Abstract: Ecology and environment of arid Central Asia is fragile and sensitive to climate change. However，its tempo-spatial pattern of climate change is uncertain. Owing to the sparse distribution of meteorological stations and the lack of high-precision meteorological data，using a single data source to study the regional climate change may introduce great uncertainty. Therefore，it is highly recommended to combine site observation meteorological data with reanalysis climate data to explore the tempo-spatial pattern of climate change in Central Asia. In this paper，data of 31 meteorological stations （OBS），the CRU meteorological interpolation data and three high-resolution reanalysis data：CFSR，ERA-Interim and MERRA are used to analyze the temporal changes of air temperature. The applicability of the CRU，CFSR，ERA-Interim and MERRA，in Central Asia were evaluated with the observed temperature data （OBS） during 1980-2011 from 31 meteorological stations with the correlation analysis，T-test and the method of least squares. The accuracies of the reanalysis datasets were quantified with absolute error （AE），correlation coeffic- ient （CC），mean absolute error （MAE） and root mean square error （RMSE）. The evaluation results showed that all the datasets can catch the annual and seasonal temperature change well （The CC value is 0.70 to 0.98）. These indicated that all the four datasets can be used to describe the temperature change of Central Asia，especially the temperature spatial pattern. For each dataset，the CRU and ERA-Interim had an underestimation for the air temperature of OBS and the underestimate range is -1.14 to -0.31 ℃. While the MERRA data had an overestimation for the OBS，the overestimate range is 0.09 to 2.03 ℃. In addition，there are best simulation results for the annual and winter temperature: the CC values are bigger than 0.9 and the MAE is -0.5 to 1.23 ℃. However，the summer simulation results were worse than spring，fall and winter: the CC values reached 0.7-0.84 and the MAE is -1.14 to 2.03 ℃. The temperature change results indicated as follows：（1） In the last 32 years，annual average air temperature has increased significantly by 0.36-0.47 ℃·（10 a）^-1，which means that has already rose 1.15-1.50 ℃ in Central Asia during the period；（2） Air temperature rise is the fastest 0.71-0.93 ℃·（10 a）^-1 in spring，but there is no obvious change in winter；（3） The middle，the south，the southwest and the west region of Central Asia has experienced significant warming process during the last 1990s to the early 2000s，while the air temperature of other parts in Central Asia has no obvious change.

Key words: Temporal pattern analysis, Central Asia, reanalysis data, temperature change rate, Mann-Kendall Test