Abstract: Water resource is very deficient in inland river basin of Northwest China. The runoff from mountainous watershed basically presents the regional water resources. How to simulate and predict the runoff from mountainous watershed is very important to the sustainable development and ecological environment protection of the inland river basin in Northwest China. However, these mountainous basins have complex topography, versatile ground feature, different landscape elements, while the meteorological and hydrological stations in these basins are relatively scare. Therefore, it is difficult to create highly precise distributed hydrological models and other models based on physical mechanism of runoff generation. Thus, in this paper, 6 empirical models for runoff generation have been created, which are classified into 2 types, the nonlinear statistical model and the artificial neural network model (ANN). All the models work well. The NSE of statistical model 1 is 0.8342 and 0.8094 for simulation and prediction process respectively, of statistical model 2 is 0.8115 and 0.8736 respectively, and of statistical model 3 is 0.8005 and 0.8748. The NSE of ANNs model 1 is 0.9033 and 0.7778, of model 2 is 0.8971 and 0.7708, of model 3 is 0.9809 and 0.8550.For simulation process, the ANNs models run better than the statistical models, while in the predicted process, the statistical model work better. From these numerical experiments, it can be conclude that the monthly averaged runoff, the monthly averaged air temperature, the monthly evaporation capacity and monthly precipitation have highly nonlinear relationship between each other. Precipitation and air temperature both are the comprehensive factors that effect the variation of the runoff and evaporation. Especially, for runoff generation in mountainous watershed with much ice, glacier and snow, the air temperature is a very important factor. However, precipitation, air temperature and evaporation capacity, in general, could not reflect the runoff generation process completely. Other factors that have important impacts should are found out.

Key words: mountainous watershed of Heihe River Basin, monthly averaged air temperature, monthly precipitation, monthly averaged runoff, models