SWAT(Soil and Water Assessment Tool),a distributed hydrologic model integrated with geographic information system(GIS) and remote sensing(RS) to simulate the hydrological processes and stream flows,was improved and applied to a high altitude,cold,semi-arid Yingluoxia catchment located on Heihe river basin,northwestern China.In corresponding to different topography,soil types,land use/cover classifications over the 10 009 km~2 catchment,SWAT-GIS delineated the watershed based on 120 m resolution DEM into 157 simulation subwatersheds,each watershed was thought a unique Hydrological Response Unit(HRU) that owns the similar hydrological characteristics.Numerical experimental simulations on the Yingluoxia catchment suggested that the shallow storage recession factor(base flow recession factor),the altitudinal range discrimination was very essential to the well simulation of base flow and snow-melting process,and the accurate determination of shallow storage recession factor and successive simulation of snow-melting process on different altitude ranges are the key factor determining the simulation resolution.Through adjustment of the parameterization,satisfied simulation results were obtained with the accuracy of 0.88 for Nash-sutcliffe coefficient,0.91 for correlation coefficient over the 11 years continuous simulations on monthly discharges.