Enhancements of storm surge hindcasting through Earth observation data

The accuracy of numerical storm surge model simulations depends critically on the quality of the initial and boundary conditions used to initialize and force the simulation. The initial conditions are the sea surface vertical displacement and the water transport zonal and meridional components at the beginning of the simulation, while the boundary conditions are determined by the physical variables producing the surge and determining its evolution, such as the wind and the atmospheric pressure at the sea surface. The boundary conditions are usually supplied to operational storm surge models by atmospheric model forecasts, while the initial conditions of the surge field are obtained by the surge model itself as the final state of a preceding simulation. The outcome of eSurge-Venice (www.esurge-venice.eu), a project funded by the European Space Agency (ESA) and completed recently, has shown that it is possible to improve the description of the initial sea state and of the forcing atmospheric wind field using remotely sensed data of sea surface wind and sea surface height. The reliability of storm surge hindcasts in the Adriatic Sea - the focus location of eSurge-Venice -, made with initial conditions and wind forcing modified by satellite data, was greatly increased. Satellite scatterometer data were used to limit the bias between global model forecasts and observations of the surface wind, while the initial states of the surge level have been corrected assimilating into the storm surge model Total Water Level Envelope (TWLE) measurements taken by satellite altimeters. This contribution describes the methodology able to bring satellite observations into storm surge modelling. Hindcast experiments on selected storm surge events happened in the northern Adriatic Sea in the last 15 years has been conducted in several configuration of wind forcing and surge level initial conditions. The RMS error on the estimation of the maximum surge peak (observed - model) reduced by 35% using only scatterometer data, by 11 % using only altimetry data, and 40% using both. In a restricted number of cases a high-resolution, limited-area model forecast wind forcing gave performance similar to that of the global model wind modified by scatterometer observations. The technique of direct assimilation of altimetry data into the storm surge model, although the promising results achieved, needs further refinements, while the strategy of mitigation of the bias between model and scatterometer winds resulted reliable and easy to set up in the operational context.