Predição da Variação Extrema do Nível do Mar Relacionada a Tempestades Severas Utilizando Redes Neurais Artificiais

This paper presents an Artificial Neural Network (ANN) model developed to predict extreme coastal sea level variation (storm surges) on Southeast Region of Brazil, related to the passage of frontal systems associated with extratropical cyclones that cause severe thunderstorms. Tidal forcing is the main cause of sea level daily variation but the effects of meteorological phenomenon are also present in rising and lowing of the observed sea level and tend to be more drastic accordingly to the event. Hourly time series of water level were used from two tide gauge station. 6-hourly series of atmospheric pressure and wind components from NCEP/NCAR reanalysis data set were also used on some grid points over the oceanic area. Correlations were verified to define the time lag between the meteorological variables and the coastal sea level response to the occurrences of the extreme atmospheric systems. These correlations and time lags were used as input variables of the ANN model. Simulations until 48 hours were tested with the neural model. This model was compared with multivariate linear regression and presented the best performance, generalizing the effect of the atmospheric interactions on extreme sea level variations.