Aplicação de Redes Neurais Artificiais e Transformada de Ondeletas para Estimativas de Precipitação usando imagens do satélite GOES-8: Resultados Preliminares

Abstract

Artificial Neural Networks and Multiresolution Wavelet transform technique is applied to construct the nonlinear relationship between meteorological satellite and radar images. With the trained networks, the developed system gives the estimated next time precipitation image. The paper shows the method of network structure construction, input/output image pattern preparing and the sequence of the training. The statistic analysis results show the premise precipitation estimation for the São Paulo region. With the further modification, the system can be also used to estimate rainfall rate near radar cover region through the on-line update training process.