中文摘要：

水文气候因子模拟预测对气候变化研究、农业墒情预报、生态环境改善、水资源合理开发利用等具有一定参考意义。均生函数、BP神经网络及其结合改进方式在模拟预测中各有优点,被广泛应用,但仍有进一步改进空间。针对MGF、MGF-OSR、MGF-OSR-BP等方法粗选因子集、粗选集组合筛选、收敛适应性、精度控制等可改进空间,进一步发挥均生函数和BP神经网络优势,建立了MGF-BP-I模拟预测模型。利用MGF-OSR、MGF-OSR-BP、MGF-BP-I对科尔沁沙地区域平均年降水进行了模拟预测。结果表明,建模期MGF-OSR-BP、MGF-BP-I拟合效果均较好,MGF-BP-I建模阶段最优模式精度优于MGF-OSR-BP,MGF-BP-I整体同时最优模式结果也非常好。检验期,MGF-BP-I检验阶段最优及整体同时最优两种模式拟合效果最好,相比其他模式精度有所提高。MGF-BP-I考虑更加全面,充分发挥了均生函数和BP神经网络优势,精度远高于MGF-OSR和MGF-OSR-BP,MGF-BP-I整体同时最优模式更符合实际应用,效果理想,可用于水文气候因子模拟预测。

英文摘要：

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.