18-years of high-Alpine rock wall monitoring using terrestrial laser scanning at the Tour Ronde east face, Mont-Blanc massif

Since the end of the 20th century, each decade has been warmer than the previous one in the European Alps. As a consequence, Alpine rock walls are generally facing high rockfall activity, likely due to permafrost degradation. We use a unique terrestrial laser scanning derived rockfall catalog over 18 years (2005–2022) compared with photographs (1859–2022) to quantify the evolution of the east face of Tour Ronde (3440–3792 m a.s.l.) in the Mont-Blanc massif (western European Alps) that is permafrost-affected. Overall, 210 rockfalls were identified, from 1 to 15 500 m3. Forty-five events were >100 m3 while cumulated volume of events <10 m3 represents <1% of the fallen rocks. The rockfall magnitude-frequency distribution of the overall inventory follows a power law, with a mean exponent b of 0.44 ± 0.03, characterizing a high contribution of large rockfalls. The depth of failure ranges from a few centimeters to more than 20 m while 95% of the rockfalls depth is <5 m, highlighting the role of the active layer. The mean rock wall erosion rate is 18.3 ± 0.2 mm yr−1 for the 2005–2022 period and ranks in the top range of reported values in the Alps. It has greatly increased between the periods 2006–2014 and 2016–2022, probably in relation to a series of summer heat waves. The exceptional erosion rate of 2015 is driven by one large rockfall in August. Since 2006, an ice apron that covered 16 100 m2 has now almost vanished, and the surface of the glacier du Géant at the rock wall foot has lowered by several tens of meters. The retreat of these two ice masses contributed to the rock wall instability as more than 35% of the rockfall volume detached from the deglaciated surfaces.

A clear correlation between periods of high temperature and rockfalls has been shown in the European Alps (Huggel et al 2012, Stoffel andHuggel 2012).Schiermeier (2003) showed that normally frozen landscapes were destabilized because ice was thawing below 4500 m a.s.l.(all elevations are in m above sea level).In the MBM, this correlation has been demonstrated since the end of the Little Ice Age at the west face of the Petit Dru (3754 m) and on the north side of the Aiguilles de Chamonix (2300-3842 m; Ravanel andDeline 2008, 2011).More recently, Ravanel et al (2017) showed the impact of the 2003 and 2015 summer heatwaves on permafrostaffected rock walls in the MBM.In high Alpine areas, permafrost degradation (i.e.warming) due to global warming (Magnin et al 2017, Biskaborn et al 2019) is increasingly considered as a major factor for rock wall destabilization (Gruber et al 2004, Gruber and Haeberli 2007, Krautblatter et al 2013).This is in relation to a deepening of the active layer (the layer above the permafrost, subject to annual freeze/thaw cycles).
Rockfalls are a threat for alpinists (Soulé et al 2014, Mourey et al 2018), with mountain infrastructures subject to deterioration related to permafrost degradation (Duvillard et al 2019).The ongoing effects of global warming are likely to exacerbate these issues (Chiarle et al 2021), making long-term rockfall monitoring necessary for effective risk assessment and mitigation (Bommer et al 2010, Duvillard et al 2021).
Rockfall monitoring helps in both understanding steep slopes response to climatic change and quantifying rockfall volume and rock wall erosion rate.Many studies used terrestrial laser scanning (TLS hereafter) to quantify rock wall erosion rates.However, the duration of these studies is generally less than 10 years (Rabatel et al 2008, Ravanel et al 2011, Kenner et al 2011, Hartmeyer et al 2020a, 2020b, Guerin et al 2020, Draebing et al 2022).Long-term monitoring of permafrost-affected rock walls is necessary to assess frequency, as well as volume and triggering factors, of rockfalls.

Geographical and geological context
The MBM extends over 550 km 2 in the western European Alps, with approximately 30% covered by glaciers (figure 1 The MBM is formed of a complex metamorphic basement intruded by the large ( ∼225 km 2 ) and homogeneous Mont-Blanc granite pluton (Bussy and von Raumer 1994).This granite is affected by subvertical faults, shear zones of which two main clusters are oriented ∼N40 and N70 (Rossi 2005, Matasci et al 2018).These faults determine the distribution of granite peaks and couloirs (Bertini et al 1985).Seismic activity in the MBM is far from negligible with around forty events between 2005 and 2022 ranging in the magnitude between 2 and 4.9 (Cara et al 2007).Several rock avalanches occurred over the past centuries (Deline 2001, Deline andKirkbride 2009).Since the 2000s, the MBM has been increasingly affected by numerous small-scale rockfall events (Ravanel et al 2010, Deline et al 2012).TREF was one of the most active rock walls in the MBM during the period 2005-2009 (Ravanel et al 2011).
TREF is a rock wall of approximately 82 000 m 2 between 3440 and 3792 m, in the central part of the MBM, close to the Italian border (figure 1).Two major faults oriented N68E, passing by the Freshfield Pass and at the footwall of the Bernezat Spur, define three distinct areas from a fracturing point of view (figure 2(a)).To the South of the Freshfield Pass, fracturing is almost vertical and very dense.To the North, the Bernezat Spur is formed of a rather compact rock.The center of the rock wall can be divided in two parts: the upper part and lower part.The upper part, formerly covered by an ice apron (IA hereafter, figure 2(b)), defined as small ice bodies lying on slopes >40 • (Guillet and Ravanel 2020, Kaushik et al 2022, Ravanel et al 2023), depicts a complex pattern formed by a block chaos lying on a complex substratum.The lower part shows at least three fracture sets: one very steep with a N-trend, one parallel to the sub-vertical N70 • E major faults, and one almost horizontal.TREF has a mean slope angle of 62 • , varying from 45 to 85 • .
More than 350 archive photographs of the TREF from the late 1850s onwards (figure 3) show the evolution of the Glacier du Géant and the IA on the top of TREF.Analysis of these photos shows a decreasing ice coverage and the continuous thinning of the Glacier du Géant since the mid-1980 (figure 2(b), Fischer et al 2015, Vincent et al 2017).
Tour Ronde was first climbed in 1867 and its normal route is a popular and relatively easy mountaineering route on the margin of the east face (figure 2(a)).In recent years the use of this route is declining in summer due to insufficient or no snow cover and hazardous conditions, making the Freshfield ridge a safer choice (Mourey et al 2019).The Bernezat spur was another well-known route.

TLS data acquisition and processing
TLS technology is based on the acquisition of point clouds of topography using a time-of-flight distance measurement of an infrared laser pulse (see supplementary materials for details).It is a powerful method to monitor rock wall erosion (Rosser et al 2005, Oppikofer et al 2009, Ravanel et al 2014).Our TLS campaigns are most generally carried out at the end of the summer to ensure minimum snow cover on the rock walls, with data acquisition from the Glacier du Géant surface.Over the 18 year monitoring period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022), the TREF was scanned 16 times with an average acquisition of 10 000 000 points and an accuracy of 7 mm at a distance of 100 m.
The obtained individual point clouds were aligned using CloudCompare software (Girardeau-Montaut 2015) to get a full high-resolution 3D-model of the rock wall.The difference between the most recent acquisition and an older one allows morphological changes to be mapped (figure 4).Once differences resulting from changes in glacier, IA, and snow cover are removed, rockfalls are visually identified and their volumes computed using Poisson surface reconstruction algorithms (Kazhdan et al 2020).Every identified rockfall was measured using the software Cyclone 3DR (supplementary table 3; Leica Geosystems and Hexagon 2023).The erosion rate is the total volume lost between two successive field campaigns, divided by the smallest surface of acquisition of the rock wall (see supplementary materials for details).
Since only rockfalls with a minimum volume of several m 3 are relevant to study the potential effects of glacier retreat and permafrost degradation (Hartmeyer et al 2020a, 2020b, Graber and Santi 2022), rockfalls with volume <1 m 3 have been excluded.

Rockfall magnitude-frequency analysis
The frequency of rockfalls has a non-linear inverse relationship to its magnitude and follows a power law (equation ( 1 where F p (V) is the cumulative number of rockfalls in a given inventory, V the rockfall volume, p the period of interest, a the intercept and b the scaling exponent of the power law.The parameters a and b allow  the comparison of different rockfall inventories from various locations or time periods (Graber and Santi 2022).The value of a gives the overall rockfall activity while the b-value gives information on the relative contribution of small rockfalls.Linear correlations r 2 of power laws and ∆b uncertainty (Aki 1965) have also been calculated (see supplementary material, equation ( 3)).Mean detachment depths, inferred from the mean block thickness, range from 0.1 to 25 m (figure 5).95% of the detachments are <5 m thick, 3% are between 5 and 10 m and 2% are >10 m.Over the entire period, the mean depth is 2.1 m and the median is 1.2 m.For the periods 2005-2008 (first period of acquisition) and 2021-2022 (last period of acquisition), with a total of 34 and 30 rockfalls respectively, the average maximum detachment depth is 1.5 m (median 1.0 m) and 2.1 m (median 1.8), respectively.

Description of the two main rockfalls at TREF
The largest rockfall occurred in August 2015 (www.youtube.com/watch?v=O2LL6fmKXck) on the left side of TREF, in a rather gentle slope area still covered by an IA in 1859 and characterized by a complex fracture pattern (figure 2 A rockfall of 6791 ± 228 m 3 occurred on December 4th, 2018 in the lower part of the face (see supplementary material, figure 2).Tilting has been observed between September 2016 and September 2018.The collapsed pillar was 42 m high and 29 m wide, with a maximum thickness of 6 m.The scar corresponds to a large regular surface with a slope angle of 80 • .Ice occupied the entire width of the scar with a thickness from a few tens of centimeters (bottom) to several meters (top; see supplementary material, figure 2(c)).The 2021-2022 erosion rate is now almost six times higher (47.6 ± 6.1 mm yr −1 ).During the 18 years of the scanning period, the annual erosion rate at TREF has often been higher than at other Alpine rock walls, and it abruptly increased from 2015, with a minimum value of 12 mm yr −1 (figure 9).The mean value of annual rockfall volume has been multiplied by more than eight between 2005-2014 and 2016-2022 (500 and 4527 m 3 yr −1 , respectively; figure 9).

Rock wall erosion rate
The erosion rate of 205 mm yr −1 in 2015 resulted from the large rockfall of August 27.This highlights the need for long-term acquisition to consider the role of large and less frequent rockfalls.

Frequencies and volumes
The power law exponent b obtained for the overall period at TREF is 0.44 ± 0.03, slightly under most of those previously published.The most complete compilation of power laws on 32 rock walls shows a median value of the b exponent of 0.8 (Graber and Santi 2022).Hartmeyer et al (2020b) found a bvalue of 0.64 in the Hohe Tauern range, and Draebing et al (2022) b-values ranging from 0.57 to 0.76 in the Hungerly valley.In the MBM, the b-value has been estimated as 0.77 ± 0.04 (Ravanel et al 2017) but this inventory is not exhaustive as the methods used (Ravanel and Deline 2013) under-represent volumes <300 m 3 , and possibly volumes >12 × 10 3 m 3 due to the short study period.The b-value of Guerin et al (2020) and TREF, based on the same methodology, are consistent (0.48 ± 0.03 and 0.44 ± 0.03, respectively).
Since a greater the b-value corresponds to a greater relative contribution of small rockfalls, a b-value of 0.4-0.5 means that the relative contribution of large rockfalls for these rock walls is greater than on other studied rock walls in the European Alps, at least for rockfalls <10 000 m 3 .
At TREF, the b-values show a decreasing trend (figure 7), which could illustrate the increased proportion of large rockfalls in the overall inventory that could be related to deeper rock detachment: the mean thickness increased from 1.5 to 2.1 from 2005-2008-2021-2022 (figure 2(b)).It would be tempting to link this increase to a deepening of the active layer due to the climate warming, but the following detailed analysis shows that processes are more complex.

Preconditioning factors
Review on rock slope stability (McColl 2012, Hantz et al 2021) argued that the structure of the rock walls (pre-existing tectonic joints, foliation, degree of fracturing) is of critical importance.The b-value reflects processes which are scale-invariant and possibly the rock mass structure (Turcotte 1986).The MBM granite is a very hard rock with continuous and regular brittle fractures that separate blocks generally several tens of meters wide.This can explain the high proportion of large rockfalls, evidenced by our b-value (0.44 ± 0.03).Nevertheless, the downward trend of the b-value suggests an influence of climatic conditions (Graber and Santi 2022).In addition, the TREF rockfalls occurred both in the weakly fractured Bernezat Spur and in more fractured areas, suggesting that the climatic factor could sometimes override the geology.
When shrinking, glaciers and IAs free boulders and rock masses become unstable (Krautblatter et al 2013).Ravanel et al (2023) found that after the 2003 and 2015 summer heatwaves, rockfall deposits heavily accumulated at the foot of recently deglaciated areas in the MBM due to retreat of IAs.Permafrost degradation at TREF was enhanced by the reduction of the IA surface area.Eleven rockfalls occurred in this area, accounting for 17% of the total volume during 2005-2022.In detail, 27%, 70%, 9% and 93% of the total rockfall volume detached from the upper deglaciated area in 2007-2008, 2016-2018, 2018-2019 and 2019-2021, respectively.Glacial debuttressing is a significant preconditioning factor (Grämiger et al 2017, 2018): the lowering of the glacier surface induces a removal of the ice load and consequently a stress-release at the base of the rock wall (Ballantyne 2002, Cossart et al 2008).This debuttressing may lead to failure along critically stressed discontinuities (Davies et al 2001).In

Triggering factors
The seismic activity in the vicinity of Tour Ronde east face between 2005 and 2022 is far from negligible but the intensity did not exceed IV on the European Macroseismic Scale (EMS-98) at the surface and caused no damage (Bureau Central Sismologique Français 2023), thus having minimal impact on the rock wall.
Extreme meteorological conditions with high temperature amplitudes and sudden temperature changes may result in rock wall instability in high-Alpine rock walls (Hall 1999, Matsuoka and Murton 2008, Draebing and Krautblatter 2019) that are unexpected in their location, magnitude, frequency, and timing.The year 2022 has been extremely hot in the Western Alps, particularly at high elevation.During the winter and spring, Chamonix experienced a rainfall shortage of 136 mm, while May was the warmest month since 1900 (3.5 • C above the average) and the driest since 1959.In mid-June, while the snow cover was very low, the earliest recorded heat wave started with air temperatures reaching 10.4 • C at 4750 m (Blard and Agenzia Regionale Protezione Ambiente Valle d'Aosta 2022).From July, access to mountain huts and famous Alpine summits became more difficult or even too risky.On August 24, the Fourche hut (3674 m) 1.5 km West from the TREF collapsed.The year 2022 has been exceptional but followed the trend of the summers since 2015, with hot-even scorching-summers having become the norm.Their influence is clearly recorded in the abrupt increase of the TREF rockfall occurrence since 2015.
As observed in several detachment scars, fractures in permafrost-affected rock walls are likely to contain ice and experience strong changes during warming and thaw, while freeze/thaw cycles in the active layer weaken the fractures.This can be a preparat-   The increase in rockfall activity and erosion rate highlights a climatic driver.This has a significant impact on mountaineering.Such impact will likely increase with the predicted increase in heatwave duration and frequency (IPCC 2018), while high-elevated tourist infrastructure like huts and cable-cars in the European Alps may be threatened.Therefore, the TREF dataset could serve as a basis for modeling rockfall dynamics in high mountain regions for the next decades including various climatic scenarios.To support such modeling, more work needs to be done on the rock wall fracturing analysis to study the influence of fractures types and distribution on block stability.Similarly, continuous temperature data extending over a longer period would be beneficial for a better understanding of the processes.

Conclusions
, Gardent et al 2014).Permafrost covers 45% of the 86 km 2 of rock walls steeper than 40 • on the French side of the massif (Magnin et al 2015a).Air temperature rose by +2 • C at Chamonix (1042 m) between 1980 and 2020.
Mean annual rock surface temperature (MARST) modeling of the whole face is around −2 • C (Magnin et al 2015a).Sub-surface thermistors recorded temperature at TREF between 2006 and 2009 (supplementary figure 1) and showed mean temperatures of −0.5 to −1.1, −0.7 to −1.3, and −1.1 to −1.7 • C, at depth of −3, −30, and −55 cm, respectively.Temperature ranges from −19 • C to +14 • C at the surface, and from −12 • C to +9 • C at −55 cm.Compared to Aiguille du Midi (3745 m, Magnin et al 2015b), TREF has a warmer permafrost thermal regime, with likely a thicker active layer which is at least 3 m thick.Furthermore, thermal modeling performed by Legay et al (2021) at the location of the largest rockfall of TREF (August 2015) indicates a temperature of 0 • C at a depth of around 8 m just before the collapse.

Figure 2 .
Figure 2. Tour Ronde east face in 2022.(a) Fracture types and distribution.Yellow: main faults; red: mountaineering routes; white star: location of the sub-surface thermistors; (b) shrinkage of the Glacier du Géant at the foot of the rock wall and of the ice apron (IA) on the upper part of TREF between 1859 and 2022.

Figure 3 .
Figure 3. Archive photographs of TREF from the end of the little ice age (1859) to 2022.(a) Bisson brothers, Summer 1859 (detail); (b) V. Sella, Summer 1884 (detail); (c) W. Mittelholzer, Summer 1926; (d) IGN orthophoto, July 1995; (e) A. Rabatel, October 2006; (f) September 2011; (g) September 2018.These photographs particularly illustrate the evolution of Glacier du Géant located at the footwall of Tour Ronde.White triangle: summit (3792 m), red arrow: common point of reference, red polygons: pillar that collapsed on December 4th, 2018.The pillar culminates at 3510 m and the distance between TREF's summit and the pillar is 369 m.

Figure 4 .
Figure 4. Workflow of the 3D point cloud comparison method for rockfall detection and volume estimation.Blue: small changes excluded from the interpretation.Green: changes related to glacier retreat (lower part of the face) or IA retreat (upper part).Red: major changes between two acquisitions with more than 10 m distance.
(a)).Ravanel et al (2017) estimated its volume at 15 000 ± 3 000 m 3 .Our study measured a volume of 15 578 ± 188 m 3 with an average thickness of 17 m.The MARST modeled there is between −1.6 • C (Legay et al 2021) and −2.3 • C(Ravanel et al 2017).According toLegay et al (2021), the temperature averaged over one day before the collapse was 10 • C at the surface and 0 • C at around 8 m depth.Massive ice and water flows were observed in the detachment scar.
Glacier du Géant and the IA, and the fracturing pattern Photo-comparison (figure 2(b)) shows the lowering of the surface of the Glacier du Géant between 1859 and 2006, averaging 10 m between 2006 and 2022 (min: 0.9, max: 23.0 m; figure 8(a)).The surface area of the IA on the steep upper part of the face decreased by 68% (approximately 16 100 m 2 in 2006 and 5170 m 2 in 2022), hence dropping from 11%-3% of the total rock wall surface.Rockfall locations illustrated on figure 8 show that: • Before 2018, only three rockfalls detached from areas deglaciated since 2006.• During the 2018-2019 period, five and four rockfalls occurred, respectively, in the area uncovered by the IA and the Glacier du Géant since 2006, corresponding to 9% and 75% of the rockfall volume for this period.• Between 2019 and 2021, three rockfalls occurred from the area uncovered by the IA, corresponding to 93% of the period volume, whereas no rockfall occurred in the Glacier du Géant deglaciated area.• During the 2021-2022 period, six rockfalls (32% of the volume) were identified in the Glacier du Géant deglaciated area.• 18 rockfalls out of 210 (around 10% of the total rockfall volume) affected the very weakly fractured area of the Bernezat Spur.

Figure 6 .
Figure 6.Magnitude-frequency distribution for all rockfalls between 2005 and 2022 (red dots).For comparison, black squares are 66 rockfalls from Ravanel et al (2011) in the Mont-Blanc massif, including eight rock walls located between 3320 and 4165 m (total volume: 2223 m 3 ).
In the Hohe Tauern range culminating at 3203 m (Austrian Alps), Hartmeyer et al (2020b) measured a mean annual erosion rate of 1.9 mm yr −1 over six years, with a maximum of 10.3 mm yr −1 .In the Hungerli valley (Swiss Alps), Draebing et al (2022) found a maximum erosion rate of 1.7 mm yr −1 over three years for rock walls at 3100-3200 m.On rock walls in the Ravanel et al (2011) measured erosion rates ranging from 0.03 (Piliers du Frêney, over three years) to 6.3 mm yr −1 (Aiguille Blanche de Peuterey, over three years).Guerin et al (2020) found on the west face of the Drus (2730-3730 m) an erosion rate of 14.4 mm yr −1 over 11 years.The mean erosion rate of 18.3 ± 0.2 mm yr −1 at TREF is thus the highest measured in the European Alps.Rabatel et al (2008) found an erosion rate of 8.4 mm yr −1 at TREF between 2005 and 2006.

Figure 7 .
Figure 7. Power law's b-value depending on various binning groups.Black circles: class A. Blue circles: class D. Red dot: mean b for the overall period.Vertical lines: uncertainty.Grey boxes and dashed blue lines: period encompassed by the binning group.

Figure 9 .
Figure 9. Evolution of the total rockfall volume (m 3 ) and the erosion rates (mm yr −1 ) throughout the monitoring period (2005-2022).Note that the scan surface varied from one year to the other and the two curves are not homothetic.Dotted lines: mean annual cumulated rockfall volume for 2005-2014 and 2016-2022.
ory and sometimes triggering factor to the rockfall occurrence (Gruber and Haeberli 2007).Legay et al (2021) and Ravanel et al (2023) concluded that shallow rockfall sources (<4-6 m depth) would result from daily and seasonal freeze/thaw cycles within the active layer, whereas active layer thickening due to permafrost degradation would involve deeper source areas.The increase in the mean rockfall thickness from 1.5 to 2.1 m at TREF suggests a response to the 1 • C increase of the mean annual air temperature (MAAT) from 2006-2008-2021-2022. Between 2005 and 2014, the mean summer air temperature Chamonix was 16.2 • C, well above the MAAT of 7.9 • C. Between 2016 and 2022, these means rise to 17.6 • C and 8.8 • C, respectively (figure 10).The depth of 95% of TREF rockfalls is <5 m, i.e. in the active layer.They could be linked with daily surface temperature changes, or seasonal freeze-thaw cycles (Magnin et al 2023).3% of the TREF rockfalls detached below 5 and 10 m.Permafrost degradation linked to the active layer thickening could be their triggering factor (Legay et al 2021, Ravanel et al 2023), as these deep scars generally contain ice.Finally, 2% of the rockfalls are >10 m thick, i.e. much more than the estimation of the active layer depth.The largest rockfall (27 August 2015) occurred during a heat wave.Modeling gives a temperature of 0 • C at a depth of 8 m (Legay et al 2021), much shallower than the 17 m-thick collapsed rock mass.This difference suggests (i) an entrainment effect: a large proportion of the blocks of varied thickness that form the slab were in the active layer, or (ii) water percolation in the source area, resulting in heat advection (Hasler et al 2011, Magnin and Josnin 2021).The second largest rockfall at TREF (4 December 2018) occurred at the winter's onset.A first motion occurred between 2017 and 2018, with 80 cm thick ice in the source area.Thermal and mechanical ice changes acted as a preconditioning factor, whereas the deep penetration of heat during the summer and autumn would have triggered the rockfall (Magnin et al 2015b).

Figure 10 .
Figure 10.Deviation from the mean annual and summer air temperatures in Chamonix (1042 m) and erosion rates throughout the monitoring period (2005-2022).JJA: June-July-August.Dotted lines: mean deviation from the mean annual and summer air temperatures for 2005-2014 and 2016-2022.

A
18 year rockfall inventory (2005-2022) of the east face of Tour Ronde (3440-3792 m), based on 15 TLS acquisitions, provides unprecedented insights into rockfall dynamics in deglaciating and warming permafrost-affected rock walls.The main results of the analysis of this inventory are: • The mean annual rock wall erosion rate of 18.3 mm yr −1 at TREF between 2005 and 2022 is much higher than comparable rates from European Alps.Two distinct periods are evidenced, with a mean annual rockfall volume of 500 m 3 yr −1 for 2005-2014 and 4527 m 3 yr −1 for 2016-2022.• The rockfall magnitude-frequency relationship follows a power law with a b-value of 0.44 ± 0.03, which has decreased from 0.55 ± 0.07 in 2005-2009 to 0.37 ± 0.05 in 2019-2022.This decrease could be linked with the increase of the relative contribution of large rockfalls since 2005.• The glacier shrinkage at the foot of the rock wall favored rockfall activity: 75% of the total volume for 2018-2019 and 32% for 2021-2022 detached from areas freed up by a lowering of around 10 m of the glacier surface elevation during the last 15 years.The ice apron surface area, reduced by two-thirds since 2006, also increased the rockfall occurrence with 9% and 93% of the period rockfall volume originated from the upper deglaciated area for 2018-2019 and 2019-2021, respectively.• Ninety five percent of the rockfalls are <5 m thick.Thermal modeling suggests that most of the rock mass was located in the active layer before collapsing, which implies a dominant role for freeze/thaw cycles.The increase of annual rockfall volume since 2015 suggests that summer heatwaves have more influence than the increase in the average annual temperature (0.5 • C 10 yr −1 in Chamonix).

Table 1 .
Main characteristics of the rockfall inventory.

Table 2 .
Estimate (E) of the distribution of the rockfall volume during the 2011-2015 period, inferred from the cumulated volume between 2011 and 2015 and the occurrence of the large rockfall of 27th August 2015.