Atmospheric reanalyses integrate a variety of observing systems with numerical models to produce a temporally and spatially consistent synthesis of data for weather and climate variability studies. Products provided by reanalyises are the objectively analyzed fields over the globe,including regions with minimal observations, and also for fields that are rarely or never observed. In order to evaluate the applicability of MERRA products over the Qinghai-Xizang Plateau(QXP)MERRA reanalysis surface air temperature product T2m with surface observation air temperature from 60 meteorological stations over 30 years from 1981 to 2010 were compared for the seasonal variations and long tern trends of temperature over the QXP. Main results are as follows: (1)MERRA T2m shows very high correlation with observation temperature with a correlation coefficient of 0.76 on average,in which 96.7% of the total stations is significant at the 95% level and 85.0% of the total stations is significant at the 99.9% level. The correlation in winter is better than that in summer. The relatively low correlation coefficients are found at higher latitude regions of northern QXP.(2)MERRA T2m is lower than the in-situ observations in most of stations on the QXP with lower 3.2 ℃ on average,and is more close to observation temperature in relatively flat terrain and homogeneous surface features in interior and northern region of the QXP while higher bias is found in complex terrain and surface features in the southeastern and southern mountain and river valleys where land-atmosphere interactions strongly affect the physical parameterizations in the numerical models and complex terrain and heterogeneous surface features as well as persistent cloud cover may contribute to these differences. MERRA T2m standard deviation is smaller than observed as well. However,for long term trends the increase in observation temperature(0.525 ℃·(10a)~(-1))is well represented by MERRA(0.342 ℃· (10a)~(-1))though MERRA underestimates this increasing rate.(3)According to analysis using observation data, it is indicated that the annual mean surface air temperature on the QXP ranges from-5.1 ℃ to 14.8 ℃,with 3.6 ℃ on average. Over the last 30 years from 1981 to 2010,QXP has experienced obvious increase in temperature with 0.525 ℃·(10a)~(-1)),which obviously higher than 0.29 ℃·(10a)~(-1))of China and 0.13 ℃·(10a)~(-1))of global warming trends. At seasonal levels,the highest temperature increase occurs in winter with 0.724 ℃·(10a)~(-1)), followed by spring(0.523 ℃·(10a)~(-1))and autumn(0.467 ℃·(10a)~(-1)),and the lowest increase in temperature is found in summer but it still reaches to 0.389 ℃·(10a)~(-1)). These 30-year trends of seasonal temperature are well represented by MERRA T2m product although increasing rates are to some extent underestimated.(4)Based on this work and previous comparison study for surface air temperature from NCEP/ NCAR,NCEP/DOE and ERAInterim over the QXP,MERRA T2m has higher correlation with observation compared to these reanalysis products, which means MERRA surface air temperature has some advantage over the QXP compared to other reanalysis data and is one of the better ways to compensate for sparse ground observation network,to implement climate change,energy balance and water cycles related study,and to use initialization fields for regional climate models etc.