Identifying the drivers of water yield ecosystem service: A case study in the Yangtze River Basin, China

•We explored the multi-level driving factors of water yield in a large river basin.•Regions of high importance to the supply of water yield service were in the southeast.•Local changes in water yield were positively spatially autocorrelated.•Precipitation and the extent of urban area were important regional factors.•Several land composition metrics could act as indicators of local water yield. The Yangtze River Basin (YRB) is a region of vital economic and ecological importance in China, facing challenges of water scarcity and flooding underpinned by complex, yet underexplored, interactions between anthropogenic and biophysical changes. This study conducted a multi-scale analysis to (i) quantitatively evaluate the spatio-temporal patterns of the water yield ecosystem service of the whole of YRB from 2000 to 2015 using the InVEST model and (ii) investigate the interactive influence of climate, land-use, landscape, and socio-economic factors on the sub-watershed level water yield using the automatic linear and geographically weighted regression modelling. The results pinpointed regions of high and increasing water yield in the southeast of YRB, and that of low and decreasing water yield in the northwest. The YRB’s water yield was found to, in particular, be associated with the average annual precipitation and the extent of the urban area. We identified three interaction effects, which suggest that fostering connected patches in afforestation practices, maintaining slow-pace enhancement in compact urban development and improving the regularity of cropland’s shape in small-scale agriculture development are necessary to sustain the supply of water yield ecosystem service. Our study provides new insights into the joint effects of anthropogenic and biophysical factors underpinning the supply of YRB’s water yield ecosystem service, and sheds light on the indirect pathways of managing water resources via controlling local-level landscape metrics.