Fine Huainan Summer Precipitation Forecasting Methods Based On Statistical Downscaling

Based on the summer daily precipitation data of60weather stations in Anhui region from1999to2009and observation data of Anqing sounding station,14stations are select as a representative of the large-scale precipitation fileds, other46stations on behalf of the local-scale precipitation. Using the mid-low level wind of Anqing sounding station summer precipitation data is further divided, in nine different wind, local single-station precipitation, the EOF principal component and the REOF principal component of local precipitation as prediction object, associated with large-scale precipitation by0.05significant test of principal components as predictors, respectively using artificial neural network and multiple linear regression method to build downscaling forecasting model, called as BP1model, BP2model, BP3model and LR1model, LR2model, LR3model. The linear and nonlinear relationship between large scale circulation and local scale precipitation field in different mid-low level wind direction are analyzed, the analysis effect of the scale model test prediction is compared with the commonly used spatial interpolation methods. Throughout the forecast during the test, we come to the following conclusions:(1) The analysis of mean daily precipitation and distribution in five regions indicating that mean precipitation in the southwestly, westerly and northwesterly is more than other wind, the regional differences distribution of daily precipitation is obvious, the majority precipitation is concentrated in south of Anhui Province and Dabie Mountain area with the mountains and hills as key terrain.(2) Analysis of the neural network model and the linear regression model with same prediction object fitting and forecasting daily rainfall, the results show that the nonlinear model is better than the linear model in terms of fitting and forecasting.(3) BP1model and the BP3model among three artificial neural network models, according to the analysis of the forecasting for amount precipitation, error evaluation, represented stations forecasting and heavy rain process, the results are better than the corresponding linear model and the interpolating prediction method. As the number of selected predictors is fewer and cumulative variance contribution rate is small, the BP2model with the EOF principal component of the local precipitation as prediction object has poor forecasting result. (4) To further discuss the application scope of BP1model and BP3model, the both models simulation of heavy rainfall in June29-30can reflect the basic precipitation trends and local characteristics, BP1model in the regional with rivers and plains as the main terrain forecasts better, BP3model forecasting results in the area with main terrain of mountains and hills is relatively good, the BP3model forecasting for heavy rain is more accurate. In actual forecasting applications, the applicable downscaling prediction model can be selected according to the rainfall spatial and temporal distribution characteristics.