Hydrogeochemical evaluation and corresponding health risk from elevated arsenic and fluoride contamination in recurrent coastal multi-aquifers of eastern India

Abstract

Increasing groundwater pollution through toxic contamination is the primary concern for severe environmental and health hazards across the world. Groundwater resources are more vulnerable to toxic contamination compared to surface water because they require a long time to restore. Groundwater vulnerability and risk assessment is a challenging task among the researchers responsible for quality monitoring and management of groundwater. In this regard, predictive modeling for groundwater vulnerability is essential to defend the vulnerability of groundwater resources in current scenario. Contamination of groundwater with arsenic (As) and fluoride (F‾) poses a threat to human health in the Bengal delta region. This study developed a framework approach to assess groundwater vulnerability and related health hazard risks in the South 24 Parganas district of West Bengal, India. For modeling groundwater vulnerability, the Logistic regression (LR) method coupled with fifteen hydrogeochemical parameters was used for modeling. The health hazard risk of this study region was assessed using the human health hazard index. Furthermore, hydrogeochemical properties and groundwater quality were assessed through Piper, USSL, and Wilcox's diagram accordingly. The study revealed that F‾, depth, PO42− and SO42− are the highest controlling factors of groundwater vulnerability in this area. It is also found that because of the presence of alkaline organisms in groundwater, it is unfit for farming and drinking purposes as well. The evaluation matrices used in this study revealed that LR gives optimal prediction results (AUC-ROC is 0.891 and 0.861, kappa index is 0.810 and 0.720 in training and validation, respectively) in groundwater vulnerability. The applied approach in this study may be useful to other regions for predicting groundwater vulnerability and for the sustainable adoption of several management strategies. In the future, deep learning as well as other sophisticated learning algorithms should be useful for optimal prediction of groundwater resources.