Downscaling Guangxi Monthly Precipitation Forecast

In this paper, the spatial and temporal characteristic of monthly precipitation in Guangxi, and its changing regulations, and its physical cause had been studied. Based on the NCEP/NCAR reanalysis data of monthly mean 200hPa,500hPa,700hPa geopotential height fields, and monthly mean 200hPa,700hPa horizontal wind velocity fields, and the ensemble hindcast products from National Climate Center of China, considering the weather dominating systems and climate characteristics of Guangxi regions, the monthly precipitation downscaling forecast models are established respectively.The results of this research show that (1) monthly precipitation in northeast of Guangxi occurs mainly from March to July, causing by cold air and westerly trough, while monthly precipitation in southwest of Guangxi and the area along Beibu Gulf occurs mainly from June to August, causing by tropical cyclones or the intertropical convergence zone, etc. The analysis also shows that the flood peak of rivers in Guangxi closely correlate to the total precipitation in JJA. (2)There is significant correlation relationship between monthly rainfall in Guangxi from January, to December and geopotential height anomaly not only over south China but also over Europe-Asia high-middle latitude areas. It is shown that both lower geopotential height over south China and large amplitude of trough and ridge in Europe-Asia high-middle latitude areas are beneficial to rainfall in Guangxi. Monthly precipitation in Guangxi is significantly correlated to monthly mean U, V horizontal wind velocity over 700 hPa and 200hPa with small correlation areas and high correlation coefficients. To some extent monthly rainfall in Guangxi correlate closely to the U, V wind velocity over high-middle latitude areas, which are related to the strength of cold air. (3) In this work, 6 ways to compose predictors and 3 prediction methods are used for establishing downscaling forecast models for explanation and application the products of dynamical climate model to predict monthly rainfall in Guangxi. The 6 ways to compose predictors are significant correlation areas from 200hPa, 500hPa, 700hPa monthly mean geopotential height fields, making empirical orthogonal function (EOF) for significant correlation areas from 200hPa, 700hPa monthly mean wind fields, Chebyshev Polynomial coefficients and EOF time coefficients for significant correlation areas in 500hPa, vorticity and divergence fields constructed by 500hPa geopotential height according to the equations of geostrophic-vorticity and geostrophic-wind, etc. The 3 prediction methods are stepwise regression, Euclidean distance similarity, and BP neural network. Using the ensemble forecast or hindcast products of dynamical climate models from National Climate Center of China, we can predict monthly precipitation in Guangxi by these downscaling models. As a comparison, prediction models for monthly precipitation in Guangxi are established with the predictors from former sea surface temperature (SST), atmospheric circulation, and circulation characteristic data by using above statistic methods. The results of the prediction tests during 2003 to 2006 shown that the downscaling models are superior to the traditional staticstical models in predicting monthly precipitation of Guangxi.