The 3DVAR Application Experiments on Rainfall Prediction in the Summer in China
三维变分系统在我国夏季降雨预测中的应用试验

The MM5 modeling system and its three dimensional variation assimilation system (MM5/3DVAR) are employed to prediction rainfall during August 2005 in China,and then,the prediction results are analyzed.The background error used by 3DVAR system was reproduced by using "National Meteorological Center(NMC)" method and the prediction results of August 2005,which adapt to the prediction domains and horizontal,vertical resolution in the following experiments.The characteristics of reproduced background error are very similar to the global background error,and the length-scale of background error decreases with horizontal resolution.Three experiments(NCEP,LITT,3DVAR Experiments) are designed to analyze the impacts of 3DVAR system.NCEP Exp.only uses NCEP data to form initial field.The initial field of LITT and 3DVAR Experiments utilize CRESSMAN and 3DVAR method to analyze surface and sounding data on the base of NCEP data.Compare to the results of NCEP Exp.,the LITT Exp.improves the 1 mm and 10 mm rainfall prediction,but 25 mm and 50 mm rainfall prediction is not improved.3DVAR Exp.improves 10 mm,25 mm and 50 mm rainfall prediction significantly.The RMSE(Root Mean Square Error) of synoptic field prediction,except for temperature and relative humidity,in three experiments increase with height generally.All of the prediction of 3DVAR Exp.is the best among the three experiments.The significant improvement of 3DVAR Exp.maybe caused by that the analysis of mesoscale information by 3DVAR is more similar to observational data.