Water quality is a key factor in safeguarding ecosystem functions, protecting human health, and ensuring sustainable development. The characteristics of land use (including type, intensity, and landscape configuration) are important indicators of human activity and significantly impact river water quality, with these impacts varying across different spatial and temporal scales. This study focuses on the source region of the Chishui River, integrating two spatial scales: the riparian buffer and the sub-watershed. Using variation partitioning analysis and random forest modelling, we quantify the independent and combined contributions of land use type, intensity and landscape pattern to water quality. This allows us to identify the key influencing factors and their corresponding spatial scales. The results show that:Landscape pattern is the main dimension influencing water quality changes (explaining 33%-58% of the variation), followed by land use type (11%-22%) and intensity (4%-16%). Riparian landscape configuration exerts a more significant influence on water quality, while land use intensity at the sub-watershed scale provides stronger explanatory power. The proportion of built-up land and the intensities of construction and agriculture are key predictors of water quality, while forest cover and landscape connectivity play important roles in reducing total nitrogen and chemical oxygen demand concentrations. These findings suggest that water quality management in watersheds should consider land use characteristics at multiple scales in order to identify the most effective combinations of riparian and sub-watershed interventions. Specifically, we recommend prioritizing the control of industrial and domestic point-source pollution at the riparian scale and enhancing agricultural non-point source management at the sub-watershed scale to form an integrated ‘point-source and non-point source’ control system. This study provides new empirical evidence on the multidimensional interactions and scale effects of land use–water quality relationships, offering important theoretical and practical insights for the protection of watershed resources and the optimization of spatial planning.