Dynamic thresholds driven by rainfall-landscape synergy for non-point source pollution control: A zonal management framework for the Three Gorges Reservoir Area

• Developed novel rainfall-adaptive landscape thresholds as an ecological indicator. • Proposed a Rainfall-Adaptive Threshold Management (RATM) framework with distinct zonal strategies. • Advocated for source control, landscape optimization, and engineering solutions tailored to rainfall zones. Conventional, homogenous management, underpinned by static thresholds, is ill-equipped for non-point source (NPS) pollution control in heterogeneous watersheds. This study proposes a climate-adaptive paradigm to replace these inflexible thresholds. By coupling the Soil and Water Assessment Tool (SWAT) and Patch-generating Land Use Simulation (PLUS) models with logistic regression and Receiver Operating Characteristic (ROC) analysis, a continuous, rainfall-adaptive landscape threshold was derived to quantify the rainfall-landscape relationship. Results revealed a “polarization paradox”: while total TN and TP loads decreased by 45.8 % and 49.2 %, the pollution contribution efficiency (load share/area share) of Critical Source Areas (CSAs) intensified from 1.58 to 2.69 for TN and 1.65 to 2.90 for TP. Future scenarios projected this spatial concentration will persist, highlighting the insufficiency of relying solely on landscape-based strategies. This indicator was operationalized into a Rainfall-Adaptive Threshold Management (RATM) framework with three distinct zonal strategies: source control (e.g., a 16.2 % nitrogen and 16.0 % phosphorus fertilizer reduction) in the low-rainfall zone (<1034 mm); landscape optimization (e.g., converting 1122.64 km 2 of cropland) in the medium-rainfall zone (1034–1222 mm); and supplementary engineering solutions in the high-rainfall zone (>1222 mm). Crucially, comparative scenario analysis demonstrated that the RATM framework achieves superior NPS pollution reduction (41.0 % for TN and 40.1 % for TP) while requiring 18.6 % less cropland conversion than conventional single-threshold approaches. The RATM framework provides a spatially-explicit governance paradigm, offering a new, adaptable model for large reservoir watersheds worldwide.