Hydrochemical assessment of groundwater with special emphasis on fluoride in parts of Punjab and fluoride prediction using GIS and ML

Abstract

The study focuses on assessing groundwater quality, with a special emphasis on fluoride contamination, in the Muktsar, Bathinda, and Moga of Punjab, India. Groundwater being a crucial resource for the region, faces contamination from both natural processes and anthropogenic activities. The study employs advanced techniques, including Geographic Information Systems (GIS) and machine learning models to predict fluoride contamination and assess the water quality index (WQI). The groundwater samples were systematically collected from 281 locations using GIS at approximately 5 km distance to ensure uniform distribution. The study aims to predict fluoride levels, various hydrochemical parameters and WQI to identify high-risk areas. Using Inverse Distance Weighting (IDW), the distribution of fluoride level and WQI was mapped, revealing varying concentrations across the study area. From the study, the Random Forest (RF) model identified key hydrochemical parameters influencing fluoride contamination. The RF model demonstrates high predictive accuracy for fluoride contamination, using the receiver operating characteristic (ROC) curves for validation and yield area under the curve (AUC) values of 82%, 81%, and 94% for Muktsar, Bathinda, and Moga districts, respectively. The novel integration of GIS with machine learning provides a robust framework offering valuable insights for water resource management. The results showed significant fluoride contamination in many areas, posing serious health risks like dental and skeletal fluorosis. The findings highlight the importance of addressing both natural and human-induced factors in managing groundwater quality, ensuring safe drinking water, and protecting public health in affected regions.