Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Félix St-Denis has a total of 7 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Comment on egusphere-2026-4686

    by Félix St-Denis, L. Bruno Tremblay

    We implement a nonlocal ice strength regularization in the standard Viscous-Plastic (VP) and Elastic-Viscous-Plastic (EVP) models. To this end, we introduce an inhomogeneous Helmholtz equation for the ice strength that provides a physically based regularization of the ill-posed VP and EVP formulations while preserving the rate-independence of the plastic regime and the material instabilities that give rise to narrow deformation bands. A linear stability analysis of the new system of equations shows that the nonlocal regularization imposes an upper bound on the growth rates of perturbations, a necessary condition for well-posed problems. Those results are confirmed in idealized divergent numerical experiments: the standard (E)VP models yield resolution-dependent solutions with an increasing number of fractures as the spatial resolution is increased, whereas the nonlocal models produce a resolution-independent solution with a single fracture at the center of the ice plate. Numerical experiments demonstrating the ill-posedness of the hyperbolic tangent viscosity regularization previously identified through a linear stability analysis are also presented. With the advent of kilometer-scale sea ice simulations (Δx ≤ 2–3 km), a well-posed rheology is essential to ensure that fracture statistics reflect physical deformation processes rather than numerical instabilities.

  • Open Access

    Nonlocal ice strength regularization of the Viscous-Plastic sea ice rheology: resolving ill-posedness and resolution-dependent fracture patterns

    by Félix St-Denis, L. Bruno Tremblay

    We implement a nonlocal ice strength regularization in the standard Viscous-Plastic (VP) and Elastic-Viscous-Plastic (EVP) models. To this end, we introduce an inhomogeneous Helmholtz equation for the ice strength that provides a physically based regularization of the ill-posed VP and EVP formulations while preserving the rate-independence of the plastic regime and the material instabilities that give rise to narrow deformation bands. A linear stability analysis of the new system of equations shows that the nonlocal regularization imposes an upper bound on the growth rates of perturbations, a necessary condition for well-posed problems. Those results are confirmed in idealized divergent numerical experiments: the standard (E)VP models yield resolution-dependent solutions with an increasing number of fractures as the spatial resolution is increased, whereas the nonlocal models produce a resolution-independent solution with a single fracture at the center of the ice plate. Numerical experiments demonstrating the ill-posedness of the hyperbolic tangent viscosity regularization previously identified through a linear stability analysis are also presented. With the advent of kilometer-scale sea ice simulations (Δx ≤ 2–3 km), a well-posed rheology is essential to ensure that fracture statistics reflect physical deformation processes rather than numerical instabilities.

  • Open Access

    Coastal Sea Ice Concentration Derived from Marine Radar Images: A Case Study from Utqiaġvik, Alaska

    by Felix St-Denis, L. Bruno Tremblay, Andrew R. Mahoney, Kitrea Pacifica L. M. Takata-Glushkoff

    We apply the Canny edge algorithm to imagery from the Utqiaġvik coastal sea ice radar system (CSIRS) to identify regions of open water and sea ice and quantify ice concentration. The radar-derived sea ice concentration (SIC) is compared against the (closest to the radar field of view) 25 km resolution NSIDC Climate Data Record (CDR) and the 1 km merged MODIS-AMSR2 sea ice concentrations within the ∼11 km field of view for the year 2022–2023, when improved image contrast was first implemented. The algorithm was first optimized using sea ice concentration from 14 different images and 10 ice analysts (140 analyses in total) covering a range of ice conditions with landfast ice, drifting ice, and open water. The algorithm is also validated quantitatively against high-resolution MODIS-Terra in the visible range. Results show a correlation coefficient and mean bias error between the optimized algorithm, the CDR and MODIS-AMSR2 daily SIC of 0.18 and 0.54, and ∼−1.0 and 0.7%, respectively, with an averaged inter-analyst error of ±3%. In general, the CDR captures the melt period correctly and overestimates the SIC during the winter and freeze-up period, while the merged MODIS-AMSR2 better captures the punctual break-out events in winter, including those during the freeze-up events (reduction in SIC). Remnant issues with the detection algorithm include the false detection of sea ice in the presence of fog or precipitation (up to 20%), quantified from the summer reconstruction with known open water conditions. The proposed technique allows for the derivation of the SIC from CSIRS data at spatial and temporal scales that coincide with those at which coastal communities members interact with sea ice. Moreover, by measuring the SIC in nearshore waters adjacent to the shoreline, we can quantify the effect of land contamination that detracts from the usefulness of satellite-derived SIC for coastal communities.


  • Open Access

    Coastal Sea Ice Concentration Derived from Marine Radar Images: A Case Study from Utqiaġvik, Alaska

    by Félix St-Denis, Bruno Louis Tremblay, Andrew R. Mahoney, Pacifica Askitrea Leona Mai Takata-Glushkoff

    We apply the Canny Edge algorithm to the Utqiaġvik coastal sea ice radar system (CSIRS) to quantify open water and sea ice (landfast or drifting) concentration. The radar-derived sea ice concentration is compared against the 25-km resolution NSIDC Climate Data Record (CDR, one pixel closest to the radar field of view) and the 1-km merged MODIS-AMSR2 sea ice concentrations in the ∼11 km field of view for the year 2022-2023, when improved image contrast was first implemented. The algorithm was first optimized using sea ice concentration from 14 different images and 10 ice analysts (140 analysis in total) covering a wide range of ice condition with landfast ice, drifting ice and open water. The algorithm is also validated quantitatively against high-resolution MODIS-Terra in the visible range. Results show a correlation coefficient and mean bias error between the optimized algorithm, the CDR and MODIS-AMSR2 daily SIC of 0.18 and 0.54, and ∼ -1 and 0.9%, respectively, with an averaged inter-analyst error of ± 3%. In general, the CDR captures the melt period correctly and overestimates the SIC during the winter and freeze-up period, while the merged MODIS-AMSR2 better captures the punctual break-out events in winter including those during the freeze-up events (reduction of in SIC). Remnant issues with the detection algorithm include the false detection of sea ice in the presence, fog or precipitation (up to 20%), quantified from the summer reconstruction with known open water conditions. Coastal marine radars provide prove very useful in bridging the gap between (lower) resolution satellite-derived and in-situ (visual) sea ice conditions, and to quantify land contamination from lower resolution satellite-derived SIC.


  • Open Access

    Evaluating approaches for on-the-fly machine learning interatomic potentials for activated mechanisms sampling with the activation-relaxation technique nouveau

    by Eugène Sanscartier, Félix Saint-Denis, Karl-Étienne Bolduc, Normand Mousseau

    In the last few years, much effort has gone into developing general machine-learning potentials capable of describing interactions for a wide range of structures and phases. Yet, as attention turns to more complex materials, including alloys and disordered and heterogeneous systems, the challenge of providing reliable descriptions for all possible environments becomes ever more costly. In this work, we evaluate the benefits of using specific vs general potentials for the study of activated mechanisms in solid-state materials. More specifically, we test three machine-learning fitting approaches using the moment-tensor potential to reproduce a reference potential when exploring the energy landscape around a vacancy in Stillinger–Weber silicon crystal and silicon–germanium zincblende structures using the activation-relaxation technique nouveau (ARTn). We find that a targeted on-the-fly approach specific to and integrated into ARTn generates the highest precision on the energetics and geometry of activated barriers while remaining cost-effective. This approach expands the types of problems that can be addressed with high-accuracy ML potential.