17 
5 
Validation 
The operational circulation model is used by various BSH services. It is an important tool 
in the Agency’s water level prediction service and, therefore, its accuracy is checked daily 
by comparing measured and predicted water levels. Furthermore, measured and 
modelled time series of water levels - as well as of water temperatures - are used to 
evaluate the model results statistically. Comparisons of other model parameters with 
measurements are carried out as well, but they often cover shorter periods. 
In the following, examples for model validation will be presented for the model 
parameters water levels, currents, water temperature, salinity, and ice cover. All 
examples refer to the period after 01.01.1999. 
It should be noted again that all model results are forecasts. This means that errors in the 
parameters do not necessarily imply errors of the model, they often reflect uncertainties 
in the meteorological forecast data. Hindcast calculations would in many cases reduce 
the errors. 
5.1 Water levels 
The high and low water level predictions for the German Baltic and North Sea stations 
are evaluated yearly on a routine basis. Figure 7 shows the frequency distribution of 
measured and modelled high water (HW) differences at the stations Borkum and 
Cuxhaven for the year 1999. While at Borkum the bias of HW level predictions is approx. 
0 cm, the computed water levels at Cuxhaven are slightly too high (bias = 6.6 cm). The 
standard deviation at these stations is 15 cm and 18 cm, respectively. Nearly the same 
values are obtained for the standard deviation of low water (LW) differences. Higher bias 
values are found for LW (between 10 and 20 cm). The same feature is found at other 
stations in the inner Wadden Sea areas. While the stations on the outer islands in the 
German Bight show similar errors as Borkum, the water levels predicted for stations in 
the inner parts of the Wadden Sea often show higher deviations at LW. The reason is 
that even the high model resolution of 1 nautical mile in the German Bight does not 
adequately describe the complex topography in the Wadden Sea and German estuaries, 
which has a stronger effect on LW levels than on HW levels. 
Table 2 shows bias and standard deviation of high and low water forecasts for the North 
Sea stations Borkum and Cuxhaven in the years 1998 to 2000. After implementation of 
the new model version the bias has been reduced since 1999, especially for low waters. 
The reason is a better representation of the tidal signal by a better resolution of 
topography.