Thermokarst and export of sediment and organic carbon in the Sheldrake River watershed, Nunavik, Canada



[1] A spatiotemporal computation of permafrost decay covering the period from 1957 to 2009 and validated by field investigations was made over a 76 km2 river catchment straddling the tree line, in the discontinuous permafrost zone, east of Hudson Bay, in order to estimate the amounts of sediments and organic carbon released by thermokarst. Lithalsas and palsas are the dominant permafrost landforms, whereas thermokarst ponds, landslides, active layer failures, and gullies are the main features of permafrost degradation. Results show that 21% of the existing permafrost in 1957 had disappeared in 2009, resulting in a 96% growth of the thermokarst pond cover and a 46 to 217% increase of the number of active erosion landforms. An increase of stream connectivity related with the degradation of permafrost potentially allowed for an increase of sediments and carbon delivery to the main stream by a factor of 1.6. Volume of active landslides and gullies also increased by 12 to 38%, enhancing sediment and organic matter yields. Significant differences in permafrost degradation and in sediment and carbon inputs were observed along an east–west transect, from sites located at the head of the watershed near the tree line to sites located downstream close to the Hudson Bay coast. Thermokarst ponds in the forest tundra area released 2.3 times more sediments and dissolved organic carbon per unit of area in the fluvial system than in the shrub tundra area. Despite these yields by thermokarst, the Sheldrake River catchment currently does not seem to be yielding proportionally more sediments and carbon than a permafrost-free river catchment.

1 Introduction

[2] Recent climate changes have induced geomorphic perturbations in arctic and subarctic landscapes because of severe terrain disturbances caused by the thawing of ice-rich permafrost [Jorgenson et al., 2001; Jones et al., 2011]. These impacts are particularly conspicuous in the discontinuous permafrost zone as permafrost extent is decreasing rapidly, resulting in an increase in number and extent of thermokarst lakes in areas where soil drainage is poor, such as clays and peatlands [Laberge and Payette, 1995; Matthews et al., 1997; Camill, 1999; Osterkamp et al., 2000; Beilman et al., 2001; Payette et al., 2004; Vallée and Payette, 2007]. However, in spite of measured spatial reduction of permafrost areas, very little field data exist on the volumes and masses of mineral and organic sediment released to stream systems by permafrost decay.

[3] Most papers have focused on the spatial decay of permafrost at the scale of palsa bogs [Laprise and Payette, 1987; Laberge and Payette, 1995; Zuidhoff and Kolstrup, 2000; Vallée and Payette, 2007] and at the regional scale [Vitt et al., 2000a; Thibault and Payette, 2010; Sannel and Kuhry, 2011]. However, the impact of permafrost thawing on sediment release and contribution to solid discharge in river catchments is still poorly known. Some attempts were made to measure sediment and carbon outputs from stream basins in degrading permafrost regions [Kokelj et al., 2005; Bowden et al., 2008; Olefeldt and Roulet, 2012]. Both the original and the transitional basin morphology during permafrost terrain degradation are likely to exert some control on the paths and discharge of thermokarst products, particularly carbon. Such an assessment is needed as the organic matter (OM) that was stored in the permafrost is recycled in thermokarst ponds as particulate organic matter (POM) and dissolved organic matter (DOM), which are involved in the biogeochemical production of methane (CH4) and dioxide gas (CO2). Furthermore, a fraction of the released fine-grained sediments and carbon also flows into the fluvial system and, ultimately, to the sea where it likely contributes to changes into the marine food web and the sedimentary system.

[4] The main objective of this paper is to provide a better understanding of the processes of geomorphic change in a river catchment where permafrost is being impacted by thermokarst. The study investigates the release of sediments and carbon from permafrost decay in the fluvial system with emphasis on the potential export of fine sediments (silty clay) as suspension load, dissolved organic carbon (DOC), and particulate organic carbon (POC). We describe erosion and fluvial sediment transport within a catchment whose morphology, particularly stream connectivity, is being transformed by thermokarst. Our approach consists of estimating sediment and carbon volume and mass changes resulting from recent and current thermokarst and related denudation processes within the watershed. We also tested the hypothesis that the increase in number and extent of thermokarst ponds has increased the overall sediments and carbon load of the fluvial system studied.

[5] In order to obtain this new catchment-scale knowledge and to better understand regional mass transfers related to permafrost thawing, the specific objectives of the study were (1) to quantify permafrost decay between 1957 and 2009 in a river basin located in the discontinuous permafrost zone; (2) to quantify the area and size of new thermokarst ponds and erosion features since 1957; and (3) to assess the amounts and fractions of organic matter, organic carbon, and mineral sediments sourcing from recent permafrost degradation that are mobilized in the reorganized watershed.

2 Study Area and Regional Setting

[6] The Sheldrake River is 25 km long from its source in Sheldrake Lake to the shore of Hudson Bay. Inland, it flows over glacially eroded Precambrian granitic gneiss in a hilly (200–250 m above sea level (asl)) landscape partially covered by late glacial and postglacial sediments. The six last kilometers of the river run in a valley carved across Late Proterozoic sedimentary-volcanic rocks which form a chain of coastal hills (400 m asl) parallel to the arc-shaped coastline of the eastern Hudson Bay. For the final 2 km, the river flows on a gentle structural coastal slope (<3°) following the dip angle of a thick basalt layer. It flows into Hudson Bay 8 km north of the Inuit village of Umiujaq (56°37′N; 76°32′W) (Figure 1).

Figure 1.

Location of the Sheldrake River catchment showing the position of the tree line and the maximum elevation reached by the Tyrrell Sea (marine limit). BGR and VDT are informal names for two monitored lithalsas.

[7] Following the receding front of the Laurentide Ice Sheet eastward and inland, between 8000 and 7400 cal years B.P. [Lajeunesse and Allard, 2003a; Lajeunesse, 2008; Lavoie et al., 2012], the Tyrrell Sea inundated the whole Sheldrake River watershed (Figure 1). The studied valley and associated tributary valleys were filled with postglacial marine silty clay, reaching up to 85 m in thickness [Allard and Seguin, 1985]. Therefore, silty clay accounts for 86% of surface deposits (Table 1). Sandy and gravelly ice contact deltas [Lajeunesse and Allard, 2003a] mark the marine limit at an elevation of 220 m asl at the east end of the Sheldrake River basin [Allard and Seguin, 1985]. Near the mouth of the river, sand and gravel deposits anchored on the Nastapoka Hills form a fan-like apron sloping from the hills to the actual seashore [Allard and Seguin, 1985; Lajeunesse and Allard, 2003a, 2003b]. In flat and poorly drained inland valleys, peat covers silt and sand deposits with an average thickness of 1 m [Lévesque et al., 1988]. Surficial deposits cover 32% of the watershed. Bedrock and water (lakes) make up the balance.

Table 1. Surficial Deposits Distribution in the Sheldrake River Catchment
 Area (km2)Total Surficial Deposits (km2)Silty Clay (%)Sand (%)Peat (%)
Shrub tundra331188120
Forest tundra431384115

[8] The climate is subarctic, with cold winters (−24°C in January) and cool summers (10°C in August). Since 1990, mean annual air temperatures increased from −5.2°C to −3.9°C; 40% of the 550 mm average annual precipitation falls as snow [Environment Canada, 2010]. The cooling impact of Hudson Bay generates a west–east climate gradient from the shoreline inward, resulting in the tree line being parallel to the coastline, about 15 km inland (details on vegetation can be found in Payette and Rochefort [2001] and Bhiry et al. [2011]). The tree line refers to the first occurrence of isolated trees in the landscape [Payette, 1983]. Here, compared to regional estimates of the tree line position [Payette, 1983], it has been slightly moved to west to include valleys with scattered spruces.

[9] The 76 km2 studied watershed is located at the boundary between the isolated and sporadic permafrost zone (2 to 50% areal cover) and the widespread discontinuous permafrost zone (50 to 90%) [Allard and Seguin, 1987a]. According to Lévesque et al. [1988], in 1957, permafrost covered approximately 50% of surficial deposits in the Sheldrake River basin. The rest of the surficial deposits are permafrost free and consist of taliks, forested and shrubby hollows, and wetlands. Lithalsas (mineral permafrost mounds), permafrost plateaus (elongated and wide mineral permafrost landforms), palsas (peaty permafrost mounds or permafrost mounds with a peat cover), and peat plateaus (elongated and wide peaty permafrost landforms) are the only permafrost landforms (for simplicity, the term “lithalsa” is used here to include permafrost plateaus and the term “palsa” encompasses peat plateaus). All these heaved landforms stand above the surrounding terrain by several meters (generally 3 to 5 m) due to the development of ice segregation lenses formed by cryosuction of soil water in freezing fine-grained, frost susceptible soils [Pissart, 1985, 2002]. Their volumetric ice content varies between 50 and 80% [Calmels and Allard, 2004; Calmels et al., 2008]. Permafrost thickness in the frozen mounds and plateaus varies typically from 10 to 15 m [Lévesque et al., 1988]. The active layer is ~1.5 m deep on clay soils [Delisle et al., 2003; Calmels and Allard, 2004, 2008; Calmels et al., 2007] and is currently thickening (Figure 2). It is about 0.6 m in peat on tops of palsas [Marchildon, 2007]. One of our thermistor cables attests of the presence of permafrost in bedrock near the coast in the village of Umiujaq. However, regional distribution of permafrost in bedrock is unknown.

Figure 2.

Change in the maximum active layer depth in the VDT lithalsa between 2001 and 2009. Vertical axis shows the depth in centimeters below the ground surface. The dashed grey line is the linear regression line.

[10] In Northern Québec, palsas and lithalsas formed during cold periods of the Late Holocene, i.e., 1500 to 1000 B.P., and during the Little Ice Age (LIA) [Couillard and Payette, 1985; Allard and Seguin, 1987a, 1987b; Lavoie and Payette, 1995; Payette and Delwaide, 2000; Arlen-Pouilot and Bhiry, 2005; Marchildon, 2007]. Since the end of the LIA, numerous permafrost landforms have degraded due to global warming and to increased snow precipitations [Seguin and Allard, 1984; Vitt et al., 2000a; Payette et al., 2004; Arlen-Pouilot and Bhiry, 2005]. Permafrost mounds at different stages of degradation are found within the region [Calmels et al., 2007]. The main geomorphological impact associated with this decay is an increasing number of thermokarst ponds and erosion on hill slopes [Allard et al., 1987].

[11] The Sheldrake River watershed was selected for this study because several previous studies in the region provide baseline information on permafrost spatial distribution, landforms, permafrost thickness, ground ice content, and thermal regime. This strategic area straddles the tree line, in a region of intensive degradation of permafrost, and is representative of the eastern coast of Hudson Bay landscape. Meteorological stations and sites for monitoring permafrost decay, such as the BGR lithalsa (informal name for Bundesanstalt für Geowissenschaften und Rohstoffe, a German research center which collaborated with the Centre d'études nordiques in the early 2000s) and the VDT lithalsa (for Vallée des Trois, informal name given to a valley east of the village of Umiujaq), provide quantitative data that help to understand landscape adaptation to permafrost decay (see Figure 1 for locations). Finally, a water level gauge installed in 2008 2 km upstream from the Sheldrake River mouth provides useful hydrological information applicable to the catchment.

3 Permafrost Degradation Processes

[12] In the studied watershed, the first evidence of degradation is the observed recent thawing of permafrost mounds (Figure 3a). Surface and slopes of lithalsas are affected by thaw settlement, gelifluction, active layer failure, and slopewash erosion. Alteration of the thermal balance of the ice-rich mounds first creates a summit depression which deepens and enlarges summer after summer, retaining water and forming a thermokarst pond surrounded by an ice-poor ridge [Calmels et al., 2007]. Clayey sediments are brought to the expanding pond by overland flow on the bare surface of frost boils, which keep bringing fine-grained sediment to the surface through overturning (Figure 3b). During rain events, clay is eroded and channeled in furrows between the frost boils into the adjacent ponds or directly into the stream network [Seguin and Allard, 1984]. Small active layer slides along the shore of the ponds often release sediments as well [Allard et al., 1987].

Figure 3.

Permafrost decay features in the Sheldrake River catchment. (a) Typical thermokarst landscape in the forest tundra area, (b) a frost boil evolving into an active layer failure, (c) palsa peat blocks falling in a pond (photo: D. Sarrazin), (d) active layer failure detachment on the edge of a high permafrost plateau, (e) network of eroding gullies between permafrost plateaus, and (f) active layer slide occurring in spring 2010; note the frost boils on the top of the plateau (photo: D. Sarrazin).

[13] On palsas and peat plateaus, extension cracks inherited from the time of heave occur on the swelled surface of the mound and divide the peat cover into blocks that collapse in surrounding ponds (Figure 3c), streams, and rivers, as frequently observed in North America [Allard and Seguin, 1987a; Allard et al., 1987; Lévesque et al., 1988; Vitt et al., 1994; Laberge and Payette, 1995; Osterkamp and Romanovsky, 1999; Beilman et al., 2001; Payette et al., 2004; Arlen-Pouilot and Bhiry, 2005] and elsewhere [Matthews et al., 1997; Zuidhoff and Kolstrup, 2000]. In peat plateaus, ground subsidence can initiate the thermokarst pond which expands through peat block erosion [Sannel and Kuhry, 2011]. Some palsas also gradually lose their peat cover through erosion, particularly by the wind [Seppälä, 2003]. When the peat cover is totally eroded, the underlying silt is exposed at the ground surface. Since silt has a higher thermal conductivity than peat, the active layer gets thicker. Frost boils become active.

[14] During the subsidence of palsas and lithalsas, slope retreat delivers peat and sediments in the expanding ponds. Therefore, the thermokarst ponds act as sediment traps and are very turbid [Breton et al., 2009]. A pond isolated from the fluvial drainage network is a small sedimentation basin within a closed catchment [Bouchard et al., 2011]. Consolidation of postglacial silty clay underlying thermokarst ponds makes the soil nearly impermeable once the permafrost is completely thawed. Consequently, vertical water infiltration is negligible. Drainage of the pond only occurs when it overflows or when it becomes connected to the main drainage network through some newly eroded creek channels. Some mounds, however, are directly located along river banks. When they decay, the morphology of the river channel is affected [Vallée and Payette, 2007] and their sediments are then directly taken in charge as river load.

[15] Landslides occur along steep river banks, particularly in areas of permafrost plateaus, in the western part of the studied watershed. Small active layer failures occur on the steep upper edge of permafrost plateaus (Figure 3d). They are active during summer thaw [Seguin and Allard, 1984] and lead to the failure of “a thin veneer of vegetation and mineral soil and subsequent movement over a planar inclined surface” [McRoberts and Morgenstern, 1974]. Finally, gully erosion (Figure 3e) occurs between adjacent permafrost plateaus and can create networks of gullies in some permafrost fields. They are directly connected to the drainage network through elongated ravines.

4 Methodology

[16] The methodology is based on precise mapping, estimates of landforms changes in volume and mass, observations of flow regime in the catchment, and observations of active processes.

4.1 Permafrost and Thermokarst Mapping

[17] The Sheldrake River watershed was first delineated using a digital elevation model generated from a topographic map (contour lines: 20 m, Ressources Naturelles, Faune et Parcs Québec). We choose to exclude the Sheldrake Lake catchment as it has no visible signs of erosion, being surrounded by wetlands and bedrock outcrops. Only the area drained by the Sheldrake River and its tributaries was accounted for in measurements of erosion and mass computation. Surficial deposits were mapped using 1957 aerial photographs (1:40,000) (~1 m resolution), a 2009 GeoEye satellite image (0.6 m resolution), field observations, and revisions of previous maps by Lévesque et al. [1988] at a smaller scale (1:50,000). The 1957 aerial photographs were geo-referenced with the 2009 satellite image using ~ 20 control points per photographs distributed throughout the area. Control points mainly consisted of rock outcrops.

[18] Every permafrost mound and thermokarst pond in the catchment was delineated and mapped in the ArcGis software for both 1957 and 2009. This method allowed for a much higher precision than previous studies which used area estimates [Lévesque et al., 1988]. For 1957, the use of a stereoscope with enlarging lenses (3×) allowed for the detection of typical heaved permafrost landforms. For 2009, permafrost features and thermokarst ponds were precisely mapped on the high-resolution satellite image. Lithalsas are pitted with light active frost boils. Palsas have a more uniform surface and the brownish color of the peat cover is characteristic. The whitish thermokarst ponds are often surrounded by rim ridges and look like sinkholes in the landscape. Field observations, photographs, and videos taken from helicopter flights over the whole basin area helped to verify uncertain geological contacts and validate mapping. Not all thermokarst ponds are associated with a rim ridge, particularly in wetlands, in peat bogs, and in fen areas. Coalescence of several ponds leads to the creation of larger lakes and unstructured wet areas. In this study, every water body resulting from permafrost decay was considered as a thermokarst pond. The margin of error is estimated to less than 5% for the delineation of each pond and permafrost mound.

[19] In order to detect variability in changes within the catchment and to assess the relative contribution in sediment and carbon fluxes from different sectors in the watershed, the catchment area was classified into 41 numbered units of land systems. Segmentation was based on the visual criteria of permafrost percentage cover, topographic style, and type of surficial deposits. Two permafrost maps were created and subtracted to produce a map of permafrost degradation between 1957 and 2009 (Figure 4).

Figure 4.

Map of the degradation of the permafrost in percentage and location of landslides, active layer detachment failures, and gullies.

[20] The presence of permafrost in sandy areas (land systems 1, 13, 35, 36, and 40; for location, see Figure 4) is difficult to confirm because no typical periglacial landform indicates presence or absence of permafrost below the surface. Those areas are mostly wind-eroded surfaces supporting small dune ridges and were considered as permafrost areas without vegetation as suggested by Lévesque et al. [1988], who made numerous geo-electrical surveys on these terrain types. No visible signs of thermokarst have been observed in these areas, and sand deposits represent only 11% of total catchment in unconsolidated deposits. The study focused on the potential export of fine sediments and carbon as suspension load in the fluvial network. Therefore, sand deposits are shown on the map but were not included in the computations.

4.2 Landslide Identification and Sediment and Organic Matter Released

[21] The yields of sediment and carbon from decaying permafrost were estimated through geometrical measurements on air photographs supported by ground checks and measurements at typical sites. The measured landforms included landslides, gullies, and small active layer failures.

[22] Three landslides were visited in the field and seven were identified on the satellite image. They consist of non vegetated slopes in thawed and collapsed sediments along steep river banks that are constantly destabilized by fluvial toe erosion. Geometrical parameters (height, length, horizontal distance, and breath) were measured in the field and on the satellite image. Thereafter, the eroded volume in each landslide was estimated with trigonometrical functions (Figure 5). Because of the high rate of the isostatic uplift in this region [Lavoie et al., 2012], fluvial incision is still in progress and maintains steep river banks, particularly along large permafrost plateaus. Consequently, we assume that the original land surface was a cube volume without any slope.

Figure 5.

Method for estimating missing volume from landslides along steep river banks: satellite image, field photograph, and geometrical calculation.

[23] Slope of three active layer failures and one system of gullies were measured in the field and, respectively, 6 and 19 such landforms were identified on the satellite image in the basin. In order to estimate volumes of eroded sediments, active erosion surfaces (i.e., bare and unstable) were measured on the satellite image. An average scar depth of 0.75 m with a slope of 40° for active layer failures and an average scar depth of 0.9 m with a slope of 45° for gullies between two plateaus were estimated by field measurements. Erosion landforms less than 3 m in length were not considered in this study.

[24] Volume was converted into mass by assuming a porosity of 0.35 and a density of the solid particles of 2.7 t/m3, which are typical values for the glaciomarine clays [Holtz and Kovacks, 2009]. The volume to mass conversion for landslides takes into account a 1.5 m thick active layer without ice content and an underlying permafrost with an average ice content of 60% [Calmels et al., 2007, 2008; Calmels and Allard, 2008; Fortier et al., 2008]. For active layer failures and gullies, the ice content is considered null since erosion only takes place in the active layer. We assume that the totality of the volume of missing sediment is released in the fluvial system and that OM is exported in DOM or POM.

[25] A large complex of gullies is active in a scrubland sector in the center of the catchment. The abundance of shrubs since at least 1957, which retain a thick snow cover, indicates that this sector has no permafrost. This was confirmed by electrical resistivity surveys conducted by Lévesque et al. [1988]. Although morphometric changes were measured in this sector as well, this area was excluded from the computation of permafrost-induced sediment quantities since it does not actually contribute to thermokarst-induced sediment input into the river system.

4.3 Estimation of Organic Carbon and Sediment Fluxes from Connected Thermokarst Ponds

[26] All ponds connected to the drainage network were inspected to assess the volume of turbid water that is evacuated when they overflow during rain events. Connected thermokarst ponds (CTP) in 2009 were first delineated using the high-resolution GeoEye image. Sometimes, draining streams were not visible, but wet linear shrubby hollows connecting ponds to the river drainage system act as flow paths during precipitation events. This is confirmed by the presence of small deltas at the confluence with permanent streams. To assess the 1957 situation, every pond that was connected in 2009 and already present on the old air photographs was considered as CTP.

[27] To calculate the coverage of new connected thermokarst ponds (NCTP), i.e., the area of CTP created between 1957 and 2009 (area NCTP), we used the following equation:

display math(1)

where area CTP57 and area CTP09 are the areas covered (m2) by CTP in 1957 and in 2009. To assess the current release of total suspended sediment (TSS) and DOC from CTP, it is assumed that the volume of water (with known contents of DOC and TSS) overflowing into the drainage network is equal to the volume formed by area CTP09 and the thickness of a layer of cumulated rain that generates overflow. This volume must be considered as a minimum because the area of the drainage basin of each thermokarst pond was not taken into account.

[28] The hydrographs from the gauging station (Figure 6) show that every rain event >2 mm/day has an impact on the Sheldrake River water level. Therefore, during a rain event, we can assume that flow occurs through CTPs. The study focused on the period from mid-June, at the end of the freshet, to mid-October, just before freeze back. A rain event is defined here as a period of rain of one or several consecutive days leading to a peak of discharge, i.e., a water stage on the hydrograph (Figure 6).

Figure 6.

Hydrographs (water level) of the Sheldrake River from mid-June to mid-October in 2009 and in 2010. Every rain event with an impact on the river discharge is numbered. Daily rainfalls are plotted as bars.

[29] The volume of water released by NCTP in the drainage network (W) between the 15 June and the 15 October is summarized by the following expression:

display math(2)

where P is the depth of cumulated rain (m) falling on the catchment between 15 June and 15 October in 2009 and 2010.

[30] Finally, exports of TSS and DOC from NCTP were calculated as follows:

display math(3)
display math(4)

where expDOC and expTSS are the mass of DOC and TSS (tons) released by NCTP between 15 June and 15 October of each year; avTSS and avDOC are, respectively, the average concentration of TSS and DOC (g m−3) in NCTP. 10−6 is the conversion factor from grams to tons.

[31] Estimating avTSS and avDOC in the catchment is difficult given that the turbidity of ponds and streams is highly variable in space and time (~4400 ponds in 2009). We used concentration values from Breton et al. [2009]. These authors had selected in our study area 16 thermokarst ponds they had estimated representative of the variety of colors (related to sediment and organic matter concentrations) and development phases among numerous thaw ponds in the region. They have measured an average concentration of 19.8 g m−3 of TSS (16 ponds) and 4.9 g m−3 of DOC (5 ponds). The TSS is principally composed of silt and clay [Breton et al., 2009]. They also provided measurements of TSS and DOC concentrations in 17 other thermokarst ponds in the wider region of eastern Hudson Bay coast, which all fall in the same range as those taken in our study area. In the Sheldrake river catchment, concentrations of DOC range from 1.3 to 11.5 g m−3 (median: 4.3; σ: 3.05; standard error of the mean: 0.76) and concentrations of TSS range from 5.3 to 39.7 g m−3 (median: 15.5; σ: 12.9; standard error of the mean: 1.13) [Breton et al., 2009]. From these data, we simply used average concentration values of 20 g m−3 of TSS and 5 g m−3 of DOC for the total population of thermokarst ponds in our study area. This calculation assumes a linear relationship between precipitation and DOC/TSS loading, even if a strong flow dependency of export can exist [Finlay et al., 2006]. It also ignores effects of connectors, which, if vegetated, can filter out TSS and influence DOC concentration. Evaporation between rain events is not considered. Estimates are very likely to change with more detailed sampling of concentration in the future. Despite large uncertainties, this approach provides an acceptable order of magnitude though still rough and preliminary.

4.4 Hydrological Connectivity

[32] Providing a precise overview of the first-order drainage network of the whole catchment is complicated because of the discontinuous extent of surface sediments between rock outcrops and the fragmented nature of the landscape. In order to test and illustrate how thermokarst may have affected stream connectivity, an ~ 0.2 km2 flat subbasin located in the forest tundra area (land system 24) was selected. This subbasin was selected because of the presence of palsas, the rate of permafrost decay, and the increase in thermokarst pond coverage, which are similar to the mean value of the forest tundra area. The stream density (Dd) is defined as the cumulative length of water tracks (L) divided by the area of the subbasin (A) [Horton, 1945] (equation (5)).

display math(5)

5 Results

5.1 Spatiotemporal Evolution of Permafrost

[33] The mapping of the entire basin of the Sheldrake River provides an overview of permafrost conditions on the eastern coast of Hudson Bay. In 2009, permafrost in the Sheldrake River watershed was present in 20% of surficial deposits, while it was 26% in 1957. In 2009, permafrost was present in 32% of surficial deposits in the shrub tundra area, while it was present in only 10% of the surficial deposits in the forest tundra (Table 2).

Table 2. Lithalsas, Palsas, and Total Permafrost Coverage and Changes Between 1957 and 2009
 LithalsasPalsasTotal Permafrost
 Square MeterPercentSquare MeterPercentSquare MeterPercent
 Shrub tundra
 Forest tundra

[34] Complete statistics of permafrost decay and thermokarst ponds between 1957 and 2009 for the 41 land systems are available in Tables A1 and A2. Synthesized data are presented in Tables 2 and 3.

Table 3. Thermokarst Pond Coverage, Connected Thermokarst Pond Coverage and Changes Between 1957 and 2009
 Thermokarst PondsConnected Thermokarst PondsCTPPa
 Square MeterPercentSquare MeterPercentSquare Meter
  1. a

    CTTP is connected thermokarst pond from palsa collapse.

 Shrub tundra
 Forest tundra

[35] Total permafrost coverage decreased by 21% from 1957 to 2009. However, the degradation is very unevenly distributed in space, with an average loss of permafrost of only 7% in the shrub tundra area, compared to a permafrost loss averaging 43% in the forest tundra, east of the tree line (Figure 4).

[36] In 1957, thermokarst ponds covered 0.56 km2. In 2009, this value had doubled to reach 1.09 km2 (+96%). In the forest tundra, the number of ponds increased from 2958 ponds in 1957 to 4442 in 2009 (+50%) (Table 4). Logically, thawing ponds replaced permafrost mounds and the trend follows the rate of permafrost degradation.

Table 4. Number and Average Size of Lithalsas, Palsas, and Thermokarst Ponds in the Forest Tundra Area, and Changes Between 1957 and 2009
 LithalsasPalsasThermokarst Ponds
 NumberSize (m2)NumberSize (m2)NumberSize (m2)
Change (%)4−29−55−65033

[37] Flat-floored poorly drained inland valleys were the most affected areas. Permafrost degradation was generally above 50%, and the surface covered by thermokarst ponds increased by 100% or more since 1957 (see land systems 19, 27, 33, 34, and 41 in Tables A1 and A2). In the four areas where palsas and lithalsas occur together, i.e., in peat deposits, palsas have degraded by 43%, whereas lithalsas have decreased by 71%.

[38] Because of the large extent of permafrost plateaus in the western part (shrub tundra) of the basin, it is impossible to count the exact number of mounds and assess how many individual landforms have disappeared. In the forest tundra, the number of palsas had decreased by 55%. The average area occupied by a lithalsa and a palsa was 1806 and 954 m2, respectively, in 1957. These values have decreased to 1274 and 895 m2 in 2009 (Table 4). No new permafrost mound, either palsa or lithalsa, formed during this period.

[39] Statistics on landslides and gullies are presented in Table 5. Of the 38 features identified in 2009, 12 were clearly present and active in 1957 and 2009, 12 of the landforms were clearly absent in 1957, and 14 of the possible already active features in 1957 were unidentifiable because of the presence of snow patches on the aerial photographs. The number of active landslides and gullies increased from 46 to 217% between 1957 and 2009 (Table 6). This large percentage range is due to the residual snow covers masking some slopes on the 1957 air photographs.

Table 5. Characteristics and Measurements of the Landslides, the Active Layer Failures, and the Gullies of the Sheldrake River Catchment in 2009
LocationBiomeLenght (m)Horizontal Distance (m)Height (m)Breath (m)Slope(°)Surface (m2)Depth (m)Volume (m3)Mass (t)Present in 1957?
  1. a

    Shrub tundra.

  2. b

    Forest tundra.

Active Layer Detachment Failures
Total        35,65248,505 
Table 6. Number, Volume, and Mass Changes in Landslides and Gullies Between 1957 and 2009
  Total Eroded SedimentEroded Organic Mattera
Characteristics in 1957 and in 2009Number of LandformsCubic MeterMetric TonCubic MeterMetric Ton
  1. a

    The mass and volume of eroded organic matter are included in the mass and volume of total eroded sediment.

Absent in 1957, present in 2009123,7785,4224970
Present in 2009, snow covered in 1957146,02310,57178137
Active in 1957 and 20091225,85132,512336423
Total in 20093835,65248,505463631
Variations46 < × < 217%12 < × < 38%

5.2 Activity of Landslides, Active Layer Failures and Gullies Between 1957 and 2009

[40] The activity of landslides, active layer failures, and gullies increased by a factor of 1.12 to 1.38 between 1957 and 2009 (Table 6). A total of 35,652 m3of eroded sediment volume from active landslides and gullies was calculated in 2009, which corresponds to 48,505 t of fine sediments, with an average content of 1.3% of OM [Calmels and Allard, 2008], i.e., 463 m3 (631 t). The 12 newly active landslides (observed in 2009 only) correspond to an eroded volume of 3778 m3 (5422 t) of clay, of which 49 m3 (70 t) is of OM. The 14 gullies and active layer failures unidentifiable on the 1957 photographs correspond to an erosion of 6023 m3 (10,571 t) of clay, which contains 78 m3 (137 t) of OM. Landforms already active in 1957 show a lost volume in 2009 of 25,851 m3 (32,512 t) of clay with 336 m3 (423 t) of OM. By averaging landslide activity over the study period, 686 m3 yr−1 (933 t yr−1) of sediment and 9 m3 yr−1 (12 t yr−1) of OM would have been released annually by permafrost-related landslides.

5.3 Hydrological Connectivity

[41] The drainage (i.e., stream and channel) density increased in the selected subbasin from 13.2 in 1957 to 15.6 in 2009 (+18%) (Figure 7).

Figure 7.

Change in drainage density caused by lithalsas and palsas decay in land system 24.

[42] In 2009, 0.24 km2 of the 1.09 km2 covered by thermokarst ponds were connected to the main stream. 0.09 km2 of the new 0.54 km2 of water coverage, resulting from the pond expansion between 1957 and 2009, were connected to the drainage network (NCTP) (Table 3).

5.4 Potential TSS, DOC and OM flow in the System from CTPs

[43] From mid-June to mid-October, 21 rain events that increased water levels occurred in 2009 and 22 in 2010 (Figure 6). The cumulative rainfall associated with these events is 279 mm in 2009 and 402 mm in 2010. Assuming that CTP coverage stayed roughly the same in 2009 and in 2010, i.e., 0.24 km2 for CTP and 0.09 km2 for NCTP, transfers through the network of thermokarst ponds and connectors into the fluvial system are estimated at 1.01 t in 2009 and 1.46 t in 2010 for TSS and 0.34 t (2009) and 0.49 t (2010) for DOC Out of these amounts, 0.37 t in 2009 and 0.54 t in 2010 for TSS and 0.12 t and 0.18 t for DOC come from recent thermokarst ponds formed between 1957 and 2009 (Table 7).

Table 7. Yields of Sediment in Suspension and Dissolved Organic Carbon in Tons From Connected Thermokarst Ponds to the Drainage System, in 2009 and 2010
 Surface Area (km2)Cumulated Rain (m)Water yield (m3)aTSSb Yield (t)DOCc Yield (t)
  1. a

    The volume of water which overflows from the thermokarst ponds and joins the fluvial system is equal to the surface area of CTPd or NCTPe multiplied by the height of the cumulated rain falling between mid-June to mid-October 2009 and 2010.

  2. b

    Total suspended sediment.

  3. c

    Dissolved organic carbon.

  4. d

    Connected thermokarst pond.

  5. e

    Connected thermokarst pond created between 1957 and 2009.


[44] In 2009, CTP resulting from the collapse of palsas covered 21,510 m2 more than in 1957 (Table 3). Assuming that this area had an average peat thickness of 1 m with a dry bulk density of 0.1 t/m3 [Robinson and Moore, 1999; Vitt et al., 2000b], a total of 21,510 m3 equivalent to 2151 t of peat has been potentially mobilized into the drainage network as OM, i.e., POM and DOM, during the 52 years period. Of that amount, 96% was released from the forest tundra area. Assuming a constant rate, 41 t/year of OM have been released.

6 Discussion

6.1 Toward a Permafrost-free Landscape

[45] In the Sheldrake River catchment, the area covered by lithalsas and palsas has decayed by 21% and the area covered by thermokarst ponds has increased by 96% in 52 years (Table 8). But these average values conceal important differences within the catchment area. Permafrost has degraded roughly at the same rate (23%) as was observed for riparian palsas 150 km to the north near the tree line by Vallée and Payette [2007], but sensibly less than a palsa peatland to the south, near the Lac Guillaume Deslile, where palsas decreased by 83% between 1957 and 2003 [Payette et al., 2004] or lithalsas 8 km south of the study area, i.e., 40% [Fortier and Aubé-Maurice, 2008].

Table 8. Permafrost and Thermokarst Pond Changes at Different Sites in Northern Québec From References
StudiesLocationAreaType of PermafrostPeriodChange in Permafrost Extent (%)Change in Thermokarst Pond Extent (%)
  1. a

    Nine sites around the village of Umiujaq.

  2. b

    Estimate from authors results.

Payette et al. [2004]

56°11′N/75°55W0.15 km2palsa peatland1957–2004−8348 b
Fortier et Aubé-Maurice [2008]56°33′N/76°30′Wa2.25 km2lithalsas1957–2005−40175
Marchildon [2007] (Nastapoka)56°51′N/76°15W3 × 1.5 km2lithalsas and palsas1957–2005−3679

Vallée and Payette [2007]

57°45N/76°20′N14 landformslithalsas and palsas1957–2001−2376
This Study
The Sheldrake River watershed56°37′N/76°32′W76 km2lithalsas and palsas1957–2009−2196
Forest tundra56°37′N/76°32′W43 km2lithalsas and palsas1957–2009−43103
Shrub tundra56°37′N/76°32′W33 km2lithalsas1957–2009−773
Table A1. Detailed Statistics on Permafrost Coverage in 1957 and 2009, and Changes
  Surficial DepositsLithalsa Coverage in 2009Lithalsa Coverage in 1957Change in Lithalsa CoverageaPalsa Coverage in 2009Palsa Coverage in 1957Change in Palsa CoverageaTotal Permafrost Changea
ZoneBiomeSquare MeterSquare MeterPercentSquare MeterPercentbSquare MeterPercentSquare MeterPercentbSquare MeterPercentbSquare MeterPercentSquare MeterPercent
  1. a

    Between 1957 and 2009.

  2. b

    Percentage of permafrost coverage in the area covered by surficial deposits.

Silty Clay Deposits
Total 14,458,6691,946,394142,794,47019−848,076−3039,1590.3129,3790.9−90,220−70−938,296−32
Peat Deposits
Total 676,36215,066251,1878−36,121−7164,9259.6113,99816.9−49,073−43−85,194−52
TOTAL 21,212,8154,180,88619.75,184,84624.4−1,003,960−19.4112,9250.5253,4281.2−140,503−55−1,144,463−21
Table A2. Detailed Statistics on Thermokarst Ponds Coverage in 1957 and 2009, and Changes
 TPa in 1957TPa in 2009TPa Change 1957–2009CTPb in 1957CTPb in 2009CTPb Change 1957–2009CTPPc in 2009
ZoneSquare MeterPercentdSquare MeterPercent%dSquare MeterPercentSquare MeterPercentdSquare MeterPercentdSquare MeterPercentSquare Meter
  1. a

    Thermokarst pond coverage.

  2. b

    Connected thermokarst pond coverage.

  3. c

    Area covered by connected thermokarst ponds coming from palsa collapse.

  4. d

    Percentage of thermokarst pond coverage in the area covered by surficial deposits.


[46] The Hudson Bay is the principal factor driving the east–west climate gradient characterized by colder coastal temperatures than inland [Lévesque et al., 1988]. The colder coastal zone is generally characterized by cooler summers due to frequent marine fog and high cloudiness and by cold winters due to strong and dry winds once the bay is frozen. The inland climate, which corresponds to the forest tundra at ~15 km from the coast (Figure 8), is warmer and characterized by a thicker and less dense snow cover, particularly in forest stands. As a result, permafrost is patchier and decays more rapidly inland, east of the tree line, in the forest tundra than in the coastal shrub tundra.

Figure 8.

Gradient of percentage of permafrost degradation between 1957 and 2009 from the Hudson Bay landward, with position of the tree line. Each point corresponds to one land system. Sandy areas are excluded.

[47] Several environmental factors such as vegetation and surficial geology are responsible for the observed differences in thawing rates and degradation features within the Sheldrake River catchment. On this 76 km2 watershed, the range of degradation varies from 1 to 100% depending on distance of Hudson Bay, and on topographic, hydrologic, ecological, and geomorphic factors. Palsas have decayed less rapidly than lithalsas, likely because thermal properties of peat help slow down the permafrost thawing [Vitt et al., 1994].

[48] Pond coverage expansion follows permafrost decay. However, the rate of pond formation is less than the rate of permafrost decay. Whereas 1.14 km2 of permafrost has disappeared since 1957, only 0.54 km2 of thermokarst ponds has formed over the same period. Explanations are as follows: first, a pond replacing a lithalsa has a smaller diameter than the original mound due to the preservation of the rim ridge. For example, a 196 m2 typical lithalsa (50 m in diameter) was replaced by a 152 m2 pond with a 44 m2 rampart (3 m in width). Second, terrestrialization of thermokarst ponds due to colonization by sedges and mosses can reduce the rate of thermokarst pond coverage [Zuidhoff and Kolstrup, 2000; Payette et al., 2004]. This is likely the reason why the area covered by ponds in peatlands is proportionally less extensive than in the rest of the catchment. Third, permafrost fragmentation, talik formation, and subsequent drainage can lead to shrinking lakes and disappearance of thermokarst ponds through rapid (even catastrophic) drainage [Yoshikawa and Hinzman, 2003; Smith et al., 2005; Jones et al., 2011; Sannel and Kuhry, 2011]. Finally, in well-drained terrain, evidence of permafrost disappearance can be associated with vegetation growth (shrub) in hollows, while in wet areas or when a lithalsa or a palsa collapses on the banks of the river, a stream, or a lake, it merges with it and is not considered anymore as a thermokarst feature. On the other hand, ponds can also merge to form larger thermokarst lakes [Seguin and Allard, 1984].

[49] As mentioned by Vallée and Payette [2007], the area covered by thermokarst ponds increased more rapidly than the number of new ponds (96% versus 50%), indicating that a number of ponds in 1957 were associated with non-totally thawed permafrost mounds or plateaus. Since then, the lithalsa or palsa has totally disappeared and the pond occupies a larger area. However, the increasing number of lithalsa landforms and their decreasing individual average size in the forest tundra area indicate a general fragmentation of the permafrost.

[50] In 1957, thermokarst ponds covered 2.6% of surficial deposits for a cumulated area of almost 0.56 km2. Permafrost was then slowly degrading [Allard and Seguin, 1987b]. The 1.4°C warming between the end of the LIA and the 1940s in Northern Québec [Chouinard et al., 2007] is likely responsible for this early degradation. Permafrost probably reached its maximum expansion during the LIA and then started to decay [Allard et al., 1987; Payette et al., 2004; Arlen-Pouilot and Bhiry, 2005]. After a short and light cooling between 1940 and 1990, while permafrost continued to thaw essentially due to a rise in snow precipitation [Payette et al., 2004], air temperatures have increased suddenly by 1.7°C until 2005 [Chouinard et al., 2007]. In 2010, an average annual air temperature of 0.0°C was recorded at Umiujaq, whereas the long-term average of the annual air temperature noted by Environment Canada (1925–2010) is −5°C. This supports observations of a recent acceleration of the permafrost degradation processes in the forest tundra biome [Payette et al., 2004]. In areas where the losses exceeded 50% between 1957 and 2009, permafrost would have completely disappeared if the rate of degradation would have been continuous since the end of the LIA. Land system 31 is now devoid of permafrost. Considering that the increase of precipitation is predicted to persist and that temperature will continue to warm up by about 3 to 4°C until 2050 [Allard et al., 2007; Intergovernmental Panel on Climate Change, 2007; Brown et al., 2013], it is anticipated that most of the permafrost shall disappear in the forest tundra in the coming decades. Only fossil thermokarst features will remain.

6.2 Importance of the Activity of Landslides, Active Layer Failures, and Gullies

[51] Landslides are induced by fluvial erosion, which creates active layer slumping in the ice-rich permafrost and accelerates thaw and thickening of the exposed active layer. In summer, clay liquefaction leads to failure of the “newly exposed active layer.” The headwall recedes by slumping and remains parallel to the original slope angle. However, these landslides do not expose the underlying permafrost since the active layer on the slope is too deep. The landslides are concentrated along steep river banks, particularly in areas of permafrost plateaus, in the western part of the studied basin (see photograph on Figure 5). The released sediments are rapidly evacuated by river flow. No ground ice exposure was ever observed despite the fact that drilling only a few meters behind the cliff head revealed abundant ground ice [Allard et al., 1996].

[52] Most of these landslides show signs of high activity. In summer 2009 and 2010, the four landslides showed blocks of turf and clay that had recently slid from the top to the bottom of the slope. The smooth and water-saturated surface of the headwall and the absence of vegetation cover also indicate that erosion is active. A new landslide occurred during the thaw period (June) of 2010 (Figure 3f). The three landslides having the gentlest slopes (25–27°) were already active in 1957. Three slides now appear to be entering in their stabilization phase as their slope is becoming too gentle to trigger detachment, and the permafrost table has retreated several meters from the top of the headwall which is rapidly colonized by shrubs.

[53] Active layer failures occur on the steep (35–40°) upper edge of permafrost plateaus (Figure 3d). They are scattered throughout the study area. In 1957, three of them were already active, three were absent, and two were covered by snow. The two failures visited in 2009 show limited activity. As an active layer failure is triggered by annual thawing and since the slide is confined into the top part of the active layer, as noticed by Seguin and Allard [1984], the erosion form is maintained active by two processes. First, melting of snow at the end of spring and rain in summer oversaturate the surface layer of the soil and lead to liquefaction of clay, which flows on the slope. Second, frost boils oriented across the slope continuously bring liquefied or viscous fine sediments to the surface and can slide and trigger small active layer failures (Figure 3b).

[54] Gullies are widespread landforms in the watershed, either singly or as a network. Generally, gullies have symmetric slope angles along the edges of two elongated plateaus. They deepen and widen during rain events. The slope of the permafrost plateau edge increases (40–45°) leading to skin failures, which grow and deepen year after year. Eighty percent of the listed gullies were already present or covered by snow in 1957. Several former gullies now (and in 1957) colonized by a thin layer of vegetation show that incision between plateaus is not a recent process. As for active layer failures, it is likely that these erosional landforms were present in the basin since the LIA. It is also possible that some plateaus have been divided in several smaller individual mounds by gullying. Nevertheless, the gullies are still very active and should continue to erode and export significant volume of sediments during future rain events. Almost two gullies out of three were covered by snow in 1957. This suggests that snow accumulation between plateaus reduces frost penetration in winter and contributes to trigger gully formation, as emphasized in several studies on palsa collapse [Seppälä, 1990, 1994; Matthews et al., 1997] and degradation of permafrost plateaus [Allard et al., 1996].

6.3 Increase of Sediments and OM Released by Landslides, Active Layer Failures, and Gullies

[55] The higher number of landslides observed in 2009 (+46 to +217%) suggests that the overall geomorphic activity on slopes and in gullies has increased, resulting in higher volumes of eroded sediments in 2009 than in 1957 (+12 to +38%). It is in accordance with several other studies, which noticed an increasing number of active thaw slumps in the Mackenzie River delta [Lantz and Kokelj, 2008], in Hershel Island [Lantuit and Pollard, 2008] and on the Aklavik Plateau [Lacelle et al., 2010]. However, it seems that the small size of landslides and gullies results in small inputs in the Sheldrake River fluvial system. The total volume of eroded sediments in the catchment from the 38 erosion landforms related to permafrost disturbances is only 35,652 m3. For comparison, changes in gully size in the only nonpermafrost complex of erosion in the center of the basin (land system 16) suggest a loss of 26,600 m3 of clay. The only study in Northern Québec that quantifies landslide scars reveals that seven rotational landslides that occurred in unfrozen clay of the Great Whale River estuary have exported 17 × 106 m3 of sediments [Bégin and Filion, 1987]. Thaw slumps in continuous permafrost can affect several thousands of square meters of terrain [Lantuit and Pollard, 2008]. Therefore, landslides induced by permafrost decay in the Sheldrake River basin are a small source of sediment in comparison with erosion landforms and mass wasting assessed elsewhere, both in permafrost and in nonpermafrost areas.

[56] The volume of lost permafrost terrain in about 50 years inevitably means that sediment and carbon amounts released in the fluvial ecosystem have increased. It is likely that a part of these products is temporarily stored in the aquatic system, as bottom sediments in ponds and streams. However, eroded sediment and carbon will ultimately reach Hudson Bay.

6.4 Impact of an Increasing Connectivity on Organic Carbon and Mineral Sediment Yields

[57] The important increase in catchment internal connectivity, reflected by increases of 58% in CTP and 18% in stream density (Figure 7) from 1957 to 2009, represents a major environmental and hydrological change. Despite large methodological uncertainties, degradation of permafrost and the development of thermokarst ponds currently release a significantly higher amount of DOC and TSS in the fluvial ecosystem than previously. No comparison with other studies is available yet to our knowledge.

[58] This increasing connectivity also results in a more efficient drainage network and a fragmentation of the remaining permafrost. Consequently, in discontinuous permafrost, as in the continuous permafrost [Rowland et al., 2010], recent climate warming has altered the water balance. Thermokarst was already occurring in 1957. Results clearly indicate that the rate of permafrost decay has recently increased in the studied area, probably since 1990, as observed elsewhere in the region [Payette et al., 2004; Fortier and Aubé-Maurice, 2008]. This increasing rate of permafrost degradation probably occurred following the significant increase of air temperature in northeastern Canada during the 1990s [Chouinard et al., 2007].

[59] CTPs in the forest tundra area export 2.3 times more TSS and DOC per unit of area than the CTPs in the shrub tundra area. This amount increases to a factor of 2.6 for NCTPs. The first explanation is likely the difference in the ratio of the CTP coverage per km2 that is 3.6 times superior in the forest tundra. Conversely, landslides, active layer failures, and gullies are more abundant in the shrub tundra area. They currently contribute 70% of the total sediment and carbon released by landslides in the river catchment. Therefore, the tree line is a major boundary between two regimes of thermokarst processes.

[60] The measurement of CTP inputs based on rain events was done in summer conditions only. Thus, this approach does not include snowmelt and breakup during which ponds overflow and hydrological connectivity is temporarily higher [Bowling et al., 2003; Woo and Guan, 2006]. An important fraction of annual POM and DOC is likely transferred through the fluvial system during the melting period. For example, in the Great Whale River, 100 km south of the area, 23% of annual POC and 11% of annual DOC are exported during the breakup [Hudon et al., 1996].

[61] The primary source of old POM in the basin is the peat released from the decay of palsas which have degraded by 55% since 1957 [Tarnocai et al., 2009; Kuhry et al., 2010]. Not all of the released OM from palsa decay may reach the fluvial network. However, it is likely that a nonnegligible fraction of the eroded peat, which collapsed in CTP, is temporarily stored in the ponds to be thereafter exported in the river by overflow during rain events.

7 Conclusion

[62] Permafrost has considerably degraded east of Hudson Bay, as elsewhere in the discontinuous zone in circum-arctic regions. Palsas and lithalsas of the Sheldrake River catchment have generally decayed by 21% between 1957 and 2009. The rate of thaw was much higher in the catchment head (forest tundra area) than near the basin outlet on Hudson Bay coast (shrub tundra area). The tree line is a definite climate and ecological boundary that significantly impacts on the rate of permafrost decay and on the volume of sediment and carbon yields. Climatic projections for the next decades indicate that this trend will continue until permafrost disappears.

[63] Increase in activity of landslides and in thermokarst pond cover between 1957 and 2009 resulted in significant sediment and carbon release in the fluvial system. However, the total computed amounts of currently released sediments and carbon from the Sheldrake River basin remains small compared to the Great Whale River, a permafrost-free catchment where landslides occurred in the same type of clay deposits. This illustrates that beyond a certain stage of degradation (now 20% left above the tree line and 10% below), the impact of residual permafrost degradation on sediments and carbon release in fluvial systems declines quickly.

Appendix A

[64] Details of the statistics summarized in Table 2. Each land system is listed with corresponding biome (forest tundra or shrub tundra), area in surficial deposits (m2), area (m2) and percentage of surficial deposits covered by lithaltas, palsas, and total permafrost (lithalsas and palsas) in 1957 and 2009. Changes in lithalsas, palsas, and total permafrost coverage in area and percentage, between 1957 and 2009, are also given. Sandy areas are excluded.

[65] Details of the statistics summarized in Table 3. Each land system is listed with corresponding coverage (m2 and percentage of surficial deposits) of thermokarst ponds and connected thermokarst ponds in 1957 and 2009. Changes in thermokarst ponds and connected thermokarst ponds coverage in area and percentage, between 1957 and 2009, are also given. Areas covered by connected thermokarst ponds originating from palsa collapse are also shown. Corresponding biomes and surficial deposits can be found in Table A2.


[66] This project was supported by grants from Arcticnet and the Natural Science and Engineering Research Council of Canada (discovery grant to M. Allard). The Centre d'études nordiques of Université Laval provided important logistical support. We thank Denis Sarrazin for the installation and maintenance of all the data-gathering equipment in the study region. The field assistance of Catherine Falardeau-Marcoux and the help in mapping by René-Charles Bernier, Emmanuel L'Hérault, and Carl Barrette were appreciated. The suggestions of Warwick Vincent and Mickael Lemay on a preliminary version of the manuscript are appreciatively acknowledged. The paper was significantly improved by the reviewing process. We are also grateful to the Inuit community of Umiujaq for its generous hospitality.