Source apportionment, drinking water quality prediction and health risk appraisal of groundwater nitrate using hydrochemistry, machine learning and Monte Carlo simulation - A case study from the Suruliyar River basin, South India

The main objective of this study is to evaluate the nitrate contamination in the groundwater of the Suruliyar River basin using modern technologies such as machine learning (ML) and Monte Carlo Simulation (MCS) for a cost-effective and time-efficient assessment of areas at risk. In the basin, nearly 152 samples collected in both seasons (pre-monsoon (PRM), (76) and post-monsoon (POM), (76) seasons) with an area extent of 1332 km² and 1486 km² of land area show nitrate level exceeding 45 ppm during the PRM and POM seasons. The source for this contamination is the unregulated use of fertilizers, with minor contributions from sewage and livestock waste. The Nitrate Pollution Index (NPI) indicates that, 898 km² and 1326 km² sampled areas fall under significant to very significant pollution categories during the PRM and POM seasons. Machine learning predictions of nitrate level were most accurately predicted using the Support Vector Machine (SVM) model, which achieved accuracies of 87.50% and 81.25% in the PRM and POM seasons. Traditional health risk assessment reveals that 83% and 89% of samples pose risk to children, by Hazard Quotient (HQ) > 1 during the PRM and POM seasons. The MCS results further support this finding, showing maximum 95th percentile HQ values of 5.5510 and 7.4938 for children in the respective seasons, confirming their higher vulnerability to nitrate contamination compared to other age groups. This research provides critical insights that can support policymakers and authorities in implementing measures to reduce nitrate pollution and its health complications, to guarantee the Sustainable Development Goal (SDG) 3 and 6 for the sustainable development.