Analysis of geoelectric data through machine learning algorithms for waste leachate detection

Electrical resistivity tomography (ERT) is an effective method for detecting the leachate plume due to the plume's very low resistivity values. However, it is well-known that identifying contaminated areas in landfill sites based only on the distribution of electrical resistivity values is highly ambiguous, especially in the presence of clayey soils, given the low resistivity values that generally characterize both wet/saturated clays and contamination plumes. To overcome this problem, the ERT method is usually combined with the induced polarization method to derive useful information on leachate detection from the resistivity, chargeability, and ratio values. In this study, we developed a tentative methodology for leachate detection based on clustering analysis of geoelectrical data. The k-means algorithm was applied to perform a cluster analysis of the inverted resistivity and chargeability data acquired in a landfill site in the Campania region (southern Italy). This site is in a geological context characterized by silty-clayey deposits, with intercalations of graded sandstones from the Miocene age. Therefore, it represents a meaningful test bench for investigations integrating different geophysical datasets.