中文摘要：

利用参与IPCCAR4的25个全球气候模式的模拟结果与观测资料，通过分类集合检验气候模式对东亚大陆气温和降水的模拟能力。结果表明：（1）高、中、低分辨率以及所有模式的集合（分别记为I、Ⅱ、Ⅲ、Ⅳ类）结果均能较好地再现1901-1999年温度的增温趋势，其中II类的线性趋势与观测值最为接近，但年际变率均小于观测值；（2）I、Ⅱ、Ⅲ、Ⅳ类均能很好地模拟出气候态温度的月变化特征以及空间分布型，但模拟值偏小，其中I类的冷偏差最小；（3）对于降水的时间演变特征，模拟效果明显偏差、不确定性显著；（4）4种集合方式均能较好模拟出气候态降水的空间分布型，但模拟值普遍偏高，其中I类的误差最小。可见，在温度和降水的时间变化方面，高分辨率的集合结果并没有体现出相应的优势，但在空间分布上模拟效果明显要好于其他3种方式。

英文摘要：

Based on the observation and simulation results about surface temperature and precipitati- on derived from the outputs of coupled atmosphere-ocean general circulation models participating in the IPCC AR4, the assessment of global models＇ simulation capabilities in East Asia continent has been done by the methods of multi-model ensembles according to different horizontal resolution of atmospheric model. The results are as follows ： （ 1 ） the results of high, medium, low resolution models and ensemble of all models （ referred to I , II, Ⅲ, Ⅳ Ensemble respectively） are able to reproduce the increasing trend of temperature from 1901 to 1999 and the linear trend of 11 Ensemble is closer to the observed tempera- ture than that of other results, but the interannual variability of all four ensemble results is smaller than that of observation. （2） The four ensemble results can also accurately simulate the monthly variation and spatial distribution characteristic of temperature from 1980 to 1999, but the simulation value is lower than the observation. In addition, the I Ensemble＇s error compared to the observation＇s is the least. （3） There is a significant deviation in simulation results of time evolution characteristics of precipitation, so the simulation uncertainty about precipitation is obvious. （4） The four ensemble results can reproduce spatial distribution characteristic of precipitation from 1980 to 1999, and the I Ensemble＇s error compared to the observation＇s is the least. However, the simulation value is higher than the observation＇s. There- fore, the I Ensemble has no advantage on the simulation of temporal variation for temperature and precipitation, but the simulation performance of spatial variation is obviously better than that of the other three ways.