil Engineering

In establishingthepeculiarnicheof a volumeof conference proceedingswithinthescientific

1.1 The GregynogConference

Likewise,residentialvenues(Greygnogis a remotestudycentrekindly madeavailableby

1.2 The Proceedings

Copyright American Geophysical Union

Part II- BaroclinicDynamics

Theorderof thefive papers in thissectionconvenientlyfollowsa sequenceof (firs0some

This sectionincludes7 papersconcerned with circulationbothhorizontal andvertical.The

The extentof the emphasison sedimentation withinthisconference reflectsthe increasing

Copyright American Geophysical Union

the turbiditymaximumin the Tamar; using a numericalmodel in conjunctionwith

Thissection includessixpapersthataddress variousaspects

Part VI- AppliedStudies

Copyright American Geophysical Union

tidal modelof Australia'sCentralGreatBarrierReef to studythe exchangeof coastalwaters.

Hopefullytheabovesinglesentence summaries helpin bridgingthegapbetweenthetitlesand

Lest anyonehasmissedthe discernible trendin estuarineresearchawayfrom traditional

Copyright American Geophysical Union

The influence of coastally trapped waves

JervisBay is a small(15 km x 8 km) coastalembaymentin New SouthWales,Australia.As