Development of a new technique for quantifying the uncertainty in coastal sea level forecasts due to errors in meteorology

If we want to forecast sea level at a point on the coast, such as Sheerness near the Thames Barrier, then it is important to know what the sources of forecast error are. For example, small errors in the forecast wind may lead to errors in the forecast sea level and these effects may accumulate in the hours leading up to a flooding event. However, it is not straightforward to guess what pattern of wind error may cause a particular change in coastal sea level several hours later. This is similar to knowing how to blow over a cup of tea in order to make the fluid level peak at a single point at a particular time. Here, we developed a new technique called adjoint modelling which is able to answer this question. By mimicking the evolution of sea level around the UK using a simplified computer model, the adjoint may be used to test the sensitivity of sea level to the wind forcing. The adjoint model relies on mathematical relationships which govern the simplified system and its evolution in time. Using this approach we identified a region of approximately 300 km near the coast and within 3 hours preceeding an extreme sea level event at Sheerness for which errors in the wind stress matter most for the sea level forecast. It is the forecast sea level at Sheerness that is most important in determining whether the Thames Barrier should be closed. Therefore, our study suggests that reduction of meteorological forecast error in this space-time window may be most important for improving the accuracy of extreme sea level forecasts at Sheerness. See paper: "Tide-surge adjoint modeling: A new technique to understand forecast uncertainty". J. Geophysical Research. 2013.