Using machine learning, positive matrix factorization and kernel density estimation to assess source-oriented health risk of groundwater from a coal-mining area of Southwestern China

A typical coal mining karst area in southwest China. Groundwater resources face significant threats due to the accumulation of major and trace elements in aquifers in resource-rich yet environmentally vulnerable coal mining regions. The study aims to determine natural background levels, identify multiple contamination sources, and assess human health risks associated with these elements to support effective groundwater protection strategies in such areas. In the study, 22 groundwater samples were collected from a representative area of coal mining. Ca-HCO 3 was identified as the dominant hydrochemical type. Major and trace elements were analyzed using Natural Background Levels (NBLs) estimation, Positive Matrix Factorization (PMF), and a human health risk model. Preliminary results indicated that the groundwater in the study area was predominantly alkaline, with NO₃⁻ as the primary ion, of which 27.27 % of samples exceeded regulatory standards. Elements such as K + , HCO 3 – , NO 3 – , H 2 SiO 3 , and Sr appeared sensitive to both natural processes and anthropogenic activities, as reflected by their relatively high mean/TV ratios (58.3 %, 59.99 %, 74.5 % and 70.4 % respectively). Groundwater chemistry was primarily influenced by natural geochemical processes, including karstification (about 34.96 %), dissolution of evaporites (3.42 %) and silicates (55.12 %), and the presence of trace element-bearing minerals (e.g., 18.06 % Li and 2.38 % Fe), as well as human-induced contamination from mining operations (27.52 %). Health risk assessment revealed that children may be more vulnerable to health risks than adults, with 22.7 % of groundwater samples showing Hazard Index (HI) values exceeding 1 for children. NO 3 – posed a significant threat to human health, as 18.1 % of the sampling sites had HQ values exceeding 1. And populations relying on groundwater in the central and northeastern parts of the study area should be prioritized for risk mitigation. These findings should be regarded as exploratory, given the limited sample size and single sampling campaign, but they nonetheless provide useful insights into groundwater protection and offer a methodological reference for similar regions. • Groundwater natural background levels of K⁺, HCO 3 − , NO 3 ⁻, H 2 SiO 3 , and Sr are raised. • Six sources were identified for major and trace elements by CA, PCA, and PMF. • Risk assessments showed that nitrate poses a serious threat to children’s health. • Monte Carlo simulation using KDE method simulates data characteristics of elements.