中文摘要：

利用第五次耦合模式比较计划（Phase 5 of Coupled Model Intercomparison Project,CMIP5）提供的30个全球气候模式模拟的1961~2005年的夏季逐月环流场资料及同期NCEP再分析资料,引入泰勒图及各种评估指标,探讨全球气候模式对东亚夏季平均大气环流场的模拟能力,寻求具有较好东亚夏季环流场模拟能力的气候模式。结果表明：1）全球气候模式能够模拟出东亚夏季平均大气环流的基本特征,CMIP5模式的总体模拟能力较第三次耦合模式比较计划（CMIP3）有较大程度的提高,如CMIP5模式对东亚大部分地区夏季海平面气压（Sea Level Pressure,SLP）场的模拟偏差在6 h Pa以内。2）模式对不同层次环流场的模拟能力存在差异,500 h Pa高度场的模拟能力最强,其次为100 h Pa高度场、850 h Pa风场,SLP场最弱;对东亚夏季主要环流系统的模拟对比发现,模式对印度热低压及东伸槽强度指数的模拟能力最好。3）综合CMIP5模式对东亚夏季各层次平均环流场以及主要环流系统的模拟能力,发现模拟较好的5个模式为CESM1-CAM5、MPI-ESM-MR、MPI-ESM-LR、MPI-ESM-P和Can ESM2。4）相对于单一模式,多模式集合平均（MME）模拟能力较强,但较优选的前5个模式集合平均的模拟能力弱。

英文摘要：

Based on 1961-2005 monthly reanalysis data of atmospheric general circulation from the NCEP data, this paper evaluates the summer atmospheric general circulations in East Asia （EA） simulated by 30 climate models in the Phase 5 of Coupled Model Intercomparison Project （CMIP5） historical simulation experiments. Taylor figures and various assessment indicators have been used to find the models which are better in simulating summer atmospheric general circulation in EA. Results show that： 1） Global climate models are able to simulate the basic features of summer average atmospheric general circulation over EA. Compared with CMIP3 models, the simulation ability of CMIP5 models have made a great improvement. Taking sea level pressure （SLP） for example, the simulated deviations are less than 6 hPa over most areas of EA. 2） Modeling abilities of general circulation for different levels are different. The simulation of 500 hPa geopotential height is the best, followed by 100 hPa geopotential height and 850 hPa wind, however, SLP is relatively poor. Compared with simulations of main circulation systems, the Indian hot low pressure and eastward extending trough intensity index are well simulated. 3） According to the performance of 30 climate models, the authors find five models which simulate the average atmospheric general circulation and the main circulation system better. They are CESM1-CAM5, MPI-ESM-MR, MPI-ESM-LR, MPI-ESM-P, and CanESM2. 4） The performance of multi-model ensemble is better than any single one, however, it is weaker than the best five models ensemble mean.