HYDRO-GEOTECHNICAL MAPPING OF MANAGED AQUIFER RECHARGE SITES: INTEGRATING GIS AND MACHINE LEARNING FOR ENHANCED GROUNDWATER REMEDIATION STRATEGIES

Abstract

Recharging depleted aquifers, reducing degradation, and increasing storage are all possible outcomes of MAR. There are a number of factors that determine the efficiency of MAR, including site selection, hydrogeological conditions, soil and subsurface features, source water supply, contaminant risk, land use, and the hydraulic response of the receiving aquifer. It is possible that regional site selection approaches that make use of a single hydrogeological parameter or multi-criteria decision analysis based on expert opinion could be effective; however, these methods are unable to take into account the nonlinear interactions that occur between environmental and hydro-geotechnical elements. An integrated hydro-geotechnical mapping framework combines geographic information systems (GIS), remote sensing, hydrogeological characterisation, geotechnical research, machine learning (ML), and groundwater-quality assessment in order to identify, rank, and validate MAR sites. It is proposed that lithology, soil texture, hydraulic conductivity, infiltration capacity, aquifer thickness, depth to groundwater, recharge potential, slope, drainage density, lineament density, land use/cover, electrical conductivity, ground water quality, contamination vulnerability, source water quality, and infrastructure proximity all be taken into consideration. GIS has the ability to provide spatial information structure, harmonise data, spatially analyse data, map constraints, and visualise data. On the other hand, machine learning models such as Random Forest, Support Vector Machine, Gradient Boosting, Extreme Gradient Boosting, and Artificial Neural Networks have the capability to model non-linear relationships between environmental predictors and groundwater-response indicators. A recent study has demonstrated that machine learning has the potential to enhance groundwater mapping prediction provided it is taught and evaluated independently. In addition to the accuracy of the models, the technique that has been suggested takes into account the quality of the data, cross-validation, uncertainty analysis, explainable machine learning, and field verification in order to take into account the hydrogeological plausibility and the uncertainty of the models. Following a transparent and open preliminary screening procedure, a decision framework that is based on GPS makes use of machine learning in order to significantly enhance spatial predictions. Additionally, the model takes into account the attenuation of groundwater contaminants, the compatibility of source water, the duration of transit, the storage of aquifers, the risk of clogging, and the efficiency of recovery. In order to select MAR sites, develop experimental recharge methods, and devise an adaptive groundwater management strategy, water resource planners can make use of hydro-geotechnical suitability maps. Those metropolitan areas that are experiencing groundwater contamination and depletion can benefit from the technique described above.