Using observational data spanning the period from February to December 2009 and recorded at the Suli station in Qinghai Province,the land-surface model CLM3.0 was employed to simulate the freezing and melting of soil.The results indicate that the simulated soil temperature is higher than the observed soil temperature and the ultimate thawing date is earlier than the observed date during the melting period.During the freezing period,the simulated soil temperature is lower than the observed soil temperature and the ultimate freezing of the deep soil is earlier than that observed.Overall,the simulation of freezing is better than that of melting,and the simulation of a shallow layer is better than that of a deeper layer.In the original CLM3.0,it is assumed that frozen soil begins to melt when the soil temperature exceeds 0C,which is inconsistent with observations.The critical freeze-thaw temperature was calculated according to thermodynamics equations and the freeze-thaw condition was modified.In this work,the melting rate for frozen soil was reduced using the modified scheme,and the simulated soil temperature was lowered. Meanwhile,the refreezing of soil during the melting season was well simulated and more closely matched observations.Additionally,it was found that the rates of melting and freezing differ,with the former being slower than the latter,but refreezing during the melting season is rather quick. Using observational data spanning the period from February to December 2009 and recorded at the Suli station in Qinghai Province, the land-surface model CLM3.0 was employed to simulate the freezing and melting of soil. The results indicate that the simulated soil temperature is higher than the observed soil temperature and the ultimate thawing date is earlier than the observed date during the melting period. During the freezing period, the simulated soil temperature is lower than the observed soil temperature and the ultimate freezing of the deep soil is earlier than that observed . Overall, the simulation of freezing is better than that of melting, and the simulation of a shallow layer is better than that of a deeper layer. In the original CLM 3.0, it is assumed that frozen soil begins to melt when the soil temperature超过 0C, which is inconsistent with observations. critical critical-thaw temperature was calculated under thermodynamics equations and the freeze-thaw condition was modi fied.In this work, the melting rate for frozen soil was reduced using the modified scheme, and the simulated soil temperature was lowered. Meanwhile, the refreezing of soil during the melting season was well simulated and more closely matched observations. Additionally, it was found that the rates of melting and freezing differ, with the former being slower than the latter, but refreezing during the melting season is rather quick.