Evaluation of water quality characteristics, spatiotemporal variations, and pollution sources apportionment in the Beijing-Hangzhou Grand Canal, Eastern China

• DO, COD Mn , COD, BOD 5 , NH 3 -N, and TP were comprehensively evaluated across the Beijing-Hangzhou Grand Canal. • The modified Water Quality Index (WQI-DET) reflected the spatiotemporal dynamics of water quality along the Grand Canal. • Study identified the critical variables causing water quality deterioration at 20 monitoring sites. • SHAP interpretation combined with RFR model was applied to identify the primary driving factors influencing water quality. Effective management of large urban rivers requires accurately identifying the spatiotemporal drivers of key indicator exceedances and apportioning pollution sources from the latest monitoring data, thereby providing actionable insights for watershed administrators. The Beijing-Hangzhou Grand Canal, as the world’s longest artificial river and a vital “golden waterway” traversing several central cities in eastern China, plays an essential role in regional development. This study systematically evaluated its water quality patterns and spatiotemporal dynamics using a novel Water Quality Index (WQI-DET), which can better reflect water quality conditions and identify the causes of water quality deterioration, incorporating six key indicators: dissolved oxygen (DO), permanganate index (COD Mn ), chemical oxygen demand (COD), biochemical oxygen demand (BOD 5 ), ammonia nitrogen (NH 3 -N), and the total phosphorus (TP). Our analysis uncovered marked spatial and seasonal fluctuations, with water quality deteriorating notably during summer. The WQI-DET values also revealed that key indicators influencing water quality deterioration exhibited distinct seasonal and spatial variability. DO emerged as the dominant deterioration factor in summer, while TP played the leading role in other seasons. Spatially, COD Mn -related pollution was more severe in the northern section of the canal, whereas NH 3 -N contamination predominated in the southern reaches. This study adopted Random Forest Regression (RFR) coupled with SHAP (SHapley Additive exPlanations) method to identify the primary drivers of water quality dynamics. To facilitate effective and actionable watershed management, this study considered various variables (30 in total), including environmental, socioeconomical, and industrial structure factors. The results highlighted that irrigated area, fertilizer application, temperature, precipitation, and livestock rearing (pigs and sheep) were the dominant contributors to exceedance across all six indicators. These findings enhance comprehension of the mechanisms underlying water quality variation along the Grand Canal, offering valuable insights for targeted and effective regional water quality management.