An intelligent system to real time rainfall prediction using radar data

This work presents a new approach for rainfall measurements making use of weather radar data for real time application to the radar systems operated by institute of Meteorological Research (IPMET) - UNESP - Bauru - SP-Brazil. Several real time adjustment techniques has been presented being most of them based on surface rain-gauge network. However, some of these methods do not regard the effect of the integration area, time integration and distance rainfall-radar. In this paper, artificial neural networks have been applied for generate a radar reflectivity-rain relationships which regard all effects described above. To evaluate prediction procedure, cross validation was performed using data from IPMET weather Doppler radar and rain-gauge network under the radar umbrella. The preliminary results were acceptable for rainfalls prediction. The small errors observed result from the spatial density and the time resolution of the rain-gauges networks used to calibrate the radar.