Abstract: The watersheds in arid regions have the characteristics of large area and runoff supplied by mountains snowmelt and precipitation. The aim of this paper is to find the suitable method to simulate the runoff in arid area. The Soil Conservation Service Curve Number(SCS-CN) developed by the U.S. Department of Agriculture National Resources Conversion Service(NRCS) is the most popular and widely applied model for direct runoff estimation. This method is modified by accounting for the static portion of infiltration and the antecedent moisture. Based on the water balance equation and the curve number CN is derived from the tables given in the National Engineering Handbook for catchment characteristics, such as soil texture, land use, hydrologic condition, and initial soil moisture condition. The variability of these components can cause the instability of model parameters, which are sensitive to environmental changes. The method allows the relationship among the rainfall depth(P), the retention parameter(S), the curve number CN, the initial abstraction(Ia). In practice, a variability of S and CN is observed, as a result of seasonal and storm morphology changes. To describe this variability and assess errors in the estimation of CN, many statistical studies were performed. Based on the spatiotemporal difference of watershed, international and domestic scholars have brought up different methods to improve the SCS-CN model. Some of them are to improve it mechanism, and some is to improve it parameter calculation method. Based on large room for improvement of SCS model, the major problem of this paper to solve is to discuss the improving method for parameter algorithm fitting the typical watershed in arid region with large-scale precipitation-snowmelt mixed supply. In this paper, Wenquan Watershed in source regions of Bortala River, Xinjiang, China was taken as the case with the following paths to improve the SCS model:(1) to use the 16m resolution remote sensing image of high-definition earth observation satellite "Gaofen-1" for the first time to provide a high degree accuracy data of the land use classification for SCS-CN to determine the curve number. The classification accuracy is 86.43% and kappa coefficient is 0.837; the paper used the world soil database to supply the soil texture data for the CN value of SCS model, which reflected the advantage of GIS and RS in application of hydrological model. (2) Primary model of SCS is a pure study on rainfall and runoff. In order to satisfy the characteristics of mixed supplied runoff, the paper revised the precipitation to the sum of rainfall and snowmelt, and the snowmelt equivalency was calculated with degree-day model.(3) Comprehensive using of above methods to simulate the runoff of the 21 rainfall events during 2013-2014 of study area, the forecast the precision in calibration and verification period relative errors of improved parameter algorithm of SCS-CN model were 7.3% and 12.8%, and the main reason of simulation error of SCS-CN model is temperature owing to the temperature of mountain gap station are higher than that mountain surface. So the precision of the simulation results of the model needs to be improved. As to the large-scales and complex underlying surface conditions, the simulation results were in a reasonable range. The simulation results of the improved method show that it has feasibility and practicability for the study area, which can provide a reference for runoff simulation of similar watershed.

Key words: SCS-CN model, China-made GF-1 satellite remote sensing, degree-day model, runoff simulation