Mechanisms, volumetric assessment, and prognosis for rapid coastal erosion of Tuktoyaktuk Island, an important natural barrier for the harbour and community

Warming of the Arctic climate is imposing substantial change on the already retreating coast of the southeastern Beaufort Sea. The high sensitivity of coastal erosion to a warming climate has been documented over extensive coastal regions of the Arctic Ocean (Irrgang et al. 2022; Nielsen et al. 2022) and changes in the drivers of shoreline retreat are clearly evident along the western Canadian Arctic coast.

The Arctic is warming at twice the global mean rate (IPCC 2013) and the onset of this “Arctic amplification” can be traced to the 1970s (Nielsen et al. 2022). The warming is contributing to higher sea-surface and ground temperatures, changes in storminess, reduced ice in the Arctic seas (more prolonged open-water seasons, wider open-water fetches, and increased wave energy), and a global rise in sea level (Forbes 2011; Forbes and Hansom 2011; Overeem et al. 2011; Irrgang et al. 2022). In the Mackenzie Delta region, mean annual air temperatures have increased at three times the global mean rate (Zhang et al. 2019) and increases in mean annual ground temperature have been documented (Burn and Kokelj 2009), leading to greater depths of seasonal thaw (O’Neill et al. 2019). The rate of sea ice loss has been more rapid than previously predicted (Stroeve et al. 2012; Barnhart et al. 2014a), exposing wide parts of the ice-rich Arctic Ocean coasts to extensive open water on an annual basis (Barnhart et al. 2014b) and increasing the potential for damaging storm surges (Vermaire et al. 2013). There is now widespread evidence of accelerated erosion rates in Yakutia, Chukotka, Alaska, and in the Canadian Beaufort Sea (e.g., Overeem et al. 2011; Jones et al. 2009, 2020; Günther et al. 2013; Barnhart et al. 2014b; NOAA 2015; Irrgang et al. 2018; Farquharson et al. 2018; Maslakov 2019; Berry et al. 2021).

Processes and rates of coastal change in the Canadian Beaufort Sea have been the focus of studies dating back almost 60 years (Mackay 1963, 1986; McDonald and Lewis 1973; Lewis and Forbes 1975; Forbes and Frobel 1985; Harper et al. 1985; Harper 1990; Ruz et al. 1992; Solomon et al. 1994; Solomon and Covill 1995; Dallimore et al. 1996; Hill and Solomon 1999; Solomon 2005). Comparable early studies were undertaken along the Alaska coast (e.g., Hume and Schalk 1967; Hume et al. 1972; Kobayashi 1985; Reimnitz et al. 1985; Reimnitz and Aré 1998), where a marked acceleration of erosion rates was first documented (Mars and Houseknecht 2007; Jones et al. 2009).

More recently, there has been a rapid growth in research activity along the entire mainland Beaufort Sea coast, leading to substantial advances in understanding of the rates and processes of land loss (e.g., Jones et al. 2008; Hoque and Pollard 2016; Barnhart et al. 2014a, 2014b; Radosavljevic et al. 2016; Fritz et al. 2017; O’Rourke 2017; Ramage et al. 2017; Obu et al. 2017; Couture et al. 2018; Irrgang et al. 2018; Cunliffe et al. 2019; Lim et al. 2020a, 2020b; Berry et al. 2021).

Some progress has been made in the development of models to represent the combined effects of evolving drivers and interacting thermal and mechanical erosion processes. These range from detailed process representation (e.g., Kobayashi et al. 1999; Hoque and Pollard 2009; Ravens et al. 2012; Barnhart et al. 2014a; Bull et al. 2020) to empirical approaches scalable to more extensive segments of coast (e.g., Kupilik et al. 2019, 2020; Nielsen et al. 2022) and initial consideration of artificial intelligence strategies (Irrgang et al. 2022).

Extensive surveys of coastal retreat along the Beaufort Sea coastline from Barrow to Tuktoyaktuk reveal highly variable rates both in time and space, with extremes up to 22 m/year (Solomon 2005; Gibbs and Richmond 2015; Jones et al. 2018). Some of the highest erosion rates are associated with large ice-rich cliffs and breached thermokarst lakes (Mars and Houseknecht 2007; Jones et al. 2018, 2020; Lim et al. 2020b).

Arctic coastal communities face ongoing challenges with erosion (Lamoureux et al. 2015; Liew et al. 2020), with negative impacts on the Inuit way of life (Irrgang et al. 2019; Brady and Leichenko 2020). Progressive and accelerating erosion (averaging <1 m/year but >2 m/year in many areas and locally up to 10 times more) is widespread across the Beaufort Sea coast from Barrow to Tuktoyaktuk (Solomon 2005; Jones et al. 2009, 2018; Lantuit et al. 2012; Gibbs et al. 2019; Obu et al. 2017) and rates of coastal retreat in Kugmallit Bay, west and north of Tuktoyaktuk, are no exception (Fig. 1).

The community of Tuktoyaktuk has for many years faced severe threats from coastal erosion and thawing permafrost, exacerbated by climate-driven change, such that present shore protection measures are not working and the shoreline continues to recede (Fig. 2; Wolfe et al. 1998; Lamoureux et al. 2015). This has led directly to the destruction or relocation of homes on the waterfront. Hamlet officials and other residents have repeatedly sought action to stabilise the seaward coast of the peninsula; several alternative adaptation measures are currently being considered (Baird 2019).

One of the most important pieces of coastline in the community is Tuktoyaktuk Island, which serves as a natural breakwater between Tuktoyaktuk Harbour and the Beaufort Sea. Its northwesterly exposure contributes to a high baseline rate of erosion, exacerbated by the impacts of storm waves and surges that can cause up to 10–15 m of erosion in a single event (Public Works Canada 1971; Rampton and Bouchard 1975; Solomon and Covill 1995). In the face of this inexorable erosion, community residents would like to see action taken to stabilise Tuktoyaktuk Island. If the island were to erode away entirely, the harbour would become vulnerable to increased wave action. This would limit both local use of the harbour for traditional pursuits and larger-scale sealift and other economic development opportunities, which were a driver for the recently completed Inuvik–Tuktoyaktuk highway (Lamoureux et al. 2015).

A better understanding of the short- and long-term sediment exchange dynamics of Tuktoyaktuk Island in this rapidly changing environment is critical to developing appropriate adaptation strategies for the community. Here we present the first comprehensive assessment of Tuktoyaktuk Island, using newly collected information from in situ sampling and ground surveys, combined with high-resolution photogrammetry using remotely piloted aircraft systems (RPAS) often referred to as uncrewed aerial vehicles (UAVs) or drones.

The objectives of the paper are (1) to provide the first volumetric assessment of coastal erosion at Tuktoyaktuk Island, (2) to evaluate the mechanisms that drive erosion across the island, and (3) to consider the implications of ongoing coastal erosion at current or higher rates on the long-term survival of the island and its sheltering of the harbour and community. Our results contribute to knowledge on erosional processes and vulnerability of Tuktoyaktuk Island, which is needed for the future viability of Tuktoyaktuk as a sustainable and resilient maritime community and port facility.

The hamlet of Tuktoyaktuk (2021 population of 937, Statistics Canada 2022) is located on the southeastern Beaufort Sea coast, facing onto Kugmallit Bay east of the Mackenzie Delta, in the Northwest Territories, western Arctic Canada (Fig. 1). Tuktoyaktuk lies within the land claim boundary of the Inuvialuit people, under terms of the 1984 Inuvialuit Final Agreement.

The community of Tuktoyaktuk was established around a natural harbour that is sheltered from the open ocean by Tuktoyaktuk Island. The harbour originated as a river valley initiated by proglacial meltwater following the Last Glacial Maximum, draining to a lower relative sea level (RSL), and was later flooded during the postglacial and ongoing RSL rise (Rampton 1988; Forbes et al. 2014). The rate of marine transgression has varied with the rate at which sea level has risen over the last 8000 years (Hill et al. 1993). The mean rate of RSL rise measured at the Tuktoyaktuk tide gauge (1961–1997) was 3.5 ± 1.1 mm/year (Manson and Solomon 2007). We return to a consideration of accelerated sea level rise under climate change later in the paper.

The surficial geology of the Tuktoyaktuk region is dominated by glaciofluvial terrace and lacustrine plain facies with numerous interspersed lake and organic (peat) deposits (Rampton and Bouchard 1975). The coastline surrounding the community of Tuktoyaktuk consists of a diverse landscape of low sand spits, ice-rich tundra cliffs (1–10 m in height), breached and drowned thermokarst lakes, as well as anthropogenic shore protection in the community. The shoreface is shallow and irregular, an eroding permafrost seabed with local depressions marking the sites of former thermokarst lake basins (Forbes et al. 2014). The shoreface slope is very low as a result of the rapid shoreline recession and the shallow offshore prodelta seabed of Kugmallit Bay (Fig. 1). The foreshore is predominantly sand and pebble beach material, often littered with driftwood storm deposits (Forbes and Frobel 1985; Harper et al. 1988). Beach berms and washover deposits indicate active storm remobilization and deposition. The majority of the beach material fronts low tundra cliffs and ranges from 10 to 25 m wide; however, the immediate area around Tuktoyaktuk also has several significant spit complexes (at Pingo Canadian Landmark, South Spit, North Spit) (Fig. 2). Typically sand spits are wide and low with beach crests less than 1 m above mean sea level (MSL) and are subject to frequent washover events (Forbes and Frobel 1985; Harper 1990; Héquette et al. 2001; Forbes et al. 2014). Tuktoyaktuk Island is approximately 1550 m long, with varying width from 36 m at its current narrowest point to 450 m on the western flank. Rampton and Bouchard (1975) used field observations and airphoto interpretations to characterise the island as a combination of lacustrine plain, glaciofluvial terrace, and littoral deposits.

Tuktoyaktuk is highly vulnerable to storm surges. The tidal range is minimal, with a spring (mean) range of 0.7 m (0.3 m) (Canadian Hydrographic Service 2022). However, a large proportion of the variance in water level is driven by winds and atmospheric pressure. Storm surges up to 2 m or more above MSL have been documented, with severe impacts. The record storm of September 1970 caused 13 m of shoreline retreat at Tuktoyaktuk, widespread flooding, and loss of life (Public Works Canada 1971). Over the past two decades, there has been an increase in storminess, defined as the frequency and duration of high winds >36 km/h (Scharffenberg et al. 2020; Berry et al. 2021), many over open water, producing storm waves that attack the coast.

According to Solomon (2005), the mean coastal erosion rate for the entire Mackenzie Delta region of the Beaufort Sea (including Tuktoyaktuk Peninsula) was 0.61 m/year for the period 1974–2000. More recently, O’Rourke (2017) calculated a mean erosion rate of 0.64 m/year (1950–2004) for the region of Kugmallit Bay. Using the digital shoreline data presented by Solomon (2005) and Hynes et al. (2016), the mean coastal erosion rate for the southeast corner of Kugmallit Bay (approximately 5 km on either side of Tuktoyaktuk) was calculated to be 0.34 m/year for 1974–2000 (Fig. 1). There are only two areas in Tuktoyaktuk that are eroding at rates greater than 1.0 m/year: a short section of the mainland shore between North and South spits (where houses were removed in 2020) and the outer shore of Tuktoyaktuk Island (Fig. 2).

The detailed coastal retreat assessment for the island focused on the rapid retreat of shoreline and (or) cliff-line vectors over recent decades. This is consistent with other methods used for the Mackenzie–Beaufort region (Solomon 2005; O’Rourke 2017; Lim et al. 2020a, Clark et al. 2021; Berry et al. 2021). We derived cliff-edge and shore positions from historical air photos, combined with satellite imagery, high-precision differential GPS surveys, airborne Light Detection and Ranging (LiDAR), and more recent images from RPAS photogrammetry to provide the most complete coverage to date for an assessment of coastal change on Tuktoyaktuk Island. Shore- and cliff-line positions were digitised using a total of 19 air photos available between the years 1947 and 2004. After rectification, the root mean square error of images was less than 5 m with a pixel resolution of 2.5 m. Sources of positional uncertainty include the geospatial orientation and resolution of the air photo. The digital resolution combined with the lighting conditions at the time of acquisition limited our ability to accurately identify the shore- and cliff-line positions in some instances. These data were used as part of the coastal assessment presented by Solomon (2005) and are available as part of a more recent GIS data compilation (Hynes et al. 2016).

Cliff positions after 2000 were derived from the following satellite image acquisitions: Quickbird (2007), Worldview 2 (2008), and GeoEye (2010) with positional accuracies of <1 m and spatial resolution between 0.5 and 1 m. Real-time kinematic GPS ground surveys have been conducted periodically in Tuktoyaktuk to help validate imagery, but also to record accurate cliff-line positions. Cliff-line positions acquired from differential GPS surveys were used from the years 2003, 2004, 2012, 2014, and 2016. These surveys were tied to the Canadian Geodetic Survey monument M039008 located in Tuktoyaktuk, with precision of less than 0.05 m. Lastly, four digital elevation models (DEMs) have been used in this study to help determine cliff-line position, a 2004 DEM (1 m grid size) created from airborne LiDAR data, a 2016 DEM, 2017 DEM, and a 2018 DEM (0.5 m grid size) all created from RPAS imagery. Accuracies of the DEMs were tied to the geodetic benchmark and were deemed to be less than 0.15 m.

A linear regression analysis was performed across the historical cliff line vectors (1947–2017). Using ArcGIS and the Analyzing Moving Boundaries Using R (AMBUR) toolbox, shore-normal transect lines were created at 10 m spacing around the island to measure the distance and rates of shoreline change between all of the intersecting lines (Fig. 3). Distance between the intersect points along each transect was calculated and regressed against the subsequent years. The overall rate of change was established from the slope of linear regression.

The DEMs used to calculate the volume change were derived from four separate data sources: 2004 LiDAR, 2016 RPAS photogrammetry, and 2018 (3 August and 7 August) RPAS photogrammetry. The 2004 DEM was derived from a LiDAR point cloud and represented the ground surface elevation data. Data were gridded using a natural neighbour gridding algorithm to 1 m pixel size. Classification inaccuracies of the LiDAR ground data were minimal due to the sparse vegetation coverage and density along the cliff edge. The horizontal and vertical accuracy based on the World Geodetic System 1984 (WGS-84) horizontal datum and the Canadian Geodetic Vertical Datum 1928 (CVGD28) was 0.15–0.30 m. The value of MSL referred to throughout the paper is referenced to the projected value on the CVGD28 vertical datum. At Tuktoyaktuk, this older datum is 5.2 cm below the new geodetic vertical datum CGVD2013 (HTv2.0, NAD83, epoch 2010.0; Canadian Geodetic Survey 2021a).

DEMs were created by stereo-matching images across the island using the “structure from motion” algorithm in Pix4D software. This technique was applied to over 7200 aerial photographs in 2016, 6500 photographs in 2017, and 6000 photographs in 2018, to generate high-resolution orthomosaics of the land surface. The aerial image (RPAS) mosaics were processed using 15–20 ground control targets placed across the island to create the high-precision (<0.10 m) DEM. For this study, the raster grid was resampled at 1 m to match and allow for comparison with the LiDAR data set.

The horizontal and vertical accuracy of both systems and methodologies incorporated long-term geodetic measurements and ground control points based on the Canadian Geodetic Survey (2021b) Active Control System monument located in Tuktoyaktuk (M039008). GPS beach profiles were acquired as validation for the LiDAR and RPAS surveys and results were within the required specification (±0.15 m for LiDAR and ±0.10 m for RPAS).

Volume change was computed in ArcGIS by differencing the elevation values of each cell across the overlapping DEMs. Positive or negative values (m³) represent volume gain or loss, respectively. The measured volume loss along the cliff face was adjusted for ice melt using the observed excess ice percentage by stratigraphic unit.

It is challenging to quantify the volume of large ice bodies, such as ice wedges, which intersect the cliff but are rarely fully exposed. Ice wedges are common across the island; however, the lack of continuous surface expression visible from aerial imagery along the clifftop makes it difficult to determine the location and orientation of wedges. Despite the lack of visible surface expression, ice wedges have been observed sporadically in the upper parts of newly eroded cliff faces. From these observations, the nominal depth and width ranges representative of the typical morphology of ice wedges on the island were determined.

In this study, the future evolution of Tuktoyaktuk Island is projected by extrapolation of the accelerated shore recession rates observed since 2000. Future shoreline positions are determined by extrapolation of recession rates transect by transect, continued to the point of island breaching.

We use the mean rate at each transect for 2000–2018. Along the northwest-facing cliff, the mean rate was 1.80 ± 0.02 m/year (Table 1), a 14% increase over the 1947–2000 mean recession rate and 22% greater than for the 15 years preceding (1985–2000). On the southeast-facing shore (harbour shore), where rates were much lower, the 2000–2018 rate was 2.8 times faster than 1947–2000 and showed a five-fold increase over 1985–2000 (Table 1).

In the absence of a mechanistic model and projections of the main erosion drivers, which are expected to intensify, we consider the implications of further acceleration in the rate of recession using a simple one-line computation with combined outer and inner shore retreat rates at the point of breaching, resulting in shorter times to breaching as a linear function of erosion rates. As the rates increase, the remaining time contracts to the point of limiting further acceleration, while also narrowing the island to the point at which an individual major storm could accelerate the removal of the final link.

The western side of the island is characterised by low‐lying ice wedge-polygons within beach material (Fig. 4). Adjacent to this segment is a low tundra area elevated no more than 2.5 m above MSL. The main portion of the island starts where the low tundra slopes upward to the upland tundra terrace, at an elevation of approximately 10 m (above MSL), which extends eastward 610 m. It is this region where most of the erosion occurs. East of the upland tundra area, the elevation drops at a drained lake and rises again to upland tundra before dropping off to the beach at the eastern end of the island (Fig. 4). The long axis of the island is parallel to the outer shore, facing northwest, the modal orientation of winds and waves in the largest storm events (Manson and Solomon 2007). The northwest-facing shore forms a ∼10 m high cliff in ice-bonded sediments (Fig. 5). Erosion occurs by thermo-abrasional undercutting during storm-surge events, sometimes aided by exposure and melting of ice wedges, leading to block toppling (Fig. 6B). The landward side of the island, being much more protected, has a lower backshore slope and erosion rates are 20% or less of those on the outer shore (see below).

The stratigraphy of the northwest-facing cliff face is represented in Fig. 5. For the most part, the cliff face is almost vertical, but is typically buried by slump material (Fig. 6D). On rare occasions after storms, the base of the cliff may become exposed and accessible for examination and sampling. Two retrogressive thaw flow (RTF) slumps 60–100 m wide are also present (Fig. 5), providing a conduit for thawed material and meltwater to move downslope. Sediments comprising the exposed cliff face on Tuktoyaktuk Island were mapped in detail, allowing for the delineation of five lithological units (Fig. 5).

On a textural basis, the stratigraphy of the island includes silty clay, sandy silt, laminated sand and pebbles, massive sand, and a cap of peat (Fig. 5). The basal unit is massive ice at the western end and ice-bonded clayey silt toward the east. The massive ice body exposed at the base of the cliff during storm events is up to 2 m thick (Fig. 6D) and is shown here to extend over 800 m across the front of the cliff (Fig. 5). The presence and lateral depth into the cliff of the massive ice has been demonstrated by Lapham et al. (2020), with drill core acquired in 2018. This unit has been assigned an ice content value of 100% for sediment computations. The overlying clayey-silt unit is 2.0–2.5 m thick and contains an abundance of ice bands and veins (Fig. 6C). Sampling of this unit revealed excess ice content ranging from 45% to 66% (Table 2). At the top of the clayey-silt unit, there is an erosional unconformity, overlain by stratified sandy silt, 4.0–4.5 m thick, with 27%–50% excess ice. Two coarse sand units (grain size 0.5–1.0 mm) occur above the sandy silt unit. Both sand units are in the range of 29%–37% excess ice. The lower sand unit, 2.0–3.1 m thick, contains an abundance of pebbles and cobbles that form parallel and subparallel laminations. The laminated sand and pebble unit is overlain by a massive sand unit (0.6–0.9 m thick) that provides the base for the thin peat unit at the surface, a ubiquitous feature of the landscape in this region (Rampton 1988). The peat is up to 1 m thick and contains black fibrous peat and root and wood fragments. This unit is considered the active layer of seasonal thaw and, for the purpose of this assessment, we consider it to be almost ice-free (5%), as it does contain small amounts of ice (up to 29% measured at base of unit) and is intersected by ice wedges. O’Neill et al. (2019) documented rates of 0.2–0.8 cm/year subsidence due to deeper seasonal thaw at 10 sites across the region, with loss of 5–20 cm of shallow excess ground ice in the last 25 years. Given that the island has eroded back 40–50 m in that time period, the apparent volume loss due to thaw subsidence would be a very small proportion.

Field observations indicate that ice wedges may be up to 1 m wide at the top and extend at least 1.5 m into the substrate (Fig. 6A). The true three-dimensional shape and distribution of the ice wedges on the island is hard to verify without the use of geophysical surveys and detailed ground truth data along the top. Based on our observations and high-resolution satellite imagery, it is clear that thermal cracking of the terrace surface in the upland tundra area can occur in irregular shapes up to 25 m in diameter. Combining the general dimensions of ice wedge features observed along the cliff face, their contribution to the total volume of cliff material being eroded on an annual basis is estimated to be 5% (Table 3).

The mean rate of coastal recession on the island over the entire record 1947–2018 was 0.80 ± 0.05 m/year (Table 1). Eighty percent of transects experienced net erosion, 5% deposition, and the remaining 15% a minor change (within the error of measurement). The northwest-facing cliff exposed to Kugmallit Bay exhibited the highest rate of change (1.63 ± 0.04 m/year), while the southeast cliff (facing the harbour) retreated at a mean rate of 0.26 ± 0.04 m/year during this time period (Table 1).

This study provides the first comparison of shoreline change rates over four quasi-equal time periods (1947–1966, 1966–1985, 1985–2000, 2000–2018) for both the front and back of the island over the 71 years 1947–2018 (Table 1). Mean erosion rates prior to 2000 on an island-wide basis were 0.77 ± 0.07 m/year (1947–1966), 0.63 ± 0.06 m/year (1966–1985) and 0.71 ± 0.06 m/year (1985–2000). The overall mean was 0.69 ± 0.06 m/year from 1947 to 2000, after which it increased to 1.12 ± 0.07 m/year (2000–2018), a 62% jump in recession rate. Since 2000, 93% of the island transects show net erosion, leaving only small pockets of accretion highlighted by the growth of the eastern spit (Fig. 7).

Only the front of the island (northwest-facing cliff) has experienced consistent erosion across the whole section since 1947. Other regions along the eastern spit and northwest complex of ice wedge polygons and beach washover show a less consistent pattern of change (Fig. 7). The mean erosion rate along the northwest-facing cliff for the entire 71 years (1947–2018) was calculated to be 1.63 ± 0.04 m/year. Prior to 2000, annual mean rates were 1.59 ± 0.04 m (1947–1963), 1.64 ± 0.06 m (1963–1985), and 1.47 ± 0.04 m (1985–2000), giving a multidecadal mean of 1.58 ± 0.05 m/year (1947–2000). This increased to 1.80 ± 0.02 m/year after 2000, a 14% increase (Table 1).

Coastal change rates along the southeast-facing shore remained low throughout, aside from two or three sections where erosion rates have increased since 2000 (Fig. 8). Very little shoreline change has occurred in the southwest section of low tundra aside from a few instances in the 1947–1966 and 2000–2018 time periods (Fig. 8).

The dominant and consistent erosion along the northwest-facing cliff has resulted in the island maintaining its relative shape while decreasing in width. An exception to this process is the breach of the southern upland tundra peninsula located on the back side of the island (Fig. 7). The breach of the narrow peninsula was created between 2016 and 2018 and created a small island that is no longer connected to the main island. The removal of this region reduces the island’s footprint and is a contributor to the marked recent acceleration of area loss seen in Fig. 9.

A comparison of elevations between 2004 and 2018 along the northwest-facing cliff can be seen in cross-shore profiles (Fig. 10). Each profile represents the typical erosion pattern along the northwest-facing cliff across the three physiographic units (1, lowland tundra; 2, upland tundra terrace; 3, lacustrine basin; Fig. 3). Profiles 1 and 2 exhibit similar and consistent patterns of elevation and slope change across the timespan. Also, the recession along the water line (beachface) is similar to rate at the cliff base and cliff top. The lowland tundra region (Profile 1) has lost 14–15 m of material at the water line and cliff top. In the upland region (Profile 2), there is a 20–21 m loss across the beach and cliff top. Profile 3 highlights the non-uniform retreat and elevation change at the lacustrine basin transect to the west. In this area, there is 14–15 m of loss along the beach front and a change in the cliff top height due to the slope away from the ocean.

A comparison of elevations on Tuktoyaktuk Island from digital surface models between 2004 and 2018 shows that erosion accounted for 88% of the total island volume change contributing 9200 m³ of sediment to the ocean on an annual basis. Sediment accumulation accounted for the remaining 12% of the island change during that time period (see areas of increased island volume in Fig. 11.

A one-year comparison of elevations from 2017 and 2018 DEMs reveal similar results, with 8900 m³ of material removed by erosion, accounting for 69% of the total island volume change. The remaining 31% results from shoreline accumulation along the eastern sand spit complex and beach over-wash (more than likely driftwood) in the northwest complex. It is also clear that the erosion along the northwest-facing cliff face accounts for the majority of the volume change (7772 m³ per year).

Table 3 is a comparison of the area and volume change for each stratigraphic unit along the northwest-facing cliff face. The calculated sediment contribution for each unit takes into account the mean excess ice content. The volume of eroded material assumes the exposure areas of the stratigraphic units on the cliff face have remained uniform over time. Using the long-term average erosion rate, the island has exported 6107 m³ of sediment into the ocean on an annual basis for the past 70 years. From 1947–2000 the island lost on average 5663 m³/year. This number has increased to 6663 m³/year since 2000 (Table 3).

The acceleration in erosion of Tuktoyaktuk Island is driven by both long-term increases in air and water temperature and short-term forcing events that are increasing in intensity and duration (O’Rourke 2017; Lim et al. 2020a; Clark et al. 2021; Berry et al. 2021; Scharffenberg et al. 2020) during prolonged open-water seasons. However, it should be noted that these drivers of change are dramatically heightened during periods of increased water levels from storm surge and relative sea-level rise (Atkinson et al. 2016). We expect the island will continue to erode at the same or increased rates in the coming years as experienced by other sites in the region. Berry et al. (2021) and Scharffenberg et al. (2020) point out that the region has experienced an increase in the duration of summer storms during this time period. There is a direct correlation between the increased duration of storms and coastal erosion along northwest-facing cliffs (Berry et al. 2021). In one storm (August 2000) the island has receded up to 15 m (Solomon 2005), while Rampton and Bouchard (1975) reported that portions of the community at Flagpole Point (now known as North Spit) had eroded up to 45 feet (13.7 m) during a 1971 storm. Comparisons of the island position and morphology before and after a storm event in 2018 reveal a considerable change and loss of sediment at the base of the cliff (Fig. 12). This is further evidence that erosion along the northwest-facing cliff of Tuktoyaktuk Island is driven and exacerbated by coastal storms.

Increasing temperatures and diminished sea ice extent and duration, both of which promote more rapid thermal and mechanical abrasion of ice-rich coasts, are increasingly well documented in the region (e.g., Berry et al. 2021). Arctic amplification of temperatures emerged in the 1970s, and will drive more rapid acceleration of thermal abrasion, which may account for a 10%–20% increase in Arctic-mean erosion rates by 2100, according to semi-empirical modelling by Nielsen et al. (2022). Combined with more extensive open water and increased wave energy, this study projected continuing acceleration of Arctic coastal erosion through the 21st century, without accounting for relative sea-level rise. These authors did acknowledge that higher sea level will increase the effectiveness of thermo-abrasion processes.

Tuktoyaktuk lies in a region near the margin of the Laurentide Ice Sheet at the Last Glacial Maximum, and is subject to long-standing glacial-isostatic subsidence (forebulge collapse), with measured vertical motion of −1.50 ± 0.68 mm/year (ITRF2005); Canadian Geodetic Survey 2021b), consistent with the newly defined NAD83v70VG Canadian crustal velocity model (Robin et al. 2020). This is almost half of the measured RSL rise at Tuktoyaktuk (1961–1997) (Manson and Solomon 2007), suggesting that RSL is now rising more rapidly.

Projections of relative sea-level rise for the coming century at this site are available from data sets provided in James et al. (2021). Based on IPCC AR5 projections for global MSL and incorporating crustal motion from the NAD83v70VG model, these provide localised RSL projections under two global greenhouse gas emission scenarios defined as Representative Concentration Pathways RCP4.5 and RCP8.5 (IPCC 2013). Using median values, these project a sea-level rise at Tuktoyaktuk by 2100 amounting to 0.526 m (RCP4.5) or 0.721 m (RCP8.5) relative to the mean of 1986–2005 water level. Moving forward from 2020, there is little difference between scenarios prior to 2050, with projected median RSL rise of 17–20 cm (Fig. 13). This shows a clear acceleration from 2006 to 2020 and a current rate of RSL rise (2020–2030) of 7.15 ± 1.1 mm/year, double the late 20th century rate.

Thermal niche formation is promoted by storm surge flooding of the cliff base, during which slump material is removed and warm seawater is in direct contact with ice-bonded sediment (Kobayashi 1985). Rising sea levels not only raise the level of storm-surge flooding, but increase the frequency of extreme high water levels (Lamoureux et al. 2015). Manson and Solomon (2007), using a realistic RSL projection, showed that a storm surge of 2.65 m Chart Datum (2.61 m MWL) at Tuktoyaktuk in 2000 (taken to be a 100-year return event) would have a return period of <25 years by 2050 and <5 years by 2100. Combined with other evidence for increased storm-surge frequency with diminished sea ice (e.g., Vermaire et al. 2013; Lim et al. 2020a), it is clear that RSL will exacerbate current rates of cliff recession.

Tuktoyaktuk Island contains three types of ground ice: massive segregated ice, interstitial ice, and ice wedges, all of which influence both the total volume of the island and the total volume of material being eroded into the water column and deposited onto the seabed.

The massive ice body located at the base of the cliff (Fig. 5 and picture in Fig. 6) contains horizontal banding, suggesting it is epigenetic segregation ice formed in situ, the same origin as interpreted for similar banded massive ice at nearby coastal exposures of the Pingo Canadian Landmark (Mackay 1989; Mackay and Dallimore 1992). This would be consistent with a downward freezing front beneath the low-permeability fine-grained clayey-silt overlying unit (Fig. 5). This is 2.0–2.5 m thick (Table 2), with an abundance of excess ice. A drill hole landward of the cliff top encountered the massive ice body ∼8.2–9.4 m below the upper terrace surface (Lapham et al. 2020), consistent with the cliff-face exposure (Fig. 5).

Other massive ice bodies underlie the mainland peninsula in the hamlet nearby. Almost 4 m of ice observed in the walls of an excavated ice cellar contain <20% sand in narrow stringers separated by almost pure ice This shows recumbent folding, suggesting deformation by overriding glacier ice (Rampton and Mackay 1971). A number of boreholes within the hamlet and along the west-facing outer shore also show extensive massive buried ice beneath a thin cover of sand and gravel (Wolfe et al. 1998).

The sandy-silt stratigraphic unit above the clayey silt at Tuktoyaktuk Island is ∼4.0–4.5 m thick and accounts for over 40% of the island volume (Fig. 5). This and the underlying clayey silt contain 27%–66% interstitial ice in lenses and veins (Table 2).

Ice wedges form in cracks that develop as a result of seasonal thermal contraction. They tend to extend only as deep as the strength of material will allow. Cracks often terminate abruptly just below stratigraphic boundaries (Lachenbruch 1963). Therefore, we can assume that the ice-wedges on Tuktoyaktuk Island extend down to the upper contact of the sandy-silt unit (∼3.3 to <5 m below the surface, Table 2). Visual observations of the cliff face revealed ice-wedge spacing (rudimentary polygon dimensions) of ∼25 m. For this reason, we have estimated the ice-wedge volume in the upper part of the section to be 5%. This number is slightly less than values reported by Pollard and French (1980) and Murton (2005), but similar to ice-wedge volume percentages reported by Dallimore et al. (1996) on nearby Richards Islands in an area of similar stratigraphy. The large volume of massive segregated and interstitial ice along this coast makes it increasingly vulnerable to a warming climate.

Not only is the excess ice content an important contributor to the erosion of the cliffs, it is also an important consideration when calculating the volume of material being eroded on an annual basis. By assessing the volume and composition of sediments eroded from the cliff face, it is possible to determine the contribution of sediments to the nearshore region on an annual basis (Table 3). These new data have implications for understanding the changing nearshore sediment dynamics, which affect navigation channels and influence the formation of shorefast ice, as well as for quantifying carbon and contaminant release. The volume of sediment eroded from Tuktoyaktuk Island is the dominant source of new sediment influx to the nearshore system at the harbour mouth. Although harder to quantify, the Mackenzie Delta plume, coastal runoff from storm surges, and other remote sources of sediment transported through longshore drift also contribute to the sediment budget.

The front of Tuktoyaktuk Island lost a volume of approximately 5663 m³/year of sediment and ice between 1947 and 2000, and an additional 6662 m³/year from 2000 onwards (Table 3). Of this volume, 34% is attributed to ice (lost through meltwater discharge) and 66% to sediment. Of the eroded sediment, 16% is clay, 40% silt, and 44% sand. The upper portion of the island is predominantly sand and accounts for most of the volume of eroded sediments and the bulk of the sediment transported alongshore to feed the eastern spit and infill the harbour-entrance tidal channel (Forbes et al. 2014).

The new understanding of the stratigraphy and volume change of the island presented here enables an improved assessment of the impacts of the sediment influx to the nearshore region. The coarse material (sand) will fall out of suspension rapidly and be dispersed in the shore zone. The 46% of finer material (silt and clay) will remain in suspension and be transported further alongshore, offshore, or into Tuktoyaktuk Harbour. Nearshore sedimentation of the already shallow bathymetry (Forbes et al. 2014; Boike and Dallimore 2019) leading into Tuktoyaktuk Harbour already blocks access for vessels with larger draft.

It has long been documented that coastal erosion is a key driver of organic material to the Arctic Ocean (Macdonald et al. 2015; Wegner et al. 2015). Lapham et al. (2020), working on Tuktoyaktuk Island, provide information on the potential for frozen material to produce greenhouse gas emissions prior to erosional exposure and thaw. However, the direct biogeochemical impacts of carbon flux and transport to marine environments (Bruhn et al. 2021; Tanski et al. 2021) in an accelerated erosional landscape, such as Tuktoyaktuk Island, are harder to quantify. This study provides the necessary baseline information for future studies on the release of organic material entering the ocean and atmosphere, a globally significant process occurring across eroding Arctic coastlines.

As of 2018, the width of the island was reduced to 36 and 39 m at its two most narrow points (eastern and middle portions of the upland tundra terrace, respectively). This raises the question about how long it will be before the island is breached and divided, creating an opening and potential exposure for the harbour. We have projected future evolution of the shoreline (technically the cliff line) by extrapolation using the post-2000 accelerated recession rate, on a transect-by-transect basis (Fig. 14). This figure also shows the alongshore distribution of erosion rates from 2000 to 2018. As noted earlier, these are accelerated relative to pre-2000 recession rates by 14% along the outer (northwest-facing) cliff and a factor of 2.8 along the harbour shore (“SE cliff” in Table 1).

The projected loss of land in the upland tundra region is dominated by continuing retreat along the northwest-facing cliff on the seaward side of the island at the mean rate of 1.8 m/year (with alongshore variability as shown for individual transects in Fig. 8 and more generally in Fig. 14). In addition, since 2000, the southeast-facing harbour side of the island has also started to erode at a much more rapid rate, particularly in two “hotspots”, near the eastern point and at the west end of the narrow portion of the island. Progressive retreat of both shores at their local recession rates brings the outer and inner cliff lines together in 2044 (Fig. 14). The point of convergence is not either of the current narrowest locations, but a location that today is slightly wider (57 m) but exposed to more rapid erosion, particularly on the harbour side.

This linear extrapolation of the current recession rates is obviously a very simple approach. Although it uses much higher rates than the long-term (71-year) mean, thereby accounting for enhanced drivers of erosion, it likely underestimates the mean erosion rates over the next 26 years from 2018. To estimate the sensitivity of the time to breaching to varying rates of acceleration, we consider a cross-island transect at the point of breaching and apply linear front and back erosion rates increased beyond the rates employed above. Using those rates, our “one-line” model reaches convergence and breaching in 25 years (just a year short of the more sophisticated analysis in Fig. 14). Increasing the rates as of 2018 by a further 14% on the outer shore, to a value of 2.11 m/year and a very modest 50% on the inner shore, to 0.72 m/year, leads to convergence and breaching in 20 years (2038). Increasing the outer shore erosion by 50% and doubling the inner shore rate gives a time to breaching of 16 years (2034).

Given observations of very dramatic increases in erosion rates on shores exposed to similar drivers and eroding primarily in the same way (e.g., Jones et al. 2018), the 2034 scenario is not implausible. We suspect that the high sand content in Tuktoyaktuk Island, the shallow bathymetry of Kugmallit Bay (Fig. 1), limiting wave height, and possibly less extreme change in open-water fetch compared with the western Beaufort coast in Alaska (which was at one time a choke-point for vessels transiting the coast between the Beaufort and Chukchi seas), may limit the rate of acceleration over the coming 25 years. However, a doubling of relative sea-level rise to >7 mm/year (Fig. 13) will reduce the depth limitation of wave energy in the bay off Tuktoyaktuk and cause more frequent inundation of the cliff base massive ice exposure, potentially pushing recession rates higher.

On balance, we believe that the rates of recession will continue to rise, with some variability resulting from year-to-year fluctuations in the drivers, but that breaching within 15 years is unlikely. Without intervention, it will occur before mid-century and it will be important to continue monitoring the site to detect any changes in erosion trends.

The western portion of the island located in the lowland tundra and low-lying ice-wedge polygons is projected to continue to erode back, but will largely remain intact. Sediment deposition along the spit located in the eastern end of the island will continue. This is the one area of the island that may increase in size in the coming years.

The first breach point of the island is located at the transition between the lower tundra and the upland tundra terrace. Expansion of the breach is expected to follow, widening the opening, developing either a third inlet or a wide sand flat subject to high-energy overwash in storms. The future evolution is difficult to predict, as opening of the island beach may affect the tidal flow and stability of the present harbour entrance.

Given the short time-frame identified in this paper, this site (with its wealth of historical data) is a prime candidate for more detailed, process-based modelling with appropriate drivers, to clarify the relative importance of temperatures, water levels, and wave energy and to refine the estimate of the likely time of breaching.

This study documents 71 years of long-term progressive erosion of an ice-rich permafrost coast at Tuktoyaktuk Island. It identifies the importance of lithostratigraphic data (sediment properties and excess ice content) for an evaluation of erosion processes, rates of volume loss, and sediment delivery to the marine environment. It is clear that drivers of coastal erosion, including temperatures, open-water season, open-water fetch, wave energy, and storm-surge frequency are increasing in this region, leading to acceleration of coastal erosion as much as 20 years ago. At the same time, RSL projections indicate that the rate of RSL rise this decade is more than double the late 20th century rate.

Detailed aerial imagery, positional surveys, and a sampling program have led to the following insights: (1) Erosion of the exposed northwestern shore has accelerated over the last two decades (from 1.58 ± 0.05 m/year to 1.80 ± 0.02 m/year). (2) The island lithology consists of 44% sand, 40% silt, and 16% clay overlying a massive ice body. (3) Excess ice content accounts for ∼37% of the island material and estimates of sediment delivery from the island need to adjust for the loss through ice melt. (4) Erosion accounts for 88% of the total island volume change (9200 m³/year), with the majority taking place along the northwest-facing cliff. (5) At the current rate, Tuktoyaktuk Island will be breached by 2044 (and likely sooner). The loss of this important natural barrier will not only increase wave agitation inside the harbour, but also may lead to increased harbour shore erosion and sedimentation affecting marine eco-systems and navigation.

The authors declare there are no competing interests associated with this work.

Data generated or analyzed during this study are available from the corresponding author upon request.

The authors acknowledge the financial support of the Climate Change Geoscience Program (Natural Resources Canada), the NERC Arctic Office UK–Canada Bursary scheme, CIRNAC BRSEA, CIRNAC CCPN, the Polar Continental Shelf Program (Natural Resources Canada), and Transport Canada (Northern Transportation Adaptation Initiative: NTAI). We are grateful to the field crews, community members, and NRCan staff who worked on this project during the 2013–2018 field surveys, in particular Jeremy Bentley, Angus Robertson, Patrick Potter, and Eric Patton for their contributions to the field logistics, data collection, and processing. We also acknowledge the community of Tuktoyaktuk, the Tuktoyaktuk Community Corporation, and the Tuktoyaktuk Hunters and Trappers Committee for their continued support. Permitting for this project fell under the NWT Science Licences 15607, 15915, 16073, and 16286. Inuvialuit land administration land use permit: ILA18TN005. This is Natural Resources Canada contribution 20210176 (Canadian Crown copyright reserved).