Abstract:

Monthly precipitation forecast for the upper Yangtze River is very essential to the water resources management for the entire Yangtze River basin.Three typical meteorological stations were selected respectively in three different climatic zones.All the selected stations contained nearly 60 years of monthly precipitation records in the upper Yangtze River.This paper estimated the month of precipitation and precipitation time delay parameter,and established monthly precipitation forecasting model using backpropagation neural network,radial basis function neural network,generalized regression neural network and multiple linear regression method respectively,to predict the precipitation of coming month.Then,the mean square error and coefficient of determination were used to verify the simulation accuracy of various models and the model simulation results.The results show that artificial neural network prediction model is superior to multiple linear regression in general.Especially,the performance of the backpropagation neural network is better than the others.It can be determined as an effective monthly precipitation methods for the upper Yangtze River after determining reasonable input variables and network structure.〖