Abstract
This study presents a novel approach to investigating the spatial relationship between land use and river water quality by applying Geographically Weighted Regression (GWR), which explicitly accounts for the nested structure of sub-watersheds-a factor that has been frequently overlooked in previous studies. The Yura River watershed in Japan was selected as the study site, and electrical conductivity (EC) was used as a comprehensive indicator of water quality. To reflect local land-use impacts, we introduced the difference in EC between upstream and downstream sampling points (ΔEC) and allocated it to individual sub-watershed polygons. By analyzing both irrigation and non-irrigation seasons, the study found that key land-use types, such as paddy fields, water bodies, and evergreen broadleaved forests, exert varying influences on water quality depending on the season and location. The GWR model outperformed global regression models in capturing spatial heterogeneity and reduced residual spatial autocorrelation, thereby validating its effectiveness in watershed-scale environmental analysis. Importantly, this study is the first to integrate GWR with ΔEC while considering the hierarchical structure of sub-watersheds. This framework enables more accurate identification of localized land-use effects on water quality, which are often masked in global models. The findings underscore the need for region-specific land-use management and offer methodological insights for improving watershed conservation strategies in heterogeneous landscapes. By highlighting both seasonal variation and spatial dependency, this study provides a useful toolset for environmental monitoring and supports the development of targeted, evidence-based watershed policies.