14.12.2012 Views

GUIDE WAVE ANALYSIS AND FORECASTING - WMO

GUIDE WAVE ANALYSIS AND FORECASTING - WMO

GUIDE WAVE ANALYSIS AND FORECASTING - WMO

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

give large differences in the comparisons. Further, it<br />

must be remembered that in situ measurements are only<br />

estimates and individual spectral components may have<br />

quite large uncertainty due to high sampling variability<br />

in the estimates.<br />

For these reasons, model validation studies are<br />

usually performed in a statistical sense using all available<br />

data, and are based on comparisons of derived<br />

parameters. A common approach is to perform a regression<br />

analysis on a set of important parameters such as<br />

H s, T – z (or T p) and, if directional wave data are available,<br />

the mean direction (for different frequency bands). In<br />

addition, it may be useful to compare mean wave spectra<br />

or cumulative distributions (of, for example, H s) for<br />

simultaneous model results and measurements. Since the<br />

wind field that drives the model is intimately related to<br />

the model wave field, most evaluation studies also<br />

include verification of wind speed and direction.<br />

An important requirement for evaluation of a wave<br />

model is the availability of reliable sea-state measurements<br />

and related weather data. Most of the evaluation<br />

studies reported in the last 15 years have used buoy data<br />

for wind and wave measurements and have occasionally<br />

used analysed weather maps for additional wind<br />

information. A few earlier studies (e.g. Feldhausen et al.,<br />

1973) have used visual ship reports for wave height to<br />

make a qualitative evaluation of several wave hindcasting<br />

procedures. More recently, satellite altimeter data<br />

have been used to validate wave model results. For<br />

example, Romeiser (1993) validated the global WAM<br />

model (WAMDI Group, 1988; Komen et al., 1994) for a<br />

one-year period using global data from the GEOSAT<br />

altimeter.<br />

Most verification studies have attempted to calculate<br />

several statistical parameters and analysed the<br />

magnitude and variation of these parameters to determine<br />

the skill of a wave model. Among the parameters<br />

that are most commonly used are:<br />

• The mean error (ME) or bias;<br />

• The root-mean-square error (RMSE);<br />

• The Scatter Index (SI) defined as the ratio of RMSE<br />

to the mean observed value of the parameter; and<br />

• r, the sample linear correlation coefficient between<br />

the model and the observed value.<br />

A few studies have considered other statistical measures<br />

to evaluate wave model performance such as the slope of<br />

the regression line between model and observed values<br />

or the intercept of the regression line on the y axis. The<br />

four parameters as listed above are the best indicators of<br />

the performance of a wave model.<br />

Verification of operational wave models at<br />

national Services of many countries is an ongoing<br />

activity. At most Services, the initial verification of a<br />

wave model is performed during the implementation<br />

phase and the verification statistics are updated periodically<br />

so as to monitor the model’s performance.<br />

Table 6.1 shows the performance of a few wave models<br />

which are presently in operational use at various<br />

OPERATIONAL <strong>WAVE</strong> MODELS 73<br />

national (international) organizations in different parts<br />

of the world. The table presents verification statistics<br />

for wind speed, wave height and peak period (where<br />

available) in terms of four statistics, namely: ME,<br />

RMSE, SI and r as defined above. The table includes a<br />

variety of models developed over the last twenty years,<br />

ranging from first generation (1G) to the recent third<br />

generation (3G) models (see Section 5.5 for a discussion<br />

of model classifications).<br />

Verification statistics for many more models have<br />

been summarized elsewhere by Khandekar (1989).<br />

Based on these statistics, it may be concluded that a firstgeneration<br />

wave model driven by winds obtained from a<br />

weather prediction model can provide wave height<br />

simulations with an RMS error of about 1.0 m and a<br />

scatter index of around 35 per cent. The second- and<br />

third-generation wave models, driven by winds from<br />

operational weather prediction models, can provide wave<br />

height simulations with RMS errors of about 0.5 m and<br />

scatter indices of about 25 per cent. The third-generation<br />

WAM model (see Section 5.5.4), which has run<br />

operationally since early 1992 at the European Centre<br />

for Medium-Range Weather Forecasts (ECMWF) in the<br />

United Kingdom, generates wave height values with<br />

RMS errors ranging from 0.4 m to 0.7 m in different<br />

parts of world oceans, while the bias (or mean error)<br />

ranges from 0.05 m to 0.2 m. Since these statistics were<br />

generated further improvements have been made both to<br />

the determination of wind fields and the wave models<br />

themselves, so it is expected that the results of the verifications<br />

will also be continually improving.<br />

A sample wave height plot generated by the US<br />

Navy’s GSOWM at a buoy location in the Gulf of<br />

Alaska is shown in Figure 6.6. The figure shows wave<br />

height variation in the Gulf of Alaska as measured by the<br />

buoy and as modelled by the GSOWM. The GSOWM<br />

was driven by winds from the GSCLI (Global Surface<br />

Contact Layer Interface) boundary-layer model. The<br />

figure also shows wave height variation simulated by the<br />

WAM model, which was configured to run on a 1° x 1°<br />

grid and was driven by surface wind stress fields generated<br />

by the Navy Operational Global Atmospheric Prediction<br />

System (NOGAPS). Also shown along with the<br />

wave plots are wind and wave error statistics covering<br />

the one-month period from 20 February to 20 March<br />

1992. The GSOWM has a high bias of about 1 m in<br />

its wave height simulation on the US west coast. This<br />

bias is due to the model’s tendency to retain excess<br />

low-frequency energy from swell waves which often<br />

travel from the central Pacific to the US west coast.<br />

Figure 6.7 shows the verification of the global wave<br />

height field generated by the WAM model at ECMWF<br />

against wave heights measured by the radar altimeter<br />

aboard the satellite ERS-1 (see Section 8.5.2). The altimeter<br />

wave heights plotted along the sub-satellite tracks<br />

cover a six-hour window centred around the time for<br />

which the model wave height field is valid. In general, the<br />

model wave heights show an excellent agreement with

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!