Abstract: Analysis of water conservation functions of mountainous areas and small watersheds, as well as the spatio-temporal patterns and correlations between water conservation and climate and human activities, is the foundation for regional ecological restoration and water resource management. We analyzed the spatial and temporal variations of water conservation in Xihanshui Basin based on the water production module in the InVEST model, combined with the Pearson correlation coefficient. We used scenario simulation and geoDetector modeling to explore the response of water conservation to climate and human activities and spatial differentiation driving factors. The results showed that water conservation of Xihanshui Basin showed an overall upward trend from 1990 to 2020, with an average value of 38.35 mm. There was a spatial distribution pattern of high in the northwest, high in the southeast, low in the northeast, and high in the mountainous areas and low in the loess hilly areas. Both precipitation and water conservation showed a trend of first decreasing and then increasing. Precipitation, potential evapotranspiration, and temperature had highly significant positive, significant negative, and non-significant positive correlations with water conservation. The water conservation was mainly contributed by forest, grassland and cultivated land, accounting for more than 60%. Water conservation per unit area of each class was 61.66, 41.28 and 33.31 mm·km^-2, respectively. The explanatory power of the pair interaction of different driving factors on the water conservation was higher than that of any single factor. The explanatory power of the interaction between the comprehensive index of land use degree and potential evapotranspiration, the comprehensive index of land use degree and NDVI, was relatively high. With the increases of uncertainty of future climate change, the decrease of precipitation would become the key factor limiting water conservation. In addition, promoting the conversion of unused land into eco-logical land types would promote the steady improvement of regional ecological environment quality.

Key words: water conservation capacity, InVEST model, geoDetector, Xihanshui Basin