Spatio-Temporal Evolution and Driving Factors of Agricultural Non-Point Source Pollution in the Upper Yangtze River Economic Belt Incorporating Ecological Regulation Functions

Abstract

Agricultural non-point source pollution (ANPSP) presents a critical threat to water security in the upper reaches of the Yangtze River Economic Belt (UYREB). Traditional export coefficient models (ECMs) often fail to capture the spatially heterogeneous processes of pollutant transport and ecological attenuation in complex mountainous terrains. Here, we constructed an improved ANPSP risk assessment framework by integrating ecosystem regulating functions (water yield, soil retention, and habitat quality) into the ECM. By coupling the Patch-generating Land Use Simulation (PLUS) model and Geographically and Temporally Weighted Regression (GTWR), we evaluated the spatiotemporal risk evolution from 2000 to 2020, projected risk configurations for 2030 under three scenarios (Natural Development, ND; Cropland Protection, CP; Ecological Protection, EP), and identified the spatiotemporal non-stationary drivers. The key findings are: (1) Independent validation using observed river water-quality data further demonstrated that the modeled risk index was significantly positively correlated with contemporaneous TN and TP concentrations (Pearson’s r = 0.673, p < 0.05), supporting its ability to identify the spatial distribution of relative ANPSP risk while accounting for regional differences in ecosystem regulation. (2) Historically, the integrated ANPSP risk index exhibited a “rise-then-decline” pattern, peaking in 2010 with extreme risk areas covering 12.1% of the region. Under the 2030 EP scenario, high-risk areas contract by 15.3% compared to 2020, demonstrating superior mitigation compared to ND and CP scenarios. (3) GTWR reveals that cropland proportion is the dominant positive spatial associate (0.800 − 0.852), while GDP shows a negative spatial association in economically developed sub-regions, reflecting potential environmental management co-benefits rather than direct causation. This framework provides an effective relative risk assessment tool for mountainous watersheds, highlighting that land use optimization under ecological regulation may contribute to reducing potential pollutant transport risk.