PREDICTIVE MODELING FOR WATER QUALITY ASSESSMENT: INTEGRATING MACHINE LEARNING WITH LULC, GROUNDWATER LEVELS, RAINFALL DISTRIBUTION, AND HEALTH IMPACTS

Abstract

To guarantee the safety of drinking water sources in the future, it is crucial to develop a comprehensive understanding of the current state of groundwater quality and pollution levels. Accurate water quality prediction is pivotal in effectively controlling water pollution and enhancing water management practices. This study focused on predicting groundwater quality for drinking and irrigation purposes in Madhya Pradesh, a central state in India. The dataset used in the study comprised groundwater samples of monitoring stations across 52 districts of Madhya Pradesh from May 2000 to May 2018. Various machine learning models, including Deep Neural Network (DNN), Artificial Neural Network (ANN), Random Forest (RF), and XGBoost, were used to predict different water quality indices such as Sodium Absorption Ratio (SAR), Soluble Sodium Percentage (SSP), Kelly's Ratio (KR), Permeability Index (PI), Magnesium Hazard (MH), and Water Quality Index (WQI). The performance of these models was compared using minimal input variables. The results showed that DNN and SAR best predicted WQI, while ANN achieved the highest accuracy for SSP, KR and PI predictions. RF and XGBoost also yielded satisfactory results for all the indices. Furthermore, the study assessed the impact of nitrate and fluoride contamination on human health through a risk assessment. The findings revealed that infants were at the highest risk, followed by children, adult females, and adult males, emphasizing the need for addressing nitrate and fluoride contamination in the study area. Additionally, the study examined the influence of land use/land cover, rainfall, and groundwater level on groundwater quality. The land use/land cover classification revealed various classes, including forest, cropland, built-up land, range land, flooded vegetation, and water bodies. The results showed that cropland had the most deteriorated groundwater quality due to excessive usage of fertilizers and pesticides. Built-up areas and water bodies also exhibited poor groundwater quality due to improper wastewater treatment and disposal. The study also analyzed the relationship between groundwater quality and normal rainfall distribution, indicating that lower rainfall was associated with poorer groundwater quality. The analysis revealed that as the groundwater depth increases, the proportion of poor to unsuitable water quality also increases. In conclusion, this study provides insights into the prediction of groundwater quality using machine learning models and highlights the risks posed by nitrate and fluoride contamination on human health. The findings underscore the importance of addressing land use practices, rainfall patterns, and groundwater management strategies in Madhya Pradesh to ensure sustainable groundwater resources.