A new forecast method of the field soil moisture in dry season is built based on the artificial neural networks of the backlpropagation model. The results showed that the forecast accuracy is high enough as compared with the measured values. In general, this method is suitable for routine forecast of the soil moisture in computator condition of IBM PC/386. This new method can be used to study the rational utilization of the soil moisture resources and to decrease the waste of agrometerogic resources of the light and heat for the dry farming regions and dry season in China. It is likely to be a new approach to use valuable observational data of soil moisture.