A Neural Network approach to the problem of recovering lost data in a network of marine buoys

Neural Network (NN) technology provides several reliable tools for analysis in many science and technology applications. In particular NN are often applied to the development of statistical models for intrinsically non-linear systems, since NN usually behave better than ARMA or GARCH models in complex conditions. A member of this class of problems is the analysis of time series of significant wave heights from a network of buoys. A project is being carried out by the Italian DSTN-SIMN (Technical Surveys Dept. -National Hydrological and Marine Survey) and the Dept. Of Physics of the University of Rome "La Sapienza", in order to reproduce the time series collected by the Italian SWaN network of buoys (Sea Wave monitoring Network). Aim of the project is the determination of the best way to fill gaps and long periods of missing data with the best accuracy by means of a reanalysis of the whole ten years' data set of the SWaN Here a NN model is proposed for time-space analyses of the marine data. Main feature of the tool is the ability to reproduce long time series of data without any increase of the error. The method is based on a preliminary spatial analysis of the wave climates in order to classify the degree of overlapping of information from different stations. This overlapping, where possible, led to an optimal and selective training of the AW by means of data collected in different, nearby, locations. NN numerical simulations of some important historical storm are compared with the data originally observed at the stations of Crotone (Ionian Sea), Pescara (Adriatic Sea) and Monopoli (Adriatic Sea).