Verification of water level forecasts

Winters 2002 – 2005 and 2006 – 2009

Verification by the Storm Surge Warning Service (SVSD) shows that the quality of storm surge forecasts is difficult to maintain. Especially, the winters of 2006 – 2009 seem to show behave poorer than 2002 – 2005.

This web site presents the verification of the operational DCSM model forecasts, both for all surges and for surges more than 80 cm.

Interpretation

This verification supports the conclusions of the SVSD, although the data (surges versus water levels) are different.

Considerations

In order to achieve more firm conclusions several aspects will have to be addressed:

1. What tells the verification of the meteorological input, especially the 10-m wind?
2. The verification is for extremes. A ‘good’ model will always underestimate extremes, because there are many possibilities to produce a less extreme forecast than a more extreme forecast, so an error will on average tend to less extreme forecasts.
   The latter period of this verification seems, on first sight, to contain more extremes than the earlier period.
3. The SVSD verification is for extreme observed surges. How does this compare with the verification of forecasts for extreme surges? (The difference being that in the first case the extreme surge might not be forecast, and in the second case it might not have happened.)
4. Are there enough cases for the conclusions to be significant?