中文摘要：

选取1个暴雨实例，利用统计方法对同期风云2号卫星红外云图资料进行温、湿度反演，通过像素点灰度筛选、灰度反演值订正方法对反演场实施质量控制，并将温、湿度反演场与温、湿度常规分析场实行变分同化。最后为研究反演资料在降水预报中的作用，设计3种引人反演资料的方案与MM5常规方案进行比较，并用TS评分客观评价了上述方案的预报效果。

英文摘要：

In one rainstorm process we firstly statistically retrieved temperature and humidity of the atmosphere from the FENGYUN-2 Infrared cloud image data, then carded out quality control by means of pixel filter and retrieval correct, and implemented the variational assimilation between retrieval fields and analysis fields by means of SOR method. At last, in order to find out the action of the retrieval data for heavy rainfall forecast, we designed four kinds of schemes to perform the comparative experiments in MM5 model, and introduced Threat score to evaluate the rainfall forecast results of the different schemes.