Simulation and prediction of hydrological and climate factors is very significant for climate change research, soil moisture forecasting, ecological environment improvement, reasonable development and utilization of water resources. Mean generating function method, BP neural network method and their combination are widely used in simulation and prediction. Each of these methods have their own have advantages, but there is still room for further improvement. As for rough selection of factor set,selection of the factor set combination and accuracy control of MGF, MGF-OSR, MGF-OSR-BP, a simulation and forecast model MGF-BP-I was built for taking the advantages of mean generating function and BP neural network. The mean annual precipitation in the Horqin Sandy Land was simulated and forecasted by using MGF, MGF-OSR-BP,MGF-BP-I. The results show that, in the modeling period, MGF-BP-I and MGF-OSR-BP have better fitting effect, optimization mode accuracy of MGF-BP-I is better than that of MGF-OSR-BP, and global optimization mode of MGF-BP-I is very good. In the verification period, MGF-BP-I verification phase optimization and MGF-BP-I global optimization mode of simulation results are best. MGF-BP-I takes the advantages of mean generating function and BP neural network, its accuracy is much higher than those of MGF-OSR and MGF-OSR-BP. MGF-BP-I global optimization model is more consistent with the practical application, the effect is ideal, can be used in the simulation and forecast of hydrological and climate factors.