Based on the micro-climatic observation data by the automatic weather stations installed in 6 kinds of farmland local conditions. The relationship between the soil temperature of vegetables field, air temperature in the fruit tree garden, water temperature of fishpond and the observations at weather station are analyzed. The prediction models of soil temperature of vegetables field, air temperature of fruit tree garden, water temperature of fishpond as the function of the air temperature of weather station are developed using the stepwise progression method. And then with Visual Basic 6.0, an agricultural weather forecast system is set up on windows platform.