Abstract - This paper presents an empirical model to forecast floods in urban areas. The methodology is based on indexes controlled by meteorological radar data. The model has been used in one of the biggest cities of the world, the city of São Paulo, Brazil. The São Paulo flood forecasting system (SAISP) is operated using a S-Band radar and a telemetric network. The model was developed to attend many public agencies detaching the civil defense and the traffic control. Due to the lack of data an empirical approach was adopted based on the present, accumulated and forecasted rainfall. The S-Band radar produces CAPPI´s (Constant Altitude Plan Position Indicator) every five minutes from the which a nowcasting model can produce forecasting for the next three hours. The relationship between these three variables defines the hydrological situation for an specific area. This information triggers a set of field activities. The model was applied to many small urban catchments. The paper presents results obtained in the Cabuçu de Baixo river in the last rainy season 2002/2003. The results shows an excellent performance. In general the model can forecast floods between one and three hours ahead even with the occurrence of convective storms.