International Snow Science Workshop Proceedings - Montana State University Library

Item: Before machine learning: validating historical SNOWPACK simulations for avalanche hazard forecasting in Glacier National Park, B.C., Canada.

• • Title: Before machine learning: validating historical SNOWPACK simulations for avalanche hazard forecasting in Glacier National Park, B.C., Canada.

    Proceedings: Proceedings, International Snow Science Workshop, Whistler, BC, Canada, 2026

    Authors:

    • Benjamin Imbach  [ GEOSLAB, Department of Biology, Chemistry and Geography, University of Quebec at Rimouski, Canada ]  [ GRIMP, Department of Applied Geomatics, University of Sherbrooke, Canada ]  [ Center for Nordic studies, Laval University, Canada ]
    • Francis Meloche  [ Institute of Geotechnical Engineering, ETH Zürich, Switzerland. ]  [ GEOSLAB, Department of Biology, Chemistry and Geography, University of Quebec at Rimouski, Canada ]  [ GRIMP, Department of Applied Geomatics, University of Sherbrooke, Canada ]
    • Jean-Benoit Madore  [ GRIMP, Department of Applied Geomatics, University of Sherbrooke, Canada ]
    • Paul Billecocq  [ GRIMP, Department of Applied Geomatics, University of Sherbrooke, Canada ]
    • Francis Gauthier  [ GEOSLAB, Department of Biology, Chemistry and Geography, University of Quebec at Rimouski, Canada ]  [ Center for Nordic studies, Laval University, Canada ]
    • Alexandre Langlois  [ GRIMP, Department of Applied Geomatics, University of Sherbrooke, Canada ]  [ Center for Nordic studies, Laval University, Canada ]
    • Philippe Gachon  [ Department of Geography, Étude et Simulation du Climat à l'Échelle Régionale center (ESCER), University of Quebec in Montreal, Canada ]

    Date: 2026-09-28

    Abstract: Accurate avalanche hazard forecasting depends on reliable snow and weather information, which is often unavailable in the complex mountainous terrain where hazards are greatest. This limits the availability of training data for emerging machine learning approaches to avalanche hazard assessment.This study lays the groundwork for such a model for Glacier National Park (GNP), British Columbia, Canada, by validating snow cover simulations produced by the MeteoIO/SNOWPACK/Alpine3D model chain driven by four gridded meteorological forcing datasets spanning reanalyses, a regional climate model, and a numerical weather prediction model. Each dataset is downscaled using subgridding and terrain-adapted interpolation techniques. Modelled snow cover properties are then compared against an extensive record of snow profiles and instability tests collected by the Avalanche Control Section between the 1999-2000 and 2022-2023 seasons, using the profile similarity approach. This work establishes a validated historical simulation framework upon which future machine learning approaches to avalanche hazard forecasting in GNP can be built.

    Object ID: ISSW2026_O2.3.pdf

    DOI: https://doi.org/10.15788/1790098554

    Language of Article: English

    Presenter(s): Benjamin Imbach

    Keywords: Avalanche forecasting, SNOWPACK/Alpine3D, Model validation, Weak layer identification, Meteorological forcing data, Glacier National Park

    Page Number(s): 932 - 939

  • Get Full Article

Back to Home page