Application of machine learning algorithms for seasonal and annual perchlorate risks in groundwater of the Arjunanadi River Basin (India): drinking water quality assessment and human vulnerability

Machine learning algorithms were used to analyze and predict the seasonal perchlorate content in groundwater of the Arjunanadi River Basin (ARB) in South India and associated health risks. The CaMgCl type of majority samples reflected secondary salinity from prolonged rock-water interaction and prominent anthropogenic influence. Perchlorate concentrations ranged from 0.001 to 0.21 ppm. In post-monsoon season (POMS) 2023, 24.4% of the groundwater samples were classified as at risk and 8.5% as high risk, with corresponding area mapped to an extent of 280 and 80 km2, respectively. Similarly, in pre-monsoon season (PRMS) 2024, 22.5% and 1.0% of the samples were classified as at risk and high risk with corresponding area mapped to an extent of 244 and 2.5 km2. In POMS 2024, 29.8% and 8.5% of the samples were classified as at risk and high risk with corresponding area mapped to an extent of 274.5 and 46 km2. In PRMS 2025, 23.4% and 3.2% of the samples were classified as at risk and high risk with corresponding area mapped to an extent of 242 and 34 km2. Among all the machine learning algorithms, Random Forest (RF) showed the moderate performance with a mean accuracy of 71.5% compared to others. Health hazard assessment indicated that children are more vulnerable to perchlorate exposure, with 72% (POMS 2023), 67% (PRMS 2024), 87% (POMS 2024), and 80% (PRMS 2025) of samples showing a total hazard index (THI) >1. Additional managed aquifer recharge (MAR) structures in the higher concentration area away from the pre-existing check dams would be helpful to raise the static level and improve the water quality. This study provided valuable understanding on perchlorate levels among the seasons and a trend of health risk over a year with a view to implement effective and safe water management practices under the sustainable development goals (SDG) of 3 and 6.