When there is not much information on how the interaction specifically works between multiple meteorological parameters and soil moisture, we tend to get help from artificial neural network (ANN), which has a great ability in function simulation. In creating a medium-sized BP network with some routine meteorological parameters as input variables and soil moisture the output, we find that the network could give a very good simulation with nearly an arbitrary precision in the phase of net construction, though validation results are not so satisfied as these trained, there is hope to get the network improved with a large number of samples and also farther exploration into the mechanism of these real relationships.