Improving storm surge forecast in Venice with a unidimensional Kalman filter

Abstract

This article presents the application of a post-processing method to correct forecasts produced by a hydrodynamical storm surge model. The model runs over the Mediterranean Sea and into the Venice lagoon. It provides storm surge forecasts in the Mediterranean Sea and total sea-level forecasts inside the Venice lagoon. Sea level observations outside and inside the lagoon are used to correct these forecasts in some locations, by means of a unidimensional Kalman filter. This method, not very known in ocean sciences, is explained here in details. Its advantages consist of easy implementation, low computational times and in the need of few input data. In the present application, the forecast improves up to about two days ahead, even if the best results are obtained in the first 6 hours. The method does not require any long database to be calibrated and can be used also to provide hourly updates to the initial forecast. In the case of extreme events, the possibility of hourly updates is essential to keep a small error in the short-term forecast. These qualities make this method very useful for improving the prediction of a deterministic model which, normally, is performed only once or twice a day. Finally, its generality makes it adaptable for use with other variables and in association with non-local methods of data assimilation.