A data-driven framework for planning-oriented decision support in integrated watershed management: insights from machine learning in Northern China

Spatial heterogeneity and anthropogenic disparities contribute to varying pollution challenges across global water bodies, highlighting the importance of understanding regional patterns and key pollution issues to support watershed management strategies tailored to local conditions. However, current research on watershed management decision-making is limited, with insufficient emphasis on the relationship between regional characteristics and pollution issues. To address the gap, this study developed a hybrid framework for identifying regional patterns and key pollution issues by coupling multiple machine learning models. This framework integrates K-means clustering with the extreme gradient boosting (XGBoost) classification model and employs shapley additive explanations (SHAP) analysis to enhance the classification and recognition of regional feature patterns. Furthermore, six prediction models were evaluated to predict key pollution drivers, with the gradient boosting machine (GBM) showing superior performance (Coefficient of determination, R² = 0.839; Mean squared error, MSE =0.00757). The results identified four distinct city clusters with divergent urban characteristics, including high pollution levels, well-developed agriculture, water shortage, and underdeveloped economies. Further analysis revealed specific pollution risks in different clusters, supporting the need for differentiated control priorities. The application of this framework in northern China demonstrates its effectiveness in identifying regional patterns and key pollution drivers, aiding governments and practitioners in efficiently conducting pre-planning for watershed pollution control based on regional characteristics. Positioned as the initial stage of a multi-layered decision-making architecture for sustainable watershed governance, the framework provides a valuable perspective and emphasizes the importance of developing a full-process decision support system.