Abstract: Global warming in the last 100 years has caused tremendous impacts on global and regional climate changes. As one of the largest arid areas in the world，Central Asia covers the territory of five countries （Kazakhstan，Kyrgyzstan，Tajikistan，Turkmenistan and Uzbekistan）and part of northwest China （Xinjiang Province）.Central Asia is a typical temperate desert steppe continental climate，precipitation is scarce，temperature changes severely and ecosystem is fragile. Freshwater mainly comes from precipitation in Central Asia. However，systematic researches for climate change of Central Asia under global warming are rare. As Xinjiang of China is a part of the Central Asia，so the research for climate in Central Asia is necessary to understand the climate change in China. Based on the CRU （Climatic Research Unit）dataset and the outputs of the historical and 4 representative concentration pathways （RCPs）future projection experiments from 10 climate models of Coupled Model Intercomparison Phase 5 （CMIP5），the climatic feature，interannual variability，and linear trend of the annual rainfall over Central Asia during 1951 to 2005 have been studied firstly. Then the statistical parameters of the 10 models and CRU dataset such as mean field，interannual variability and linear trend were used to calculate [NSTD]，spatial correlation and [RRMSE] between model results and CRU dataset to evaluate these models’ performance，after this assessment some models with relatively better performance among the 10 models are selected to establish a multi-model ensemble to project the spatial and temporal distribution of the annual rainfall over Central Asia during 2011 to 2060 under different greenhouse gas emission. The results showed that most models well simulated the spatial distribution of annual precipitation over Central Asia. The annual precipitation slightly increased but not significant over Central Asia during 1951 to 2005，and the simulation of MRI-CGCM3，MIROC5 and CNRM-CM5 performed relatively better. According to the assessment，4 models with relatively better performance among the 10 models were selected to establish a multi-model ensemble. Comparison of the 4 models’ ensemble mean precipitation with the CRU data show that the modeled location and extent of the high/low precipitation centers are in good agreement with CRU data during 1951-2005. The precipitation projected by the 4 models’ ensemble display that the spatial distribution of the annual rainfall over Central Asia are very similar under different representative concentration pathways （RCPs）during 2011-2060. In the future 50 years，the annual precipitation over Central Asia will increase significantly，especially in the southern part of Xinjiang，China （the low value area will change into a high value area）. Overall，in the next 50 years the annual precipitation increasing rates and the extent of regions with significant precipitation increasing rates over Central Asia expand with the increased RCPs.

Key words: Central Asia, precipitation, CMIP5, model Evaluation, RCPs