Using the Simple Biosphere Model (SiB2),soil thermal properties (STP) were examined in a Tibetan prairie during the monsoon period to investigate ground surface temperature prediction.We improved the SiB2 model by incorporating a revised force-restore method (FRM) to take the vertical heterogeneity of soil thermal diffusivity (k) into account.The results indicate that (1) the revised FRM alleviates daytime overestimation and nighttime underestimation in modeled ground surface temperature (Tg),and (2) its role in little rainfall events is significant because the vertical gradient of k increases with increasing surface evaporation.Since the original formula of thermal conductivity (λ) in the SiB2 greatly underestimates soil thermal conductivity,we compared five algorithms of λ involving soil moisture to investigate the cause of overestimation during the day and underestimation at night on the basis of the revised FRM.The results show that (1) the five algorithms significantly improve Tg prediction,especially in daytime,and (2) taking one of these five algorithms as an example,the simulated Tg values in the daytime are closer to the field measurements than those in the nighttime.The differences between modeled Tg and field measurements are mostly within the margin of error of ±2 K during 3 August to 4 September 1998.