Service Overview and Validation

A part of BMT in Energy and Environment
Overview of the service and

validation of the database
Reference: RP_A870
Date: June 2011
www.waveclimate.com
BMT ARGOSS
Overview of the service and validation of the database
_____________________________________________________________________________________________________
Document status sheet

Title
Reference
RP_A870
Issue
11
Date
June 28, 2011
Author(s)
Peter Groenewoud
Reviewed by
Cees de Valk, Martin Williams
Revision history:
Issue
11
10
9
8
7
6
5
4
3
2
1
Date
June 2011
April 2010
June 2009
May 2008
June 2007
May 2006
June 2005
April 2004
October 2003
April 2002
October 2001
Comment
Updated model data validation 1992-2009. Re-organised this report.
Included satellite data 2009 in the validation
Included satellite data 2005 and model data 1992-2004 in the validation
Included satellite data 2003 in the validation; model data 1998-2002
Introduced model hindcast data 1998-1999
Validation of the initial version of the satellite database
More information on versions can be found online at the history page of waveclimate.com.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
BMT ARGOSS
_____________________________________________________________________________________________________
Contents
1.
INTRODUCTION ............................................................................................................................. 1
1.1.
1.2.
1.3.
1.4.
1.5.
2.
OVERVIEW OF WAVECLIMATE.COM .......................................................................................... 6

2.1.
2.2.
2.3.
2.4.
3.
ERROR STATISTICS OF RAW SATELLITE DATA PER MISSION AND PER YEAR ...................................... 9
CALIBRATION OF SATELLITE DATA WITH BUOYS ........................................................................... 13
CALIBRATION COEFFICIENTS FOUND FOR SATELLITE DATA ........................................................... 15
ERROR STATISTICS OF SATELLITE DATA AFTER CALIBRATION WITH BUOYS .................................... 16
SAR DATA .................................................................................................................................... 20

4.1.
4.2.
5.
CONTENT OF THE DATABASE ........................................................................................................ 6

DATA PROCESSING...................................................................................................................... 6
DATA ACCURACY......................................................................................................................... 7
CLIMATE STATISTICS PROVIDED BY THE ONLINE SERVICE ............................................................... 8
ALTIMETER AND SCATTEROMETER DATA ............................................................................... 9

3.1.
3.2.
3.3.
3.4.
4.
OBJECTIVES OF THIS DOCUMENT.................................................................................................. 1

EXECUTIVE SUMMARY ................................................................................................................. 1
THE ONLINE SERVICE IN RELATION TO IN-HOUSE CONSULTANCY ..................................................... 3
FITNESS FOR USE OF THE ONLINE SERVICE ................................................................................... 4
STRUCTURE OF THIS DOCUMENT .................................................................................................. 5
ERROR STATISTICS OF RAW SAR DATA PER REGION ................................................................... 20

CALIBRATION OF SAR DATA ...................................................................................................... 22
WAVE MODEL DATA ................................................................................................................... 24

5.1.
5.2.
5.3.
5.4.
CALIBRATION OF MODEL DATA.................................................................................................... 24

ERROR STATISTICS OF SATELLITE DATA USED FOR MODEL CALIBRATION ....................................... 26
THE EFFECT OF MODEL CALIBRATION WITH SATELLITES ACCORDING TO BUOYS ........................... 30
THE APPLICABILITY OF MODEL CALIBRATION WITH SATELLITES WORLDWIDE .................................. 38
APPENDICES ....................................................................................................................................... 41
APPENDIX A- BUOYS USED FOR VALIDATION .............................................................................. 42
APPENDIX B- PARAMETERS USED FOR ERROR STATISTICS ..................................................... 44
APPENDIX C- GLOBAL AND REGIONAL WAVE MODELS ............................................................. 46
APPENDIX D- THE SHALLOW WATER WAVE RAY MODEL........................................................... 48
APPENDIX E- FRAME OF REFERENCE ............................................................................................ 49
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
BMT ARGOSS
_____________________________________________________________________________________________________
List of tables
Table 1 Wind and sea state data contained in the global data base at BMT ARGOSS. ........................ 6
Table 2 Accuracy of satellite and wave model wave height provided by the online service. ................. 7
Table 3 Accuracy of satellite and wave model wind speed provided by the online service. .................. 7
Table 4 Available statistics and corresponding data sources. ............................................................... 8
Table 5 Error statistics of raw wave height from altimeter based on all buoys. ...................................... 9
Table 6 Error statistics of raw wind speed from altimeter based on all buoys. ..................................... 10
Table 7 Error statistics of raw wind speed from scatterometer based on all buoys. ............................. 10
Table 8 Error statistics of raw wave height from Topex altimeter based on all buoys. ......................... 11
Table 9 Error statistics of raw wind speed from Topex altimeter based on all buoys. .......................... 11
Table 10 Calibration coefficients found for wave height from altimeter based on all buoys. ................ 15
Table 11 Calibration coefficients found for wind speed from altimeter based on all buoys. ................. 15
Table 12 Calibration coefficients found for wind speed from scatterometer based on all buoys. ......... 15
Table 13 Error statistics of calibrated wave height from altimeter based on all buoys. ........................ 16
Table 14 Error statistics of calibrated wind speed from altimeter based on all buoys. ......................... 16
Table 15 Error statistics of calibrated wind speed from scatterometer based on all buoys. ................. 16
Table 16 Statistics of raw wave height based on SAR/ECMWF spectra per region. ........................... 20
Table 17 Statistics of raw mean period based on SAR/ECMWF spectra per region............................ 20
Table 18 Statistics of raw zero-crossing period based on SAR/ECMWF spectra period per region. ... 21
Table 19 Statistics of raw height of long waves based on SAR/ECMWF spectra per region. .............. 21
Table 20 Altimeter-buoy error statistics for wave height per processing level ..................................... 26
Table 21 Altimeter-buoy error statistics for wind speed per processing level ...................................... 26
Table 22 Scatterometer-buoy error statistics for wind speed per processing level ............................. 29
Table 23 Relative model-buoy error per region before and after calibration with satellites ................. 30
Table 24 Bias model-buoy per region before and after calibration with satellites ............................... 30
Table 25 List of NOAA buoys used for validation. ............................................................................... 43
Table 26 BMT ARGOSS hindcast models, grid resolutions and wind data sources. ........................... 46
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
BMT ARGOSS
_____________________________________________________________________________________________________
List of figures
Figure 1 Global and regional hindcast models........................................................................................ 1
Figure 2 Location and average significant wave height of NOAA buoys. ............................................... 2
Figure 3 Bias and STD of error of raw wave height per year of Topex altimeter and all buoys. .......... 11
Figure 4 Bias and STD of error of raw wind speed per year of GFO altimeter and all buoys. ............. 12
Figure 5 Bias and STD of error of raw wave height per year of GFO altimeter and all buoys. ............ 13
Figure 6 Least squares fit of raw wave height for Topex altimeter and all buoys. ................................ 14
Figure 7 Bias and STD of error of raw wave height per year of GFO altimeter and all buoys. ............ 17
Figure 8 Bias and STD of error of corrected wave height per year of GFO altimeter and all buoys. ... 17
Figure 9 Least squares fit of raw wave height for Ers-1 altimeter and all buoys. ................................. 18
Figure 10 Least squares fit of corrected wave height for Ers-1 altimeter and all buoys. ...................... 18
Figure 11 Least squares fit of raw wind speed for Jason-2 altimeter and all buoys. ............................ 19
Figure 12 Least squares fit of corrected wind speed for Jason-2 altimeter and all buoys. ................... 19
Figure 13 Error in original altimeter wave height over the years .......................................................... 27
Figure 14 Error in calibrated altimeter wave height over the years ...................................................... 27
Figure 15 Prob. of exceedance of raw altimeter and buoy wave height ............................................... 28
Figure 16 Prob. of exceedance of calibrated altimeter and buoy wave height ..................................... 28
Figure 17 Prob. of exceedance of raw model and buoy wave height for Atlantic ................................. 32
Figure 18 Prob. of exceedance of altimeter-calibrated model and buoy wave height for Atlantic ........ 32
Figure 19 Prob. of exceedance of raw model and buoy wave height for NFL ...................................... 33
Figure 20 Prob. of exceedance of altimeter-calibrated model and buoy wave height for NFL ............. 33
Figure 21 Prob. of exceedance of raw model and buoy wave height for GOM .................................... 34
Figure 22 Prob. of exceedance of altimeter-calibrated model and buoy wind speed for GOM ............ 34
Figure 23 Prob. of exceedance of satellite-calibrated model and buoy wind speed for GOM .............. 35
Figure 24 Prob. of exceedance of altimeter and buoy wave height for GOM ....................................... 35
Figure 25 Prob. of exceedance of raw model and buoy wave height for PAC ..................................... 36
Figure 26 Prob. of exceedance of altimeter-calibrated model and buoy wave height for PAC ............ 36
Figure 27 Prob. of exceedance of raw model and buoy wave height for HAW .................................... 37
Figure 28 Prob. of exceedance of altimeter-calibrated model and buoy wave height for HAW ........... 37
Figure 29 Bias in global model wave height after calibration with satellites ......................................... 38
Figure 30 Bias in global model wind speed after calibration with satellites .......................................... 38
Figure 31 Bias in model wave height in the Mediterranean after calibration with satellites ................. 39
Figure 32 Bias in model wind speed in the Mediterranean after calibration with satellites .................. 39
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
BMT ARGOSS
_____________________________________________________________________________________________________
1.
Introduction
1.1.
Objectives of this document

Our internet service www.waveclimate.com provides worldwide wave and wind climate
information based on wave model computations and satellite measurements stored in a
database at BMT ARGOSS. This report gives an overview of the products of the online
service and of the processing and quality of the data stored in the database.
1.2.
Executive summary
At BMT ARGOSS metocean consultancy products and related web services are primarily
based on hindcast data from our in-house database, currently covering the years 1992-2010.
BMT ARGOSS runs a global wave hindcast model in all major ocean basins as well as local
models in semi-closed basins such as the Mediterranean (see Figure 1 below).
Figure 1 Global and regional hindcast models.
Wave model data are calibrated with satellite data to remove any systematic error. The
satellite data are calibrated with buoy data. The positive effect of this calibration is
substantiated by comparing the model data to true buoy measurements before and after the
calibration with satellite data.
With reference to buoys, it is demonstrated that the quality of satellite data has increased
after each step of the processing chain: the relative error in wave height is reduced from 15%
to 11%. The resulting best satellite wave height and wind speed observations are practically
un-biased. Proof is given that, with reference to buoys, calibration with these best satellite
observations does indeed improve the quality of our wave model. The systematic error in
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 1
BMT ARGOSS
_____________________________________________________________________________________________________
model wave height is reduced by 50% or more. The final bias in model wave height relative to
buoys is at most 5cm while wind speed is off by less than 20cm/s.
Basic processing of satellite data is performed under the responsibility of the space agencies
that supply the data. Sensors are altimeter (measuring significant wave height and wind
speed), scatterometer (measuring the wind vector) and SAR Wave Mode (providing spectral
wave information such as periods and directions). Satellites include GeoSat, ERS-1/2,
Topex/Poseidon, GFO, Jason-1/2, Envisat, Quikscat and MetOp-A.
Buoy observations come from deep-water NOAA buoys around North America and Hawaii
shown below in Figure 2.
Figure 2 Location and average significant wave height of NOAA buoys.
Validation and, if necessary, calibration of wind and wave data in the database is done by
BMT ARGOSS after each update of the database, usually once a year.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 2
BMT ARGOSS
_____________________________________________________________________________________________________
1.3.
The online service in relation to in-house consultancy

The Waveclimate.com online service was developed to provide easily obtainable, entry level
metocean information for situations where the user does not need detailed expert analysis.
Note: as stated on the service website the user should always consider whether a more
extensive appraisal is appropriate, in which case BMT ARGOSS metocean specialists can
assist.
As demonstrated in this document, BMT ARGOSS carries out a level of automated quality
control, validation and calibration of the data underpinning the online service, that is
systematic, robust and commensurate with the above service objective. It is for the user to
decide whether or not this automated processing is sufficient for his needs (if in doubt through
direct contact with BMT ARGOSS), but to assist in the decision-making process this section
describes some key differences in methods and tools used in the online service and in the
more detailed consultancy projects.
The online service provides information about the normal climate only and use of
the online data for extreme value analysis is not advised, particularly if no additional
site specific calibration is carried out. BMT ARGOSS metocean experts routinely
undertake this work in consultancy projects, and can provide advice and assistance.
Information on tropical storms is not provided in waveclimate.com. Satellite sampling
of tropical storms is sporadic and very high wind speeds and wave heights in affected
areas (e.g. Central America and the USA, the western Pacific) are unreliable. These
events are also not well represented in model hindcast data (other than dedicated
storm hindcasts). Waveclimate.com issues a warning when an area has been
selected where tropical storms occur; in such cases the user should contact BMT
ARGOSS for a more detailed analysis based on storm track data not available via the
online service.
Waveclimate.com provides model data for the global grid ( 1x5/4), the
Mediterranean, the Black Sea, the Red Sea, the Caspian Sea and the Persian Gulf
(1/4x1/4). High resolution EU shelf data (1/6 x 1/6) is available for use outwith the
online service. See Figure 1 or Appendix C- Global and regional wave models.
The online service uses a traditional method to distinguish wind-sea and swell
components of a sea state. The wind-sea part of the wave spectrum is represented
by a Donnellan-Pierson spectrum and thus related to the corresponding wind. For
consultancy, wave steepness is used to separate wind-sea and swell: a distinct peak
of the wave spectrum with steepness above 0.03 will be classified as wind-sea.
Normally, at most two distinct spectral peaks are taken into consideration and
classified as either swell or wind-sea; additional swell peaks can be resolved if
required, for example off West Africa.
Altimeter and scatterometer data available through the online service are calibrated
separately for each mission, resulting in relative errors of 12% for altimeter wave
height and 15% for scatterometer wind speed after calibration with buoys (Table 2
and Table 3 in section 2.3). Recently, we created an improved set of satellite data
that we use to calibrate the wave model data available online; we also use this new
set of satellite data for consultancy. This improvement involves more advanced
processing of altimeter data (section 5.1) such as spike removal and the creation of
one set of altimeter data consistent over time. Altimeter data from all missions are
calibrated with a master satellite, merged and then calibrated with buoys. The
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 3
BMT ARGOSS
_____________________________________________________________________________________________________
relative error in the resulting best altimeter wave height now reduces to 11% (Table
20 in section 5.2).
As noted above, consultancy projects provide BMT ARGOSS metocean specialists
with the opportunity to undertake more detailed analyses than are carried out for the
Waveclimate.com service. These projects are planned in consultation with the client
and may include, for example, additional calibration using site specific, in-situ
measurements, extreme value analysis and the derivation of metocean conditions
(particularly waves) in sheltered or shallow water locations.
The use of the shallow water models available through the online service also
requires a certain level of expertise, for instance the choice of the offshore boundary
for the wave ray model and pragmatic interpretation of the results.
To optimise the web site performance, waveclimate.com
1.4.
Calibrates integrated wave model parameters (significant wave height and wave
period as explained in section 5.1), whereas in a typical project the wave spectra are
calibrated and hence all wave parameters, including direction are adjusted.
Retrieves encoded and compressed wave spectra from binary files; one file contains
the complete time series for one grid point. Compression introduces minor loss of
accuracy.
Does not append a high-frequency tail to the wave spectrum. As a consequence,
zero-crossing wave period in waveclimate.com is up to 10% higher in wind-sea
dominated areas.
Uses 25 spectral frequencies and 12 spectral directions, whereas 30 frequency and
24 direction bins are routinely analysed in consultancy projects.
Fitness for use of the online service

The information provided through waveclimate.com is intended for the preliminary appraisal of
metocean conditions, and without additional verification is not appropriate for engineering
design. If you are in any doubt as to the suitability of the information for your purpose, or you
would like to discuss more extensive metocean services, please contact us
(info@bmtargoss.com) and an experienced metocean advisor will be pleased to assist you.
Please also take note of the following important guideline for use of waveclimate.com:
The offshore climate data represent the average climate over the selected area or at the
selected location, so they are suitable for fully exposed sites in deep water. The
nearshore climate option, preferably the wave ray model (see Appendix D- The shallow
water wave ray model), should be used wherever sheltering occurs and on shallow water.
The shallow water models are meant to be applied in coastal areas of limited size, say of
up 200 kilometres wide. Translating offshore wave conditions over greater distances
might frustrate (better) wave propagation of the global model.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 4
BMT ARGOSS
_____________________________________________________________________________________________________
1.5.
Structure of this document

In chapter 2 an overview given of the content of the global database, data processing and,
based on these data, products offered by the online service. The processing and accuracy of
altimeter and scatterometer satellite data presented by waveclimate.com are explained in
chapter 3. SAR data is addressed in chapter 4. The processing and error statistics of wave
model data is described in chapter 5. Here, special attention is paid to the set of satellite data
used for model calibration (this dataset is based on but differs from the set of satellite data
described in chapter 3).
In the appendices A-E, detailed information is given concerning the set of buoys used (A), the
parameters used for the error statistics (B), the deep-water wave model (C), the shallowwater wave model (D) and the definition of wave and wind parameters mentioned in this
report (E).
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 5
BMT ARGOSS
_____________________________________________________________________________________________________
2.
Overview of waveclimate.com
The content of the database, data processing, data accuracy and the products offered by the
online service are explained in the next sections.
2.1.
Content of the database

The global database contains the wind and sea state data listed in Table 1 below.
Variable
Wave spectral density and mean
direction per frequency band,
together with coincident wind
speed and direction
speed and direction
Wave hindcast model

Global third generation
wave model Wavewatch III
driven by NCEP surface
wind analyses
Regional third generation
wave model Wavewatch III
driven by high-resolution
ECMWF surface wind
analyses
Variable
Significant wave height
Significant wave height and wind
speed
Sensor
Radar altimeter
Radar altimeter
Wind speed and direction
Scatterometer

speed and direction
Synthetic aperture radar

(SAR) and scatterometer
Grid
1 x 1 degrees
all major ocean
basins
Period covered
1992-2010
x degrees
- Mediterranean
- Red Sea
- Black Sea
- Caspian Sea
- Arabian Gulf
Satellite
Geosat
ERS-1/2
Topex/Poseidon
Jason-1/2,GFO,
Envisat
ERS-1/2,
Quikscat,
Metop-A
ERS-1/2
1992-2010
Period covered
Mar 1985 - Dec 1989
Aug 1991 - Dec 2010
Mar 1992 - Dec 2010
Apr 1993 -Jun 2003
Table 1 Wind and sea state data contained in the global data base at BMT ARGOSS.
2.2.
Data processing
Data processing concerns quality control, correction and calibration as explained in chapter 35 in this report. In summary:
Ambiguity in scatterometer wind direction has been removed at the supplying agency
by applying constraints on the spatial characteristics of the output wind field, such as
on rotation and divergence. At BMT ARGOSS, initial quality control of altimeter
(significant wave height and wind speed) and scatterometer (wind speed and
direction) data from our suppliers involves various automated procedures such as
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 6
BMT ARGOSS
_____________________________________________________________________________________________________
2.3.
range checks, checks for error flags, detection of outliers, check for consistency
between wind speed and wave height and for consistency in space. Next, significant
wave height (altimeter) and wind speed (altimeter and scatterometer) are corrected
for bias for each mission separately using in-situ data from buoys obtained from
NOAA and Environment Canada. See chapter 3.
SAR spectra with very low signal-to-noise levels and spectral features related to
surface slicks are removed. Some information on short waves, short swells in
particular, is missing in wave spectra retrieved from SAR spectra. This information is
obtained from ECMWF global wave model spectra. In the wave climate data
products, significant wave heights and wave periods derived from SAR data are
calibrated on-the-fly using altimeter wave height observations obtained over the same
area. See chapter 4.
Calibration coefficients for the wave model are determined offline for each model
point using the improved set of altimeter and scatterometer data (based on but
different from the dataset presented by the online service- mentioned in section 1.3
and described in 5.1). Calibration coefficients for wind speed are applied offline by the
wave model to the driving winds fields. The online service applies the calibration
coefficients determined for wave height, thus removing the systematic error from
wave height and wave period generated by the wave model. The creation of the set
of satellite data used for model calibration and the effect of this calibration are
explained in chapter 5.
Data accuracy
The accuracy of significant wave height and 1-hourly wind speed (at 10m above sea level)
obtained from the data sources in the database (listed in Table 1) is summarised below in
Table 2 and Table 3. Statistics are based on comparisons against buoy data (buoys are
depicted in Figure 2 and listed in Appendix A- Buoys used for validation).
The quality of the data is summarised in terms of the (relative) root-mean-square (RMS) error
explained in Appendix B- Parameters used for error statistics.
The two tables below list averages over the specified period. Apart from SAR wave height,
the figures reported relate to calibrated wave height and wind speed: altimeter and
scatterometer data were calibrated with buoy data before being used for model calibration.
Source
Altimeter
SAR
Wave model
Period
1985-2009
1993-2003
1992-2009
#Samples
34412
1317
1166792
Buoy mean (m)

2.16
2.31
2.08
RMS error (m)

0.30
0.44
0.42
Error (%)
12
17
17
Table 2 Accuracy of satellite and wave model wave height provided by the online service.
Source
Altimeter
Scatterometer
Wave model
Period
1992-2009
1992-2009
1992-2009
#Samples
31651
82901
1134695
Buoy mean (m/s) RMS error (m/s)

7.21
1.45
7.10
1.15
6.87
1.58
Error (%)
18
15
21
Table 3 Accuracy of satellite and wave model wind speed provided by the online service.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 7
BMT ARGOSS
_____________________________________________________________________________________________________
2.4.
Climate statistics provided by the online service

Statistical information is provided about the overall sea state (significant wave height, zerocrossing wave period, mean wave period, peak period, mean wave direction) but also about
wind-sea only, or swell only. Wind-sea consists of the waves having crests moving no faster
than 1.2 times the wind speed, so they are growing. Longer, and therefore faster moving,
waves are called swell. Statistics of wind-sea and swell are derived from SAR and wave
model data.
The wave climate can be derived offshore, i.e. at a fully exposed location or area in relatively
deep water, or nearshore, where the water is shallow and sheltering might occur. Translation
of offshore wave conditions to any nearshore location of interest can be done with a very
simple model or with a more advanced wave ray model (see Appendix D- The shallow water
wave ray model).
The wind and wave climate can be also determined for a particular season or month of
interest.
The table below indicates the various statistics which can be provided and the source of the
data from which the statistics are derived.
Statistic for waves or wind
Source of data
Offshore
Nearshore
Monthly distribution
Wave height (total sea state only)
Wind speed
Hindcast/altimeter
Hindcast/altimeter
Hindcast
Hindcast
Histogram
Wave height
Wave period
Wind speed
Hindcast/altimeter
Hindcast/altimeter
Hindcast/SAR
Hindcast
Hindcast
Hindcast
Scatter diagram 2D
Wave height vs. wave direction
Wave height vs. wave period
Wave height vs. wind speed
Wind speed vs. wind direction
Hindcast/scatterometer
Hindcast/altimeter
Hindcast/SAR
Hindcast/SAR
Hindcast
Hindcast
Hindcast
Hindcast
Scatter diagram 3D
Wave height vs. wave period per wave direction sector
Wave height vs. wind speed per wind direction sector
Wave height vs. wave period per wind speed class
Hindcast
Hindcast
Hindcast
Hindcast
Hindcast
Table 4 Available statistics and corresponding data sources.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 8
BMT ARGOSS
_____________________________________________________________________________________________________
3.
Altimeter and scatterometer data

The set of altimeter and scatterometer satellite data described in this chapter is available via
waveclimate.com. This dataset currently covers the years 1985-2010. Validation results of
altimeter are based on co-located altimeter and buoy data from Jan 1985 until Dec 2009. For
scatterometer, results are based on data from Mar 1992 until Dec 2009. Comparison is done
separately for the different missions. Satellite data from SAR is described in the next chapter.
The parameters compared to buoy observations are significant wave height (from altimeter)
and wind speed (from scatterometer and altimeter). First, error statistics of altimeter and
scatterometer data before calibration with buoy data are addressed per mission and per year
(paragraph 3.1). Next, the calibration method is explained (3.2) and the calibration coefficients
found for each mission are given (3.3). Finally, error statistics of the satellite data after
calibration with buoy data are presented (3.4).This set of satellite data calibrated per mission
with buoys is used by the online service.
3.1.
Error statistics of raw satellite data per mission and per year
Error statistics per mission
Below, the quality of the raw significant wave height obtained by the altimeters of the various
satellites is summarised. The Topex/Poseidon and the Geosat satellites deliver the most
accurate wave height estimates. Ers-1 altimeter data is somewhat less accurate. Table 5
shows the quality of the raw wave height obtained by the altimeters of the various satellites.
Satellite
No. of
samples
Buoy
Mean [m]
STD
Error [m]
Topex
Poseidon
Geosat
Ers-1
Ers-2
Jason-1
GFO
Envisat
Jason-2
7696
524
1598
2706
7433
5421
3849
4241
944
2.30
2.27
2.31
2.35
2.07
2.05
2.17
2.06
1.95
0.26
0.32
0.29
0.37
0.36
0.31
0.26
0.27
0.31
Correlation
Coefficient
[-]
0.98
0.97
0.97
0.96
0.96
0.96
0.98
0.98
0.96
Relative
RMSE
[%]
10
12
11
22
16
14
14
12
14
Bias
[m]
0.02
0.02
-0.03
-0.46
0.16
0.08
0.23
0.07
-0.03
Table 5 Error statistics of raw wave height from altimeter based on all buoys.
Raw wind speed is most reliably estimated by Quikscat and the Ers scatterometers. Wind
speed from Geosat altimeter is much less accurate than wind speed from other sources.
Therefore, Geosat wind speed is not used by the online service Table 6 and Table 7 show the
quality of the raw wind speed obtained by the various missions.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 9
BMT ARGOSS
_____________________________________________________________________________________________________
Satellite
No. of
samples
Buoy
Mean [m/s]
STD
Error [m/s]
Topex
Poseidon
Geosat
Ers-1
Ers-2
Jason-1
GFO
Envisat
Jason-2
7696
524
1598
2706
7433
5421
3849
4241
944
7.50
7.86
7.74
7.82
7.05
6.99
7.01
6.98
6.84
1.55
1.70
2.47
1.59
1.57
1.47
1.43
1.37
1.55
Correlation
Coefficient
[-]
0.91
0.90
0.77
0.91
0.91
0.91
0.93
0.92
0.90
Relative
RMSE
[%]
19
20
30
19
21
19
19
18
22
Bias
[m/s]
0.06
0.10
-0.69
-0.51
0.37
0.21
0.37
0.12
0.64
Table 6 Error statistics of raw wind speed from altimeter based on all buoys.
Satellite
No. of
samples
Buoy
Mean [m/s]
STD
Error [m/s]
Ers-1
Ers-2
Quikscat
9402
8778
64721
7.85
7.46
6.94
1.20
1.15
1.14
Correlation
Coefficient
[-]
0.94
0.94
0.94
Relative
RMSE
[%]
15
16
15
Bias
[m/s]
-0.54
-0.56
0.25
Table 7 Error statistics of raw wind speed from scatterometer based on all buoys.
Error statistics per year

We also analysed the statistics per satellite-sensor combination per year for all buoys in order
to search for irregularities over time. We provide some examples to illustrate the procedure.
Bias (satellite observation minus buoy observation) and standard deviation of the error in raw
Topex wave height is depicted below.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 10
BMT ARGOSS
_____________________________________________________________________________________________________
Figure 3 Bias and STD of error of raw wave height per year of Topex altimeter and all buoys.
From the figure above, we see that the bias in Topex raw wave height increases significantly
in 1998. This appears to be caused by a performance degradation of the chirp generator. A
switchover to the backup altimeter was successfully realised on Feb 9, 1999. Therefore, we
derived separate calibration coefficients for 1998 for wave height and wind speed from Topex.
Table 8 and Table 9 show the error statistics of Topex data for 1998 and for the years 19922005, excluding 1998.
Period
1998
1992 2005
(excl. 1998)
No. of
samples
Buoy
Mean [m]
STD
Error [m]
Relative
RMSE
[%]
14
Bias
0.32
Correlation
Coefficient
[-]
0.97
498
2.41
7198
2.29
0.25
0.98
10
0.04
[m]
0.21
Table 8 Error statistics of raw wave height from Topex altimeter based on all buoys.
Period
1998
1992 - 2005
(excl. 1998)
No. of
samples
Buoy
Mean [m/s]
STD
Error [m/s]
1.45
Correlation
Coefficient
[-]
0.91
Relative
RMSE
[%]
17
498
7.96
7198
7.46
Bias
[m/s]
0.12
1.59
0.91
19
0.05
Table 9 Error statistics of raw wind speed from Topex altimeter based on all buoys.
The quality of wind speed measured by Geosat follow-on (GFO) deteriorates as of 2007 as
shown below in Figure 4. For this reason, the online service only uses GFO wind speed from
the years 2002-2006.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 11
BMT ARGOSS
_____________________________________________________________________________________________________
Figure 4 Bias and STD of error of raw wind speed per year of GFO altimeter and all buoys.
The next plot does show however that wave height observations from GFO are reliable for all
years.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 12
BMT ARGOSS
_____________________________________________________________________________________________________
Figure 5 Bias and STD of error of raw wave height per year of GFO altimeter and all buoys.
3.2.
Calibration of satellite data with buoys

Least square fits (see Figure 6 below) of linear relationships between raw satellite data and
buoy data of wave height and wind speed were determined per satellite-sensor pair. Satellite
data are on the vertical axis, buoy data on the horizontal axis. For reference, we also plotted
the y=x line and the sorted satellite observations against the sorted buoy observations.
The fitting procedure applied is Total Least Squares, minimising the sum of squares of the
residuals measured orthogonal to the fitted line. Minimising distances in y-direction only, e.g.
assuming noise in satellite data only, proved to be sensitive to switching x and y, indicating
that this procedure would result in seriously biased estimates.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 13
BMT ARGOSS
_____________________________________________________________________________________________________
The figure below (Figure 6) shows the fit of raw Topex wave height against buoy
measurements.
Figure 6 Least squares fit of raw wave height for Topex altimeter and all buoys.
Calibration coefficients derived from these fits, i.e. slope () and intercept (), were found per
satellite-sensor combination in order to calibrate the raw satellite data:
H icorrected H iraw
The use of non-linear curve fitting did not improve the fits significantly. Note that (the
smoothed version of) the sorted satellite versus buoy observations curve is such a non-linear
curve.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 14
BMT ARGOSS
_____________________________________________________________________________________________________
3.3.
Calibration coefficients found for satellite data

Calibration coefficients for Topex are taken year-dependent. For the other satellites,
calibrations are valid for all relevant years. The following calibration coefficients were found:
Satellite
Topex
Topex
Poseidon
Geosat
Ers-1
Ers-2
Jason-1
GFO
Envisat
Jason-2
Period
1998
1992-2005
excl. 1998
1992-2002
1985-1989
1991-1996
1995-2009
2002-2009
2002-2008
2003-2009
2008-2009
[-]
1.07
1.06
[m]
-0.40
0.09
1.02
0.98
1.14
1.06
1.05
1.08
1.05
1.02
-0.05
0.09
0.20
0.04
0.03
0.07
-0.19
-0.00
Table 10 Calibration coefficients found for wave height from altimeter based on all buoys.
Satellite
Topex
Topex
Poseidon
Geosat
Ers-1
Ers-2
Jason-1
GFO
Envisat
Jason-2
Period
1998
1992-2005
excl. 1998
1992-2002
1985-1989
1991-1996
1995-2009
2002-2009
2002-2006
2003-2009
2008-2009
[-]
1.00
0.93
[m]
-0.15
0.51
0.95
0.94
1.03
0.90
0.93
0.87
0.95
0.88
0.34
1.11
0.27
1.03
0.66
0.61
0.22
0.23
Table 11 Calibration coefficients found for wind speed from altimeter based on all buoys.
Satellite
Ers-1
Ers-2
Quikscat
Period
1992-1996
1996-2000
2000-2009
[-]
1.07
1.06
1.00
[m/s]
0.02
0.10
-0.24
Table 12 Calibration coefficients found for wind speed from scatterometer based on all buoys.
We apply the above calibration to the raw satellite data except for Geosat winds. Winds from
Geosat are too inaccurate, so they are not used by the online service. From Table 10-Table
12, we see that calibration increases wave height by up to 14% (for Ers-1 altimeter). They
also confirm that, at least in open sea ares, scatterometer winds are more reliable than
altimeter winds.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 15
BMT ARGOSS
_____________________________________________________________________________________________________
3.4.
Error statistics of satellite data after calibration with buoys

As a result of applying the above calibration coefficients (Table 10-Table 12) to the raw
satellite data, we get the following error statistics of the calibrated satellite data:
Satellite
No. of
samples
Buoy
Mean [m]
STD
Error [m]
Topex
Poseidon
Geosat
Ers-1
Ers-2
Jason-1
GFO
Envisat
Jason-2
7696
524
1598
2706
7433
5421
3849
4241
944
2.30
2.27
2.31
2.35
2.07
2.05
2.17
2.06
1.95
0.25
0.32
0.29
0.36
0.36
0.31
0.26
0.27
0.30
Correlation
Coefficient
[-]
0.98
0.97
0.97
0.96
0.96
0.96
0.98
0.98
0.97
Relative
RMSE
[%]
10
12
11
14
15
13
10
11
13
Bias
[m]
0.00
0.02
0.02
0.01
-0.01
-0.01
-0.01
-0.01
-0.00
Table 13 Error statistics of calibrated wave height from altimeter based on all buoys.
Satellite
No. of
samples
Buoy
Mean [m/s]
STD
Error [m/s]
Topex
Poseidon
Geosat
Ers-1
Ers-2
Jason-1
GFO
Envisat
Jason-2
7696
524
1598
2706
7433
5421
3849
4241
944
7.50
7.86
7.74
7.82
7.05
6.99
7.01
6.98
6.84
1.51
1.64
2.39
1.61
1.46
1.40
1.26
1.33
1.41
Correlation
Coefficient
[-]
0.91
0.90
0.77
0.91
0.91
0.91
0.93
0.92
0.90
Relative
RMSE
[%]
18
19
28
19
19
18
16
17
19
Bias
[m/s]
0.02
0.04
-0.00
-0.01
-0.01
-0.03
0.02
-0.01
-0.03
Table 14 Error statistics of calibrated wind speed from altimeter based on all buoys.
Satellite
No.
of
samples
Ers-1
Ers-2
Quikscat
9402
8778
57408
Buoy
Mean [m/s]
7.85
7.46
6.94
STD
Error [m/s]
1.22
1.16
1.12
Correlation
Coefficient
[-]
0.94
0.94
0.94
Relative
RMSE
[%]
14
14
15
Bias
[m/s]
-0.01
0.03
0.03
Table 15 Error statistics of calibrated wind speed from scatterometer based on all buoys.
Relative RMSE of the calibrated satellite data (listed in Table 13-Table 15) is indeed smaller
than relative RMSE of the raw satellite data (listed in Table 5-Table 7). Error in wave height
from Ers-1 altimeter (22% to 14%) and GFO altimeter (14% to 10%) reduces significantly.
Comparable error reductions are found for altimeter wind speed from GFO (19% to 16%) and
Jason-2 (22% to 19%).
As a confirmation of the correction procedure, example plots are given on the next pages to
illustrate the effect of calibration. The next two figures illustrate the removal of bias from GFO
wave height through calibration.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 16
BMT ARGOSS
_____________________________________________________________________________________________________
Figure 7 Bias and STD of error of raw wave height per year of GFO altimeter and all buoys.
Figure 8 Bias and STD of error of corrected wave height per year of GFO altimeter and all buoys.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 17
BMT ARGOSS
_____________________________________________________________________________________________________
The plots below show fits of Ers-1 wave height and buoy data before and after the calibration.
Figure 9 Least squares fit of raw wave height for Ers-1 altimeter and all buoys.
Figure 10 Least squares fit of corrected wave height for Ers-1 altimeter and all buoys.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 18
BMT ARGOSS
_____________________________________________________________________________________________________
The effect of the calibration of Jason-2 wind speed is illustrated below.
Figure 11 Least squares fit of raw wind speed for Jason-2 altimeter and all buoys.
Figure 12 Least squares fit of corrected wind speed for Jason-2 altimeter and all buoys.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 19
BMT ARGOSS
_____________________________________________________________________________________________________
4.
SAR data
4.1.
Error statistics of raw SAR data per region

The set of SAR satellite data described in this chapter is available via waveclimate.com.
Validation results are based on SAR data and buoy data from Apr 1993 until Jun 2003.
Satellite data from altimeters and scatterometers is described in the previous chapter.
For the validation, the buoys are grouped into 5 geographical regions: Atlantic, Gulf of
Mexico, Newfoundland, Pacific and Hawaii. The wave climate within a region is more or less
uniform. The buoys and regions are depicted in Figure 2 and listed in Appendix A- Buoys
used for validation.
The results of the most important statistic parameters are given for the significant wave
height, mean period and zero-crossing period. The spectral SAR data are especially useful for
wave climates that exhibit large waves (wavelength > 200m and period > 12 sec). Therefore
we also considered the wave height for long waves (wave period exceeds 12 seconds).
The statistics for all buoys and years are collected in the tables below for the merged
SAR/ECMWF spectra.
Atlantic
N
Mean [m]
Bias [m]
StDev [m]
Rrmse [%]
Corr
348
1.97
0.08
0.50
22
0.89
Gulf of Mexico Newfoundland

193
1.29
0.06
0.33
22
0.92
69
2.85
0.07
0.59
19
0.91
Pacific
Hawaii
454
2.95
0.04
0.46
14
0.95
313
2.25
0.08
0.33
14
0.88
Table 16 Statistics of raw wave height based on SAR/ECMWF spectra per region.
Atlantic
N
Mean [s]
Bias [s]
StDev [s]
Rrmse [%]
Corr
348
7.20
0.14
0.99
14
0.68

193
5.92
0.34
1.60
29
0.42
69
8.61
0.26
0.90
11
0.79
Pacific
Hawaii
454
9.32
0.29
0.73
9
0.93
313
8.80
0.21
0.63
8
0.93
Table 17 Statistics of raw mean period based on SAR/ECMWF spectra per region.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 20
BMT ARGOSS
_____________________________________________________________________________________________________
Atlantic
N
Mean [s]
Bias [s]
StDev [s]
Rrmse [%]
Corr
348
5.88
0.00
0.75
13
0.70

193
4.95
0.01
1.27
25
0.40
69
6.98
0.25
0.83
13
0.74
Pacific
Hawaii
454
7.30
0.24
0.61
9
0.92
313
6.80
0.14
0.53
8
0.92
Table 18 Statistics of raw zero-crossing period based on SAR/ECMWF spectra period per region.
Atlantic
N
Mean [m]
Bias [m]
StDev [m]
Rrmse [%]
Corr
348
0.40
0.08
0.36
71
0.64

193
0.17
0.13
0.22
135
0.64
69
1.05
0.14
0.50
41
0.85
Pacific
Hawaii
454
1.47
0.12
0.44
25
0.94
313
0.94
0.03
0.30
27
0.90
Table 19 Statistics of raw height of long waves based on SAR/ECMWF spectra per region.
The best results are obtained in the Pacific and Hawaii and the worst results in the Gulf of
Mexico. This can be explained by the fact that SAR data is especially useful for a wave
climate that includes long waves, which are lacking in the Gulf of Mexico. This is also
reflected by the results in Table 19, i.e. the mean wave height for long waves in the Gulf of
Mexico is a fraction of the mean wave heights for the other regions.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 21
BMT ARGOSS
_____________________________________________________________________________________________________
4.2.
Calibration of SAR data

Wave heights and wave periods derived from SAR are calibrated on-the-fly by the online
service with altimeter wave heights, in order to ensure consistency of the wave height
distributions from both data sources. The correction is performed on-the-fly as it depends on
the offshore area under investigation. Corrections of a few percent are typical. Maximum
correction for SAR wave height is about 15%. Note that this calibration has not been taken
into account in the validation results for SAR wave height presented in the previous section.
It is assumed that wave heights from altimeter and SAR are statistically correlated (altimeter
and SAR wave height samples used come from the same area but are not collocated in time).
Correction of total SAR wave height (based on the total spectrum) is done as follows:
H icorrected H i2 si
where i counts the SAR samples, H icorrected and H i denote the corrected and total SAR wave
height respectively, is slope and is intercept. The factor s i ensures a smooth transition
near zero. It is found as
si 1 exp( H i2
To correct height of wind-sea or swell, the ratio of corrected and uncorrected total wave
height is used:
ri
H icorrected
Hi
Hswellicorrected ri Hswelli
Hseaicorrected ri Hseai
where Hswell and Hsea satisfy
Hswelli2 Hseai2 H i2 .
Calibration of all wave periods (zero-crossing and mean wave period, either corresponding to
the total spectrum, the wind-sea part or the swell part), makes use of the same ratio:
Ti corrected ri Ti
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 22
BMT ARGOSS
_____________________________________________________________________________________________________
Slope () and intercept () relate to the coefficients of the linear regression fit of SAR total
wave energy quantiles to altimeter wave energy quantiles:
2
2
H SAR
a H ALT
b
with a residual. Regression of SAR energy on altimeter energy assumes that altimeter wave
height is more accurate than SAR wave height, which is what we find in the validation against
buoys. In the regression, we minimize the sum of squares of
max( 0.25, H ALT ) , so we
assume that the standard deviation of the residual is proportional to the altimeter wave height.
In the correction of total SAR wave height above,
1 / a and b / a .
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 23
BMT ARGOSS
_____________________________________________________________________________________________________
5.
Wave model data

In waveclimate.com, only calibrated wave model data is available. Validation results of the
wave model presented in this section are based on model data, re-processed satellite data
and buoy data from 1992 until 2009.
The satellite data used for model calibration described in this section is used for our
consultancy projects but not (yet) by waveclimate.com. Waveclimate.com uses satellite data
calibrated for each mission separately by buoy data as described in chapter 3 of this
document.
The first section below (5.1) explains how model calibration coefficients are created offline,
based on satellite measurements. Starting off with quality-checked, but still un-calibrated
satellite data (described in section 3.1), it is shown how the various processing steps improve
the quality of the satellite data to be used for model calibration (5.2). Finally, the effect of
model calibration with satellite data is measured against buoys (5.3).
5.1.
Calibration of model data

In this context, model calibration means calibration of model parameters by means of satellite
observations in order to remove any systematic model error (bias) at all grid points. Relevant
model parameters are significant wave height, wave periods, and wind speed. Measurements
of significant wave height are supplied by altimeters; observations of wind speed come from
both altimeters and scatterometers. This set of quality-checked, but still un-calibrated, satellite
data (described in chapter 3) is referred to as original satellite data. The calibrated satellite
data, used for model calibration, are referred to as best satellite data;
The following processing steps are taken to derive calibration coefficients for model wave
height and wind speed, based on best satellite observations:
1. Add model data, satellite altimeter and scatterometer data to a matrix of spatial bins,
one month at the time. Add buoy data and compute error statistics for the original
data over the years. The size of spatial bins is set to 5/8x1/2 for the global grid and
to 1/4x1/4 for the Mediterranean. In this way, grid points coincide with the center of
a bin.
2. Improve the quality of altimeter measurements
a. Remove spikes trough median-filtering
b. Apply mission-dependent a-priori corrections, e.g. removal of extreme winds
and waves from Jason-2 reported during the first month of the mission (July
2008)
c.
Calibrate each satellite with a master satellite (Topex/Jason-2) and merge

the set of altimeter data to get a consistent data set over the years
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 24
BMT ARGOSS
_____________________________________________________________________________________________________
d. Calibrate the merged altimeter data with buoy observations (co-locate within
50km and 30 minutes)
3. Calibrate scatterometer wind speed per satellite with buoys measurements (co-locate
within 50 km and 30 minutes)
4. Merge calibrated altimeter and calibrated scatterometer wind speed observations
Based on the number of measurements available, scatterometer data will be
dominant in the calibration on open sea whereas altimeter will gradually take over
towards the coast.
5. Compute model-satellite error statistics, including calibration coefficients (scale and
intercept from linear regression analysis). Model-satellite samples are co-located
within 50km. In time, model data are interpolated onto satellite observation dates.
Model wave height is calibrated with altimeter data while model wind speed is
calibrated with merged altimeter and scatterometer measurements.
6. Export calibration coefficients (scale and intercept to correct model wave height and
wind speed) per model grid point, based on best satellite data. During this step,
special attention is paid to potentially suspect points like boundaries, points with too
few co-located samples and points that provide large corrections. At boundary points,
calibration coefficients are copied from the nearest internal point. No calibration is
applied to points with too few samples. Scale and intercept are bounded at points for
which the ratio of corrected and original model mean is too small or too large.
A minimum of 100 co-located satellite samples is required for calibration of model
wave height or wind speed. For wind speed, from both scatterometer and altimeter,
there is an extra demand: the samples need to be within 15 km. This additional
demand ensures that satellite wind speed samples are centered about the center of
the spatial bin. For the same reason, calibration coefficients at boundary points are
copied from the nearest internal point.
The above steps result in sets of calibration coefficients for model wave height and wind
speed based on best satellite data.
Wind speed is calibrated before it is used by the wave model, whereas wave parameters are
calibrated afterwards. Waveclimate.com only uses calibrated model data. The calibration
coefficients for significant wave height and wind speed are computed offline and per wave
model grid point. Linear regression is performed on wave model and satellite samples
collocated in space and time. Slope and intercept are applied as follows:
H icorrected j H i j ,
where i is the sample index and j the grid point index. Calibration of height of wind-sea, height
of swell and wave periods re-uses the ratio of calibrated and un-calibrated total significant
wave height, similar to the method used for SAR (described in section 4.2 above).
Buoys are grouped into the same 5 geographical regions as used for validation of SAR
(section 4.1). In the next sections, original and calibrated satellite and model data are
compared to true measurements from buoys.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 25
BMT ARGOSS
_____________________________________________________________________________________________________
5.2.
Error statistics of satellite data used for model calibration

The tables below summarize the effect of each processing step (steps 2a-d, explained in the
previous section) in terms of altimeter-buoy error statistics.
Avg (m)
2.03
2.04
2.09
2.07
2.15
Bias (m)
-0.12
-0.11
-0.06
-0.08
0.00
Std (m)
0.36
0.31
0.29
0.28
0.28
Rrms (%)
15.00
13.30
11.80
11.40
11.00
Nr
94450
93024
93004
93004
93004
original
filtered
a priori
master
best
Table 20 Altimeter-buoy error statistics for wave height per processing level
Avg(m/s)
7.21
7.25
7.19
7.45
7.06
Bias (m/s)
0.17
0.20
0.15
0.41
0.02
Std (m/s)
1.62
1.60
1.56
1.55
1.43
Rrms (%)
20.90
20.60
20.00
20.60
18.30
Nr
91555
90144
87604
87604
87604
original
filtered
a priori
master
best
Table 21 Altimeter-buoy error statistics for wind speed per processing level
Each processing step improves the quality of the altimeter wave height and wind speed. For
significant wave height, the relative error reduces from 15% for original data to 11% for best
data (see Table 20). Note that the removal of spikes by median filtering significantly reduces
the variability of altimeter wave height. For both wave height and wind speed, biases are
mainly removed by a priori corrections and (of course) by buoy calibration. Altimeter wind
speed error is reduced from 21% to 18% (see Table 21).
The above process yields one set of altimeter data consistent over time. This is illustrated by
the plots below that show the error in original and calibrated altimeter wave height relative to
buoys over the years (Figure 13 and Figure 14).
Calibration of altimeter waves improves the quality over the whole range of wave heights as
illustrated by the probability of exceedance (PoE) plots below (Figure 15 and Figure 16).
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 26
BMT ARGOSS
_____________________________________________________________________________________________________
The figures below show the error in raw and best altimeter wave height over the years.
Error in wave height of alt minus buoy per class
1
Bias and STD of error in wave height (m)
0.8
Bias
STDE
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09
Classes
Figure 13 Error in original altimeter wave height over the years
Error in wave height of alt minus buoy per class

1
Bias and STD of error in wave height (m)
0.8
Bias
STDE
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09
Classes
Figure 14 Error in calibrated altimeter wave height over the years
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 27
BMT ARGOSS
_____________________________________________________________________________________________________
The next two plots focus on the effect of calibration on more extreme altimeter waves.
PoE of wave height for 1992-2009
10
buoy
alt
-1
Fraction of time exceeded
10
-2
10
-3
10
-4
10
-5
10
6
8
wave height [m]
10
12
14
Figure 15 Prob. of exceedance of raw altimeter and buoy wave height
10
buoy
alt
-1
10
-2
10
-3
10
-4
10
-5
10
6
8
wave height [m]
10
12
14
Figure 16 Prob. of exceedance of calibrated altimeter and buoy wave height
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 28
BMT ARGOSS
_____________________________________________________________________________________________________
Scatterometer data is still calibrated with buoys per mission (unlike the altimeters, there is
hardly any overlap in time between the various missions). The reduction of the relative error in
wind speed (see Table 22) is less spectacular than for altimeter: 15.6% for original
scatterometer winds and 15.2% error for the buoy-calibrated winds. Apparently, original
scatterometer winds match buoy winds quite well already (at least much better than altimeter
winds with errors around 21%).
Avg(m/s)
7.12
7.12
Bias (m/s)
0.13
0.05
Std (m/s)
1.20
1.17
Rrms (%)
15.60
15.20
Nr
158800
158800
original
best
Table 22 Scatterometer-buoy error statistics for wind speed per processing level
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 29
BMT ARGOSS
_____________________________________________________________________________________________________
5.3.
The effect of model calibration with satellites according to buoys

The next table (Table 23) summarizes the effect of model calibration with satellites in terms of
the relative buoy-model error for significant wave height and wind speed. The abbreviations in
the first column relate to the areas of interest. The second column gives the relative error in
original model wave height (raw). Errors of altimeter-calibrated wave height are given in the
next column (best). The last two columns give the relative error in model wind speed before
(raw) and after calibration (best) with best merged scatterometer and altimeter data.
Rel. error
Wave height
Wind speed
(%)
(%)
(%)
(%)
GOM
21.6
21.2
22.0
22.4
ATL
22.5
20.6
24.1
23.3
NFL
20.7
20.9
25.7
24.6
PAC
16.6
16.0
19.7
18.9
HAW
15.2
14.1
17.4
17.3
averaged
18.4
17.4
21.3
20.7
raw
best
raw
best
Table 23 Relative model-buoy error per region before and after calibration with satellites
From Table 23 it becomes clear that, on average and for most regions, satellite calibration
reduces the model-buoy error in wave height and wind speed. Significant improvements are
found for waves at buoys in the Atlantic and around Hawaii (more than the average 1% error
reduction). This improvement in relative error is mainly caused by reduction of the biases
(variation does not change much), shown in the table below (Table 24).
Bias
Wave height
(m)
(m)
GOM
-0.06
ATL
Wind speed
(m/s)
(m/s)
0.04
0.00
-0.08
-0.19
-0.05
0.55
0.18
NFL
-0.13
0.00
0.71
0.04
PAC
-0.03
-0.02
0.36
-0.06
HAW
0.05
0.00
-0.09
-0.12
averaged
-0.07
-0.01
0.27
-0.01
raw
best
raw
best
Table 24 Bias model-buoy per region before and after calibration with satellites
The above table (Table 24) shows that the systematic bias in model wave height is reduced
for all regions with about 50% or more. Bias in wind speed is strongly reduced in most areas.
The final bias in model wave height relative to buoys is 5cm or less while wind speed is
biased by less than 20cm/s.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 30
BMT ARGOSS
_____________________________________________________________________________________________________
The plots on the next pages (Figure 17-Figure 28) illustrate the results printed in Table 24.
The positive effect of model calibration with satellites is most apparent at buoy locations in the
Atlantic (Figure 17 and Figure 18) and offshore New Foundland (Figure 19 and Figure 20).
In the Gulf of Mexico (Figure 21 and Figure 22), hurricanes seem to be a problem. Extreme
model waves, say higher than 10m, are too high and this is also true for wind speed above 20
m/s (Figure 23). Calibration of model wave height with altimeter cannot change this because
there are no buoy-altimeter co-locations above 10m (Figure 24). Note that model waves
below 5m, corresponding to about 99.8% of the samples, do improve thanks to calibration
with altimeter (compare Figure 21 and Figure 22).
In the Pacific (Figure 25 and Figure 26) calibration with altimeter does not change much. At
the buoys near Hawaii (Figure 27 and Figure 28), model calibration improves the lower swell
(swell below 4m representing roughly 98% of the data), which is too high, at the expense of
the higher swell.
The exceedance plots below (Figure 17-Figure 28) confirm that calibration with satellites
improves normal model waves in all regions. In general calibration with satellites improves
(Atlantic, New Foundland) or does not affect (Gulf of Mexico and Pacific) model extremes. But
there are exceptions (Hawaii).
Again it must be emphasized that the online service focuses on the normal wave and wind
climate and that extreme value analysis demands additional expertise as stated earlier in
section 1.3.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 31
BMT ARGOSS
_____________________________________________________________________________________________________
The graphs of probability of exceedance (PoE) below clearly show the improvement of model
wave height due to calibration with altimeter according to the buoys in the Atlantic.
10
atl
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
6
8
wave height [m]
10
12
14
Figure 17 Prob. of exceedance of raw model and buoy wave height for Atlantic
10
atl
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
10
15
wave height [m]
Figure 18 Prob. of exceedance of altimeter-calibrated model and buoy wave height for Atlantic
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 32
BMT ARGOSS
_____________________________________________________________________________________________________
The figures below show the positive effect of calibration of model wave height with altimeter
according to the buoys offshore New Foundland.
10
nfl
model
-1
10
-2
10
-3
10
-4
10
-5
10
10
wave height [m]
15
20
Figure 19 Prob. of exceedance of raw model and buoy wave height for NFL
10
nfl
model
-1
10
-2
10
-3
10
-4
10
-5
10
10
wave height [m]
15
20
Figure 20 Prob. of exceedance of altimeter-calibrated model and buoy wave height for NFL
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 33
BMT ARGOSS
_____________________________________________________________________________________________________
The next plots show the effect of model calibration of model with altimeter according to the
buoys in the Gulf of Mexico. Waves below 5m improve. Waves above 10m remain too high.
10
gom
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
10
wave height [m]
15
20
Figure 21 Prob. of exceedance of raw model and buoy wave height for GOM
10
gom
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
10
wave height [m]
15
20
Figure 22 Prob. of exceedance of altimeter-calibrated model and buoy wind speed for GOM
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 34
BMT ARGOSS
_____________________________________________________________________________________________________
Extreme model winds (>20m/s) are also too high in the Gulf of Mexico (Figure 23). Co-located
altimeter-buoy wave heights do not cover the model extremes above 10m (Figure 24).
PoE of wind speed for 1992-2009
10
gom
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
10
20
30
wind speed [m/s]
40
50
Figure 23 Prob. of exceedance of satellite-calibrated model and buoy wind speed for GOM
10
gom
alt
-1
10
-2
10
-3
10
-4
10
-5
10
6
8
wave height [m]
10
12
Figure 24 Prob. of exceedance of altimeter and buoy wave height for GOM
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 35
BMT ARGOSS
_____________________________________________________________________________________________________
The next two pictures show the (negligible) effect of calibration of model wave height with
altimeter according to the buoys in the Pacific.
10
pac
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
6
8
10
wave height [m]
12
14
16
Figure 25 Prob. of exceedance of raw model and buoy wave height for PAC
10
pac
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
6
8
10
wave height [m]
12
14
16
Figure 26 Prob. of exceedance of altimeter-calibrated model and buoy wave height for PAC
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 36
BMT ARGOSS
_____________________________________________________________________________________________________
The plots below show the effect of calibration of model wave height with altimeter according
to the buoys near Hawaii. Lower swell improves at the cost of higher swell.
10
haw
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
4
6
wave height [m]
10
Figure 27 Prob. of exceedance of raw model and buoy wave height for HAW
10
haw
model
-1
10
-2
10
-3
10
-4
10
-5
10
-6
10
4
6
wave height [m]
10
Figure 28 Prob. of exceedance of altimeter-calibrated model and buoy wave height for HAW
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 37
BMT ARGOSS
_____________________________________________________________________________________________________
5.4.
The applicability of model calibration with satellites worldwide

Model wave height is calibrated by altimeter and model wind speed by merged altimeter and
scatterometer winds (see section 5.1). The plots below show the remaining bias in model
wave height and wind speed after calibration with satellites for the global and the local grids.
Figure 29 Bias in global model wave height after calibration with satellites
Figure 30 Bias in global model wind speed after calibration with satellites
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 38
BMT ARGOSS
_____________________________________________________________________________________________________
Figure 31 Bias in model wave height in the Mediterranean after calibration with satellites
Figure 32 Bias in model wind speed in the Mediterranean after calibration with satellites
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 39
BMT ARGOSS
_____________________________________________________________________________________________________
The figures above show that, except for the polar region where ice causes problems (e.g. lack
of samples), calibration with satellites reduces the systematic model error equally in all of the
worlds oceans and seas (Figure 29 and Figure 30), including the semi-closed basins (Figure
31 and Figure 32), to less than about 5 cm for wave height and to less than 25 cm/s for wind
speed. This means that there isnt any significant variation in satellite performance and so we
may assume that the calibration of the model with satellites in regions with buoys (around
Northern America) is also applicable to any other (non-polar) region on the globe, including
the semi-closed basins.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 40
BMT ARGOSS
_____________________________________________________________________________________________________
Appendices
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 41
BMT ARGOSS
_____________________________________________________________________________________________________
Appendix A- Buoys used for validation

The set of buoys depicted in Figure 2 has been divided into 5 regions with a more or less uniform
wave climate:
The Gulf of Mexico (GOM, buoy numbers start with 42)
The northern Atlantic east of Northern America (ATL, buoy numbers start with 41 or 440)
Offshore Newfoundland (NFL, buoy numbers start with 441)
The northern Pacific (PAC, buoy numbers start with 46)
The region around Hawaii (HAW, buoy numbers start with 51)
The buoys report hourly wind speed, significant wave height, zero-crossing wave period and
mean wave period. Some buoys provide spectral information.
The table below lists each buoys position, region, observation period and the number of
observations available.
Buoy
41001
Nobs
149032
Lat
34.68
Lon
-72.64
Region
ATL
Begin
01jan85
End
08jun08
41002
150946
32.27
-75.19
ATL
27feb85
30nov08
41006
99259
29.30
-77.40
ATL
26may82
21apr96
41010
250045
28.90
-78.53
ATL
10nov88
31dec09
42001
196010
25.93
-89.65
GOM
01jan85
31dec09
42002
193012
25.89
-93.57
GOM
01jan85
31dec09
42003
181573
25.94
-85.91
GOM
01jan85
31dec09
42039
64900
28.80
-86.06
GOM
01jan02
31dec09
42040
62004
29.21
-88.20
GOM
01jan02
05oct09
42041
23485
27.50
-90.46
GOM
08may02
16mar05
44004
158983
38.46
-70.69
ATL
01jan85
08mar08
44005
162817
42.90
-68.94
ATL
01jan85
31dec09
44008
63575
40.50
-69.43
ATL
01jan01
06dec09
44011
59007
41.08
-66.58
ATL
01jan01
22sep09
44137
36354
41.30
-61.40
NFL
30nov88
15oct97
44139
33325
44.30
-57.40
NFL
02dec88
21nov97
44140
25371
42.70
-50.60
NFL
05sep90
19nov96
44141
42892
42.10
-56.10
NFL
05sep90
08dec97
46001
179334
56.29
-148.18
PAC
01jan85
31dec09
46002
156312
42.53
-130.26
PAC
25jan85
28jul09
46003
100426
51.85
-155.92
PAC
01jan85
11aug99
46004
52622
50.90
-135.90
PAC
04aug88
31dec97
46005
158951
46.08
-131.00
PAC
01jan85
18dec08
46006
139094
40.84
-137.49
PAC
01jan85
21nov08
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 42
BMT ARGOSS
_____________________________________________________________________________________________________
46035
167028
56.91
-177.81
PAC
13sep85
31dec09
46036
69879
48.30
-133.90
PAC
22sep87
31dec97
46047
66517
32.43
-119.53
PAC
01jan02
31dec09
46059
111158
37.98
-130.00
PAC
19oct94
09jan09
46066
46695
52.70
-155.00
PAC
01jan02
31dec09
46184
61634
54.00
-138.80
PAC
20sep87
31dec97
51001
180005
23.40
-162.30
HAW
01jan85
24dec09
51002
180215
17.20
-157.80
HAW
01jan85
31dec09
51003
179918
19.10
-160.80
HAW
01jan85
31dec09
51004
173696
17.40
-152.50
HAW
13feb85
07oct09
51026
30383
21.40
-157.00
HAW
16jan93
23nov96
51028
62081
0.00
-153.90
HAW
29oct97
14apr08
Table 25 List of NOAA buoys used for validation.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 43
BMT ARGOSS
_____________________________________________________________________________________________________
Appendix B- Parameters used for error statistics

The error is defined as the difference between samples of two data sources, e.g. between
satellite observations and buoy measurements, so for wave height:
ei H isatellite H ibuoy
with H isatellite the wave height of sample no. i retrieved from satellite data, and Hibuoy the
coincident buoy measurement of the wave height. Coincident means that the retrieved
satellite observation is located within 50 km around the buoy and that the observations do not
differ more than 30 minutes in time. Bias, standard deviation of the error and correlation
coefficient are defined as
bias e n 1
i 1,..,n
stde (n 1) 1
(e
i 1,..,n
correlatio n
SB
S B
B buoy (n 1) 1
where
S , B , SB
e )2
(H
i 1,..,n
buoy
i
buoy 2
H mean
)
denote the standard deviation of the satellite, the standard deviation of
the buoy and the co-variance of the satellite and the buoy respectively.
Apart from bias, standard deviation of the error and correlation coefficient, the quality of the
data can be expressed in terms of the relative root-mean-square error (RRMSE). The rootmean-square error of say, significant wave height, is
RMSE n 1
satellite
i
H ibuoy
i 1,..,n
It is similar to the standard deviation, but also includes the bias error.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 44
BMT ARGOSS
_____________________________________________________________________________________________________
The relative error is the root-mean-square error normalised by the root-mean-square value of
the buoy wave height:
RRMSE
H isatellite H ibuoy
i 1,..,n
buoy
i
i 1,..,n
The overall quality of satellite/model data can be conveniently expressed by this one
measure: the relative root-mean-square error (%) which incorporates both bias (offset) and
standard deviation (variability) of the differences between two data sources.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 45
BMT ARGOSS
_____________________________________________________________________________________________________
Appendix C- Global and regional wave models

BMT ARGOSS runs a global wave hindcast model in all major ocean basins as well as local
models in the Mediterranean, the Black Sea, the Red Sea, the Caspian Sea and the Persian
Gulf (outlined in Figure 1). Models, corresponding grid resolution, area covered and the
supplier of the driving winds are listed in the table below.
Model
Resolution
Corner NW
Corner SE
Wind
Global
0115' x 0100'
7800'N , 18000W
7800'S, 17845'E
NCEP
EU Shelf
0010' x 0010'
6600'N , 01500'W
4000'N, 03100'E
ECMWF
Mediterranean
0025' x 0025'
4530'N, 00515'E
3030'N, 03545'E
ECMWF
Black Sea
0025' x 0025'
4700'N, 02745'E
4100'N, 041 45'E
ECMWF
Caspian Sea
0025' x 0025'
4500'N, 04315'N
3645'N, 05345'E
ECMWF
Red Sea
0025' x 0025'
2800'N, 03345'E
1300'N, 04315'E
ECMWF
Persian Gulf
0025' x 0025'
3000'N, 04815'E
2415'N, 05645'E
ECMWF
Table 26 BMT ARGOSS hindcast models, grid resolutions and wind data sources.
The wave model is a 3rd generation wave prediction model, based on the WAVEWATCH III
code. A 3rd generation wave model involves representation of the spectrum on a discrete
frequency-direction grid, and explicit computation of non-linear wave-wave interaction to redistribute energy.
WAVEWATCH III (Tolman, 2002) is a third generation wave model developed at
NOAA/NCEP. It differs from earlier models in many important points such as the governing
equations, the model structure, the numerical methods and the physical parameterisations.
The model solves the spectral action density balance equation for wavenumber-direction
spectra. The implicit assumption of this equation is that properties of medium (water depth
and current) as well as the wave field itself vary on time and space scales that are much
larger than the variation scales of a single wave. Also, the modeling of physical processes
does not address conditions where the waves are strongly depth-limited. These two basic
assumptions imply that the model can generally by applied on spatial scales (grid increments)
larger than 1 to 10 km, and outside the surf zone.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 46
BMT ARGOSS
_____________________________________________________________________________________________________
Physical features
The governing equations include refraction and straining of the wave field due to temporal
and spatial variations of the mean water depth and of the mean current (tides, surges
etc.), when applicable.
Parameterizations of physical processes (source terms) include wave growth and decay
due to the actions of wind, nonlinear resonant interactions, dissipation (`white capping')
and bottom friction.
Wave propagation is considered to be linear. Relevant nonlinear effects such as resonant

interactions are, therefore, included in the source terms (physics).
The model includes

o Several alleviation methods for the Garden Sprinkler Effect
o Sub-grid representation of unresolved islands
o Dynamically updated ice coverage.
o Multiple source term packages
o A full nonlinear interaction source term option (for research)
Numerical features
The model is written in ANSI standard FORTRAN 90, fully modular and fully allocatable.
The model uses a regularly spaced longitude-latitude grid (longitude and latitude
increment do not need to be equal) and, optionally, a Cartesian grid.
Wave energy spectra are discretized using a constant directional increment (covering all
directions), and a spatially varying wavenumber grid that corresponds to an invariant
logarithmic intrinsic frequency grid.
Both a first order accurate and third order accurate numerical scheme is available to
describe wave propagation.
The source terms are integrated in time using a dynamically adjusted time stepping
algorithm, to concentrate computational efforts in conditions with rapid spectral changes.
The model can optionally be compiled to include shared memory parallelisms and a
distributed memory environment.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 47
BMT ARGOSS
_____________________________________________________________________________________________________
Appendix D- The shallow water wave ray model

The near shore wave transformation takes the following processes into account:
1
Sheltering of certain wave directions by island and capes
Refraction of waves by varying bathymetry
Wave shoaling due to changing water depth
Wave breaking due to the limited water depth or steepness of waves in deep water
Local fetch-limited growth
The first step in the wave transformation is to calculate the wave trajectories which end in the
near shore location of interest. A wave ray is the trajectory in space followed by a wave
packet with a particular frequency and initial propagation direction. It is determined by the
spatial variation of depth and current. The present model only accounts for the effect of depth
variation; the effect of current variation is ignored. A wave ray bends (wave refraction)
where there is a gradient in depth perpendicular to the wave direction, which causes a
gradient in propagation speed along a wave crest.
Starting from the near shore location, ray back-tracing is used to compute the wave rays for a
discrete set of frequencies and near shore propagation directions. The frequencies are the
same as those of the wave hindcast model providing the offshore wave conditions, and the
wave directions are on a regular grid of 5 degree spacing or finer. An Ordinary Differential
Equation solver is used to solve the coupled equations for position and propagation direction
backwards, while deriving the wave number magnitude from the local depth. The depth is
determined from a realistic digital bathymetry of the region of interest of sufficient resolution,
based on nautical Charts. The ray curvature is limited by an uncertainty principle to prevent
unrealistic wiggles. The solver adapts its step length to guarantee sufficient accuracy.
When the wave rays are computed, the near shore spectra are calculated from the offshore
spectra multiplied by weights determined from the Action Balance over the ray, and reduced
to satisfy the breaking criteria when necessary. For paths traced back to the shore, weights
are set to zero and fetch-limited growth is estimated from the local wind to account for waves
from these directions. Fetch-limited growth is computed using a growth curve and a DonelanPierson spectral shape. Finally the propagated and fetch-limited spectra are combined.
Offshore wave spectra from several different grid points of the hindcast model along an
offshore boundary chosen by the operator are used as offshore boundary conditions.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 48
BMT ARGOSS
_____________________________________________________________________________________________________
Appendix E- Frame of reference

(a) Definitions and notation
Wave height H
Crest-to-trough wave height of an individual wave (between two consecutive
up-crossings of the still water level).
Significant wave height Hs

Averaged wave height H of the 1/3 highest waves. Except on very shallow
water, Hs is accurately approximated by Hm0, defined as 4 times the
standard deviation of the vertical surface displacement (4 times the square
root of spectral moment m0, see below).
Zero-crossing period T
Time elapsed between two consecutive up-crossings of the still water level.
Mean zero-crossing period Tz

The average of the zero-upcrossing period T for a particular sea state. Tz
is approximated by Tz Tm0,2 (see Moment-based wave period below).
Spectral moment mp
p
For any integer p, mp is the integral over frequency f of f multiplied by the
wave spectrum, with f frequency in cycles per unit time. Remark: m0 is the
total variance of sea surface elevation.
Spectral density of sea surface waves S (wave spectrum)

The spectral density describes how the variance of the sea surface elevation
is distributed over frequency f. It is often referred to as wave spectrum.
Wave period based on spectral moments Tmp,q

1/(q-p)
Tmp,q= (mp/ mq)
with mp and mq spectral moments, and p and q two
distinct integers.
Wave peak frequency Fp

This is the frequency where the wave spectrum reaches its maximum.
Wave peak period Tp

This is the wave period corresponding to the wave peak frequency.
Wave peak direction Pd

This is the wave direction corresponding to the wave peak frequency.
Wave length
The horizontal distance between two consecutive up-crossings of the still
water level in the direction of wave propagation.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 49
BMT ARGOSS
_____________________________________________________________________________________________________
Wave steepness parameter s

A dimensionless parameter, defined as the ratio of significant wave height
Hs to the deep-water wave length corresponding to the wave period Tm-1,0,
2
i.e., s= (2/g) Hs/(Tm-1,0)
Principal or mean wave direction Hsd

The direction derived from the first-order directional Fourier moments (sine
and cosine-weighted moments) of the directional wave spectrum. Wave
direction is defined as coming from. It can also be defined for (a) limited
range(s) of frequencies and represented as a function of frequency.
Wind-sea and swell (online service)

Wind-sea consists of the waves having crests moving no faster than 1.2
times the wind speed, so they are growing. Longer, and therefore faster
moving, waves are called swell.
Wind-sea and swell (offline consultancy)

Wind-sea is found as a component (distinct peak) of the wave spectrum with
wave steepness s >0.03. Note that this engineering definition does not
consider the wind; only wave steepness. Wind-sea parameters are found by
applying the definitions of these parameters only to the wind-sea component
of the spectrum. Swell is defined as the component (distinct peak, or peaks)
of the wave spectrum which is not steep enough to qualify as wind-sea
Wind speed u10 and wind direction u10d

Sustained wind speed at 10m above the (sea) surface and associated
direction. Wind direction is defined as coming from. Sustained means
averaged over 1 hour.
Gravitational acceleration g
2
On Earth, taken equal to 9.81 m/s
(b) Units and conventions
Wind and wave directions are defined as coming from relative to true north positive
clockwise.
Units are expressed using the SI convention if not stated otherwise:
- length or distance (wave height, surface elevation, water depth) in
metres,
- time (wave periods) in seconds,
- speed in metres per second,
- direction in degrees clockwise from North.
_______________________________________________________________________________________________________
A870
June 2011
BMT ARGOSS
Page 50

Service Overview and Validation

Uploaded by

Document Information

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

Service Overview and Validation

Uploaded by

Copyright:

Available Formats

A part of BMT in Energy and Environment

Overview of the service and

Document status sheet

Overview of the service and validation of the database

June 28, 2011

Cees de Valk, Martin Williams

OVERVIEW OF WAVECLIMATE.COM .......................................................................................... 6

SAR DATA .................................................................................................................................... 20

CONTENT OF THE DATABASE ........................................................................................................ 6

ALTIMETER AND SCATTEROMETER DATA ............................................................................... 9

OBJECTIVES OF THIS DOCUMENT.................................................................................................. 1

ERROR STATISTICS OF RAW SAR DATA PER REGION ................................................................... 20

WAVE MODEL DATA ................................................................................................................... 24

CALIBRATION OF MODEL DATA.................................................................................................... 24

Objectives of this document

Figure 1 Global and regional hindcast models.

Figure 2 Location and average significant wave height of NOAA buoys.

The online service in relation to in-house consultancy

To optimise the web site performance, waveclimate.com

Fitness for use of the online service

Structure of this document

Content of the database

Wave hindcast model

Wind speed and direction

Wave spectral density and mean

Synthetic aperture radar

Mar 1992 - Dec 2010

Apr 1993 -Jun 2003

Buoy mean (m)

RMS error (m)

Buoy mean (m/s) RMS error (m/s)

Climate statistics provided by the online service

Table 4 Available statistics and corresponding data sources.

Altimeter and scatterometer data

Error statistics per year

Calibration of satellite data with buoys

Calibration coefficients found for satellite data

Error statistics of satellite data after calibration with buoys

The effect of the calibration of Jason-2 wind speed is illustrated below.

Error statistics of raw SAR data per region

Gulf of Mexico Newfoundland

Gulf of Mexico Newfoundland

Gulf of Mexico Newfoundland

Gulf of Mexico Newfoundland

Calibration of SAR data

max( 0.25, H ALT ) , so we

Wave model data

Calibration of model data

Calibrate each satellite with a master satellite (Topex/Jason-2) and merge

Error statistics of satellite data used for model calibration

Bias and STD of error in wave height (m)

Figure 13 Error in original altimeter wave height over the years

Error in wave height of alt minus buoy per class

Bias and STD of error in wave height (m)

Figure 14 Error in calibrated altimeter wave height over the years

Fraction of time exceeded

Figure 15 Prob. of exceedance of raw altimeter and buoy wave height

PoE of wave height for 1992-2009

Fraction of time exceeded

Figure 16 Prob. of exceedance of calibrated altimeter and buoy wave height

The effect of model calibration with satellites according to buoys

Fraction of time exceeded

PoE of wave height for 1992-2009

Fraction of time exceeded