Abstract: Defining the driving factors of water yield in arid areas can provide a scientific basis for regional water resources optimization and sustainable development. In this paper, the InVEST model and geo-detectors were used to explore the spatial distribution of the average annual water yield in the Shule River Basin, Gansu Province, China and the driving forces at different spatial scales in 2015, based on the MODIS vegetation index, HWSD soil dataset, and meteorological data. Results show that the average annual water yield of the Shule River Basin was in an order of south > north > middle. At the basin scale, the dominant driving force of the spatial pattern of water yield is precipitation. The interaction of slope and precipitation is the most significant for water yield. At the regional scale, the dominant interpretation factors in the southern Qilian Mountains, northern Ma Zong mountains area, and the central plains are precipitation, human disturbance intensity, and precipitation, respectively. Most significant factor interaction is between human disturbance intensity and other factors. In different land uses, the slope in cultivated land is the dominant driving force, while precipitation of other land types is the main influencing factor of water yield. The interaction between precipitation and other factors in various categories has greatly enhanced the single factor driving force. Therefore, the research at the multi- scale driving mechanism of water yield in arid regions is crucial to the sustainable management of regional water resources.

Key words: water yield, InVEST model, geo-detector, Shule River Basin