Abstract: Based on the CRU (Climatic Research Unit) dataset and the output of the historical simulations from 20 CMIP5 models, their simulation capabilities for the climatology and climate variability of the annual mean surface air temperatures in Central Asia have been examined by using Taylor diagram, the trend analysis, and EOF methods. The results show that the 20 CMIP5 models well simulated the significant rise trend, temporal evolution and spatial distribution of the annual mean temperature over Central Asia during 1951-2005, especially the high and low value centers and contour value distributions. The Taylor diagram shows that the root mean square errors are about 0.5 for most models, the spatial correlation coefficients range from 0.85 to 0.90 and the standard deviations range from 0.5 to 1.0. The EOF analysis results show that the 20-model ensemble can better simulate the features of the spatially consistent rise in annual mean temperature and the significant north and south reverse phase fluctuations over Central Asia.

Key words: air temperature, climate model, Taylor diagram, Central Asia, assessment