Clustering analysis of a solid waste landfill using ERT and TDIP geophysical data.

Environmental contamination from unregulated landfills poses significant challenges, requiring accurate subsurface characterization. Non-invasive geophysical techniques like electrical resistivity tomography (ERT) and time domain induced polarization (TDIP) provide valuable insights but their interpretation often remains subjective. This study addresses the need for more objective and efficient approaches to identify and delineate contaminated zones within landfills. By integrating ERT and TDIP surveys with K-means clustering, we successfully partitioned the subsurface into five distinct, quantitatively characterizable clusters based on resistivity and phase parameters. This achieved an overall clustering matching rate of 87.1 %, while the chromium-containing sludge zones showed an accuracy of 84.1 %, a recall of 84.6 %, a precision of 88.0 %, and an F1 score of 86.3 % when validated against borehole data. Our results demonstrate that machine learning-enhanced geophysics can autonomously distinguish between domestic waste and chromium-containing sludge without prior labels, establishing clearer contaminant boundaries and reducing interpretation bias compared to traditional methods. This approach not only improves the reliability of contamination mapping but also offers a robust framework for data fusion in complex environmental settings. The methodology presents a powerful tool for landfill characterization and remediation guidance, with potential applications extending to various contaminated site investigations and environmental management strategies.