Urban aquifer health assessment and its management for sustainable water supply: an innovative approach using machine learning techniques

Abstract

Water scarcity is a threat to resilient and sustainable cities as well as the resident population's well-being. This research represents a machine learning (ML) and geospatial data coupling along with groundwater sustainability management strategy (GSMS) components in a stepped methodological approach for urban, peri-urban aquifer sustainability management in Vizianagaram extended city limit in Southern India. Random forest (RF) and support vector machine (SVM) are two ML models in combination with geospatial data (hydrogeological and geo-environmental controlling factors) coupling predicted probable groundwater occurrence by categorizing different groundwater potential zones (GWPZ). The prediction accuracy determined by the area under the receiver operation curve (AUROC) demonstrated accuracy of >80 %, with the RF model being the highest at 88.40 %. Integration of GWPZ with policy and legislation, stakeholder involvement, sensitization, and linkage, along with strengths, weaknesses, opportunities, and threats (SWOT) analysis in the form of GSMS, can help policymakers, planners, and government authorities achieve sustainable measures near to reality and future adaptability.