Abstract: In the development of distribution hydrological models, finding a way to produce distribution precipitation data is a hotspot in hydrological research. And it is not only a hot issue, but also a difficulty that develops a model calibration method based the hydrological and physical processes for the practice of the PUB. In this study, the distributed hydrological model-SWAT is applied in the upstream of the Yili River basin with scarce data for the key issue about its water resources evaluation. The catchment area of the basin is nearly 57 000 square kilometers, only four hydrological stations and two meteorological stations have. Firstly, the input data are important for hydrological processes of the basin，especially the precipitation. The spatial distribution of precipitation in this basin is appropriately described by dividing the basin into eight regions, and the precipitation lapse rates in these regions are different each other, varying from 7.9 mm/100 m to 19.6 mm/100 m. Furthermore, there are many elevation bands according to the variation of the elevation in each subbasin. Then, a new suitable method for model calibration is adopted using multitemporal scales, including annual scale, monthly scale, flood and drought period, multivariables, such as precipitation, evaporation, total runoff, surface runoff and ground water, and multisites incorporates the natural and social characteristics of the basin, which include topography, geological structure, precipitation mechanism, general circulation, agricultural irrigation and reservoirs. This application obtained a preferable simulation results, all indexes reaching the standard. Both NSE and r2 are above 0.6, the highest of 0.83, and RE are all within 8%. It indicates that the model represents the hydrological processes of the basin properly; the results can provide references for the evaluation and management of water resources of the basin. Besides, to improve the model's simulation, the following studies are needful: to perfect the research on hydrologic mechanism, to improve the observation level of precipitation, and to investigate the related parameters of underlying surface including soil and vegetation.

Key words: SWAT model, precipitation, hydrologic mechanism, model calibration, datascarce basins