Introduction

Rock slopes adjacent to glaciers may be particularly prone to paraglacial catastrophic failure due to glacial erosional oversteepening of valley sides, climate-forced debuttressing after glacier thinning and recession, and thawing permafrost in steep mountain slopes (e.g., Evans and Clague 1994; Coe 2020; Evans et al. 2021; Deline et al. 2022). Frequent rockfalls may happen at alpine glaciers in high mountains on an annual basis, whereas the infrequent occurrence of rock avalanches (here defined as flow-like motion of rock fragments from a large rockslide; Hungr et al. 2014) may alter mountain landscapes and pose geohazards for down-valley communities and infrastructure (Deline et al. 2022). The magnitude and frequency of landslides in glacierized regions may, however, be influenced by changes in seismic activity or climate (e.g., Huggel et al. 2012; Hartmeyer et al. 2020).

Most research on landslide-glacier interactions has been biased towards large rock avalanches (> 1 Mm3), while smaller rock avalanches onto glaciers have received less attention. In general, rock avalanches onto glaciers have higher speed and longer runout than rock avalanches of similar magnitude in non-glacial environments (Deline et al. 2022), although contrasts in mobility are observed (Evans and Clague 1999; Deline 2009). Low apparent coefficient of friction between debris and glacier ice (Aaron and McDougall 2019), generation of basal pore pressure by frictional heating (Sosio et al. 2012; He et al. 2023), and entrainment of snow and ice (Bessette-Kirton et al. 2018) may allow long travel distances, even beyond the glacier terminus. Deposition of large amounts of debris on a glacier may protect buried glacier ice from ablation (Østrem 1959; Reznichenko et al. 2010; Laha et al. 2023) and cause acceleration of glacier ice velocity due to the increased debris-ice thickness (e.g., Jamieson et al. 2015; Wyk et al. 2022). The reduced ablation may lead to stagnation of the debris-covered glacier front and formation of supraglacial thermokarst features (Deline and Kirkbride 2009; Irvine-Fynn et al. 2017). A particular geomorphological challenge occurs when rock avalanche debris is deposited as a cross-valley terminal dump moraine ridge, which may be difficult to differentiate from ordinary terminal moraine ridges formed during glacier advances or periods of glacier stagnation (Hewitt 1999, 2009; Schulmeister et al. 2009; Vacco et al. 2010). Thus, the interpretation of deposits from rock avalanches as terminal moraines may be misleading in a paleoclimatic reconstruction context (Hewitt 1999; Yde and Paasche 2010; Reznichenko et al. 2016). Yet, although research on rock avalanches onto glaciers is increasing, there is still limited knowledge on the variability and processes of landslide-glacier interactions.

Rock avalanches onto glaciers are common in certain glacierized regions, while almost absent in others. For instance, in the mountain ranges of Alaska, Yukon Territory, and British Columbia numerous rock avalanches have been triggered by large earthquakes such as the “Good Friday” earthquake (M 9.2) on March 27, 1964 (Tuthill 1966; Bull and Marangunic 1967; Reid 1969; Shugar and Clague 2011); the Denali Fault earthquake (M 7.9) on November 3, 2002 (Shugar and Clague 2011; Schulz 2022); and the rock and ice avalanche on Steele and Hodgson Glaciers (M 5.2) on July 24, 2007 (Lipovsky et al. 2008). However, other large rock avalanches onto glaciers in the same region have been triggered by other drivers such as the Mount Munday rock avalanche at Ice Valley Glacier in July 1997 (Delaney and Evans 2014); the Taan Fiord rock avalanche on October 17, 2015 (Dufresne et al. 2018); and the Lamplugh rock avalanche on June 28, 2016 (Bessette-Kirton et al. 2018; Dufresne et al. 2019). Other regions where rock avalanches onto glaciers are widespread include the Karakoram Himalaya (Hewitt 1999, 2009; Berthier and Brun 2019), Iceland (Sigurðsson and Williams Jr 1991; Ben-Yehoshua et al. 2022), New Zealand (Shulmeister et al. 2009; Cox et al. 2015; Dunning et al. 2015; Cody et al. 2020), the Caucasus (Haeberli et al. 2004; Huggel et al. 2005), and the European Alps (Deline 2005, 2009; Huggel et al. 2012; Deline et al. 2015; Mergili et al. 2020). In contrast, it is striking that rock avalanches onto glaciers appear to be very rare in high-latitude glacierized regions such as Greenland, Svalbard, Norway, the Canadian High Arctic, and the Antarctic Peninsula. The reason for this heterogenous spatial distribution has not been studied in detail.

In western Norway, there exists a record of several large rock slope failures, and rock avalanches and their secondary effects are considered a serious geohazard. The most devastating historical event was the September 1345 Gauldalen event where approximately 500 people were killed when a landslide dammed the river Gaula, forming a 14-km-long lake that finally breached and subsequently flooded the downstream valley (Rokoengen et al. 2001). More recently, the tsunamigenic rock avalanches from the mountain Ramnefjellet into Lake Lovatnet near the village Loen have been the most disastrous. The January 1905 Loen rockfall (0.05 Mm3), which remobilized the downslope scree deposits (0.3 Mm3), caused a 40.5-m-high tsunami wave that killed 61 people. In September 1936, another 73 people were killed by the Loen rock avalanche (1 Mm3) that generated a 74-m-high tsunami wave (Grimstad 2006; Waldmann et al. 2021). Rock avalanche activity has also occurred at the mountain Vora, where 11 successive rock avalanches occurred, adding up to a total volume exceeding 100 Mm3 (Aa et al. 2007). However, no rock avalanches have been released from Vora since 3600 cal. years BP (Aa et al. 2007). Despite that rock slope failures occur frequently in western Norway, the geological record holds limited evidence of rock avalanches onto glacier surfaces (Blikra et al. 2012). The very few examples include a Late Pleistocene rock avalanche (31 Mm3) at the mountain Skarfjellet (Schleier et al. 2015) and a post-Little Ice Age rock avalanche (0.3 Mm3) onto a small unnamed cirque glacier at the mountain Tverrbytthornet (Owen et al. 2010). The latter deposit has been dated by lichenometry to approximately 1900 CE (Owen et al. 2010).

Here, we present an investigation of a 2010 rock avalanche onto the glacier Brenndalsbreen. We confine the timing of the slope failure, identify the detachment area, assess the mobility on the glacier surface, and consider preparatory and triggering factors. We conduct a back-analysis using a Voellmy flow model to confirm the location of the detachment area and volume estimate. Hitherto, there have been no published studies focusing on Brenndalsbreen, so we set the rock avalanche into context with pre- and post-event glacier changes and discuss potential geomorphological implications of the rock avalanche. Finally, we discuss how our approach and findings have implications for studies of rock avalanches onto glaciers in other parts of the world.

Geographical and geological setting

Brenndalsbreen (61°41′N, 6°53′E) is a 20.55 km2 (in 2019; Andreassen 2022; Andreassen et al. 2022) outlet glacier of Jostedalsbreen ice cap, the largest ice cap in mainland Europe (Fig. 1). Lower Brenndalsbreen (glacier atlas ID 2301; Andreassen and Winsvold 2012) is a regenerated part of Brenndalsbreen and covers an area of ~ 0.57 km2 with an elevation range between 530 and 1310 m a.s.l. and an average slope of 22°. A regenerated glacier is a type of glacier that is reconstructed of falling ice from one or more larger glaciers located above the regenerated glacier, and it is a relatively common glacier type in glacierized regions with high-altitude plateaus worldwide. Lower Brenndalsbreen has its own secondary ablation and accumulation areas with a secondary equilibrium line altitude at around 750 m a.s.l. The secondary accumulation area is constituted by a snow and ice avalanche cone that currently receives sufficient deposition to maintain positive accumulation. Due to atmospheric deposition of dust, suspended sediment in meltwater draining from Upper Brenndalsbreen and debris in snow avalanches, the snow surface in the secondary accumulation area darkens during the summer season. The result is that the secondary ablation area has distinct annual stratification bands where the thin dark bands represent the summer surface. These bands should not be confused with annual ogive bands that form beneath icefalls (Goodsell et al. 2002). Lower Brenndalsbreen has a bergschrund crevasse that extends across the glacier in its upper part, marginal crevasses along the sides, and splaying crevasses at the front.

Fig. 1
figure 1

Location map of Lower Brenndalsbreen and Upper Brenndalsbreen in the valley Brenndalen, a tributary valley to Oldedalen. The positions of nearby meteorological stations are shown with a square or, if located outside the map, arrows with UTM (EPSG:32632) coordinates. The location of the Tungøyane farmhouses that were destroyed during the Little Ice Age maximum is shown with a black dot. The fixed marks used for marking the front positions of Brenndalsbreen and Briksdalsbreen in 1900 are shown with yellow dots

Full size image

The glacier front of Upper Brenndalsbreen (19.98 km2; glacier atlas ID 2305; Andreassen and Winsvold 2012) consists of a 950-m-long ice cliff that rests at the edge of the steep valley sides above Lower Brenndalsbreen. Upper Brenndalsbreen has an elevation range between 1170 and 1950 m a.s.l., where the minimum elevation is at the southern valley side. The shortest vertical distance between the upper and lower parts is 80 m at the headwall in the innermost part of the valley. From Upper Brenndalsbreen, ice from dry calving and meltwater falls onto Lower Brenndalsbreen, where it deposits in the accumulation area and enters the subglacial drainage system, respectively.

Lower Brenndalsbreen is situated in the east–west–oriented valley Brenndalen, which is a hanging tributary valley to the larger north–south–oriented valley Oldedalen. As most other large valleys in western Norway, Oldedalen has a Holocene record of rock avalanches and shallow landslides, snow avalanches, and flooding (Nesje and Aa 1989; Nesje 2002; Vasskog et al. 2011). These oversteepened, glacially eroded valleys are characteristic for the fjord and valley landscape of western Norway (Gjessing 1965; Nesje and Whillans 1994). Brenndalen has a length of 4.4 km, where the inner 2.0-km is covered by Lower Brenndalsbreen. The bedrock consists of quartz monzonite belonging to the Western Gneiss Region of Norway (Hacker et al. 2010). The strike orientation of valleys in the area reflects the trend of major fault and fracture zones (Hestnes et al. 2022). Mid-valley, Brenndalen is divided into two distinct areas by a steep rock cliff. The lower area of Brenndalen is characterized by dense vegetation and depositional landforms such as moraine ridges and glaciofluvial outwash plains. In contrast, the upper proglacial area has very sporadic vegetation and is defined as an erosional landscape with landforms such as roche moutonnées and plastic scouring (p-forms) including potholes, glacial striations, and Nye-channels. The presence of these erosional landforms indicates that Lower Brenndalsbreen is a temperate glacier with basal sliding. The valley walls are almost vertical and at places even overhanging, separating the valley floor (minimum elevation 220 m a.s.l.) from the peneplain plateau (1200–1585 m a.s.l). Brenndalsbreen is located just 13 km south of Ramnefjellet and 16 km east of Vora.

The climate is subarctic to tundra according to the Köppen-Geiger climate classification (Beck et al. 2018). The two nearest meteorological stations, Oldedalen (Met station no. SN58480 at 44 m a.s.l.) and Nordre Oldevatn (NVE station no. 88.30.0 at 34 m a.s.l.), are located c. 3 km west-northwest and 13 km north of Brenndalsbreen and measure precipitation and temperature, respectively (Fig. 1). Here, the mean annual precipitation is 1421 mm (1991–2020), and the mean annual air temperature is 7 °C (2009–2022). Another meteorological station, Steinmannen (Statkraft station at 1633 m a.s.l.), is located on a nunatak 13 km east of Brenndalsbreen. At Steinmannen, the mean annual air temperature is − 3 °C (2009–2022), but there are no precipitation measurements from this station. The climate around Brenndalsbreen is expected to get warmer and wetter in the future. Hanssen-Bauer et al. (2017) reported a modeled increase in temperature in western Norway of c. 2 °C (4 °C) for RCP4.5 (RCP8.5) from 1971–2000 to 2071–2100. During the same period, precipitation intensity on days with heavy rain in the region of Sogn og Fjordane, where Brenndalsbreen is located, was estimated to increase by almost 10% (20%) for RCP4.5 (RCP8.5).

Methods and data

Field observations were conducted at Brenndalsbreen on July 11 and September 9, 2021 and on September 20, 2022. The primary focus of the fieldwork was to assess the sedimentological characteristics of rock avalanche deposits and landform formation near the glacier front. The assumed detachment area of the rock avalanche is inaccessible. Interviews with people living in Oldedalen were carried out to obtain local knowledge about the rock avalanche, including gaining access to photographs taken in the years following the event.

Available national orthophotographs from several years and satellite imagery from 2010 (Image © 2010 Planet Labs PBC) were investigated to constrain the timing of the rock avalanche event on Lower Brenndalsbreen and to track the location of the deposits on the glacier surface (Table 1). We used orthophotographs from the years 1966, 2004, and 2010 to construct digital terrain models (DTMs) using Agisoft Metashape software. In addition, orthophotographs from 2017 and an existing LiDAR DTM from 2020 were applied (Table 1). We calculated DTMs of Difference (DoDs) to compare DTMs for the years 1966–2004, 2004–2010, and 2010–2020 to identify and characterize the detachment area. In photogrammetric surveys, objects such as target marks, which are viewed from a restricted range of angles, result in precisions that are reduced along the viewing axis (e.g., Kraus 1993; Luhmann 2000). In the viewing direction, the theoretical precision, \({\sigma }_{z}\), can be estimated by

$${\sigma}_{z}=\frac{{\overline{D}}^{2}}{bd}{\sigma }_{i}$$
(1)

where \(\overline{D}\) is the mean object distance, \({\sigma }_{i}\) is the precision of image measurements (assumed to be a half pixel), b is the distance between the camera centers (i.e., the stereo base), and d is the principal distance of the camera. According to Luhmann (2000), the achievable object precision parallel to the image plane, \({\sigma }_{x}\) (= \({\sigma }_{y}\)), can be estimated by

$${\sigma }_{x}={m}_{b}{\sigma }_{i}$$
(2)

where \({m}_{b}\) is the image scale number (calculated by \(\frac{\overline{D}}{d}\)). With regard to the intersecting geometry of the imaging configuration, this object precision can be weighted by considering a design factor q (Fraser 1996), which may be up to a value of 3 for weak imaging configurations (Luhmann 2000).

Table 1 Summary of airborne data acquisition and theoretical precision estimates. GSD is ground sampling distance
Full size table

We back-calculate the rock avalanche with the Debrisflow module of the numerical simulation model RAMMS (Christen et al. 2010; RAMMS group 2022), which previously has been used for modeling of rock avalanches (e.g., De Pedrini et al. 2022). Whilst we only show a small selection of simulation runs, a total of 42 simulation runs have been performed to test sensitivity to the input data and parameters. For the modeling, we used two DTMs from 2010 and 2020. Resolutions of the calculation grid between 1 and 10 m were tested, and a 5-m resolution was found to be an acceptable compromise between computing time and detail in the depositional pattern. The density was set to 2600 kg m−3, as we assume higher density for shattered rock than for soils. The applied variations of the Voellmy parameters range from 0.15 to 0.3 for the Coulomb-type friction (µ) and from 150 to 1000 m s−2 for the viscous turbulent friction coefficient (ξ). The Debrisflow module in RAMMS offers two alternative ways of defining the starting conditions, by defining a release area (block model) or discharge parameters (input hydrograph model; RAMMS group 2022). We compare results from both approaches. In the block model, values for the release depth range from 8 to 15 m yielding approximate initial volumes between 90,000 and 175,000 m3. For the hydrograph, the flow volume was set to 105,000 m3, maximum discharge to 5000 m3 s−1, and velocity to 30 m s−1.

The 2010 Brenndalsbreen rock avalanche

The timing of the rock avalanche

The exact time of the rock avalanche at Brenndalsbreen is unknown, as there were no known eyewitnesses to the event. The last photograph of a debris-free glacier surface was taken on November 17, 2009 (Fig. 2a). At this time, the valley was snow-free. The first evidence of landslide debris on the glacier surface is a satellite image (Planet RapidEye-5) from June 4, 2010 (Fig. 2b). However, the ice stratigraphy of the regenerated glacier allows a better constraint of the timing. Before the rock avalanche, the surface of Lower Brenndalsbreen was debris-free and the annual stratification bands were visible below the secondary accumulation area (Fig. 3a). As rock avalanche debris was deposited both in the ablation area and accumulation area of the regenerated glacier (Fig. 3b), the debris in the ablation area remained at a supraglacial position while the debris in the accumulation area disappeared into the glacier as englacially transported debris. The result is a well-defined melt-out line across the glacier at the proximal end of the landslide debris, where sediment buried in the accumulation area progressively melts out as the landslide debris is moving down-glacier (Fig. 3c). This transverse supraglacial melt-out line coincides with the position of the transverse dark band that represents a summer surface in the ice stratigraphy. From this, we can deduce that the Brenndalsbreen rock avalanche must have happened either before Brenndalsbreen became permanently snow-covered during the winter 2009–2010 or after the formation of the snow avalanche cone for the winter 2009–2010.

Fig. 2
figure 2

a The last known image of Lower Brenndalsbreen before the rock avalanche (November 17, 2009; photograph by Hallgeir Elvehøy). b The first known image of Lower Brenndalsbreen after the rock avalanche (June 4, 2010; Planet RapidEye-5)

Full size image
Fig. 3
figure 3

Airborne orthophotographs of Lower Brenndalsbreen from (a) August 12, 2004, (b) September 27, 2010, and (c) September 23, 2017. a Before the rock avalanche, the surface of Lower Brenndalsbreen was debris-free, and stratification bands formed annually. b On September 27, 2010, the rock avalanche deposit was superimposed on winter 2009/2010 snow and ice deposits, except for minor post-landslide snow avalanche deposits in the upper secondary accumulation area. c In 2017, the rock avalanche deposit had moved down-glacier, and entrained sediment was melting out along a transverse supraglacial melt-out line that coincided with the summer surface of 2010 in the ice stratigraphy

Full size image

Satellite imagery and photographs from summer 2010 show that the rock avalanche deposit is positioned on top of the snow avalanche cone and visible on the surface of the lower part of the secondary accumulation area (Figs. 3b and 4a). The upper part of the secondary accumulation area is covered by snow and ice avalanche deposits that do not extend to the equilibrium line. Photographs from the subsequent summer 2011 show that the winter 2010–2011 snow avalanche cone covered the rock avalanche deposits above the equilibrium line (Fig. 4b). Thus, the glacier ice layer immediate above the rock avalanche melt-out line represents the accumulation season of 2010–2011 in the ice stratigraphy. Below the equilibrium line, each annual ice band has a length of 50–75 m along the central flowline on the glacier surface, so the possibility that an annual ice band forms each accumulation season is very likely. Therefore, we conclude that the rock avalanche must have happened after the formation of the 2009–2010 snow avalanche cone.

Fig. 4
figure 4

a Photograph from June 24, 2010, showing rock debris exposed high up on the northern side of the snow avalanche cone of Lower Brenndalsbreen (red circle; photograph by Stefan Winkler). b During the summer 2011, snow and ice from the 2010/2011 accumulation season covered the upper part of the rock avalanche deposit (photograph by Jon Olav Kvamme)

Full size image

Snow observations from the meteorological station in Oldedalen (44 m a.s.l.) show that all but 6 days of snow in the 2009–2010 season were measured between December 27, 2009, and March 23, 2010, with the maximum snow depth measured on March 16, 2010. Using temperature observations from the meteorological station Nordre Oldevatn (34 m a.s.l.) and assuming an average lapse rate of 6.1 °C km−1, the area around Lower Brenndalsbreen is estimated to have stayed below the freezing point during the entire period between December 10, 2009, and March 18, 2010. Hence, Brenndalsbreen was likely covered by snow from late-December 2009 to at least late-March 2010.

Based on the information available, we conclude that it is very likely that the rock avalanche happened between mid-March and June 4, 2010. As it is very unlikely that the rock avalanche occurred in 2009, from this point forward, we recommend referring to the event as the 2010 Brenndalsbreen rock avalanche. Later, we will discuss potential triggering factors that may pinpoint the timing of the rock avalanche even further.

Identification of detachment area

An intriguing aspect of the rock avalanche event is the lack of an apparent landslide scar. This contrasts with the easily identifiable light markings on the mountain walls left by minor rock falls further down-valley in Brenndalen. After comparing photographs and satellite imagery before and after the rock avalanche event, we find no visible signs of a detachment area. The most likely reason is that the detachment area is located on an almost vertical mountain wall within the inner narrow gorge in the upper part of Lower Brenndalsbreen. We used a set of DTMs from 1966, 2004, 2010, and 2020 to identify a potential detachment area. The DTMs were subtracted to form three DoDs: a DoD after the rock avalanche (2010–2020; Fig. 5a), a DoD that covered the rock avalanche period (2004–2010; Fig. 5b), and a DoD from before the rock avalanche (1966–2004; Fig. 5c). A general characteristic for these DoDs is that they include significant image artifacts on very steep and shady parts of the slope, and especially along the escarpment. However, a potential detachment area is identified on the inner part of the northern mountain wall right above the accumulation area of Lower Brenndalsbreen (1110–1200 m a.s.l.). This observation is supported by the presence of supraglacial debris along the northern mountain wall on an orthophoto taken 24 June 2010 (Fig. 4a) and by potential fresh scars identified on the mountain wall. However, the area is too steep for using the DoDs to constrain a solid rock volume change estimate.

Fig. 5
figure 5

Three DTMs of Difference (DoDs) from a 2010–2020, b 2004–2010, and c 1966–2004 of Lower Brenndalsbreen. The black arrow indicates the likely detachment area, and the background grayscale image is a 2010 orthophotograph

Full size image

Alternative hypotheses for potential detachment areas include the mountain wall beneath the ice cliff of Upper Brenndalsbreen. This area shows volume gain and loss in the DoDs (Fig. 5), but inspection of photographs and satellite imagery before and after the rock avalanche reveals no visible changes on the rock wall surface, so the elevation differences are likely artifacts in the DTM. Snow avalanches from the less-steep southern mountain wall seem to be common, and they leave avalanche tracks similar in extent to the rock avalanche. However, we do not find any visible changes on the southern mountain side, and the DoDs do not indicate any changes in elevation. Finally, a plausible explanation for the lack of an obvious detachment area could be that the supraglacial debris on Lower Brenndalsbreen derives from a sediment-rich outburst event from Upper Brenndalsbreen triggered by extreme melting, extreme precipitation, or sudden drainage of a glacial lake. No ice-marginal or supraglacial lakes exist on Upper Brenndalsbreen, and it is unlikely to have a large subglacial lake or long-term damming of the subglacial network at a temperate glacier located on a plateau. Extreme melting does not occur during this time of year, and we do not observe any extreme precipitation events in the meteorological records. Also, the large angular boulders found in the supraglacial debris are inconsistent with what we would expect from subglacial drainage, and mosses were observed on some rock debris surfaces (Fig. 6). This leaves us with the most likely detachment area being located on the northern mountain wall in the inner part of the gorge. We test this assumption by back-analyzing the rock avalanche with the numerical simulation model RAMMS.

Fig. 6
figure 6

A large angular boulder on the glacier surface deriving from the 2010 rock avalanche. The red arrow shows an ice axe for scale, while yellow arrows indicate patches of moss (September 20, 2022; photograph by Siri H. Engen)

Full size image

Morphology of the supraglacial debris

The assumed detachment area had its head at approximately 1200 m a.s.l., while the base coincided with the glacier surface at 1110 m a.s.l. As the fall height is relatively small, rock fragmentation due to impact may have been limited compared to the fragmentation that occurred due to collisions, friction, and stress during transport. The travel speed of the debris was likely high when the debris was confined in the inner narrow gorge where the glacier surface is steep (> 22°). As the debris came out of the narrow gorge and spread across the wider glacier area, the travel speed decreased rapidly.

A prominent feature was the separation of the debris into two distinct branches after the debris emerged from the gorge (Fig. 3b). This separation was caused by the convex transverse profile of the glacier surface. The northern branch extended to the lateral glacier margin where debris was dumped off the glacier. There are no signs of debris run-up on the mountain side, so the travel speed must have been low when the debris reached the margin. At this site, a meltwater stream from Upper Brenndalsbreen enters Lower Brenndalsbreen, so landslide debris may have been transported into the subglacial drainage system by the meltwater stream. The western branch was characterized by three main lobes and covered a larger area from midway on the glacier surface to the southern lateral glacier margin (Fig. 3b). Only a small proportion of the debris reached the lateral margin. Longitudinal flowbands on the supraglacial debris surface were visible on the 2010 satellite images, particularly on the southern branch, but flowbands became less pronounced in subsequent years. Flowbands are a characteristic landform for landslides on glacier surfaces, where the basal friction is low (Dufresne et al. 2019). However, the flowbands on Brenndalsbreen appeared to follow the pre-existing glacier surface topography.

The supraglacial debris was poorly sorted and ranged from fine-grained sediment to large angular boulders up to 20 m long (Fig. 6). Many of the larger boulders were deposited close to the lobe margins. For instance, one boulder travelled to the margin of the northern branch, where it was dumped off the ice. Other boulders were aligned parallel with the flow direction at boundaries between lobes on the western branch. The average sediment thickness of the supraglacial deposit was difficult to estimate due to large variations and partly because field observations were conducted more than a decade after the event, when a proportion of the finer material had been washed away and the deposits in the secondary accumulation area had disappeared into the glacier. For estimating a rock volume, we use an average sediment thickness of 0.40 ± 0.20 m, well knowing that this estimate is uncertain and may be lower than the actual average sediment thickness.

Volume and path geometry

The deposit volume is estimated to 0.130 ± 0.065 Mm3. This estimate is based on a footprint area of 326,000 m2 of the visible deposits and secondary accumulation area on a national orthophotograph from September 27, 2010, and the estimated average thickness. Assuming a bulking factor of 20% related to the grain spacing after transport and fragmentation (McSaveney 1978; Delaney and Evans 2014), the solid rock volume was 0.109 ± 0.054 Mm3. This is a rough estimate as the assumed bulking factor is unknown and may be higher if snow and ice were entrained into the rock avalanche debris during transport over the glacier surface.

The total drop height (H) from the head of the scar (1200 m a.s.l.) to the toe of the deposit (670 m a.s.l.) was 530 m, while the runout length (L) of the rock avalanche path was 1190 m. The height-to-length ratio (H/L; also referred to as Heim’s ratio or apparent coefficient of friction) was 0.45, which is relatively high compared to most existing H/L ratios from rock avalanches onto glaciers (Evans and Clague 1999; Geertsema et al. 2006; Deline et al. 2022), but similar to the mean H/L ratio of supraglacial rock avalanches in the Glacier Bay National Park, Alaska (Smith et al. 2023). A relatively high H/L ratio may reflect limited data from small rock avalanches (< 1 Mm3) onto glaciers and that we expect the H/L ratio to decrease with landslide volume (Evans and Clague 1999; Deline et al. 2022), as it does for landslides not affected by runout on glaciers (e.g., Parez and Aharonov 2015). The excessive travel distance (LE = L − H/0.62) was 335 m, indicating that the runout was longer than expected for rockslides in general. This simplified parameter assumes a constant H/L ratio of 0.62 for rockslides, and it does not take lithology, topography, or environmental setting into account (Hsü 1975). The low basal friction on the hardened snow surface in the secondary accumulation area and on the snow and ice surface in the secondary ablation area may not have been the only cause for a long runout. The confinement within the narrow gorge may have helped in channelizing the debris onto the lower part of Brenndalsbreen, and the steepness of the glacier surface of 22° may have compensated for the low vertical fall height (0–90 m) at the detachment.

The fahrböschung, defined as tan−1 (H/L), is a parameterization of the travel angle and another way of expressing the relative mobility of a rock avalanche. The 2010 Brenndalsbreen rock avalanche had a fahrböschung of 24°, which is only slightly higher than the surface slope of the glacier and a high fahrböschung compared to reported fahrböschungs of rock avalanches onto glaciers (e.g., Geertsema et al. 2006; Jiskoot 2011; Shugar 2011; Sosio et al. 2012). As for the H/L ratio, we would expect that the fahrböschung is controlled by volume rather than drop height (H).

Back-analysis of the 2010 rock avalanche

As the depositional pattern suggests a flow behavior of the inferred rock avalanche, we back-analyze the event using the Voellmy flow model in RAMMS::Debrisflow. The modeled flow behavior is highly sensitive to the quality of the available versions of DTMs, as well as changes in the surface topography of glacier-covered areas, which vary on annual and seasonal timescales. The 2020 DTM is the only available LiDAR-based dataset, which provides seamless elevation data for the entire glacier, but it stems from 10 years after the event. During this decade, the glacier surface in the impact area was lowered by 18.2 m on average. This will result in an overestimation of the velocities in the early phase of the flow, and we therefore had to abandon all modeling attempts based on the 2020 DTM.

All presented model runs (Fig. 7) are based on a 5-m DTM derived from image matching of the September 2010 orthophoto campaign, thus obtained only few months after the rock avalanche occurred. For the block model (Fig. 7a and b), the initiation volume is calculated as the product of the inclined area times the release depth. As the detachment area is almost vertical, a release depth of 8 m is sufficient to create a release volume of 105,000 m3. The flow volume for the presented input hydrograph model runs (Fig. 7c and d) is therefore set to the same value. The modeling run with standard Voellmy parameters (µ = 0.2, ξ = 200 m s−2; Fig. 7a) shows reasonably realistic runout lengths for both models. However, the pattern of deposition does not match exactly with the real deposition pattern and is highly sensitive to annual and seasonal changes in glacier surface topography. The patchy character of the deposits seen in Fig. 7 a, b, and d could only be prevented with higher Coulomb-type friction (e.g., block release with µ = 0.35 and ξ = 300 m s−2; Fig. 7c), which gives a too short runout. In order to match the observed runout length, the Coulomb-type friction value had to be lowered to µ = 0.15, which is in line with a relatively smooth glacier surface giving less resistance to motion, in particular when the flow begins to slow down. Variations in the viscous turbulent friction coefficient ξ were found to have only minor effects on runout length and depositional pattern. Small values of ξ are sometimes associated with granular flow whereas larger values are reported for muddy flows (RAMMS group 2022). Independent of the uncertain weather conditions during the rock avalanche, we assume that the impact and basal friction led to readily available meltwater, so that our chosen value for ξ is slightly higher than the standard value (ξ = 300 m s−2; Fig. 7b–d). As we assume that it takes time before the rock slab fully disintegrates, we have introduced an area of high cohesion (yield stress = 2000 Pa) in some model runs. This leads to slightly more deposition early in the rock avalanche path without any significant changes in runoff length or pattern. Overall, the back-analyses agree with the previously determined detachment area.

Fig. 7
figure 7

Selected modeling runs based on a 5-m DTM from 2010. a A block release simulation with standard Voellmy parameters µ = 0.2 and ξ = 200 m s−2. b A block release simulation with Voellmy parameters µ = 0.15 and ξ = 300 m s−2. c An input hydrograph simulation with Voellmy parameters µ = 0.3 and ξ = 300 m s−2. d An input hydrograph simulation with Voellmy parameters µ = 0.15 and ξ = 300 m s−2

Full size image

Brenndalsbreen before the 2010 rock avalanche

The post-Little Ice Age history of Brenndalsbreen has received very limited attention compared to many other outlet glaciers from Jostedalsbreen ice cap. Here, we provide a brief overview of the glacier changes since the Little Ice Age to address the effects of the glacier on the detachment area.

The Little Ice Age maxima of the outlet glaciers from the Jostedalsbreen ice cap culminated around the mid-eighteenth century, when most outlet glaciers reached their maximum Neoglacial extent (Nesje et al. 2008a, b; Nesje 2009; Carrivick et al. 2022; Gjerde et al. 2023). Historical accounts show that the farms in Oldedalen, including the farm Tungøyane located beneath the mouth of Brenndalen (Fig. 1), suffered from severe damage and starvation caused by floods and snow avalanches as the climate deteriorated and the glaciers advanced in the decades prior to the Little Ice Age maximum (Nesje 1994). At its maximum extent, Brenndalsbreen covered the entire hanging valley Brenndalen and its terminus was situated at the mouth of Brenndalen above the farm Tungøyane. On December 12, 1743, an ice avalanche from the glacier front destroyed the farmhouses at Tungøyane, killing all people and animals except for a servant, a boy, and two cows (Nesje 1994).

As climate became warmer after the Little Ice Age maximum, the front of Brenndalsbreen receded into Brenndalen. Figure 8 presents relative changes of the position of the glacier front since 1900 (Andreassen et al. 2005, 2020). The front variation record shows that Brenndalsbreen experienced minor readvances between 1905 and 1911 and again between 1923 and 1930. After 1930, the glacier retreated rapidly until 1950, which is consistent with the regional climate warming in the 1930s in Norway (Hanssen-Bauer 2005). It was likely during the 1940s that the upper and lower parts of Brenndalsbreen became separated. Photographs from the 1900s and 1930s show Brenndalsbreen as coherent, whereas an orthophoto from 1966 shows that Lower Brenndalsbreen had detached from the upper part (Fig. 9). During this period Brenndalsbreen was referred to as Åbrekkebreen. Like the adjacent Briksdalsbreen and other outlet glaciers from Jostedalsbreen ice cap, Brenndalsbreen slowly readvanced from the 1950s until a more rapid advance, known as the Briksdalsbre Event (Nesje and Matthews 2012), peaked in the 1990s (Winkler et al. 2009; Winkler 2021; Seier et al. 2024; Fig. 8). The Briksdalsbre Event is evident from a small terminal moraine, a lateral moraine on the north side of Lower Brenndalsbreen, where deposits from an alluvial fan and a pre-existing lateral moraine have been reworked, and distinct trimlines on the mountain sides. Because the main glacier tongue advanced over the exposed bedrock of the central valley floor, morainic evidence of the Briksdalsbre event is comparatively subdued (Winkler 2021). Since around 2000, the front of Brenndalsbreen has receded rapidly (Figs. 8 and 10).

Fig. 8
figure 8

Cumulative frontal changes of Brenndalsbreen between 1900 and 2022 (black dots). In 1900, the glacier front of Brenndalsbreen was located in the central part of the valley Brenndalen, 53 m from a fixed mark established by Rekstad (1902). Frontal changes of adjacent Briksdalsbreen are shown for comparison (gray dots). Data from NVE (Andreassen et al. 2005, 2020; https://glacier.nve.no/glacier/viewer/ci/en/)

Full size image
Fig. 9
figure 9

Photographs from a September 16, 1900, and b the 1930s show that Brenndalsbreen is coherent (photographs by John B. Rekstad and unknown person, respectively; University of Bergen photography archive MARCUS). c Orthophoto from July 21, 1966, showing Brenndalsbreen as a regenerated outlet glacier

Full size image
Fig. 10
figure 10

Biennial photograph series of Lower Brenndalsbreen from 1995 to 2009 prior to the rock avalanche (photographs by Stefan Winkler)

Full size image

The detachment area of the 2010 rock avalanche was glacier-covered at least until the 1940s (Fig. 9). However, it is questionable whether the detachment area ever was covered by ice again. In 1966, the glacier extent in the inner part of the secondary accumulation area was only slightly larger than the glacier extent in 2010 (Figs. 3b and 9), and in 2004, the glacier extent in the inner part of the secondary accumulation area was almost similar to 2010 despite that the glacier front receded more than 400 m between 2004 and 2010 (Fig. 8). This suggests that the glacier advance and thickening of the secondary ablation area during the Briksdalsbre Event were partly compensated by a lowering of the secondary accumulation area. Hence, most of the detachment area was likely exposed for 45–70 years before the rock avalanche occurred.

Post-rock avalanche debris transport and deposition

In the years following the rock avalanche, the supraglacial debris was transported towards the glacier front and lateral margins (Fig. 11). The northern lobe gradually disappeared as the melt-out line approached the margin, and the rock avalanche debris was dumped off the glacier as a supraglacial melt-out till on top and slightly proximal of deposits from the Briksdalsbre Event. The western branch moved towards the glacier front, where the distal part of the branch reached the front in 2019 and 2020. Any surface features related to landslide dynamics such as flowbands vanished within a few years and were replaced by a hummocky surface where movement and stability of supraglacial debris were controlled by crevasse formation, supraglacial streams, and gravity flow. This surface resembled the typical geomorphology of debris-covered glaciers with glaciomorphological landforms such as ice cliffs and bowls (e.g., Westoby et al. 2020; Kneib et al. 2023).

Fig. 11
figure 11

Annual photograph series of Lower Brenndalsbreen from 2010 to 2023 following the rock avalanche event (photographs by Stefan Winkler)

Full size image

Two scenarios concerning the fate of the rock avalanche debris can be considered: (1) Either a moraine-like ridge of rock avalanche debris will form across the valley, or (2) the debris will be deposited gradually as the glacier front continues to recede. The first scenario may happen either when the distal end of the debris lobe reaches the glacier front and dumps off large boulders along the ice margin or when the debris layer thickens on the glacier front as the proximal end of the debris lobe approaches the glacier front. The latter may cause stagnation and detachment of the glacier front, forming an ice-cored moraine, which may leave a moraine-like ridge. The second scenario is likely to occur when glacier recession and thinning is high, and the thickness of the debris cover is unable to protect the ice beneath from surface melting.

At Brenndalsbreen, we observed boulders deposited along a small ridge at the northwestern corner of the glacier front, but a cross-valley dump moraine ridge had not formed when the debris lobe reached the glacier front (Fig. 12a). Instead, individual large boulders were dumped off the glacier and laid scattered near the ice margin, and finer material was washed off the glacier surface and deposited as the glacier front receded. Based on our field observations, it seems most likely that the rock avalanche debris will be deposited as a supraglacial melt-out till during continuous glacier recession. Small ridges may form when accumulations of boulders are dumped together, but we do not expect development of any cross-valley ridges.

Fig. 12
figure 12

a Boulders dumped off at the front of Lower Brenndalsbreen as the glacier terminus is receding (July 11, 2021; photograph by Siri H. Engen). b Subglacial sediments have been exposed at the glacier front since 2013 and are subsequently overlayered by debris from the 2010 rock avalanche. The large boulder at the central part of the glacier front is from the 2010 rock avalanche (September 9, 2021; photograph by Jacob C. Yde). c Esker-like longitudinal ridge emerging at the glacier front (September 9, 2021; photograph by Marthe Gjerde)

Full size image

A meter-thick deposit of angular coarse-grained debris with a fine-grained matrix started to emerge from beneath the glacier front in 2013 (Fig. 12b). This occurred long before the distal end of the debris lobe reached the glacier front, and thus, the deposition of this landform is unrelated to the 2010 rock avalanche. The presence of a sedimentary deposit is much in contrast to the bare proglacial foreland with scattered rocks lying on the glacially eroded bedrock. The origin of this deposit is interpreted to be rock avalanche debris, which was overridden by the glacier during pre-Little Ice Age glacier growth. The deposit has a carapace of boulders capping a diamicton, which is consistent with the sedimentology of rock avalanche deposits (Reznichenko et al. 2011; Shugar and Clague 2011; Dufresne et al. 2018). It is unclear whether the rock avalanche happened in front of the glacier or onto the glacier surface further up-glacier and transported supraglacially to the site of deposition. This older rock avalanche deposit may serve as an example for what the deposition of the 2010 rock avalanche may look like. However, the debris from the 2010 rock avalanche is deposited on top of the older rock avalanche deposits, making it a composite landform with multiple sedimentary facies.

Another interesting landform has emerged from the subglacial environment at the glacier front since around 2018. It is an esker-like longitudinal ridge (91 m long, 11 m wide, and 3–4 m high in September 2021) formed up-glacier of a large boulder (Fig. 12c). It is a clast-supported diamicton composed of poorly sorted angular boulders with a fine-grained matrix, and the sedimentology looks similar to the rock avalanche deposit except that it lacks a distinct upper carapace layer. It may have formed as an erosional landform derived from the old rock avalanche deposit, where subglacial meltwater streams have removed material from either side of the ridge. However, an alternative explanation of the origin may be that debris from the 2010 rock avalanche entered the subglacial drainage system at the meltwater inlet where the northern lobe dumped off material during the rock avalanche or fell into crevasses, and subsequently got deposited in a subglacial channel leading to the glacier front.

Discussion

Potential preparatory processes

Several processes may have been involved in destabilizing the detachment area prior to the rock slope failure. Glacial debuttressing, defined as the loss of slope support provided by glacier cover (Deline et al. 2022), has been inferred as a major control on rock slope instability after deglaciation (e.g., Bovis 1990; Evans and Clague 1994; Holm et al. 2004; Cossart et al. 2008; Ballantyne et al. 2014), particularly in combination with subglacial water pressure dynamics and paraglacial thermomechanical stresses (Grämiger et al. 2017, 2018, 2020). A causal connection between the glacial debuttressing, which happened between the 1940s and 1966 at Brenndalsbreen, and the paraglacial slope failure that occurred 45–70 years later seems likely. The chain of events may have been as follows. First, thinning of Brenndalsbreen removed the compressive stress on the glacially over-steepened rock wall that may have been glacier-covered since the pre-Little Ice Age expansion of Jostedalsbreen ice cap in the seventeenth and eighteenth centuries (Gjerde et al. 2023; Nesje et al. 2023). Then, the subsequent stress relaxation along the rock slope surface may have led to extension along pre-existing fractures or even the development of new exfoliation joints, which are common geomorphological features on the steep rock slope surfaces in Brenndalen. The role of the “Briksdalsbre” advance during the 1990s is unclear, but it may have caused temporary glacier-cover on the lower part of the detachment area for some years.

Glacial debuttressing may, however, not have been the only preparatory process. Subglacial erosion plays an important role in steepening the almost vertical rock slopes and may have caused undercutting of the detachment area after the area was debuttressed, and thus increasing the instability of the detachment area. Subglacial erosion caused by glacier sliding below Lower Brenndalsbreen is evident by roche moutonnées in the glacier foreland and glacial striation on the rock walls. However, the subglacial erosion rate may be further enhanced along the rock walls in the narrow gorge due to input of glacial meltwater with a high load of suspended sediment from Upper Brenndalsbreen. This effect of subglacial water erosion is also observed in the glacier foreland as p-forms and potholes eroded into bedrock.

Deglaciation has a significant effect on the rock slope temperature regime. As long as the detachment area was covered by temperate glacier ice, the ice-rock interface at the detachment area had an almost constant temperature at the pressure melting point, with some minor fluctuations caused by flow of meltwater along the interface during the ablation season or during winter rainfall events. In other words, the glacier cover prevented the rock slope from being frozen prior to deglaciation. This rock temperature regime changed to a much more dynamic system after deglaciation. Thus, we do not suspect permafrost degradation to be an important process, as it is unlikely that the rock slope was frozen prior to deglaciation. Instead, propagation of a freezing front and sporadic ground ice formation in the exposed rock slope may have occurred after deglaciation. Using the mean annual lapse rate (2009–2022) between Nordre Oldevatn (34 m a.s.l.) and Steinmannen (1633 m a.s.l.) of approximately 6.1 °C km−1 together with the mean annual temperature at Nordre Oldevatn, the estimated mean annual air temperatures near the detachment area head (1200 m a.s.l.) and base (1110 m a.s.l.) are around 0 °C. This is a rough estimate due to the spatiotemporal variability of the lapse rate and local microclimatic conditions. In fact, the actual temperature may be colder as the location is in the shade, on a near-vertical slope with no snow cover, and affected by descending cold air masses from Upper Brenndalsbreen. At least as a first-order approximation, this estimated mean annual air temperature indicates the potential for sporadic formation of permafrost water-to-ice expansion of cracks and microcracks.

After deglaciation, the detachment area was also subjected to thermal stress due to temperature changes and freeze–thaw cycling. As a proxy for the intensity of freeze–thaw cycling, we use the average number of the daily cycles above and below 0 °C at the two meteorological stations, Nordre Oldevatn and Steinmannen. For the 14 years with data from both stations (2009–2022), the average numbers of the daily atmospheric freeze–thaw cycles at Nordre Oldevatn and Steinmannen were 62 and 56 days, respectively. These numbers are consistent with observations at other meteorological stations in coastal western Norway (Kerguillec 2015) and indicate that frost weathering is an active process in crack expansion in combination with other processes, although it is an unlikely trigger mechanism of a rock avalanche of this magnitude.

Potential triggering processes

Although the 2010 Brenndalsbreen rock avalanche may have happened without the occurrence of a specific trigger event, it is worth elucidating potential trigger mechanisms within such environments. Western Norway is a tectonically stable region with rare high-magnitude seismic activity. According to the Norwegian National Seismic Network (https://nnsn.geo.uib.no/nnsn/#/), no earthquakes were recorded near Brenndalsbreen between November 2009 and July 2010; thus, a seismic trigger can be excluded.

Intense and heavy precipitation events are common triggers of shallow landslides in western Norway (e.g., Rüther et al. 2022), particularly in combination with snowmelt (Bondevik and Sorteberg 2021). They may also generate high pore pressure in rock slopes and thereby trigger rock avalanches in glacial environments (e.g., Fischer et al. 2010). Based on the precipitation record at the meteorological station Oldedalen, two heavy precipitation events occurred during the time window between November 17, 2009, and June 4, 2010. The first event happened on November 20–21, 2009 (daily precipitation 23 mm; air temperature 2.5–10.5 °C) before any snow covered the glacier, while the second event occurred on March 18–19, 2010 (daily precipitation 34 mm; air temperature 1.0–2.0 °C) and was followed by an air temperature increase of up to 9 °C starting on the afternoon of March 19, 2010. Based on our analysis of the position of the rock avalanche deposits in the ice stratigraphy, the November 20–21, 2009 event can be excluded as a triggering event whereas the March 18–19, 2010 is a possibility. After several months of sub-zero air temperatures, an air temperature increase above 0 °C followed heavy precipitation over western Norway. Several roads were temporarily blocked by snow avalanches and rockfalls (Bergens Tidende 2010). Assuming that intense precipitation combined with subsequent snowmelt would have caused high porewater pressure in rock fractures and thereby leading to triggering of rock slope failure, the heavy precipitation event on March 18–19, 2010, is the most likely candidate for a meteorological trigger.

Thawing of seasonally frozen ice during spring warming may also cause high porewater pressure within rock fractures, particularly if combined with rainfall and intense freeze–thaw cycling. This process may result in fracture propagation along mechanically weak structures in the rock (Zhang et al. 2020) and lead to failure when exceeding a stability threshold.

Finally, the role of physical impacts by snow and ice avalanches, or even smaller rockfall, on lower elevated parts of the rock wall may be considered as a potential trigger. Ice blocks from dry calving along the glacier front of Upper Brenndalsbreen may either fall directly onto Lower Brenndalsbreen or be mobilized by snow avalanches. It seems likely that repeating direct impacts of ice-rich snow avalanches on an already unstable rock wall will have sufficient force to trigger failure. However, knowledge about the frequency, magnitude, and ice content of snow avalanches at Brenndalsbreen is lacking, and modeling studies are required to test whether the force of an impact is enough to induce a rock avalanche of 0.130 ± 0.065 Mm3.

Rock avalanches onto glaciers in Norway

The 2010 Brenndalsbreen rock avalanche is a rare case of a paraglacial rock avalanche onto a glacier in mainland Norway. This may seem surprising given the total number of 5260 glaciers larger than 0.01 km2 (Andreassen 2022). However, many glaciers and ice caps in Norway are located on plateaus with few or no surrounding rock slopes. Numerous glaciers are small (less than 1 km2) and have rock walls with limited capability of producing rock avalanches. Thus, based on topography the glaciers most prone to rock avalanches are located in alpine landscapes, such as the Jotunheimen mountain range east of Jostedalsbreen ice cap, or confined within incised valleys in western Norway such as Lower Brenndalsbreen.

The main reason for explaining the rarity of rock avalanches onto glaciers in Norway compared to other glacierized regions is, however, likely related to the lack of large earthquakes capable of triggering rock avalanches (larger than M 6.0–6.5 according to Blikra et al. 2002). However, the regional clustering of large rock avalanche deposits in Norway indicates that seismic activity may have been important as a rock avalanche trigger in pre-historic times when the frequency of high magnitude earthquake may have differed from the present conditions (Blikra et al. 2002, 2012) due to that the isostatic rebound after Norway was covered by the Fennoscandian Ice Sheet during the Weichselian glaciation. In general, the stress-releasing effect of glacial debuttressing appears to be a slow process, allowing glaciers beneath unstable slopes to disappear before a major failure occurs (Blikra et al. 2012). For instance, if the failures at Ramnefjellet (Grimstad 2006; Waldmann et al. 2021) and Vora (Aa et al. 2007) near Brenndalsbreen had any relation to glacial debuttressing, the time-delay was thousands of years. The 2010 Brenndalsbreen rock avalanche may be an uncommon event, where pre-existing structures in the bedrock combined with several preparatory processes allowed failure with a time-delay on decadal scale.

Implications for studies on rock avalanches onto glaciers in other parts of the world

This study of the 2010 Brenndalsbreen rock avalanche may serve as an example of how landslide researchers can collaborate with glaciologists and glacial geologists to obtain a comprehensive analysis of a rock avalanche onto a glacier. This includes not only a detailed description and modeling of a specific rock avalanche event, but also adds to an understanding of how rock slope processes interacted with glacier and climate changes prior to the rock avalanche event, and how the rock avalanche debris was transported and deposited at the glacier margin during the years following the rock avalanche event.

Our approach for identifying the detachment area of the 2010 Brenndalsbreen rock avalanche is universal and may be used as a framework in other parts of the world, where there is a lack of direct observations of a rock avalanche onto a glacier and no apparent fresh scars of a detachment area on the rock slopes. In particular, the combination of constructing pre- and post-event DoDs and using numerical simulations to back-analyze the rock avalanche provided valuable support for the presumptions of a likely detachment area based on available photographs and satellite imagery.

This study adds to the existing knowledge on the diversity of rock avalanches in different glacial settings worldwide. Most rock avalanches onto glaciers occur in high-altitude Alpine landscapes such as in Alaska, British Columbia, the European Alps, the Himalayas, and New Zealand, where mountain peaks may be exposed to permafrost degradation (e.g., Bottino et al. 2002; Cox et al. 2015; Dufresne et al. 2019; Friele et al. 2020). The 2010 Brenndalsbreen rock avalanche, however, originated from a low-altitude steep rock slope where fast-flowing ice caused significant erosion as it descended through an icefall from an upper plateau glacier to a lower glacier tongue until the two parts of the glacier became separated 45–70 years before slope failure. This landscape setting with glacierized plateaus is more common in the Arctic, making the observations in this study relevant for analysis of landslides and assessment of landslide risks associated with climate-forced glacier recession in high-latitude regions such as Greenland, Svalbard, Russian High Arctic, Canadian High Arctic, and Antarctica.

The findings of the study of the 2010 Brenndalsbreen rock avalanche also contribute to the evaluation of potential trigger mechanisms for non-seismic rock avalanches in glacierized environments. As opposed to many non-seismically triggered Alpine rock avalanches, permafrost degradation is not a significant preparatory factor for the 2010 Brenndalsbreen rock avalanche because the temperate thermal regime at the glacier base prevented subglacial permafrost formation. Instead, the transition from unfrozen stable thermal conditions during glacier cover to intense freeze–thaw cycling and varying rock surface temperatures after deglaciation may have contributed to failure in combination with other processes such as glacial debuttressing and hydromechanical stress. This thermal transition in rock slopes has been termed “paraglacial thermal shock” by Grämiger et al. (2018) and may be more significant for paraglacial rock damage than previously thought.

Other implications of our study are related to the post-event glacial transport and deposition of rock avalanche debris. For instance, we show how ice stratification can be used to determine the timing of a rock avalanche event and to track a rock avalanche deposit as it is being transported towards the glacier front. Our observations suggest that the rock avalanche debris at Brenndalsbreen will be deposited as a supraglacial melt-out till during continuous glacier recession rather than as a moraine-like ridge. Landslide deposits and moraines are found together in many glaciated valleys worldwide. This co-occurrence has led to a debate on whether specific moraines are the result of a glacier advance triggered by a large rock avalanche or represent a paleoclimatic signal due to a climate-forced mass balance perturbation. Perhaps the most famous example is the Late Glacial Waiho Loop moraine of Franz Josef Glacier, New Zealand (for details and further discussion, see Tovar et al. 2008; Shulmeister et al. 2009; Vacco et al. 2010; Alexander et al. 2014; Winkler 2018). The complexity of interpreting the potential impacts of rock avalanches on moraine formation and of glacial weathering of rock avalanche deposits is also illustrated by the emergence of glacially overridden older rock avalanche deposits at the front of Brenndalsbreen, forming a composite rock avalanche landform with multiple sedimentary facies. The rock avalanche activity at Brenndalsbreen may serve as a modern analogue and improve our understanding of glacial deposition of rock avalanche debris.

Conclusions

The 2010 Brenndalsbreen rock avalanche represents a rare case of a rock avalanche onto a glacier in the modern history of mainland Norway. The deposit volume is estimated to be 0.130 ± 0.065 Mm3, equivalent to a solid rock volume of 0.109 ± 0.054 Mm3. The position of entrained rock avalanche debris within the ice stratigraphy and photographic documentation from the summer 2010 allowed a reduction of the time window of the occurrence of the rock failure to between mid-March and June 4, 2010. The detachment area is not immediately visible on the rock slopes surrounding the regenerated Brenndalsbreen. Using several DTMs, we suggest that the most likely location of the detachment area is on the northern, almost vertical rock slope in the narrow gorge above the accumulation area of Lower Brenndalsbreen. This is supported by back-analyses using a Voellmy flow model to confirm the location of the detachment area and volume estimate.

The drop height (H) was 530 m, while the runout length (L) was 1190 m. The H/L ratio and fahrböschung were 0.45 and 24°, respectively. These values are relatively high compared to most other rock avalanches onto glaciers, but it may reflect a bias in data towards large rock avalanches (> 1 Mm3).

An analysis of glacier changes prior to the rock avalanche indicates that Brenndalsbreen separated into an upper and lower part in the 1940s. The detachment area may have been deglaciated for a period of 45–70 years before the rock avalanche happened. This indicates that although glacial debuttressing is likely to have been a preparatory factor, it is associated with a decadal-scale lag time. Subglacial erosion beneath the detachment area and commencement of freeze–thaw cycling at the rock surface after deglaciation may also have added to destabilizing the rock slope. A potential triggering event may have been a heavy snow and rainfall event on March 18–19, 2010, but it is also plausible that the rock avalanche happened without a specific trigger.

The supraglacial rock avalanche debris was deposited in two debris branches, each characterized by frontal lobes. A distinct distal melt-out line across the glacier developed, representing the equilibrium line on Lower Brenndalsbreen in 2010. The largest boulders were angular and up to 20 m long. The supraglacial debris reached the glacier front in 2019 and 2020, where it was deposited proglacially while Lower Brenndalsbreen continued to recede. The 2010 Brenndalsbreen rock avalanche debris was partly deposited on top of what appears to be an old rock avalanche deposit that has been melting out from beneath the glacier front since 2013.

This study of the 2010 Brenndalsbreen rock avalanche adds to the growing knowledge on rock avalanches onto glaciers and represents a relatively small rock avalanche on a regenerated glacier from a region where rock avalanches onto glaciers rarely occur.