Understanding the efficiency and uncertainty of water supply service assessment based on the Budyko framework: A case study of the Yellow River Basin, China

• Assessing efficiency and uncertainty of water supply in the Yellow River Basin. • Uncertainty in precipitation, ET 0 , Kc inputs, and ω value affected model performance. • Integration with the lumped model yielded more robust simulation results. • Climate change emerged as a key factor driving water supply service dynamics. Freshwater ecosystem services (ESs) analyses are increasingly employed to address water resource management challenges. However, few studies have systematically examined the efficiency and uncertainty of such assessments, limiting their applicability for decision-making. In this study, the InVEST water yield model was applied to assess water supply service in the Yellow River Basin (YRB) from 2000 to 2022. We evaluated the model’s sensitivity to climate variables and the parameter ω. Six sets of data/parameter input combinations were constructed to drive the model independently. Spatiotemporal trends were compared with observed data from 33 hydrological stations along the Yellow River mainstem and tributaries to assess the model performance and uncertainties. Finally, the response of water supply to climate change and vegetation dynamics was further discussed. The results showed that precipitation exhibited the highest sensitivity, and errors in precipitation inputs were the primary source of data input uncertainties. Compared to the raster-scale ω-value calculation method, the method combined with the lumped model delivered the most robust simulation results (R2, RMSE, and MAE for mainstream basins: 0.91, 50.08 mm, and 38.85 mm; for tributary basins: 0.89, 6.43 mm, and 4.04 mm, respectively). Climate change, particularly changes in precipitation, emerged as a key factor driving water supply service dynamics. These findings enhance the understanding of efficiency and uncertainty in water-related ESs assessments and offer valuable insights for applications in other regions.