4. Did the forecasts by KNMI-MMD (maritime meteorlogical service) of wave height for the North Sea improve?
It has been determined above, that the most important quality gain was obtained at short term forecasts (fp = 6 and 12 hours) of Hs10, low-frequency wave height, more than for Hm0, significant wave height. For Hm0 the effect is less clear, except for the 'analysis' (fp = 0) and for the very short term (fp = 6). So far, we have looked at the behaviour of wave model NEDWAM only. Now the effect on the forecasts issued by the Maritime Meteorological Service of KNMI (MMD) will also be considered. We do this by looking at figures 3c and 4c, showing monthly RMS errors and bias respectively of the forecasts by the MMD of Hm0 for fp = 0, 12 and 36 hours at Europlat.
figure 3c: forecasts of significant wave height at Europlat (issued by MMD), compared with observations: monthly RMS error
figure 4c: forecasts of significant wave height at Europlat (issued by MMD), compared with observations: monthly bias
This data has been obtained from the UVS database, maintained by Martin Stam. (The data can be inspected through the KNMI Intranet internally; both for MMD in general and the individual forecasters). The bias of fp = 0 has clearly diminished, see figure 4c. There is something remarkable about forecasted values, fp = 12 and 36 hours, actually parallel to the results of NEDWAM (figure 4b): the MMD forecasts have become lower than the analysis (fp = 0) since June 2000 , while forecasted wave heights were greater than before this time!
The RMS error of forecasted Hm0 does not show such clear
changes. The analysis looks somewhat better, but the forecasts did not improve, see figure 3c.
Comparing this with the results of wave model NEDWAM
(figure 3b), we see that the curves have become
almost similar since June 2000. This suggests that the forecaster is intervening less in the
model output than before.
The UVS database does not contain similar data of Hs10 as forecasted by the MMD.
However, due to the nature of this parameter (determined from the low-frequency part of the
model spectrum) the forecaster is depending more on model output than in case of
significant wave height Hm0.
We may conclude, that the MMD benefits from the introduction of
data assimilation of wave spectra , particularly in the occurrence of northerly
swell, provided that real-time directional spectra of observations in the northern part of
the North Sea are available. The forecasts of low-frequency wave height in the southern
North Sea have particularly improved.