U-Net-based Precipitation Predict by Cloud Map

Abstract

Many traditional precipitation prediction methods in meteorology nowadays require many types of data to be input as parameters. This research is to investigate ways to use deep learning techniques for precipitation prediction using only input cloud maps. This paper establishes a technical route for predicting rainfall through cloud map data using U-Net, and experiments. Rainfall models were successfully trained using U-Net and predicted.