Probabilistic risk assessment of heavy metals sources by integrating interpretable machine learning with receptor models.
Source: PubMed, NCBI / U.S. National Library of Medicine
Accurate source apportionment of heavy metals (HMs) in surface water and the development of a source-based risk assessment system is essential for formulating effective pollution mitigation strategies. However, the combined contamination patterns of multiple HMs and the quantitative linkage between pollution sources and health risks remain inadequately understood, particularly in high-altitude regions. This study selected Gangba County, a representative transboundary headwater area located on the northern flank of the Himalayas in the southern Qinghai-Tibet Plateau, as the study area. An integrated framework incorporating APCS-MLR, PMF, XGBoost-SHAP, human health risk assessment (HRA), and Monte Carlo simulation (MCS) was employed to investigate HMs contamination and its associated health risks in surface water. The results indicated that overall water quality was generally satisfactory; however, arsenic (As) exhibited localized enrichment, reaching a maximum concentration of 96.9 μg/L. Three distinct pollution sources were identified: a geogenic sources (As, Hg); an integrated environmental dynamic source (Cd, Pb, Ni); and an anthropogenic activity source (Cr). Health risk assessments revealed that non-carcinogenic risks remained within acceptable thresholds, whereas the mean carcinogenic risk from As exceeded the guideline value of 1×10, affecting 32.23% of children and 29.38% of adults. Notably, probabilistic risk estimates were marginally lower than those deriv
Abstract
Accurate source apportionment of heavy metals (HMs) in surface water and the development of a source-based risk assessment system is essential for formulating effective pollution mitigation strategies. However, the combined contamination patterns of multiple HMs and the quantitative linkage between pollution sources and health risks remain inadequately understood, particularly in high-altitude regions. This study selected Gangba County, a representative transboundary headwater area located on the northern flank of the Himalayas in the southern Qinghai-Tibet Plateau, as the study area. An integrated framework incorporating APCS-MLR, PMF, XGBoost-SHAP, human health risk assessment (HRA), and Monte Carlo simulation (MCS) was employed to investigate HMs contamination and its associated health risks in surface water. The results indicated that overall water quality was generally satisfactory; however, arsenic (As) exhibited localized enrichment, reaching a maximum concentration of 96.9 μg/L. Three distinct pollution sources were identified: a geogenic sources (As, Hg); an integrated environmental dynamic source (Cd, Pb, Ni); and an anthropogenic activity source (Cr). Health risk assessments revealed that non-carcinogenic risks remained within acceptable thresholds, whereas the mean carcinogenic risk from As exceeded the guideline value of 1×10, affecting 32.23% of children and 29.38% of adults. Notably, probabilistic risk estimates were marginally lower than those derived from deterministic methods, suggesting that the latter may overestimate actual exposure risks. Natural geological processes were determined to be the predominant drivers of As accumulation. This study offers a scientific foundation for targeted management and mitigation of HMs pollution in high-altitude surface water environments.