Rock avalanches in northeastern Baffin Island, Canada: understanding low occurrence amid high hazard potential

Abstract

Rock avalanches in fjord environments can cause direct catastrophic damage and trigger secondary submarine landslides and tsunamis. These are well-documented in Greenland, Norway, and Alaska but have gone largely unreported in the extensive fjord terrain of the eastern Canadian Arctic. We provide the first inventory of rock avalanche deposits in northeastern Baffin Island—a region characterized by moderate to high seismic hazard, steep and high-walled fjords and glacial valleys, active deglaciation, and observed climate warming. Over a broad study area of ~60,000 km², one sixth of the terrain had sufficient slope height and gradient to potentially generate rock avalanches. Within that hazard zone, we identified eight rock avalanche deposits at six locations. Only three rock avalanche deposits at two locations are dated, using aerial imagery (1958-present), to the last century while five deposits at four locations are inferred as syn- to post-glacial, likely occurring shortly after local debuttressing. These total numbers fall well below documented inventories from Greenland, Norway, and Alaska. We hypothesize that (1) continuous permafrost persists throughout this region and continues to act as a stabilizing factor and (2) rock mass quality is high in areas of most extreme relief contrast within the study region relative to analogous high-latitude fjord systems such as those in southwestern Greenland. We suggest that Baffin Island is currently in a period of quasi-stability that follows the intense instability during initial deglaciation, yet precedes the higher anticipated slope instability that may occur during permafrost degradation.

Introduction

Large landslides in glaciated environments cause primary catastrophic damage and, when proximal to fjord waters or glacial lakes, can trigger the secondary cascading hazards of subaqueous landslides and displacement waves. Displacement waves triggered by landslides into fjords are among the most damaging as steep and narrow fjord walls effectively channel wave energy, increasing runup heights (Løvholt et al. 2015). In a 2018 review of fourteen “megatsunamis” in the last century (i.e., with runup heights exceeding 50 m), ten were triggered by subaerial landslides into fjords or lakes in glaciated mountainous terrain (Higman et al. 2018). Among these are the historic and deadly Loen (AD 1905 and 1936) and Tafjord (AD 1934) disasters of Western Norway (Hermanns et al. 2014), the 2015 rock avalanche and displacement wave in Taan Fiord, Alaska that generated a 193-m displacement wave (Higman et al. 2018), and the 2017 rock avalanche that triggered a 90-m displacement wave in Karrat Fjord devastating the coastal western Greenland village of Nuugaatsiaq (Gauthier et al. 2018; Svennevig et al. 2020). Most recently, in November 2020 a rock avalanche in Elliot Creek, British Columbia, Canada, triggered a cascade of hazards including tsunami, outburst flood, and sediment plume (Geertsema et al. 2022). Rock avalanches, exemplified in these cases, are the most common landslide type to trigger displacement waves in these fjord and alpine environments given their characteristically large volume, high energy, and mobility (Roberts et al. 2014; Hermanns et al. 2014).

Despite the recognized tsunami hazard (Leonard et al. 2014), the landslide-displacement wave hazard has not been thoroughly assessed on Baffin Island, Canada’s largest island, as few rock avalanches have been reported. However, as with other portions of the Eastern Canadian Arctic Rim, Baffin Island’s extensive northeast coastline is deeply incised by fjords characterized by long (> 1.6 km) steep slopes (Tremblay et al. 2015; Gosse et al. 2020). Numerous high-latitude rock avalanches are observed in the fjords of Norway (Blikra et al. 2006; Hermanns et al. 2014, 2017; Schleier et al. 2015; Böhme et al. 2015), Alaska (McSaveney 1975; Higman et al. 2018; Bessette-Kirton and Coe 2020), and Greenland (Kelly 1979; Svennevig 2019; Svennevig et al. 2020). Combined with active onshore and offshore seismicity, ongoing deglaciation (Gilbert et al. 2017; Noël et al. 2018), and accelerated climate warming (Rapaić et al. 2015; Meredith et al. 2019), the steeply incised region of northeastern Baffin Island would be considered vulnerable to rock avalanche hazards according to established hazard paradigms (Evans and Delaney 2015; Gariano and Guzzetti 2016; Coe 2020). Coastal communities on Baffin Island are generally situated within 60 m above sea level with much of the infrastructure at lower elevations leaving them uniquely susceptible to the threat of rock avalanche-triggered displacement waves and highlighting the need for improved hazard identification and risk assessment. Nevertheless, assessment of rock avalanches and their associated displacement wave hazards has been largely overlooked in this region.

The objective of this study is to provide the first extensive evaluation of the evidence for past rock avalanches within a swath of coastal land that includes both coastal slopes and inter-fjord regions. Evaluating the historical record of rock avalanches on Baffin Island has previously been challenging due to the extensive area, remote and difficult access, and limited high quality remotely sensed data owing to its high latitude and long snow season. However, with the increasing availability of high-resolution optical imagery and digital elevation models (ArcticDEM), mapping of broad swaths of this region has become feasible. We leverage these data to develop a systematic inventory of preserved onshore rock avalanche deposits in northeastern Baffin Island to (1) determine the prevalence and distribution of post-glacial rock avalanches in the region and (2) identify potential spatial correlations with seismicity, bedrock geology, topography, and glacial position that may elucidate their causal mechanisms. These results help address the following questions: Are rock avalanches occurring, but unreported, in this remote coastal region? If not, why are these hazards comparatively less common on Baffin Island relative to other northern hemisphere fjord systems despite common destabilizing factors such as high slopes, over-steepened terrain, ongoing deglaciation, warming permafrost, and active seismicity?

The results of this study have motivated a conceptual time-transgressive deglacial stability model and a hypothesis that Baffin Island has been experiencing a temporary period of quasi-stability that followed a hypothesized period of intense slope failure during initial deglaciation (i.e., rapid erosion, debuttressing, and stress release), yet precedes a future period of potential instability due to warming-induced extensive permafrost degradation and increased precipitation (leading to thermomechanical degradation and increased pore water pressures).

Northeastern Baffin Island

A study area of approximately 60,000 km² along the northeastern coast of Baffin Island was delineated for rock avalanche mapping (Fig. 1) owing to its high coastal relief, ongoing deglaciation, high seismic hazard, and relevance to coastal hamlets of the Qikiqtani region. The area extends from Pond Inlet (Mittimatilik) in the north to approximately 150 km south of Clyde River (Kangiqtugaapik) and encompasses all land surface from the outer coast near the fjord mouths to the fjord heads and their intersection with the interior highland plateau.

Fig. 1

A Study area (outlined in black) along the northeastern coast of Baffin Island with key geographic features. Basemap Imagery: Earthstar Geographics (ArcGIS^® Pro powered by Esri). B Simplified geology of Baffin Island (St-Onge et al. 2009). HB, Home Bay; CP, Cumberland Peninsula; CS, Cumberland Sound; HP, Hall Peninsula; MP, Meta Incognita Peninsula; FP, Foxe Peninsula; FB, Foxe Basin

Physical setting and geology

The study area is situated in the Davis Highlands and Coastal Lowlands physiographic regions of the eastern Canadian Arctic Rim, characterized by deeply incised, steep to near-vertical, exposed bedrock cliffs with heights reaching 1000 m above sea level (asl), local peaks reaching 2000 m asl, and steep-walled fjord basins extending to depths of 400 m below sea level (bsl). Surface material in the area includes a mixture of exposed bedrock, regolith and felsenmeer, variably thick glacial, glaciomarine, and glaciofluvial deposits, and secondary solifluction and colluvial deposits. Extant isolated valley and cirque glaciers, tidewater glaciers, and ice caps persist but are rapidly diminishing in volume (Gardner et al. 2012). Bedrock geology is dominated by felsic to intermediate gneisses, felsic intrusive rocks, and metasediments and metavolcanics (Mary River Group) of the Archean Rae Craton throughout much of the study area (de Kemp and Scott 1998; St-Onge et al. 2022). The metasediments of the Paleoproterozoic Piling Group are exposed in the southernmost region (Henderson 1985; St-Onge et al. 2009), encompassing less than 5% of the total study area. Mafic dykes of the Franklin dyke swarm are exposed on the coastline, at a relatively low occurrence. Paleozoic and Cenozoic sedimentary rocks primarily occur offshore, with none forming the bedrock cliffs within the study area. Exhumation of the plateau region has been relatively fast compared to central and Foxe Basin side of Baffin Island, in part owing to possible Cenozoic rift-flank uplift (Yaehne 2008; Creason and Gosse 2014) and to late Cenozoic isostatic rebound associated with deep fluvial and glacial incision on the Baffin Bay coast.

The region is classified as a moderate to high seismic hazard zone by the Geological Survey of Canada (Basham et al. 1977; Lamontagne and Bent 2021) referred to by some as the Baffin Island seismic zone (Basham et al. 1985; Bent 1996) where the largest recorded seismic event (M_w 6.1) on Baffin Island occurred in 1963. The Baffin Island seismic zone is situated proximal to the offshore Baffin Bay seismic zone where the largest recorded earthquake above the Arctic Circle (M_w 7.3) occurred in Baffin Bay in 1933 (Bent 2002). To date, the zones of high seismicity on Baffin Island and in Baffin Bay have not been associated with rupture on any mapped fault structures (Bent and Voss 2022). Several have recognized glacial isostatic adjustment (GIA) as a possible trigger for earthquakes in this region (Stein et al. 1979; Wu and Hasegawa 1996). However, this correlation remains unresolved (Bent and Voss 2022), as are moment tensors for most events. The historical catalogue of seismicity in this region extends from 1925 to present with variable precision (International Seismic Catalogue 2021). From 1925 to 1957, seismometer coverage in the region was absent and no earthquakes lower than M_w 5.5 were recorded. In 1957, improved seismometer coverage allowed earthquakes > M_w 3.7 to be recorded. It was not until 1973 onward that lower magnitudes of > M_w 2.0 were included in the historical dataset. The full dataset reveals variable clusters of onshore seismicity separated by largely aseismic regions within the study area throughout the last century (Fig. 1) albeit aftershocks have not been distinguished. To our knowledge, peak ground acceleration has not been calculated for the Baffin Island seismic dataset, perhaps owing to low signal quality among the sparse seismometer network.

Northeastern Baffin Island is situated in the continuous permafrost zone (Heginbottom et al. 1995) above the treeline of northern Canada. Regional modeling based on remotely sensed data suggests mean annual ground temperatures range from −14 to −10 °C throughout most of the study area with slightly warmer ground temperatures of −5 to −10 °C modelled in the outermost Coastal Lowlands (Westermann et al. 2015). These values are consistent with a global review of permafrost change over the last several decades (Smith et al. 2022) that shows an average ground temperature of −12 °C circa 1980 gradually warming to −11 °C in 2020 in the Canadian high-latitude Arctic, which is notably colder than other Cold Arctic regions (Smith et al. 2022). Ground ice contents are broadly estimated to be < 10% in the upper 10–20 m of the ground (Heginbottom et al. 1995); however, isolated ice-rich ground conditions have been observed in low elevation environments such as coastal flood plains where surficial materials primarily consist of sediments and exhibit dynamic permafrost landforms such as patterned ground, thermokarst, and distinct fluvial patterns (3vGeomatics Inc. and BGC Engineering Inc., 2011).

The eastern Canadian Arctic has experienced a warming trend of > 0.3 °C per 10 years from 1981 to 2010 with anomalous “hot spots,” including northern Baffin Island, where autumnal warming has locally exceeded 5 °C, largely due to declining sea-ice concentrations (Rapaić et al. 2015). Currently, mean annual air temperatures in northeastern Baffin Island vary from −10 to −15 °C with average monthly precipitation of approximately 20 mm (Syvitski and Normandeau 2023). Peak monthly average air temperatures have varied from 4 to 8 °C in July and August from 2005 to 2020; however, daily highs above 10 °C are increasingly observed in these peak summer months over the last decade (data extracted from Environment and Climate Change Canada, Historical Climate Data website https://climate.weather.gc.ca/index_e.html on January 25, 2023).

Paleogeography

The maximum extent and precise retreat chronology of the Laurentide Ice Sheet (LIS) within the NE Baffin Island study area remain the subject of ongoing study and debate (Miller et al. 2002; Briner et al. 2007; Dalton et al. 2023). Most recently, analysis of marine bathymetric and geophysical data from western Baffin Bay suggests that the maximum limit of the LIS may have extended beyond the shelf break (Brouard and Lajeunesse 2017). Retreat of the ice margin may have been episodic in places and catastrophic in others with periods of relative stability on the continental shelf (Kelleher et al. 2022), although ages for the mapped glacial landforms on the shelf are sparse. Similarly, despite recent detailed mapping of glacial and post-glacial submarine landforms within several fjords of northeastern Baffin Island (Brouard and Lajeunesse 2019; Couette et al. 2023), the lack of age control on these features limits the interpretation of the precise timing of deglaciation within the fjord systems, with few exceptions (Kelleher et al. 2022). Based on continent-scale reconstruction, the ice retreat within the fjords appears to have occurred between 12 and 8 ka (Dalton et al. 2023). All ages in this paper are calibrated to IntCal20 (Reimer et al. 2020) and Marine20 (Heaton et al. 2020).

Despite rising global eustatic sea level since the early Holocene, the relative sea level of northeastern Baffin Island has been lowering at rates as high as −8 to −12 mm/year in the early Holocene owing to regional glacio-isostatic rebound (Syvitski et al. 2022). Local marine limits range from 73 m asl in Cambridge Fjord in the northern region of the study area (Andrews 1987) and from 40 to 88 m asl in fjords of the Home Bay area, in the southernmost region of the study area (Andrews et al. 1970), as evidenced from ubiquitous topset-foreset contacts in raised deltas throughout the study region.

Previous landslide research in the study area

Systematic rock avalanche deposit inventories are sparse in Canada (Schwab et al. 2003; Geertsema et al. 2006; Mitchell et al. 2020), and to date, none has been carried out in the eastern Canadian Arctic. This work builds on recent efforts to address current knowledge gaps in the seismic and tsunami hazard of the northeastern Canadian Arctic, which include a preliminary landslide overview (Gosse et al. 2020), post-glacial geohazard assessment of Pond Inlet and Eclipse Sound (Broom 2019; Gosse et al. 2020), and an oral history of felt earthquakes, observed slope failure, and displacement waves from the Kangiqtugaapik community (Gosse et al. 2020). Collectively, these studies have demonstrated that earthquakes, rock avalanches, and displacement waves have occurred in northeastern Baffin Island, some in recent history. However, the recurrence rate, correlation, causal mechanisms, and hazard potential remain poorly understood.

Terminology

Rock avalanches are a specific category of rock slope failure defined by a combination of large volume (generally > 1 × 10⁶ m³), high mobility (reach angle > 32°, H/L ratio < 0.625), and intense internal fragmentation that together result in high-energy, high-velocity granular flow (Hermanns et al. 2022). Rock avalanches are spontaneous, extremely fast, high-energy release events, and more destructive than other types of rock slope failures such as rock slides, rock collapses, and rockfalls (Evans et al. 2006), which are also observed throughout the study area. This study focuses on rock avalanches specifically as the landslide type most likely to trigger catastrophic displacement waves in fjords or lakes given their large volume and energy. We use the term “displacement wave” to describe those waves generated from subaerial mass movements into bodies of water in lieu of the term “tsunami,” which is more appropriately applied to earthquake-triggered waves (Hermanns et al. 2013).

To discuss the controlling mechanisms for rock slope failures, we employ the framework of McColl (2015), which differentiates preconditions, preparatory factors, and triggers otherwise defined in Hermanns et al. (2006) as “static boundary conditions,” “dynamic boundary conditions,” and triggers. Preconditions are defined as innate factors that impact the strength and stability of a slope, and which are, in human terms, constant over time (i.e., lithology, structure, rock mass strength). Preparatory factors are those that can gradually affect the stability of a slope over time, but do not necessarily lead to imminent failure (i.e., glacial debuttressing, rock fatigue, permafrost degradation); however, by progressive action of these factors, critical stability conditions can be reached such that failure can occur without any external trigger. Triggers are the immediate factors that drive slope condition from a critical state to failure (i.e., heavy precipitation, individual earthquakes) (McColl 2015; Blondeau et al. 2021). In the absence of direct observation, the identification and characterization of specific triggers are often untenable (Eberhardt et al. 2004), whereas preconditions and preparatory factors can be inferred through systematic studies (Blondeau et al. 2021).

Methodology

To achieve the research objectives, a regional-scale inventory of onshore rock avalanche deposits was developed using publicly available, high-resolution, remotely sensed data. This study focusses on the mapping of terrestrial rock avalanche deposits. A similar systematic assessment of potential rock avalanche deposits in the marine environment is not yet possible owing to the limited availability of full-coverage high-resolution multibeam sonar data throughout the study area.

Spatial data

Spatial data used in this study can be divided into two categories: (1) data used directly for systematic mapping and (2) supplemental interpretative data. Spatial data used directly for mapping consisted primarily of high-resolution optical imagery and digital elevation models (DEMs). Optical imagery was made available through the ArcGIS Pro World Imagery base map by Esri and used herein under license. The dataset consists of a mosaic of high (0.3 to 0.6 m horizontal) resolution Maxar Worldview satellite images providing full coverage of the study area. The World Imagery dataset is a dynamic base map built on multi-temporal mosaics. As a result, the optical imagery used in this study ranges from acquisition dates of 2014–2022 and in lieu of being an instantaneous view, was considered a “contemporary” snapshot of rock avalanche deposits within the study area. It is recognized that there may be sources of error associated with using multi-temporal spatial data. For elevation data, the 2-m resolution Arctic DEM (Porter et al. 2018) provides full coverage of the study area with occasional isolated data gaps in areas of extreme vertical relief or overhangs. The ArcticDEM data for this area were created in 2018. They were downloaded in 50 km × 50 km tiles and mosaicked over the breadth of the study area.

Supplementary interpretative data used in this study included aerial photographs, bedrock geological maps, and seismological data. Digital mapping was supplemented with the review of a hard-copy collection of overlapping 1:49,000 (1958, 1961) air photos from the National Air Photo Library, which provide approximately 85–90% coverage of the study area. The air photos were frequently referenced to investigate the historical geomorphology of landforms and deposits identified during mapping. Other supplementary data, not directly used for systematic mapping but referenced throughout the analysis, included a compilation of 1:250,000 scale bedrock geology maps prepared by the Geological Survey of Canada (Jackson and Davidson 1975; Jackson and Morgan 1978; Jackson et al. 1979; Henderson 1985), and a catalogue of seismic records with approximate epicenter locations and magnitudes (events > M3.0) that records the regional historical seismicity from 1917 to 2018 (Storchak et al. 2013, 2015; Di Giacomo et al. 2018; International Seismological Centre 2021).

Mapping procedure

All digital spatial data were compiled into a geographical information system (GIS) database using ArcGIS® Pro Version 2.8.5 software by Esri. Optical imagery, DEM, and available multibeam sonar bathymetry were projected in both 2D map view and 3D scene view to facilitate inventory mapping. Bedrock geology and the seismological catalogue were stored exclusively in 2D map view for reference and analysis.

A systematic mapping approach was implemented based on a methodology modified from susceptibility mapping by the Norwegian Geological Survey (Böhme et al. 2023). The study area was subdivided into 50 km × 50 km mapping frames and mapped at scales of 1:10,000 and 1:50,000 using both 2D and 3D synchronous views in ArcGIS^® Pro. Mapping was focused by outlining areas with a minimum slope height of 200 m and slope gradients > 25°, which are considered common criteria for the generation of rock avalanches and slope deformations (Böhme et al. 2023). Rock avalanche deposits were identified by searching for large lobate deposits of variably fragmented rock. In supraglacial environments, a high contrast in hue and intensity was an effective indicator of a deposit. However, where deposits occurred against bedrock, talus, or other ice-free terrain, a textural contrast in debris size and morphology, the slope of the deposit relative to adjacent slopes, and elevation asperities were used as visual indicators. A minimum surface area of 0.1 km² was determined as the lower limit criterion for a potential rock avalanche deposit, which approximates the minimum volume cutoff of 10⁶ m³ for a rock avalanche deposit (Hsu 1975). Evidence of potential high mobility of the deposit was also evaluated using maximum runout distance, H/L ratio, and Fahrböschung (from German, translated to “reach” in English) angle (Hungr 2006; Hermanns et al. 2022). Where possible, potential source zones in the adjacent bedrock slopes were identified (i.e., smooth slip surfaces, where no patina or lichen cover suggests recent failure/exposure) and mapped for each deposit.

The rock avalanche deposits were outlined as polygons in ArcGIS Pro. Measurements and estimates for each deposit included position (latitude/longitude), surface area (m²), maximum travel distance, H/L ratio, and travel (Fahrböschung) angle. The centroid of each deposit was estimated to produce latitude and longitude coordinates. The estimated location of the source area was identified for all deposits with varying levels of certainty owing to optical occlusion in the satellite imagery and shadow zones in the DEM. Volume estimates are provided by approximating an average thickness for each deposit and multiplying by the measured surface area. These volumes are provided with a high degree of uncertainty given the lack of pre-event topography and are included herein to provide a general sense of scale for these deposits.

Results

Rock avalanche deposits

Within the limits of the overall study area (~60,000 km²), a sub-area of approximately 10,000 km² was identified using the terrain criteria to have the potential to generate rock avalanches. In this area, only six rock avalanche deposits were identified, three of which are interpreted to represent multiple rock avalanche events at the same location.

The deposits have been labelled numerically in order from north to south with the prefix RAD (rock avalanche deposit). A map of the spatial distribution of the deposits relative to physiographic elements of note is provided in Fig. 2. Table 1 summarizes the characteristics of each deposit, including coordinates, surface area, H/L ratio, angle of reach, volume estimates, and interpreted chronology where available. Where noted, H/L ratios and Fahrböschung angles are not provided because they are not representative of the potential extent and mobility of the original event as the distal limits of several deposits have either been eroded by glaciofluvial processes or are submerged below fjord waters. In other words, where glaciofluvial processes have reduced the footprint of the deposit, the resulting H/L ratio and Fahrböschung will be artificially high for typical rock avalanche mobility. A detailed description of each rock avalanche deposit is provided below, and selected images of each of the deposits are provided in the corresponding Figs. 3, 4, 5, 6, 7, and 8.

Fig. 2

Distribution of rock avalanche deposits (RAD-1 to RAD-6) throughout the study area relative to modern seismicity. Seismic catalogue (1917–2020; ISC, 2021). Basemap: Earthstar Geographics (ArcGIS^® Pro)

Table 1 Rock avalanche deposit summary—spatial and geometric characteristics

Fig. 3

Initial 2008 RAD-1 deposit observed with optical imagery (dated August 2014) draped over ArcticDEM to create 3D Scene in ArcGIS Pro. A Birds-eye overview of RAD-1 looking westward. B RAD-1 deposit looking southward. C RAD-1 side valley overview of source zone; however, the specific source area for the 2008 event has not been identified with available imagery. Imagery does not illustrate the exact source block and sliding planes for failure. Images are prepared in ArcGIS Pro powered by Esri

Fig. 4

Second RAD-1 rock avalanche event overview: A Sentinel-2 L1C satellite imagery dated August 2016, B helicopter fly-over photograph of RAD-1 deposit dated August 2019, and C helicopter fly-over photograph of RAD-1 source zone dated August 2019, which shows a clear source zone and sliding plane (yellow arrow)

Fig. 5

Daily temperature highs (°C) from Clyde River, Nunavut (solid line), and Pond Inlet; Nunavut (dashed line) weather stations for summer period from June 1 to September 30, 2008, and 2016. Rock avalanche event time ranges are illustrated by blue arrow, showing both occurred within days of peak summer temperatures. Data extracted from the Environment and Climate Change Canada Historical Climate Data website (https://climate.weather.gc.ca/index_e.html) on January 25, 2023

Fig. 6

RAD-2 rock avalanche overview: A plan view of deposit, Maxar World Imagery (dated August 2017) draped over ArcticDEM in ArcGIS Pro (powered by Esri), which shows drag of the deposit approximately 1800 m from the original position, and B 1958 air photo view of RAD-2 deposit

Fig. 7

RAD-3 overview, A Maxar World Imagery (dated August 2020) draped over ArcticDEM, B zoom-in of source zone and daylighting slip plane, and C zoom-in of deposit morphology with potential secondary adjacent deposit. Images are prepared in ArcGIS Pro powered by Esri

Fig. 8

RAD-4 overview, A Maxar World Imagery (dated August 2021) draped over ArcticDEM, B imagery of the source zone (yellow dashed line denotes intersection of failure plane and vertical release planes) and daylighted thrust fault (white dotted line), and C imagery of primary deposit morphology with potential secondary adjacent deposit or rock glacier (orange arrow). Images are prepared in ArcGIS Pro powered by Esri

RAD-1

RAD-1 is a multi-event supraglacial deposit located in the northernmost region of the study area, at the foot of a currently glaciated side valley. The deposit was previously identified during field mapping of the ULLINIQ project (Gosse et al. 2020), and their interpretation has been substantiated as a rock avalanche consisting of a mixture of glacier ice, colluvium, and bedrock. Our analysis of available optical satellite imagery indicates that this deposit formed during at least two distinct failure events. An initial rock avalanche occurred between July 20 and 31, 2008 (Fig. 3A–C), followed by a larger rock avalanche event over the same area that mobilized between July 27 and August 14, 2016 (Fig. 4A–C). The initial landslide deposit covered an area of approximately 265,000 m² with a H/L ratio of 0.37 and a maximum runout length of 2.0 km (Fig. 3A, C). It formed a lobate geometry with lateral levees and an apparent “hollow” core likely formed by the melting of ice debris and snow. The second larger event covered an area of 485,000 m² with a similar lobate geometry but lacking the hollow core of the initial event. The latter deposit has an H/L ratio of 0.33 and maximum runout length of 2.2 km (Fig. 4A, B).

The RAD-1 events were sourced from an east-facing glacier side valley that hosts Archean banded migmatites (Jackson and Davidson 1975). The precise source zone and mechanism of failure for the events are difficult to discern from optical satellite imagery due to a combination of shadowing and limited resolution in the steep side valley. During a helicopter-supported site investigation, Gosse et al. (2020) reported foliated felsic gneiss bedrock in the probable source zone, with foliation dipping approximately 40° down slope and an intrusive mafic lens in the upper cliff section (Fig. 4C). Comparison of satellite images before and after the two events suggests that the cliff retreated approximately 150 m and 100 m, respectively, during the failure events (Gosse et al. 2020). Basal detachment is interpreted to have originated in the upper 1000–1200 m of bedrock backslope, though accumulations of colluvium, snow, and ice within the valley were likely mobilized with each event.

Daily air temperature data at the Pond Inlet (~200 km from RAD-1) and Clyde River (~1100 km from RAD-1) weather stations were extracted from the Environment and Climate Change Canada Historical Climate Data website (https://climate.weather.gc.ca/index_e.html) on January 25, 2023. Evaluation of data from the time periods of the interpreted rock avalanches (summer 2008 and 2016) shows that both failure events occurred within days following peak summer temperatures, which reached 20 °C in the region (Fig. 5). Precipitation data during these time periods are unavailable for weather stations within the study area.

RAD-2

RAD-2 is another supraglacial deposit, located approximately 20 km south of RAD-1. This deposit was also identified during field mapping of the ULINNIQ project (Gosse et al. 2020) and is similarly interpreted to be formed from a rock avalanche composed of bedrock, colluvium, and ice. In recent imagery, the deposit is lobate, curved, and stretched northward over approximately 1.8 km of supraglacial transport from its interpreted side valley source area (Fig. 6A, B). This indicates that RAD-2 is comparatively older than RAD-1, which remains proximal to its side valley source zone. Based on review of aerial photographs, the failure occurred prior to 1958; however, given the proximity and position of the deposit relative to the assumed source zone in an east-facing side glacial valley, it likely occurred within a period of a few months to years prior to the 1958 air photo (Fig. 6C). Measurements of the deposit from its 1958 position, considered most representative of initial failure, suggest a surface area of 340,000 m², a H/L ratio of 0.32, and maximum runout of 2.0 km.

Pre- and post-failure imagery is unavailable for this deposit to evaluate changes to the interpreted source zone within the nearby glacier side valley. Review of the earliest imagery (1958 air photos) corroborates that the deposit is likely a combination of snow, ice, colluvium, till, and bedrock. An alternative interpretation of this event as a glacier detachment from the side valley was considered; however, till in these side valleys generally consists of a thin veneer and is not likely sufficient to solely account for the volume of material observed in the deposit, suggesting appreciable bedrock detachment from the valley walls, which consist of Archean felsic-dominated gneiss (Jackson and Davidson 1975). A probable source zone in the bedrock escarpment on the north wall of the side valley has been identified where a persistent bedrock foliation dips steeply out of the side valley slope (Fig. 6B).

RAD-3

RAD-3 consists of two distinct, but adjacent, coastal deposits located near the head of Royal Society Fjord at the intersection of a side valley glacier and the main southeast fjord wall (Fig. 7A). The deposits are interpreted as rock avalanches, with no evidence of ice or colluvium entrainment within the debris. The volumetrically larger and likely primary deposit is boulder-rich and lobate, with a total measured surface area of 120,000 m² (yellow outline Fig. 7C). The distal limit of the primary deposit is well defined and shows evidence of active glaciofluvial erosion, which suggests that the original deposit was likely larger than that observed today. The proximity of the deposit to the fjord wall may be the result of this material being deposited at a time when a glacier more closely abutted the fjord wall, forcing much of the debris accumulation proximal to the slope. Subsequent erosion has removed the distal limits of the deposit.

In the same area, a secondary, more subdued, and lower-relief deposit extends northward from the primary RAD-3 deposit, abutting the distinctive lateral moraine of the adjacent retreating valley glacier (orange outline Fig. 7C). This deposit appears thinner but covers a similar overall surface area to the primary mapped deposit. Interpretation of the available ArcticDEM and optical imagery suggests that this deposit occurred after the primary deposit, flowing over its northern flank. The limit of the secondary deposit appears intact, with no evidence of significant erosion, which suggests that the original deposit footprint is likely intact and that deposition occurred after a full glacier retreat in this area. This is exemplified by the observed lower H/L ratio of 0.51, corresponding to its preserved longer runout at the foot of the fjord wall slope.

The source zone of both rock avalanche deposits is interpreted as a wedge-shaped scar in the NW-facing fjord wall, approximately 700 m asl. The failure area is hosted in a faulted contact zone between Archean granodioritic migmatite and the Archean Mary River Group metasedimentary and metavolcanic rocks (Jackson et al. 1979; de Kemp and Scott 1998). High-resolution imagery in the source zone is heavily shadowed inhibiting a clear interpretation of the local lithology, structure, and potential failure mode. However, based on the morphology of the crest and apparent anisotropies in the exposed bedrock, a strong structural control on the slope failure is certain (Fig. 7B).

RAD-4

RAD-4 is identified near the head of Clark Fjord at the intersection of the southern fjord wall and a large side valley. The deposit is a raised plateau of variably fragmented bedrock material that is distally truncated by a modern glacial meltwater stream along which significant active erosion and oversteepening of the deposit toe is apparent (Fig. 8A). As with RAD-3, the present-day deposit is interpreted as a partial deposit, and therefore, the total volume, maximum runout, and mobility of the original event are underestimated with recent imagery. The deposit has a measured surface area of 280,000 m². H/L ratio and maximum runout are not measured for this deposit given the clear erosion of the distal portion. Material in the deposit ranges from apparently fine-grained debris to large boulders measuring up to 10 m in maximum dimension and is assumed to be sourced primarily from bedrock in the overlying slope, with no evidence of ice entrainment, though a possible component of valley side till or previous colluvium cannot be excluded.

The primary deposit is sourced from a large block at the crest of the adjacent east-facing bedrock slope, approximately 1000 m above the valley floor. A persistent through-going structure, mapped as a thrust fault separating the lower Archean/Paleoproterozoic gneiss from overlying Paleoproterozoic granitoids (Jackson and Morgan 1978), cuts the slope approximately 500 m above the valley dipping out of the slope (~45°) and forms the basal sliding plane of the source block with two steeply dipping, intersecting fractures forming the back release planes (Fig. 8B). While the ultimate trigger cannot be determined for this event, bedrock structure is considered a first-order control on this event. A smaller secondary deposit immediately north of the primary deposit is interpreted as a potential second rock slope failure at the site, or a remobilization of some material from the initial rock avalanche debris. The secondary deposit exhibits solifluction lobes on its surface, suggesting post-depositional gravity deformation of the likely permafrozen debris, caused by its position on relatively steeper valley slopes (Fig. 8C).

The deposits are evident in air photos, with very little morphological changes compared to present-day imagery beyond distal erosion, which indicates the failure events are more than decades older than the 1958 imagery; however, a specific age estimate for the event is untenable with the available data. The most recent ice margin chronologies for this region, based on available radiocarbon ages (see Dalton et al. 2020 for a summary), suggest that the fjord heads in this region of NE Baffin Island were largely ice-free circa 8 ka, providing an upper-limiting age for the deposit.

RAD-5

RAD-5 is situated approximately 2 km south-southeast of RAD-4 along the same east-facing glacier side valley slope. Glaciofluvial erosion of the limits of the deposit from the adjacent glacial meltwater stream is apparent and therefore (as with RAD-3 and RAD-4) is considered a partially preserved deposit. The deposit is the largest in the study area with a measured surface area of 600,000 m² and a lateral extent of 1.5 km along the side wall. It has an observed maximum runout distance of 1.5 km. The debris field currently consists of a series of boulder-rich lobes that host a complex pattern of internal lobate surface ridges that suggest post-failure gravity deformation and/or solifluction (Fig. 9A, C). Debris slopes at the distal limits of the deposit lack vegetation or lichen cover suggesting that the material is at the angle of repose and is undergoing active displacement and erosion. It is unlikely that the present-day limit represents the true maximum runout of the deposit as post-glacial erosion has removed some of the material from the initial event. The relatively steep gradient (15–25°) at the foot of the bedrock cliffs has also promoted gravity deformation within the deposit, which likely masks characteristics of its initial geometry.

Fig. 9

RAD-5 overview, A Maxar World Imagery (dated August 2021) draped over ArcticDEM, B imagery of the source zone—exposed glacier valley side walls form the steeply dipping sliding plane, C imagery of deposit morphology including solifluction lobes, and distal glaciofluvial erosion of the deposit. Images are prepared in ArcGIS Pro powered by Esri

The source zone of the deposit is demarcated by a large area of notable lighter color on the adjacent exposed bedrock slope, which has been interpreted as a series of steeply dipping (~50–70°) fractures that form a persistent basal sliding surface or surfaces of the rock avalanche event (Fig. 9B). The backslope of this event coincides with the same mapped thrust fault that has been mapped through the source zone of RAD-4 where overlying Paleoproterozoic granodiorite is thrust upon Archean/Paleoproterozoic gneiss (Jackson and Morgan 1978). The interpreted source zone is the largest of all recorded events and measures up to 500 m in height and 1100 m in width. As with the proximal RAD-4, the specific timing of the event is untenable; however, review of aerial imagery highlights that it occurred prior to 1958, with few morphological changes observed between 1958 and modern high-resolution imagery. It is similarly bound with a maximum age of approximately 8 ka when the ice sheet limits are modelled to have retreated from the fjord head region in this area (Dalton et al. 2023).

RAD-6

RAD-6 is situated on the northern fjord wall at the head of Tingin Fjord. This is the only rock avalanche deposit observed within the Piling Group metasedimentary sequence. The deposit is fan-like with steep margins with a limit that is obscured beneath the fjord-head waters. It is boulder-rich, with the largest boulder measuring up to 15 m in apparent maximum linear dimension. Boulders positioned directly along the shoreline of the fjord headwaters suggest that part of the deposit remains submerged beneath the fjord water (Fig. 10A, C). As a result, the total area estimate, H/L ratio, and maximum runout distance are considered conservative estimates and underrepresent the magnitude of the event. Deposit morphology is blocky with glaciofluvial channels observed along its flanks. Lineations observed in the upper slope of the deposit measure 80–100 m in length and are oriented semi-perpendicular to the rock avalanche flow path. These are interpreted as tension cracks reflecting post-depositional deformation in the deposit.

Fig. 10

RAD-6 overview, A Maxar World Imagery (dated September 2020) draped over ArcticDEM, B imagery of the source zone fjord walls illustrating a wedge-shaped scarp with steeply dipping foliation in the side walls, and C imagery of deposit morphology, variable grain size distribution, and truncation at the fjord waters. Images are prepared in ArcGIS Pro powered by Esri

The fjord walls in the source zone of RAD-6 are mapped as interlayered pelitic and psammitic schists and gneisses of the Paleoproterozoic Piling Group with local intrusions of Paleoproterozoic granites (Henderson 1985). Visual assessment of the optical imagery and DEM in the source zones highlights a large, wedge-shaped scar in the source zone (Fig. 10B). The southern slope of the back scarp exhibits steeply dipping (~70°) foliation consistent with the interpreted interlayered metasediments. This foliation dips out of the slope and likely forms a primary kinematic sliding plane. Folding might be visible in the upper portion of the slope; however, a more detailed analysis of the source zone is inhibited by intense shadowing in the optical imagery.

Spatial distribution

The limited number of deposits (n = 6) identified in the study area precludes the use of robust spatial statistics to evaluate event distribution and potential correlative relationships with other geographic factors. The deposits are broadly situated throughout the study area with observations in the northernmost, central, and southernmost sectors. While statistically significant spatial clustering or geographic patterns are not defensible, it is worth noting that in two separate areas, two deposits are collocated. Supraglacial deposits RAD-1 and RAD-2 in the northernmost region of the study area are separated by approximately 20 km. These deposits are also the only events dated to the last century suggesting a temporal correlation in addition to spatial. In the central portion of the study area, RAD-4 and RAD-5 are separated at their center points by only 2 km and are sourced from the same glacial valley side wall.

Discussion

Supraglacial rock avalanches

Two deposit locations (RAD-1 and RAD-2) are classified as supraglacial rock avalanches or rock-ice avalanches, interpreted as catastrophic slope failures involving glacier ice, colluvium, and bedrock. The supraglacial rock avalanche deposits are noted to have the highest runout length and lowest H/L ratio of the inventory, reflecting a notably higher mobility than the fjord wall rock avalanches. The mobility of these deposits is likely enhanced for two reasons: (1) due to the presence of interstitial ice, snow, and water, which reduces friction within the deposit and increases travel distance, and (2) due to deposition over the low-friction glacier surface (Schleier et al. 2015; Deline et al. 2022).

The supraglacial rock avalanches are the youngest events observed in the study area, both occurring within the last century. The collocation of RAD-1 and RAD-2 in the northernmost extent of the study area, and similar east-facing side valley source zones, points to a potential common causal mechanism. RAD-1 represents the youngest, and the only deposits with well-constrained ages determined by bookending satellite imagery, which shows initial failure in July 2008 followed by a second event in July/August 2016. Peak summer temperatures for the region hit anomalously high levels (> 20 °C) within days prior to the events (Fig. 5), which suggest that warm temperatures and the resultant slope changes such as thermal expansion, snowmelt, and buildup of pore pressures in discontinuities with permafrost thaw may have destabilized the slope, acting as a trigger or key preparatory factor for failure.

While conjectural without more precise age control on the event, we hypothesize that RAD-2 may also have been driven by warmer summer temperatures that induced increased pore water pressures and potential permafrost thaw given the physical similarities of the deposits and their source zones. Rock-ice avalanches are notably more common in summer months and in years with above average mean annual air temperature where basal meltwater results in shear stress reduction (Kääb et al. 2021; Leinss et al. 2021).

A correlation between an increase in the frequency and magnitude of mass wasting in glaciated terrain and the effects of climate change has been well-documented (Davies et al. 2001; Geertsema et al. 2006; Huggel et al. 2012; Böhme et al. 2015; Hermanns et al. 2017). Rising mean annual air temperatures, and temperature extremes, result in permafrost degradation and an increase in meltwater percolation throughout bedrock joint systems, which can drive deep, rapid thaw and subsequent rock mass degradation (Gruber and Haeberli 2007). Permafrost degradation as a primary control on rock slope destabilization throughout the late Pleistocene and Holocene has been well illustrated by evaluating paleoslip rates of rock slope deformation (Böhme et al. 2019; Hilger et al. 2021; Etzelmüller et al. 2022; Penna et al. 2023). While the present study shows that rock avalanche events remain rare in northeastern Baffin Island, the relatively recent occurrence of these supraglacial events and the projected warming trend in this region could indicate that more large-scale slope failures may occur in the future, in agreement with the findings from analogous cryospheric terrain in Svalbard and Greenland (Kuhn et al. 2021; Svennevig et al. 2022).

Fjord wall rock avalanches

Four of the interpreted rock avalanche deposits (RAD-3 to RAD-6) are observed proximal to or within fjords. These deposits are derived from the mobilization of bedrock material from the fjord and side valley walls with no evidence of significant glacier ice or colluvium entrainment. Kinematic structural control in the source zone of these events was determined based on the evaluation of the source zone lithology, structure, and rock mass condition in optical imagery. At each site, persistent (100s of meters) bedrock structure in the source zone was observed, which resulted in the formation of kinematic planar slabs, blocks, and/or wedges as noted in the “Rock avalanche deposits” section. A common observation of unstable rock slopes of varying sizes throughout Baffin Island is an association with slope-scale anisotropies such as lithological contacts between felsic and mafic rocks resulting from mafic intrusions or foliations within gneissic lithologies (Gosse et al. 2020).

These four rock avalanche deposits all pre-date the 1958 air photo acquisition period, which denotes the earliest available imagery for the study area. As a result, establishing a precise age for these deposits is untenable with the use of remotely sensed data. Based on a visual assessment of the surface condition and lichen cover of the deposits in optical imagery and in consideration of the long-term glaciofluvial erosion evident along the distal limits of several deposits, the timing of these events is inferred to be during or shortly after local deglaciation (approx. 7–8 kyr).

The identification of potential causal mechanisms for these events is complicated by the limited age control on their occurrence. Bedrock structure (foliations, through-going faults, and joints) is considered a first-order pre-condition for the occurrence of RAD-3 through RAD-6 where persistent sliding and block/wedge release planes were identified in all source zones and is a well-documented pre-condition in the literature (Hermanns and Strecker 1999; Brideau et al. 2005). This is most clearly exemplified in the source zones of RAD-4 and RAD-5 where a large, interpreted fault daylights in the glacial valley wall, likely contributing to the basal sliding plane of both rock avalanche events.

Spatial correlations

Seismicity

This project was initiated with an original hypothesis that mapped rock avalanche deposits may exhibit a positive spatial correlation to the zones of most active historical (1914–2020) seismicity in northeastern Baffin Island, reflecting a potential relationship between earthquakes and rock avalanche events. In an oral history collected from community members of Kangiqtugaapik (Clyde River), one example of a weakly felt earthquake that knocked a tea cup off a boulder and was followed by an observed rock slope failure moments later was recalled by several individuals on a multi-family hunting trip in the early 1980s (Gosse et al. 2020).

The results of this mapping yield an insufficient number of deposits to develop statistically valid spatial relationships between preserved rock avalanche deposits and seismic events. Visual comparison of the rock avalanche deposit locations in relation to clusters of seismic events exceeding magnitude 3.0 shows no strong correlation. Earthquake records with magnitudes less than 3.0 were not considered in the evaluation as these lower magnitude events are unlikely to individually trigger rock avalanche events (Keefer 2002). For each rock avalanche deposit identified, a proximal region of 200 km radius was investigated for recorded historical earthquakes that may have been coincident to the slope failure. No evidence of potential coseismicity with any of the identified rock avalanche events was observed.

However, it is important to highlight here that the seismic record for this region is sparse due to limited historical seismograph coverage (Basham et al. 1977; Bent and Voss 2022). The available catalogue within the limits of this study area encompasses seismic events extending only from 1925 to present for events >M5.5, from 1957 to present for events >M3.7, and from 1973 to present for events >M2.0 (International Seismological Centre 2021). Recent advances in seismograph coverage and focal mechanism solutions are improving our understanding of the nature of seismicity in eastern Baffin Island (Bent and Voss 2022); however, many questions remain, and the temporo-spatial distribution of post-glacial seismicity, which may have impacted the oldest rock avalanche events observed here, remains elusive. As a result, seismicity is not precluded as a potential causal mechanism for the oldest syn- to post-glacial events.

Physical setting

Other spatial characteristics of the study area were visually interpreted for potential relationships to the mapped rock avalanche deposits. A common spatial characteristic of the fjord wall rock avalanches (RAD-3 to RAD-6) is their position inland from the outer coast, proximal to the fjord heads (Fig. 11). In fact, no deposits were observed less than 60 km from the fjord mouths despite the ubiquitous presence of potential slopes and source zones that meet the minimum slope height and angle criteria defined for this study. Near the fjord heads, bedrock slopes are generally at their highest and steepest, suggesting that there may be a topographic control on the generation of rock avalanche events where potential energy of the failure would be at its highest.

Fig. 11

Rock avalanche distribution in northeastern Baffin Island relative to key physiographic province and projected ice stream positions of the Laurentide Ice Sheet during the Last Glacial Maximum (De Angelis and Kleman 2007)

In an evaluation of large rock slope failures in the British Mountains, Jarman and Harrison (2019) suggested that “concentrated erosion of bedrock,” where glaciers and ice sheets breach paleodivides (i.e., transition from highland plateaus to trough heads) and rapidly incise bedrock, may have induced sufficient stress redistribution in bedrock slopes to provoke slope failure, which could explain otherwise puzzling clusters of rock slope failures. Of interest to this theory, the four oldest rock avalanches (RAD-3 to RAD-6) identified in northeastern Baffin Island all occur proximal to major trough heads where fast-flowing ice streams within the LIS have been interpreted during the LGM (De Angelis and Kleman 2007). At these positions, the proposed “concentrated erosion of bedrock” may have plausibly occurred in the early Holocene resulting in localities of higher rebound stress. However, in a follow-up evaluation of rock slope failure distribution in the Scottish Highlands, Ballantyne and Stone (2013) highlighted that numerous glacial breaches and trough heads host no evidence of rock slope failure and that a more likely control of these events is simply overall slope height and gradient, with steep gradients perhaps playing the most significant role in determining failure conditions.

Rock avalanche source zones were evaluated for common features. Notably, the two modern supraglacial rock avalanche deposits (RAD-1 and RAD-2) were both sourced from east-facing glacier side valleys separated by a distance of less than 20 km. East-facing slopes are more commonly associated with colder ground regimes given their exposure to the cooler temperatures of morning solar radiation in comparison to the generally warmer south- or west-facing slopes. It is therefore unclear how important the roles of solar radiation and aspect are to these modes of failure. The four structurally controlled rock avalanches (RAD-3 to RAD-6) are situated within northeast-southwest trending fjord systems; however, the orientations of the fjord walls at the source zone vary from northwest-facing to east- and southeast-facing sidewalls, suggesting slope orientation and solar aspect are not key controls for these failures. In a review of a preliminary inventory of rock avalanche deposits in Norway, a similar finding was made with no discernible correlation between slope azimuth and rock avalanche frequency or magnitude (Penna et al. 2022).

Evaluation of other well-documented physiographic factors that may influence the distribution of rock avalanche events such as bedrock lithology, glacier proximity, and glacier velocity yielded no clear insights for the potential causal mechanisms of the mapped deposits. A notable qualitative observation is that the six observed RAD are in close proximity to the thickest extant glaciers in the study area, which may be responsible for a greater frequency or magnitude of glacier quakes or isostatic adjustments.

On the sparsity of rock avalanches in northeastern Baffin Island

In three counties of Norway with the highest rock avalanche activity, which comprise an area equivalent to roughly 70–80% of the northeastern Baffin Island study area, hundreds of post-glacial rock avalanches and their deposits have been documented, with over 30 noted to have triggered displacement waves (Blikra et al. 2006; Hermanns et al. 2014). In the past century in Norway, ten rock avalanches have occurred at four locations (Furseth 2006; Hermanns et al. 2021), the latest of which occurred in the mid-1980s. Preliminary landslide mapping of Greenland (Svennevig 2019) has identified 564 total landslides with 376 of those mapped within the Nuussuaq Basin on the west coast, directly across Baffin Bay from the study area. Dozens have been identified as rock avalanches, specifically, and in historic time, six documented rock avalanches have occurred at three localities (Svennevig et al. 2020, 2023). A multi-decadal inventory from the St. Elias region of Alaska identified 220 rock avalanches over a period of 36 years from 1984 to 2019 in an area of 3700 km², approximately 5% of the total study area evaluated here (Bessette-Kirton and Coe 2020). When compared to these other high-latitude fjord regions, the number of mapped rock avalanche deposits in this large study area is notably lower despite the near ubiquitous presence of several known preconditions, preparatory factors, and potential triggering mechanisms.

Here, we present two hypotheses to explain the relative sparsity of rock avalanche evidence in northeastern Baffin Island: (1) the persistence of permafrost in the study area continues to act as a stabilizing factor, and (2) overall rock mass competence is high in areas of most extreme relief contrast and slope grade in the study area, delaying rock fatigue through stress release.

Hypothesis 1—permafrost stabilization

The persistence of continuous permafrost in bedrock slopes acts as a compound stabilizing factor. Frozen ground conditions serve to both increase rock mass cohesion and reinforce joint strength. Conversely, warming of the thermal regime in a rock mass and subsequent permafrost degradation results in both thermomechanical dilation coupled with hydrothermally induced strength reduction (i.e., freeze-thaw processes) in joint systems (Davies et al. 2001; Gruber and Haeberli 2007; Hasler et al. 2012; Krautblatter et al. 2013) and has been proposed as a first-order control on rock avalanches and deep-seated deformation in rock slopes (Frauenfelder et al. 2018; Magnin et al. 2019; Bessette-Kirton and Coe 2020; Hilger et al. 2021; Etzelmüller et al. 2022; Penna et al. 2023; Kuhn et al. 2023).

As noted in the “Physical setting and geology” section, detailed studies and site investigation of bedrock permafrost condition throughout Baffin Island remain sparse and have been focused primarily in and around coastal communities at low elevations. The spatial variability of permafrost in the steep slopes of the fjords and glacial valleys remains uncertain. However, based on regional ground temperature modeling (Westermann et al. 2015), the presence of continuous permafrost with mean annual ground temperatures from −14 to −10 °C is considered representative for the steep, high-relief terrain that was the focus of this study. It is anticipated that this continuous permafrost persists throughout the study area; however, uncertainty remains in the potential development of active layers during the seasonal thaw, particularly on south-facing fjords walls that would be subject to higher solar incidence throughout the summer months.

A recent study of bedrock permafrost from western Greenland (Marcer et al. 2022) suggested that bedrock permafrost in north-facing slopes remains persistent to sea level; however, south-facing slopes are experiencing degradation with discontinuous permafrost conditions up to approximately 400 m asl. This permafrost degradation has been suggested as a key preparatory factor for several large slope failures in western Greenland (Strzelecki and Jaskólski 2020; Svennevig et al. 2020, 2022). This active permafrost degradation, anticipated to be warming ahead of the northeastern Baffin Island coast, may be one factor to explain the larger number of rock avalanche events observed in western Greenland relative to this present study.

Hypothesis 2—bedrock lithology and rock mass quality

Bedrock geology within the study area is dominated by the orthogneisses, granites, and supracrustal rocks of the Archean/Paleoproterozoic Rae Craton and, to a lesser extent, the metasediments of the Paleoproterozoic Piling Group in the south, near Home Bay (see Fig. 1, St-Onge et al. 2009). Fjords with the greatest slope height and near-vertical gradient are observed in the generally stronger and more competent basement gneisses and granite throughout the study area. Slope heights and angles generally diminish in the metasedimentary bedrock in the southernmost region of the study area. While stronger, more competent, bedrock lithology does not preclude rock avalanche activity, it is likely that the observed higher rock mass quality (strength and deformability) of the gneisses throughout much of the study area may contribute to the relatively few rock avalanches observed. In comparison, the region of western Greenland with the highest density of rock avalanche activity (i.e., the Vaigat Strait and Disko Bay regions) (Svennevig 2019) is situated within Early Paleozoic sediments, equivalents of which are not observed in the Baffin Island study area, and Proterozoic metasediments broadly equivalent to the Piling Group on Baffin Island. Combined with the impacts of permafrost degradation noted in the “Hypothesis 1—permafrost stabilization” section, the in situ rock mass quality of the fjord walls in the sedimentary and metasedimentary lithologies is anticipated to be relatively lower and potentially preconditioned for failure relative to those in the older basement gneissic lithologies.

Rock fatigue is defined as the weakening of a rock mass from cyclic loading, resulting in damage propagation and expansion of internal micro and macro fracture systems (Liu and Dai 2021). Rock fatigue in high-relief glaciated terrain is attributed to a combination of paraglacial stress release (Ballantyne 2002; Cossart et al. 2008; McColl 2012) and in situ stresses induced by the interaction of topography, tectonic history, and erosional dynamics of a rock mass (Miller and Dunne 1996). The effect of glacial debuttressing on slope destabilization is most significant at the onset of deglaciation and glacial thinning, particularly where bedrock structure is oriented to form kinematic slabs, blocks, or wedges. However, geomechanical modeling suggests that long-term rock fatigue attributed to paraglacial stress release can also span several thousand years and is a key preparatory factor for many contemporary slope deformations and failures (Guglielmi and Cappa 2010). This time-dependency of rock fatigue is governed by both external processes (i.e., those driving stress redistribution) and internal factors—primarily rock mass strength (Wyrwoll 1977). Rock mass strength, a combination of intact rock strength and bedrock structure, is a critical control of slope stability (Norrish and Wyllie 1996). As a result, where rock mass quality is high and bedrock structure is favorably oriented for kinematic stability, the effects of rock fatigue may be delayed.

We propose that the rock avalanche failures observed in this study represent those that were structurally preconditioned to failure along persistent, intersecting discontinuities that created kinematic blocks and therefore occurred immediately or soon after local deglaciation where glacial buttressing is removed and paraglacial stress release is highest. While it is recognized that rock fatigue associated with paraglacial stress release is actively ongoing throughout the study area, the generally strong and competent rock masses that form the fjord walls, coupled with the permafrozen ground conditions, are likely delaying the effects of rock mass degradation and sustaining high and steep walls longer than those incised into the generally weaker and more anisotropic sedimentary, volcanic, and metasedimentary units that dominate the areas of highest landslide activity on the western Greenland coast.

A period of quasi-stability

Several models have been proposed to explain post-glacial rock slope failure frequency in mountainous terrain. Two of the earliest assertions include a constant frequency model, in which the frequency of post-glacial rock slope failure does not change over time after deglaciation, and an exhaustion model, in which the overall probability of failure diminishes exponentially over time (Cruden and Hu 1993). Subsequent research pointed to a more plausible “rapid response” model where most rock slope failures occur within a short period after deglaciation followed by a much lower, and relatively constant frequency (Ballantyne and Stone 2013; Hermanns et al. 2017). However, in light of warming global temperatures, further studies have begun evaluating permafrost degradation as a potential late-stage perturbation to the long-term “constant frequency” proposed in the rapid response model. Permafrost degradation, with the associated thermomechanical damage and increased pore water pressures, has been established as a probable first-order control for accelerating rock slope deformation and landslide activity (Frauenfelder et al. 2018; Böhme et al. 2019; Hilger et al. 2021; Svennevig et al. 2023).

We speculate that northeastern Baffin Island likely experienced a relatively higher frequency of rock avalanche activity during and immediately after deglaciation of the fjords, driven primarily by the early effects of glacial debuttressing and paraglacial stress release, in line with the “rapid response” model. The deposits of these events would have largely been transported supraglacially and redistributed by glacial or glaciofluvial processes throughout valleys and fjord sea floors, unrecognizable as rock avalanche deposits in a systematic mapping program. Examples of supraglacial rock avalanche deposit transport have been described from Norway (Schleier et al. 2015), Alaska (Bessette-Kirton and Coe 2020), New Zealand (Reznichenko et al. 2011; Dunning et al. 2015), Spain (Carrasco et al. 2013), and Tierra del Fuego (Jackofsky 2000; Evenson et al. 2009), with many pointing to local stress release during the onset of deglaciation as a key factor.

After this initial period of slope instability immediately after deglaciation, a period of relative stability may be expected while permafrost ground conditions continue to function as a ubiquitous stabilizing factor. Permafrost degradation can variably lag deglaciation by millennia depending on local ground conditions, physiography, and climate influences (Jones et al. 2023). With the acknowledged role of permafrost thaw on slope destabilization, a secondary period of higher frequency rock avalanche activity may then be anticipated after this period of relative quiescence if climate warming reaches a threshold required to raise ground temperatures and degrade existing permafrost. We offer a conceptual model of post-glacial rock slope failure frequency in tectonically quiescent Arctic regions (Fig. 12), following the “rapid response model” (Ballantyne and Stone 2013; Hermanns et al. 2017) of high rock slope failure frequency that is anticipated soon after local deglaciation, which quickly tapers to a lower constant rate of failure. We propose that there is a lagging perturbation to that model where the climatic effects of permafrost thaw and increased pore water pressures from snow/ice melt and increased rain/snow ratio will increase rock slope failure frequency.

Fig. 12

Conceptual model of post-glacial rock slope failure frequency in tectonically quiescent Arctic regions. The curve is normalized to local preconditioning factors such as rock mass quality and climate. The sword bars depict the progress through the slope failure frequency curve for three representative regions: Norway, western Greenland, and Baffin Island

One proof of concept of this model is the observed changes in acceleration of a large deep-seated gravity deformation in Norway, which has not yet catastrophically failed (Böhme et al. 2019). Early Holocene slip rates were fast following deglaciation, slowed through the mid-late Holocene with cooler temperatures, and then increased again with rising temperatures since 1900 (Böhme et al. 2019). Monitoring data from hundreds of unstable rock slopes in Norway indicate that higher contemporary slip rates relative to historic rates are being observed at failure scarps where permafrost is mapped as sporadic discontinuous (Hermanns et al. 2021).

We propose that northeastern Baffin Island may currently be in a period of relative stability with respect to the formation of rock avalanches—i.e., between the two regimes of early-onset slope instability due to paraglacial stress release and the lagging instability from permafrost degradation. As the climate continues to warm, with accelerated temperature rise at higher latitudes (Ballantyne et al. 2010, 2013), extensive permafrost thaw may eventually lead to a higher frequency of rock avalanche activity on Baffin Island, more comparable to that observed in the fjord terrain of Norway and Greenland. This proposed model may impact current risk assessments of landslides and associated displacement waves in regions that are currently undergoing both deglaciation and permafrost degradation, including Greenland, Tierra del Fuego, the Canadian Arctic, and Alaska.

Limitations of the study

Given the lack of full-coverage systematic bathymetric data throughout the study area, this study does not include a detailed assessment of potential marine deposits in the inventory presented. In the fjords that have been partially surveyed with multibeam sonar, the fringe zone proximal to the fjord side walls is generally excluded due to the limits of multibeam sonar coverage and research vessel access. This inhibits mapping in zones where large portions of coastal rock avalanche deposits may be situated. However, review of the available data and published fjord seafloor studies (Brouard and Lajeunesse 2019; Bennett et al. 2020) suggests that evidence of submarine rock avalanche deposits remains similarly sparse within the study area.

Through remote sensing alone, the differentiation of the deposits of glacial and periglacial processes from those of slope processes in glaciated terrain is notoriously challenging (Deline and Kirkbride 2009; Hartvich et al. 2017). Rock glaciers and debris cones representing the accumulation and subsequent creep of large rockfall from overlying fjord walls over 100s to 1000s of years can be difficult to discern from relatively instantaneous failure events (Clark et al. 1998; Hartvich et al. 2017). Alternative interpretations for some deposits identified herein may be plausible with further field investigation and ground truthing. Potential supraglacial deposits that may have occurred near the terminus of tidewater glaciers in this region could be underrepresented in optical imagery as they would be transported and dispersed in the marine environment. However, given the relatively low glacier velocity throughout northeastern Baffin Island, it is anticipated that the number of these deposits would be few and that, as exemplified in RAD-2, supraglacial deposits may be visible in imagery for up to a century.

Conclusions

The purpose of this study was to establish the most likely preconditions and preparatory factors controlling the frequency and distribution of rock avalanches in this region and to better understand the apparently low documented record of rock avalanche deposits on Baffin Island in comparison to analogous fjord regions of Greenland and Norway. Here, we present the first systematic inventory of preserved post-glacial rock avalanche deposits in northeastern Baffin Island. Evaluation of the low occurrence and the spatiotemporal distribution of the deposits has provided insights into the prevalence and controls of rock avalanche events in this region. These can be summarized as follows:

• There are few (n = 8 at 6 locations) preserved subaerial rock avalanche deposits observed in northeastern Baffin Island despite the presence of several preconditions, preparatory factors, and potential triggers in the area. The observed deposits are far fewer than comparable inventories carried out in analogous high-latitude fjords systems such as Norway, Greenland, and Alaska.

• The rock avalanches observed can be classified as either (1) supraglacial rock avalanches that involved the entrainment of bedrock, colluvium, and ice/snow and (2) fjord wall rock avalanches that involved primarily bedrock material and exhibited a strong structural preconditioning in the source zone.

• The supraglacial rock avalanches occurred within the last century and are interpreted to have climatogenic controls. The most recent rock avalanche deposit consisted of at least two distinct failure events (2008 and 2016) that occurred within days of peak summer temperatures.

• The fjord wall rock avalanche deposits likely date to a period during or shortly after local deglaciation, though precise age control is untenable. These events all occurred in proximity to fjord heads where slope heights and gradients are at a maximum, suggesting some topographic control on rock avalanche occurrence in this region.

• Two hypotheses are suggested to explain the sparsity of rock avalanches in the terrestrial record of northeastern Baffin Island: (1) persistent permafrost in the region continues to act as a stabilizing factor, and (2) overall rock mass competence is high in areas of most extreme relief contrast.

• We recognize that some deposits may remain masked underwater in deep and steep-walled fjords where bathymetric data are either lacking or too sparse for systematic mapping; however, review and mapping of the available bathymetry suggests that coastal-derived submarine deposits are similarly sparse.

This work is intended to be a foundational inventory and research program from which more detailed analyses can build the contemporary understanding of rock avalanche processes in the eastern Canadian Arctic. As the climate warming trends continue, it will be increasingly important to continue these studies to assess the hazard of rock avalanches and their secondary geohazards in fjord environments.