由极端降水引发的灾害占中国自然灾害的70%，准确的模拟和预估极端降水事件对于防灾减灾至关重要。该工作从观测资料和模式模拟两方面着手，首先对比极端降水在观测资料和再分析资料中的差别；在此基础上，科学评估上一代气候模式对中国极端降水的模拟能力；最后评估最新高分辨率全球气候耦合模式FGOALS-f对极端降水的模拟能力，并分析模式空间分辨率对模拟结果的可能影响。 结果表明：与站点观测和热带卫星观测（TRMM）相比，两种格点资料中CN05.1比APHRO能够更准确的描述中国极端降水；三大再分析数据均无法体现200mm/day以上的极端降水，其中ERA-Interim最接近观测，但极端降水总量和频次均偏低70%-90%。上一代全球气候耦合模式（CMIP5）中，CNRM-CM5对极端降水模拟能力最好，但依然无法模拟出200mm/day以上降水，频次和总量均偏低30%-80%；最新的FGOALS-f耦合模式对中国极端降水的模拟能力显著提高，表现为：降水总量和降水频次上均更接近实际观测，总量和频次偏低10%-60%，并能够较为准确的模拟出200mm/day以上的极端降水。在模式的动力框架、参数化方案等相同的情况下，分辨率的提高能够一定程度上提高极端降水的模拟能力。

Disasters caused by extreme precipitation account for 70% of the natural disasters in China, so accurate simulation and prediction of extreme precipitation events are very important for disaster prevention and reduction. This work first compared the difference of extreme precipitation among observation and re-analysis, and then evaluated the simulation of extreme precipitation in CMIP5 models, finally investigated if the extreme rainfall simulation is improved in the new-released high resolution coupled climate model (FGOALS-f) and if the spatial resolution affects its performance. Compared with station data and satellite data (TRMM), CN05.1 gridded observation exhibited better performance than APHRO, but both underestimated the extreme rainfall. Three re-analysis data all underestimated the extreme rainfall in both frequency and amount, and they can’t simulate the rainfall with intense above 200mm/day. Even though ERA-Interim shows better performance than the other two, the amount and frequency are underestimated up to 90%. In CMIP5 output, CNRM-CM5 has the best performance in heavy rainfall simulation over China, but still can’t simulate the rainfall above the 200mm/day. The extreme precipitation performance has been much improved in FGOALS-f, which has both better spatial pattern and more realistic intensity-frequency/amount distribution. Particularly the precipitation above 200mm/day can be well reproduced in FGOALS-f. It is also found that increasing spatial resolution can get more accurate simulation of extreme rainfall.