Based on the characters of semi-tropical climate,topography,vegetation and the NOAA/AVHRR data in Guangxi,the split-window algorithms for retrieving the LST(land surface temperature)from satellite are analyzed and compared.The results suggest that the Becker&Li algorithm is suitable to retrieve the day-time LST in Guangxi.The clear-sky day-time-averaged LST in Guangxi is imaged by using the Becker&Li algorithm.We find out the approach of calculating annual mean LST by employing appropriate templets.For the cloudy region,the pixel values are substituted by the image data of same period and close time after being calculated. The calculation is based on the correlation of clear sky image with cloudy one.At last,the clear-sky day-time-averaged LST image of Guangxi is obtained.Analysis results indicate that the LST image made by the satellite remote sensing in this research is more extensive,more particular and clearer than that from the second agroclimatic regionalization in describing the distribution of thermal resource.LST is a more effective factor of thermal regionalization.