中文摘要：

利用我国全球中期数值预报业务模式T213L31及其三维变分同化系统SSI，对全球NO—AAl6、17的ATOVS资料进行了同化试验。对2007年7月16-18日发生在川渝地区的一次暴雨过程进行数值模拟研究，对比分析了仅同化常规资料与加入卫星资料后的同化结果及模拟结果。试验表明：T213L31模式加入同化AMSU资料后，可以改善降水预报，尤其是降水强度。通过连续滚动同化卫星资料，模式对于大尺度环境场改进明显，模拟效果更加接近实况。AMSU资料的使用可以改进温度场、湿度场以及风场，对于降水落区的预报有很好的指示意义。同化AMSU资料虽然一定程度上增加了降水区的湿度，但是该区域为一个稳定维持的低湿中心占据，是造成降水落区出现偏差的原因。

英文摘要：

The global mesoscale numerical forecast system T213L31 and its three-dimensional variational data assimilation system SSI are used in this paper to assimilate ATOVS data from NOAA-16 and 17 satellites. A heavy rainfall event in Eastern Sichuan Province and Chongqing during 16--18 July 2007 is simulated to compare and analyze the assimilation and simulation results based on conventional data only with those based on ATOVS data also. The results show that with AMSU data, the T213L31 Model is improved in precipitation forecast accuracy, especially the precipitation intensity. By continuous assimi- lation, the weather situation fields are obviously improved and are closer to the reality. It is found that AMSU data can improve the temperature, humidity and wind fields, which are significant for the forecast of precipitation location. In the precipitation area, although the humidity is increased by assimilating the AMSU data, the stay of the low centre of humidity is the reason for the deviation of the precipitation location.