Mapping Redox Conditions and pH in Indian Drinking Water System Using Machine Learning Models

1. Introduction

Comprehending the regional occurrence of contaminants in aquifers is vital for efficient groundwater resource management and assessing the risk to drinking-water supplies. The spatial distribution of various contaminants affecting human health and ecological well-being relies on the geochemical conditions within aquifers, specifically pH and redox conditions. Despite being fundamental groundwater characteristics, pH and redox conditions are often excluded from the evaluation of groundwater quality for vulnerability or risk assessments due to a lack of sufficient information regarding their spatial distribution within aquifers. The focal point of this study stems from the escalating concerns regarding water pollution, particularly the surge in nitrate, magnesium (Xu, W., & Su, X., 2019), and lead levels in groundwater, particularly in the Indian region. The objective is to anticipate safe water conditions by evaluating redox conditions and pH levels in groundwater. Elevated pH levels (M.Larocque et al., 2019) contribute to heightened concentrations of calcium and magnesium carbonates in pipelines. Although these levels may not directly pose health hazards, they can lead to skin issues like dryness, itchiness, and irritation. Conversely, a decrease in pH intensifies water acidity, rendering it toxic. By knowing the redox state of ground water, we can say that whether the water is toxic or anoxic and also helps us to determine whether it contains elevated levels of many contaminants including arsenic, nitrate and even some man made contaminants. The concentrations of arsenic and manganese are more likely to be present at levels that exceed human health bench marks in anoxic ground water, and concentrations of uranium, selenium and nitrate are more likely to exceed this benchmarks in toxic groundwater. As stated in (S.M’nassri et al., 2022) the redox conditions of the ground water is an important factor in predicting the contaminants and constituents might be present in groundwater at levels of concern for human health. In this proposed work, gradient boosted regression tree has been applied as the main machine learning model for the application. The analysis of the north eastern U.S.A aquifers are collected for the data. The application of many of machine learning (Ryuji Sakata et al., 2018) have been applied to compare the efficiency of all models. Hence, it was finalized to visualize the coastal mapping of north-eastern U.S.A has been shown in results.