The concept of mapping aquifer vulnerability to pollution dates back to the late 1960s, when French researchers first attempted to cartograph the inherent susceptibility of groundwater bodies to contamination. But the modern era of AVA truly began in 1987, when the United States Environmental Protection Agency standardized the DRASTIC method—a weighted index model whose acronym spells out its seven hydrogeological parameters: Depth to water table, net Recharge, Aquifer media, Soil media, Topography, Impact of the vadose zone, and hydraulic Conductivity. Each parameter is rated and weighted, and the resulting vulnerability index is mapped across a study area using geographic information systems (GIS). Six years later, in 2001, the GALDIT method emerged as the coastal counterpart, tailored specifically to seawater intrusion. Its six indicators—Groundwater occurrence, Aquifer hydraulic conductivity, Level of groundwater above sea level, Distance from shore, Impact of existing status of seawater intrusion, and Thickness of the aquifer being mapped—capture the physics of the freshwater-saltwater interface, a boundary that rising seas and over-pumping are steadily pushing landward.

The weakness of both classic frameworks, the review explains, lies in their subjectivity. The ratings and weights assigned to each parameter were originally fixed by expert judgment, meaning two analysts applying the same method to the same aquifer could produce noticeably different vulnerability maps. Over the following decades, researchers therefore attacked the problem from several complementary directions. Statistical techniques allowed weights and ratings to be calibrated against real-world contamination data, most commonly nitrate concentrations in agricultural regions or chloride levels in coastal aquifers. Logistic regression, frequency ratio, and single-parameter sensitivity analysis emerged as workhorses of this approach, quantifying which parameters actually drive contamination risk in a given setting rather than relying on generic literature values.

Multicriteria decision-making (MCDM) methods constitute a second major branch of improvement. Among these, the analytic hierarchy process (AHP) has become the most widely used tool for modifying DRASTIC indices. AHP structures expert knowledge into a hierarchy of pairwise comparisons, from which mathematically consistent weights are derived; extensions such as fuzzy AHP, SWARA (stepwise weight assessment ratio analysis), and ordered weighted averaging (OWA) further refine how parameter importance and uncertainty are handled. According to the review, AHP, genetic algorithms, and fuzzy logic are the three techniques most frequently applied to DRASTIC. Entropy weighting—derived from the information content of the data itself rather than expert opinion—has been paired with these subjective methods to create hybrid objective-subjective weighting schemes that outperform either approach alone.

Optimization algorithms form the third pillar. Genetic algorithms, which evolve candidate solutions through selection, crossover, and mutation, have proven especially effective at tuning DRASTIC weights against observed contamination records. The review documents a rich ecosystem of nature-inspired metaheuristics now deployed in this domain: particle swarm optimization, differential evolution, invasive weed optimization, firefly algorithms, ant colony optimization, and biogeography-based optimization have all been used to search the vast space of possible weight combinations for the configuration that best reproduces measured vulnerability patterns. For GALDIT, the dominant optimization techniques identified are single-parameter sensitivity analysis, genetic algorithms, and support vector machines. Comparative studies of multiple metaheuristics applied to the same coastal aquifers have shown that even modest weight optimization can substantially sharpen the agreement between predicted and observed seawater intrusion, and optimized vulnerability maps have in turn been used to design groundwater quality monitoring networks and delineate capture zones.

The fourth and arguably most transformative branch is machine learning and, more recently, deep learning. Supervised committee machines that combine fuzzy logic models, artificial neural networks trained on nitrate or chloride observations, support vector machines and support vector regression, random forests, decision trees, and adaptive neuro-fuzzy inference systems (ANFIS) have all been coupled with DRASTIC or GALDIT frameworks. These hybrid architectures allow the underlying index model to supply physically meaningful input parameters while the learning algorithm discovers the nonlinear relationships between those parameters and actual contamination. Bayesian model averaging has been used to quantify and propagate the uncertainty inherent in combining multiple models, and ensemble approaches—bagging several optimized DRASTIC frameworks, for instance—have proven robust in delimiting contamination-prone zones. At the frontier, deep learning neural networks coupled with optimization algorithms and convolutional neural networks formulated for spatial mapping are now producing aquifer vulnerability maps that learn directly from the spatial structure of the data, with studies reporting clear gains in predictive accuracy over classical overlay methods.

The review also catalogues a proliferation of purpose-built alternatives and refinements. DRASTIC has spawned DRASTIC-LU (adding land use), DRASTICA (adding anthropogenic impacts), and numerous regional variants validated against nitrate and sulfate concentrations. GALDIT has been extended into GALDIT-SUSI, which accounts for surface water bodies in seawater intrusion assessment, and GALDIT-NUTS, designed for island aquifers. Karst systems—where water can traverse fissured carbonate rock in hours rather than years—have inspired a parallel family of methods including EPIK, PI, RISKE, KARSTIC, REKS, PaPRIKa, and the PRESK and DRISTPI frameworks, all reflecting the “European approach” that considers origin of infiltration, pathway properties, and protective cover. Newer schemes such as RIVA (a statistically grounded intrinsic vulnerability method) and IKAV (an integrated karst vulnerability approach tested in Mexico’s Yucatán) demonstrate that methodological innovation remains very much alive across every aquifer type.

Why does this thirty-year technical evolution matter now? The authors anchor their analysis firmly in the era of climate change. Global warming is reshaping recharge patterns, intensifying droughts that force aquifer over-extraction, and raising sea levels that drive saltwater into freshwater lenses in densely populated coastal regions from the Mediterranean to South and East Asia. As surface water becomes less reliable, the demand for groundwater is climbing, which means the consequences of contaminating a once-vulnerable aquifer are becoming more severe and less reversible. Vulnerability maps are the practical tool through which land-use planners, water managers, and protection-zone designers decide where agriculture, industry, and urban development can safely coexist with drinking water supplies. If those maps are built on arbitrary weights, the resulting protection strategies inherit that arbitrariness; if they are calibrated, optimized, and validated with modern statistical and machine learning tools, they become defensible instruments of policy.

The review’s conclusions offer both a scorecard and a roadmap. For DRASTIC-based studies, AHP, genetic algorithms, and fuzzy logic dominate the toolkit; for GALDIT-based coastal assessments, single-parameter sensitivity analysis, genetic algorithms, and support vector machines lead. Hybrid frameworks—combining statistical validation, MCDM weighting, metaheuristic optimization, and machine learning—consistently outperform any single technique, and the trend toward ensembles, uncertainty quantification, and deep learning appears set to continue. The authors emphasize that future work should integrate climate projections directly into vulnerability assessment, so that maps reflect not only present-day hydrogeology but also the shifting recharge regimes, sea levels, and land-use trajectories of the coming decades. The research was supported by the Basic Research Laboratory Program of the National Research Foundation of Korea, funded by the Ministry of Science and ICT.

For the millions of communities that depend on groundwater every day, the message is clear: the invisible infrastructure beneath our feet can be protected, but only if the maps guiding that protection are as rigorous and adaptive as the pressures threatening it. In a warming world, the science of aquifer vulnerability has moved from static expert judgment toward dynamic, data-driven, and increasingly intelligent assessment—and this review captures, in a single synthesis, how far that transformation has come and where it must go next.