You are on page 1of 8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

Section9:StatisticalAnalysisofStreamGaugeData
Ifthegaugingrecordcoversasufficientperiodoftime,itispossibletodevelopaflowfrequencyrelationby
statisticalanalysisoftheseriesofrecordedannualmaximumflows.Thedesignercanthenusetheflowfrequency
relationinoneoftwoways:
Ifthefacilitysiteisnearthegaugingstationonthesamestreamandwatershed,thedesignercandirectly
usethedischargeobtainedfromtheflowfrequencyrelationforthedesignAEP.
Ifthefacilitysiteisonthesamestream,butnotproximatetothegaugingstation,itmaybepossibleto
transposegaugeanalysisresults.
WidelyacceptedandappliedguidelinesforstatisticalanalysesofstreamgaugedataarepublishedinGuidelines
forDeterminingFloodFlowFrequency,Bulletin#17B(IACWD1982).ProceduresfromBulletin#17B,with
someTexasspecificrefinements,asoutlinedinthismanual,arerecommended.Theyinclude:
Obtainingasufficientlylargesampleofstreamflowdataforstatisticalanalysis,
UsingthelogPearsontypeIIIdistributionfittingprocedure,
Usingaweightedskewvalue,
Accommodatingoutliers,
Transposinggaugeanalysisresults,ifnecessaryandappropriate.

DataRequirementsforStatisticalAnalysis
Thegreatestchallengeinapplyingthestatisticalanalysisofstreamgaugedataisobtainingasufficientlylarge
sampleofstreamflowmeasurements(orestimates)sothatthesampleisrepresentativeoftheentirepopulationof
flows.Twotypesofdatamaybeconsidered(IACWD1982),systematicdataandhistoricaldata.
Systematicdataareflowrecordsgeneratedfromadefinedsetofrulesandrecordedonaregularbasis.For
example,theUnitedStatesGeologicalSurvey(USGS)annualmaximumflowrecordforagaugeconsistsofthe
maximuminstantaneousflowvalueforeachyear,recordedeveryyearoveragiventimeperiod.Ifannual
maximumflowvalueswererecordedonlyforyearsinwhichlargeeventsoccurred,thentherecordwouldno
longerbesystematic.Gaps(missingyears)inthesystematicrecorddonotprecludeuseofsuchdatasolongas
thegapsaretheresultofmissingdata,andnottheresultoffilteringthedatabasedonflowmagnitude.
Historicaldataareflowestimatesforeventsnotincludedinthesystematicrecord.Thesedatatypicallyconsistof
historicallysignificantevents,andthusareasampleofextremeeventsobservedbylocals.Historicaldatashould
beincludedintheanalysiswhenpossible.Incaseswhereonlyashortsystematicrecordisavailable,historical
dataareparticularlyvaluable.Useofhistoricaldataalsoensuresthattheresultsoftheanalysiswillbeconsistent
withtheexperienceofthelocalcommunity(IACWD1982).
Forhighwaydrainagedesignpurposes,astatisticalanalysisofstreamgaugedataistypicallyappliedonlywhen
adequatedatafromstreamgaugingstationsareavailable.ThedefinitionofadequatedatacomesfromUSGS
practiceandisprovidedinTable43.
Table43:RecommendedMinimumStreamGaugeRecordLengths(DalrympleandBenson
1960)
Desiredpercentchanceexceedance(ARI)

Minimumrecordlength(years)

10year

25year

10

50year

15

100year

20

http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

1/8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

100year

20

ForTxDOTapplication,sourcesforannualpeakflowdatainclude:
USGSNationalWaterInformationSystem(NWIS).
USDepartmentoftheInterior,USGSWaterResourcesDataTexas,SurfaceWater.Theseareprepared
annuallyandcontainrecordsfor1wateryearperpublication.Asaresult,abstractingannualpeaksfora
longrecordistimeconsuming.
InternationalBoundaryandWaterCommissionwaterbulletins.
Iftheavailabledatasourcesallowthedesignertoconstructasufficientlylargesampleofannualpeakstreamflow
values,thenthefollowingconditionsmustalsobesatisfiedoraccountedforbeforeundertakingthestatistical
analysis:
Thedatamustberepresentativeofthedesignconditionofthewatershed.
Thedatamustnotbesignificantlyaffectedbyupstreamregulation.
Thesystematicrecordmustbestationary,withnogeneraltrendofincreasingordecreasingflows
resultingfromchangestothewatershed.
Thedatamustbehomogeneous,withflowvaluesresultingfromthesametypesofevents.Ifannual
peakflowscanresultfromeitherrainfallorsnowmelt,thenamixedpopulationanalysismaybe
required.
Errorsinflowmeasurementsmustnotbesignificantrelativetootheruncertaintiesintheanalysis.

LogPearsonTypeIIIDistributionFittingProcedure
ThelogPearsontypeIII(LPIII)statisticaldistributionmethodisrecommendedinBulletin#17Bandisthe
standardofpracticeforestimatingannualprobabilityofexceedanceofpeakflows.Anoutlineofthismethod
follows.However,thedesignerisnotlimitedtousingthismethod,especiallyiftheresultingflowfrequency
relationdoesnotseemtofitthedata.
ThefollowinggeneralprocedureisusedforLPIIIanalyses.RefertoBulletin#17Bforfurtherinformation.
1.Acquireandassesstheannualpeakdischargerecord.
2.Computethebase10logarithmofeachdischargevalue.
3.Computethemean,standarddeviation,and(station)skewofthelogflowvalues.
4.Computetheweightedskewcoefficientfromthestationskewandregionalskew.
5.Identifyhighandlowoutliersfromthesampleset.
6.Recomputethemean,standarddeviation,andstationskewofthelogflowvalueswithoutliersremoved
fromthesampleset.
7.ComputeflowvaluesfordesiredAEPs.
WiththeLPIIImethod,thelogarithmofthedischargeforanyAEPiscalculatedas:

Equation42.
Where:
http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

2/8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

=meanofthelogarithmsoftheannualpeakdischarges
Qp=floodmagnitude(cfsorm3/s)ofAEPp
K=frequencyfactorforAEPpandcoefficientofskewappropriateforsite
SL=standardofdeviationoflogarithmsoftheannualpeakdischarges
Seethespreadsheetfreqfrac.xlsforvaluesofK,basedonstationskewcoefficient.
ThethreestatisticalmomentsusedtodescribetheLPIIIdistributionarethemean,standarddeviation,andskew.
Estimatesofthesemomentsforthedistributionoftheentirepopulationofflowsarecomputedfortheavailable
sampleofflowswiththeequationsbelow.
Themeanisgivenby:

Equation43.
Where:
=meanofthe(base10)logarithmsoftheannualpeakdischarges
X=logarithmoftheannualpeakdischarge
N=numberofobservations
Thestandarddeviationisgivenby:

Equation44.
Where:
SL=standarddeviationofthelogarithmsoftheannualpeakdischargeNandXaredefinedasabove
Thecoefficientofskew(stationskew)isgivenby:

Equation45.
http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

3/8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

Where:
G=coefficientofskewoflogvaluesN,X,andSLaredefinedasabove
SkewrepresentsthedegreeofcurvaturetotheflowfrequencycurveasshowninFigure43.InFigure43theX
axisscaleisprobability(symmetricataboutAEP=0.5)andtheYaxisscaleisbase10logarithmicflow.Askew
ofzeroresultsinastraightlineflowfrequencycurve.Anegativeskewvalueproducesaflowfrequencycurve
withlesserflowsthanthezeroskewline,andapositiveskewproducesaflowfrequencycurvewithgreaterflows
thanthezeroskewline.

Figure43.Skewofdischargeversusfrequencyplots
Thefollowingcasesrequirespecialconsideration.Bulletin#17Bprovidesfurtherguidance:
Recordisincompleteflowsmissingfromrecordbecausetoosmallortoolargetomeasure(flows
filteredfromrecordbasedonflowmagnitude).
Recordcontainszeroflowvaluesstreamwasdryallyear.
Recordcontainshistoricalflowsnotrecordedinasystematicfashion.Examplesareextremeevents
recordedpriortoorafterinstallationofastreamgauge.TheseareindicatedbycodeinUSGSannual
peakdischargedata.
Flowsaretheresultoftwodistincttypes(amixedpopulation)ofhydrologiceventssuchassnowmelt
andrainstorms.

Skew
Bulletin#17Brecommendsusingtheweightedskewvalue,GW,todeterminefrequencyfactorvaluesinEquation
42.
Tocalculateweightedskew,usethisequation,andfollowthestepsprovidedinBulletin#17B:

Equation46.
http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

4/8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

Where:
GW=weightedskewvalue
=regionalskewfromFigure44
G=stationskewfromEquation45
=meansquareerrorof forTexas,is=0.123(RMSE=0.35)(Juddetal.1996),which
replacesthevalueof0.302(RMSE=0.55)presentedinBulletin#17B.
MSEG=meansquareerrorofG.MSEGisafunctionofGandperiodofrecord

Figure44.GeneralizedskewcoefficientsforTexas(Judd1996)( )

http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

5/8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

Equation47.
WhereNistherecordoflengthand
A=0.33+0.08|G|for|G|<0.90
A=0.52+0.30|G|for|G|>0.90
And
B=0.940.26|G|for|G|<1.50
B=0.55for|G|>1.50

AccommodationofOutliers
Thedistributionofalltheannualandhistoricalpeakdischargesdeterminestheshapeoftheflowfrequencycurve
andthusthedesignpeakdischarges.TheshapeofthefrequencycurvegeneratedbyalogPearsontypeIII
analysisissymmetricalaboutthecenterofthecurve.Therefore,thedistributionofthehigherpeakdischarges
affectstheshapeofthecurve,asdoesthedistributionofthelowerpeakdischarges.
FloodingiserraticinTexas,soaseriesofobservedfloodsmayincludeannualpeakdischargeratesthatdonot
seemtobelongtothepopulationoftheseries.Thevaluesmaybeextremelylargeorextremelysmallwithrespect
totherestoftheseriesofobservations.Suchvaluesmaybeoutliersthatshouldbeexcludedfromthesetofdata
tobeanalyzedortreatedashistoricaldata.Bulletin#17Bcallsforidentificationoftheseoutliers.
Designflowsaretypicallyinfrequentlargeflows.Therefore,itisdesirabletobasethefrequencycurveonthe
distributionofthelargerpeaks.Thisisaccomplishedbyeliminatingfromtheanalysespeakdischargeslowerthan
alowoutlierthreshold.Thevalueforthelowoutlierthreshold,therefore,shouldexcludethosepeaksnot
indicativeofthedistributionforthehigherpeaks.Thisvalueischosenbyreviewingthesequentiallyranked
valuesforallpeakdischargesusedintheanalysis.
Equation48providesameansofidentifyingthelowoutlierthreshold(Asquithet.al1995):

Equation48.
Where:
LOT=estimatedlowoutlierthreshold(cfs)
=meanofthelogarithmsoftheannualpeakdischarge(seeEquation43)
SL=standarddeviationofthelogarithmsoftheannualpeakdischarge(seeEquation44)
G=coefficientofskewoflogvalues(stationskew,seeEquation45)
a=1.09
b=0.584
c=0.140
http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

6/8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

d=0.799
ThisequationwasdevelopedforEnglishunitsonlyanddoesnotcurrentlyhaveametricequivalent.
Highoutlierthresholdspermitidentificationofextremelyhighpeakdischargeswithprobabilitysmallerthan
indicatedbytheperiodofrecordforastation.Forexample,ifatrue1%percentchanceexceedance(100year)
peakdischargeweregaugedduringa10yearperiodofrecord,thefrequencycurvecomputedfromthe10years
ofrecordwouldbeundulyshapedbythe1%percentchanceexceedancepeak.
TheUSGShasmadeeffortstoidentifyhighoutliers,referredtoashistoricalpeaks,byidentifyingand
interviewingresidentslivingproximatetothegaugingstations.Inmanycases,residentshaveidentifieda
particularfloodpeakasbeingthehighestsinceaprevioushigherpeak.Thesepeaksareidentifiedasthehighest
sinceaspecificdate.
Inothercases,residentshaveidentifiedaspecificpeakasthehighestsincetheyhavelivedproximatetothe
gaugingstation.Thosepeaksareidentifiedasthehighestsinceatleastaspecificdate.Thehistoricalpeaksmay
precedeorbewithintheperiodofgaugedrecordforthestation.
Equation49providesameansofidentifyingthehighoutlierthreshold(Bulletin#17B):

Equation49.
Where:
HOT=estimatedhighoutlierthreshold(logarithmofflow)
N=numberofsystematicpeaksremaininginsampleafterpreviouslydetectedoutliershavebeen
removed
=meanofthelogarithmsofthesystematicannualpeakdischarges,withpreviouslydetectedoutliers
removed
SL=standardofdeviationofthelogarithmsoftheannualpeakdischarges
KN=frequencyfactorforsamplesizeNfromAppendix4ofBulletin#17B
AllknownhistoricalpeakdischargesandtheirassociatedgaugeheightsanddatesappearontheUSGSTexas
WaterSciencewebsite.
ToincorporatehighoutlierinformationwhenfittingtheLPIIIdistributionaccordingtoBulletin#17Bprocedures,
thedesignerwill:
UseEquation49todefinethehighoutlierthreshold.
Collectsupportinginformationabouttheidentifiedhighoutlyingflows.
Retainaspartofthesystematicrecordanyhighoutlyingflowsfoundnottobethemaximumflowof
record.
Extendtheperiodofrecordfortheanalysistoincludetheflowiftheflowsvalueisfoundtobethe
maximumflowofrecordandliesoutsidethesystematicrecord.Ifthevaluedoesliewithinthe
systematicrecord,theperiodofrecordisnotextended.Inbothcases,thedesignershallrecomputethe
LPIIIparametersfollowingtheproceduredescribedinSectionV.A.9andAppendix6ofBulletin#17B.
Thoroughlydocumentdata,interviews,decisions,andassumptionsusedtojustifytheidentificationof
highoutliersandrecomputationofLPIIIparameters.
http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

7/8

10/14/2016

HydraulicDesignManual:StatisticalAnalysisofStreamGaugeData

TxDOTrecommendstheuseofhydrologicstatisticalanalysiscomputerprogramsthatcandetectoutlyingvalues
andrecomputedLPIIIparametersconsistentwithBulletin#17Bprocedures.

TranspositionofGaugeAnalysisResults
Ifgaugedataarenotavailableatthedesignlocation,dischargevaluescanbeestimatedbytranspositionifapeak
flowfrequencycurveisavailableatanearbygaugedlocation.Thismethodisappropriateforhydrologically
similarwatershedsthatdifferinareabylessthan50percent,withoutletlocationslessthan100milesapart.
FromtheresearchofAsquithandThompson2008,anestimateofthedesiredAEPpeakflowattheungaugedsite
isprovidedbyEquation410:

Equation410.
Where:
Q1=EstimatedAEPdischargeatungaugedwatershed1
Q2=KnownAEPdischargeatgaugedwatershed2
A1=Areaofwatershed1
A2=Areaofwatershed2
Transpositionofpeakflowisdemonstratedwiththefollowingexample.Adesignerrequiresanestimateofthe
1%AEPstreamflowatanungaugedlocationwithdrainageareaof200squaremiles.Anearby(within100miles)
streamgaugehasahydrologicallysimilardrainageareaof450squaremiles.The1%AEPpeakstreamflowatthe
gaugedlocationis420cfsbasedonthepeakflowfrequencycurvedevelopedforthatlocation.Substitutinginto
Equation410resultsin280cfsasanestimateofthe1%AEPpeakdischargeattheungaugedlocation:

Ifflowfrequencycurvesareavailableatmultiplegaugedsites,Equation410canbeusedtoestimatethedesired
peakAEPflowfromeachsite.Then,withjudgmentandknowledgeofthewatersheds,thoseestimatescouldbe
weightedtoprovideanestimateofthedesiredAEPflowattheungaugedlocation.Thisprocessshouldbewell
documented.
Designofastoragefacility,suchasadetentionpond,mayrequireestimatesofAEPflowsforlongerdurations.If
aflowfrequencycurveforlongerflowdurationisavailableatanearbygaugedlocation,thenEquation411,
basedonananalysisofmeandailyflows(Asquithetal.2006),maybeusedfortransposition:

Equation411.

http://onlinemanuals.txdot.gov/txdotmanuals/hyd/statistical_analysis_of_stream_gauge_data.htm

8/8

You might also like