Land surface temperature(LST) is the key indicators to assess land surface models (LSMs). CLM4.5 has attracted much attention in mainland China. However, there have been few comprehensive LST assessments of CLM4.5 that used abundant latest long-term observation data from mainland China and considered land-atmosphere coupling. Therefore it is difficult to evaluate its performance for an actual climate simulation. In this work, LST data from the recent 30 years were collected from 809 Chinese meteorological stations, and the simulation capability of CLM4.5 for LST was comprehensively assessed for the first time. Then, in order to improve the model, sensitivity tests of soil thermal conductivity (STC) were carried out. Although CLM4.5 could accurately simulate the spatial distribution character of LST, there was a cold bias of 4.5{degree sign}C for all of mainland China. Seasonally, larger bias was observed in summer and autumn, which had more precipitation and greater soil moisture than other seasons. Deviation increased from southeast to northwest, but varied greatly between seasons. There was a significant linear regression relationship between two LSTs, with annual correlation coefficients of the two LSTs for all stations between 0.75 and 0.9 (P < 0.001). LST increased at a rate of 0.058{degree sign}C/a. Though it was successfully simulated, the trend value was smaller. The bias of CLM4.5 was better than that of ERA-interim but slightly worse than that of ERA-interim/Land. Assessment of three different STC schemes showed that the Lu-Ren scheme was the most one that suitalbe for LST siumulation in mainland china.