Supervised Classification of Groundwater Potential Mapping Using Integrated Machine Learning and GIS-Based Techniques

Abstract

Addressing the global water depletion challenge, this study integrates five supervised machine learning algorithms (MLAs) with GIS-based techniques to assess groundwater potential. The employed MLAs include Ensemble Boosted Trees (logic-based learners), Naive Bayes (NB; statistical learning algorithms), Support Vector Machines (SVM), Multi-Layer Perceptron (MLP; Artificial Neural Networks), and k-Nearest Neighbors (kNN; instance-based learners). These MLAs were utilized to generate groundwater potential maps (GPMs) based on seven influential variables: aquifer unit types, transmissivity, lineament density, slope, soil type, land use/land cover, and drainage density. Classifier performance was evaluated using metrics such as True Positive Rates (TPR), False Negative Rates (FNR), Positive Predictive Values (PPV), False Discovery Rates (FDR), and the Area Under the Curve (AUC) of Receiver Operating Characteristic (ROC) curves. Results indicate that kNN-based learners outperformed other methods, achieving a validation accuracy of 90.70% and an AUC of 1, which corresponds to 100% accurate predictions. Ensemble Boosted Trees, MLP, SVM, and NB followed, with validation accuracies of 89.7%, 79.4%, 77.6%, and 75.7%, respectively. The methodology developed in this study can be applied to estimate and manage potential groundwater resources in regions facing water scarcity issues.