IMPACT OF METEOROLOGICAL ELEMENTS AND SPATIAL RESOLUTION IN DEEP LEARNING RAINFALL PREDICTION

Abstract

 In recent years, there has been a need to improve the accuracy of precipitation prediction using deep learning due to the increasing number of heavy rainfall events. However, due to the lack of spatial resolution, meteoro-logical factors other than precipitation were not learned. Therefore, we constructed a virtual observation point and investigated whether meteorological factors other than precipitation contribute to the learning process when the spatial resolution is increased. As a result, we found that meteorological elements other than precipitation are noisy and worsen the prediction accuracy in case of low rainfall, but improve the prediction accuracy in case of high rainfall. In addition, this tendency became stronger as the spatial resolution was increased.