Analysis of the complex correlations between land use and water quality using the PLS-SEM method combined with fluorescence fingerprinting data.

Accurate analysis of surface water pollution mechanisms is critical for effective environmental restoration and protection. However, evaluation methods for small watersheds with dense populations and complex pollution sources remain limited. This study integrates partial least squares structural equation modeling (PLS-SEM) with fluorescence fingerprinting data from excitation-emission matrix-parallel factor analysis (EEM-PARAFAC) to investigate nutrient sources in rivers of southeastern China. The findings reveal that land use intensity (LUI) significantly influences pollutant concentrations, but the presence of outliers underscores complex pollution mechanisms. Using EEM-PARAFAC components as mediators, the C5 component, representing sewage-derived substances, was identified as a key driver, fully mediating nitrogen (β = 0.953, p < 0.001, VAF = 117.5%) and phosphorus (β = 0.921, p < 0.001, VAF = 113.2%) levels. In contrast, agricultural non-point sources (C1 and C2: β = -0.270, p > 0.05) had negligible direct effects on nutrient concentrations, emphasizing the need to prioritize domestic sewage control. Additionally, components C1 and C2 exerted strong direct effects on dissolved organic carbon (β = 0.495, p < 0.001), surpassing the influence of sewage (β = 0.380, p < 0.001). These results demonstrate that the combined use of PLS-SEM and EEM-PARAFAC is a robust approach for identifying pollution sources in data-limited small watersheds, supporting cost-effective aquatic environmental restoration strategies.