Artificial neural network hydrologic model for urban watersheds

An artificial neural network (ANN) hydrologic model was developed to forecast floods on a heavily urbanized watershed within the Metropolitan Area of São Paulo (MASP), Brazil. The Tamanduateí watershed is fustigated with frequent flash floods during the summer season. Recent results have indicated that convective events tend to produce higher rainfall accumulation over the MASP region due to the local heat island and sea breeze circulation. Thus, this simple yet conceptually sound model can be used to anticipate flood events. The ANN was trained with streamflow and weather radar derived rainfall accumulation data from several cases. The model was verified with a larger number of independent flash flood events.