Abstract: Surface temperature is an important geophysical parameter in energy and substance exchange processes of land-atmosphere interactions, and is a crucial element of global climate system. Time series of observed surface temperature data from meteorological stations are discrete and lack continuity in space. Satellite retrieval of land surface temperature is continuous in space, but with short time series. The simulation of surface temperatures using land surface models is an effective way of obtaining surface temperature data of high temporal resolutions. However, significant differences have been noted between different versions of land surface models, a large part of which lies in the physical processes of key land surface parameters. Thus a comparative evaluation of the performance of surface temperature for 1981-2004 was conducted over Inner Mongolia in this paper using NCAR (National Center for Atmospheric Research) Land Surface Models (CLM3.0, CLM3.5, CLM4.0 and CLM4.5) forced by NCEP (National Center for Environmental Prediction) and ground observation data. The study compared and analyzed differences in the results between models, tested the capability of models to simulate surface temperature over Inner Mongolia and provided guidance and reference for future improvements in the models. The results of the analysis of temporal and spatial variations in surface temperature over Inner Mongolia showed that:NCAR/CLMs simulations could be used to reproduce spatial and temporal variations that were in good agreement with data observed in ground stations. CLM4.5 model was the best as it had the highest correlation coefficient and lowest average deviation and RMS error for the Inner Mongolia due mainly to improvements in the calculation of roughness in the model. The simulation results for surface temperature by different CLM models were generally lower than the observed values. The mean deviation between the CLM simulated results and the ground observations was minimum during winter. Summer bias increased, especially for the eastern region. The deviation in temperature for the eastern part during summer was above 3℃. This indicated that the simulation capability of maximum surface temperature for the eastern and central regions was significantly lower than that for the western region. The differences between various versions of the model for the western region were not as obvious as those for the eastern and central regions. This was related to the improvements in snow patterns and hydrological processes of CLM4.0 and CLM4.5 model versions. In summary, CLM4.0 and CLM4.5 versions were fully applicable in Inner Mongolia, but the simulated values of surface temperature were lower than the measured ones. While the deviations were small in winter and high in summer, the in the eastern region were greater than those in the central and western regions.