Key points are not available for this paper at this time.
Rapid economic development and population growth within the Qiantang River Basin intensified environmental pressures, leading to deteriorating water pollution and severe eutrophication at the basin’s outlet. To address these challenges, we developed a novel framework to identify the most sensitive subbasins impacting water quality downstream. The framework captured both direct influences from immediate upstream subbasins and indirect influences from more distant ones along nutrient flow paths. Three machine learning algorithms, including Random Forest (RF), Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost) were employed to simulate nutrient variability at subbasin outlets. RF performed best (R 2 = 0.92 for TN and = 0.88 for TP) and was selected for further analysis. Variable importance derived from the RF models suggested that nutrient levels depended strongly on those from the previous month, highlighting legacy effects and retention processes. Node influences analysis revealed significant upstream controls, especially from tributary nodes with higher nutrient loads. Based on these influences, we developed a subbasin-level sensitivity index and combined with stability assessment to identify critical source areas (CSAs) and classify their stability. This approach identified parts of the Jinhua river subbasins as stable CSAs causing sever water quality problems, while those in the Puyang River were classified as unstable CSAs. Different management strategies should be applied to CSAs with different stability levels. This study demonstrates the value of integrating spatial connectivity, temporal nutrient dynamics, and hydrological structure into nutrient management frameworks, providing a scalable and site-specific approach more targeted and effective water management strategies.
Xu et al. (Sat,) studied this question.