Abstract: In order to resolve the deficiency of micro-climatic observation data in the work of agricultural weather forecast, a case study of Mianyang had been researched. The author used the date of average temperature and land surface temperature from 1981 to 2009, the relationship between them had been analyzed. Based the results, the forecast model that using the temperature as the independent variable had been established by the method of stepwise regression analysis, and the result had been tested .It showed that: most of deviation between predictive value and measured value were less than ±1.0℃, and the average relative error less than 10%. The result showed that forecast model had high precision, and that could be use to prediction of land surface temperature.