Combining modelled snowpack stability with machine learning to predict avalanche activity

• Published in: The Cryosphere
• Main Authors: Viallon-Galinier, Léo, Hagenmuller, Pascal, Eckert, Nicolas
• Other Authors: Centre d'Etudes de la Neige (CEN), Centre national de recherches météorologiques (CNRM), Institut national des sciences de l'Univers (INSU - CNRS)-Observatoire Midi-Pyrénées (OMP), Institut de Recherche pour le Développement (IRD)-Université Toulouse III - Paul Sabatier (UT3), Université de Toulouse (UT)-Université de Toulouse (UT)-Institut national des sciences de l'Univers (INSU - CNRS)-Centre National d'Études Spatiales Toulouse (CNES)-Centre National de la Recherche Scientifique (CNRS)-Météo-France-Institut de Recherche pour le Développement (IRD)-Université Toulouse III - Paul Sabatier (UT3), Université de Toulouse (UT)-Université de Toulouse (UT)-Institut national des sciences de l'Univers (INSU - CNRS)-Centre National d'Études Spatiales Toulouse (CNES)-Centre National de la Recherche Scientifique (CNRS)-Météo-France-Centre National de la Recherche Scientifique (CNRS)-Institut national des sciences de l'Univers (INSU - CNRS)-Observatoire Midi-Pyrénées (OMP), Université de Toulouse (UT)-Université de Toulouse (UT)-Institut national des sciences de l'Univers (INSU - CNRS)-Centre National d'Études Spatiales Toulouse (CNES)-Centre National de la Recherche Scientifique (CNRS)-Météo-France-Centre National de la Recherche Scientifique (CNRS)-Observatoire des Sciences de l'Univers de Grenoble (OSUG ), Institut national des sciences de l'Univers (INSU - CNRS)-Université Savoie Mont Blanc (USMB Université de Savoie Université de Chambéry )-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE)-Université Grenoble Alpes (UGA)-Météo-France-Institut national des sciences de l'Univers (INSU - CNRS)-Université Savoie Mont Blanc (USMB Université de Savoie Université de Chambéry )-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE)-Université Grenoble Alpes (UGA), Institut des Géosciences de l’Environnement (IGE), Institut de Recherche pour le Développement (IRD)-Institut national des sciences de l'Univers (INSU - CNRS)-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE)-Université Grenoble Alpes (UGA)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP ), Université Grenoble Alpes (UGA)
• Format: Article in Journal/Newspaper
• Language: English
• Published: HAL CCSD 2023
• Subjects: snow avalanche; machine learning; mechanical stability indices; snow cover modelling; cross-validation; avalanche forecasting; [SDE]Environmental Sciences; The Cryosphere
• Online Access: https://hal.science/hal-04116830; https://hal.science/hal-04116830/document; https://hal.science/hal-04116830/file/tc-17-2245-2023.pdf; https://doi.org/10.5194/tc-17-2245-2023

International audience Predicting avalanche activity from meteorological and snow cover simulations is critical in mountainous areas to support operational forecasting. Several numerical and statistical methods have tried to address this issue. However, it remains unclear how combining snow physics, mechanical analysis of snow profiles and observed avalanche data improves avalanche activity prediction. This study combines extensive snow cover and snow stability simulations with observed avalanche occurrences within a random forest approach to predict avalanche situations at a spatial resolution corresponding to elevations and aspects of avalanche paths in a given mountain range. We develop a rigorous leave-one-out evaluation procedure including an independent evaluation set, confusion matrices and receiver operating characteristic curves. In a region of the French Alps (Haute-Maurienne) and over the period 1960–2018, we show the added value within the machine learning model of considering advanced snow cover modelling and mechanical stability indices instead of using only simple meteorological and bulk information. Specifically, using mechanically based stability indices and their time derivatives in addition to simple snow and meteorological variables increases the probability of avalanche situation detection from around 65 % to 76 %. However, due to the scarcity of avalanche events and the possible misclassification of non-avalanche situations in the training dataset, the predicted avalanche situations that are really observed remains low, around 3.3 %. These scores illustrate the difficulty of predicting avalanche occurrence with a high spatio-temporal resolution, even with the current data and modelling tools. Yet, our study opens perspectives to improve modelling tools supporting operational avalanche forecasting.