Spatiotemporal patterns, driving mechanisms, and threshold responses of watershed ecosystem services from a supply-demand flow perspective

The Dongting Lake Basin, a hydrologically sensitive and ecologically strategic watershed in the middle reaches of the Yangtze River, plays a key role in flood regulation, water provision, and regional ecological security. As a typical river-lake coupled system, the basin exhibits strong upstream-downstream hydrological connectivity and pronounced seasonal variability in runoff redistribution. However, rapid urbanization and increasing human disturbance have progressively altered ecosystem service (ES) supply-demand relationships and disrupted service transmission processes across the basin. This study examines the spatiotemporal patterns, driving mechanisms, and threshold responses of watershed ecosystem services from a supply-demand flow perspective. We quantified the supply, demand, and ecosystem service supply-demand ratio (ESDR) of carbon sequestration (CS), water yield (WY), and habitat quality (HQ) from 2000 to 2020. Ecosystem service flow (ES flow) trajectories were simulated using prevailing wind fields for CS, explicit river network connectivity for WY to represent hydrologically constrained upstream-downstream transfer, and circuit-based ecological connectivity for HQ. An Extreme Gradient Boosting model integrated with Shapley Additive Explanations (XGBoost-SHAP), together with partial dependence analysis, was employed to identify dominant drivers, interaction effects, and nonlinear threshold intervals, which were subsequently synthesized to support spatial zoning. From 2000–2020, CS supply remained relatively stable, while demand nearly tripled, leading to an increase in the Carbon Sequestration Supply-Demand Ratio (CESDR) from 0.019 to 0.075. In contrast, sustained demand pressure caused the Water Yield Supply-Demand Ratio (WESDR) to decline from −0.004 to −0.007 and the Habitat Quality Supply-Demand Ratio (HESDR) to decrease from 0.494 to 0.480, indicating a progressive intensification of hydrological supply-demand imbalance despite stable or increasing water generation capacity. CS flow increased approximately threefold and predominantly followed southwest-northeast wind directions, whereas WY flow intensified strictly along upstream-downstream river networks, highlighting the dominant control of hydrological connectivity on water-related service redistribution at the basin scale. HQ flow networks became denser but less effective as supply patches declined and demand patches expanded, indicating increasing fragmentation. Driving mechanisms showed pronounced temporal reorganization, with human disturbance increasingly dominating CESDR and persistently constraining WESDR and HESDR. Nonlinear threshold analysis further reveals that water-related services exhibit abrupt deterioration once landscape fragmentation and cumulative human disturbance exceed critical hydrological connectivity thresholds, underscoring the vulnerability of basin-scale water regulation processes. Integrating these threshold responses enabled the identification of conservation, transition, and human-activity zones, each characterized by distinct supply-demand coordination patterns and flow stability, providing a process-oriented basis for improving watershed water regulation and hydrological resilience. • Identified ES supply–demand mismatches across the Dongting Lake Basin. • Simulated ES flows via wind, hydrology, and habitat connectivity models. • Detected nonlinear ESDR thresholds using XGBoost–PDPs framework. • Several dominant drivers exhibit clear recovery intervals beyond thresholds. • Developed spatial zoning strategies based on ESDR threshold responses.