Abstract: Water resources system is very fragile and the climate change in arid region is sensitive to the global climate, so it is necessary to enhance the water resources system in arid region of northwest to address climate change and allocate limited water resources rationally. It is absolutely significant to quantify the water resource in inland river basins especially in the arid zone of northwest China. This paper took the mountainous watershed of Urumqi River as a study area to simulate the hydrological processes of mountainous watershed in inland river basins by using soil and water assessment tool(SWAT)and sequential uncertainty fitting(ver.2)algorithm, and selected 25 parameters that were closely associated with the hydrologic processes to execute parameter sensitivity analysis. The sensitivity analysis results showed that ALPHA_BNK, SOL_K, GW_DELAY, TLAPS, PLAPS and some snow melt parameters(SMFMX, SFTMP, SMTMP)played a very important role in hydrological processes in the mountainous watershed and it is crucial to add elevation bands when the SWAT model was used in the mountainous watershed. SWAT daily simulation results showed that both the Nash-Sutcliffe efficiency and the determination coefficient values of the calibration period(2002-2006)and validation period(2007-2011)were higher than 0.75(monthly simulation results surpassed 0.85), and the percent bias was controlled within ±25%, which indicated that the simulation results were satisfactory. In uncertainty analysis, the higher P-factor and lower R-factor also showed that the model uncertainty was relatively small. All these statistics make sure that it is feasible to use SWAT as a flow simulator in the upper reaches of Urumqi River. Also it could provide reference for other similar researches in the mountainous watershed of northwest China and support the further research in this river basin.

Key words: SWAT model, Sufi-2 algorithm, sensitivity analysis, uncertainty analysis