Numerical Sea level forecasts

Error sources

Errors in water level forecasts for the North Sea coasts can be attributed to causes described below.

Errors in the meteorological forecasts

Errors in the meteorological input for the storm surge model are usually the main error source for water level forecasts in storm situations. These errors can be errors in the position and intensity of depressions. But as the resolution of meteorological models decreases, also small-scale intense phenomena come into play. Predictability of such phenomena is intrinsically limited.
Near the coast, the land influence in the wind over sea should be carefully dealt with.

A solution to deal with the uncertainty in meteorological input is the use of ensemble forecasts. They give information on the possible ranges for the meteorological forecasts, and, applied to the storm surge model, also for the water levels. For small scale phenomena they can be useful for assessing different possible scenarios.

Uncertainty in the air-sea interaction

Apart from the meteorological forecasts themselves, there is also an uncertainty in the air-sea interaction. Usually a charnock-type drag relation is used, but there is evidence that this is not correct anymore for high wind speeds.

Parametrisations which follow from a newly developed wind-over-waves theory are currently being developed and made available for storm surge forecasting.

Errors in the astronomical tide

One ingredient for storm surge forecasts is the astronomical tide. With increasing accuracy, it becomes clear that the prediction is not prefect. Especially in quiet summer periods, periodic spikes of up to 20 cm in the difference between observation and astronomical prediction can be identified.

The effect on forecasts of high surges is limited, but it does enter into the assimilation process of observed water levels.

Limited accuracy of the storm surge model

The accuracy of storm surge forecasts is also limited by the accuracy of the storm surge model itself. Although the model has undergone an extensive tuning process, there are always the effects of limited resolution, especially near the coast.
In the North Sea stratification is not important, especially not in winter. Therefore a depth-integrated model is generally sufficient for water level forecasts.

Erroneous observations for data assimilation

If observed data are assimilated into the storm surge model, good quality control is essential. At present an automatic quality control algorithm is used which properly rejects all of the bad observations.