Accelerating Thermokarst Transforms Ice-Cored Terrain Triggering a Downstream Cascade to the Ocean

Abstract

Recent climate warming has activated the melt-out of relict massive ice in permafrost-preserved moraines throughout the western Canadian Arctic. This ice that has persisted since the last glaciation, buried beneath as little as 1 m of overburden, is now undergoing accelerated permafrost degradation and thermokarst. Here we document recent and intensifying thermokarst activity on eastern Banks Island that has increased the fluvial transport of sediments and solutes to the ocean. Isotopic evidence demonstrates that a major contribution to discharge is melt of relict ground ice, resulting in a significant hydrological input from thermokarst augmenting summer runoff. Accelerated thermokarst is transforming the landscape and the summer hydrological regime and altering the timing of terrestrial to marine and lacustrine transfers over significant areas of the western Canadian Arctic. The intensity of the landscape changes demonstrates that regions of cold, continuous permafrost are undergoing irreversible alteration, unprecedented since deglaciation (~13 cal kyr B.P.).

1 Introduction

Low Arctic circumpolar landscapes underlain by warm ice-rich permafrost are considered highly susceptible to climate change and are undergoing increased thaw and thermokarst degradation (Grosse & Romanovsky, 2011; Lacelle et al., 2015; Lantz & Kokelj, 2008). In the western Canadian Arctic, this recent regional and local intensification of thaw slumping indicates a dramatic equilibrium shift in postglacial landscape evolution (Kokelj et al., 2017). However, the cascade of fluvial, sedimentological, and geochemical inputs from landscapes to lakes and oceans remains poorly understood. Using satellite imagery and historical aerial photographs, we demonstrate the inherent climatic sensitivity of ice-cored permafrost terrain on Banks Island, Canada, an area underlain by cold, continuous permafrost (< −10°C) and highlight the downstream consequences of abrupt and widespread acceleration of thermokarst.

The study area on eastern Banks Island is situated within the Jesse moraine belt, a prominent late glacial (13.75–12.75 cal kyr B.P.) readvance of the Laurentide Ice Sheet (Lakeman & England, 2012) (Figure 1). It onlaps the coastlines of Banks and Victoria islands for 5–60 km along the length of the Prince of Wales Strait (~300 km, > 20,000 km²) and is widely cored by remnant glacial ice (Lakeman & England, 2012) that in places exceeds 25 m thickness (Smith & Farineau, 2015). The landscape is characterized by nested, broad, ice thrust controlled moraines (Evans, 2009) and, most commonly, undulating till plains with prominent orthogonal tundra polygons (Dyke et al., 1992).

Figure 1

Map showing thaw slumps, impacted fluvial networks, and fluvial-ocean linkages. A 2009 SPOT mosaic image within the Jesse moraine belt (see inset), and the regional watershed divide (dashed black line) and Johnson Point catchment (solid black line). RTS (red circles) mapped from the 2009 SPOT mosaic and the 2015 ArcticDEM data show points of slope sediment mobilization with the context of fluvial networks. Highlighted stream segments are fourth order or higher, and entry points to the ocean are identified by yellow and brown triangles. Impacted rivers are classified according to the proportion of upstream RTS with yellow (<10 RTS) and brown (>10 RTS) triangles.

During the past decade there has been a substantial increase in the magnitude and frequency of thermokarst processes in ice-cored terrain and ice-rich sediments, specifically retrogressive thaw slumps (RTS) (Kokelj et al., 2015; Ramage et al., 2017). RTS are dramatic manifestations of permafrost thaw (Figure 2) and have been documented throughout the circumarctic (Bowden et al., 2008; Kokelj et al., 2017; Leibman, 1995). They can be initiated through fluvial or wave erosion (Lantuit et al., 2012), changing permafrost and lake configurations (Kokelj et al., 2009) or exceptional thaw and/or precipitation events (Kokelj et al., 2015; Lacelle et al., 2010), and vary with the geomorphic setting and local climate (Kokelj et al., 2017; Segal et al., 2016). The scale and rapidity of recently observed changes in ice-cored terrain within areas of cold permafrost, formerly considered to be more resilient to climate change (Smith et al., 2010), signifies that these areas are instead particularly susceptible to disruption by recent climatic changes (Rowland et al., 2010; Segal et al., 2016). The regional and circumarctic distribution of RTS suggest ice marginal landscapes have preserved vast stores of relict glacial ice (Kokelj et al., 2017), and the recent climate-driven acceleration of thaw slumping suggests that a geomorphic threshold has been crossed, triggering a cascade of sedimentary and geochemical effects. Within these regions, variability in landscape responses is poorly understood and the impacts to downstream fluvial systems, including coastal ecosystems, remain unexplored.

Figure 2

Retrogressive thaw slumps (RTS) near Johnson Point, Banks Island. (a) Headwall exposure of deformed and sheared glacial ice overlain by <1.5 m of sediment. (b) Headwall exposing buried glacial ice in active RTS, and liquefied slump floor material. Note angular (and striated) glacial erratic boulders released from ice. (c) Oblique aerial view of RTS in Figure 2b showing extent and confluence of individual RTS that have developed since 2009 and connectivity between the disturbed slope and stream. Photographs by S.F. Lamoureux, July 2015.

The recent and abrupt climate-driven alteration of a permafrost-preserved deglacial land system (Dyke & Evans, 2003; Dyke & Savelle, 2000) signals intensification of postglacial landscape adjustment and amplification of the downstream impacts of permafrost thaw (IPCC, 2013; Liljedahl et al., 2016). The thaw of relict glacial ice and massive syngenetic ground ice (formed postglacially in glaciolacustrine and raised glaciomarine deposits) and the hosting sediments generate significant new sediment, solute, and water inputs, fundamentally altering summer hydrological, sedimentary, and solute regimes of streams. Our data indicate that if unabated, ongoing thermokarst will transform the hydrological and fluvial sedimentary regime, amplifying and altering the timing of terrestrial to marine and lacustrine transfers over significant areas of the western Canadian Arctic and elsewhere.

2 Materials and Methods

2.1 Data Sources and Image Preparation

A time series of 1961 stereo black and white aerial photographs, 2008–2009 SPOT imagery, and a 2015 WorldView-2 (WV2; acquired 10 August) satellite image were used to assess changes in the extent of RTS within the Johnson Point watershed study area (Figure 1, hashed area: 230 km²) between 1961 and 2015 (supporting information Table S1). Broader, regional perspectives of RTS and other thermokarst-affected landscape extents were interpreted from SPOT imagery and 2015 ArcticDEM data (Figure 1). Thaw slumps were mapped directly on the aerial photographs using a mirror stereoscope, and then, 600 dpi (~4.2 m pixel resolution) scans of the aerial photographs and linework were georeferenced to the WV2 satellite imagery using ArcGIS v10.2.2. Thaw slumps were individually mapped based on morphological and visual characteristics, including the presence of a distinct headwall and contrast between disturbed and undisturbed material. The identification and delineation of thaw slumps is limited by the pixel resolution of each type of imagery (supporting information Table S1). To account for this, the minimum slump size mapped using SPOT imagery was used as the baseline, and slumps <0.3 ha were not included in the final aerial photograph and WV2 RTS inventories. At the scale of imagery and mapping used, the generally small feature size, small slump headwalls and floors (~2–10 m), and their occurrence on generally low-relief terrain are likely to minimize horizontal/vertical distortions associated with the various orthorectified (SPOT and WV2) and georeferenced (aerial photographs) image types. Nonetheless, discrepancies in data resolution and accuracy are acknowledged.

Tiles from the 1:50,000 Canadian Digital Elevation Data data set were mosaicked to create a digital elevation model (DEM), and stream vectors and lake boundaries were obtained from the National Hydro Network. DEM preprocessing was performed prior to watershed delineation to fill sinks, which are common errors in elevation data where neighboring cells all have higher elevation values leading to inaccurate flow direction. All data sets were converted to a common NAD83 UTM Zone 11N projection. A series of nested watersheds were then delineated using methods outlined in Rudy et al. (2015) for the main river in the catchment that discharges at Johnson Point. A legacy of Landsat images acquired from Landsat 5, Landsat 7 ETM+, and Landsat 8 OLI TIRS were acquired to identify the presence/absence and timing of a summer sediment plume into the ocean (supporting information Table S2).

2.2 Historical Climate and Precipitation Data

Hydroclimatic drivers and RTS activity were compared using trends in air temperature and summer precipitation for the period 1956–2015. Homogenized historical temperature and precipitation records were acquired from climate stations in Mould Bay (Prince Patrick Island, NT, ~380 km north) and Sachs Harbour (Banks Island, NT, ~240 km southwest) (supporting information Figure S1).

2.3 Water Sampling and Analytical Methods

Water samples were collected from narrow (1–5 m wide), shallow (<50 cm deep), ephemeral, braided streams in all second- and third-order watersheds as well as along the main river channel (supporting information Figure S2) and analyzed for total suspended sediments (TSS), total dissolved solids (TDS), and water stable isotopes. TSS and TDS were analyzed at Taiga Laboratories, Yellowknife. Massive ice samples collected from the headwalls of four RTS and rainfall samples were analyzed for stable water isotopes. Within RTS sample sites, at least 10 cm of the ice was removed from the surface to reduce the possible mixing with surface melt and refrozen water. Water samples were collected from surface ponds that formed immediately after a large rainfall event. The ¹⁸O/¹⁶O and D/H ratios of the samples were determined using a liquid water analyzer (Los Gatos Research). All measured water samples were calibrated and normalized using internal laboratory water standards that had been previously calibrated relative to Vienna standard mean ocean water (VSMOW) and Standard light Antarctic Precipitation (SLAP). The results are presented using the standard δ-notation with respect to VSMOW and plotted with respect to the Global Meteoric Water Line (Craig, 1961). Analytical reproducibility for δ¹⁸O and δD is ±0.3‰ and ±1‰, respectively. Using massive ground ice and rainfall as end-members in a two-component mixing model, stream and river samples were separated into their putative components.

3 Results and Discussion

3.1 Accelerating Thermokarst

On eastern Banks Island, both the density of RTS and their disturbed area has substantially increased since 1961 (supporting information Table S3 and Figure 3). Post-2009, the density of thaw slumps increased by 131% with the disturbed area reaching 76 ha/100 km². In contrast, slump density increased by only 35% during the period 1961–2009. This acceleration of disturbance is supported by similar mapping in a region ~100 km farther south on Banks Island (Segal et al., 2016). Expansion rates were estimated for large slumps (>2 ha) that initiated post-2009. The average longitudinal length of these post-2009 RTS in our study was ~200 m which equates to expansion rates of ~30 m/a, considerably higher than maximum rates measured from aerial photographs on southern Banks Island between 1952 and 1962 of 12.2 m/a (Lewkowicz, 1987).

Figure 3

(a) Temporal comparison of the increase in total disturbed area associated with RTS and (b) the increase in the frequency of large RTS in the Johnson Point watershed.

The size distribution and frequency of RTS shows a clear and substantial increase post-2009 (Figure 3). The changes we observe in this watershed are corroborated with the appearance on Landsat imagery of a prominent sediment plume in 2011 where the Johnson Point river enters the ocean, its persistence each year thereafter, and its absence (excepting 1989) between 1986 and 2010 (Figure 4 and supporting information Table S2).

Figure 4

(a) Observations of sediment plumes were more frequent post 2010. (b) Percentage of Landsat images between 1986 and 2015 with plumes sharply increasing in 2011. Historical Landsat images were obtained from the USGS Global Visualization Viewer (glovis.usgs.gov) (supporting information Table S2).

The intensification of RTS has occurred in concert with increasing mean annual summer (June–August) temperatures (MAST) in the western Canadian Arctic Archipelago. Higher summer air temperatures increase active layer thickness, thawing near surface permafrost and triggering terrain instability where ice-rich sediments occur (Hinkel et al., 2001). Climate stations in Mould Bay and Sachs Harbour show statistically significant (p < 0.01) long-term (1956–2015) warming trends in MAST (supporting information Figure S1) (Environment Canada, 2017). At Sachs Harbour, a relatively stable landscape from 1961 to 2009 is associated with a MAST of 4.3°C that increased to 6.1°C from 2010 to 2016. The summers of 2010–2012 were particularly warm with MAST >7°C, peaking at 8.4°C in 2012—the warmest year on record (1956–2015).

During the 1956–2015 interval of increasing summer temperatures at Sachs Harbour there was neither an increase in total summer precipitation nor an increase in the intensity of precipitation, two factors identified elsewhere in the low Arctic in promoting the expansion of RTS (Kokelj et al., 2015; Segal et al., 2016). This contrasts with more recent local observations of intense rainfall events and thunderstorms in Sachs Harbour phenomena considered rare in the settlement's history (1956 to present) (Jolly et al., 2002). However, due to its location 240 km to the southwest of Johnson Point, the climate record at Sachs Harbour may not accurately reflect rainfall at Johnson Point. We propose that the prominent intensification of RTS reported here post-2009 records an increase in active layer thaw prompted by summer warming acting in concert with increased precipitation.

3.2 Downstream Impacts of Thermokarst

RTS magnify instability of ice-rich landscapes, coupling slope, and fluvial processes that accelerate sediment transfer from their catchments (Kokelj et al., 2015). To determine the downstream impact of RTS, samples were collected from tributaries and the main channel of the Johnson Point watershed during a 5 h interval on 10 July 2015 (Figure 5 and supporting information Figure S2). The impact of RTS during sampling was clear—disturbed watersheds had mean total suspended sediment (TSS) loads 3 orders of magnitude greater than undisturbed watersheds (100 000 versus 10 mg/L) and total dissolved solids (TDS) that were 2 times higher (500 versus 250 mg/L). Furthermore, increasing watershed order (size) minimally diminished the impact of physical disturbance, with TSS and TDS values at the outlet of the main river (sixth order, 230 km² catchment) reaching 3,030 and 320 mg/L, respectively (Figure 5). There is a significant positive relationship between TSS, TDS, and the percentage of the catchment disturbed (TSS: Spearman correlation = 0.68, p < 0.001, TDS: Spearman correlation = 0.76, p < 0.001) (Figures 5b and 5c). The connectivity of slumps to the stream channels as well the distance between the disturbance and the sampling points contribute to the variability in downstream TSS. Regardless, these relationships indicate that sedimentary and geochemical impacts to fluvial systems will increase with the intensification of climate-driven thermokarst. Notably, a number of undisturbed tributaries were dry; hence, downstream water chemistry changes resulted from sustained discharge primarily traced from disturbances. Concentrations of TSS measured in this study are comparable to observations from other slump-impacted ice marginal streams in northwestern Canada (Kokelj et al., 2013), but they are remarkably higher than in studies conducted in Alaska (Bowden et al., 2008) and from the Canadian High Arctic (Lamoureux et al., 2014). These contrasts reflect the slope-stream connectivity that develops in association with increasing thaw slump intensity.

Figure 5

(a) Johnson Point watershed, with second- through sixth-order stream watersheds delineated. Watersheds with a high density of RTS are associated with the highest total suspended sediments (TSS) and total dissolved solids (TDS). The relationship between (b) TSS and (c) TDS and the proportion of catchment area disturbed. Outlets from headwater lakes exhibited minimal discharge at the time of sampling. (d) Isotopic proportions of ground ice (grey) and rainfall (blue) samples classified by watershed order.

In order to quantify the contribution of buried massive ground ice melt to regional streamflow, we analyzed samples of ice from the headwalls of four RTS as well as rainfall on 8 July 2015, for stable isotope ratios of oxygen (¹⁸O/¹⁶O) and hydrogen (D/H) (supporting information Figure S3). The buried (glacial) ice has δ¹⁸O values between −22.4 and −29.2‰ and δD values from −177.1 to −232.4‰. Two of the massive ice samples were clustered with δ¹⁸O values of −28.8 and −29.2‰ and δD values of −228.9 and −232.4‰, substantially lower values than regional rainfall (δ¹⁸O, 14‰ ± 2.2 and δD, 121.1‰ ± 21.5, n = 4) and consistent with other studies of massive ground ice in this region identified as buried glacier ice (French & Harry, 1988; Worsley, 1999). The remaining two samples indicate possible contamination with modern snow melt derived water but could reflect a complex glacial origin; we remain cautious about their stratigraphic context.

Rain water and ground ice were used as end-members in a two-component isotope mixing model to separate river water samples into putative sources from ice versus atmospheric precipitation. Results indicate that the proportion of relict ground ice meltwater constitutes a greater proportion of discharge in third-order streams where flow from several thermokarst-affected second-order tributaries converge (Figure 5d). Thermokarst-affected watersheds have a proportionally higher inferred ground ice contribution than undisturbed watersheds, whose water is otherwise supplied by rainfall and at the time of sampling, melt of aerially small and spatially limited late lying snowpacks. Notable exceptions include four undisturbed watersheds (20% of second-order watersheds) where isotopic composition indicates >85% ground ice contribution to runoff. We interpret these results to indicate that active layer thaw is now truncating widespread relict massive ice deposits. Each of these four watersheds show evidence of thermokarst ponding and exhibit extensive areas of orthogonal ice wedge polygons characteristic of ice-cored terrain (Dyke et al., 1992; Lakeman & England, 2012). Undisturbed catchments exhibiting high isotopically defined relict ice-sourced meltwater contributions to streamflow represent the emergence of a new source of subsurface water that is now being released by climate change. This isotopic streamflow signal within undisturbed catchments may provide predictive measure for assessing future landscape destabilization and intensification of fluvial sedimentary impacts. While the proportion of ground ice in the main river is diminished compared to the third-order channels upstream, the outlet of the river includes a ~ 35% contribution from relict ground ice meltwater demonstrating that permafrost degradation now contributes significantly to summer baseflow.

4 Conclusions

Permafrost disturbance has accelerated substantially in this region, where cold climate (< −10°C mean annual ground temperature) has preserved relict glacial ice beneath as little as 1 m of sediment overburden (Lakeman & England, 2012). Increased active layer thaw is truncating relict massive ice, significantly increasing the prevalence of RTS. These observations suggest that landscape disequilibrium has been triggered by recent climate change.

There are yet few reports of climate-induced destabilization of High Arctic terrain. In contrast, widespread thawing and thermokarst has been reported during comparable warming of several discontinuous permafrost zones, such as the Tuktoyaktuk Peninsula, Canada, and Siberia (Grosse & Romanovsky, 2011). On Banks Island, within the Jesse moraine belt, we conclude that the recent pulse of exceptional summer temperature, and likely anomalous heavy rainfall, is accelerating permafrost thaw highlighting the vulnerability of this expansive, ice-cored terrain.

Buried glacial ice (and other massive ice deposits) can be important morphological components of permafrost landscapes. Climate-driven thaw can renew the melt-out of relict massive ice in permafrost-preserved moraines. Our isotopic evidence demonstrates that a major contribution to baseflow discharge in ice-cored, High Arctic terrain is now being provided from ground ice melt. The sedimentary and geochemical impacts of thawing relict ice on downstream environments are significant, and our data indicate that fluvial effects will increase with intensification of thermokarst. Hydrological impacts from intensification of RTS and top-down thaw truncation of massive ice is now augmenting summer runoff and ephemeral drainage in small High Arctic watersheds underlain by relict massive ice. Accelerated thermokarst is also significantly increasing sediment and solute transport downstream (>20 km) to both lacustrine and coastal environments. This has important implications for reordering the summer hydrological regime in freshwater systems, expanding the seasonality of transfers to lakes and oceans, and affecting the ecological integrity of these water bodies. The intensity of the reported landscape changes on Banks Island demonstrates that regions of cold, continuous permafrost are undergoing irreversible alteration, unprecedented over centennial to millennial timescales based on the melt of buried ice that has persisted since ~13 cal kyr B.P. Its scale, abruptness, and environmental impact rank among the most dynamic examples of permafrost degradation in the world.

Acknowledgments

Research was supported by NSERC (S. F. Lamoureux) and the W. Garfield Weston Foundation (A. C. A. Rudy), the Geological Survey of Canada Geo-Mapping for Energy and Minerals Program, Polar Continental Shelf Program, NWT Cumulative Impacts Monitoring Program (CIMP), and the NWT Geological Survey. Chemical analyses were provided by Taiga Laboratories, Yellowknife. We greatly appreciate the support from the community of Sachs Harbour and field assistance by Charleton Haogak. Data listed in the tables will be made available on the Polar Data Catalogue. We are thankful for the constructive manuscript reviews of Steve Grasby (NRCan—internal review) and from journal reviewer M. Gooseff and a second anonymous reviewer. Research was conducted under NWT Scientific Research License 15687 and is NRCan contribution #20170086 and NTGS contribution #0106.