1. Introduction
Mountain glaciers are at risk of extinction, which is complete disappearance, due to global climate change. The imbalance between the current climate and glacier geometry varies greatly between regions. Scandinavia is a region projected to experience substantial losses regardless of the degree of future warming (Zekollari and others, Reference Zekollari2025). The consequences of losing mountain glaciers span biodiversity loss, tourism decline, impacts on societal health and livelihood, and cultural heritage (e.g. Purdie, Reference Purdie2013; Ezzat and others, Reference Ezzat2025; Burrill and others, Reference Burrill, Dannevig and Brendehaug2025; Howe and Boyer, Reference Howe and Boyer2025).
Norwegian glaciers span a wide geographical range from the northernmost glaciers at 70° North in Finnmark County to the southernmost glaciers at 59° North in Rogaland County (Andreassen and others, Reference Andreassen, Winsvold, Paul and Hausberg2012). Norway has a rich record of glaciological and geodetic mass balance observations, as well as frontal variation measurements on many glaciers (Andreassen and others, Reference Andreassen, Elvehøy, Kjøllmoen and Belart2020). However, most glaciers lack field observations and have only been studied through repeat glacier inventories (e.g. Østrem and others, Reference Østrem and Ziegler1969; Reference Østrem, Dale Selvig and Tandberg1988), which limits the depth and scope of possible analysis. An increasing number of historical aerial photographs are being scanned and made available through the Norwegian Mapping Authority’s web portal, Norgeibilder.no as orthophotos. These datasets offer higher spatial resolution for analysis and allow for interpretations predating the satellite era. A comparison of orthophotos from 1978 with the most recent 2019 glacier inventory revealed that Breifonn has greatly reduced in area over the 41 year period (Andreassen, Reference Andreassen2022). Breifonn was therefore included in the Global Glacier Casualty List (GCCL) launched in August 2024 (Boyer and Howe, Reference Boyer and Howe2024). The GCCL is a platform dedicated to documenting and commemorating glaciers that have disappeared or soon-to-disappear due to climate change (Raup and others, Reference Raup, Andreassen, Boyer, Howe, Pelto and Rabatel2025).
In this paper, we document changes in the area of Breifonn from its ‘pre-industrial’ Little Ice Age (LIA) extent to 2024, using geomorphological mapping, glacier inventories, historical maps, aerial photographs and recent satellite imagery. We derive digital terrain models (DTMs) from historical photos from 1955 and 1978 and make a new DTM from our own survey using Uncrewed Aerial Vehicle (UAV) during a field visit to the glacier in August 2024 to assess elevation changes and geodetic mass balance. We also assess the associated methodological uncertainties, which arise primarily from limited geomorphological evidence and challenges in delineating glacier outlines and reconstructing Breifonn’s ice surface topography due to persistent snow cover. Our aim is to document the profound changes the glacier has undergone over the industrial period and discuss whether Breifonn can still be considered a glacier.
2. Setting
Breifonn (59.75°N, 6.89°E) is a small glacier on the border between Rogaland and Vestland counties (Fig. 1). It is the southernmost glacier in Norway and has been included in all the published glacier inventories of southern Norway (Liestøl, Reference Liestøl, Hoel and Werenskiold1962; Østrem and others, Reference Østrem and Ziegler1969; Reference Østrem, Dale Selvig and Tandberg1988; Andreassen and others, Reference Andreassen, Winsvold, Paul and Hausberg2012; Winsvold and others, Reference Winsvold, Andreassen and Kienholz2014; Andreassen and others, Reference Andreassen2022). Some ice bodies further south were included in the recent glacier inventories, but since they were smaller than 0.1 km2 they were categorised as snow or ice patches (Andreassen and others, Reference Andreassen, Winsvold, Paul and Hausberg2012; Andreassen, Reference Andreassen2022). Breifonn has also been spelt as Breidfonn, Bredfond and Breifond on historical maps. To our knowledge, no glaciological investigations of Breifonn exist beyond the national glacier inventories.
Figure 1. Location map showing (a) Breifonn in southern Norway, south of the largest glaciers Folgefonna, Hardangerjøkulen and Jostedalsbreen, and (b) Breifonn on the border between the counties Vestland and Rogaland. The glacier outlines and selected glacier IDs (3097 and 3099) are from 2019 (Andreassen and others, Reference Andreassen, Nagy, Kjøllmoen and Leigh2022). Coordinates are geographical (a) and UTM Zone 32 N (b).
Information on Breifonn’s extent and evolution prior to the 1950s is sparse and fragmented. Southern Norwegian glaciers generally advanced to more extended positions during the LIA (Grove, Reference Grove2004), offering a useful proxy for their ‘pre-industrial’ state. Although there is no direct dating of Breifonn’s LIA advance, data from the nearby Folgefonna ice cap suggest a maximum LIA extent between the 1870s (Nussbaumer and others, Reference Nussbaumer, Nesje and Zumbühl2011) and the 1930s (Tvede, Reference Tvede1973). On historical maps, Breifonn appears in full only on a hand-drawn portfolio map sheet (porteføljekart) by Staib (Reference Staib1860), part of Norway’s main map series produced between approximately 1817 and the 1860s (Harsson and Aanrud, Reference Harsson and Aanrud2016). The section of the map featuring Breifonn is based on field surveys carried out in 1856–57, documented in two rectangular survey maps (rektangelmålinger) by Scharffenberg (Reference Scharffenberg1856) and Naeser (Reference Naeser1857). In these maps, Breifonn is shown as covering substantial portions of the plateau summit. Several adjacent mountain plateaus appear similarly covered, as indicated by the distinctive turquoise colouration used for these summit areas. The accompanying field survey reports, which provide detailed topographic descriptions of the mapped area, refer to these summits as ‘sneefjelder’ (‘mountains covered with snow year-round’). This suggests that most of these features likely represented mountain plateaus covered with perennial snowfields, rather than actual ice masses. Mountaineers’ accounts in Norwegian Tourist Association yearbooks provide early 20th-century descriptions and photographs of the glacier (Wareberg, Reference Wareberg1927; Johannessen, Reference Johannessen1928; Anda and Sagen, Reference Anda and Sagen1941). Johannessen (Reference Johannessen1928) dubbed Breifonn ‘Norway’s South Pole’ and claimed an area of 30 km2, likely an overestimate influenced by the exceptionally cold summer and widespread snow that year. Anda and Sagen (Reference Anda and Sagen1941) describe Breifonn as a ‘gently curved shield […], blending seamlessly into the surrounding landscape without any abrupt drop’ (p. 99; translated from Norwegian), with a smooth, crevasse-free surface.
Breifonn is located west of the main water divide in an area of high precipitation in southwestern Norway. It has an annual precipitation of about 2000 mm and a mean annual air temperature of −2.5°C (estimated for 1579 m a.s.l.) for the reference period 1991–2020 (SeNorge, 2025). In 2003, Breifonn spanned an elevation range from 1417 to 1601 m asl, with a slope of 7 degrees and a north-facing aspect (Andreassen and others, Reference Andreassen, Winsvold, Paul and Hausberg2012). The bedrock of the plateau summit on which Breifonn rests is composed of mica gneiss (Sigmond, Reference Sigmond1975).
3. Data and methods
3.1. Glacier inventories
All published glacier inventories of southern Norway/Norway have included area estimates on Breifonn. In the following, we detail the information and sources used. In the first ‘List of areas and numbers of glaciers’, Breifonn is listed with an area of 3.2 km2 with reference to 1955 (p. 50, Liestøl, Reference Liestøl, Hoel and Werenskiold1962). The same images or map were used in the first inventory of glaciers in South Norway (Østrem and Ziegler, Reference Østrem and Ziegler1969), in which Breifonn is divided into two parts, units 7 and 8. The northwestern part has an area of 2.65 km2, whereas the part facing southeast has an area of 0.64 km2, totalling 3.27 km2. In the inventory of South Norway published in 1988, Breifonn is listed as Breidfonn, as one unit, with an area of 2.61 km2 (Østrem and others, Reference Østrem, Dale Selvig and Tandberg1988). None of these inventories is available with digital outlines, only tabular data. Digital outlines exist from a Landsat 1988–97 glacier inventory (Winsvold and others, Reference Winsvold, Andreassen and Kienholz2014). Here, Breifonn is mapped from a Landsat 5 TM 1988 image, and the area of the main part is 1.91 km2 (Fig. 2c). Breifonn is also covered using a Landsat 5 TM scene of 9 August 2003 (Andreassen and others, Reference Andreassen, Winsvold, Paul and Hausberg2012). Breifonn is listed with ID 3099 and with several small units around it (Fig. 2d). The area of Breifonn (3099) is listed as 1.37 km2. In the latest inventory made using Sentinel-2 imagery, the image covering Breifonn was taken on 27 August 2019; the glacier had disintegrated in several parts (Fig. 2e) with the largest remaining part (3099) having a size of just 0.26 km2 (Andreassen, Reference Andreassen2022).
Figure 2. The demise of Breifonn from 1955 to 2024 was displayed using orthophotos (Ortho) and Satellite imagery (Landsat 5 TM and Sentinel-2). All outlines are from the year of mapping. Note the lake in the upper left corner that has emerged as the glacier has now retreated from it.
Another digital dataset that digitised glacier outlines from first-edition 1:50,000 topographical maps in the N50 series from the Norwegian Mapping Authority (Winsvold and others, Reference Winsvold, Andreassen and Kienholz2014). Breifonn was in that study digitised from the N50 sheets 1314-1 Røldal and 1314-2 Suldalsvatnet that were based on aerial photographs acquired in 1953, 1955 and 1961. Likely it is the 1955 photos used for the part covering Breifonn. The outline on the maps is patchy and seemingly includes many snowfields (Fig. 3b). The total area of the continuous part of Breifonn in the N50 product is 5.5 km2.
Figure 3. Glacier outlines of Breifonn from the ‘pre-industrial’ LIA to 2024. (a) LIA and 2019 outlines shown with a Sentinel-2 satellite image in false colours from 6 September 2024 (Copernicus Sentinel data). The dark grey-brownish zone (in the centre of the figure) southeast of the Breifonntjørna lake is interpreted as the footprint left by glacial coverage during the LIA. (b) Outlines shown as coloured polygons with a terrain shadow as background (Norwegian Mapping Authority). The 2024 outline was manually digitised from UAV orthophoto (Figure 4b).
3.2. Aerial photos
Several aerial photos of Breifonn exist for the period 1955–2019, but not all are available digitally from the Norwegian Mapping Authority’s web portal Norgeibilder.no as orthophotos. All orthophotos available from Norgeibilder.no were checked as part of this study. Many of the orthophotos are from a time of severe snow conditions and so are not optimal for assessing glacier extent or condition. In the photos from 1981, 2012 and 2013, the glacier and surroundings were almost fully snow-covered; in the photos from 1955, most of the glacier was snow-covered (Fig. 2a), whereas the images from 1978 (Fig. 2b) and 2019 (not shown) had the best/least snow conditions. We therefore ordered the 1955 and 1978 photos as individual photos from the Norwegian mapping authorities for further data processing to generate DTMs and derive elevation changes, and compute geodetic mass balances.
3.3. Digital terrain models
We constructed several new DTMs for this study using a DTM from 2019 from the Norwegian Mapping Authority as a reference. In the following, we describe each DTM.
3.3.1. 2019
A photogrammetry-derived point cloud, as well as 1 and 10 m DTMs, is available from the Norwegian Mapping Authority based on the 2019 aerial photographs. The exact acquisition dates of the 2019 survey are not known, as the images were taken over a range of days. The 2019 DTM at 1 m resolution was used as a reference for other DTMs in this work. No airborne laser scanning is available for this region.
3.3.2. 1955 and 1978
Aerial imagery from both 1955 (five images, mean altitude 3690 m a.s.l.) and 1978 (five images, mean altitude 2430 m a.s.l.) was aligned using Structure-from-Motion (SfM) photography in Agisoft Metashape. Camera calibration reports were not available, so the camera’s mathematical model was reconstructed by identifying the fiducial markers in each image and subsequently using automatically collected tie points between the images. In the absence of any control point data for the historical images, they were initially aligned in an arbitrary coordinate system, but then georeferenced to manually-identified relatively fixed points (e.g. large bedrock features) using the 2019 orthoimagery and accompanying elevation model for reference. Using these coordinates as manual Ground Control Points (GCPs), the 1978 model was aligned using the nine GCPs with a mean RMS error of 2.34 m. The same process was followed for the 1955 images using eleven GCPs with an RMS error of 2.40 m. DTMs were generated without interpolation to ensure only valid image matches were used in the DTM differencing. DTMs for the 1955 and 1978 surveys were exported at 1 m resolution, while orthomosaics were exported at 0.4 and 0.2 m in 1955 and 1978, respectively.
3.3.3. 2024 UAV
On 28 August 2024, we conducted a UAV-based SfM photogrammetry field survey covering the extent of the Breifonn glacier to provide an updated and high-resolution assessment of the extent and to construct a DTM (Smith and others, Reference Smith, Carrivick and Quincey2016). Aerial imagery was acquired using a DJI Mavic 3 Enterprise UAV. Ground control was unavailable during the field survey; however, the imagery was directly georeferenced using a real-time kinematic positioning (RTK) module communicating with a DJI D-RTK 2 high-precision real-time kinematic positioning (GNSS) receiver acting as a base station, located at the ice margin. While not as accurate as an independent GNSS survey of GCPs located within the study area (James and others, Reference James, Robson and Smith2017), the centimetre-level positioning was deemed sufficient for these mapping purposes, and the relatively lightweight field equipment was more optimal for a remote field location. In total, 363 images were collected from manual flight lines with a minimum 80% sidelap and 60% frontlap between successive images. Additional off-nadir images were taken along flight lines to minimise any doming effects arising in the image alignment process.
SfM photogrammetry was performed using Agisoft Metashape (version 2.0.2) following the workflow of James and others (2017). The mean image location accuracy was <0.02 m. In total, 347 images aligned successfully, with a 3D root mean square (RMS) error of 0.068 m. Dense point clouds were created and rasterised to produce a 0.227 m resolution DTM. An orthomosaic image was created at 0.0057 m resolution. Both data products were exported in the ETRS 1989 UTM Zone 32 N coordinate system. Errors are well constrained within the glacier outline that was the focus of the field survey and used to calculate glacier geometry; however, errors will be larger on the ice-free surfaces outside of this region.
3.4. Glacier outline delineation and LIA reconstruction
Glacier outlines were digitised from orthophotos from 1955, 1978, 2019 and 2024. The existing N50 outline digitised from the topographic maps was very patchy (Fig. 3b). Most of the glacier was snow-covered, and the outline digitised at the time of mapping likely mapped both snow and ice. The 1955 outline was derived using the N50 outline as reference and edited to fit the 1955 orthophoto excluding many of the likely snowfields (Fig. 2a). The 1978 outline was already digitised from orthophoto (Andreassen, Reference Andreassen2022) and here edited slightly by removing some of the patchy outlines in the southern part (Fig. 2b). We also used the DTM differencing results to aid the outline editing. The 1988, 2003 and 2019 outlines from satellite imagery were used without modifications. These outlines were derived using a semi-automated method from satellite images (Andreassen and others, Reference Andreassen, Winsvold, Paul and Hausberg2012, Reference Andreassen2022; Winsvold and others, Reference Winsvold, Andreassen and Kienholz2014). The 2024 outline was manually digitised from the UAV orthophoto covering the main part of the present glacier (Fig. 4b).
Figure 4. Elevation change (m a-1) for Breifonn from 1978 to 2019 and 2019 to 2024. Background orthophoto in (b) is from the UAV survey of 28 August 2024. Note the difference in scale, the box in (a) is the extent of (b).
We estimated Breifonn’s ‘pre-industrial’ LIA glacier extent by adopting an approach used elsewhere in Norway, including for Jotunheimen (Baumann and others, Reference Baumann, Winkler and Andreassen2009), Hardangerjøkulen (Weber and others, Reference Weber, Boston, Lovell and Andreassen2019), Langfjordjøkelen (Weber and others, Reference Weber, Lovell, Andreassen and Boston2020) and Jostedalsbreen (Carrivick and others, Reference Carrivick, Andreassen, Nesje and Yde2022). This approach involved the visual interpretation of aerial photographs available in ArcGIS Pro base images, Norgeibilder (with the 2019 imagery also available as colour-infrared imagery) and a hill-shaded image of the 2019 DTM to identify glacial landforms around Breifonn. Specifically, the landforms identified were predominantly glacial lineations (grooves or ridges formed by glacier movement, showing ice flow direction), erosional boundaries (the transition between recently ice-moulded bedrock and more weathered and/or lichen-covered bedrock), glacial drift limits (the boundary between terrain covered by glacially transported sediments and adjacent surfaces of different composition or older glacial drift), as well as ice-marginal moraines (ridges deposited at the glacier margin). Typically, landforms and glacial drift deposits relating to the LIA appear relatively fresh, clean, unvegetated, unweathered and uncolonised by lichen. Mapping of LIA glacial landforms was further directed by the historical glacier inventories, and a field visit in August 2024 was used to inform some of the mapping. The rather vague, disparate (and undated) erosional and drift boundaries, as well as ice-marginal moraines, were joined to create a continuous glacier outline for the LIA. This outline is a best guess, digitised from one glacial landform to the next, following topography logically.
The satellite-derived glacier inventories have an estimated overall area uncertainty of ± 3%, but with larger relative uncertainties for smaller glaciers (Andreassen and others, Reference Andreassen, Winsvold, Paul and Hausberg2012, Reference Andreassen2022; Winsvold and others, Reference Winsvold, Andreassen and Kienholz2014). Due to Breifonn’s smaller size and uncertainties related to snow around the perimeter, we therefore estimate an area uncertainty of ± 10% for the satellite-derived outlines (1988, 2003, 2019) as well as the orthophoto-derived outlines (1955, 1978, 2024). As noted by Paul (Reference Paul2013), the precision of the derived glacier area is not necessarily higher when using high-resolution data. In the case of Breifonn, the snow-covered parts cause the greatest uncertainty. For the reconstructed LIA glacier extent, we estimate a relative area uncertainty of 20%. This is based on the studies by Carrivick and others (Reference Carrivick, Andreassen, Nesje and Yde2022) and Reinthaler and Paul (Reference Reinthaler and Paul2023), who both consider uncertainty from subjective interpretation of geomorphological features and from variability in digitising glacier outlines.
3.5. DTM differencing
The DTMs from 1955, 1978, 2019 and 2024 were co-registered and differenced using the XDEM Python library (Hugonnet and others, Reference Hugonnet2024). For each successive DTM pair, the older elevation model was co-registered relative to the more recent elevation model (except for the 2024 model, which was co-registered against the 2019 dataset). Non-linear co-registration biases were corrected using an Iterative Closest Point alignment, which can account for rotations between datasets. Subsequently, we followed the method set out by Nuth and Kääb (Reference Nuth and Kääb2011), which seeks to minimise the RMSE slope normalised elevation biases over stable terrain to solve sub-pixel linear alignments. The co-registration shifts are summarised in Table S1.
The resulting elevation change rasters had voids over areas with low image contrast, such as fresh snow or shadows. These voids were filled using a third-order polynomial hypsometric binning approach (after McNabb and others, Reference McNabb, Nuth, Kääb and Girod2019) with a bin size of 10 m. It was deemed that the voids on the 1955 DTM, however, were too substantial to fill, and as such this DTM was used only for comparing thinning rates within the small part (9.5%) covered.
We calculated the geodetic mass balance, Bgeod:
\begin{equation}{B_{geod}} = \Delta V \times {f_{\Delta V}}/A\end{equation}where A is the average glacier area of the two surveys, assuming a linear change in time, and ΔV is the volume change. We applied a density conversion factor, fΔV, of 850 ± 60 kg m−3 (Huss, Reference Huss2013). Glacier elevation-change uncertainty was estimated with the xDEM implementation of Hugonnet and others (Reference Hugonnet, Brun, Berthier, Dehecq, Mannerfelt, Eckert and Farinotti2022), which models heteroscedastic, spatially correlated errors via variograms from stable terrain and terrain slope/curvature, producing gridded error fields. These spatial errors were propagated and combined with the density conversion term and a 10% area uncertainty to obtain the total geodetic mass-balance uncertainty.
4. Results
4.1. Area change
Our LIA reconstruction indicates that the ‘pre-industrial’ glacier extent of Breifonn was 5.8 ± 1.2 km2 (Table 1). Breifonn was covering the bowl-shaped central part of the plateau summit area, including the area now occupied by the Breifonntjørna lake (Fig. 3a). The terrain surface in this area is ice-moulded and draped with a thin layer of fresh-looking glacial drift that is densely streamlined and fluted, providing strong evidence of recent LIA glacial coverage. This glacial imprint is readily identifiable in the 2019 aerial imagery as a brownish, rust-coloured zone that contrasts sharply with the surrounding greyish terrain. In the Sentinel-2 imagery shown in Figure 3a, this area appears as a darker grey-brownish zone, clearly distinct from the lighter brownish surroundings. However, the transition between these two zones is often indistinct and diffuse, making it challenging to pinpoint and map the exact location of erosional and drift boundaries. This introduces some degree of uncertainty regarding the lateral and frontal positions of the reconstructed LIA glacier margin. Areas of the plateau summit lying outside this zone are largely clean, lacking lichen colonisation, and exhibit a grey, almost ‘bleached’ appearance. This suggests that these plateau summit surfaces remained permanently snow-covered during the LIA. Furthermore, our reconstruction indicates the presence of five very small cirque glaciers along the plateau edges, with a combined area of approximately 0.7 ± 0.2 km2. Of these, glacier ID 3097 is included in the N50, 1988, 2003 and 2019 inventories, while glacier ID 4088 appears only in N50 and 2019 inventories (Fig. 3b).
Table 1 Area changes of Breifonn from published data and produced in this study. The 1955 outline is used as a reference after the breakup of Breifonn into the main glacier part (ID 3099) and other fragmented parts (other) after 1978. AP, aerial photos; map, topographical map; LIA, Little Ice Age maximum extent. Area estimates are reported with a ± 10% uncertainty range for outlines derived in this work, except for the LIA outline, which we estimate to 20%; see main text for details.
In 1955, the area of Breifonn was 3.3 ± 0.3 km2, and in 2024, the area of the largest remaining part was only 0.17 ± 0.02 km2. Counting all glacier parts within the 1955 outline shows a reduction in glacier area from 3.3 ± 0.3 to 0.22 ± 0.02 km 2 in the period 1955–2024, resulting in an overall area reduction of 93% or 1.4% a−1 over the 69 years. The orthophotos and satellite imagery show that Breifonn split into several parts between 1978 and 1988 (Figs. 2b,c and 3b).
4.2. Elevation change and geodetic mass balance
The DTM differencing shows that Breifonn has thinned continuously over all time periods studied. Between 1978 and 2019 (41 years), the glacier thinned by an average of −0.40 ± 0.02 m a-1, with the largest thinning occurring at the centre of the glacier, with a total thinning of up to 45 m (Fig. 4). The geodetic mass balance over the 1978–2019 period is −0.58 ± 0.07 m w.e. a−1. Over the last period 2019–24 (5 years), the main remnants of Breifonn (ID 3099 and 4080) experienced average thinning of 0.75 ± 0.11 and 0.90 ± 0.15 m a−1, respectively, with maximum thinning of nearly 12 m over the five-year period. The two other smaller parts (4081 and 4083) covered by the 2024 survey showed less thinning of 0.10 ± 0.16 and 0.03 ± 0.16 m a−1, respectively. The geodetic mass balance of Breifonn (3099) in the 2019–24 period is −0.78 ± 0.25 m w.e. a−1. As mentioned, we were not able to derive a glacier-wide estimate for 1955–78 due to significant voids in the 1955 DTM. For the portion of the DTM that had elevation values, the glacier thinned on average by 0.77 ± 0.09 m a−1 between 1955 and 1978. Given that the accumulation area was not covered, these values are not considered representative of the entire glacier.
5. Discussion
5.1. Factors influencing area estimates
As the glacier outlines of Breifonn are partly covered in snow, the interpretation of snow fields influences the resulting areas and area changes. For instance, the area was greatly reduced by using the 1955 orthophoto instead of the N50 outlines, from 5.5 to 3.3 km2. Our estimate is, however, similar to the glacier inventories of Liestøl (Reference Liestøl, Hoel and Werenskiold1962) and Østrem and others (Reference Østrem and Ziegler1969) using the same photos. Furthermore, small edits of the southern part of the 1978 extent of Breifonn reduced the area from 2.51 to 2.38 km2, a difference of 5% showing that the area will vary depending on human interpretation of orthophotos in line with previous studies (e.g. Paul and others, Reference Paul2013). Two orthophotos of Breifonn in 2019 are available on Norgeibilder.no. One from 26 July 2019 has several snow fields resulting in an area of 0.58 km2 for the biggest part (3099) when including these in the outline. The one from 21 September 2019 is fully covered in a fresh layer of snow and was therefore not used for area estimates. The Sentinel-2 outline we use for our calculations gives an area of 0.26 km2 for Breifonn on 27 August 2019. This emphasises the need for checking sources and selecting orthophotos or satellite data with little seasonal snow cover if available. In the case of Breifonn and many other glaciers in Norway, cloud conditions and seasonal snow are the main hindrances to obtaining accurate data (e.g. Paul and others, Reference Paul, Andreassen and Winsvold2011).
The reconstructed LIA outline represents our best estimate, with accuracy constrained by limited and subdued landform evidence. Geomorphology-based LIA reconstructions are most effective where glaciers left abundant landforms, typically latero-frontal moraines and trimlines (e.g. Osipov and Osipova, Reference Osipov and Osipova2019; Lee and others, Reference Lee, Carrivick, Quincey, Cook, James and Brown2021). In such cases, LIA extents are delineated by extending modern outlines down-glacier, following trimlines to the outermost LIA moraine limit (e.g. Carrivick and others, Reference Carrivick, James, Grimes, Sutherland and Lorrey2020; Tielidze and others, Reference Tielidze, Mackintosh, Gavashelishvili, Gadrani, Nadaraia and Elashvili2025). Up-glacier sections are often left unmodified, either assuming negligible post-LIA change in accumulation areas or because cirque- and valley-type glaciers are topographically confined (e.g. Stokes and others, Reference Stokes, Andreassen, Champion and Corner2018; Reinthaler and Paul, Reference Reinthaler and Paul2023; Zhang and others, Reference Zhang, Xu, Sun, Li and Xu2024). In contrast, Breifonn’s unconstrained plateau setting and flat, sheet-like geometry required reconstruction around its entire perimeter, not only down-glacier. Given the limited geomorphological evidence at Breifonn, the ± 20% area uncertainty we estimated may represent a lower bound, as interpretation uncertainty dominates the total error (Reinthaler and Paul, Reference Reinthaler and Paul2023). Nevertheless, the availability of a wide range of high-resolution remote sensing datasets allowed us to map the subtle features at Breifonn and so reduce uncertainties related to landform visibility.
5.2. Breifonn’s classification as a glacier
In this paper, we document the rapid shrinking of Breifonn. The field visit in August 2024 revealed that it now consists only of fragmented ice patches. Can we still classify Breifonn as a glacier? Definitions of glaciers vary in the literature and are used differently. Cogley and others (Reference Cogley2011) define a glacier as a ‘perennial mass of ice, and possibly firn and snow, originating on the land surface by the recrystallisation of snow or other forms of solid precipitation and showing evidence of past or present flow’. Thus, it is sufficient to have evidence of past flow according to this definition. Some studies use minimum size thresholds that can depend on the spatial resolution of data sources used for the inventories or be suitable when assessing large glacier regions. However, in regions where ice bodies are sparse, even small glaciers may be of significant interest. When mapping with high-resolution imagery (<1 m) with minimal seasonal snow cover, glaciers <0.05 km2 and even <0.01 km2 are readily identifiable, and using a minimum size threshold may be inappropriate (Leigh and others, Reference Leigh2019). To address this, Leigh and others (Reference Leigh2019) proposed a scoring system (with a possible maximum score of 20 points) to enable the identification of very small glaciers and classify them as ‘certain’ (11–20 points), ‘probable’ (6–10 points) or ‘possible’ (2–5 points), perennial snow (1 point) based on the following diagnostic features indicative of glacier motion: crevasses, bergschrund, visible ice, evidence of past or present flow such as flow features and deformed stratification. This classification system follows the glacier definition of Cogley and others (Reference Cogley2011).
The orthophotos from 1955, 1978, 2006, 2019 and 2024 reveal glacier ice and evidence of past flow. The 2019 orthophotos have the highest spatial resolution and best coverage of the recently deglaciated glacier foreland. This proglacial area has a densely fluted surface with a small number of ice-marginal moraines. With a total score of 18 out of 20 points according to the Leigh and others (Reference Leigh2019) classification system, Breifonn can be categorised as a ‘certain’ glacier. The only criterion it does not meet is the bergschrund, due to its initial formation as a plateau glacier, rather than a cirque or mountain glacier. Although Breifonn appears thin and downwasting now, our results reveal that its central parts were up to 50 m thicker in 1978 and thus were thick enough for the ice to deform and flow. However, the aerial photos from 1978 reveal supraglacial drainage channels typical of less active glaciers.
The 2024 field visit revealed that both the detached parts and the ice remnants of Breifonn contained clear blue glacier ice, crevasses and evidence of past flow (Fig. 5). However, Breifonn appeared too thin to sustain any significant flow. The visit also showed that even in late summer, seasonal snow can remain around the glacier margins, making it difficult to precisely define the actual outline, and thus the area, of the ice bodies. Breifonn has diminished rapidly in recent years and is expected to continue to shrink. How long it will take to completely vanish depends on both future summer melting and winter precipitation. With current rates of retreat, it may take less than 10 years. Snow-rich years may temporarily slow the loss and prolong the survival of some of Breifonn’s remnants.
Figure 5. Photographs of Breifonn showing (a) a historical view of the glacier and its surroundings from afar, taken between 1947 and 1949 (Owner: Nasjonalbiblioteket); (b) the middle section of the present-day glacier; (c) one of its detached parts. Both (b) and (c) were taken on 28 August 2024 by Liss M. Andreassen.
Breifonn presently spans a narrow elevation range of less than 200 m. Glaciers with small elevation ranges are particularly susceptible to climate warming as they have less potential to retreat to higher elevations (Zekollari and others, Reference Zekollari2025). In contrast to Breifonn, glacier ID 3097 appears to have experienced little change in area since its ‘pre-industrial’ extent (Fig. 3b). Although it has retreated from its LIA position, its cirque-like setting against a cliff face likely helps to retain its size due to snow drift, its protected location and northern exposure (Fig. S1).
6. Summary and conclusions
We document the demise of Breifonn, Norway’s southernmost glacier, from its maximum LIA extent to its present (2024) state as a vanishing glacier. Ice-marginal snow covers part of the glacier perimeter in all available photographs and images of Breifonn and complicates glacier outline mapping and change assessments. Reconstructing the ‘pre-industrial’ LIA extent of the glacier proved challenging due to the sparsity of geomorphological evidence. Despite this, our results clearly reveal that Breifonn is rapidly shrinking and has changed from an active glacier to a vanishing glacier. From the ‘pre-industrial’ LIA to 1955, it nearly halved in size from about 5.8 ± 1.2 km2 to 3.3 ± 0.3 km2. Between 1955 and 2019, Breifonn lost more than 93% of its area and disintegrated into several smaller ice patches. As of 2024, the largest remaining part with ID 3099 measured only 0.17 ± 0.02 km2. The central part of the glacier experienced surface lowering of up to ∼1 m a−1 over the period 1978–2024. We estimate a geodetic mass balance of −0.58 ± 0.07 m w.e. a−1 between 1978 and 2019 and −0.78 ± 0.25 m w.e. a−1 between 2019 and 2024. Breifonn formed in a region with high winter precipitation, but the current warming is taking its toll on Breifonn. The current small altitudinal range of less than 200 m will make it very difficult for the ice remnants of Breifonn to survive much longer and may soon be reported to the GLIMS list of extinct glaciers (Raup and others, Reference Raup, Andreassen, Boyer, Howe, Pelto and Rabatel2025).
Supplementary material
The supplementary material is available at https://doi.org/10.1017/aog.2025.10029.
Data availability
Geodetic mass balance results will be submitted to WGMS. The orthophotos from 1978 and 2019 are available at Norgeibilder.no. The original glacier inventory outlines from N50, 1988, 2003 and 2019 are available in the GLIMS database. The new glacier outlines from LIA, 1955, 1978 and 2024 are available from Nasjonalt Vitenarkiv (https://doi.org/10.58059/wazv-zv52).
The processed DTMs and elevation change rasters are available upon request. Scanned versions of the NVE glacier inventories are available at: https://www.nve.no/vann-og-vassdrag/vannets-kretsloep/bre/publikasjoner-publications/breatlas-glacier-inventories/. The historical photograph of Breifonn (Fig. 5a) is available from Nasjonalbiblioteket (https://urn.nb.no/URN:NBN:no-nb_digifoto_20170324_00052_NB_MIT_KNR_04218)
Acknowledgements
We thank two anonymous reviewers and scientific editor Etienne Berthier for valuable comments that helped us improve our manuscript. This work is a contribution to the NVE projects ‘Morfologisk basert breutbredelse LIA’ and ‘NVE Copernicustjenester’, and to the ‘Klima i Norge 2100’ report. Publications costs are covered by Norsk klimaservicesenter. We thank Isaac Dawson and Jenna Sutherland for assistance on the field visit to Breifonn. We are grateful to Sidsel Kvarteig (Kartverket – Norwegian Mapping Authority) for sharing her expertise on the historical maps of Breifonn and for providing access to historical survey reports.
Author contribution
MS and BR processed aerial photographs and digital terrain models. LMA and BK digitised glacier outlines. PW reconstructed the LIA outline. LMA and BR analysed DTM data. LMA made figures and tables. LMA wrote the manuscript with contributions from all authors. All authors commented on and contributed to the final manuscript.
Open access
