Improved method of sea level forecasting at Venice (Northern Adriatic Sea)

In this paper a forecasting method for the extremely dangerous aqua alta phenomenon is developed. Thecity of Venice, which is located in the northeastern part of Italy, is often subjected to intense flooding, dueto increasing of the sea level. The classical methods used in Geophysics generally failed in forecasting thisphenomenon, because of the complexity of the physical modelling. Therefore, a method based on a nonlinearauto-regressive moving average (ARMA) approach is proposed here. It is found that the time series,corresponding to the sea level data, present a Gaussian distribution, but during short periods, of order offew days, a strong non-Gaussian behaviour is evident, which is concomitant with aqua alta events. Thissuggests the development of a non-linear version of the usual ARMA models. An empirical orthogonalfunction technique is applied in order to avoid numerical pathologies of the model due to the particularcharacteristics of the time series. The encouraging results indicate that such technique is a good tool forforecasting and alarm purposes.