Research

We monitor Arctic sea ice as it changes under a warming climate. Our work tracks ice type, thickness, roughness, and the timing of freeze-up and melt, which together describe the state of the ice cover through the season. As the Arctic shifts toward a thinner, more dynamic first-year ice cover, these properties are changing quickly, and existing operational products do not always capture them at the scales that matter. We work from the local scale, where communities make travel decisions, to the pan-Arctic scale that climate research depends on. The aim is to make sea ice observable and interpretable across both.

Radar is our primary tool because it sees through cloud and darkness and responds to the physical properties of snow and ice. We use synthetic aperture radar (SAR) across frequencies to characterize sea ice. No single frequency resolves the full range of ice conditions, so we combine frequencies and pair them with field measurements to interpret what the radar signal actually represents. This multi-frequency approach lets us separate ice types and stages that look similar to any one sensor. It also prepares our methods for the next generation of radar satellites now coming into service.

We develop machine-learning methods for sea ice classification and property retrieval from radar imagery. Our focus is on physics-constrained models that encode the physics of radar scattering, rather than relying on statistical patterns alone. Building physical constraints into a model keeps its outputs physically consistent and prevents it from finding correlations that violate what we know about how radar interacts with ice. This matters most when a model is applied to ice conditions or regions it has never seen, where purely data-driven methods tend to fail. Done well, this approach reduces the need for large labelled training sets, which are scarce in the Arctic.

We detect and characterize hazardous and newly formed ice, including thin and refrozen ice, landfast polynyas, and other conditions that put Arctic travellers at risk. Many of these hazards are dangerous precisely because they are hard to see. Freeze-up is a particular focus, since it is the least observed phase of the sea ice cycle and the stage where new ice forms and is most difficult to distinguish. We use multi-frequency radar to separate these conditions and, where possible, to characterize how new and hazardous ice develops through the season.

We work with northern communities and weave Inuit Qaujimajatuqangit together with our sea ice science. Communities hold detailed knowledge of local ice conditions and of the hazards that matter for travel and hunting, and that knowledge shapes what we choose to measure and map. Weaving local knowledge into radar observations makes the resulting information meaningful for on-ice travel decisions, not just scientifically accurate. We also co-design how results are shared, so that maps and products reach communities in formats and on timelines they can use. This engagement runs through our projects from the start rather than being added at the end.