Precipitation Prediction Modeling using Neural Network and Empirical Orthogonal Function Base on Numerical Weather Forecast Production

摘要

Base on numerical weather forecast (NWF) products, a new prediction method using Artificial Neural Network (ANN) and Genetic Algorithm (GA) is proposed for model establishment by means of making a low-dimension ANN learning matrix through empirical orthogonal function (EOF). The example of application is based on the T213 numerical weather forecast (NWF) products from China Meteorological Administration (CMA) and products from the Japanese fine-mesh NWF model, and three ANN prediction models for daily precipitation are established for three different areas in Guangxi province. It is shown from the contrast analysis that TS scores of the three ANN models for moderate or even heavier rain are 0.57, 0.45, and 0.3 respectively, which are obviously higher than those of the T213 and fine-mesh NWF models.