Researches carried out over the past 18 years in the MBM have been funded by several main research programs: EU Alcotra PERMAdataROC, PrévRisk Haute-Montagne, AdaPT Mont-Blanc, and PrévRisk CC projects; EU AlpineSpace PermaNET project; ANR-14-CE03-000 VIP Mont-Blanc and ANR-19-CE01-0018 WISPER projects. The authors would like to thank the Communauté de Communes de la Vallée de Chamonix Mont-Blanc, the Compagnie du Mont-Blanc, the Compagnie des Guides de Chamonix and the École Nationale de Ski et d’Alpinisme for their support since 2005, as well as the numerous colleagues who contributed to these projects, notably, but non-exhaustive: Paolo Pogliotti, Umberto Morra di Cella and Eduardo Cremonese from Agenzia Regionale per la Protezione dell’Ambiente Valle d’Aosta (Aosta, Italy), Jeannette Noetzli now in Institute for Snow and Avalanche Research (Davos, Switzerland), Stephan Gruber now in Carleton University (Ottawa, Canada), Michael Krautblatter now in Technical University of Munich (Germany), and Jan Beutel now in University of Innsbruck (Austria). The authors also acknowledge the editorial committee of the Journal of Alpine Research for handling the manuscript, the invited editor (Marta Chiarle), the anonymous reviewers and Lothar Schrott for their constructive comments and corrections on the manuscript.

Introduction

1Since the end of the 1980s, climate change has initiated profound changes in landscapes, ecosystems, economic resources, and natural hazards in the high mountain. Landscapes dominated by bare rock and talus are replacing glaciers, with accumulations of unconsolidated material, steep rock slopes and new lakes forming in glacier bed overdeepening or behind moraine dams (Haeberli et al., 2016). At the same time, periglacial rock faces have become increasingly unstable, potentially causing infrastructure destabilisation (e.g., refuges, ski lifts, Duvillard et al., 2019), disappearance or alteration of mountaineering routes (Mourey et al., 2019), and major disasters through high-magnitude events and possible cascading hazards (Haeberli et al., 2016). These emerging threats urge for the development of in-depth knowledge to set up targeted mitigation and adaptation strategies.

2While glacial retreat is a well-known factor of slope destabilisation, notably through the process of paraglacial adjustment (Ballantyne et al., 2014), the role of permafrost degradation—i.e., the thawing of terrain whose temperature remains permanently < 0 °C—in response to atmospheric warming, has gradually become the focus of scientific concerns (e.g., Gruber et al., 2004). Lacking consideration of rock wall permafrost (i.e., permafrost affecting rock walls) until the early 2000s, primarily because of its invisibility, resulted in a crucial gap in the systemic understanding of high alpine morphodynamics. A return to early studies on frost weathering processes (e.g., Walder, 1985) and on cold environment engineering (e.g., Mellor, 1971), combined with the development of specific studies on rock wall permafrost (e.g., Gruber et al., 2004) and permafrost-affected rock walls, i.e., rock walls that are likely affected by permafrost (e.g., Ravanel et al., 2010), has contributed to theoretical assessments of the role of permafrost in the morphodynamics of steep rock slopes.

3Several thermo-hydro-mechanical processes related to permafrost are considered as possible triggering mechanisms (V > 100 m³) (Krautblatter et al., 2013). They operate in different depths and possess different response times to climate forcing ranging from seasonal time scales such as active layer thickening (i.e., shallow layer that thaws in summer and refreezes in winter), to millennial time scales with the degradation of permafrost formed during the Holocene at greater depth (Noetzli and Gruber, 2009). Frozen rock and ice-infill in cracks are mechanically weakened during warming and thawing, which makes frozen rock walls highly sensitive to atmospheric warming. But because normal stress control prevails in slope failures, the effectiveness of the various processes associated with thawing depends on the rock overburden. Detachment depths < 20 m are thought to involve failures along ice contact in cracks while detachment depths > 20 m are more likely to be caused by changes in rock mechanics (Krautblatter et al., 2013). Explaining failures occurring along pre-existing cracks is challenging because they induce a variety of non-linear thermal, hydrogeological and mechanical processes. Heat propagates from the surface to depth through progressive conduction in the rock mass, but cracks favour air ventilation and heat advection from water infiltration, as well as latent heat exchange under freezing and melting of the ice/water-infill. In addition, water circulation in ice-sealed bedrock can lead to the development of a water column forming high hydrostatic pressure, so that changes in the thermal regime are not the only potential triggers. The frequent presence of ice and water in failure scars (e.g., Deline et al., 2013) provides evidence of the plausibility of complex triggering mechanisms combining warming and melting of the ice-infill, as well as water-related thermal and mechanical processes. As a result, linking these theoretical explanations to a technical understanding of observed rock slope failures bears many challenges for the development of operational solutions for risk assessment.

4Following the pioneering studies on rock wall permafrost launched in Switzerland after the hot summer of 2003, marked by a numerous rockfalls (Gruber et al., 2004), research began in France in the Mont Blanc massif (MBM) in 2005. MBM naturally emerged as a privileged study area, notably because of its unique combination of high-elevated peaks and pillars, and the rich tradition of researches on rock avalanches (Deline et al., 2012). Several research projects have been carried out, largely initiated or directed by Philip Deline and are led now by several active researchers in the EDYTEM laboratory. This contribution summarises the first 18 years of research in the MBM on gravitational processes driven by the degradation of the cryosphere in steep slopes—here referenced as “cryo-gravitational processes”, and which address the following questions: where and when do cryo-gravitational hazards occur? What are the physical processes involved in their triggering?

The Mont Blanc Massif: A Privileged Open-Air Laboratory for Investigating Morphodynamics and Cryosphere of Alpine Rock Faces

General Settings

5MBM is one of the external crystalline massifs of the western European Alps extending over 550 km² across France, Swiss and Italy. It is a window of the Variscan basement within the Penninic and Helvetic nappes, and is composed of two main lithological units: a 453 ± 3 Ma Variscan metamorphic series (mostly gneisses and mafic schists) intruded by a 304 ± 3 Ma calc-alkaline granite and aplitic veins (von Raumer and Bussy, 2004). Oriented SW-NE, the Variscan batholith is bounded to the NW by a wide shear zone and by thrusts in the E and SE. Faults and shear zones are mostly oriented NW-SE with subvertical dipping organised with a fan geometry across the range that outlines its main topographical features.

6Strongly asymmetric, the average slope angle between Mont Blanc summit at 4808 m ASL and Courmayeur (1230 m) on its SE flank is about 40°, while it is 15° on its NW flank between the top and Chamonix at 1040 m. This asymmetry is also marked by the climate settings as the main crest line with 28 summits above 4000 m ASL acts as a climate border where Mediterranean and Atlantic fluxes meet. As a result, the Italian side is typically warmer and drier than the French side.

7In the late 2000s, about 100 km² of the MBM were still glaciated (Gardent et al., 2014; Fig. 1), but affected by pronounced glacier shrinkage since the 1980s (Gardent et al., 2014), while the morphodynamics of its steep rock faces is accelerating, posing hazards and risks for the infrastructure and population (Deline et al., 2012).

Main Data and Knowledge Gathered Over the Last 18 Years

8Permafrost investigations started in autumn 2005, with the first temperature sensors installed at the near surface (3 to 55 cm depth) of Aiguille du Midi (AdM, 3842 asl) rock walls (Magnin et al., 2015) while a few months before, the Bonatti pillar collapsed (292 000 m³; Guerin et al. (2020) and drew attention to the recent morphodynamics of MBM rock walls. Since then, a variety of data have been collected from field observations and created from models to better understand permafrost and rock wall dynamics and formulate hypotheses about the development of cryo-gravitational hazards (Fig. 2).

9Analyses of the temporal distribution of rockfalls in MBM over millennial to seasonal time scales showed that phases of gravitational activity are concomitant to warm periods, such as the Holocene Climate Optimum (Gallach et al., 2020), the 1940s and every decade since the 1990s (Ravanel and Deline, 2011) or the hot summers of 2003 and 2015 (Ravanel et al., 2017). Since 2007, a network of observers reports rockfalls in the central part of the massif and contributes to the establishment of a detailed rockfall database (Fig. 1).

10Concurrently, rock surface temperature data collected at AdM were included in an Alpine-wide data collection (Cremonese et al., 2011) that has been used to calibrate a multiple linear regression model—the rock model (Boeckli et al., 2012)—relating the Mean Annual Rock Surface Temperature (MARST) to the Mean Annual Air Temperature (MAAT) and the Potential Incoming Solar Radiation (PISR). The model was used to map permafrost over the entire Alpine range and was later applied to MBM on a 4-m-resolution Digital Elevation Model (DEM). The same model implemented with the 1981–2010 MAAT measured in Chamonix and spatialised with a 0.55 °C/100 m lapse rate is displayed in Figure 1.

11Since 2010, permafrost investigations are strengthened with temperature monitoring in 5 boreholes: 4 boreholes reaching a depth of 10-m at AdM (last one installed in 2021), and a 17-m-deep borehole installed in 2017 at Aiguille des Grands Montets.

12Rock face monitoring based on repeated terrestrial laser scanning such as at Tour Ronde (3792 m asl, Rabatel et al., 2008) has shed light on the possible relation between the melting of “ice aprons” (IAs, surface and immobile cold ice bodies) and rockfall occurrence. In their recent inventory, Kaushik et al. (2022) complete former glacier inventories (Gardent et al., 2014) that have been used to model glacier-bed overdeepenings to appraise potential future lakes (Cathala et al., 2021) that could be a source of high-magnitude chain-reaction hazards in the coming decades but at the same time could represent new economic resources (Haeberli et al., 2016).

13The most recent developments focus on bridging these different datasets to develop a comprehensive approach of cryo-gravitational hazards in order to support the development of operational solutions for land planners, geotechnicians and mountain practitioners (Fig. 2). Progress has been achieved in defining where and when cryo-gravitational hazards occur, but the main challenges are related to understanding how they develop and how they are triggered.

Spatio-temporal Distribution of Cryo-Gravitational Hazards

14From 2007 to 2021, 1389 rockfalls > 100 m³ were inventoried with accurate coordinates, topographical settings, date of failure or detection, and if possible, a variety of additional characteristics such as scar depth or ice observation. Most events are shallow (median and average scar depth of approx. 4 m) and their occurrence, if related to permafrost dynamics, is therefore most likely due to failure at the ice contact in cracks (Krautblatter et al., 2013). Analysis of this dataset uncovers key features in the spatial and temporal distribution of rockfalls to link their occurrence with possible triggering mechanisms.

Are Rockfalls Linked to the Degradation of the Cryosphere?

15Figure 3 shows the distribution of the 1152 events that occurred in the area covered by the MARST map (Fig. 1). Ignoring transient and 3D heat fluxes (Noetzli et al., 2007; Noetzli and Gruber, 2009), this 30-year period steady state MARST calculated with the rock model is somewhat representative of expected permafrost conditions at depth. However, absolute values of MARST are partly biased by the limitations of the rock model: it does not take snow into account and its implementation on a DEM smooths out sharp reliefs. Therefore, we interpret these results in terms of isolated to cold permafrost, which is a qualitative interpretation of the calculated MARST. With all precautions taken, rockfalls appear to be strongly related to permafrost; no events have been observed in unfrozen rock walls. The peak of rockfall frequency is found in warm permafrost areas and fewer events occurred in cold as well as isolated permafrost conditions.

16The wide range of permafrost conditions under which rockfalls occurred support the hypothesis that a variety of permafrost-related processes are likely to trigger rockfalls, from the warming of isolated ice-filled cracks that had remained due to the retarding effect of latent heat consumption when interstitial ice melts (Ravanel et al., 2023a), to water infiltration into cold permafrost areas (Hasler et al., 2011; Fig. 3). It also highlights that, areas close to the melting point (warm permafrost) are likely the most active ones.

17In addition, 11.5% of rockfalls occurred in slopes that were covered by IAs in the 1950s (Ravanel et al., 2023a). This quantification is limited because it does not take into account glacier shrinkage at the foot of rock walls and because of the possible mismatch between digitalised IAs and rockfall referencing. But this preliminary verification suggests that a portion of rockfall is directly related to thermal and mechanical disturbances associated with glacier retreat that include debuttressing (Ballantyne et al., 2014), permafrost formation (Wegmann et al., 1998), or active layer formation (Hartmeyer et al., 2020).

Are Rockfalls Caused by Seasonal Thawing?

18Rockfalls mainly occurred during the thawing season: the June-September period gathers 96% of the events. Since 2015, the yearly number of rockfalls has dramatically increased, coinciding with a rise in mean air temperature (MAT) during the thawing season; as a general feature, the warmer the thawing season, the higher the rockfall frequency (Fig. 4a).

19As a potential link between higher MAT and more frequent rockfalls, the deepening of the active layer thickness (ALT) and the subsequent warming and melting of ice-filled cracks (Fig. 3) may be considered as the main triggering mechanism. However, current field data on this matter is ambiguous and does not support a straightforward correlation. For example, in 2019 nearly twice as many rockfalls were observed as in 2020 (Fig. 4a), yet ALT has deepened between 2019 and 2020 in 2 of 3 boreholes at the AdM (Fig. 4bc) in coherence with higher MAT (Fig. 4a). Most rockfalls generally occur well before the maximum ALT is reached, (it is especially the case for 2020). Such finding has already been reported by comparing the rockfall frequency and modelled ALT for the 2003 summer season in Switzerland (Gruber et al., 2004), which supports the hypothesis of water-related processes either accelerating the heat transfer or causing pressure that cannot be depicted through 1D measurements or a simple heat conduction model. Such hypothesis is also supported by recent estimation of the potential water infiltration calculated from snow melting and rainfall that showed that more water was potentially available for infiltration in 2019 than in 2020, especially from rainfall (Ben-Asher et al., 2022).

20Appraising water-related processes has thus become a major challenge to better understand both permafrost degradation processes and rockfall triggering mechanisms (Krautblatter et al., 2012).

Current Developments to Understand the Physical Processes behind Failures

21To address the thermal and mechanical effects of water infiltration and circulation in permafrost rock walls, some attempts to couple thermal and hydrogeological models with advanced numerical modelling tools have been carried out (Magnin and Josnin, 2021). The ultimate aim would be to address processes acting at fine spatial and temporal scales and provide relevant data for mechanical models. Current limits are associated with (i) the ability to constrain the key physical parameters that drive and control hydrological and hydrogeological processes in mountain permafrost environment, (ii) the ability of conventional hydrogeological models to simulate the specific processes occurring in permafrost, and (iii) the lack of field measurements to constrain the models and validate simulations (Fig. 5).

Main Challenges in Quantifying Water Infiltration

22The melting of snow that heterogeneously accumulates on rock walls throughout the years represents a major source of water in high mountains (Fig. 5). Yet, linking snowfall to water infiltration in such an environment is challenged by highly complex processes. Quantitative information on snow accumulation on rock walls is rare and measurement techniques bear large errors and uncertainties (e.g., Haberkorn et al., 2017). Furthermore, infiltration of meltwater can occur days, weeks, and months after a snowfall event. During this period, snow may be redistributed by wind and gravity, sublimate or melt and refreeze, forming an impermeable ice crust (Phillips et al., 2016).

23In MBM, characterisation of snow accumulation was performed on the AdM S face using time-lapse photogrammetry, high-resolution photogrammetric models, and numerical modelling. About 25% of the fallen snow accumulates on the 55° steep and rugged rock face (Ben-Asher et al., 2022), and that average snow depth steadily decreases between 40° slope angle, where it is approx. 1 m (up to > 2.5 m at point scale), and 70°, a threshold above which snow no longer accumulates. By coupling advanced climate, energy balance and heat transfer modelling tools, the same study quantified snow water equivalent and infiltration timing. It provided information on surface temperature and consequently on permafrost distribution, as well as input water data for further analysis with coupled thermal and hydrogeological models. These recent developments broaden the understanding gained in previous studies, showing the crucial control of snow on local rock wall permafrost conditions (Magnin et al., 2017b) which is a key challenge for future studies. In addition to snow, the melting IAs, as seen during the hot summer of 2022, is another possible source of water with unclear influence on the energy balance of the permafrost-affected rock faces (Ravanel et al., 2023b). Further challenges in determining water infiltration are associated with the limited knowledge on the hydraulic conductivity of fractured permafrost-affected rock slopes.

Main Challenges in Appraising Hydrogeological Processes

24In low permeability rock, water flow pathways are mostly concentrated in cracks whose hydraulic conductivity can vary depending on the saturation and degree of freezing of the interstitial water, which can change seasonally. This can effectively influence the macro-scale capability of water to flow, and hence alter hydraulic conductivity.

25The use of numerical tools to support the overall understanding of hydrogeological features and processes faces several difficulties when applied to frozen rock. The relevant spatial and temporal scales of the aforementioned processes span over several orders of magnitude and require strong numerical capabilities (CPU time spanning several weeks). Physical equations describing hydrological processes in unsaturated conditions such as those prevailing in permafrost-affected fractured media are also strongly non-linear, which leads to numerical instabilities. Last but not least, phase changes at subfreezing temperature are not always fully implemented in all conditions in hydrogeological models: for instance, the PiFreeze module that deals with freezing water in the Feflow® program is not implemented to work on discrete feature elements, that would be the best suited to describe fractures. Yet, software improvement and enhanced computing capability could overcome current limitations.

26While these numerical challenges are tackled, efforts are carried out in parallel to gain direct observations of water flow characteristics. Indeed, observations of hydrogeological processes in mountain permafrost rock walls are scarce and mostly attained indirectly as temperature anomalies (e.g., Phillips et al., 2016). At AdM, galleries have been equipped since summer 2022 with a system measuring and sampling water flow to gain insights on water quantity, discharge, age as well as fracture connectivity. Geophysical surveys also bear hope to appraise hydrogeological processes with promising perspectives for geotechnical purposes.

Towards Operational Solutions for Mitigating Cryo-Gravitational Hazards

27Cryo-gravitational hazards pose different types of risks in MBM, whose mitigation urgently needs appropriate anticipatory approaches. For geotechnics, in-situ diagnostics are challenged by the difficulty to assess the ground ice-content, distribution and characteristics. For mountain practitioners, determining the most dangerous periods is needed while for land planners, anticipating the spatial extent and magnitude of hazards is necessary. Three main research directions are thus outlined that could feed into future operational solutions.

Developing Diagnosis Strategies

28In MBM, 60 infrastructure elements, including ski lift components, avalanche gas exploders, and mountain huts, are settled on permafrost ground (e.g., Duvillard et al., 2019) encompassing rock walls, highly fractured and intermediately steep bedrock, and sometimes non-consolidated materials. While permafrost conditions can be roughly assessed, geotechnical maintenance would benefit from a detailed understanding of the thermal state and ground ice content, distribution and potential evolution. Core or destructive boreholes allow such quantitative assessment but their spatial representativeness is poor. Furthermore, current modelling approaches to map or to simulate permafrost evolution at depth are not designed for bedrock with a high degree of fracturing and potentially a relatively high ice content. To overcome these limitations, geophysical methods appear to be promising.

29Despite their indirect character (i.e., measurement of permafrost proxies), combined electric resistivity tomography and induced polarisation measurements offer new opportunities to access relevant ground parameters such as temperature and liquid water content (Duvillard et al., 2021). These techniques have the main advantage to be non-invasive and non-destructive and more spatially representative thanks to 2D to 3D inversions (Fig. 6ab). Coupled with recent advances in petrophysics models (Fig. 6c) and compared to thermal models (Fig. 6d), water and ice content could be interpreted quantitatively (e.g., Wagner et al., 2019).

30The joint inversion of galvanometric/electromagnetic data with seismic and ground-penetrating radar data as well as the fusion with petrophysical models and thermal modelling will be key to obtain 3D distribution of the relevant cryo/hydromechanical properties of alpine settings and their potential evolution towards critical states (Wagner et al., 2019). Future challenges also include the development of joint inversion methods that use terrain or geophysical data, or direct observations from deep boreholes as well as time-guided inversions by combining time-lapse geophysical data, thermal modelling, and mechanical modelling results. As many mountain sites are hardly accessible, the use of drone swarm doing geophysical time-domain electromagnetic measurements remotely provides another perspective.

Enhancing Predictability of Cryo-Gravitational Hazards

31The increasing rockfall frequency and volume threaten the sustainability of high mountain activities (Mourey et al., 2019). Recent statistical analysis of simulated surface and scar-depth temperature of inventoried rockfalls showed high statistical significance between higher than usual surface and scar-depth temperatures and rockfall occurrences, showing the promise of rockfall probability forecast (Legay et al., 2021) that could support mountaineers in their risk analysis. Several improvements could be considered to strengthen the predictability of rockfall occurrence. First, the study could be updated with recent events (2016–2022) as this would improve the statistical robustness. In addition, the statistical significance could be improved by comparing the results of temperature simulations at the real events to some random events defined according to the main topographical features of the real events (see Legay et al. [2021] for more details).

32Another source of improvement lies in the accuracy of the predicted surface temperature that could be better achieved by using energy balance modelling taking into account snow processes (Ben-Asher et al., 2022). It would also be interesting to consider other variables than the daily air temperature to determine the parameters associated with failure occurrences, such as the date of snow melt which partly controls the surface temperature but can also be used as a proxy for water infiltration, or the maxima and minima of air temperature, for example. This would be specifically relevant for rockfalls occurring in the early summer as snow often persists at high elevation until then. The peak of rockfall occurrence in MBM in July 2019 (Fig. 4b), that was reached far before the maximum ALT developed, could for example be related to the rather high amount of available water from snowmelt and rainfall in the early summer (Ben-Asher et al., 2022).

Mapping Current and Future Cryo-gravitational Hazards Susceptibility

33Integrative approaches coupling permafrost assessment, melting of ice faces, glaciers retreat and induced landscape changes are important for land planners who need to anticipate the spatial mutation of hazards in the changing climate and cryosphere conditions, as well as potential economic resources (Haeberli et al., 2016).

34The spatial distribution of inventoried rockfalls can be used to map potential release areas in a hazard assessment approach such as in Figure 7 (not published yet) that shows all rock walls with predicted MARST < 0 °C and the potential runout areas of rockfalls in the MBM and predicted future lakes in the Bionnassay watershed (Cathala et al., 2021). This watershed clearly appears as a hot spot with numerous lakes that could form or are currently forming, and which are possibly exposed to rock or ice falls. Land planners may survey this area with great attention in order to anticipate potential high magnitude events. This area would also be particularly interesting to investigate the relationship between permafrost and hanging glaciers. A key question here is whether hanging glaciers are melting or destabilising due to permafrost degradation taking place in surrounding rock walls?

Conclusions

35Over the past three decades, steep rock slopes in cold environments have become increasingly unstable, with more frequent rockfalls and large-scale events sometimes causing devastating cascading hazards downslope, promoted by the storage of unconsolidated and unstable materials left behind retreating glaciers. In the European Alps, MBM has naturally emerged as an open-air laboratory for studying permafrost-affected rock faces because of the acceleration of its morphodynamics since the 1990s, the uniqueness of its high-elevated rock peaks, the variety of its glacial systems, and the pronounced societal implications of its environmental dynamics.

36Within the past 18 years, a baseline dataset has been established to better understand when and where rock wall destabilisation occurs. The acquired data suggests a dominant influence of permafrost-related processes on rock slope destabilisation. Main research directions are targeting:

1. the deepening of statistical analyses that could serve the development of forecasting tools in order to assess periods and areas of high risk,

2. the advancement of physics-based models to couple thermal processes with hydrogeological processes and ultimately with mechanical processes, that, together with geophysical surveys, could offer new diagnosis approaches for geotechnical purposes,

3. the investigation of the interaction between surface ice bodies (IAs and hanging glaciers) and permafrost and their relative control on each other,

4. and finally, the development of integrative approaches combining state-of-the-art knowledge in order to anticipate landscape changes and support land planners in their evaluation of risks and economic resources.

37These research directions will deepen the work initiated nearly 20 years ago that has made the MBM a world-renowned field for the observation of permafrost-affected rock walls, of rock wall permafrost, and more generally of the cryosphere of steep alpine rock walls and its geomorphological and societal implications.