Health Risk Assessment of Physicochemical and Heavy Metal Exposures of Groundwater in Linggi River Basin, Malaysia with Machine Learning Approach

Abstract

Groundwater is an increasingly important water source in Malaysia due to climate variability, population growth, and rising demand. However, contamination by heavy metals poses significant risks to public health and water sustainability. This study aimed to assess the physicochemical quality of groundwater in the Linggi River Basin and evaluate health risks associated with heavy metal exposure. Eleven sampling wells were selected between 2022 and 2024, representing diverse land-use activities. Major heavy metals (Cd, Cu, Fe, Pb, Mn, Ni, Zn) were analyzed using ICP-MS and compared against World Health Organization standards. Statistical analyses and machine learning models, specifically Random Forest (RF) and Multiple Linear Regression (MLR), were applied to predict metal concentrations, while Shapley Additive Explanations (SHAP) identified key geochemical drivers. Health risks were quantified using Hazard Quotient (HQ), Hazard Index (HI), and Excess Cancer Risk (ECR) for adults and children via oral and dermal pathways. Results revealed severe contamination, with Fe (mean 164.53 mg/L) and Mn (mean 6.93 mg/L) far exceeding WHO limits, while Pb and Cd also surpassed permissible levels at several sites. RF outperformed MLR in predictive accuracy, and SHAP confirmed the influence of oxidation-reduction potential, bicarbonate, and temperature on metal mobility. Health risk assessment showed oral ingestion as the dominant exposure pathway, with children more vulnerable than adults. Cd emerged as the primary carcinogenic agent, contributing over 90% of cumulative cancer risk. Agricultural and mixed-use zones exhibited the highest risk indices, reflecting combined geogenic enrichment and anthropogenic inputs. In conclusion, this study highlights the urgent need for groundwater monitoring, pollution control, and child-focused health interventions in high-risk zones. By integrating machine learning, health risk assessment, and GIS mapping, the research provides a scalable framework for early detection, hotspot identification, and improved groundwater governance in Malaysia.