Machine learning based evaluation of fluoride contaminated groundwater and health risks in the Pambar River basin, South India

Excess fluoride in groundwater has become a significant health issue across several regions globally, yet comprehensive studies integrating seasonal variation, human health implications, and predictive modeling remain limited. This study examines fluoride ranges and adverse health effects in groundwater of Pambar River basin, South India, employing machine-learning techniques to classify groundwater based on its aptness for consumption. Totally, 100 groundwater samples were collected during North-East monsoon 2024, Post-monsoon, Pre-Monsoon and South-West Monsoon of 2025. Hydrochemical investigation indicated predominantly alkaline water, with sodium and bicarbonate as dominant ions. The overall mean fluoride concentration, calculated from four seasonal datasets, was 1.38 mg/L, and 43% of samples exceeded the World Health Organization guideline of 1.5 mg/L. Correlation analysis revealed calcium negatively influenced fluoride levels due to Calcium-Fluoride precipitation. Principal Component Analysis explained 64.9%-68.3% of variance, highlighting geochemical processes as primary controls, with secondary influence from agricultural runoff and waste leaching. Entropy-based water quality evaluation revealed 36.75% of samples were safe for drinking, while 58.75% required treatment. Among machine-learning models, support vector machines achieved the best predictive performance, with random forest and extreme gradient boosting also performing well under limited seasonal datasets. Evaluation of Human health hazard indicated potential fluoride-related risks, particularly for children (50%), teens (45%), women (44%), and men (43%). These findings provide a baseline for future groundwater management and underscores the importance of implementing sustainable measures to mitigate fluoride pollution in Pambar's groundwater resources.