This paper presents a soil temperature simulation scheme of using the conventional meteorological station data,thus improving the soil temperature simulation of the land surface process model of BATS proposed by Dickinson et al.This scheme can simulate the annual,seasonal and diurnal soil temperature variations at different depth fairly well.The results show that the simulated soil temperature of multi-layers by using the heat diffuse equation is comparable with the observed data;and the model is able to simulate the annual,seasonal and diurnal temperature variation very well at various levels.The temperature of deep layer is higher than that of surface layer in winter,while the upper layer temperature is higher than that of the deeper layer in summer.The times when the temperature gradient begins to reverse are about in April and October,which is in agreement with the feature observed.The model correctly simulates the time lags of soil temperature variations of various layers.