3.1. Distribution of ice aprons

A combination of very high-spatial resolution images like the orthophotos available from www.geoportail.gouv.fr (0.4 m resolution), panchromatic images from SPOT 6 (1.5 m resolution) and the high-resolution images available from Google Earth were accurately used in order to manually digitize IAs in the MBM under a GIS environment (Kaushik and others, Reference Kaushik2021, Reference Kaushik, Ravanel, Magnin, Trouvé and Yan2022a). Since IAs have a slope >40° (slope angle regularly used as a minimum value to define rock walls), they might be confused with avalanching glaciers but they do not have a vertical front from which blocks of ice (‘séracs’) break off, nor do they have crevasses. The upper boundary of an IA is generally indicated by the ice–rock transition. The lower boundary is either marked by such an ice–rock transition (when IA is located within a rock face) or by a bergschrund (when the IA stands on top of a glacier of a different type; see Schweizer, Reference Schweizer1988) (see e.g. Fig. 9), i.e. a crevasse that forms between the relatively stable IAs and the more mobile glacier downstream.

A total of 423 IAs were identified by Kaushik and others (Reference Kaushik, Ravanel, Magnin, Trouvé and Yan2022a) (Fig. 3), corresponding to a total of 4.21 km² which is ~0.8% of the massif surface area and ~2.6% of the total surface area covered by ice (162.4 km² according to Kaushik and others, Reference Kaushik, Ravanel, Magnin, Trouvé and Yan2022a). Regarding the surface area of IAs, the statement of Guillet and Ravanel (Reference Guillet and Ravanel2020) defining IA <0.1 km² is confirmed since the values in the MMB are from 0.001 to 0.05 km². This upper limit does not seem restricted to the MMB since very large IAs like the main one on the Shivling north face (6543 m a.s.l.; Garhwal Himalaya, India; Fig. 1f) or on the west face of Tocllaraju (6034 m a.s.l.; Cordillera Blanca, Peruvian Andes; Fig. 1g) have a surface area ~0.1 km². Problems associated with the lack of coverage over the entire MBM, presence of cloud cover, illumination and shadow area issues, which make visual interpretation difficult and the presence of seasonal snow cover were all overcome using images acquired through different modes (Kaushik and others, Reference Kaushik, Ravanel, Magnin, Trouvé and Yan2022a).

According to the same authors, 87.5% of IAs are located above 3100 m a.s.l. (Fig. 4a), i.e. above the regional equilibrium line altitude (Rabatel and others, Reference Rabatel, Letréguilly, Dedieu and Eckert2013). The lowest ice bodies are generally located at the foot of cold north faces. IAs are present in all orientations (Fig. 4b) with a maximum on northwest faces (19.4%), then north (16.3%) and, counter-intuitively, south (14.4%), where the IAs are located at higher altitude than the north-facing IAs. In total, 58% of the IAs have a slope angle between 50° and 65° (Fig. 4c) like IAs of Triangle du Tacul, Tour Ronde or Grandes Jorasses. For some, slopes are >75°. Comparable work outside the MBM and the alps is necessary to validate and complete the results of this first regional survey, in particular to verify the relative part of IAs within the ice-covered areas.

Fig. 4. Topographical characteristics of IAs in the MBM (Kaushik and others, Reference Kaushik2021, modified): (a) elevation; (b) aspect and (c) slope angle; values correspond to the number of IAs. The analysis was performed using ASTER global DEM available at 25 m resolution.

3.2. Four types of occurrence of ice aprons

To better understand the location of IAs on rock walls and their spatial relationships with glaciers, four types of IA formation contexts can be observed in the MBM based on the survey of Kaushik and others (Reference Kaushik, Ravanel, Magnin, Trouvé and Yan2022a):

1. (1) type 1 is ‘long-standing exposed’ IAs ((1) in Fig. 1), i.e. IAs having continuously existed independently of any glacier outside ice ages; they are overall receding (Section 4.1);

2. (2) the retreat of glacier bodies releases IAs possibly previously buried under the ice (type 2), including at very high altitudes (Fig. 1b); for example, retreat of valley or slope glaciers frees sectors at the foot of steep back-walls ((2) in Figs 1b, c), or lateral rock slopes on which ice remains ((2) in Fig. 1d);

3. (3) type 3 corresponds to an evolution of what Galibert (Reference Galibert1965) called ‘north faces with quasi-continuous snow/ice shells’: these pretty steep slopes that used to be snow-covered all year long become steeper as the snow and then the ice melt; the surface of the IA is then completely freed from snow, temporarily or permanently ((3) in Fig. 1);

4. (4) type 4 corresponds to IAs that overlook hanging, slope or cirque glaciers; these IAs tend to lose their snow cover; it often allows the distinction between steep IAs ((4) in Figs 1b, c) and glaciers, the latter being gentler, snow-covered and whose mobility is shown by crevasses.

4.1. Variations in ice apron surface area since the 19th century

Using oblique historical imagery and a specifically developed photogrammetric inverse perspective method for extracting physical landscape measurements (see Guillet and others, Reference Guillet, Guillet and Ravanel2020), Guillet and Ravanel (Reference Guillet and Ravanel2020) documented the variations in surface area for six IAs from four north-oriented steep rock walls within the MBM since the termination of the LIA (see Fig. 3): one at Triangle du Tacul (3970 m a.s.l.) between 3570 and 3690 m a.s.l. (0.004 km² in 2018; slope angle: ~59°); two at Tour Ronde (3792 m), 3343–3594 m and 3600–3750 m (0.011 km² in 2018 and 0.005 km² in 2012, respectively; both ~55°); two at Grandes Jorasses (4208 m), 3220–3650 m and 3500–4000 m (0.016 km² in 2016 and 0.028 km² in 2017, respectively; both ~63°) and one at Aiguille des Grands Charmoz (3445 m), 2907–3185 m (in 2015; 0.0105 km² in 2012; ~52°).

Shrinkage is documented over all the considered IAs from the end of 1850s to the mid-to-late-1960s (Fig. 5). This period of shrinkage is followed by a short period of IA steadiness or expansion, from the early 1970s to the early 1990s. The Triangle du Tacul IA, for example, gained 11% of surface area between 1966 and 1988. This is certainly due to the significant precipitation and cooler temperatures than today (Auer and others, Reference Auer2007) which allowed alpine glaciers to gain volume (see e.g. Vincent, Reference Vincent2002). Since the beginning of the 21st century, the study confirms the observations of the mountaineers, namely an accelerated shrinkage over all the six IAs. The shrinkage recorded since the termination of the LIA strongly varies between the different IAs. The upper IA of Grandes Jorasses experienced the lowest shrinkage, ~20% of the post-LIA surface area (−7400 ± 200 m²), while other IAs show higher relative losses: 35% for Triangle du Tacul (−1940 ± 158 m²), 40% for the upper IA of Tour Ronde (−3500 ± 147 m²), 50% for the lower one (−1100 ± 127 m²) and 60% for the lower IA of Grandes Jorasses (−28 440 ± 550 m²) (Fig. 5). The most dramatic shrinkage affected the IA of Grands Charmoz, which melted completely during the summer of 2017 (Guillet and Ravanel, Reference Guillet and Ravanel2020).

Fig. 5. Variations of surface area of six IAs in the MBM since the termination of the LIA (Guillet and Ravanel, Reference Guillet and Ravanel2020). Error bars represent the 95% confidence interval. Elevations are in m a.s.l.

Using very high-resolution optical images, Kaushik and others (Reference Kaushik2021) have quantified the surface area of the IAs in the MBM for 2012 (Pleiades 1B PAN and ortho-images IGN – Institut national de l'information géographique et forestière) and 2019 (Pleiades 1B PAN and Sentinel-2). Although manual mapping procedures are time-consuming, they enabled to map outlines of surface ice with the highest precision, an operation which remains difficult with automatic mapping, especially for very small snow/ice features (e.g. Paul and others, Reference Paul2013). The work has recently been completed by Kaushik and others (Reference Kaushik2022b) by including the years 1952 and 2001 from ortho-images IGN. The total surface area of IAs in 1952 was 7.93 km² and then dropped to 5.91 km² in 2001, 4.92 km² in 2012 and 4.21 km² in 2019 (Fig. 6). From 1952 to 2019, IAs have lost ~47% of their surface area, corresponding to an average rate of surface area loss of 0.78 km² decade⁻¹. The area loss was 25.4% from 1952 to 2001 compared to 30.6% from 2001 to 2019 (Fig. 6).

Fig. 6. Evolution of the total surface areas of IAs in the MBM over seven decades.

4.2. Significant recent changes in ice apron thickness

IAs are small ice masses located in mainly concave topographies (Kaushik and others, Reference Kaushik2022b). The loss of surface is therefore related to the loss of thickness.

Mountaineers have widely experienced this loss through: (1) the disappearance of ice in many climbing routes (Mourey and others, Reference Mourey, Marcuzzi, Ravanel and Pallandre2019) or even on an entire rock wall, as in the cases of the north face of Aiguille des Grands Charmoz in 2017 (see Section 4.1) or the famous niche in the north face of Les Drus (3754 m a.s.l.) of which only a curved rest of ice remained in the most shaded area at the end of the recent hot summers of 2019 and 2022; (2) the emergence of spurs and boulders within IAs that were recently still continuous (e.g. at Triangle du Tacul where a spur appeared in 2003–04 in the centre of the lower IA and where new rocks now appear almost every year; see black dashes in Fig. 8); (3) the disconnection between current IA surfaces and mountaineering equipment used until the very beginning of the 2000s (Mourey and others, Reference Mourey, Perrin-Malterre and Ravanel2020). In the Gabarrou-Marquis route at Triangle du Tacul for example, ~6.2 m of ice had melted in 2020 since the early 2000s when a belay, now inaccessible, was still used on the right rim of the IA (Fig. 7a).

Fig. 7. Two examples of the thickness loss of IAs. (a) Loss of ~6.2 m of ice thickness at the Triangle du Tacul between the beginning of the 2000s, when permanent mountaineering equipment could be reached from the surface of the ice, and 2020. (b1) Left sidewall of the Argentière Glacier in 1952 (IGN orthophoto); (b2) in 2019 (Spot7 image); (b3) part of the IGN topographic map 3630OT Chamonix 1 : 25 000 that indicates (brown contours) the shrinkage of several IAs.

Such ice loss can also be revealed locally when topographic maps are updated. Since 2012, the IGN 3630OT Chamonix map 1: 25 000 (IGN, 2017) contains a new brownish tint corresponding to ‘glacier retreat between 1980 and 2009’ derived from aerial images of 2008 (Fig. 7b). Although not usable for our study given the steepness of the slopes, the new colours suggest the loss of IA area and thickness, confirmed by the comparison of aerial photos (Fig. 7b).

To quantify the IA-wide thickness loss and its spatial variability, we determined the annual mass balance for 2017 using stakes in the lower Triangle du Tacul IA (Fig. 8; aspect: 355° N, surface area: 0.04 km²; slope angle: 59°). Ten 2 m-long fibreglass stakes (1 m in the ice to measure the probable melt and 1 m outside to measure the possible accumulation) were installed on 29 March 2017. They were checked on 25 September 2017 and 16 March 2018. The hot summer of 2017 consistently melted 42–48 cm of ice over the whole IA (excluding the area covered by snow in March 2017 and areas where snow avalanches and boulder falls destroyed the stakes; Fig. 8), i.e. a mean melting of 0.4 ± 0.03 m w.e. No mass gain or loss was detected between September 2017 and March 2018. Thus, we measured over 1 year a significant summer loss and no accumulation. This monitoring and previous observations (Guillet and Ravanel, Reference Guillet and Ravanel2020) suggest that summer melting combined with the absence of accumulation has probably been a fairly general trend since the early 2000s. The efficiency of the summer melting is corroborated by Kaushik and others (Reference Kaushik2022c) who investigated the evolution of the physical backscatter properties and surface changes of IAs by exploiting time series of X- and C-band synthetic aperture radar (SAR) images from PAZ and Sentinel-1 satellites. SAR data indeed provide information about the seasonal behaviour of IAs since physical changes of IA surfaces modify the backscattering of radar waves. The analysis of the temporal variations of the backscatter coefficient illustrates the effects of summer air temperatures on the surface of the IAs with a strong decrease in coefficient values indicating an increase in the liquid water content and thus the melting. Accumulation on IAs is severely hindered by surface topography, as only snowfall occurring within the 0 to −5°C temperature range can accumulate on slopes steeper than 30° (Guillet and Ravanel, Reference Guillet and Ravanel2020; Eidevåg and others, Reference Eidevåg, Thomson, Kallin, Casselgren and Rasmuson2022). With the growing number of summer (Beniston and others, Reference Beniston2004; Zampieri and others, Reference Zampieri2016) and winter (Beniston, Reference Beniston2005; Colucci and others, Reference Colucci, Giorgi and Torma2017) heat waves, such conditions have become rarer while smooth ice surfaces are not conducive to snow deposition under windy conditions. IAs are thus becoming particularly distinguishable (grey ice) in the MBM from other types of less steep and still snow-covered ice masses (see Fig. 1). The evolution of the colour of the IAs and their very high retreat rates over recent period contribute greatly to the darkening of the high-mountain slopes and give rise to fears of numerous disappearances in the near future.

Fig. 8. Effects of a hot summer on the evolution of IAs: the example of the summer 2017 thickness loss in Triangle du Tacul. Elevation in m a.s.l. Outside areas covered with snow in early summer, mass balance was −0.4 ± 0.03 m w.e (only ablation occurred during the year).

4.3. Short-term effects on rock slope stability

Rockfalls >100 m³ have been surveyed in the MBM since 2007 (Ravanel and Deline, Reference Ravanel and Deline2013; Ravanel and others, Reference Ravanel, Magnin and Deline2017a). Over the recent years, many small events have occurred in recently deglaciated areas due to retreat of IAs: following 2015, 2017, 2018, 2019, 2020 and 2022 summer heat waves, rockfall deposits have heavily accumulated at the foot of ice-covered rock walls (Fig. 9). Besides boulders formerly embedded into the ice and released as the ice melts, a high rockfall frequency from the margins of the shrinking IAs is observed. Ice vanishing often exposes perennially frozen and heavily fractured rock that the disappearance of the ice in the form of a cover or joints no longer allows to maintain on the slope (Davies and others, Reference Davies, Hamza and Harris2001; Krautblatter and others, Reference Krautblatter, Funk and Günzel2013; Mamot and others, Reference Mamot, Weber, Schröder and Krautblatter2018; Kenner and others, Reference Kenner, Arenson, Grämiger and Shroder2022). At the scale of larger alpine faces like Monte Rosa east face (4634 m a.s.l., Italian Alps), Fischer and others (Reference Fischer, Eisenbeiss, Kääb, Huggel and Haeberli2011) showed that the sequence of the largest slope failures occurring during the last few decades is spatially connected to disappearing surface ice and further demonstrated the coupling between permafrost-related instability and the condition of adjacent ice masses. The case of the north face of Tour Ronde in 2020 sheds light on this point. An automatic camera was used to monitor the evolution of the face throughout the summer season (Fig. 10). Totally snow covered at the beginning of the summer, the IA underwent notable melting in July and a first rockfall occurred from a recently deglaciated area on 31 July. A second event occurred on 4 August, from an area directly upstream which had still been covered by ice a decade earlier. Later, a third rockfall occurred on 27 August just upstream of the previous one, in an area also covered by ice several decades ago. The two last events exposed massive ice in their scar (Fig. 10), confirming a permafrost context (Cremonese and others, Reference Cremonese2011). This ice suggests a rockfall related to permafrost degradation (e.g. Gruber and Haeberli, Reference Gruber and Haeberli2007; Ravanel and others, Reference Ravanel, Magnin and Deline2017a). A rockfall of ~12 000 m³ (estimation from measurements on a 3-D model acquired by terrestrial LiDAR) finally took place on 13 August 2022 after three long heat waves in the western European Alps (the summer of 2022 was the hottest ever recorded in France and in Chamonix – data Météo France) causing a probably unprecedented IA melting and a very significant heat input for the permafrost. While the melting of IAs likely destabilizes rock volumes almost immediately, the retreat of IAs also implies the development of a new direct air/rock connection in deglaciated areas – for the very first time in several millennia (see Section 6.2) –, generating permafrost degradation (active layer development, i.e. the formation of a sub-surface layer that thaws each year, and a deep and long-term warming from the surface), then triggering rockfalls (see also Haeberli and Gruber, Reference Haeberli, Gruber and Margesin2009; Hartmeyer and others, Reference Hartmeyer2020). The subsurface response of permafrost can be all the faster as the ice gives heat to the surrounding rock while melting (Hasler and others, Reference Hasler, Gruber, Font and Dubois2011). IAs are thus first-order protective layers for permafrost (see e.g. Wegmann and others, Reference Wegmann, Gudmundsson and Haeberli1998) since the thermal conductivity of the rock is much better than that of the ice and a thermal inertia arises from latent heat (Noetzli and others, Reference Noetzli, Gruber, Kohl, Salzmann and Haeberli2007; Harris and others, Reference Harris2009; Hasler and others, Reference Hasler, Gruber, Font and Dubois2011). Heat therefore does not penetrate the ice well. Thus, the temperature variability at ~8 m depth (see Section 5) is much lower under the ice of the Tour Ronde IA than in the rock at the Aiguille du Midi (Magnin and others, Reference Magnin, Deline, Ravanel, Noetzli and Pogliotti2015b) with a rather similar orientation and altitude.

Fig. 9. Rockfall deposits at the foot of: (a) the north face of Rocher du Petit Plateau (3680 m a.s.l.; photo 13 August 2019); (b) the north side of Col du Diable (3770 m; photo 8 August 2018); (c) the north face of Pointe Lachenal (3565 m; photo 13 August 2019) and (d) along the Brenva ridge on the French–Italian border (SPOT6 image, late summer 2018). Areas between the upper limits of the IAs in 1978 (blue) and 2018 (red) are the main source of the rockfalls.

Fig. 10. Evolution of the north face of Tour Ronde during the summer 2020 and in 2022. The lower IA, covered with snow at the beginning of the summer 2020 is almost snow-free at the end. Three rockfalls occurred from the upper limit of the IA: on 31 July and on 4 and 27 August (stars: sources). The last two exposed massive ice (arrows) in their scar. Light blue and blue dashed lines: IA limits in 2009 and the 1880s, respectively. A much larger rockfall finally occurred on 13 August 2022.

5.1. First two boreholes drilled in ice aprons

In high mountains, temperature measurements in boreholes have been used routinely to study the thermal regime of glaciers (Fisher, Reference Fisher1955; Haeberli and others, Reference Haeberli, Schmid and Wagenbach1988, Reference Haeberli, Frauenfelder, Kääb and Wagner2004; Haeberli and Funk, Reference Haeberli and Funk1991; Lüthi and Funk, Reference Lüthi and Funk2001; Suter and others, Reference Suter, Laternser, Haeberli, Frauenfelder and Hoelzle2001; Suter and Hoelzle, Reference Suter and Hoelzle2002; MacDonell and others, Reference MacDonell, Kinnard, Mölg, Nicholson and Abermann2013), rock glaciers (Arenson and others, Reference Arenson, Hoelzle and Springman2002; Haeberli and others, Reference Haeberli2006; PERMOS, Reference Noetzli, Pellet and Staub2019) and rock wall permafrost (Harris and others, Reference Harris2009; Magnin and others, Reference Magnin, Deline, Ravanel, Noetzli and Pogliotti2015b; Pogliotti and others, Reference Pogliotti2015). In the MBM, recent ice temperatures (Vincent and others, Reference Vincent2007a, Reference Vincent2007b) have been compared with older data (Lliboutry and others, Reference Lliboutry, Briat, Creseveur and Pourchet1976; Vincent and others, Reference Vincent2020), while measurements have been carried out at the Tête Rousse and Taconnaz Glaciers because of risks for the valleys (Gilbert and others, Reference Gilbert, Vincent, Wagnon, Thibert and Rabatel2012, Reference Gilbert, Vincent, Gagliardini, Krug and Berthier2015). In order to better understand the thermal regime of IAs and its relationships with the underlying permafrost, two boreholes in IAs were drilled to bedrock and equipped (Figs 11a, b) with a Geoprecision thermistor chain (Dallas DS18S20, accuracy: ±0.25°C, resolution: 0.065°C): on the north face of Tour Ronde (nine sensors up to 880 cm deep, installed in December 2016), and at Triangle du Tacul (four sensors up to 100 cm deep, installed in March 2017) (Ravanel and others, Reference Ravanel2017b). These chains were installed in high-resistance polyurethane (PU) tubes cased in October 2016 in boreholes drilled normal to the ice surface using a portable steam-driven ice drill (Heucke, Reference Heucke2003). Foam pads were inserted between the sensors to prevent heat advection by air. Temperature was recorded every 4 h by a Geoprecision logger (Dallas M-Log5W). Setting up required significant mountaineering skills, but maintenance was even more challenging. At Triangle du Tacul, the summer 2017 heat wave melted almost half of the thickness of the IA (see Section 4.2) at the borehole location, exposing three sensors to free air. At Tour Ronde, the early heat wave of summer 2017 produced rockfalls from the upper margin of the IA that strongly damaged the upper part of the borehole. The following winter, an avalanche destroyed the logger.

Fig. 11. Thermal regime of two IAs in the MBM. (a, b) borehole location and thermistor chain at Tour Ronde and Triangle du Tacul. (c, d) Tour Ronde and Triangle du Tacul daily temperatures at different depths; at Tour Ronde, the jump in temperature at 8.8 m depth in June 2017 (red box) is certainly linked to liquid water flowing to the bottom of the borehole, following its damage by a rockfall; at Triangle du Tacul, three of the four sensors came out of the ice due to its melting. (e, f) Tour Ronde and Triangle du Tacul modelled daily temperature from surface to 8.8 and 1.0 m deep, respectively.

5.2. Cold ice stuck to the permafrost

At Tour Ronde, we acquired a bit more than 10 months of data (2 December 2016–11 October 2017; Fig. 11c). Near-surface IA temperature is heavily correlated with air temperature, while only seasonal air temperature signal affects ice temperature below 5 m depth. The lowest temperature at 5 m depth is recorded in the early summer (Fig. 11e). From 2 December 2016 to 30 May 2017, sensors at −10, −30 and −50 cm displayed very high variability, with two periods (21 March–13 April; 25 April–24 May) of similar temperatures with reduced variability that suggests the setting of a snow cover (see e.g. Magnin and others, Reference Magnin2017), confirmed by photos automatically taken from the nearby Aiguille du Midi cable-car station. Influx of warm air affects the MBM from the second 10 d period of May, evidenced by the largely positive temperature measured by the 10 cm-deep sensor in the rock on the left margin of the IA, at the borehole altitude. The available heat melted the snow, generating a loss of insulation (see e.g. Magnin and others, Reference Magnin2017), and warmed the exposed ice. This surface change lowered the surface albedo during the time of maximum incoming solar radiation, which at this IA with its exposure of azimuth N12° is especially important during hours of the morning sun. Following 30 May, the three shallowest sensors recorded significant warming of the ice, with temperatures close to the melting point. Between 7 and 12 June, these temperatures became positive as the sensors were exposed to free air. This significant melting was favoured by high air temperature and still the exposure of the IA to direct sunlight. On 15 June, the ice at the level of the sensor −100 cm reached temperature close to the melting point. The sensor seems exposed to free air on 7 July, suggesting a melting of ~1 m of ice between 7 June and 7 July. Initially positioned at −150 cm, the fifth sensor indicated a temperature of 0°C from 8 July but remained buried in the ice (no positive temperature recorded).

The lowest mean temperature (−6.6°C) for the available period was found for the sensor initially located at −250 cm. At the ice–rock interface, the mean 10 month temperature was −4.8°C, confirming the presence of cold permafrost in the rock slope. At this depth (originally −880 cm), the lowest temperature is recorded on 13 June (−6.0°C) while the peak temperature was not yet reached on 11 October 2017 (−4.2°C), when the equipment was damaged. On 2 December 2016, this temperature was −4.4°C, before decreasing: maximum temperature at the base of the IA is therefore likely attained in November. At this depth, a jump in the temperature curve is recorded several days from 12 June 2017 (Fig. 11e), related to a rockfall detected with the Aiguille du Midi camera. It is conceivable that the rockfall damaged the PU tube, allowing liquid water to flow down to the bottom of the borehole, thus temporarily modifying the temperature.

At Triangle du Tacul, temperature was continuously recorded between 1 April 2017 and 31 March 2018 (Fig. 11d). Because of the IA thinness, ice temperature is extremely dependent on air temperature. The sensor at −10 cm recorded significant ice warming from 3 May, leading to its free air exposure on 16 May. The sensor initially at −30 cm indicated ice temperature close to the melting point around 28 May, and was exposed on 12 June. The −50 cm sensor indicated a temperature of 0°C from 12 June, and positive temperature from 16 July, as the mass balance (cf. Section 4.2) indicates a loss of thickness of 46–48 cm near the borehole (Fig. 8). In view of the strong documented ice losses (see Fig. 11f), only the mean annual temperature at the rock–ice interface can be considered: −8.1°C. The annual temperature range is very significant there due to the very thin ice: 0°C on 6 August 2017 (ice temperature almost reaches the melting point during two short periods: 3–5 August and 26 August–3 September), and −20.2°C on 20 February 2018.

These recordings confirm that – outside warm periods during which ice temperature close to the surface can reach 0°C eventually before melting – the IAs are cold ice settled on terrain with negative subsurface temperatures, i.e. stuck to the permafrost-affected bedrock. They are part of the permafrost-related periglacial environment. The basal sliding processes are therefore largely excluded as a consequence of cold-bed conditions, explaining the stability of ice on such steep slopes.

6.1. Simple shear as the main regime for the deformation of ice-apron ice

To observe lattice preferred orientation, as well as grain shape and size, thin sections were sampled in one of the ice cores, using standard protocols for thin section preparation (see e.g. Schwarz and others, Reference Schwarz1981; Durand and others, Reference Durand2006). Thin sections were typically 4–5 cm wide and 0.4–0.5 cm thick. Guillet and others (Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021) used an Automatic Ice-Texture Analyser (Wilson and others, Reference Wilson, Russell-Head and Sim2003) to measure lattice preferred orientation.

Results from Guillet and others (Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021) are presented on pole diagrams in stereographic projections in Figure 12 and quantify the statistical distribution of c-axis orientations (the optical birefringence axis, normal to the basal plane of the ice crystallographic structure; see: Alley and others, Reference Alley, Gow and Meese1995). Guillet and others (Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021) described c-axis distributions clustered around a single vertical maximum with most grains in the analysed samples oriented parallel to the core axis (reddish colours in Fig. 12). Several tilted grains (greenish colours in Fig. 12) could be observed within the ice core. Guillet and others (Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021) further reported a sharp transition in ice lattice preferred orientation at 5 cm above bedrock, with an increase in the concentration of tilted grains, which they interpreted as resulting from the increase in the concentration of impurities in the ice, preventing effective recrystallization. This phenomenon has been widely observed and described in the scientific literature (see e.g. Thorsteinsson and others, Reference Thorsteinsson, Kipfstuhl and Miller1996; Weiss and others, Reference Weiss2002). Statistical distributions of c-axis orientations in ice have been vastly documented and strong lattice preferred orientation are often attributed to the deformation of ice under simple shear and dynamic recrystallization under glacier conditions (see e.g. Hudleston, Reference Hudleston, Saxena, Bhattacharji, Annersten and Stephansson1977; Hudleston, Reference Hudleston2015). As the ice is deformed in simple shear, the single maximum distribution of c-axis orientations is typically oriented normal to the shear plane. This is reinforced by recrystallization (Journaux and others, Reference Journaux2019). According to Guillet and others (Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021), the single maximum pattern observed in Triangle du Tacul ice likely results from a quasi-stationary shear regime, with a plane parallel to the surface slope of the IA.

Fig. 12. c-Axis lattice preferred orientation pole figures and orientation colour-coded images displaying the c-axis orientations for all the thin sections of the Triangle du Tacul ice core (Guillet and others, Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021). Crossed out sections represent destroyed portions of the ice core. Note that depth representation is not to scale and depths have been rounded to the nearest value. In most pole figures, statistical distributions of c-axis orientations are clustered around a single maximum. Second maxima on section 2 is the consequence of the relative weight of the two greenish crystals in the statistical distribution of the sample. c-Axis orientations on the orientation colour-coded images are given by the colour scale on the top left. The colour code on the pole figures represents the density of c-axis orientations at the pixel scale (1 pixel = 20 μm).

6.2. Ice aprons as a glacial heritage

The above considerations and the significant melt of IAs during the recent decades suggest the outcropping ice to be of millennial age. Considering previous work focusing on the age of alpine glaciers (see e.g. Jenk and others, Reference Jenk2009; Gabrielli and others, Reference Gabrielli2016; Uglietti and others, Reference Uglietti2016; Zipf and others, Reference Zipf, Merchel, Bohleber, Rugel and Scharf2016; Feng and others, Reference Feng2019; Preunkert and others, Reference Preunkert2019a, Reference Preunkert2019b; Bohleber and others, Reference Bohleber, Schwikowski, Stocker-Waldhuber, Fang and Fischer2020), Guillet and others (Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021) expected the age of near-bedrock ice in IAs to be ranging between several hundred and several thousand years. A dating method specific to ice containing no vegetal leftover and suitable for potential ice ages older than a few hundred years was therefore considered. The authors specifically used ¹⁴C in particulate organic carbon (POC) associated with a recent sample preparation method successfully applied to near-bedrock ice from cold-based glaciers at high altitude (Hoffmann and others, Reference Hoffmann2018; Preunkert and others, Reference Preunkert2019a) and lower altitude (Bohleber and others, Reference Bohleber, Hoffmann, Kerch, Sold and Fischer2018). They focused on the lower part of the ice cores to avoid potential near-surface age-biasing artefacts, did however omit the lowest few centimetres of the cores to avoid macroscopic impurities – in greater concentration near bedrock (Guillet and others, Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021). Sample preparation (ice decontamination, melting, filtering and combustion) is described by Preunkert and others (Reference Preunkert2019a). POC average concentration of the samples was 330 ± 140 ng C g⁻¹ (Guillet and others, Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021), i.e. 15 times higher than values found in the lowest metres of the neighbouring Col du Dôme Glacier (Preunkert and others, Reference Preunkert2019b), located at 4250 m a.s.l. and 5 km southwest of Triangle du Tacul. This difference was interpreted as resulting from the altitude difference and the presence of biogenic material near the IA during its formation (Guillet and others, Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021). Radiocarbon analyses were performed at the mass spectrometer accelerator at the Curt-Engelhorn Centre Archaeometry (Manheim, Germany). Three ages were obtained: (1) a mean calibrated ¹⁴C-age of 2640 ± 130 year cal bp on the 2018 core between 32 and 52 cm depth, averaging the ice age over 20 cm (Guillet and others, Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021), corresponding to Göschener I, a 3000–2300 year bp alpine cold period (Zoller and others, Reference Zoller, Schindler and Röthlisberger1966; Boxleitner and others, Reference Boxleitner2019); (2) 630 ± 55 year cal bp between 30 and 40 cm depth and 2840 ± 175 year cal bp between 40 and 49 cm on the 2019 core (Guillet and others, Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021). Therefore, the mean 2019 sample age is 1735 year cal bp (with a mean high time gradient of 220 years cm⁻¹), i.e. 950 years younger than the 2018 sample. This difference was explained by the authors as resulting from an imperfect overlap of the covered time periods within the two sampling depths of the two cores due to (1) an inhomogeneous spatial accumulation distribution at this IA, affecting the age–depth model, (2) variability of isochronology distance to bedrock due to sub-glacial topography and (3) spatial variability of the ice melt (ablation). Using the age–depth calculation from the 2019 ice core, Guillet and others (Reference Guillet, Preunkert, Ravanel, Montagnat and Friedrich2021) further suggest an age of ~5500 year bp as a first crude estimate of the very near bedrock ice at Triangle du Tacul.

This age is consistent with other ages of basal ice in high-altitude cold-based alpine glaciers: 7000 year cal bp at Alto dell'Ortles (3905 m a.s.l., central alps, Italy) (Gabrielli and others, Reference Gabrielli2016), 5600 year cal bp at Col du Dôme (Preunkert and others, Reference Preunkert2019b) and 5000 year cal bp at Chli Titlis (3030 m a.s.l, central alps, Switzerland) (Bohleber and others, Reference Bohleber, Hoffmann, Kerch, Sold and Fischer2018) while older ages exist as >15 000 cal bp at Colle Gnifetti (4455 m a.s.l., Monte Rosa, Switzerland) (Jenk and others, Reference Jenk2009; Hoffmann and others, Reference Hoffmann2018; see also: Gäggeler and others, Reference Gäggeler, von Gunten, Rössler, Oeschger and Schotterer1983; Lüthi and Funk, Reference Lüthi and Funk2000). Unlike the aforementioned glaciers, where basal ice is located tens of metres below the surface, IAs possibly offer a quick access with appropriate mountaineering skills to thousands of years old ice.

In the coming years, to better appreciate the palaeo-environmental potential of IAs through the establishment of an age model and the detection of chemical tracers of air pollution sources, to model their effects on the underlying permafrost (new borehole temperature measurements and modelling) and to ensure continuous monitoring of thickness variations (LiDAR or glaciological mass balance) is imperative while, in the meantime, strengthening the effort towards the understanding of IA dynamics (rheology, accumulation/ablation and related local factors) in the alps and outside is needed.