Empowered machine learning algorithm to identify sustainable groundwater potential zone map in Jashore District, Bangladesh

Abstract

Groundwater is a crucial natural resource that supplies potable water to the local population in Jashore, Bangladesh. In recent years, the district has experienced groundwater scarcity during the summer season, significantly affecting both domestic and agricultural usage, thereby impacting the local populace. However, accurately assessing its availability and quality remains a challenge. Our objective was to identify groundwater prospect zones in Jashore and provide valuable insights for management and decision-making through Explainable Artificial Intelligence (XAI). This study used 15 groundwater conditioning factors for potential zone mapping. The k-Nearest Neighbors (KNN), Neural Networks (NNET), Gradient Boosting Machines (GBM), Random Forest (RF), and Support Vector Machines (SVM) models were applied to assess groundwater suitability, resulting in the identification of five distinct areas with different levels of potential. The areas with very low suitability covered 39.37, 14.95, 107.23, 291.59, and 300.24 km2, while areas with very high suitability covered 1255.61, 969.49, 1221.99, 1958.85, and 1211.03 km2, respectively. Notably, the RF and NNET models exhibited higher accuracy, with Area Under Curve (AUC) values of 90%. The KNN, GBM, and SVM models achieved AUC values of 85%, 86%, and 88% respectively. The study reveals that the northeastern region has the greatest groundwater potential, while the southern and southwestern regions have lower capacities. The researcher's also utilized Partial Dependence Plots (PDP) technique, to analyze parameter behavior in predicting groundwater suitability. This approach provided valuable insights into the impact of various factors on groundwater potential, aiding management decisions. The study findings inform to policymakers and water resource managers in Jashore, enabling sustainable groundwater management by identifying high-potential areas. This knowledge optimizes resource allocation, ensuring sufficient supply for sustainable development.