Modeling stream water quality variance and interactions across watersheds: a multilevel and spatial approach

Abstract

As land use pressures intensify across watersheds, understanding how landscape features affect stream water quality has become increasingly urgent for effective environmental management. By integrating hierarchical linear modeling (HLM) with geographically weighted regression (GWR), this study captures both broad-scale trends and localized spatial variability in the relationship between water quality (biochemical oxygen demand [BOD], total nitrogen [TN], and total phosphorus [TP]), land use, and watershed topography—addressing the limitations of traditional statistical models that assume uniform effects and overlook spatial hierarchy. Our findings indicate that urban and agricultural land use were positively associated with elevated BOD and TP concentrations, but that these effects were generally weaker in areas with steeper topography. BOD interacted significantly with watershed slope in both urban and agricultural areas, while TP showed a significant interaction only in agricultural watersheds. In contrast, TN showed no significant interaction with watershed slope, suggesting that nitrogen dynamics are primarily influenced by hydrological processes and land management practices rather than topography. These results highlight the importance of multilevel modeling in water quality assessments, demonstrating that watershed characteristics can amplify or constrain land use impacts on stream ecosystems. Covering the entire national territory of South Korea, our findings provide valuable insights for land-use planning, watershed management, and national water quality policies, supporting more effective and targeted pollution control strategies.