Flood Regulation Service Responses to Urban Green Space Change in a Plateau Valley City: A Case Study of Lhasa, China

Plateau valley cities are increasingly exposed to flood risk because topographic constraints, ecological fragility, and rapid urbanization jointly intensify runoff sensitivity. Although dynamic and process-based assessments of flood regulation service (FRS) have advanced, limited attention has been given to how changes in urban green space (UGS) components are associated with marginal runoff-retention responses in a plateau valley setting. Taking central Lhasa as a case, this study combined Sentinel-2-based UGS mapping, intPLUS land-use scenario simulation, Soil and Water Assessment Tool (SWAT) runoff modeling, XGBoost, SHapley Additive exPlanations (SHAP), and accumulated local effects (ALE) to evaluate the FRS responses under multiple scenarios. A pressure–response overlay was further used to diagnose spatial mismatches between runoff pressure and modeled regulation response. Results showed that UGS expanded by 5212.66 ha from 2017 to 2025, indicating a marked greening trend, but the natural development scenario projected a 1403.80 ha decline by 2033. In contrast, the ecological protection scenario projected a 915.98 ha increase and produced a stronger modeled runoff-retention response, although the improvement remained spatially uneven across subbasins. Model-based interpretation identified forest land as the most influential predictor of runoff-depth-based FRS response, while grassland, park green space, and waterfront green space were generally associated with runoff-depth reduction and showed distinct nonlinear response patterns. Specifically, forest land and grassland showed diminishing marginal effects after higher increment ranges, park green space exhibited delayed effectiveness before clearer regulation benefits emerged, and waterfront green space reached effective response ranges at relatively low proportional increments. Spatial diagnosis further showed that 19 subbasins, accounting for 18.88% of the study area, remained response-lagged, mainly along urban expansion fringes where runoff pressure was not matched by sufficient modeled regulation response. The integrated framework provides a transferable, spatially explainable basis for plateau valley FRS assessment, whereas the identified thresholds and zoning implications remain case-specific and require local validation.