Abstract: Based on the meteorological data of ten days and soil moisture data in three representative stations in Shanxi Province,the linear regression prediction equations of ten days for soil relative moisture in the different soil layers were built by using the statistical software SPSS,to improve the local agricultural meteorological services. According to the soil moisture balance equation,the prediction factors were filtered by analyzing the factor affected the soil moisture variability. The forecast model was validated by the historical data and data in 2007 for the different soil layers with the relative error under 30%. It was showed that physical statistics was a simple and easy method to forecast the soil moisture in semi-arid and semi-humid climate zones for the rain-fed farmland.

Key words: Statistical method, Statistical method, Soil relative moisture, Forecast