1. Introduction
Small glaciers and ice caps (hereafter referred to as glaciers) cover ∼133 400 km2 around the periphery of Antarctica and represent 18% of global glacier area, excluding the major ice sheets, of which ∼77 600 km2 are found around the periphery of the Antarctic Peninsula (Pfeffer and others, Reference Pfeffer2014; RGI 7.0 Consortium, 2023). However, despite such a large glacierised area, few glaciological data are available for land-terminating glaciers in Antarctica compared to other world regions. Antarctic Peninsula glaciers receded at accelerating rates until the end of the 20th century (Davies and others, Reference Davies, Carrivick, Glasser, Hambrey and Smellie2012), and several ice shelves collapsed (Cook and Vaughan, Reference Cook and Vaughan2010), coincident with air temperatures warming among the most rapidly of any place on Earth (Vaughan and others, Reference Vaughan2003). Glacier recession rates then slowed down during the early 21st century, coincident with decreased air temperatures (Oliva and others, Reference Oliva2017). Nonetheless, following an increase in air temperatures since the mid-2010s (Carrasco and others, Reference Carrasco, Bozkurt and Cordero2021), Antarctic Peninsula glaciers are once again receding at an enhanced rate (Engel and others, Reference Engel, Láska, Kavan and Smolíková2023). Furthermore, recent extreme warming across the Antarctic Peninsula has led to exceptional melt rates in some places (Siegert and others, Reference Siegert2023), most notably on the George VI and Larsen C ice shelves (Bevan and others, Reference Bevan, Luckman, Hendon and Wang2020; Banwell and others, Reference Banwell2021; Xu and others, Reference Xu2021).
Exceptional melt events release large volumes of freshwater, as well as sediments and solutes, into proglacial streams, lakes and the Southern Ocean (Carrivick and Tweed, Reference Carrivick and Tweed2021; Kavan and others, Reference Kavan, Hrbáček and Stringer2023; Stringer and others, Reference Stringer2024). More generally, the loss of glacier mass can increase meltwater flux and the availability of sediment (Carrivick and Rushmer, Reference Carrivick and Rushmer2009). This affects both water temperature (Carrivick and others, Reference Carrivick, Brown, Hannah and Turner2012a) and water quality, and thus the fragile marine and lacustrine ecosystems of Antarctica (Nedbalová and others, Reference Nedbalová, Nývlt, Kopáček, Šobr and Elster2013; Gonçalves and others, Reference Gonçalves2022), as well as contributing to glacier flow accelerations (Bell and others, Reference Bell, Banwell, Trusel and Kingslake2018; Tuckett and others, Reference Tuckett2019).
Recently recorded mass losses from the peripheral glaciers of Antarctica (2000–19: −20 Gt yr−1) are the fifth greatest of any world region, but small in comparison with other glacierised world regions when considered per unit glacierised area (−1.5 × 10−4 Gt yr−1 km−2 compared to a global average of −3.7 × 10−4 Gt yr−1 km−2; Hugonnet and others, Reference Hrbáček, Engel, Kňažková and Smolíková2021; RGI 7.0 Consortium, 2023). Projected increased melt rates across the Antarctic Peninsula are uncertain. Some models indicate mean annual surface melt will increase in the northern Antarctic Peninsula by >2 m water equivalent (w.e.) yr−1 (Garbe and others, Reference Garbe, Zeitz, Krebs-Kanzow and Winkelmann2023), though other projections show this may be offset by increased snowfall accumulation under high-emission scenarios (Edwards and others, Reference Edwards2021). Therefore, there is a pressing need to understand how recent air temperature increases, changes in precipitation (Carrasco and Cordero, Reference Carrasco and Cordero2020) and an increase in extreme warm events affect Antarctic Peninsula glaciers.
This study presents and evaluates glacier changes in the James Ross Archipelago in response to recent and ongoing extreme air temperature warming recorded at the Johann Gregor Mendel (JGM) Czech Antarctic Station (57.9°W, 63.8°S). Specifically, we provide satellite-derived data on changes in glacier area (Landsat-7, Landsat-8 and Landsat-9) and albedo (MODIS) for the James Ross Archipelago since 2010. Additionally, we present coincident in situ measurements of ablation and accumulation for Lookalike Glacier and Davies Dome on the Ulu Peninsula, James Ross Island, collected as part of a long-term study in the region (Engel and others, Reference Engel, Láska, Smolíková and Kavan2024).
1.1. Study site
The James Ross Archipelago is situated off the northeast coast of the Antarctic Peninsula, across the Prince Gustav Channel and 10 km eastwards from Trinity Peninsula (Fig. 1). The largest islands of the archipelago are James Ross Island, Vega Island, Snow Hill Island and Seymour Island, and they are home to 156 glaciers (Fig. 1, Table 1). The bedrock across the archipelago is primarily composed of Cretaceous to Paleogene mudstones and sandstones, with upland regions of Neogene basalts, hyaloclastite breccias and tuffs (Smellie, Reference Smellie2013; Mlčoch and others, Reference Mlčoch, Nývlt and Mixa2019). The archipelago has a semi-arid polar continental climate, with a mean annual air temperature of −7°C at JGM Station (10 m a.s.l.) (Kaplan Pastíriková and others, Reference Kaplan Pastíriková, Hrbáček, Uxa and Láska2023) and a mean lapse rate of 0.43°C 100 m−1 measured over glacierised sites (270–540 m a.s.l. between 2013 and 2016) (Ambrozova and others, Reference Ambrozova, Laska, Hrbacek, Kavan and Ondruch2019). Mean annual precipitation is estimated at between 400 mm and 700 mm w.e. (Palerme and others, Reference Palerme, Genthon, Claud, Kay, Wood and L’Ecuyer2017), although high wind speeds mean the effective precipitation is lower than this (Nývlt and others, Reference Nývlt2016; Hrbáček and others, Reference Hrbáček, Engel, Kňažková and Smolíková2021). To consider the regional variability in climatic conditions (Morris and Vaughan, Reference Morris and Vaughan2003), we split the glaciers into two groups: those closer to the considerably warmer Ulu Peninsula (north western sector) and those closer to the ice-bound Weddell Sea (south east).
Figure 1. Map of glacier outlines derived from the GLIMS dataset. Glaciers in the north-western sector of the island are in pale blue, with those in the south-eastern sector in white. Those glaciers with in situ measurements (Lookalike Glacier and Davies Dome) are coloured in red; these have GLIMS IDs of G301945E63889S and G302049E63932S, respectively. We have also labelled glaciers that have experienced remarkable changes: Whisky Glacier (G301946E63935S), Kotick Glacier (G301659E64016S) and Swift Glacier (G302228E64270S). The inset shows the location of James Ross Archipelago with respect to the AP (highlighted red).
Table 1. Total number of glaciers analysed by island. Glacier outlines were derived from the GLIMS dataset, with some erroneous ones excluded. Areas are those presented in the results section for 2023. See the Methods for further details
Since the Last Glacial Maximum, when ice from the Antarctic Peninsula coalesced with James Ross Island (Glasser and others, Reference Glasser2014), the most prominent ice mass on James Ross Island is the Mount Haddington ice cap (Fig. 1), which is estimated to be between 200 and 300 m thick and drains over steep cliffs to outlet glaciers that are predominantly marine-terminating (Skvarca and others, Reference Skvarca, Rott and Nagler1995). Numerous small valley and cirque glaciers also exist beneath near-vertical bedrock cliffs on the Ulu Peninsula of James Ross Island and have changed in thermal regime but relatively little in geometry during the late Holocene (Carrivick and others, Reference Carrivick, Davies, Glasser, Nývlt and Hambrey2012b). Vega Island is largely covered by two plateaux ice caps that feed small tidewater glaciers (Davies and others, Reference Davies, Carrivick, Glasser, Hambrey and Smellie2012). Seymour Island is free of glacier cover. In contrast, Snow Hill Island is almost entirely covered by a marine-terminating ice cap (Davies and others, Reference Davies, Carrivick, Glasser, Hambrey and Smellie2012).
In this study, we present data for Lookalike Glacier (a land terminating glacier) and Whisky Glacier (marine terminating), among others. In previous studies, Lookalike Glacier has also been referred to as ‘Whisky Glacier’ or ‘IJR-45’, causing confusion. Therefore, we use the name Lookalike Glacier to refer to the small land-terminating glacier on the Ulu Peninsula, while Whisky Glacier refers to the larger marine-terminating glacier that terminates in Whisky Bay, as suggested by the Antarctic Place Names Committee in 2015.
2. Methods
2.1. In situ measurements
2.1.1. Air temperature
We used air temperature data collected for each glaciological year (defined as beginning of March to end of February; Engel and others, Reference Engel, Láska, Smolíková and Kavan2024; note that this is shifted by 1 month with respect to the standard glaciological year of the southern hemisphere, which runs from the beginning of April to the end of March) between 2009/10 and 2022/23 from the automatic weather station located near JGM Station (10 m a.s.l.). Air temperature was measured using a Minikin TH datalogger and EMS33H probe (EMS Brno, Czech Republic) with an accuracy of ±0.15°C, installed at 2 m above the ground in a multi-plate radiation screen. From the hourly air temperature observations, we have calculated the annual sum of positive degree days at JGM Station (Hock, Reference Hock2003). We conducted a linear regression to calculate the average annual change in temperature and report the standard error of the slope of the regression line (similar to other studies, e.g. Carrasco and others, Reference Carrasco, Bozkurt and Cordero2021).
2.1.2. Glaciological measurements
Glacier ablation and accumulation was measured using stakes placed on Davies Dome Glacier and Lookalike Glacier, which form part of a long-term measurement study (Engel and others, Reference Engel, Láska, Nývlt and Stachoň2018, Reference Engel, Láska, Smolíková and Kavan2024). These data have mostly been previously published (Engel and others, Reference Engel, Láska, Smolíková and Kavan2024), although this study examines the individual point data in greater detail. The ablation stakes are representatively distributed across the surface of the glaciers except for the crevassed sea-terminating outlet of Davies Dome. In the case of Davies Dome, we have selected to report measurements from ablation stakes along a W-E transect because they are less affected by snowdrifts and enhanced accumulation at the northeastern slope of the dome. We have split these ablation measurements into 100 m altitude bins.
The height of the stakes above the glacier surface was measured with an accuracy of ±0.01 m using a standard tape measure during the austral summer, typically in early February. Although the accuracy of our measurements was taken to be ±0.01 m, this does not account for uncertainty that may arise from melt at the base of the stake or possible compaction of snow/firn/ice at the base of the stake, or variability in snowpack density (Thibert and others, Reference Thibert, Blanc, Vincent and Eckert2008). We converted the ablation stake values in the ablation and accumulation zones to mass changes using densities of 900 and 500 ± 90 kg m–3, respectively (Engel and others, Reference Engel, Láska, Nývlt and Stachoň2018). Huss and others (Reference Huss, Bauder and Funk2009) estimate point surface mass balance data to have an uncertainty of ±0.1 m w.e. in the ablation area and ±0.3 m w.e. in the accumulation zone. For simplicity, we therefore assume a central estimate of 0.2 m w.e. as our uncertainty. For elevation bins with two measurements, we propagate the uncertainty using Equation (1):
\begin{align}\frac{{\sqrt {{{0.2}^2} + \,{{0.2}^2}\,} }}{2} = 0.12\end{align}2.2. Glacier area
Glacier outlines for years 2011, 2017 and 2023 were delimited by manually editing GLIMS glacier outlines (Raup and others, Reference Raup, Racoviteanu, Khalsa, Helm, Armstrong and Arnaud2007) to digitise glacier extent visible in Landsat images (see Table 2). We used glacier outlines produced between 2005 and 2014 available from the GLIMS dataset but found it necessary to manually remove some erroneous polygons that seemed to correspond to frozen lakes, snow-filled surface depressions, shadows and a few spurious digitising/topology errors. The years 2011, 2017 and 2023 were chosen for image clarity (lack of clouds and overall snow cover). We then quantified area change for each period (2011–17 and 2017–23) and converted this into a percentage change relative to the initial glacier area. Error in glacier outline position is assumed to be ±15 m, or half the size of one Landsat pixel (Paul and others, Reference Paul2017). Our conservative estimate of the uncertainty in glacier area, assessed by creating a ±15 m buffer on each glacier and calculating the standard deviation in those areas, produces a mean uncertainty (1σ) in a glacier area of ±0.13 km2, or 7% of the glacier area, which is comparable to previous studies (Malmros and others, Reference Malmros, Mernild, Wilson, Yde and Fensholt2016; Taylor and others, Reference Taylor, Quincey, Smith, Potter, Castro and Fyffe2022). Other sources of errors can be introduced by misidentifying debris cover or snow-covered nunataks/headwalls, which we consider to be negligible and therefore accounted for within our conservative approach.
Table 2. List of satellite (Landsat) images used to delimit glacier outlines. The specific image date is coded into the image ID
2.3. Albedo
We used the MODIS Snow Cover Daily Global product (MOD10A1 V6.1, 500 m resolution) to extract albedo values for all of the study glaciers with an area of > 2 km2 (n = 99), using Google Earth Engine. This product has been widely used in previous studies (e.g. Dumont and others, Reference Dumont2012). Mean albedo was calculated for each glacier for every day of January and February (i.e. the austral summer) between 2010 and 2023. These daily values were then averaged to give a single annual albedo value per glacier (mean standard deviation, σ, of each glacier over the study period of 0.16). The analysed glaciers were then separated into northwest (mean σ = 0.18) and southeast sectors (mean σ = 0.14) (Fig. 1), and a median annual value calculated to represent each sector.
While not specifically assessed in this study, previous work has shown the root mean squared error between in situ and satellite measurements of albedo to be small (e.g. 0.026, Traversa and others, Reference Traversa, Fugazza, Senese and Frezzotti2021). However, given this study focuses on interannual comparisons of albedo, relative changes in albedo are more important than true values. Uncertainties typically stem from saturation problems over snow-covered areas, although this is difficult to quantify and is assumed to be negligible (Naegeli and others, Reference Naegeli, Huss and Hoelzle2019).
3. Results
3.1. Temperature and albedo changes
Both mean annual air temperature and the sum of positive degree days have increased during the study period (Fig. 2a). The lowest annual temperature in the period was in 2009/10, at −8.59 ± 0.15°C, and the highest of −3.61 ± 0.15°C in 2022/23. The average warming trend was 0.24 ± 0.08°C yr−1 (r 2 = 0.44, p = 0.01). The sum of positive degree days have also increased over the study period, with an average rate of increase of 15.0 ± 3.8 K d yr−1 (r 2 = 0.56, p < 0.01). This trend is most clear from 2018/19 when the sum of positive degree days increased from 213 K d to 406 K d in 2021/22. Despite a slight decrease in the sum of positive degree days in 2022/23 relative to the previous year, with a value of 391 K d, it had the second highest value in our study period. The lowest sum of positive degree days was in 2009/10, at 127 K d.
Figure 2. (a) Mean annual air temperature (MAAT) and annual sum of positive degree days (PDD), measured at JGM; (b) violin plot of % glacier area change 2011–17 and 2017–23, NB: the width of the plots is proportional to the number of glaciers (n = 156); (c) change in albedo for the north west (NW) and southeast (SE) glaciers, NB: shaded area shows inter-quartile range.
The rate of glacier area change (decline) increased in absolute terms in the period from 2017 to 2023 compared to between 2011 and 2017 (Fig. 2b). The median rate of glacier area loss between 2011 and 2017 was 0.06% yr−1, but then increased to 0.38% yr−1 between 2017 and 2023. While the upper quartile value for area change was maintained at 0.00% yr−1, the lower quartile rose from an area loss of 0.39% yr−1 to 0.94% yr−1. The largest changes in area are typically associated with smaller ice masses (see Supplementary Figure SI 3). Given an uncertainty in the glacier area of 1.2% yr−1, 23 glaciers receded by a magnitude greater than the uncertainty between 2011 and 2017, compared to 38 glaciers between 2017 and 2023. All of these statistically significant glacier recessions were in glaciers smaller than 6 km2 in area.
Our data suggest that median albedo has decreased through the study period (Fig. 2c) at an average rate (linear regression) of −0.07 ± 0.02 decade−1 (south eastern sector, R 2 = 0.47, p = 0.01) and −0.07 ± 0.03 decade−1 (north western sector, R 2 = 0.27, p = 0.06), and this decline is most pronounced since 2019. This general negative trend is punctuated by pronounced minima in 2012 (SE = 0.72, NW = 0.62), 2016 (0.60, 0.69) and 2023 (0.66, 0.59); these minima are most pronounced for the north-western sector (Fig. 2c).
3.2. Glacier area changes
While most glaciers have decreased in surface area during the period 2011–23 (Fig. 2b), there have been some notable exceptions. Two glaciers have gained in areal extent: Whisky Glacier (Figs 1 and 3b) and Kotick Glacier (Figs 1 and 3a). Whisky Glacier slightly reduced in area, from 28.7 ± 0.4 km2 to 28.2 ± 0.4 km2 (a loss of 1.8%) between 2011 and 2017 before the terminus advanced by ∼800 m, increasing its area to 29.5 ± 0.4 km2 (an increase of 4.7%). Kotick Glacier advanced by >800 m during the period between 2011 and 2017 when it increased in area from 5.0 ± 0.1 km2 to 5.5 ± 0.1 km2 (an increase of 11.0%). Since 2017, the terminus position of Kotick Glacier has remained approximately stable, although the glacier as a whole decreased in size to 5.2 ± 0.1 km2 (a loss of 6.0% of its area), with that area loss occurring primarily close to its headwall. In contrast, Swift Glacier’s eastern terminus position receded by 4.2 km (Figs 1 and 3c), 3.3 km of which occurred since 2017. Swift Glacier, which has its accumulation area on Mount Haddington ice cap according to the GLIMS outlines (Raup and others, Reference Raup, Racoviteanu, Khalsa, Helm, Armstrong and Arnaud2007, Fig. 1), decreased in size from 178.4 ± 0.6 km2 in 2011 to 175.9 ± 0.6 km2 in 2017 (a loss of 1.3% of its area) and was further reduced to 161.0 ± 0.6 km2 in 2023 (a loss of 8.5% of its area).
Figure 3. Area changes in glaciers on James Ross Island, including the advance of Kotick Glacier (a), Whisky Glacier (b) and the remarkable loss of area of Swift Glaciers (c). GIFS of these changes are available in the Supplementary material.
3.3. Site-specific albedo and glacier stake data
Across the entire archipelago, there has been a reduction in albedo through time (Figs 2c and 4a). On Davies Dome and Lookalike Glacier (Fig. 4a), albedo has followed the same trend as the rest of the island, with a broadly negative trend (Davies Dome = −0.05 ± 0.04 decade−1, Lookalike = −0.06 ± 0.03 decade−1) punctuated by minima in 2012, 2015, 2016 and 2023. Minima in albedo correspond with measurement of negative changes in stake height. While the relationship (Pearson’s test) between albedo and point surface mass balance is not statistically significant (p > 0.05), albedo has a closer relationship with point surface mass balance at 200–300 m (p = 0.08), compared at higher elevation (p = 0.26).
Figure 4. (a) Mean albedo for Lookalike Glacier and Davies Dome; (b) changes in point surface mass balance at Lookalike Glacier and Davies Dome glaciers at different altitudes (error bars show uncertainty).
In the years 2020/21, 2021/22 and 2022/23, negative changes in point surface mass balance were measured at all altitudes of both glaciers. Since 2009/10, this was observed only in 2011/12 and 2015/16. There has been consistent ablation measured below 300 m altitude since 2012/13, except for 2018/19 when some stakes showed positive mass balance, and in the 40–500 m bin since 2019/20. Since 2009/10, only 2 years have seen positive changes at all altitudes (2010/11 and 2012/13). The greatest change in point surface mass balance at every altitude was in 2022/23, which was recorded as −1.39 ± 0.12 m w.e. below 300 m, −0.83 ± 0.12 m w.e. at 300–400 m, −1.01 ± 0.20 m w.e. at 400–500 m and −0.60 ± 0.20 m w.e. above 500 m.
4. Discussion
4.1. Observations from other studies
In addition to the data collected in this study, we present datasets of surface mass balance, accumulation area ratio and equilibrium-line altitude data collected in other recent studies that cover our study period (2010–23), specifically for Lookalike Glacier, Davies Dome and Glaciar Bahia del Diablo (Engel and others, Reference Engel, Láska, Nývlt and Stachoň2018, Reference Engel, Láska, Smolíková and Kavan2024; WGMS, Reference Zemp, Gärtner-Roer, Nussbaumer, Welty, Dussaillant and Bannwart2023). The studies by Engel and others (Reference Engel, Láska, Nývlt and Stachoň2018, Reference Engel, Láska, Smolíková and Kavan2024) calculated annual surface mass balance using the glaciological method between January and mid-February for Lookalike Glacier and Davies Dome. The change in height of stakes was measured and converted to m w.e. (using densities for ablation and accumulation of 900 and 500 ± 90 kg m–3, respectively, and then interpolated using the nearest-neighbour technique to create a set of surface mass balance isolines). The results from Vega Island are also calculated using the glaciological method, but interpolations were conducted using a linear approach (Marinsek and Ermolin, Reference Marinsek and Ermolin2015; WGMS, Reference Zemp, Gärtner-Roer, Nussbaumer, Welty, Dussaillant and Bannwart2023). Equilibrium line altitude and accumulation area ratio were then derived from the surface mass balance isolines.
These datasets from other studies support the patterns of change identified in this study (Fig. 4b). For example, glaciological changes at Lookalike Glacier, Davies Dome and Glaciar Bahia del Diablo all demonstrate a negative trend in surface mass balance and accumulation area ratio, and an increase in equilibrium line altitude (Fig. 5). Equilibrium line altitude has increased at an average rate of 12.9 ± 3.8 m yr−1. Mean glacier-wide surface mass balance has decreased at an average rate of −0.06 ± 0.02 m w.e. yr−1 between 2010 and 2023, with accumulation area ratio also decreasing at an average rate of −3.6 ± 1.3% yr−1. The highest surface mass balance values observed at Lookalike Glacier and Davies Dome, within the 301–400 m elevation band (Fig. 4b), confirm the spatial pattern of surface mass balance described by Engel and others (Reference Engel, Láska, Nývlt and Stachoň2018). The increased accumulation in the NE section of Davies Dome was attributed to the removal of snow from the flat top of the dome and its redistribution by wind onto the leeward slope. At Lookalike Glacier, the zone of enhanced accumulation extends from the flat upper part along the eastern glacier margin down to lower elevations, influenced by snowdrifts. Decreased accumulation at the highest section of this glacier may be attributed to its steep slopes below the ice divide and snowdrift from the divide to the adjacent glacier.
Figure 5. Glaciological datasets for Glaciar Bahia del Diablo (WGMS, Reference Zemp, Gärtner-Roer, Nussbaumer, Welty, Dussaillant and Bannwart2023), Lookalike Glacier (Engel and others, Reference Engel, Láska, Nývlt and Stachoň2018, Reference Engel, Láska, Smolíková and Kavan2024) and Davies Dome (Engel and others, Reference Engel, Láska, Nývlt and Stachoň2018, Reference Engel, Láska, Smolíková and Kavan2024). These depict: (a) annual glacier-wide surface mass balance (SMB); (b) equilibrium-line altitude (ELA) and (c) accumulation area ratio (AAR).
4.2. Drivers of recent melt
The mean annual air temperature increase of 0.24 ± 0.08°C yr−1 recorded at JGM Station (using the glaciological year March to February) during the study period (2010–23) and an increase in the sum of positive degree days of 15.0 ± 3.8 K d yr−1 are exceptional, both by Antarctic standards (Supplementary Figure SI 1; see Turner and others (Reference Turner, Marshall, Clem, Colwell, Phillips and Lu2020) and their table 3) and in a global context, where temperatures have been rising rapidly since 1971 (up to 0.05°C yr−1) (Fan and others, Reference Fan, Duan, Shen, Wu and Xing2020; Osborn and others, Reference Osborn2021). Temperature records from other Antarctic stations in the region indicate that the long-term (1953/54–2023/24) increase in temperature has been a magnitude smaller than those recently recorded for James Ross Island, with Turner and others (Reference Turner, Marshall, Clem, Colwell, Phillips and Lu2020) reporting warming of 0.03°C yr−1 and 0.02°C yr−1 at the nearby Marambio and Esperanza bases, respectively. When compared directly to data acquired Esperanza Station (∼60 km NE of JGM) from the BAS Met READER dataset (Colwell, Reference Colwell2013), we find that mean annual air temperature (measured over the glaciological year of March to February) has been rising at an average rate of 0.19 ± 0.06°C yr−1 between 2009/10 and 2023/24, indicating the temperature rises observed at JGM Station are representative of the wider region. Indeed, the temperature increases and subsequent lengthening of the melt season recorded at JGM Station are in excess of observed rapid changes in the Arctic (0.1°C yr−1 on Svalbard), which is said to have undergone the Earth’s fastest rate of warming (Adakudlu and others, Reference Adakudlu2019; Arndt and others, Reference Arndt2019; England and others, Reference England, Eisenman, Lutsko and Wagner2021).
Our data show that albedo varies interannually (Fig. 2c), likely dependent on the combined effects of snowfall (accumulation), aeolian transport of dust, as well as possible debris sourcing from surrounding rocks (Naegeli and Huss, Reference Naegeli and Huss2017; Johnson and Rupper, Reference Johnson and Rupper2020; Davies and others, Reference Davies2024). Although a rise in air temperature (Fig. 2a) would be expected to increase the rate of glacial melt, a decrease in albedo would further exacerbate melt rates (Naegeli and Huss, Reference Naegeli and Huss2017; Johnson and Rupper, Reference Johnson and Rupper2020; Davies and others, Reference Davies2024). Changes in albedo induce a positive feedback, whereby dust and other debris lower glacier albedo, leading to melt and thinning which reveals more debris, which acts to keep albedo low (Naegeli and Huss, Reference Naegeli and Huss2017). We interpret the difference in albedo between the north west and south east sectors of the archipelago to highlight the importance of the Ulu Peninsula: a large glacier-free region, a source of dust that, upon aeolian deposition, lowers albedo because large dust storms are observed frequently (Kavan and Nývlt, Reference Kavan and Nývlt2018). Dust deposits have been observed to have accumulated on Triangular Glacier (Kavan and others, Reference Kavan, Nývlt, Láska, Engel and Kňažková2020; Engel and others, Reference Engel, Láska, Kavan and Smolíková2023) and Lookalike Glacier (Kavan and others, Reference Kavan, Nývlt, Láska, Engel and Kňažková2020) on the Ulu Peninsula.
More broadly, the James Ross Archipelago is home to several large proglacial areas which are abundant in fine dust (Davies and others, Reference Davies, Carrivick, Glasser, Hambrey and Smellie2012), which is likely to contribute to darkening glacier surfaces in the north-west sector, lowering albedo and thus increasing melt rates, particularly in years with less snowfall (Fassnacht and others, Reference Fassnacht, Cherry and Venable2015; Kavan and others, Reference Kavan, Nývlt, Láska, Engel and Kňažková2020). We contend that the combination of rising temperature, culminating in very high temperatures in 2022/23 and low surface albedo (Fig. 2c), is likely the cause of enhanced glacier area loss and the exceptional glacier ablation (Figs 2b and 5b) in recent years. This is consistent with modelling work, which has shown that glaciers in the northern Antarctic Peninsula region are highly sensitive to relatively small changes in air temperature, especially during periods when air temperature fluctuates around 0°C (Jonsell and others, Reference Jonsell, Navarro, Bañón, Lapazaran and Otero2012). This sensitivity has been observed across the Antarctic Peninsula (Costi and others, Reference Costi2018). This loss of glacier mass is acutely shown by in situ stake data (Fig. 4c), which records consistent ablation for 3 years. Similarly, there was a consistent thinning in both periods on Snow Hill Island and to the east of Vega Island, as well as around the periphery of James Ross Island, particularly on its southern coast (Fig. 4b). Furthermore, consistent warming since the 1950s (Supplementary Figure SI 1), means that short-term extreme weather events are likely to have been compounded by decades of warming, which in turn is likely to further decrease glacier longevity in the future (Zekollari and others, Reference Zekollari, Huss and Farinotti2020)
4.3. Drivers of glacier area change
While most glaciers on the archipelago have undergone moderate decreases in their area since 2011 (median values of 0.06% yr−1 2011–17 and 0.38% yr−1 since 2017), three glaciers have shown large aerial changes (Fig. 2b). Swift Glacier receded from a marine-terminating environment to a land-terminating setting between 2011 and 2023. This will likely mean that the region experiences a decrease in the supply of sediment (nutrients into the sea), as well as a reduction in ocean mixing and a decrease in albedo (Chu and others, Reference Chu, Smith, Rennermalm, Forster, Box and Reeh2009; Riihelä and others, Reference Riihelä, Bright and Anttila2021; Meire and others, Reference Meire2023).
Successive years of summer Landsat images (available as GIF images in the Supplementary material) reveal that the rate of Swift Glacier’s terminus recession was non-linear, with the majority (3.3 km) of this 4.2 km recession occurring in the austral summer of years 2018/19. This rate of recession is comparable to the highest rates (∼5 km yr−1) recorded for large tidewater glaciers in the Antarctic Peninsula region (Wallis and others, Reference Wallis, Hogg, van Wessem, Davison and van den Broeke2023). It is two orders of magnitude higher than most glacier recession rates recorded in the Arctic, which are typically between 10 and 35 m yr−1 (Rachlewicz and others, Reference Rachlewicz, Szczuciński and Ewertowski2007; Kavan and Strzelecki, Reference Kavan and Strzelecki2023); specifically, it is one order of magnitude greater than the highest recession rates observed on Svalbard and other glaciers in the Barent’s Sea region (∼300 m yr−1, Błaszczyk and others, Reference Błaszczyk2021; Carr and others, Reference Carr, Murphy, Nienow, Jakob and Gourmelen2023), and more akin to exceptional retreat rates observed in northern Greenland (>1 km yr−1, Carr and others, Reference Carr, Stokes and Vieli2017). We propose that this frontal recession of the Swift Glacier ablation tongue could be due to a combination of warming air and ocean temperatures (Cook and others, Reference Cook, Holland, Meredith, Murray, Luckman and Vaughan2016), and record low sea ice extent in the Weddell Sea in the 2018/19 season (Jena and others, Reference Jena, Bajish, Turner, Ravichandran, Anilkumar and Kshitija2022). In addition to this, rising equilibrium line altitudes (Fig. 5b) indicate that the equilibrium line altitude of Swift Glacier is likely to regularly be high enough to intersect with the glacier head wall. This would result in most of Swift Glacier’s area being in the ablation area, with any snowfall it receives completely melted by the end of the season. Swift Glacier’s topographic setting, flowing from the Mount Haddington Ice Cap into a lower elevation cirque (Fig. 1), means it is vulnerable to disconnection from the Mount Haddington Ice Cap (Rippin and others, Reference Rippin, Sharp, Van Wychen and Zubot2020; Davies and others, Reference Davies2022, Reference Davies2024), although without any direct observations of avalanching it is not possible to say with certainty that this has occurred. A sudden decrease in nourishment to the glacier tongue, brought on by a disconnection of the outlet glacier from its ice cap, has previously been observed to cause the rapid recession of glaciers in North America (Davies and others, Reference Davies2024), and we interpret the cause of Swift Glacier’s recession to be mechanistically similar. This loss of ice flow through to the glacier tongue has resulted in substantial thinning (as highlighted by data from Hugonnet and others (Reference Hrbáček, Engel, Kňažková and Smolíková2021); Fig. 6a, b). Consequent lowering of the terminus has revealed exposed bedrock and thermokarst features (see Supplementary Figure SI 2) that are characteristic of stagnant, degrading ice masses elsewhere (e.g. Schomacker and Kjær, Reference Schomacker and Kjær2008; Błaszkiewicz and others, Reference Błaszkiewicz, Andrzejewski, Dudek, Sobota and Czarnecki2023). While this has been interpreted as a substantial recession, disconnected ice masses often experience increased debris cover due to the diminished influx of new ice and the resulting ice stagnation (Davies and others, Reference Davies2022) and further research should focus on collating velocity data to verify this interpretation.
Figure 6. Rate of surface elevation change for (a) 2010–14 and (b) 2015–19 (Hugonnet an others Reference Hrbáček, Engel, Kňažková and Smolíková2021).
Whisky Glacier and Kotick Glacier are both unusual due to experiencing large advances in their terminus positions. Having receded between 2011 and 2017, the Whisky Glacier terminus appears to have advanced slightly (∼200 m) between 2017 and 2022, followed by a 400 m advance in a single year between 2022 and 2023. While this terminus advance may represent a surge event, in the absence of surface elevation and velocity data it is difficult to make this interpretation with any certainty. This advance occurred despite calving (visible in satellite images) and thinning occurring at the terminus of the glacier (Fig. 6). Further attention should be given to Whisky Glacier in the future to determine the cause of the advance.
The terminus advance of Kotick Glacier coincides with a period for which there is ice surface velocity data (NASA MEaSUREs ITS_LIVE project data (Gardner and others, Reference Gardner2018) and surface elevation data (Hugonnet and others, Reference Hrbáček, Engel, Kňažková and Smolíková2021)). The terminus of Kotick Glacier advanced by 350 m in 2014/15 and 300 m in 2015/16 (Fig. 3a) and this corresponds to an ice surface velocity increase from 41 ± 29 (1σ) m yr−1 to 130 ± 29 m yr−1 (Fig. 7a). This suggests that Kotick Glacier may be a surging glacier. This interpretation is further supported by the pronounced surface elevation gain (often referred to as ‘bulging’) present at the glacier terminus (Fig. 7b), which is shown by an increase in the rate of glacier thickening at the glacier terminus, although it should be noted that these elevation changes are associated with relatively large uncertainties (Hugonnet and others, Reference Hrbáček, Engel, Kňažková and Smolíková2021, Fig. 7). This phenomenon is frequently observed as glaciers reach the active phase of a surge cycle (Clarke and Blake, Reference Clarke and Blake1991). This occurred prior to the terminus advance and coincided with a reduced rate of thinning upstream of the bulge. Following the terminus advance in 2015, the rate of thinning on the bulge increased (Fig. 7c). These changes occurred coincidentally with a decrease in glacier velocity in 2016 to 121 ± 33 m yr−1 and more obviously to 70 ± 22 m yr−1 in 2017 as it relaxed back into a state of relative quiescence. Therefore, to our knowledge, this is the first surge-type glacier to have been identified in Antarctica from velocity and surface elevation change. We note that many glaciers in the Antarctic Peninsula region have been observed to suddenly increase in velocity (De Angelis and Skvarca, Reference De Angelis and Skvarca2003; Glasser and others, Reference Glasser, Scambos, Bohlander, Truffer, Pettit and Davies2011); however, these events were associated with the collapse of the Larsen A, Larsen B and Prince Gustav Ice shelves and are likely to be the consequence of de-buttressing on the glaciers (Rignot and others, Reference Rignot, Casassa, Gogineni, Krabill, Rivera and Thomas2004; Joughin and others, Reference Joughin, Shapero, Smith, Dutrieux and Barham2021). Previous work (Sevestre and Benn, Reference Sevestre and Benn2015) has highlighted that glacier geometry, in particular glacier length, is correlated with surging behaviour. With a length of 4 km, a mean annual air temperature measured at the nearby JGM of −7°C and a mean annual precipitation of up to 700 mm yr−1, the geometry of Kotick Glacier and the ‘climatic window’ are suitable for surging to occur (Sevestre and Benn, Reference Sevestre and Benn2015; Benn and others, Reference Benn, Fowler, Hewitt and Sevestre2019). Indeed, Carrivick and others (Reference Carrivick, Davies, Glasser, Nývlt and Hambrey2012b) commented that, considering the environment, it was surprising that no surge-type glaciers had been observed on James Ross Island.
Figure 7. (a) Velocity is described by the median value on the bulge evident at the front of the glacier, with the shading showing the standard deviation. Insets show the difference in the average annual surface elevation change between (Hugonnet and others, Reference Hugonnet2021) tiles for (b) 2005–09 (median uncertainty ±1.89 m) and 2010–14 (±2.31 m) and (c) 20–14 (±2.31 m) and 2015–19 (±5.90 m).
While the velocity and surface elevation data, as well as the glacier’s climatic setting and geometry, make it likely that this is a surge-type glacier, we are mindful that we have not observed any cyclicity due to an absence of velocity data for this region before 2012 and the low availability of satellite images from before 1990. It is feasible that recent temperature increases in the region may have made this surge event more likely by delivering more meltwater to the bed (Tuckett and others, Reference Tuckett2019; Benn, Reference Benn2021), although other glaciers in the region maintained a positive mass balance in 2015 when this surge event occurred (Engel and others, Reference Engel, Láska, Nývlt and Stachoň2018). Possible geomorphological evidence for surging has previously been described on the Antarctic Peninsula, but never directly observed (Nichols, Reference Nichols1973; Wellman, Reference Wellman1982), and further research looking for geomorphological evidence for surging around Kotick Glacier may be useful to determine if there is evidence of cyclicity.
4.4. Outlook
Although the recession rates revealed in this study are lower than those documented in the paleo record (Batchelor and others, Reference Batchelor2023), it is clear that some of the glacier changes observed on James Ross Island have been rapid. Given that glacier mass loss is projected to increase in the coming years (DeConto and Pollard, Reference DeConto and Pollard2016), we find it instructive to consider the longevity of the James Ross Archipelago glaciers to future melt by using annual equilibrium line altitude data (Fig. 5b) in relation to glacier hypsometry.
Parts of glaciers below the minimum mean equilibrium line altitude recorded over our study period (238 ± 115 m a.s.l., n = 3) are vulnerable to future melting (Fig. 8) and include large proportions of Snow Hill Island, the southern coast of Vega Island and the periphery of James Ross Island (which have already experienced thinning in recent years, Fig. 6). While many of these regions will remain dynamically active, those to the southeast of James Ross Island are in danger of markedly accelerated recession and thinning, much like Swift Glacier (Fig. 3c). Given that equilibrium line altitudes intersect the ice cap plateau edges, and in some places sit on the plateau itself, further disconnections are very likely as ice fluxes to the glacier tongues in the lower elevation cirques reduce. As areas at the foot of the steep headwall become increasingly exposed, these glaciers risk becoming dynamically, and subsequently physically, disconnected from their accumulation zone and will become dead-ice without any further mass input from either snowfall or avalanche from the plateaux. Additionally, the surface lowering of ice to the north-west of James Ross Island and on Vega Island also appears to reveal steep topography at the glacier bed that would make several additional glaciers in the archipelago vulnerable to disconnections in the more distant future. This would imply that many glaciers on the James Ross Archipelago are likely to substantially reduce in size in the coming decades, in addition to those already identified as vulnerable (Engel and others, Reference Engel, Kropáček and Smolíková2019).
Figure 8. Map of regions with elevation below the minimum and maximum ELAs (equilibrium line altitude, see Figure 5) according to REMA (Howat and others, Reference Howat, Porter, Smith, Noh and Morin2019). The minimum of these values was 238 m, and the maximum was 550 m.
Sites below the maximum recorded equilibrium line altitude are comparable to those most likely to face substantial melt, with some parts of this region expected to be glacier-free by 2100 under medium- and high-emission scenarios, according to modelling (using a temperature-index approach) conducted by Lee and others (Reference Lee2017). Our data suggest that significant melting will occur on many glaciers across the archipelago, perhaps indicating the projection of future proglacial area by Lee and others (Reference Lee2017) may be an underestimation if glacier disconnections occur and glaciers recede at a non-linear rate. In addition to disconnections, several other non-linear feedbacks should also be considered in the context of rising equilibrium line altitude. Firstly, a rise in equilibrium line altitude will, due to the low slope of the accumulation area of these ice caps, significantly reduce the size of the accumulation zone and lead to enhanced melt (Åkesson and others, Reference Åkesson, Nisancioglu, Giesen and Morlighem2017). As equilibrium line altitude rises, so does the area of bare ice, which has a lower albedo than snow and exacerbates melt further (Johnson and Rupper, Reference Johnson and Rupper2020; Davies and others, Reference Davies2024). Indeed, the initial melt of the glacier may itself be enhanced by dust deposits that reduce albedo, as may be the case for the glacier thinning observed at Davies Dome and Lookalike Glacier (Fig. 4). Given the low albedo (Fig. 2c) measured across the James Ross Archipelago, in combination with recent increases in equilibrium line altitude, these non-linear feedbacks are likely to affect many glaciers across the region and will lead to substantial melt, particularly if equilibrium line altitude rises above the maximum elevation of some glaciers (McGrath and others, Reference McGrath, Sass, O’Neel, Arendt and Kienholz2017). The majority of the Mount Haddington Ice Cap is currently above the maximum equilibrium line altitude and so at low risk of ablation. However, its relatively low relief means that it is sensitive to rising air temperatures and rising equilibrium line altitudes, and if equilibrium line altitudes continue to rise, it could experience widespread melt (Boston and Lukas, Reference Boston and Lukas2019), similar to that occurring on Snow Hill Island (Fig. 5). Furthermore, melt of the ice cap would mean the exposed ice would be at a lower altitude in warmer air temperatures, further decreasing the surface mass balance and creating a surface mass balance-elevation positive feedback; this could exacerbate the impact of rising equilibrium line altitudes, the implication of which is irreversible and accelerating glacier recession, as observed in other parts of the world (Davies and others, Reference Davies2024). However, it should be noted that the recession of glaciers onto land is likely to somewhat stabilise the surface mass balance of the glacier by reducing sub-glacial melt, perhaps mitigating some of this recession rate (Carrivick and others, Reference Carrivick, Smith, Sutherland and Grimes2023).
As with Swift Glacier, many more marine-terminating glaciers are likely to recede onto land as temperatures increase, as has been the case on Svalbard (Kavan and Strzelecki, Reference Kavan and Strzelecki2023). Consequently, the proglacial area of the archipelago is likely to increase in size over the coming years. This will impact marine ecosystems (Szeligowska, Reference Szeligowska2021) and increase the availability of dust, likely accelerating future glacier melt (Oerlemans and others, Reference Oerlemans, Giesen and Van Den Broeke2009). Additionally, glacier recession will change the available habitats in the region; in the Arctic, the loss of tidewater glaciers has been detrimental to the populations of many mammals and seabirds (Lydersen, Reference Lydersen2014). While the exposure of new bedrock is likely to be positive for lichens and plants (depending on the availability of liquid water), some of these species may be invasive (Heller and Zavaleta, Reference Heller and Zavaleta2009; Golledge and others, Reference Golledge, Everest, Bradwell and Johnson2010; Olech and Chwedorzewska, Reference Olech and Chwedorzewska2011).
5. Summary and conclusions
Recent rapid warming across the James Ross Archipelago has been manifested in increased mean annual air temperature of 0.24 ± 0.08°C yr−1 between 2010 and 2023. That rate is exceptional compared to the long-term average (1953–2023) of 0.03 ± 0.01°C yr−1 and has increased the number of days where melt occurs (sum of positive degree days of 15.0 ± 3.8 K d yr−1), which may explain the rate of glacier area reduction. The median loss of glacier area between 2011 and 2017 was 0.06% yr−1, but then increased to 0.38% yr−1 between 2017 and 2023, with small glaciers shown to be particularly vulnerable. These very high air temperatures, combined with decreasing albedo on glacier surfaces (average rate SE sector = −0.07 ± 0.02 decade −1; NW sector = −0.07 ± 0.03 decade −1), have caused enhanced melt rates since 2020; the most negative point surface mass balance change was −1.39 ± 0.12 m w.e. at 200–300 m a.s.l. on Lookalike Glacier and Davies Dome in 2023. Data from the World Glacier Monitoring Service (WGMS, Reference Zemp, Gärtner-Roer, Nussbaumer, Welty, Dussaillant and Bannwart2023) and Engel and others (Reference Engel, Láska, Smolíková and Kavan2024) highlight that equilibrium line altitude has also been gradually increasing and reached 550 m a.s.l. (n = 1) in 2023, up from a minimum of 238 ± 115 m a.s.l. (n = 3) in 2012. The low albedo values recorded in 2023 are likely to be the consequence of low rates of snowfall and the occurrence of localised dust storms in the archipelago’s extensive proglacial regions, which will likely exacerbate future melt rates, especially if albedo continues to decline.
Some remarkable changes to glaciers have occurred recently on the James Ross Archipelago. Swift Glacier has receded by 4.2 km since 2011, 3.3 km of which occurred in 2018/19. This is among the fastest glacier recession rates observed anywhere globally. We interpret this recession to be the result of rising air and ocean temperatures, as well as a disconnection from the Mount Haddington Ice Cap. This dramatic loss of ice highlights the non-linearity of glacier recession that can be experienced by ice cap outlet glaciers draining from plateaux edges. It illustrates the vulnerability of other glaciers on the southern and eastern coasts of James Ross Island to becoming disconnected from the Mount Haddington Ice Cap, which will fundamentally change outlet glacier dynamics and in combination reduce glacier longevity from that previously supposed.
Kotick Glacier increased in velocity in the terminus region from 41 ± 29 m yr−1 to 130 ± 29 m yr−1 between 2014 and 2015 (Gardner, Reference Gardner2018). This tripling of frontal velocity, combined with the rapid terminus position advance and formation of a bulge (increase) in ice surface elevation near the terminus, leads us to conclude that this is, to the best of our knowledge, the first surge-type glacier to have been identified in Antarctica from velocity and surface elevation change data. The cause of an advance of the terminus position of Whisky Glacier is less clear, and further research should be conducted to monitor future changes in its frontal position and velocity.
Overall, future research should seek to monitor albedo and equilibrium line altitude in situ. More attention should be given to the glaciers on the south of James Ross Island that are at risk of glacier disconnections. Future numerical modelling efforts should include the possibility of (more) extreme air temperatures in the future and of non-linear threshold processes and feedbacks associated with glacier recession (e.g. through the use of ice-dynamical models instead of surface mass balance models based on a fixed geometry).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/jog.2025.10075.
Data availability statement
The code to calculate albedo is available here: https://code.earthengine.google.com/80a2062a5a05c33c31406027e01e2e04. Other data from this study are available in the supplementary material, including the glacier outlines produced.
Acknowledgements
This work was supported by the Leeds-York-Hull NERC Doctoral Training Partnership (DTP) Panorama under grant number NE/S007458/1. The Czech Science Foundation (project number GC20-20240S) funded the glaciological and meteorological monitoring. The Czech Antarctic Research Programme provided logistics and accommodation to support C.D.S., Z.E., M.M. and K.L. at JGM in 2022 and M.M. in 2023. We also thank the reviewers for their constructive and insightful comments.
Author contributions
C.D.S. designed and led the study, produced the albedo results and prepared the manuscript. M.W.M. produced results. J.L.C. produced area results, provided ongoing advice to C.D.S. and, with D.N., edited the first full draft of the manuscript. Z.E. produced the ablation results. K.L. and M.M. produced the meteorological results. C.H. provided velocity data. D.N., D.J.Q. and B.J.D. provided text edits of the manuscript before submission. C.D.S. completed this work as part of a PhD, supervised by J.L.C., D.J.Q. and D.N. All authors contributed to writing the final version of the manuscript.
Open access
