中文摘要：

基于中尺度暴雨预报模式AREM和三维变分同化系统GRAPES-3dvar对2003年7月8日发生在长江中游的一场特大暴雨过程，进行了武汉、宜昌多普勒雷达风场资料的三维变分同化研究，结果表明：（1）增加了武汉、宜昌多普勒雷达反演风场资料后，模式对湘-鄂交界处特大暴雨区的模拟效果改善非常显著。（2）分别单独增加武汉、宜昌多普勒雷达风场资料的三维变分同化后，降水模拟都有明显改进，但要差于对两部雷达风场资料同时进行变分同化的结果，表明同化的中尺度初始风场信息越多，初始场的质量越高，降水模拟效果越好。（3）雷达风场资料的三维变分同化，改善了分析场中尺度结构的描述，从而减轻了spin-up现象，使得模式在积分初期就能模拟出与实况相近的强降水。

英文摘要：

catastrophic torrential rain occurred in the middle reaches of Yangtz River during the period of 08：00 BST 8 July 2003 to 08：00 BST 9 July 2003. Retrieval wind data from Wuhan and Yichang Doppler radar are assimilated into Advanced Regional η-coordinated Model （AREM） with Three Dimensional Variational Assimilation of Global and Regional Analysis Prediction System （GRAPeS-3DVAR）. Results show as follows. （ 1 ） When retrieval wind data from Wuhan and Yichang Doppler radar are assimilated into AREM model with 3DVAR, remarkable improvements are made in the simulation of torrential rain band in the connecting area of Hunan and Hubei Province. （2） When retrieval wind data from Wuhan or Yichang Doppler radar are assimilated into the model separately, each simulated result of rainfall is better than control experiment. But it is worse than that of assimilating two radar wind data into the model at the same time. It shows that the more initial meso-scale wind field is assimilated, the higher the quality of the initial field is and the better the precipitation simulation result is. （3）The 3DVAR assimilation of radar wind field has improved the meso-scale structure description of analysis field, which contains richer and more reasonable meso-scale information. The spin-up phenomenon is therefore alleviated and the model can simulate the heavy rain close to observations at the initial integration stage.