Mapping the spatial distribution and temporal variations of a leachate plume in a landfill based on a geoelectrical survey and using machine learning tools
Geophysical prospecting methods provide complete characterizations of contaminated sites, unlike the fragmented information obtained from conventional invasive and costly borehole drilling practices.
In an early collaboration between ATOMS Laboratory and Humber Institute of Technology and Advanced Learning, a geoelectrical study was combined with measurements of surface methane emissions for a closed landfill in southern Ontario. The resulting 2D and 3D tomographic images of the subsoil show clear contrasts of resistivity that could be used to monitor the transport and accumulation of groundwater contaminants. In addition, the possible coupling between methane emissions and leachate’s geoelectrical proxies was determined via optimization tools, giving rise to a statistically significant inference that describes the causal relationship between the shared effects of the multiple parameters in the waste stabilization zone and methane concentrations in the biogas. A deterministic approach was also tested in this contaminated site to check such a relationship using process modeling and simulation.
Combining Machine learning, geophysical and geochemical environmental data to determine levels of carcinogenic compounds in soil and water for a community affected by heavy industrial activities
In this same line of research, the ATOMS Laboratory has initiated another collaboration with Humber College (Prof. Maria Jacome), Fort William First Nation (FWFN), and U of T Departments of Human Biology and Indigenous Studies (Prof. Melanie Jeffrey), Applied Chemistry and Chemical Engineering (Prof. Daniela Galatro), and Industrial and Mechanical Engineering (Prof. Jason Bazylak). This cooperative venture funded by CIHR and NSERC grants addresses some health concerns of a First Nation community in northern Ontario by integrating near-surface geophysical methods (geoelectrical, ground penetrating radar/GPR, magnetic analysis of soil samples) and machine learning (ML) tools with the information contained in a series of Environmental Site Assessment (ESA) reports (geochemical data). One of the main goals of this multidisciplinary partnership is to characterize levels of hazardous compounds in soil, ground, and surface water around a residential area encircled by heavy industries that act as point sources of contamination.
In addition to precisely describing and monitoring the reach and evolution of a leachate plume, geoelectrical surveys can also be applied to detect resistivity contrasts linked to biodegradation induced by carcinogenic Petroleum Hydrocarbons (PHC), Polycyclic aromatic hydrocarbons (PAH), and volatile organic compounds (VOC). The high-resolution Ground-penetrating radar (GPR) technique complements the geoelectrical surveys by identifying low impedance contrasts that may be associated with the location of a contaminant plume. GPR is also used to determine the position of subsurface utilities, underground storage tanks, and associated leaks of organic compounds. Besides, the magnetic properties in contaminated soils can be linked to the presence of hydrocarbons due to the chemical alteration they induce on primary Fe-oxides/sulfides.
The existing ESA reports for the area of concern, containing analysis of soil and water metals, general chemistry, PAHs and PHCs, herbicides, pesticides, and VOCs, are presented in a way that is not amenable to decision-making regarding land use planning. Therefore, an essential part of this research depends on appropriately managing all this information in an accessible dataset. Exploratory data analysis in collaboration with community partners is an important initial step to maximize data insights defined by their structure and outliers, classification, and suggested models that correlate variables.
After obtaining the data description that provides the basis for the model formulation and predictive capabilities, the interpretation of the multiple layers of geophysical and geochemical information for the different ground media (i.e., soil and water) is quantified through the development and application of methods such as a Gaussian Mixture Model classifier, Learning Vector Quantization Neural Networks, Principal Component, and Factor Analysis. Thus, it is possible to generate maps of integrated environmental indicators that define distinct contaminated zones. These maps communicate relevant information and better visualize and rank the polluting processes in a complex setting affected by numerous sources of contamination. Identifying which sources significantly impact the area of concern is a key piece of evidence for further causality studies linking specific contaminants with cancer types overrepresented in the community.
Together with making pollutant profiles visible and manageable, this research has the added value of using a mixed methodology that goes beyond a pure engineering approach bringing scientific and community members together to work in a culturally grounded Integrated Translation Knowledge framework. The results of this kind of study are meant to build local and Indigenous capacity to monitor and remediate lands, enabling community leaders to develop a regulatory framework based on Indigenous values for land use and to make evidence-based decisions to protect the health of current and future generations.