Sulfur dioxide ground level concentrations forecasting by means of neural networks

We present the preliminary results on the use of neural networks to forecast SO/sub 2/ concentration levels in the industrial area of Ravenna. Ground level concentrations of pollutants were analyzed in the area, in particular the high levels of SO/sub 2/ occurring during relatively rare episodes. These events are typically correlated with many different aspects, such as: complex local meteorology, topography, and industrial emissions parameters. Here we propose a neural network model trained with a Bayesian learning scheme to overcome the failure of deterministic models (e.g. Gaussian models) in explaining the high ground level concentrations.