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

Area of Science:

• Environmental Geophysics
• Machine Learning Applications
• Geophysical Data Fusion

Background:

• Unregulated landfills present environmental contamination challenges requiring precise subsurface characterization.
• Non-invasive geophysical methods like electrical resistivity tomography (ERT) and time domain induced polarization (TDIP) offer insights but often suffer from subjective interpretation.
• Objective and efficient methods are needed for identifying and delineating contaminated zones within landfills.

Purpose of the Study:

• To develop and validate a machine learning-enhanced geophysical approach for objective subsurface characterization of landfills.
• To autonomously distinguish between different waste types, specifically domestic waste and chromium-containing sludge, without prior labeling.
• To improve the accuracy and reliability of contaminant boundary delineation and reduce interpretation bias in landfill investigations.

Main Methods:

• Integration of electrical resistivity tomography (ERT) and time domain induced polarization (TDIP) geophysical surveys.
• Application of K-means clustering algorithm to partition subsurface data into distinct clusters based on resistivity and phase parameters.
• Validation of clustering results against borehole data to assess accuracy, recall, precision, and F1 score.

Main Results:

• Successful partitioning of the subsurface into five distinct, quantitatively characterizable clusters.
• Achieved an overall clustering matching rate of 87.1% when validated against borehole data.
• Demonstrated high performance for chromium-containing sludge zones with an accuracy of 84.1%, recall of 84.6%, precision of 88.0%, and F1 score of 86.3%.

Conclusions:

• Machine learning-enhanced geophysics can autonomously distinguish between domestic waste and chromium-containing sludge, reducing interpretation bias.
• The integrated ERT, TDIP, and K-means clustering approach provides a reliable framework for landfill characterization and contamination mapping.
• This methodology offers a powerful tool for environmental management, with potential applications in various contaminated site investigations.