Supershear snow avalanches

Numerical simulations and full-scale avalanche measurements revealed a transition from sub-Rayleigh anticrack to supershear crack propagation during the release of snow slab avalanches. This fracturing process is similar to that reported in rare high magnitude earthquakes.
Supershear snow avalanches

My group and I are interested in the understanding and modeling of gravitational mass movements with a particular focus on snow and avalanche mechanics. We were particularly interested in one type of avalanche responsible for most avalanche-related accidents and damage: dry-snow slab avalanches. This type of avalanche releases due to the failure of a highly porous weak snow layer buried below a cohesive snow slab. Once a failure is initiated in the weak layer, for example due to the overload of a skier, this failure may propagate across the slope leading to the release of the snow slab. In the last decade, a lot of progress has been made in the understanding of these crack propagation mechanisms in snow. In particular, the important effect of the volumetric collapse of the weak layer was highlighted. This collapse leads to the bending of the overlying slab and thus a “house of cards” type of propagation which can occur even on flat terrain, as reported in the field. This type of fracture, with closing crack faces, is often reported as ‘anticrack’ in the literature and is notoriously difficult to simulate.

In 2017, we developed in collaboration with UCLA, a three-dimensional model based on the so-called “Material Point Method” (MPM) to simulate both the initiation and dynamics of snow avalanches in an integrated manner taking into account the anticrack mechanism. At this stage, the model was successfully validated based on small scale snow fracture experiments only and its up-scaling revealed initially surprising and disturbing results. On the one hand the crack propagation speed in small scale experiments and simulations matched very well and was typically lower than 60 m/s. On the other hand large scale simulations suggested speeds that could increase up to values higher than 100 m/s. Such large values have not only never been reported experimentally, but also appeared to be larger than the snow slab shear wave speed, which we thought was not possible. But what was at first worrying, turned out to be an interesting discovery with practical relevance for hazard assessment, as I will mention later.

A supershear avalanche accidentally triggered by Mathieu Schaer in Wallis (c) Ruedi Flück

After some literature review outside our field of research, namely in earthquake science, we found that crack propagation with such intersonic speeds were reported in high-magnitude strike-slip earthquakes and in some rupture laboratory experiments. This started to become extremely interesting and exciting. In fact, the term ‘anticrack’ mentioned above, was initially introduced to explain deep earthquakes! Perhaps avalanches and earthquakes had even more in common than we thought? We needed more. We needed experimental evidence! In 2019, a friend of mine, Mathieu Shaer, a professional snowboarder who I often go snowboarding with, and coincidentally one of my EPFL students, accidentally triggered a deep snow slab avalanche from which he successfully and impressively escaped. The accident occurred while filming for a snowboard movie. The video of the accident was filmed with a high-quality and high-speed camera which Mathieu sent to me for analysis, together with detailed information (location, avalanche danger, snow profile). In parallel, we knew from a seminar happening a few weeks before that our colleague Ron Simenhois was working in the US on a video analysis technique based on the change in pixel intensity as a proxi of slab deformation that could allow him to compute crack propagation speeds. We sent him the video and a few days later, he sent us his analysis: in the cross-slope direction the crack speed is sub-Rayleigh while it is intersonic in the down-slope direction! Stars aligned! We later sent him three other avalanche videos further confirming these findings. A multidisciplinary team as well as the combination between advanced numerical simulations and full-scale avalanche measurements allowed us to reveal this important transition in the avalanche release process: for short propagation distances, the crack is rather slow (sub-Rayleigh) and driven by the volumetric collapse of the weak layer (anticrack mechanism); for propagation distances typically larger than 2 to 5 meters, the crack becomes intersonic and is driven by weak layer shear failure, a process known as supershear fracture in earthquake science. Why did we miss this process in previous research? Because we were focused on simulating snow fracture experiments which are typically 1 to 2 meters long. This result recently motivated our colleague Alec van Herwijnen at the WSL Institute for Snow and Avalanche Research SLF in Davos to perform last winter the first large scale snow fracture experiments to analyze this process in greater details.

This finding has important practical consequences for the evaluation of the avalanche release size which is important quantity for avalanche forecasting and required as input of avalanche dynamics models used for hazard mapping. If we are interested in the release of large avalanches, a pure shear failure model can be assumed for the weak layer, which facilitates the development of more efficient models based on shallow water equations. Indeed a full 3D avalanche simulation may require several days of computation, whereas a depth-integrated simulation only takes a few minutes.

A final anecdote: in 2014,  Dave Hamre et co-workers made a presentation at the International Snow Science Workshop in Banff. Based on the analysis of slab fractures in avalanche videos, they suggested that crack propagation speeds could be way larger that those measured in snow fracture experiments. At that time I was not convinced and thought in particular that their results could be influenced by the explosives used to trigger the avalanches. It appears that I was wrong.

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in