Abstract
The maximum ‘Little Ice Age’ (LIA) glacier extent provides a significant baseline to assess long-term glacier change and to place currently observed rates of glacier recession in a broader temporal context. To that end, we examine the evolution of the plateau icefield Hardangerjøkulen since the LIA. First, we reconstruct Hardangerjøkulen’s maximum LIA extent (~AD 1750) and subsequent recession based on the glacial landform record and aided by historical map interpretation. Ice-marginal moraines, glacial drift limits, trimlines, and identifiable erosion and weathering boundaries provide evidence of a LIA icefield with an area of 110 km2. Existing LIA model simulations of Hardangerjøkulen are not yet fully able to reproduce our reconstructed extent. Second, we compile a set of remotely sensed icefield outlines from successive time points in the 20th and 21st century to calculate icefield area and length change since the LIA. This reveals a substantial reduction in icefield size, with a total area loss of 41 km2 (37%; 2% 10 a–1) by 2010 and a cumulative frontal retreat averaging 1.3 km (29%; 5 m a–1) by 2013. Icefield recession has been greatest since the end of the 20th century, when rates of areal shrinkage increased to 6.5–10% 10 a–1 in 1995–2010, and the rate of average terminus retreat accelerated to 17 m a–1 in 2003–2010. Third, we present a relative dating approach, based on the known age of the different icefield outlines, that allows bracketing ages to be assigned to all ice-marginal landforms between any two outlines. This approach shows that episodes of moraine formation vary temporally between individual outlet glaciers of Hardangerjøkulen, suggesting that the moraine record of a single outlet glacier alone may not be sufficient to derive an icefield-wide picture of past ice advances, and thereby climate fluctuations.
Keywords
Introduction
Since the end of the 20th century, worldwide recession of glaciers and ice caps because of climate warming is occurring at the highest recorded rates and contributing significantly to sea-level rise (Intergovernmental Panel on Climate Change (IPCC), 2014; Vaughan et al., 2013; Zemp et al., 2015, 2019). Global records of glacier length fluctuations and mass balance extend back to the end of the 19th century and late 1940s, respectively, and measurements of glacier area changes have become widespread since the 1980s following the availability of satellite remote sensing (Zemp et al., 2014, 2015). These observational records can be reconstructed back into pre-industrial times using historical documents, old maps, paintings, aerial and terrestrial photographs, and glacial landform evidence (e.g. Leclercq et al., 2014; Nussbaumer et al., 2011). Long-term data series are important in order to allow the magnitude of contemporary glacier (and thereby climate) change to be placed into a centennial-scale context.
Norway has one of the longest length change (since AD 1899) and mass balance (since 1949) records in the world (Norges vassdrags- og energidirektorat (NVE) – Norwegian Water Resources and Energy Directorate; Andreassen et al., 2005), and complete inventories of the spatial extent of Norwegian glaciers have been compiled for several time points since ~1950 (Andreassen et al., 2012b; Winsvold et al., 2014). Reconstructions of glacier fluctuations go back to the ‘Little Ice Age’ (LIA), when Norwegian glaciers experienced their last major expansion (Grove, 2004). The timing of the LIA maximum differs widely across Norway, ranging from the early 18th century to the early 20th century (e.g. Grove, 2004; Matthews, 2005; Winkler, 2003; Wittmeier et al., 2015), and also varies across outlet glaciers of the same ice mass (Bakke et al., 2005a, 2005b; Bickerton and Matthews, 1993; Tvede, 1973). Nevertheless, the maximum LIA glacier extent provides an ideal baseline for assessing long-term glacier change. For this purpose, glacier area is a crucial variable because unlike glacier length it is not based on localised point data with limited spatial coverage (cf. Zemp et al., 2014). To date, however, LIA glacier area has only been quantified for very few regions in Norway (e.g. Baumann et al., 2009; Stokes et al., 2018), resulting in a lack of digitally available glacier outlines for long-term change analyses. Reconstructions of glacier fluctuations since the LIA are typically biased towards length changes of individual mountain and outlet glaciers (Grove, 2004; Nesje et al., 2008; Nussbaumer et al., 2011), often only indirectly expressed through the mapping of moraine patterns (e.g. Erikstad and Sollid, 1986).
Monitoring long-term glacier change through multi-temporal inventories of glacier area is of high importance for plateau icefields, which make up a significant portion of Norwegian glaciers. These ice masses are particularly sensitive to climate variations because of their top-heavy hypsometry. A rise of the equilibrium line altitude (ELA) to the plateau, where the bulk of these low-gradient ice masses is situated, can lead to a substantial expansion of the ablation area, triggering rapid outlet glacier and icefield recession (cf. Oerlemans, 2012). This sensitivity has been demonstrated by case studies and model simulations at plateau icefields in Norway (e.g. Åkesson et al., 2017; Andreassen et al., 2012a; Giesen and Oerlemans, 2010; Nesje et al., 2008; Oerlemans, 1997) and elsewhere in the Northern Hemisphere (e.g. Jiskoot et al., 2009; Zekollari et al., 2017). In particular, Åkesson et al. (2017) simulated the late-Holocene evolution of the southern Norwegian Hardangerjøkulen icefield (Figure 1) and found that a rise in ELA of 100 m will result in a ~17% area reduction of the present-day icefield, compared with ~10% at Nigardsbreen (southern Norway), ~6% at the Vatnajökull ice cap (Iceland) and ~1.5% at Franz Josef Glacier (New Zealand). Giesen and Oerlemans (2010) modelled the future evolution of Hardangerjøkulen and projected that, under a temperature increase of 3°C between 1961–1990 and 2071–2100, the icefield will have almost entirely disappeared by the year 2100. Predicting the future of these ice masses requires detailed knowledge of their past and present behaviour.
Topographic map of the Hardangerjøkulen icefield and its outlet glaciers (1:130,000; Coordinate System: ETRS 1989 UTM Zone 32 N; Projection: Transverse Mercator; Map data from Kartverket – Norwegian Mapping Authority). RB: Ramnabergbreen; BB: Bukkaskinnsbreen (informal name); MB: Midtdalsbreen; BI: Blåisen; TF: Torsteinsfonna; AL: Austra Leirebottsskåka; VL: Vestra Leirebottsskåka; IS: Isdøleskåka; RS: Rembesdalskåka. The inset shows the location of Hardangerjøkulen (H) along with other ice masses in southwestern Norway. F: Folgefonna; JB: Jostedalsbreen; JH: Jotunheimen glaciers (Glacier inventory data from Andreassen et al., 2012b).
Against this backdrop, we use geomorphological mapping and historical maps to examine rates of area and length change at Hardangerjøkulen since the LIA maximum, in order to further understand the icefield’s post-LIA evolution. The specific aims are (1) to reconstruct Hardangerjøkulen’s maximum LIA extent and subsequent recession using the glacial landform record, with additional guidance provided by historical maps; (2) to quantify icefield area and length change since the LIA maximum; and (3) to establish a relative age chronology for the glacial landforms and recession patterns.
Study area
Hardangerjøkulen is the sixth largest ice mass in Norway (Andreassen et al., 2012b), covering an area of ~69.2 km2 and ranging in altitude from 1856 m a.s.l. on the icefield summit to 1066 m a.s.l. at the terminus of the Rembesdalskåka outlet glacier (Figure 1). Other key outlets are Ramnabergbreen in the north of the icefield; Bukkaskinnsbreen (informal name), Midtdalsbreen and Blåisen in the northeast; Torsteinsfonna in the east and Austra and Vestra Leirebottsskåka in the south (Figure 1). The icefield lies in the southern Scandinavian Mountains on the boundary between the maritime west coast climate and the more continental climate of southeastern Norway. After ice-free conditions in the mid-Holocene, glacier activity on the plateau restarted by ~4.8–3.4 ka BP and reached its peak during the LIA (Dahl and Nesje, 1994, 1996; Nesje et al., 1994). The maximum LIA position of Midtdalsbreen and Blåisen has been mapped and lichenometrically dated to ~AD 1750 by Andersen and Sollid (1971). Nesje and Dahl (1991) calculated a LIA ELA of 1560 +40/−45 m a.s.l. for Midtdalsbreen, which is approximately 130 m below the modern-day ELA of Rembesdalskåka at 1689 m a.s.l. (mean ELA value for the period 1963–2017 calculated from direct mass balance measurements at Rembesdalskåka using data published in Kjøllmoen et al., 2017). Following the LIA maximum, the glacial landform record of Midtdalsbreen and Blåisen (Andersen and Sollid, 1971) indicates that Hardangerjøkulen, in parallel with other glaciers in southern Norway, entered a phase of slow net retreat, which continued into the 1930s–1940s (Nesje et al., 2008). Frontal position measurements at Rembesdalskåka since 1917 show that icefield recession intensified after 1940 (Andreassen et al., 2005). However, in the 1990s, the icefield outlets, along with many maritime glaciers in western Norway, readvanced in response to a period of increased winter precipitation (Andreassen et al., 2005; Nesje et al., 2008). Since the culmination of this advance around 1995, Hardangerjøkulen’s outlet glaciers have undergone rapid 21st-century retreat (Andreassen et al., 2005).
Methods and data
Mapping of glacial landforms and surficial deposits was carried out remotely in ArcGIS from digital colour aerial photographs and verified during extensive field campaigns, following methods outlined in Chandler et al. (2018). The aerial photographs have a spatial resolution of 0.25 m and were captured on 20–22 July 2013 (acquired from http://norgeibilder.no/). All references in this paper to the ‘current’ glacier margin or the ‘present day’, therefore, refer to the icefield dimensions in July 2013. Field mapping took place at all key outlet glaciers and along the southwestern plateau flank in July and August of 2016 and 2017 using a handheld GPS device with a maximum accuracy of 3–4 m. The landform features mapped around the icefield include ice-marginal moraines, glacial drift limits, trimlines and identifiable erosion and weathering boundaries (Table 1). Typically, the surficial deposits and landforms relating to the LIA and later are comparatively fresh-looking, characterised by limited vegetation and lichen cover and unweathered surfaces. In addition, visual analysis of historical maps from the mid-19th to the early 20th century was employed to help identify Hardangerjøkulen’s LIA dimensions. Once the maximum LIA extent of Hardangerjøkulen was reconstructed, remotely sensed icefield outlines from different time points in the 20th and 21st century were used to estimate glacier change and to date glacial landforms. The former icefield outlines are based on published glacier inventories (Andreassen et al., 2012b; Winsvold et al., 2014) or were extracted from topographic maps and vertical aerial photographs (Table 2). The methods for these assessments, including the separation of the outlines into glacier units and the generation of individual centrelines, are covered in detail in the respective sections of the paper.
Glacial landforms mapped at the margins of Hardangerjøkulen.
LIA: ‘Little Ice Age’.
Overview of remotely sensed icefield outlines of Hardangerjøkulen.
NVE: Norges vassdrags- og energidirektorat.
The icefield outlines were used to assess glacier change since the LIA and assign relative ages to the ice-marginal moraines formed at the margins of Hardangerjøkulen.
Mid-19th to early-20th-century dimensions of Hardangerjøkulen based on historical maps
Historical maps can provide valuable information on former glacier extents (e.g. Cullen et al., 2013; Tennant et al., 2012; Winsvold et al., 2014). Hardangerjøkulen first appears in large scale (1:100,000) on hand-drawn mid-19th-century rektangelmålinger (‘rectangle survey maps’) and porteføljekart (‘portfolio maps’) (Figure 2), published by Norges Geografiske Oppmåling (Norwegian Geographical Survey; now: Kartverket) (Harsson and Aanrud, 2016). These maps depict the northern half of the icefield as a uniform, quasi-circular ice cap covering the plateau, without distinguishable outlet glaciers (Figure 2a and b). One interesting detail to note, however, is that the northern icefield margin seems to impinge on and partly envelop the Dyrhaugane massif (1584 m a.s.l.) (Figure 2a). By contrast, the southern half of Hardangerjøkulen is shown in more detail with a number of outlet glaciers extending from the icefield, including Austra Leirebottsskåka (Figure 2c). Many of these outlets are in locations that today are free of glacier ice (Figure 2d), indicating substantial glacier retreat since the mid-19th century. Curiously, the cartographers appear to have erroneously placed Vestra Leirebottsskåka on the western instead of the eastern side of the Moldenutane massif (1386 m a.s.l.), and the nameless lake that today fills the valley bottom in front of this outlet (1076 m a.s.l.) does not appear to have existed at the time (Figure 2c).
(a) Northwestern quadrant of Hardangerjøkulen between Rembesdalskåka in the west and Finsevatnet in the northeast, as shown on rektangelmålingen 23B 3, 4, 7, 8 (1:100,000; mapped by F. Lowzow; 1864; published by Norges Geografiske Oppmåling; available from Kartverket). (b) Northeastern quadrant of Hardangerjøkulen between Finsevatnet in the north and Torsteinsfonna in the east, as depicted on rektangelmålingen 24A 1 (1:100,000; mapped by L. Broch; 1848; published by Norges Geografiske Oppmåling; available from Kartverket). (c) Southern half of Hardangerjøkulen between Matskornipen (1639 m a.s.l.) in the southeast and Skytjedalsfjellet (1424 m a.s.l.) in the southwest, as shown on porteføljen no. 28 (1:100,000; produced by E. Lund and F. Sejersted; 1860; published by Norges Geografiske Oppmåling; available from Kartverket). (d) Outlet glaciers indicated in (c) and (e) are related to a modern-day topographic map of the same area (1:70,000; map data from Kartverket); solid lines: certain agreement; dotted lines: inferred. (e) The glacier-filled Isdøleskåka cirque, as depicted on gradteigskartet D33 Vest Hardangerjökulen (1:100,000; unknown cartographer; 1932; published by Norges Geografiske Oppmåling; available from Kartverket).
Between 1923 and 1929, the Hardangerjøkulen area was re-surveyed for the 1:100,000 scale gradteigskartene (‘quadrangle maps’) (Harsson and Aanrud, 2016). As part of the field mapping campaign around Hardangerjøkulen, the surveyors produced detailed written descriptions and photographs of the landscape, against which the mapping can be checked and verified. These records demonstrate that the icefield and its extent was mapped with high accuracy. The Hardangerjøkulen gradteigskartet shows that the Isdøleskåka cirque at the southwestern flank of the plateau was still filled by a sizable outlet glacier in the 1920s (Figure 2d and e). The icefield outline depicted on this map was used for the glacier change assessment and relative landform dating presented in the ‘Glacier change assessment’ and ‘Relative chronology of moraine formation’ sections, respectively (Table 2).
Geomorphological evidence and identification of LIA limit
Here, we describe the geomorphological record exposed during glacier recession since the LIA for each outlet glacier and the plateau summit, starting with Midtdalsbreen in the northeast and proceeding in a clockwise fashion around the icefield (Figure 1). The results of our mapping are presented in the supplementary materials, available online, as a large-format 1:20,000 scale map of the glacial geomorphology and surficial geology of Hardangerjøkulen. Sections of this map, illustrating the glacial geomorphology of Hardangerjøkulen’s key outlet glaciers, are presented in Figure 3.
Summary geomorphological maps of key outlet glaciers of Hardangerjøkulen (data for base maps, including icefield outline, from Kartverket): (a) Midtdalsbreen and Blåisen (1:32,000), (b) Austra Leirebottsskåka (1:31,000), (c) Vestra Leirebottsskåka (1:34,000), (d) former cirque outlet glaciers along the southwestern plateau flank (1:37,000) and (e) Rembesdalskåka (1:48,000). The inset shows the location of the outlet glacier forelands.
Midtdalsbreen
Midtdalsbreen (Figure 3a) extends down from the plateau to an altitude of 1403 m a.s.l. The glacial geomorphology of the outlet has previously been mapped at a scale of 1:5000 by Sollid and Bjørkenes (1977), with the position of the glacier margin shown in 1975. Their mapping was re-examined and the majority incorporated into our work. The focus of the field mapping undertaken as part of this study was primarily on the immediate glacier foreland. Midtdalsbreen’s maximum LIA extent is delimited by a cover of fresh-looking, sparsely vegetated glacial drift that stretches down to the southern slope of the 1397 m–high foothill between the Finsevatnet lake (1212–1215 m a.s.l.) and the plateau. Along its margin, the drift sheet attains its greatest thickness and is bordered by moraine ridges. The ridges are particularly pronounced in the east, where a belt of bouldery moraines runs along the lateral drift margin. These vary in morphology from well-defined, single-crested and sinuous ridges to hummocks and mounds. Many inactive meltwater channels occur along the ice-proximal side of the moraine belt. Towards the centre of the glacier foreland, scattered moraines have a semicircular arrangement around a bowl-shaped depression containing glaciofluvial deposits, ice-moulded bedrock and extensively fluted glacial drift. The radial arrangement of the flutings reflects divergent ice flow towards the margin during the LIA. The area around the current glacier front exhibits two distinct landform types. First, an assemblage of sandy-gravelly deposits and ice margin parallel ridges lies at the southeastern end of the ice front, where the glacier margin is mantled in debris. The features result from differential melting of the glacier beneath the debris cover, leading to ice-cored ridges of sorted sands and gravels, which eventually become detached from the glacier (‘controlled moraine’; cf. Evans, 2009; see also Reinardy et al., 2019). These were described by Andersen and Sollid (1971) under the term ‘stratified moraine’. By contrast, the area in front of the northwestern ice margin is dominated by densely spaced moraine ridges, between which flutings are abundant. The close spacing of the ridges points to an annual formation rate, which is supported by observations made by Andersen and Sollid (1971). Building on these observations, Reinardy et al. (2013) link annual moraine development to a process operating on a seasonal cycle: in winter, sediments become frozen to the base of Midtdalsbreen and are transported with the advancing ice front. This is followed by sediment melt-out and deposition in spring/summer.
Blåisen
Blåisen (Figure 3a) is located between Nordre (1620 m a.s.l.) and Søre (1740 m a.s.l.) Kongsnuten and flows down the plateau flank to 1424 m a.s.l. An expansive, radially fluted drift sheet with a relatively fresh appearance surrounds the outlet glacier. The drift sheet is separated from that of Midtdalsbreen by a narrow corridor of weathered bedrock terrain. It extends northwards across an undulating area of higher ground to the edge of a prominent escarpment, eastwards to the nameless lake at 1321 m a.s.l., and southwards to the 1407 m–high plateau foothill. We interpret this drift sheet to represent the maximum LIA extent, which accords with the mapping of Andersen and Sollid (1971). The northern part is densely covered with moraine ridges, which generally decrease in size towards the glacier margin. The outermost moraine along the bedrock corridor stands up to 8 m high and transitions from a sinuous, single-crested ridge at the base of Nordre Kongsnuten into a broad, multi-crested moraine belt with numerous intervening depressions and hollows away from the plateau. Subangular boulders up to 6 m wide litter the moraine belt surface. Inside this LIA moraine limit lies a series of smaller recessional moraines (up to 4 m high) that are clearly defined and continuous. These ridges are sinuous and have smooth, flat to rounded crests with a large number of well-embedded boulders. Near Blåisen’s main meltwater stream, they become fragmentary and hummocky. By contrast, the moraine ridges east of the meltwater stream, along the edge of the escarpment and on the central glacier foreland, are often relatively subdued and difficult to distinguish from sediment draped over rolling bedrock humps. Sequences of very small, less than 1 m high moraines, presumed annual in origin, are nested around the northern half of the current glacier margin. Blåisen’s easternmost LIA position is marked by an area of frontal moraines on the far side of the lake situated at 1321 m a.s.l. The moraines have a highly variable morphology and height, occurring as flat morainic bands, distinct ridges with well-defined, round-topped crests, and as areas of ridged hummocks. Boulders of up to 4 m in diameter are strewn over the otherwise finer grained and smoother surfaces of the moraines. To the southeast of Blåisen, arcuate recessional moraines are grouped around a proglacial lake (1372 m a.s.l.) in front of a small, southeast-oriented ice apron that is attached to the northeast-facing present-day glacier front. The ridges are short and fragmented around the LIA limit, but become progressively longer and more continuous towards the modern margin. Overall, the moraines inside Blåisen’s LIA limit, which often occur in compact sequences, delimit successive ice positions retreating towards the plateau. Beyond the LIA limit, Andersen and Sollid (1971) mapped a pre-LIA moraine (based on lichenometry) to the northeast of the drift sheet; however, we interpret this feature to be one of several colluvial boulder deposits along the foot of the escarpment.
Torsteinsfonna and the eastern plateau flank
Torsteinsfonna in the east of Hardangerjøkulen is a wide, apron-shaped outlet presently covering the upper plateau flank. Numerous nunataks fragment the glacier snout. A sweeping, semi-elliptical sheet of fluted glacial drift extends over the flat foreland of the plateau up to the eastern shore of the Brattefonnvatnet lake (1381 m a.s.l.). The boundary of this drift cover is very sharp (Figure 4a) and also delineated by moraine ridges, recording Torsteinsfonna’s maximum LIA extent. To the northeast of Matskornipen (1639 m a.s.l.), the outline of another former LIA outlet is imprinted on the eastern plateau flank in the form of ice-moulded bedrock and a thin blanket of fluted glacial drift.
(a) Glacial drift sheet with distinct LIA drift limit (indicated by arrow) at Torsteinsfonna. (b) Black arrows indicate the (assumed LIA) erosional/weathering boundary on the plateau summit between Isdøleskåka and Store Tresnuten (1677 m a.s.l.). The insets show the location of the photographed areas around the present-day icefield (July 2013 vertical aerial photographs acquired from http://norgeibilder.no/).
Austra Leirebottsskåka
Austra Leirebottsskåka (Figure 3b) is an icefall in the southeast of Hardangerjøkulen. At present, this outlet terminates at 1312 m a.s.l. on a bedrock terrace midway down the steep flank of the plateau. On the eastern side of the terrace, a group of tightly curved moraines is emplaced onto a lobate sheet of fresh-looking, unvegetated glacial drift, outlining a very small former ice lobe that is inferred to have protruded from Austra Leirebottsskåka during the LIA. On the western end of the terrace, a lateral moraine extends down into a cluster of nested latero-frontal moraine ridges, which were deposited on thick glacial drift on the higher ground to the west of the 1248 m-high foothill of the plateau. This landform assemblage defines a localised lobe in the former LIA glacier margin. The moraine cluster links up with arcuate moraine ridges in the valley below. These valley-floor moraines form a series of loops that set off from two points at the foot of the plateau and curve towards the Skåltjørna lake (1136 m a.s.l.). Many of these ridges are discontinuous and fragmented, or occur as disjointed mounds, which are particularly pronounced around Skåltjørna where they form ridged islets in the lake. The valley-floor moraines to the west of Skåltjørna sit on a sheet of thick glacial drift deposits, which forms a sharp boundary to the weathered and vegetated bedrock beyond. We interpret the moraine loops and the glacial drift limit to represent the maximum LIA position of Austra Leirebottsskåka and ice front fluctuations immediately following the LIA maximum (cf. Evans, 2003; Evans and Twigg, 2002). During the LIA, Austra Leirebottsskåka extended from the plateau all the way down to the valley bottom, where it spread out into a piedmont lobe. The glacier foreland between the moraine loops and the present-day ice margin is dominated by exposed, ice-moulded bedrock. Only isolated patches of drift and very few moraine fragments are present here, most prominently below the current ice margin. To the west of Austra Leirebottsskåka, ice-moulded bedrock and glacial drift limits demarcate the extent of another, smaller, LIA icefall, before the LIA drift limit rises and continues westwards along the plateau edge towards Vestra Leirebottsskåka.
Vestra Leirebottsskåka
Vestra Leirebottsskåka (Figure 3c), Hardangerjøkulen’s major southern outlet, flows from the plateau as an icefall onto a relatively gentle, U-shaped bedrock slope, terminating at 1275 m a.s.l. The ice-marginal landform record preserved on this slope is highly asymmetrical, with a complex system of lateral moraines descending the eastern flank of the slope, while the western flank and central part of the slope are largely devoid of ice-marginal landforms. On the lower part of the eastern slope, short frontal moraine segments are also present. In the valley below, the lateral moraines fan out into four major morainic belts, each consisting of numerous fragmented and often mound-like ridges. These latero-frontal belts curve towards the lake that fills the valley bottom in front of Vestra Leirebottsskåka. The southeastern side of the lake is characterised by an ice-moulded and striated bedrock plain, which in places is draped with extensive boulder blankets or thin veneers of drift. Frontal moraine segments and mounds are developed on these surface layers. The ice-polished bedrock, partially covered by sheets of glacial drift, marks the maximum LIA extent of Vestra Leirebottsskåka, while the morainic belts around the lake are evidence for minor readvances or episodes of glacier standstill (cf. Evans, 2003; Evans and Twigg, 2002) in the time immediately following the LIA maximum.
Former cirque outlet glaciers along the southwestern plateau flank
Steep-sided cirques with overdeepened, lake-filled floors are cut into the southwestern flank of the plateau (Figure 3d). The four deepest and most pronounced of these cirques are, from east to west, Juklanutane (informal name), Isdøleskåka (official name), Skytjedalsfjellet–Store Tresnuten and Træet (both informal names). Ice-marginal moraines relating to more than one glacial episode exist here (cf. Liestøl, 1963), of which pre-LIA landforms are typically surrounded by blankets of grey-whitish boulders. Most of the cirques are presently ice-free, but historical maps and old aerial and terrestrial photographs show that they all hosted minor outlet glaciers in the 19th century up until the early 20th century (Figure 2). Between the cirques, the LIA icefield margin follows the edge of the plateau summit, as evidenced by distinct erosional boundaries on bedrock (Figure 4b) and glacial drift limits.
The dominant landform feature of the Juklanutane cirque is an enormous multi-crested lateral moraine ridge (up to 80 m high) on the eastern side of the cirque mouth. Its highest crest peaks at 1418 m a.s.l. A ~200 m-long trimline is clearly visible along the western side of the cirque mouth at elevations of between ~1350 and ~1380 m a.s.l. This is at a lower elevation than the top crest of the lateral moraine on the opposite side, suggesting that the two features are not contemporaneous. The bedrock on the cirque floor below the trimline appears to be ice-moulded, onto which a small pile of relatively sorted, presumed glaciofluvial, material was deposited. Beyond the cirque, hummocks and patches of glacial drift occur on the gently sloping plateau foreland. The area between the hummocks is paved with grey-whitish boulders. Bounding the entire zone is a frontal and a lateral moraine ridge approximately 900 m to the west of the cirque mouth. We interpret the trimline and glacially eroded bedrock within the cirque basin to document the extent of the former LIA outlet. All other glacial landforms likely demarcate a more extensive ice advance predating the LIA (cf. Liestøl, 1963) because the reconstructed glacier would have been out of proportion to the LIA dimensions of Hardangerjøkulen’s other outlets.
At Isdøleskåka, arcuate bedrock ridges around the cirque mouth create a more or less closed, elongated basin. This depression is occupied by three lakes, which are separated from each other by bedrock sills. The rock floor within the cirque basin has a gently undulating topography and is draped with patches of sediment of varying thickness, making it often difficult to judge whether moraine ridges are present here or whether bedrock hillocks are blanketed by sediment. Sequences of curved latero-frontal moraines are developed to the west of the outer lake (1243 m a.s.l.) as well as on the bedrock sill between outer and middle lake. These moraines often consist of only piles and short fragments. More pronounced, bouldery lateral moraines occur to the east of the outer lake as well as to the west of the inner lake (1262 m a.s.l.). A clearly defined, but fragmented, frontal moraine runs along the shoreline of the inner lake. The 1932 gradteigskartet (Figure 2e) and a historical photograph of Isdøleskåka taken in 1928 during the gradteigskartene land surveys show that the ice margin terminated on the bedrock sill between the inner and middle lake. This suggests that the frontal moraine around the inner lake was formed during the retreat of the LIA outlet around, or not long after, this time. Consequently, the outer moraines on the cirque floor may reflect the pre-1928 LIA extent and subsequent recession of Isdøleskåka. The bedrock ridge bounding the cirque mouth in the northwest is topped at 1421 m a.s.l. by a distinct double lateral moraine. The considerable elevation of the moraine indicates the margin of an outlet glacier that must have extended well beyond the confines of the cirque basin (cf. Liestøl, 1963), and is therefore assumed to be of pre-LIA age.
The Skytjedalsfjellet–Store Tresnuten double cirque opens towards a confined upland basin between the Store Tresnuten summit (1677 m a.s.l.) and the Skytjedalsfjellet foothill. A small icefall is currently occupying the headwall of the northwestern cirque. The floor of the basin is filled by two lakes, which are enclosed by a semicircular ridge system of fragmented latero-frontal moraines. The moraine complex has a mature appearance, often with flat-topped crests and massive, up to 10–15 m wide clasts incorporated, and is surrounded by grey-whitish boulder blankets. This is taken as evidence for a pre-LIA age (cf. Liestøl, 1963). A sequence of densely spaced latero-frontal recessional moraines curves around the sides of the northwestern lake (1380 m a.s.l.) and the northwestern tip of the southeastern lake (1383 m a.s.l.). The ridges are only sparsely vegetated, revealing a sedimentary composition that changes from mainly openwork bouldery at the foot of the plateau flank to finer grained and matrix-supported towards the middle of the basin. This landform assemblage delineates the maximum LIA extent of the northwestern cirque outlet. By contrast, evidence for a LIA outlet in the southeastern cirque is sparse. The shore of the southeastern lake below the cirque headwall is flanked by colluvium. A lateral moraine ridge at the edge of the plateau summit above the cirque as well as ice-moulded bedrock and fresh-looking, unvegetated glacial drift in the upper part of the cirque headwall are indicators of a small hanging glacier during the LIA.
In the Træet cirque, a lobate sheet of sparsely vegetated glacial drift comes down the side of the plateau. This drift cover is bounded by a small number of subtle frontal moraine ridges at the plateau base, which mark the maximum LIA extent of this former outlet lobe. In the upper part of the plateau flank, sets of recessional moraines represent stages in the retreat of the LIA cirque outlet. On the flat area of higher ground in front of the cirque, bouldery moraines can be found beyond the LIA drift sheet, ranging from well-defined, round-crested ridges to openwork bands of boulders. These ridges are surrounded by blankets of grey-whitish boulders and interpreted to be of pre-LIA age.
Rembesdalskåka
Rembesdalskåka (Figure 3e) is the icefield’s largest glacier unit (Andreassen et al., 2012b). The outlet glacier dams a northern side valley to form the Nedre Demmevatnet lake (~1240 m a.s.l.), from which frequent jökulhlaups have been recorded since before ~AD 1800 (Kjøllmoen et al., 2017; Liestøl, 1956). Rembesdalskåka’s terminus lies on the plateau flank above a deep upland basin containing the Rembesdalsvatnet lake (Figure 5a). Rembesdalsvatnet was dammed for hydropower generation in the late 1970s (lowest and highest regulated water level: 860 and 905 m a.s.l., respectively), submerging much of the landform evidence relating to the glacier’s maximum LIA position. Low water levels in early summer 2017 allowed a frontal moraine segment and suites of lateral moraines to be mapped on the lake bottom (Figure 5b), which can also be identified on old aerial photographs (Figure 5a). The lateral moraines in the northeastern part of the lake basin can be seen to climb from the lake bed onto the bedrock slope between Rembesdalsvatnet and the present-day glacier terminus. This slope is highly eroded by glacier action, contrasting with the weathered and vegetated terrain around it (Figure 5a and c). We interpret the ice-moulding to be the result of LIA glacier erosion and the lake-bed moraines to represent LIA recessional moraines, indicating that Rembesdalskåka covered the entire lake basin at its LIA maximum. The lateral margins of the ice-moulded bedrock slope are fringed by pronounced trimlines and erosional boundaries (Figure 5c), and groups of moraines (mainly lateral) that are often located on thick deposits of glacial drift. The greatest concentration of moraines occurs on the southeastern side of the upland basin, where an intricate network of discontinuous, meandering lateral moraine ridges was deposited down the slope. The ridges are very closely spaced, often pushed into each other’s flanks to form multi-crested ridge complexes. While the moraines become fresher-looking and more sharp-crested downvalley, their height and sedimentary composition vary considerably both along the length of individual ridges and between ridges. The moraines record successive post-LIA retreat positions of Rembesdalskåka. Around the present-day ice margin, a ~50–200 m wide zone of extremely shiny, freshly ice-moulded and striated bedrock documents the 1990s glacier readvance (Figure 5d). The maximum extent of this readvance is delimited by moraine ridges on the central glacier foreland around Rembesdalskåka’s main meltwater stream as well as on a reverse bed slope to the northwest of the current ice front (Figure 5d). The moraine on the southern side of the stream is a single-crested ridge, primarily made up of an openwork framework of large boulders, whereas those on the northern side of the stream and the reverse bed slope are finer grained, matrix-supported ridges with multiple crests. On their ice-proximal side, these ridges are adjoined by a multitude of tightly spaced moraines, which are often very small, short and sinuous. The compact spacing of both LIA and recent moraines may suggest that they might be annual in origin. Further up-glacier, lateral moraines run parallel to the glacier tongue. The moraines above the northern ice margin are often constructed of huge, angular boulders that have been arranged by the glacier into linear bands along the southern flank of Luranuten (1649 m a.s.l.) (Figure 5d). By contrast, those above the southern ice margin are coarse-grained, bouldery sediment ridges along the steep valley side, each pushed into the proximal flank of the next higher moraine, with only the topmost ridge exhibiting a pronounced distal flank. In the Nedre Demmevatnet basin, glacial drift limits and a handful of moraine ridges, one of which is submerged, define a former ice lobe that extended into the side valley at the LIA maximum (Figure 5e). Another LIA ice lobe of Rembesdalskåka existed on a flat terrace area of the plateau flank above Nedre Demmevatnet, as shown by a lobate sheet of unvegetated glacial drift with numerous moraine ridges.
Glacial geomorphology at Rembesdalskåka. (a) Vertical aerial photograph of Rembesdalsvatnet and Rembesdalskåka from 19 September 1961 (Sortie: WF-1237; Owner: Kartverket) showing the lake before the dam was constructed. Black arrows indicate major moraine ridges that are associated with the maximum LIA advance and are today submerged. Note the bright, ice-moulded bedrock between the lake and the 1961 glacier margin. (b) Westward view across the northeastern part of the lake basin, showing post-LIA recessional moraines exposed on the bottom of Rembesdalsvatnet. (c) Eastward view towards the ice-moulded foreland of Rembesdalskåka, showing the LIA trimline above Rembesdalsvatnet (indicated by white arrow) and the same post-LIA recessional moraines as in (b). (d) Rembesdalskåka in October 2017 (Photo: Hallgeir Elvehøy, NVE). The freshly ice-moulded zone in front of the ice margin marks the 1990s glacier readvance, which is clearly delineated by pronounced moraine ridges on either side of the meltwater stream. Also note the LIA lateral moraines above the lateral glacier margin in the distance to the left. (e) Exposed lake bottom of Nedre Demmevatnet in August 2014 after a jökulhlaup event (Photo: Hallgeir Elvehøy, NVE). Frontal moraines are clearly visible in the foreground of the photo. A lateral moraine ridge can be seen on a bedrock terrace in the distance to the left (indicated by white arrow). Together, the ridges outline a small LIA ice lobe that extended into the side valley. Note the distinctly ice-moulded bedrock of the valley sides.
Pre-LIA moraines are also preserved around Rembesdalskåka (cf. Liestøl, 1963), which can be differentiated from LIA moraines by their very large size and extent. A prominent latero-frontal moraine ridge (up to ~40 m high) curves along the western flank of Luranuten towards the edge of the cliff overlooking Rembesdalsvatnet, indicating a glacier advance of pre-LIA age from the north, not sourced from the plateau. A pre-LIA, more extensive position of Rembesdalskåka is recorded along the northern side of the upland basin by a set of latero-frontal moraines on the slope of the Luraskor mountain gap, and along the southern rim of the upland basin by a system of meandering, lobed moraines that stretches from the foot of the Tresnuten ridge across the summit area of Skoranuten (1147 m a.s.l.) to the head of the Simadal valley. On the opposite side of the valley, the moraines continue further northwestwards for another 9 km or so (cf. Figure 1 in Liestøl, 1963), presumably outlining a past ice sheet margin. Two other systems of pre-LIA moraines are developed in the Skytjedal valley and around the Svolnosvatnet lake (1075 m a.s.l.).
Ramnabergbreen and the northwestern plateau flank
From Rembesdalskåka, the LIA drift limit trends northwards along the western edge of the plateau towards Ramnabergnuten (1729 m a.s.l.). Two small ice bodies on Ramnabergnuten’s northern flank were likely confluent with the icefield during the LIA maximum, as still seen on the 1932 gradteigskartet and evidenced by a continuous cover of fluted drift between the mountain and the plateau summit. The northern LIA maximum limit over Ramnabergnuten is denoted by a densely fluted drift sheet around the northeastern end of the Ramnabergvatnet lake (1371 m a.s.l.). The radial pattern of the flutings on this foreland points to divergent LIA ice flow from a local ice dispersal area centred on Ramnabergnuten. There are also what seem to be curved segments of subaqueous moraine ridges in the lake, which can be seen on the July 2013 aerial imagery. The Ramnabergnuten drift sheet vanishes at the northwestern arm of the large, unnamed ice-marginal lake (1412 m a.s.l.) abutting Ramnabergbreen. Instead, a raised shoreline runs along the northern lake side, indicating a previously higher lake level. The drift cover reappears south of the Dyrhaugane massif, where it is extensively fluted and interspersed with short recessional moraines. Its northern boundary is demarcated by glacial drift limits and a prominent elongated boulder blanket, which places the maximum LIA extent of the outlet glacier halfway between the present-day ice front and the Dyrhaugane ridge. The terrain beyond the drift sheet is weathered bedrock, and there is no geomorphological evidence that Ramnabergbreen advanced further northwards, up the slope and around Dyrhaugane, during the LIA, as depicted on the 1864 rektangelmålingen (Figure 2a). An assemblage of short, linear ridges was laid down on the lower northern flank of the 1585 m–high plateau foothill between two detached ice bodies. The ridges are aligned downhill, perpendicular to the contours of the foothill, and have an ice flow parallel to transverse orientation in alignment with the flutings that occur on the glacier foreland below. They appear (based on aerial photograph interpretation) to be made up of homogeneous fine-grained sediment. Based on these characteristics, the features are interpreted as short eskers (cf. Benn and Evans, 2010).
The north-northeastern plateau flank
Hardangerjøkulen’s north-northeastern sector comprises the midsized outlet glacier Bukkaskinnsbreen between the mountains Bukkaskinnsryggen (1690 m a.s.l.) and Bukkaskinnshjallane (1760 m a.s.l.), and a separate apron glacier on the northeastern mountainside of Bukkaskinnsryggen. On the 1932 gradteigskartet, the apron glacier and the icefield can still be seen as a contiguous ice mass. A relatively thin LIA drift sheet with glacial flutings extends in front of the glacier. Its lower end forms a thick tongue of glacial drift that stretches down the plateau slope towards an outwash plain around the western shore of Finsevatnet. Latero-frontal moraines at the downslope end of the sediment tongue mark the maximum LIA position. There are only a few other moraines on the glacier foreland, with two small groups of subdued moraine ridges present in the distal part of the drift sheet, and another concentration of moraines around the current ice front of Bukkaskinnsbreen. On the eastern side of the immediate glacier foreland, ice margin parallel ridges of sorted glaciofluvial sediment are present, resulting from drainage guided by the ice margin. Bukkaskinnsbreen’s foreland is connected to that of Midtdalsbreen by a densely fluted drift cover at the base of Bukkaskinnshjallane.
Confidence assessment of the presented LIA reconstruction
Hardangerjøkulen’s reconstructed LIA outline has a total area of 109.7 km2. The length of this outline is 82.4 km, of which we categorised individual segments into three classes of confidence (Figure 6): approximately 49.2 km (59.7%) of the outline is classed as certain based on unambiguous geomorphological evidence described in the previous sections. Approximately 26.0 km (31.5%) is interpolated over short distances between segments of unambiguous LIA outline and is thus defined as fairly certain, while 7.3 km (8.8%) of the outline is less certain and only inferred from the topography of the terrain. The LIA outline of most outlet glaciers falls into the first two categories, being clearly delineated almost throughout by ice-marginal moraines and glacial drift limits. Exceptions are the former cirque outlets Isdøleskåka and Juklanutane, where LIA moraines are sparse or entirely absent, respectively. These outlets were reconstructed by connecting the outermost LIA ice margin indicators (i.e. the trimline in the Juklanutane cirque and the outer cirque-floor moraines at Isdøleskåka) with those nearest on the plateau summit above the cirques. Their lateral margins were drawn to rise relatively symmetrically along the contour lines of each cirque.
Reconstructed outline of Hardangerjøkulen at its maximum LIA extent, classified into different levels of confidence and overlayed onto a LandsatLook Natural Colour image from 03 September 2018 (acquired from https://earthexplorer.usgs.gov/).
Along the edges of the plateau summit, the maximum LIA extent of Hardangerjøkulen is often clearly delineated by distinct glacial drift limits and marked erosional/weathering boundaries (Figure 4). Although currently pronounced, these surface features will become obscured with increasing time after deglaciation by vegetation growth and surface processes (e.g. rainwater and snow meltwater run-off, weathering, etc.). This low preservation potential means that former warm-based or polythermal plateau icefields in ancient landscapes could be either not recognised through absence of evidence or misinterpreted as cold-based (cf. Rea and Evans, 2003).
An area where the exact position of the LIA margin is less clear is the plateau summit between the 1654 and 1569 m peaks of the Juklanutane massif. Here, the LIA reconstruction is based on boundaries between areas of bright (inferred ice-moulded) and darker, rougher looking bedrock surfaces. An alternative possibility is that these boundaries could have been created by patches of perennial snow, which can presently be found in this area. However, the perennial snow patches demonstrate that snow and ice can accumulate and persist here, making it equally feasible that this area hosted LIA ice. Also, the relatively gentle, outwards-sloping summit topography of Juklanutane would have feasibly allowed LIA ice to expand to the plateau edge, as consequently mapped by this study. Applying a negative 100-m buffer to the LIA outline between Isdøleskåka and Juklanutane (Figure 6) reduces the outline area by only 1.0 km2 (0.9%).
Hardangerjøkulen’s LIA outline has been reconstructed without nunataks. The present-day nunataks in the centre of the icefield each project less than 6 m above the surrounding ice surface (2010 DEM; NVE; unpublished) and are likely to have been ice-covered during the LIA. No nunataks appear in the icefield’s centre on the 1932 gradteigskartet, where the highest elevation of the ice surface was at 1876 m a.s.l. (2010 DEM: 1856 m a.s.l.; 20 m of vertical thinning over an ~80-year period). Potential candidates for LIA nunataks are the summits of the two highest mountain peaks Bukkaskinnshjallane and Søre Kongsnuten along the northeastern plateau edge (Figure 6), which are currently 55 and 75 m above the surrounding ice surface (2010 DEM), respectively. On the 1932 gradteigskartet, this height difference was less than 21 m at Bukkaskinnshjallane and less than 58 m at Søre Kongsnuten (34 and 17 m of vertical thinning in ~80 years, respectively). Given the proximity of the early-20th-century icefield surface to the two mountaintops, and the flat-topped nature of the summit areas, we speculate that they were ice-covered during the LIA. However, any parts of the summits that did emerge from the LIA icefield would have had a negligible effect on its area (cf. ‘Areal change’ section). A nunatak-free LIA icefield is tentatively supported by signs of ice-moulding and thin glacial drift on these mountain peaks, as visible on the July 2013 aerial imagery.
Comparison with modelled LIA extents
The reconstructed LIA icefield was compared with existing LIA model simulations of Hardangerjøkulen (Åkesson et al., 2017; Giesen, 2009). Both studies employ a two-dimensional, vertically integrated shallow ice approximation (SIA), but Åkesson et al. (2017), building on the work of Giesen (2009), use a more up-to-date ice thickness data set as input.
The Åkesson et al. (2017) model (Figure 7) produced a simulation of the icefield’s LIA geometry at a resolution of between 200 and 500 m, which varied spatially based on modelled ice velocities. This places the positional accuracy of the modelled LIA outline shown in Figure 7 to ±100–250 m, with additional uncertainties relating to how well the model represents parameters such as subglacial topography, glacier dynamics and past surface mass balance. Åkesson et al.’s (2017) LIA model matches the geomorphological record presented here reasonably well in general, but underestimates ice cover in the northwest over Ramnabergnuten (by ~970 m), in the northeast around glacier unit 2959 (~750 m), as well as the extent of the outlet glaciers Blåisen (~370 m) and particularly Rembesdalskåka (~1060 m) and Vestra Leirebottsskåka (~900 m). By contrast, the model corresponds very closely to the mapped extent of the cirque outlets Isdøleskåka and Juklanutane, an area for which our proposed LIA limit is based on limited landform evidence. No LIA nunataks are predicted by the model. Our reconstructed LIA area of 109.7 km2 can be compared with the modelled area of 99.3 km2. An important factor to consider in this comparison is that the maximum LIA extent identified from the geomorphological record is time-independent, while Åkesson et al.’s (2017) model is a snapshot of the icefield at a specific point in time, that is, AD 1750. As the simulation demonstrates, Hardangerjøkulen and its outlet glaciers likely grew in a nonlinear and asynchronous fashion, so it is possible that the AD 1750 snapshot does not capture the maximum LIA extent of some outlets, which may have been reached earlier or later in time. The Giesen (2009) model overestimates the reconstructed LIA icefield extent, particularly in the north of Hardangerjøkulen.
Comparison between geomorphological (this study; shown in classes of confidence) and modelled (Åkesson et al., 2017) LIA reconstruction. The model accuracy of ±100–250 m stated by Åkesson et al. (2017) is illustrated in the form of two buffer zones around the modelled LIA outline.
For the Åkesson et al. (2017) model, only the LIA extent of Midtdalsbreen and Rembesdalskåka was available for model calibration and validation. Our icefield-wide, empirically constrained LIA reconstruction is a robust data set for improving future models of the Hardangerjøkulen icefield evolution. Comparable modelling studies by Aðalgeirsdóttir et al. (2011) of Hoffellsjökull, an outlet glacier of Vatnajökull, and by Zekollari et al. (2014) of the Vadret da Morteratsch glacier (Switzerland) have successfully used detailed LIA reconstructions to accurately match their models to. However, unlike Hardangerjøkulen, these examples are single glacier units, and Åkesson et al.’s (2017) study shows that simulating the evolution of a highly dynamic ice mass with multiple glacier units and outlet glaciers remains a challenge. Future work on dating Hardangerjøkulen’s LIA extent, which has so far only been done for the northeastern part (Andersen and Sollid, 1971), would also be useful to icefield modellers.
Glacier change assessment
In conjunction with former icefield outlines from successive time points since the mid-1920s (Table 2), we used Hardangerjøkulen’s reconstructed LIA geometry as a benchmark to quantify glacier area and length change up to the present day. One of the greatest sources of uncertainty in relation to remotely sensed glacier outlines is the inclusion of seasonal snow or snowfields as part of a glacier area (e.g. Racoviteanu et al., 2009). The possible error introduced by this is estimated by Paul and Andreassen (2009) to be 5–10% for glaciers with an area of >5 km2 and up to 25% for glaciers <1 km2 in size. We regard these estimates as an indicator of the possible error range associated with our area change analysis (i.e. 5–10% for the icefield and its outlet glaciers; up to 25% for small detached ice bodies), although we did not perform separate error calculations.
All icefield outlines were split into individual glacier units using the hydrological drainage divides from Andreassen et al. (2012b) and the same glacier ID numbers (Figure 8a). The drainage basins had to be manually extended in order to accommodate the pre-1973 icefield dimensions (Figure 8b). Also, since glacier units 2958 and 2967 were originally confluent with Ramnabergbreen (2962) and Blåisen (2966), respectively, their drainage basins were merged into their parent units in the early icefield outlines. Once the two glacier units had become detached from their parent outlets, they were treated as separate entities and glacier area change was assessed independently. Where small ice patches other than glacier units 2958 and 2967 became separated from the icefield from one outline to the next, their area was still included in that of their parent units. This may introduce a low degree of uncertainty because the inclusion of these ice patches can vary between outlines, depending on their source and creator. For instance, all previous glacier inventories generally exclude ice bodies smaller than 0.01 km2 (Andreassen et al., 2012b; Winsvold et al., 2014).
Glacier area and length change at Hardangerjøkulen since the LIA maximum (~AD 1750). (a) 2003 icefield outline with individual drainage basins. Glacier ID numbers and glacier aspect data from Andreassen et al. (2012b). (b) LIA icefield outline with extended drainage basins. Glacier ID numbers and glacier aspect information taken from Andreassen et al. (2012b). (c) Icefield recession from the LIA maximum to present using glacier outlines from successive time points. Glacier centrelines used for assessing glacier length change are shown for 13 icefield units. The insets show changes in the glacier front of Rembesdalskåka (RS) and Midtdalsbreen (MB) in greater detail, with triangles marking the locations from where in situ frontal position measurements have been or are currently carried out in the field (data: NVE). (d) Relative glacier area change since the LIA maximum at selected outlet glaciers and at Hardangerjøkulen as a whole. Decadal rates of area change per period, based on compound interest calculation, are shown as open black bars. (e) Cumulative centreline length change since the LIA maximum at selected outlet glaciers of Hardangerjøkulen. (f) Comparison between cumulative centreline and in situ length changes at Rembesdalskåka and Midtdalsbreen (data: NVE).
Absolute and relative glacier area change was calculated stepwise as differences per time interval (At0–At1; At1–At2; and so on), both for the icefield as a whole and for each individual glacier unit (Figure 8c and d; Table 3). Decadal rates of area change were computed in the same fashion using compound interest calculation (Figure 8d) (Andreassen et al., 2008; Zemp et al., 2014), and assuming AD 1750 as the timing of the icefield-wide LIA maximum.
Glacier area (km2) of Hardangerjøkulen and its glacier units since the LIA (~AD 1750).
LIA: ‘Little Ice Age’.
Aspect data from Andreassen et al. (2012b). Note that where ice bodies detach to form separate glacier units (i.e. units 2958 and 2967 in 1961 and 1973, respectively), the parent glaciers (i.e. units 2962 and 2966, respectively) record an abrupt artificial area loss, even though this area is still present within the glacier region as a whole.
Glacier centrelines were used to assess cumulative glacier length change at 13 icefield units (Table 4), excluding four units with extreme changes in glacier geometry through time because of ice-marginal snow (2958, 2959, 2967 and 2970). Average rates of length change were calculated in Table 5. Winsvold et al. (2014) employed a DEM-based least-cost path algorithm to generate centrelines for the 1973, 1988 and 2003 glacier inventories of Hardangerjøkulen. Using these as a basis, we manually re-digitised one centreline per glacier unit that was applicable to all time points (Figure 8c). The centrelines were drawn so that they followed the middle of each glacier unit between the glacier head and terminus in the least costly way, while still coinciding with the most downvalley part of the glacier front in all time points (Figure 8c).
Cumulative glacier length change at Hardangerjøkulen since the LIA (~AD 1750).
LIA: ‘Little Ice Age’.
Corrected for overestimation; recalculated without the 1923–1929 outline.
Excluding uncorrected values for Ramnabergbreen.
Comparisons of mean length changes of the 13 investigated icefield units for each measurement period.
LIA: ‘Little Ice Age’.
Excluding Ramnabergbreen and assuming AD 1750 as the timing of the icefield-wide LIA maximum.
Values are likely too low because of ice-marginal snow at some glacier units.
Using the mean cumulative change (Table 4).
We compared the findings of our change analysis with surface mass balance data from Rembesdalskåka (continuous data series since measurements began in 1963; Kjøllmoen et al., 2017), and in situ length change data based on field measurements of glacier front positions of Rembesdalskåka (since 1917, but with gaps) and Midtdalsbreen (continuous series since 1982), available from NVE.
Areal change
From an original LIA area of 109.7 km2, Hardangerjøkulen had lost 15.3 km2 (14.0%; 0.9% 10 a–1) by the mid-1920s. Historical survey reports and photographs generally confirm the 1923–1929 outline (based on historical mapping) to be accurate (cf. ‘Mid-19th to early-20th-century dimensions of Hardangerjøkulen based on historical maps’ section), but the northern margin at Ramnabergbreen is shown to extend about 150 m beyond our reconstructed LIA limit (Figure 8c). Since such an extended ice front position is not supported by the geomorphological evidence, the 1923–1929 outline is regarded as slightly overestimated in this area, although the error is only minimal (0.5 km2; 0.5% of the total 1923–1929 icefield area). An overestimation of the northern ice margin has already been noted on the 1864 rektangelmålingen (cf. ‘Ramnabergbreen and the northwestern plateau flank’ section and Figure 2). A possible reason for this error might be that the early surveyors and cartographers mapped snowfields attached to the ice margin as part of the glacier. Late-lying and perennial snow is more common on the high upland area in front of Ramnabergbreen than in many other areas of the plateau, as can be observed on multiple series of vertical aerial photographs.
Icefield recession was most substantial between 1923–1929 and 1961 when an area of 16.2 km2 (17.1%; 5.4% 10 a–1) was lost. The strong retreat might have been in response to the observed early-20th-century warming that culminated in the 1930s (cf. Hanssen-Bauer, 2005). In the periods 1961–1973 and 1973–1988, overall icefield recession was relatively minor, with a reduction of 2.8 km2 (3.6%; 3.0% 10 a–1) and 1.1 km2 (1.5%; 1.0% 10 a–1), respectively (note that the 1973 outline of Vestra Leirebottsskåka is based on 1964 aerial photographs; Table 2). This is in accordance with surface mass balance measurements at Rembesdalskåka (Andreassen et al., 2016; Kjøllmoen et al., 2016), which record a slightly negative cumulative mass balance of 1.25 m w.e. in the period 1963–1988 (Kjøllmoen et al., 2017). The prominent 1990s readvance that many glaciers along the west coast of Norway experienced in response to several preceding years of increased winter precipitation (Andreassen et al., 2005; Nesje et al., 2008) is recorded at Hardangerjøkulen in the form of an area gain of 3.9 km2 (5.2%; 7.3% 10 a–1) between 1988 and 1995. This is also reflected in Rembesdalskåka’s cumulative mass balance, which peaked in 1995 at 6.97 m w.e. (Kjøllmoen et al., 2017). An analysis of icefield elevation change for the period 1961–1995 (Kjøllmoen et al., 2001) reveals that a broad zone from west to east across Hardangerjøkulen, comprising the drainage basins of Rembesdalskåka and Blåisen as well as the icefield centre, thickened by up to 20 m. By contrast, pronounced thinning occurred at the extremities of most other icefield sectors, including across the glacier tongues of Midtdalsbreen, Torsteinsfonna and Vestra Leirebottsskåka. Thinning was particularly severe in the south of the Isdøleskåka drainage basin (up to 35 m) and across the northwestern icefield sector, including the drainage basin of Ramnabergbreen (up to 55 m) (Kjøllmoen et al., 2001).
Since the end of the 20th century, the icefield has been in recession, with area losses of 6.1 km2 (7.8%) and 3.2 km2 (4.5%) in the periods 1995–2003 and 2003–2010, respectively. Again, this is mirrored in the cumulative mass balance of Rembesdalskåka, which decreased to –2.94 m w.e. by 2010 (Kjøllmoen et al., 2017). The two periods also exhibit the highest decadal rates of area change with –10.1% 10 a–1 and –6.5% 10 a–1, respectively (Figure 8d), consistent with the global and regional trend of accelerated 21st-century glacier decline (Stokes et al., 2018; Vaughan et al., 2013). An updated ice thickness change analysis for the period 1961–2010 (NVE) shows almost icefield-wide thinning, which had increased to maximum values of 40–60 m and over 60 m across the drainage basins of Isdøleskåka and Ramnabergbreen, respectively. The 1961–1995 zone of ice thickening (Kjøllmoen et al., 2001) had reduced to a few isolated patches, located primarily in the icefield centre.
A slight area increase of 0.5 km2 (0.8%; 2.5% 10 a–1) is indicated for the final 3-year period 2010–2013. However, rather than a true area gain, this is attributed to seasonal snow around the icefield margin being included in the 2013 outline, inflating the glacier extent. In situ front measurements at both Rembesdalskåka and Midtdalsbreen in the 2010–2013 period show a clear retreat. A true increase in area is also contradicted by Rembesdalskåka’s mass balance, with predominantly negative balance years in this period (Kjøllmoen et al., 2016) and an overall negative trend in the cumulative mass balance since 1995 (Andreassen et al., 2016; Kjøllmoen et al., 2017). Consequently, icefield-wide area change since the LIA was assessed using the 2010 outline, revealing a total loss of 40.8 km2 or 37.2% of the icefield area at the LIA, with a decadal rate of recession of 1.8% 10 a–1.
The changes in Hardangerjøkulen’s area over the different time periods are generally mirrored across individual icefield units (Table 3; Figure 8d). Exceptions are the two periods 1961–1973 and 1973–1988, which both show minor icefield-wide area loss, although a combination of area gain and loss can be seen across the individual glacier units (Table 3). Rembesdalskåka (westerly aspect) shows the smallest relative area change over the whole measurement period LIA–2010 (–16.6%), while relative area change at Isdøleskåka (southwesterly aspect) is among the largest of all icefield units (–55.0%) (Figure 8d). Vestra Leirebottsskåka (southerly aspect) and Midtdalsbreen (northeasterly aspect) show moderately high relative area changes of –35.0% and –40.1%, respectively (Figure 8d). This demonstrates that aspect is not a major control on icefield change. Instead, we speculate that relative area change among individual glacier units may be a function of hypsometric factors, specifically the area ratio between outlet- and plateau-based ice. Since changes in the area of a glacier unit can only occur in the outlets, but not on the plateau where the icefield units are joined along shared drainage divides, relative area change is expected to be smaller for glacier units with a large plateau-based area percentage, and vice versa. So Rembesdalskåka has experienced little relative change because of a large plateau area, while the opposite is true for Isdøleskåka. Very small icefield units, particularly the detached ice bodies, are subject to extremely high relative area fluctuations, which can be up to ~290% and are likely caused by seasonal snow included in their outlines.
Our results compare well with other assessments of long-term glacier area change in southern Norway. Baumann et al. (2009) reconstructed the maximum LIA extent of glaciers in Jotunheimen, approximately 125 km to the north-northeast of Hardangerjøkulen (Figure 1), and compared it with 2003 glacier inventory data from Andreassen et al. (2008). They found a total glacier area reduction of 35% from the LIA to 2003, which equals area change at Hardangerjøkulen in the same period (–34.3%; area change calculated for the period LIA–2003). The timing of the LIA maximum in Jotunheimen is thought to have occurred around ~AD 1750, with LIA ages from individual glaciers varying throughout the 18th century (Matthews, 2005). Andreassen et al. (2008) assessed glacier area change in Jotunheimen between 1931–1934 and 2003 and found that glacier size had decreased by 23%. This again matches area change at Hardangerjøkulen (–23.6%; area change calculated for the period from 1923–1929 to 2003) and suggests synchronicity in the areal response of Hardangerjøkulen and glaciers in Jotunheimen to climate warming since the LIA.
Length change
Unlike icefield area, changes in glacier length were calculated up to 2013 despite the ice-marginal snow included in this outline, but which did not affect the lower lying outlet glacier termini. The mean change in cumulative centreline length of the 13 investigated icefield units from the LIA to 2013 is –1.3 km (–28.7%; –4.9 m a–1). Rembesdalskåka, Vestra Leirebottsskåka and Midtdalsbreen decreased in cumulative length by 2.6 km (23.5%), 1.8 km (25.9%) and 1.4 km (23.2%) from the LIA to 2013, respectively (Figure 8e). At the lake-terminating outlet glacier Ramnabergbreen, calving does not seem to have been a factor in accelerating frontal retreat, even though the ice-marginal lake in front of the outlet already appears on the 1932 gradteigskartet. Here, a change in cumulative length of –1.1 km (–22.9%) from the LIA to 2013 (corrected for overestimation) is below the icefield-wide average. Moreover, the post-1961 length changes at this outlet are probably a function of the observed strong thinning of the glacier unit (Kjøllmoen et al., 2001; NVE). By contrast, rapid terminus retreat by calving is inferred for Rembesdalskåka in the period between the LIA and 1923–1929 as well as at Isdøleskåka between 1923–1929 and 1961, when the two outlets receded through what is today Rembesdalsvatnet and the inner lake of the Isdøleskåka cirque, respectively, at three times their interval averages of ~0.5 km each (Table 5). Overall, frontal retreat at Hardangerjøkulen was most rapid at the beginning of the 21st century, with length changes averaging –16.7 m a–1 in the period 2003–2010, reflecting the global trend (Vaughan et al., 2013; Zemp et al., 2015). Strong retreat can also be observed from the 1920s to 1973, when the investigated icefield units receded by 12.7–15.0 m a–1 on average, possibly in response to the early-20th-century warming episode (cf. Hanssen-Bauer, 2005). Glacier length change at Hardangerjøkulen since the LIA is broadly consistent with length change in Jotunheimen, where glaciers retreated by 34% from the LIA to 2003 (Baumann et al., 2009), as compared with 23% at Hardangerjøkulen (centreline length change calculated for the period LIA–2003).
Our findings agree well with in situ length change data (Figure 8f), revealing only small differences between the two methods (cf. Winsvold et al., 2014). For Rembesdalskåka, cumulative in situ length change was calculated for the period 1921–2013 (–1331 m) and compared with a change in cumulative centreline length for the period between 1923–1929 and 2013 (–1244 m), giving a difference of 87 m. For Midtdalsbreen, cumulative in situ length change was calculated for the period 1988–2013 (–118 m), while the calculated change in cumulative centreline length was 90 m in the same period, yielding a difference of 28 m. Both data sets also display the same general trends (Figure 8f). The strong frontal retreat in the first half of the 20th century can be observed in both the centreline and in situ data from Rembesdalskåka, while the 1990s readvance can be seen in both the centreline and in situ data from Rembesdalskåka and Midtdalsbreen.
Relative chronology of moraine formation
Using the known age of the icefield outlines listed in Table 2 as time-markers, a relative chronology of moraine formation (Figure 9) and recession patterns was established. The approach presented here can essentially be employed as a relative dating technique, that is to say, the ice-marginal landforms located between two glacier outlines of known age most likely stem from that interval (taking glacier readvances into account). This allows the relative age of virtually all moraines developed since the LIA to be assessed in a comprehensive, icefield-wide and rapid manner. It is a time-efficient approach not dependent on laborious data collection in the field, as demanded by, for example, the most commonly applied dating method in LIA chronology, lichenometry (cf. Andersen and Sollid, 1971). The age control on the glacier outlines is excellent (cf. Table 2), except for the LIA outline, the age of which should be regarded as an approximation. This produces accurate, albeit only relative, dating results, which can be regarded as less prone to error or methodological weaknesses than lichenometric dating (cf. Osborn et al., 2015). The temporal resolution of the approach will likely increase where a higher frequency of sources for glacier outlines is available. We presume our dating approach to be even more effective in areas where the maximum LIA extent of glaciers occurred at or around the time of accurate topographic mapping campaigns. This might, for example, be the case in Finnmark, northernmost Norway, where the LIA glacier maximum was reached in ~1925 (Wittmeier et al., 2015), not long after the gradteigskartene surveys in this area (~1890s; Winsvold et al., 2014).
(a) Icefield outlines of known age provide age brackets for the ice-marginal moraines at Hardangerjøkulen (1:100,000). Panels (b) to (d) show moraine formation through time in greater detail at (b) Rembesdalskåka (1:30,000); (c) Midtdalsbreen (1:11,500); and (d) Blåisen (1:24,000).
At Hardangerjøkulen, a substantial number of the moraines around the icefield were produced between the LIA maximum (~AD 1750) and 1923–1929. This reflects post-LIA active recession of the outlet glaciers with minor standstills and readvances that created suites of moraine ridges, potentially on an annual basis at Rembesdalskåka and Blåisen (cf. Evans, 2003; Evans and Twigg, 2002). Numerous moraine sequences formed between the LIA maximum and the early 20th century have also been reported from the Jostedalsbreen ice cap (Bickerton and Matthews, 1993), approximately 130 km to the north of Hardangerjøkulen (Figure 1), and from Jotunheimen (Matthews, 2005), indicating very dynamic glacier behaviour across southern Norway in this period. It should be noted, however, that the timing of the LIA maximum at Hardangerjøkulen has only been dated for the northeastern sector of the icefield (Andersen and Sollid, 1971). Studies from other southern Norwegian icefields have shown that LIA ages can vary significantly across outlet glaciers of the same ice mass: for example, outlet glaciers of Folgefonna (Figure 1): ~AD 1750, ~AD 1870–1890 and ~1930 (Bakke et al., 2005a, 2005b; Tvede, 1973), and outlet glaciers of Jostedalsbreen: ~AD 1705–1860 according to lichenometric dating by Bickerton and Matthews (1993).
The limit of the 1923–1929 outline coincides not only with an arcuate line of dispersed moraine segments at Midtdalsbreen but also with moraine ridges at other outlets of Hardangerjøkulen, providing a very narrow and precise age range for these landforms. This is in line with observations made by Fægri (1935) who documented a substantial 1920s advance of Rembesdalskåka, associated with the formation of a prominent frontal moraine. Moraine ridges dating from this time have also been identified at Jostedalsbreen (Bickerton and Matthews, 1993) and in Jotunheimen (Matthews, 2005), pointing towards a regional-scale event.
In the period of strong icefield recession between 1923–1929 and 1961, moraine formation virtually ceased at all outlet glaciers, except at Blåisen and in a few lateral positions at Rembesdalskåka (Figure 9b and d). Particularly at Blåisen, moraine ridges continued to develop in densely spaced, assumed annual, sequences after 1923–1929 and into the present day. We hypothesise that differences in the degree of coupling between the glacier margins and the glaciofluvial system may explain the contrasting behaviour of the outlet glaciers (cf. Benn et al., 2003). Rembesdalskåka, Vestra and Austra Leirebottsskåka were at this point beginning to retreat from the sediment- and lake-filled valley floors in front of them onto the bedrock slopes of the plateau flank. At Midtdalsbreen, the foreland exposed in the period between the two outlines is dominated by ice-moulded bedrock, fluted drift and glaciofluvial deposits, often with distinct terrace elements (Sollid and Bjørkenes, 1977). We suggest that strong icefield recession led to the development of highly competent, well-connected meltwater streams at these outlets which efficiently transported any sediments away from the glacier margins into the proglacial environment, inhibiting moraine production (‘coupled ice margin’; cf. Benn et al., 2003). These sediments were deposited at Midtdalsbreen in the bowl-shaped depression on the central glacier foreland, while they were transferred from the other outlets into the proglacial lake basins and onwards. By contrast, the areas of the Blåisen and Rembesdalskåka forelands where moraines were deposited in this period are reverse bed slopes, and we speculate that the slope topography prevented an efficient proglacial drainage network from being established in these locations. Glacial sediments were not, or only partially, evacuated from the glacier margin, instead remaining around the ice front and becoming available for moraine formation during glacier advances (‘decoupled ice margin’; cf. Benn et al., 2003).
After 1961 (or, more precisely, 1955 according to terrestrial photographs analysed by Andersen and Sollid, 1971), moraine formation resumed at Midtdalsbreen (Figure 9c), which had then begun to retreat across a reverse bed slope. Until present, moraine ridges have been produced here on an annual basis during winter advances (Reinardy et al., 2013), signifying rapid outlet glacier response to seasonal changes. The 1990s readvance created ice-marginal moraines at all key outlet glaciers, although in highly varying quantities. At Rembesdalskåka, the readvance is seen predominantly in the form of the fresh zone of ice-moulding around the glacier margin (Figure 5d), and ice retreat from this advance initiated a brief phase of annual moraine development. Since the beginning of the 21st century, Midtdalsbreen has been the only outlet where ice-marginal moraines are still being formed in large numbers.
Our analysis demonstrates that the timing of moraine-forming events can differ between individual outlet glaciers of Hardangerjøkulen. Furthermore, the findings suggest that moraine production and distribution at the landform scale is not solely a result of climate-driven glacier fluctuations but is also determined by non-climatic, often topographic, factors (cf. Barr and Lovell, 2014; Boston and Lukas, 2019). Specifically, here we argue that foreland topography modulates proglacial drainage and sediment supply, which affects the formation and preservation potential of moraines. This raises an important question as to what extent individual moraine ridges are a reliable proxy for distinct climate signals (cf. Lukas, 2012), as utilised by other workers (e.g. Beedle et al., 2009). Our findings suggest that if moraine sequences are to be used as indicators of past icefield advances (and therefore favourable climatic conditions for icefield growth), the moraine record of a single outlet glacier alone may not be representative.
Conclusion
Geomorphological mapping of glacial landforms and surficial deposits enabled a reconstruction of the maximum LIA extent of Hardangerjøkulen. In general, the icefield reached the edges of the plateau summit, and its outlet glaciers exhibited extensive advances onto the plateau foreland, covering a total area of 109.7 km2 at the LIA maximum. Nearly 60% of the LIA outline was firmly established from ice-marginal moraines, glacial drift limits, trimlines and identifiable and erosional/weathering boundaries. The remainder was reliably interpolated over short distances between the landform-based sections of the outline. A comparison of our LIA reconstruction with a model simulation of the LIA icefield by Åkesson et al. (2017) reveals that the model, taken at a single time-step of AD 1750, underestimates ice extent of Rembesdalskåka, Vestra Leirebottsskåka, Blåisen and in parts of the northern icefield sector by 0.8 km on average. This suggests that there are unresolved challenges in modelling dynamic icefields with multiple glacier units and outlets.
By using remotely sensed outlines of the icefield from several time points in the last ~100 years, we were able to track changes in icefield area and length from the LIA to present. By 2010, Hardangerjøkulen had lost 40.8 km2 (37.2%; 1.8% 10 a–1) of its original LIA area, and by 2013, its icefield units had retreated by a cumulative average of 1.3 km (28.7%; 4.9 m a–1). Twentieth-century icefield change was characterised by rapid recession in the mid-part of the century, when Hardangerjøkulen decreased in area by 5.4% 10 a–1 in ~1930–1960 and its termini retreated by 12.7–15.0 m a–1 in ~1930–1970, while the early 1990s saw a brief re-expansion of the icefield both in area (7.3% 10 a–1) and average length (8.1 m a–1). Since the end of the 20th century, icefield recession has reached its highest rates on record, with areal shrinkage of 6.5–10.1% 10 a–1 in 1995–2010 and icefield-wide terminus retreat of 16.7 m a–1 in 2003–2010. These results are consistent with the global trend of accelerated glacier decline in the 21st century (Vaughan et al., 2013; Zemp et al., 2015).
The known age of the icefield outlines also allowed us to establish the relative age of virtually all ice-marginal landforms developed since the LIA. Most of the icefield’s moraine ridges date from the period between the LIA and 1923–1929, after which moraine formation discontinued (Midtdalsbreen, Vestra and Austra Leirebottsskåka) or was much reduced (Rembesdalskåka) until the early-1990s icefield readvance. Exceptions are Blåisen, where moraines have been continuously produced since the LIA maximum, and Midtdalsbreen, where the mid-1950s marked the beginning of an ongoing and unequalled phase of annual moraine formation. The reason for this discrepancy may be related to the availability of sediment for moraine construction, which we suggest is a function of the capability of the glaciofluvial system to transport sediments away from the outlet glacier margins (cf. Benn et al., 2003). This clear evidence for temporal variations in moraine formation across Hardangerjøkulen’s outlet glaciers has important ramifications for the use of moraine sequences as indicators of past icefield advances and attempts to link moraine spacing to changes in palaeo-climate. We therefore propose that the moraine record of a single outlet alone should not be used to make general inferences about past icefield (and climate) fluctuations.
Supplemental Material
Weber_et_al._2019__Hardangerjokulen_map – Supplemental material for Evolution of the Norwegian plateau icefield Hardangerjøkulen since the ‘Little Ice Age’
Supplemental material, Weber_et_al._2019__Hardangerjokulen_map for Evolution of the Norwegian plateau icefield Hardangerjøkulen since the ‘Little Ice Age’ by Paul Weber, Clare M Boston, Harold Lovell and Liss M Andreassen in The Holocene
Footnotes
Acknowledgements
We would like to thank Lauren Knight, Benedict Reinardy and Danni Pearce for field assistance. The Finse Alpine Research Centre kindly provided accommodation during part of the fieldwork. Thanks are also due to Henning Åkesson for sharing and discussing his modelled LIA outline of Hardangerjøkulen. P.W. gratefully acknowledges funding from the EU’s Erasmus + programme and the University of Portsmouth’s Placement Scheme for Postgraduate Researchers, enabling invaluable research stays at NVE. Two anonymous referees are thanked for their constructive reviews that helped improve the paper. This work is also a contribution to the Norwegian Copernicus Glacier Service project (Contract NIT.06.15.5). Edited by Atle Nesje.
Author contributions
P.W. contributed in the conceptualisation, methodology, investigation, writing – original draft and map preparation. C.M.B., H.L., L.M.A. contributed in the conceptualisation, methodology (L.M.A.), investigation (C.M.B.), writing – review and editing, and supervision.
Data availability
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Fieldwork at Hardangerjøkulen was supported by a number of grants. C.M.B. received INTERACT funding (project acronym PLATREAT) under the European Community’s Seventh Framework Programme. P.W. received financial support from an RGS/IBG Dudley Stamp Memorial Award and a BSG Postgraduate Research Grant.
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
