Soft and hard clustering of geoelectrical data for detection of leachate accumulation zones in municipal solid waste landfills

Abstract

Uncontrolled accumulation of leachate in municipal solid waste landfills can represent a potential hazard in municipal solid waste (MSW) landfills, since it is a liquid with a high pollutant content that can contaminate aquifers. Electrical resistivity tomography (ERT) in combination with induced polarization (IP) method can be highly effective for detecting areas with presence of leachate, given the electrical properties of leachate (highly conductive and chargeable) compared to the unsaturated waste mass and the surroundings areas (covering and bottom layers). However, the geoelectrical reconstruction expressed in terms of individual geophysical quantities (resistivity and chargeability), leaves room for ambiguities in such a complex scenario that involves several resistive-conductive transitions. In this study, we use a machine learning-based approach to integrate ERT and IP tomographic data for improving the accuracy in the detection of leachate accumulation zones. After geophysical data inversion, we perform the cluster analyses on the inverted models (resistivity, chargeability and normalized chargeability) by using both hard (K-Means) and soft (Fuzzy C-Means) clustering algorithms. We finally achieve a comprehensive integrated model of the landfill with also an assessment of the reliability of the reconstruction through computation of validation indices and, only for soft clustering, of the uncertainty by means of the membership function. We apply the proposed procedure to synthetic examples and to a real-world investigation of a MSW landfill located in Central Italy. We firstly carry out synthetic examples to test the efficiency of hard and soft clustering in imaging leachate accumulation zones, by developing multi-layer models simulating a typical MSW geometry. Then, the results of the application of the proposed approach to the real-world example show an improvement of the accuracy in the detection of leachate accumulation zones as well as an assessment of the uncertainty both in the lateral and vertical directions. Therefore, this approach can be an effective and fast tool in real cases where leachate pollution detection is required at short time intervals for waste management.