Study region: Yellow River Basin, a semi-arid and semi-humid basin in China. Study focus: Water yield (WY) is a critical indicator for water resources assessment. However, existing studies face limitations in WY validation and the exploration of nonlinear driving mechanisms. This study innovatively combines InVEST-modeled WY (validated using surface water and GLEAM_AET data) with explainable machine learning to quantify the nonlinear effects of climatic, anthropogenic, topographic, and soil factors on WY. New hydrological insights for the region: This work obtained more accurate WY and analyzed the nonlinear action mode of driving factors: (1) the new method obtained WY with equal total volume and similar spatial distribution; (2) precipitation (PRE) is the main factor affecting the spatial distribution of WY, with a contribution rate of 37.11 %, and its effect shows obvious geographical differentiation: the effect size gradually increases from north to south. In humid areas (PRE > 400 mm), WY and PRE increase linearly, while in arid areas, the response is not significant; the synergistic effect of altitude and root-restricting layer depth dominates the terrain-soil interaction process; (3) In terms of temporal evolution, the contribution rate of PRE change reached 74.22 %, and its increment was linearly correlated with WY gain. Land use change affects WY by changing evapotranspiration and root depth.