Variational data assimilation of sea surface height into a regional storm surge model: Benefits and limitations

Abstract

ABSTRACT Storm surges are coastal sea-level variations caused by meteorological conditions. It is vital that they are forecasted accurately to reduce the potential for financial damage and loss of life. In this study, we investigate how effectively the variational assimilation of sparse sea level observations from tide gauges can be used for operational forecasting in the North Sea. Novel data assimilation ideas are considered and evaluated: a new shortest-path method for generating improved distance-based correlations in the presence of coastal boundaries and an adaptive error covariance model. An assimilation setup is validated by removing selections of tide gauges from the assimilation procedure for a North Sea case study. These experiments show widespread improvements in RMSE and correlations, reaching up to 16 cm and 0.7 (respectively) at some locations. Simulated forecast experiments show RMSE improvements of up to 5 cm for the first 24 h of forecasting, which is useful operationally. Beyond 24 h, improvements quickly diminish however. Using the setup based on the shortest path algorithm shows little difference when compared to a simpler Euclidean method at most locations. Analysis of this event shows that improvements due to data assimilation are bounded and relatively short lived.