Delineation of groundwater potential zone mapping using machine learning algorithms in industrial zone

Groundwater, an essential natural resource, comprises approximately 34% of the planet’s total freshwater supply. Since the last century, factors such as over-exploitation, rapid urbanization, industrial contamination, agricultural activities, and residential pressures on groundwater reserves have escalated the global demand for freshwater. In Bangladesh, a densely populated developing country, water pollution, scarcity, and safety concerns are particularly severe, with Dhaka and its surrounding urban fringes undergoing rapid groundwater depletion. Recently, machine learning models have gained prominence in groundwater potential zone (GWPZ) mapping due to their ability to model complex hydrogeological relationships. This study develops a GWPZ map for Gazipur, a major industrial district neighboring Dhaka, by applying geospatial techniques and ensemble machine learning algorithms to integrate groundwater level data with twenty selected topographical, hydrological, meteorological, soil, geological, and land surface factors. Models employed include Logistic Regression, K-Nearest Neighbor, Random Forest, Support Vector Machine, Multilayer Perceptron, and a Voting Classifier. Model performance was evaluated using accuracy, precision, recall, F1-score, ROC–AUC, and McNemar’s test. Random Forest achieved the highest predictive performance (Test AUC = 0.911), followed by the Voting Classifier (Test AUC = 0.909). Results indicate that nearly half of Gazipur falls within low groundwater potential zones concentrated in Kapasia and Kaliakoir upazilas, while moderate to very high potential zones are predominantly situated in the southern part of the district. The comparative evaluation demonstrates the effectiveness of ensemble machine learning for groundwater potential mapping and provides practical guidance for sustainable groundwater management and informed decision-making.