中文摘要：

蒙古高原受自然与人为因素影响,生态环境面临严峻压力,研究气候变化特征及气候模式评估对该地区气候及生态环境等相关研究具有重要的参考意义。本研究以整个蒙古高原为研究区,研究蒙古高原气温与降水要素的变化趋势,对比研究中国内蒙古自治区与蒙古国地区气候变化特征,并评价了全球耦合模式比较计划第五阶段（CMIP5）模式输出的气温与降水序列在研究区的表现,利用一种改进的秩评分方法对CMIP5模式的模拟能力进行了综合评价。结果表明,蒙古高原过去几十年整体呈增温减湿的趋势,最低气温升温幅度明显大于最高气温升温幅度,中国内蒙古地区变暖趋势强于蒙古国地区;CMIP5模式模拟的平均气温普遍偏低,且低估了该区域的增温趋势,而模拟的降水量普遍偏高。不同要素间,CMIP5模式对平均气温的模拟能力最强,最高与最低气温次之,降水相对较差。模式评价结果对评价指标有较大的依赖性,因此,评估气候模式的区域效果时,建议使用多指标进行综合评价。

英文摘要：

Trends in air temperature and precipitation for the Mongolian Plateau were detected and changes between the Inner Mongolian region and Mongolian region were analyzed. Based on multievaluation indexes,Coupled Model Intercomparison Project Phase 5（CMIP5）climate models were assessed using observed station data for air temperature and precipitation over the Plateau.The overall performance of CMIP5 models for all six variables was evaluated by a revised rank score evaluation method. We found that Nor ESM1-M were the relative best models for overall performance over the region and had better ability modeling each variable. Although some models were not good enough for overall performance,they showed better ability when modeling some individual variables. CMIP5 modes tended to underestimate mean values and trends in mean air temperature and overestimate precipitation. For different variables,CMIP5 models showed the best model performance for mean air temperature over the region. The models also showed better model performance for maximum air temperature and minimum air temperature. Some models showed ability in modeling precipitation amount over the region. The assessment results were too dependent on evaluation indexes and when different evaluation indexes were used,completely opposite results were often obtained. In other words,selection of evaluation indexes was very important to assessment. For example,as for mean characteristic indexes,ACCESS1.3,CanESM2 and MIROC-ESM models showed the best model performance for mean air temperature over the region,but ACCESS1.3 and MIROC- ESM showed the worst model performance using the probability density function index,and CanESM2 showed the worst performance when using the spatial correlation coefficient index. Therefore,in the assessment of climate models at a regional scale,multi-evaluation indexes representing different characteristic are suggested for overall evaluations.