A novel framework for dynamic critical source area identification in mosaic agricultural micro-watersheds under rainfall-driven variability

Abstract

Non-point source (NPS) pollution in mosaic agricultural micro-watersheds poses significant challenges for water quality management due to dynamic interactions between fragmented land uses and rainfall variability. Traditional CSA identification methods, which rely on static parameters for pollutant load estimation and simplified hydrological distance metrics, fail to capture spatiotemporal shifts in pollution hotspots, compromising management efficacy. To address this gap, we develop a dynamic EMC-MSPA-MCR framework with two key improvements: (1) quantifies event-scale pollutant loads via dynamic Event Mean Concentration (EMC) analysis, accounting for land-use-specific runoff thresholds and rainfall scenarios, and (2) refines export efficiency using Morphological Spatial Pattern Analysis (MSPA) and Minimum Cumulative Resistance (MCR) to incorporate landscape connectivity and transport resistance. Applied to the Yankou Reservoir watershed (China), the framework achieved high accuracy (Spearman’s ρ = 0.67–0.69), identifying CSAs that contributed 65–81 % of total phosphorus (TP) loads from only 1.54–15.85 % of the area. Key results revealed rainfall-driven CSA dynamics: transportation land dominated under light rainfall (1.4–11.4 mm; 97 % of loads), croplands under moderate rainfall (11.4–24 mm; 65 %), and mixed cropland-forest sources during heavy rainfall (≥24 mm; 76 %). MSPA-MCR integration improved export efficiency predictions by 8–23 % compared to traditional methods, with core landscapes (45.6 % of CSAs) and branch/bridge zones (35–41 %) emerging as critical transport pathways under heavy rainfall. The framework’s ability to pinpoint compact, high-impact CSAs (e.g., 63 % of annual loads from 11 % of rainfall events) supports targeted interventions, offering a scalable tool for precision NPS management in heterogeneous agricultural micro-watersheds.