1. Introduction
Glacier evolution can provide relevant information about the characteristics of recent climate dynamics in remote mountain areas with scarce historical observations as they are tightly controlled by the main climatic conditions (Ohmura and Boettcher, Reference Ohmura and Boettcher2018; Del Gobbo and others, Reference Del Gobbo, Colucci, Monegato, Žebre and Giorgi2023). European glaciers have shown, in the last two decades, mass balance losses about two times higher than those observed globally (Hugonnet and others, Reference Hugonnet2021), informing of recent climate warming in European mountains (Beniston and others, Reference Beniston2018). Today, the southernmost European glaciers are located in the Pyrenees (Rico and others, Reference Rico, Izagirre, Cañadas and López-Moreno2017), with 15 glaciers (Izagirre and others, 2024) situated at elevations below the regional equilibrium line altitude (ELA; Braithwaite and Hughes, Reference Braithwaite and Hughes2020). These are classified as very small glaciers (surface < 0.5km2; Huss and Fischer, Reference Huss and Fischer2016), and their evolution is strongly controlled by topoclimatic factors, which explains a contrasted evolution in area and thickness changes despite a common regional climatic signal is observed in this mountain range (Vidaller and others, Reference Vidaller2021). However, 2022 and 2023, characterized by warmer than average temperatures (Serrano-Notivoli and others, Reference Serrano-Notivoli2023), have led to a generalized degradation of all ice bodies putting them in a critical situation.
Recent advances in remote sensing techniques as uncrewed aerial vehicles (UAVs), terrestrial laser scanners and satellite imagery have allowed an unprecedented high spatial and temporal resolution observation of very small glaciers. This has contributed to the exponential increase in the number of studies in the Pyrenees (López-Moreno and others, Reference López-Moreno2016, Reference López-Moreno2019; Vidaller and others, Reference Vidaller2021) and other mountain regions (Vincent and others, Reference Vincent2016; Fischer, Reference Fischer2018; Revuelto and others, Reference Revuelto2018; Bash and others, Reference Bash, Moorman, Menounos and Gunther2020). These studies have allowed a deeper understanding of processes driving glacier dynamics and the identification of major shortcomings that still need to be addressed. The rapid degradation of Pyrenean glaciers has triggered paraglacial and periglacial processes (López-Moreno and others, Reference López-Moreno2019). Paraglacial processes are defined as nonglacial processes that are directly conditioned by glaciation (Church and Ryder, Reference Church and Ryder1972) and thereby are beyond the glacier (Slaymaker, Reference Slaymaker2009). For clarification, this definition diverges from that of periglacial, environment of frequent freeze–thaw cycles and deep seasonal freezing; and/or a permafrost environment (Slaymaker, Reference Slaymaker2009; French, Reference French2017). Intense paraglacial processes have developed in recently deglaciated areas of the Pyrenees, with sediment redistribution and permafrost degradation affecting sediment sources and altering periglacial dynamics.
This favors the development of paraglacial landforms—debris accumulations from surrounding rockfalls, debris-covered glaciers, alluvial fans and debris-flow deposits—andgenerates a paraglacial land system very sensitive to climate shifts, such as variations in the snow regime, isotherm elevation or permafrost degradation (Serrano and others, Reference Serrano2018). Consequently, an increase in the debris-covered area extent, new soils and the formation of proglacial lakes has been observed in the Pyrenees (Vidaller and others, Reference Vidaller2023). Additionally, ice collapses, disconnection between accumulation and ablation areas, and water incisions in ice surfaces confirm that Pyrenean glaciers are in their final stage before disappearing (Serrano and others, Reference Serrano, González‐trueba, Sanjosé and Del Río2011; Martínez-Fernández and others, Reference Martínez-Fernández2023). Some of these processes are also observed in the Infiernos Glacier.
In recent years, the Infiernos Glacier has experienced significant degradation and thinning of its surface (Vidaller and others, Reference Vidaller2021). Additionally a notable accumulation of debris from the surrounding cirque walls took place in 2015 (Cuchí and others, Reference Cuchí, Pomar, Melendo, Vitaller and Jarne2017; Cancer-Pomar and others, Reference Cancer-Pomar, Fernández-Jarne, Cuchí and Del Valle-melendo2023). From these later works, which primarily provide a qualitative description of the changes observed in this glacier, it can also be inferred that while the upper area of the glacier is frequently snow free at the end of the melting period, the lower area remains snow covered. We hypothesize that the Infiernos Glacier can no longer be considered a very small glacier, while the accumulation of debris could give rise to a new paraglacial stage, such as the development of permafrost patches in the surrounding areas. With the purpose of later discussing this hypothesis, the following definitions are established. A glacier is a mass of ice on the land surface, which flows downhill under gravity and is constrained by internal stress and friction at the base and sides (Benn and Evans, Reference Benn and Evans2010). An ice patch is an ice body without movement by flow or internal action (Serrano and others, Reference Serrano, González‐trueba, Sanjosé and Del Río2011). This way, the absence of movement is the main difference between a glacier and an ice patch (Kuhn, Reference Kuhn1995). Similarly, stagnant ice (also known as dead ice) is that ice without movement. Generally, stagnant ice is formed due to detachment of/from the active glacier and topographic conditions do not allow for its movement, and it is covered with thick piles of moraine/debris, which act as insulator and protect it from quick melting (Singh and others, Reference Singh, Singh and Haritashya2011). In this work, we consider ice patch, dead ice and stagnant ice as synonyms. Permafrost is defined as ground that remains below 0°C for at least two consecutive years (Singh and others, Reference Singh, Singh and Haritashya2011).
To support or refute our hypothesis (the Infiernos Glacier can no longer be considered a very small glacier, while the accumulation of debris could give rise to a new paraglacial stage), the following objectives are established: (1) determine the existing ice thickness and quantify the ice losses observed in the last years; (2) understand the effect of the increasing debris cover on ice melt; (3) describe the regime of rockwall and ground temperatures around the current ice body. To achieve these objectives, we applied a wide variety of techniques to monitor the evolution of the Infiernos Glacier, including UAVs, airborne LiDAR, ground-penetrating radar (GPR) and ground temperature sensors. Despite the hypothesis suggesting that Infiernos is no longer a glacier but an ice patch, the term ‘Infiernos Glacier’ is maintained throughout the text.
2. Study area
The Infiernos Glacier is located in the Spanish Pyrenees, on the north face of the Infiernos peak (Fig. 1). The glacier drains to the Caldarés River, a tributary of the Gállego River (left margin of Ebro Basin). It has a mean elevation of 2834 m a.s.l (ranging from 2743 to 2970 m a.s.l) and an extent of 0.045 km2 (4.5 ha) in 2023. The Infiernos massif is composed by metamorphic rocks (quartzite/shales and marbles), on the western side (Fig. 1), with various glacial and paraglacial landforms as moraines, scree slopes, protalus lobes and lakes in over-deepened areas (Serrano-Cañadas, Reference Serrano-Cañadas1991; Serrano and others, Reference Serrano, de Sanjosé-blasco, Gómez-Lende, López-Moreno, Pisabarro and Martínez-Fernández2019).
Location of Infiernos Glacier within the Pyrenees (panels f, g). The upper left image (a) shows the glacier in August 2022 from a UAV oriented to the south, capturing the north face of infiernos peak where the glacier is located. The upper right image (b) shows the glacier in late summer 2023 with the boundaries manually identified from the orthophoto. The middle panel map (c) shows the topography around the glacier and the location of the six temperature sensors installed in august 2021. The two lower images show the glacier in late summer 2011 (d) and 2020 (e) with the boundaries manually identified from the orthophotos.
The climate of the Pyrenees is characterized by a progressive transition from Oceanic to Mediterranean conditions from East to West (Vicente-Serrano, Reference Vicente-Serrano2006; López-Moreno and Vicente-Serrano, Reference López-Moreno and Vicente-Serrano2007), and the Infiernos massif receives mixed influences. This results in wet winters and springs, and dry summers with occasional convective storms. At high elevation, annual precipitation exceeds 2000 mm, reaching 2500 mm at the highest summits (García-Ruiz and others, Reference García-Ruiz, Beguería, López-Moreno, Lorente and Seeger2001). The study area, like the rest of the Pyrenees, is characterized by high interannual variability in terms of snow accumulation (Revuelto and others, Reference Revuelto2017), which, together with summer/autumn temperatures, drives the annual glacier mass balance (López-Moreno and others, Reference López-Moreno2019). In August 2015, several hikers and also the guards of the nearest refuge (the Bachimaña hut, about 3 km away) reported the occurrence of a major rockfall over the glacier, creating a large debris-covered area (Fig. 1). Nonetheless, the presence of snow cover in the lower part of the glacier, even during the melting period, hindered a proper delimitation of this debris-covered area until recent years (2022 and 2023).
3. Data and methods
3.1. UAVs’ acquisitions
Recent advances have made UAVs reliable enough to work in remote mountain areas and observe distinct elements of the cryosphere as glaciers or the snowpack (Piermattei and others, Reference Piermattei, Carturan and Guarnieri2015; Harder and others, Reference Harder, Schirmer, Pomeroy and Helgason2016; Gaffey and Bhardwaj, Reference Gaffey and Bhardwaj2020; Revuelto and others, Reference Revuelto, López-Moreno and Alonso-González2021a). Through the application of Structure from Motion (SfM) algorithms (Snavely and others, Reference Snavely, Seitz and Szeliski2006) to nadir images acquired with the UAVs, three-dimensional (3D) point clouds representing the study area were generated at different dates. The overlap of these images must be high enough to guarantee a precise reconstruction of the surface (lateral and frontal image overlap > 70%; Goetz and Brenning, Reference Goetz and Brenning2017; Goetz and others, Reference Goetz, Brenning, Marcer and Bodin2018). Following the procedure described in Revuelto and others (Reference Revuelto, López-Moreno and Alonso-González2021a) and applied for glacier surface monitoring in the Pyrenees (Vidaller and others, Reference Vidaller2021), the surface of the Infiernos Glacier surface was generated from the annual UAV flights.
These flights were conducted at the minimum snow accumulation period (late August–September), with four acquisitions in total from 2020 to 2023. Additionally, one observation around peak snow accumulation time (early May 2023) was obtained to determine the snow depth distribution pattern over the glacier. Table 1 shows dates of acquisition, UAV platforms used and georeferencing steps. The distinct payload cameras of the UAVs had 20 Mpx. All UAV acquisitions followed the same acquisition protocol (UAV flight configuration with equivalent height flight, images overlap; see Supplementary Materials for further details, including Table S1), resulting in equivalent spatial resolutions and uncertainties (Revuelto and others, Reference Revuelto, López-Moreno and Alonso-González2021a), allowing the later comparison (in this case subtraction) of the 3D point clouds.
Acquisition dates, device flown, geolocation method and sensor characteristics of the different UAV flights exploited in this work
Note: Global positioning system post processing (GPS-PPK) is a positioning method that accurately geolocates UAV images based on the UAVs’ GPS and GPS observations from the local geodetic network. GPS + ICP method relies in standard GPS geolocation accuracies which after generating the 3D point cloud, exploit ICP algorithms to accurately geolocate these point clouds to the refence one.
With the images retrieved, the 3D point clouds were generated with the Pix4D SfM software (check Supplementary Materials for further details), all with volumetric densities higher than 27 pts/m3. The Sense Fly eBee Plus UAV platform was the only one capable of accurate images geolocation, employing GPS Post Processed Kinematic. After processing with Rinex data from the nearest GPS base station of the local geodetic network ARAGEA, the mean geolocation accuracy of the 3D point cloud was 0.052 m. Point clouds derived from DJI Mavic 3 Enterprise and DJI Mavic-2 Pro had geolocation accuracies around 5 m, lacking the PPK processing option, potentially limiting their suitability for measuring glacier surface differences. To accurately geolocate the 2021, 2022 and 2023 point clouds, we followed the procedure described in Vidaller and others (Reference Vidaller2021). These point clouds were aligned with the reference point cloud, the 2020 one, as this later point cloud had significantly lower geolocation uncertainty. After selecting common stable terrain areas around the glaciers (such as soil rocks and stable cirque walls), an iterative closest point (ICP) routine from CloudCompare software was applied to align the point clouds in these areas. The same transformation (ICP computed translation and rotation matrix) was then applied to the entire point cloud. This two steps geolocation method (first step: GPS-based geolocation from UAV images acquisition and second step: ICP) is referred hereafter as GPS + ICP. This later procedure resulted in a root mean squared error (RMSE) between the reference Sense Fly eBeePlus point cloud (from 2020) and the other point clouds being lower than 0.13 m in all cases (see Supplementary Material Table S2), which is sufficiently accurate for measuring glacier surface differences in the Pyrenean mountains (Vidaller and others, Reference Vidaller2021).
3.2. Airborne LiDAR observation
The 3D point cloud of 2011 was generated from an airborne LiDAR acquisition conducted by IGN (Spanish Geographic Institute) during a flight on 9 November 2011. The device utilized was a Leica ALS60, featuring a diode-pumped transmitter and a high-speed scanning mirror with a large aperture, operating at a wavelength of 1064 nm. The resulting geolocated point cloud had an average density of 0.35 pts/m3. Despite the late date LiDAR acquisition, which might not guarantee a snow free acquisition, no relevant snow falls occurred in previous weeks and the surface of the glacier was snow free (Vidaller and others, Reference Vidaller2021), Similar to the 2021, 2022 and 2023 UAV point clouds, the airborne LiDAR point cloud was geolocated with the 2020 UAV 3D surface, following same procedure described in previous section. In this case, the ICP alignment had an RMSE of 0.15 m. Based on this consistent ICP geolocation, we are confident that these surfaces can be compared with one another, regardless the acquisition technique.
3.3. Comparisons of the glacier surface elevation
The temporal evolution of glacier surface elevation (i.e. glacier thickness change) was computed with the Multiscale Model to Model Cloud Comparison (M3C2) tool (Lague and others, Reference Lague, Brodu and Leroux2013; James and others, Reference James, Robson and Smith2017) on CloudCompare software. The M3C2 plugin computes distances between the reference point cloud and the contrasted point cloud perpendicularly to the surface (normal to the reference surface) within a virtual cylinder of radius and height depending on the point cloud characteristics (James and others, Reference James, Robson and Smith2017). Following the same procedure, we also estimated the snow depth distribution on the glacier in May 2023.
3.4. GPR thicknesses
In July 2021, a GPR survey was conducted over Infiernos Glacier surface to measure glacier thickness in different transects (Navarro and Eisen, Reference Navarro and Eisen2009; Del Río and others, Reference Del Río, Rico, Serrano and Tejado2014). At this date, the glacier was continuously covered with snow, and there was no liquid water circulating within the ice (e.g. supraglacial channels), which could absorb electromagnetic pulses and hinder the identification of ice layers deeper than the liquid water layer. The device Mala Geoscience radar system (ProEx) with a 200 MHz unshielded antenna was used for this survey, yielding 32 radargrams from seven transects (three longitudinal and four transversal; see Fig. 2). The radargrams were processed and filtered using Reflexw version 9.1.3 (Sandmeier Scientific Software), following a systematic workflow detailed in Vidaller and others (Reference Vidaller2023). An initial estimation of snow and ice thickness was done with a propagation velocity of 0.17 m ns−1. Subsequently, the glacier thickness over the GPR transects was derived by applying propagation velocities of 0.200 ± 0.005 m ns−1 for snow and 0.163 ± 0.007 m ns−1 for ice as residual liquid water might be present during the GPR acquisition (López-Moreno and others, Reference López-Moreno2019). An ice thickness map over the glacier was made using radial basis function interpolation (Otero García, Reference Otero García2008) similar to Vidaller and others (Reference Vidaller2023). The interpolation assumed 0 m thicknesses at the glacier’s outline as glacier derived from the 2021 UAV flight.
Longitudinal (a, b and g) in N-S direction and transversal GPR transects (c, d, e, f) in E-W direction, showing signal rebound depth and time. Distribution of transects over the glacier surface is illustrated in the bottom right panel (h). The rock-ice interface and the snow-ice interface manually delignated are included in the GPR transects.
3.5. Ground and rockwall temperature
On 6 September 2021, a total of six thermometers (Fig. 1) were installed around the glacier to monitor temperatures. The sensors installed were TinyTag TGP-4017 thermistors (Navarro-Serrano and others, Reference Navarro-Serrano2019) with an acquisition frequency of 3 hours, providing records of daily and seasonal temperatures in the area. However, due to a sealing failure identified during the 2022 field campaign of sensors T3 and T6, these sensors were all replaced by Hobo waterproof sensors (UA-001-64). Sensors 1 and 2 were drilled 3-5 cm into the rock, while the others were buried approximately 5–7 cm deep among surrounding rocks from previous debris falls. Of the six sensors installed in 2021, data were recorded from only four, due to acquisition failures in the others. The Hobo sensors were replaced in October 2022, ensuring data consistency for all sensors.
4. Results
4.1. Glacier thickness and surface evolution from 2011 to 2023
The glacier area decreased by 21% from 2011 to 2023 (see Table 2). Notably, the area decreased by 5% from 2020 to 2021 and by 17% from 2020 to 2023. These numbers are subject to uncertainty because Infiernos Glacier typically retains snow for extended periods and sometimes does not melt entirely as observed in 2011, 2020 and 2021 in the lower area, hampering a precise identification of glacier boundaries for these later years.
Glacier surface extent along the study period
The interannual evolution of glacier thickness exhibited a high spatial and temporal variability in glacier shrinking (Fig. 3). From 2011 to 2020, the glacier experienced significant losses in the eastern sectors (left side of the image in Fig. 3a), with surface lowering exceeding 6 m over approximately 25% of the glacier area. In the lower and western parts of the glacier, thinning did not exceed 2 m. However, the mean thinning for this period was −3.9 m (Table 3). In the period 2020–21, there was a mean decrease of 0.6 m (Table 3), in stark contrast with 2021–22 and 2022–23, when the mean thinning was 3.2 m and 1.9 m, respectively. Overall, the glacier thinned 9.0 m from 2011 to 2023, with 5.4 m observed in the last two years. It is noteworthy that ice thinning in some parts reached 16 m during the study period.
Glacier thinning observed in different time intervals for the annual glacier minimum accumulation period. The lower (d, e, f) and upper panels (a, b, c) share, respectively, the same legend to facilitate easier inter-comparison. Panel f includes the boundaries of the rocky outcrop that emerged in 2023 and the boundaries of the debris covered area.
Mean glacier thickness changes computed in the entire glacier (boundaries of the later observation in 2023) and differences observed in the debris-covered area, computed using 2023 debris boundaries (the entire extent of the debris-covered area) and the lower debris lobe (see Figure 1)
The spatial pattern of glacier thinning exhibited notable similarities in the last three years. After 2020, the debris-covered area (Fig. 1) becomes easily identifiable, showing lower thinning compared to the rest of the glacier (left of the glacier in Fig. 3d–f). The debris-covered area covered 0.0143 km2 (32% of the glacier extent) in 2023. This area includes a lower debris lobe (see Fig. 1), with a gentler slope than the upper parties, where glacier losses were even smaller (white areas in Fig. 3d–f, from 2020 to 2023). The extent of the lower debris lobe is 0.058 km2 (11% of glacier extent) and has shown a reduction of thickness lowering about a 50% compared to the entire glacier. Finally, during the UAV acquisition of 2023, a new rock outcrop was identified in the middle of the glacier (Fig. 1). This contributed to lower ice loses in this zone (Fig. 3f, 2022–23 comparison) as very shallow ice was present in this area of the glacier.
The UAV flight in May 2023 enabled the determination of snow depth distribution over Infiernos Glacier at the end of the snow accumulation period (Fig. 4). Higher snow accumulation was observed in the lower and flatter areas of the glacier. However, snow accumulation over the debris-covered area was lower compared to the surrounding areas. Snow profiles obtained from GPR in July 2021 also indicated higher snow accumulation in the lower part of the glacier (Fig. 2). Mean and maximum snow thickness in transects across the glacier surface in May 2023 when compared to August 2020 and August 2022 ranged between 1.7–2.5 m and 2.1–4.6 m, respectively (Table 4). The snow accumulation observed in May 2023 was not enough to equate glacier surface elevation observed in late melting period 2020 (light grey in Fig. 4b), showing that the snow accumulated over a complete winter cannot compensate the dramatic ice thickness loss from 2020 to 2023.
Comparison of the snow surface in May 2023 using the glacier surface on August 2022 (a) (i.e. the snow depth map in 2023) and August 2020 as reference (b) (which is not the snow depth but the difference of the snow surface in May 2023 and ice surface in Aug 2022).
The mean and maximum thicknesses (and the transects uncertainties ‘σ’ associated to the mean values) of snow and ice observed across seven GPR transects on 21 July 2021
4.2. Glacier thickness in July 2021
The interpolated glacier thickness map (Fig. 5) indicates that shallow ice (< 7 m) covers more than 30% of the glacier extent. Conversely, there are areas where ice thickness still exceeds 20 m. The thickest part of the glacier is located in its lower section. A zone of low ice thickness is observed in the middle of the glacier (light blue surrounded by purple colors in Fig. 5). Transects A and B (Fig. 2) clearly illustrate that over distances ranging from 170 to 190 m (transect A) and from 190 to 210 m (transect B), the bedrock lies close to the glacier surface (both cases below 7 m). The map also shows a sharp a distinct transition from deep ice thicknesses in the central parts of the glacier to shallower ice near the glacier edges. The maximum ice thicknesses are observed in transects A and C (24.3 and 25.9 m, respectively, Table 4), followed by transect B (22.9 m). We estimate that the uncertainty associated with these maximum values is 0.4 m (λ/2 = 0.4 m). The GPR profiles indicate the presence of two over-deepened basins with an intermediate rock bar. The deepest ice thicknesses are recorded in both over-deepened basins; however, the surface does not reflect the substrate’s topography, displaying a beveled and rectilinear profile. One of these over-deepened areas is a bedrock gully or depression that runs along the glacier’s main vertical axis.
This map illustrates the ice thickness derived from radial basis function (RBF) interpolation of data obtained from seven ground-penetrating radar (GPR) transects conducted in July 2021. The continuous black line over the glacier shows 2023 boundaries.
4.3. Soil temperatures around the glacier
Ground and rockwall temperatures below − 3°C for relatively long time periods, located around the glacier (Fig. 1), indicate the potential presence of permafrost in certain areas. This is evidenced by temperatures below zero once the ground is insulated from air temperatures by a thick snowpack (Fig. 6). Especially, sensors T4 and T5 recorded winter temperatures reaching values close to − 5°C. These sensors are situated in a protalus lobe observed in the eastern side of the glacier (Fig. 7). Sensors T1 and T2, located higher up and directly drilled into the rockwall along with T3, show a prolonged period near 0°C, suggesting the absence of permafrost in those locations. However, the sensors are situated in shallow holes, which are susceptible to thermal conductivity due to the direct or diffuse solar radiation, among other factors. Sensor T6 exhibits a brief period near 0°C (purple dashed line in Fig. 6b, from January to March) associated with a short duration of snow cover, followed by very low ground temperatures after the snow melted, indicating the presence of seasonal frozen ground. (see more content in the Supplementary Material section).
Sensor temperature evolution around the glacier for the 2021–22 (a) and 2022–23 (b) snow seasons. For an easier interpretation, panel a shows zero curtain periods for some sensors and the potential permafrost signal with temperatures below −3°C.
The upper image (a) shows a supraglacial stream carved into the ice due to liquid water flow over the glacier surface observed in 2023 due to snow absence during the filed campaign. The lower image (b) depicts a protalus lobe located on the eastern side of the glacier obtained in the 2022 field campaign. The extent of panel a image is marked in panel b image for an easier interpretation.
The ground thermal regime typically undergoes freezing/thawing cycles during short periods, primarily in fall and spring. The activity of rockfalls, which has contributed to the debris-covered area, may be linked to freeze-up events in fall and the melting of seasonal ice in spring, probably destabilizing the walls and triggering rockfalls and landslides.
4.4. Features of recent glacier evolution
In addition to the generation of detailed 3D point clouds, UAV images have facilitated the identification of distinct features that characterize the recent evolution of the glacier. This includes a supraglacial stream carved more than 8 m deep into the ice in the lower part of the glacier (Fig. 7). UAV images have been instrumental in characterizing a protalus lobe (Fig. 7), located in the eastern side of the glacier, where temperature sensors T4 and T5 where installed (Fig. 1). Furthermore, UAV images enabled the identification of a bedrock knoll in the middle of the glacier in 2023 (Fig. 1).
5. Discussion
5.1. Ice loss from 2011 to 2023
The Infiernos Glacier has undergone rapid decline in both ice extent and thickness since 2011, with a 21% reduction in areas and an average thickness loss exceeding 9 m. Particularly high losses occurred from 2021 to 2023 with an average ice thickness decrease of over 5 m during this period. Ice loss at Infiernos Glacier from 2011 to 2020 fell within the 25th and 50th percentiles of all Pyrenean glacier losses (Vidaller and others, Reference Vidaller2021). Despite experiencing the same climatic conditions as other Pyrenean glaciers, its north-facing aspect and high snow accumulation have mitigated ice loss compared to others, highlighting topoclimatic influence (Florentine and others, Reference Florentine, Harper, Fagre, Moore and Peitzsch2018, Reference Florentine, Harper and Fagre2020; Huss and Fischer, Reference Huss and Fischer2016) in Infiernos Glacier. However, similar losses to those observed in other Pyrenean glaciers were recorded for 2021–22 and 2022–23 (Izagirre and others, 2024). These extreme years, which also severely affected Alps glaciers (Voordendag and others, Reference Voordendag, Prinz, Schuster and Kaser2023), were characterized by remarkable positive temperature anomalies in spring and summer (Serrano-Notivoli and others, Reference Serrano-Notivoli2023), as well as low and early snow accumulation in both winters compared to recent records in the Pyrenees (Gascoin and others, Reference Gascoin2015; Revuelto and others, Reference Revuelto2017).
The rise in spring temperatures and its impact on snow melting dynamics significantly affected glacial activity in Mediterranean mountains (Palacios and others, Reference Palacios, de Andrés and Luengo2003), while very hot summers can lead to substantial changes in glacier areas (Pederson and others, Reference Pederson, Fagre, Gray and Graumlich2004, Reference Pederson, Graumlich, Fagre, Kipfer and Muhlfeld2010). Under severe temperatures (hot springs and summers) and low snow accumulation during cold periods, the topographic control of very small glaciers may diminish, underscoring their sensitivity to short-term climate change (Hughes, Reference Hughes2008).
The emergence of a rocky outcrop in 2023 is expected to further accelerate glacier melting through long-wave emissivity (López-Moreno and others, Reference López-Moreno2016). This change in surface mass balance, combined with ice thickness data from a 2021 GPR survey, suggests a potential division of the glacier into two distinct bodies in the near future. Moreover, the bedrock bump observed through the GPR survey might have impacted ice flow dynamics in the last decades potentially reducing its movement capacity.
5.2. Uncertainties of changes detected
This work has applied novel close-range remote sensing techniques to characterize recent changes of Infiernos Glacier. These techniques have had a recurrent use to study the cryosphere in recent years (Deems and others, Reference Deems, Painter and Finnegan2013; Gaffey and Bhardwaj, Reference Gaffey and Bhardwaj2020) but still have shortcomings on their applicability that must be taken into account (Gindraux and others, Reference Gindraux, Boesch and Farinotti2017; Revuelto and others, Reference Revuelto, López-Moreno and Alonso-González2021a). The use of SfM algorithms to generate 3D point clouds from UAV imagery also has uncertainties, but these are usually lower than the changes observed in glaciated areas, making them suitable to study glacial and periglacial processes with error comparable to other techniques (Piermattei and others, Reference Piermattei2016). To avoid potential errors of the surface comparison, same UAV flight configuration must be used to guarantee equivalent accuracies of the final point clouds (Stark and others, Reference Stark2021). Then, after co-registering point clouds in stable terrain areas, surface differences can be computed with high certainty (Revuelto and others, Reference Revuelto2021b). These two last steps were systematically applied in this work, what guarantees reliable change detection with the UAVs. When the magnitudes of change has low magnitude (<0.1 m), M3C2 change analysis has associated uncertainties (Zahs and others, Reference Zahs, Winiwarter, Anders, Williams, Rutzinger and Höfle2022). Similarly, when M3C2 measures changes that deviates from surface normal, what usually happens in rough surfaces, change detection is less reliable (Williams and others, Reference Williams, Anders, Winiwarter, Zahs and Höfle2021). Conversely, none of these cases occur in our study area as ice loss is more than one order of magnitude higher (than 0.1 m), and glacier surface has lower roughness than the surrounding areas, doing reliable the methodology applied to characterize Infiernos Glacier changes.
5.3. Insufficient snow accumulation to generate ice
The snow depth distribution observed in late May 2023 shows a thick snowpack in the lower, flatter areas of the glacier, whereas higher areas exhibit shallower coverage. This is probably related with the slope increase observed in the upper areas of the glacier, which avoid the accumulation of a thick snowpack and trigger avalanches that accumulates in the lower and flatter areas of this ice body. Similar snow depth distribution patterns were observed in GPR transects in July 2021 (higher snow depth in lower glacier areas), showing that snow distribution is strongly influenced by local topography (Grünewald and others, Reference Grünewald2013; Revuelto and others, Reference Revuelto, López-Moreno, Azorin-Molina and Vicente-Serrano2014), resulting in consistent snow patterns annually. This pattern contrasts with the traditional glacial conceptualization based on an upper accumulation area where snow persists throughout the ablation period, and a lower elevation melting area where snow completely melts annually, differentiated by the ELA (Braithwaite and Muller, Reference Braithwaite and Muller1980; Rabatel and others, Reference Rabatel, Letréguilly, Dedieu and Eckert2013). The comparison of glacier surface elevations between August 2020 (snow free) and May 2023 (snow covered) confirms that current snow accumulation observed in 2023 is insufficient to transform into firn and subsequently into ice (Swift and others, Reference Swift, Cook, Heckmann, Gärtner-Roer, Korup, Moore, Haeberli and Whiteman2021), leading to reduced dynamism associated with gravitational ice movement from upper to lower areas (Meier, Reference Meier1962; Kuhn, Reference Kuhn1995).
5.4. Soil temperature
Over the two years with ground temperature recordings, signals of permafrost (< − 3°C) were detected at three sensors, possibly associated with buried ice (Haeberli and others, Reference Haeberli2010). These signals were observed in the protalus lobe on the eastern side of the glacier and at the lowest elevation temperature sensor. Nonetheless, the temperature records within our study area does not satisfy the permafrost definition as soil temperatures did not remain below 0°C for at least two consecutive years. In such a way, it is not possible to undoubtedly state that there is permafrost in this study area. More likely, when the ground is not isolated due to the snowpack absence (Peng and others, Reference Peng2024), air temperatures probably are heating the upper most ground layers, hampering the observation of permafrost signal. However temperatures observed in this study area adds further evidence to the occurrence of permafrost in the Pyrenees (Serrano and others, Reference Serrano, Agudo, Delaloyé and González-Trueba2001, Reference Serrano, de Sanjosé-blasco, Gómez-Lende, López-Moreno, Pisabarro and Martínez-Fernández2019; Rico and others, Reference Rico2021), encouraging future studies to determine its presence, distribution and characteristics, as well as to understand the potential implications of its evolution. Despite permafrost occurrence in the Pyrenees in global permafrost distribution models (Gruber, Reference Gruber2012; Obu and others, Reference Obu2019) is nearly negligible when compared to the extent of permafrost areas in the northern hemisphere, a deeper understanding of its distribution is needed to validate these models, what also highlights the need of further research.
5.5. Debris cover impact on glacier dynamics
The UAV surface comparisons presented here demonstrate that the debris-covered area significantly impacts glacier melt dynamics, resulting in notable reductions in ice thickness loss beneath this area. Radargrams indicate that debris cover ranges from 5 cm to 50 cm in thickness, which has been shown sufficient to insulate the ice from air temperatures and reduce ice thinning in some areas (mainly in the debris frontal lobe where thicker debris cover is present), owing to the glacier’s north-facing aspect (Owen and others, Reference Owen, Derbyshire and Scott2003; Nicholson and Benn, Reference Nicholson and Benn2006; Anderson and Anderson, Reference Anderson and Anderson2016). The high variability in the lithology of the study area (quartzites/marble contact; see Fig. 1) and future climate trajectories impacting rockwall erosion (Draebing and others, Reference Draebing, Mayer, Jacobs and McColl2022) will influence future rockfalls not only in this study area but also in the Pyrenees (Wegmann and others, Reference Wegmann, Gudmundsson and Haeberli1998; Hartmeyer and others, Reference Hartmeyer2020). If the observed rate of glacier thinning continues in the coming years, the debris-covered ice may exhibit significant relief compared to the surrounding glacier and potentially display movement characteristics typical of very small debris-covered glaciers (Securo and others, Reference Securo2024). Given the evolution of debris-free and debris-covered areas, we anticipate that the Infiernos Glacier will likely be completely covered by debris in the near future.
5.6. New paraglacial stage
Several studies have suggested that debris-covered glaciers have the potential to transform into rock glaciers (see Anderson and others, Reference Anderson, Anderson, Armstrong, Rossi and Crump2018; Securo and others, Reference Securo2024). However, these transformations are typically observed in environments with active and dynamic glaciers that accumulate large amounts of debris at their fronts, often with internal ice bodies, or in permafrost environments where debris undergoes re-icing and ice segregation. The genesis of glaciogenic rock glaciers in the Pyrenees, like the Posets rock glacier, involves cold periods of pre-Little Ice Age glacial advances and permafrost conditions (Serrano and others, Reference Serrano, Agudo, Delaloyé and González-Trueba2001). In marginal environments of temperate mountains, such as the Pyrenees, where small glaciers or stagnant ice do not generate material or transport material to their fronts, but rather accumulate rockfall from surrounding walls, debris cover is common, as has been observed in Infiernos Glacier. In areas where permafrost is present, subsurface ice and rock glaciers can be generated having their characteristic flow dynamics (Haeberli and others, Reference Haeberli, Arenson, Wee, Hauck and Mölg2024), but this is not likely the case of our study area. We consider that Infiernos Glacier cannot longer be considered a glacier but an ice patch partially covered by debris from rockfalls of the surrounding cirque walls. Thermometers show that the areas of free ice previously occupied by the Infiernos Glacier during the Little Ice Age maximum only exhibit seasonal frozen ground, maybe because the glacier was temperate or polythermal based, indicating a very low potential for rock glacier formation. This complex scenario warrants further research to understand how rock weathering and erosion will contribute to landscape changes in the study area (Sanders and others, Reference Sanders, Cuffey, Moore, MacGregor and Kavanaugh2012) utilizing the database generated in this work to track the future evolution of rockfalls (Thiele and others, Reference Thiele, Grose, Samsu, Micklethwaite, Vollgger and Cruden2017) and their impact on paraglacial processes.
UAV observations of the glacier have revealed significant degradation of the Infiernos Glacier, transitioning to a new paraglacial stage that has nearly stopped its characteristic flow (Benn and Evans, Reference Benn and Evans2010). Based on the criteria described by Izagirre and others (2024) from previous works (Leigh and others, Reference Leigh, Stokes, Carr, Evans, Andreassen and Evans2019), four of these have been clearly identified in the study area: absence of crevasses, presence of deep-water incision, lack of a snow accumulation area in the upper glacier and significant debris cover. These criteria indicate that the ice does not flow downhill. Moreover, substantial losses in ice thickness and area have been observed over the study period, and bedrock knoll has emerged in the middle of the glacier. These findings collectively confirm that the Infiernos Glacier can no longer be considered a glacier but rather an ice patch of stagnant ice which is transitioning to a new paraglacial stage. Future research in this area will be crucial to understanding how this transition is altering the landscape around Infiernos peak under warmer climatic conditions.
6. Conclusions
This study aimed to test the hypothesis that the Infiernos Glacier can no longer be classified as a glacier but rather as an ice patch transitioning to a new paraglacial stage. Between 2011 and 2023, the glacier experienced an average thinning of 9.3 m, with over 5 m lost during 2020–23 alone. In certain locations, ice thickness losses reached up to 25 m, and GPR measurements indicate multiple regions where the ice column less than 5 m. Approximately 32% of the glacier’s area is covered by debris, which, despite its relatively thin layer, effectively insulates the ice. In some parts of the debris-covered area, ice loss is reduced by a factor of two compared to the rest of the glacier.
These processes, coupled with clear indications of ice stagnation (crevasses absence, deep-water incisions, lack of snow accumulation area and significant debris cover) and the absence of new ice formation, lead to our inference that the Infiernos Glacier has already transitioned into an ice patch that could significantly diminish in size within a few years and become completely covered by debris. Ground and rockwall temperature monitoring have shown the potential presence of permafrost and seasonal frozen ground in the vicinity of the ice body, along with noticeable freezing–thawing cycles that may intensify under warmer climate conditions. These factors will be crucial in understanding the future landscape evolution of this site and other mountains where glaciers are in their final stage.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/jog.2025.22.
Acknowledgements
This work was supported by the MeltingIce project from the Leonardo program of the BBVA research foundation (‘Proyecto realizado con la Beca Leonardo a Investigadores y Creadores Culturales 2023 de la Fundación BBVA’), the EU funded LIFE project PYRENEES4CLIMATE (101104957 - LIFE22-IPC-ES-LIFE PYRENEES4CLIMA) and the Gobierno de Aragón research group ‘Procesos Geoambientales y Cambio Global’ (E02_23R). Ixeia Vidaller is enrolled in the PhD program at the University of Zaragoza (grant no. FPU18/04978). We acknowledge the help provided by Lauren Vargo (editor) and the two reviewers (Benjamin Hills and an anonymous reviewer) for the efforts on assessing and providing suggestions for improving the work.
Author contributions
JR did the analysis and wrote the first manuscript draft. JR, EI, IR, LR, ES, IV, FR-V and JIL-M. contributed on the different field campaigns with UAV, GPR and thermometers installation, all co-authors have also discussed the results and provided feedback for writing the paper.
Competing interests
Authors declare no conflict of interest. The BBVA foundation requires the inclussion of this text “La ‘Fundación BBVA no se responsabiliza de las opiniones, comentarios y contenidos incluidos en el proyecto y/o los resultados derivados del mismo, los cuales son total y absoluta responsabilidad de sus autores.’
Open access
