The study of rainfall forecast based on neural network and GPS precipitable water vapor

Water vapor and its changes directly affected the weather. It is one of key factors about severe weather formation and evolution. Accurate, and timely rainfall forecast is also important factors which increased forecast accuracy of storms, floods and other disastrous weather. In this paper, it build models of the data for training and simulation based on neural network technology, and analyzed the results of rainfall forecast by using GPS precipitable water vapor and other meteorological parameters. Through data preprocessing, BP neural network modeling and analysis it has been completed the design of rainfall forecast. With the comparison between the two-hour time prediction value of Qinhuangdao and the measured value, it has been achieved the verification of rainfall forecasting. The accuracy rate of two-hour rainfall forecast is about 92.5 percents.