Neural network models to forecast hydrological risk

River flow forecasts are required to provide basic information for reservoir management in a multipurpose water system optimization framework. Moreover, an accurate short term prediction of flow rates is crucial for practical flood forecasting. In the paper, a neural approach is used to model the rainfall-runoff process in two different river sections in the same basin. Numerical results are provided for runoff prediction in the Tirso basin at the S. Chiara and Cantoniera sections in Sardinia (Italy), by considering hour and daily time steps.