Soil and Water Pollution

3D Geophysical Mapping of Contaminated Groundwater

Non-invasive geophysical surveys are a well-established and effective approach for characterizing groundwater contamination beyond the spatial limitations of conventional borehole monitoring.
Resistivity and induced polarization (IP) geoelectrical methods are widely used to generate tomographic images that regionally delineate the geometry, depth, and continuity of contaminant plumes in complex subsurface environments.
Our geophysical work draws on applications at two closed landfills and surrounding areas in Ontario, where we conducted geoelectrical surveys to image low-resistivity and elevated IP responses associated with contaminated groundwater.
When integrated with existing hydrogeological and geochemical datasets, geoelectrical tomography serves as a practical decision-support tool for risk assessment, monitoring, and remediation planning.
This approach is applicable across diverse vulnerable environmental contexts.

Integrating Geoelectrical Data with Groundwater Flow and Contaminant Transport Modelling

Linking geophysical observations with dynamic groundwater models is essential to overcome the limitations of time-constrained static contamination mapping, which cannot fully resolve plume timing, migration direction, or controlling processes.
By integrating geoelectrical imaging with numerical groundwater flow and contaminant transport models, we explicitly resolve subsurface behavior in space and time.
At a First Nation in northern Ontario, where we carry out this research, resistivity and induced polarization (IP) data delineate conductive, high-IP zones associated with leachate-impacted groundwater.
These geoelectrical responses inform and support a transient groundwater flow and solute transport model that simulates contaminant migration over multi-decadal timescales under variable climatic conditions.
The model incorporates hydrostratigraphy, precipitation, surface-water interactions, and pollutant mitigation, capturing the influence of seasonal recharge, aquifer connectivity, and engineered controls on leachate plume evolution.
In parallel, we derive a kinematic contaminant transport model from long-term hydrogeochemical records contained in environmental assessment reports, collected from monitoring wells tested in the upper and lower aquifers.
Spatial interpolation of normalized composite pollution indices from successive monitoring years tracks the temporal evolution and displacement of the leachate plume at screened depths, providing an observation-based benchmark for direct comparison with numerical simulations.
The novelty of this approach lies in the integrated use of geoelectrical observations, process-based groundwater numerical modelling, and hydrogeochemically derived kinematic plume reconstructions.
By linking spatially continuous geophysical data with depth-resolved plume trajectories, our approach reduces reliance on dense monitoring networks and provides a transferable basis for contaminant forecasting, environmental risk assessment, and remediation planning.

Coupling Methane Emissions with Geoelectrical Proxies of Leachate Accumulation

We have explored the potential coupling between methane emissions and geoelectrical proxies associated with waste stabilization processes.
At a municipal landfill, we integrated co-located resistivity and induced polarization (IP) surveys with surface methane measurements using a supervised machine-learning approach.
This integration allowed us to resolve non-linear relationships linking methane concentrations to subsurface geoelectrical responses.
These geoelectrical responses are controlled by leachate accumulation, moisture, ionic content, and biogeochemical activity within the waste mass.
The analysis revealed significant shared subsurface controls between methane emissions and geoelectrical signatures.
Building on these previous data-driven results, we are now extending our research by examining contaminant exchange between soil, groundwater, and air.
We use probabilistic machine-learning–based mapping to identify consistent cross-media contamination pathways.

Machine Learning Mapping of Contamination Patterns

Our research moves beyond the use of conventional pollution indices by adopting machine-learning–based mapping of hydrogeochemical data.
Conventional indices compress multivariate information into single metrics, often obscuring contaminant interactions, plume architecture, and migration behavior.
Machine-learning approaches retain the full multivariate structure of groundwater chemistry, allowing contamination patterns to be defined by data-driven relationships rather than predefined aggregation rules.
Our current research focuses on integrating hydrogeochemical monitoring data using machine-learning to support spatial delineation of contamination structure and evolution in groundwater systems with sparse and uneven sampling.
So far, we have applied a suite of unsupervised multivariate machine-learning models integrated with geostatistical interpolation to identify pollution regimes, mixing regions, and migration fronts.
The resulting classifications are explicitly mapped in geographic space, enabling the discrimination of distinct contamination domains while reducing interpretive bias through multi-model comparison.
Machine-learning–derived hydrogeochemical facies maps are interpreted alongside depth-equivalent slices from 3D resistivity and induced polarization (IP) models.
By examining their geometric co-occurrence and boundary alignment, the combined hydrogeochemical–geophysical view provides complementary, mutually constraining evidence to clarify plume geometry, compositional heterogeneity, and migration fronts.
This integrated approach strengthens subsurface interpretation in contaminated environments.

Environmental Magnetism for High-Resolution Pollution Monitoring

Environmental magnetism provides a framework for characterizing anthropogenic particulate matter (PM) in settings where conventional air-quality monitoring is limited.
We have recently initiated an environmental magnetism project in collaboration with the University of Edinburgh.
A central objective of this research is to characterize the grain-size and shape distributions of magnetic minerals associated with particulate matter (PM) in surface soils and dust samples, which act as natural collectors of airborne contaminants.
Because fine and ultrafine PM can penetrate deep into the respiratory system and cross physiological barriers, particle size is a primary control on toxicity, making granulometric resolution essential for exposure and health risk assessment.
We compare First-Order Reversal Curve (FORC) magnetometry measurements from natural samples with extensive libraries of synthetic FORC curves generated through micromagnetic modelling.
Machine-learning–based inversion techniques, developed in collaboration with the Edinburgh team, enable quantitative recovery of magnetic grain-size and shape distributions by matching observed and modelled magnetic responses.
When integrated with complementary chemical data (e.g., heavy metals), these morphology-sensitive magnetic fingerprints improve air contamination mapping and source attribution, with the potential to assess links between PM size distributions and health outcomes.
The approach is non-invasive, cost-effective, and transferable to rural, remote, and underserved regions, supporting environmental monitoring and evidence-based public health decision-making.

Linking Environmental Risks to Health

We have developed a preliminary numerical framework combining average treatment effect and uplift modelling to assess causation between environmental exposure and childhood acute myeloid leukemia (AML) using synthetic data from matched case–control studies in California.
Counterfactual analysis identified causal effects for indoor benzene exposure, while outdoor exposure indicators showed no evidence of causation.
Building on this, we apply tree-based causal machine-learning methods (Random Forest, Bayesian Additive Regression Trees, and Rule Ensemble) to synthetic data from the previous regression-based model.
These methods characterize how causal risk varies across exposure levels and population characteristics by modelling heterogeneous treatment effects.
They capture nonlinear exposure–response relationships, subgroup-specific effects, and policy-relevant exposure thresholds that population-averaged methods cannot identify.
In a proposed next phase, these computational frameworks could be implemented in the Indigenous First Nation community in northern Ontario with which we have been working, using relevant community-specific environmental monitoring and health data sources.

Co-developing Knowledge Translation with Community

We advance a novel, community-driven model that integrates environmental engineering with Indigenous epistemologies through sustained dialogue, shared interpretation, and ethical co-production of knowledge.
Rather than positioning communities solely as recipients of data, the approach actively involves Indigenous leaders, Elders, youth, and community members in shaping research questions, guiding field activities, and interpreting results through Indigenous Integrated Knowledge Translation (IIKT).
Engineering activities are embedded within culturally grounded processes, such as Talking Circles and other locally defined protocols.
Indigenous Knowledge is recognized and treated as a parallel and equally valued knowledge system that informs sampling site selection, validates technical interpretations, and reframes environmental risk in relation to land, health, culture, and overall well-being.
The non-hierarchical IIKT strategy is transferable to both Indigenous and non-Indigenous contexts where trust, transparency, and meaningful participation are fundamental.
Capacity building is achieved through hands-on skill development for community members and trainees involved in field activities and data interpretation.
The approach produces accessible, visual, and community-owned data products derived from environmental monitoring practices that strengthen locally led environmental monitoring and support informed decision-making on contamination issues, health concerns, and remediation priorities.
Ultimately, this positions Indigenous communities as leaders in environmental stewardship rather than passive stakeholders.