<p>Groundwater contamination is difficult to characterize at sites where hydrogeochemical information is derived from sparse monitoring-well networks. This study presents a framework that integrates groundwater-quality records with electrical resistivity and Induced Polarization (IP) data to better understand landfill-derived leachate plume behavior. The methodological approach is applied to an upper aquifer downgradient of a bark dump and its contamination attenuation zone, where pollution is linked to ongoing health issues in an adjacent Canadian Indigenous community. Stacked Pollution Indices (SPI) are evaluated alongside unsupervised machine-learning methods (i.e., Gaussian Mixture Models, GMM; Hierarchical Agglomerative Clustering, HAC; and Self-Organizing Maps combined with K-means, SOM + K-means) to integrate and classify hydrogeochemical variables into contamination-related facies. Ordinary kriging is used to interpolate these variables between monitoring wells, generating a synthetic dataset for SOM training and improving spatial coverage across the site. Overlap in SPI ranges across GMM, HAC, and SOM + K-means clusters indicates that aggregated pollution metrics may mask intrinsic relationships among water-quality variables. Comparing 7-m depth slices through resistivity and IP cubes, corresponding to the monitoring-well screening interval in the upper aquifer, with maps of electrical conductivity, SPI, and SOM-derived facies yields a more coherent interpretation of plume geometry, migration fronts, and compositional heterogeneity than either information source could provide independently. By integrating unsupervised hydrogeochemical clustering with geoelectrical imaging, this study establishes a refined hydrogeochemical–geophysical facies zonation of the upper aquifer and leachate plume, offering a practical basis for environmental monitoring, risk assessment, and remediation planning in contaminated systems constrained by limited resources and sparse sampling.</p>