Ground ice, organic carbon and soluble cations in tundra permafrost and active-layer soils near a Laurentide ice divide in the Slave Geological Province, N.W.T., Canada

The central Slave Geological Province is situated near a divide of the Laurentide Ice Sheet and it differs from the western Canadian Arctic, where thaw-induced landscape changes in Laurentide ice-marginal environments are already abundant. Although much of the terrain in the central Slave Geological Province is mapped as predominantly bedrock and ice-poor, glacial deposits of varying thickness occupy significant portions of the landscape, creating a mosaic of conditions. Some evidence of ice-rich ground, a key determinant of thaw-induced landscape change, exists. Carbon and soluble cation content in 5 permafrost are largely unknown in the area. Twenty-four boreholes with depths up to ten metres were drilled in tundra north of Lac de Gras to address these regional gaps in knowledge and to better inform projections and generalizations at coarser scale. Excess-ice contents of 20–60 %, likely remnant Laurentide basal ice, are common in till and thaw subsidence of metres to more than ten metres is possible. Beneath organic terrain and in fluvially-reworked sediment, aggradational ice is found. The abundant ground-ice poses long-term challenges for engineering, and it makes the area susceptible to thaw-induced land10 scape change and mobilization of sediment, solutes and carbon several metres deep. The characteristics of landscape changes, however, are expected to differ from ice-marginal landscapes of western Arctic Canada, for example, based on subsurface properties. Average soil organic-carbon storage is approximately 8 and 14 kg C m−2 for the depth ranges 0–1 m and 0–3 m. The concentration of total soluble cations in mineral soils is much lower than at other previously studied locations in the western Canadian Arctic. 15 Permafrost in the study area contains much more ground ice than expected, and slightly less organic carbon and fewer soluble cations than well studied areas in the western Canadian Arctic. As these differences are strongly related to geology and glacial history, this study may inform investigations in other parts of the Slave Geological Province and its data can support scenario simulations of future trajectories of permafrost thaw at continental and circumpolar scales.

. Location of the study area in the context of permafrost zones in Canada (modified from Heginbottom et al. 1995). Yellowknife is the closest city.
The area is characterized by low relief where irregular bedrock knobs and cuestas form hills up to 50 m high (Dredge et al., 1999). Located near the ice divide of the Keewatin sector of the Laurentide Ice Sheet, it generally is a source area for sediments, unlike ice marginal locations. The spatial abundance of surface materials is derived from the surficial geology map 1:125,000 (110-112 • W, 64-65 • N) for Lac de Gras (NTS 76-D, Geological Survey of Canada, 2014). The northern part is dominated 95 by till deposits, whereas the southern half consists more prominently of bedrock (8 %) with patches of till (Hu et al., 2003).
Numerous eskers and outwash complexes (1 %), mostly composed of sand and gravel, are found in the area (Dredge et al., 1994). Till deposits are differentiated by their estimated thickness into till veneer (<2 m thick, 21 %), till blanket (2-10 m thick, 24 %), and hummocky till (5-30 m thick, 3 %). These deposits typically have a silty sand to sand matrix with low percentages of clay and 5-40 % gravel (Wilkinson et al., 2001). Organic material covers 5 % and lakes 38 % of the area. 100 Soils consist of till, glacio-fluvial sediments, or peat. Upland till surfaces are characterized by earth hummocks and organic material, visible to depths of up to 80 cm, that has been redistributed within the active layer by cryoturbation (Dredge et al., 1994). The tills derived from granitic and gneissic terrain have a silty or sandy matrix, whereas those derived from metasedimentary rocks contain a higher silt-clay content (Dredge et al., 1999). Low-lying areas are mostly comprised of colluvium or alluvium rich in organics and wet areas often have peatlands (Karunaratne, 2011). 105 The area is in continuous shrub tundra (Wiken et al., 1996) and common shrubs include northern Labrador tea (Rhododendron tomentosum) and dwarf birch (Betula glandulosa), while bog cranberry (Vaccinium vitis-idaea) and dwarf bog rosemary (Andromeda polifolia) often comprise the understory (Karunaratne, 2011). Well-drained upland areas are typically covered with a thin layer of lichens and mosses (Hu et al., 2003) (Figure 2A). Grasses and sedges with a ground cover of moss comprise the vegetation cover in valleys ( Figure 2B) and some poorly-drained low-lying areas have thick peat associated with ice-wedge polygons and sedge meadows (Hu et al., 2003;Karunaratne, 2011) ( Figure 2C). Frequently, low-lying areas have tall shrubs along small streams and at the rise of steeper slopes. Esker tops have little vegetation and are often comprised of exposed soil ( Figure 2D).

115
Soil cores with a diameter of 5 cm were obtained using a diamond drill (Kryotek Compact Diamond Sampler), sectioned into 20 cm intervals and logged (soil texture, colour, ice content and visible organic matter) in the field while still frozen (Subedi, 2016;Gruber et al., 2018a). Two soil pits were excavated within approximately 10 m of each borehole to describe and sample typical near-surface soil conditions. The depth of thaw at the time of sampling was estimated by probing. Drill core and pit samples were double-bagged for thawed shipment to the laboratory in Yellowknife. 120 5 https://doi.org/10.5194/tc-2020-33 Preprint. Discussion started: 10 February 2020 c Author(s) 2020. CC BY 4.0 License.

Laboratory
All samples were thawed and processed at ambient temperature. Samples were homogenized, poured into beakers, weighed, and allowed to settle for 12 h (cf. Kokelj and Burn, 2003). Volumes of sediment V s and supernatant water V w were recorded to estimate volumetric excess ice content (%) of the permafrost samples as known amount of deionized water was added (1:1 extraction ratio; Janzen, 1993). These samples were mixed thoroughly and then allowed to settle for 12 h. Water was collected with a syringe and filtered through 0.45 µm cellulose filter paper. The remainder of the sample was dried for 24 h at 105 • C to determine the gravimetric water content (%), expressed on a dry basis (GW C d ) and on a wet basis (GW C w ) (cf. Phillips et al., 2015).
The concentration (mg/l) of the soluble cations Ca ++ , Mg ++ , Na + and K + was determined by atomic adsorption spectropho-135 tometer at the Taiga lab in Yellowknife. Measured soluble ion concentrations C m (mg/l) were converted to an expression E using milli-equivalents per unit mass of soil (meq/ 100g of dry soil) where M e is the equivalent mass of ions (g) and M 100g w is the mass of water per 100 g of dry soil as present in the sample at the time of water extraction. Presentation of soluble cation concentrations per unit weight of dry soil facilitates comparison 140 between samples of varying moisture contents.
Organic-matter content LOI (%) is expressed on a gravimetric dry basis and was determined using the sequential loss-onignition method (Sheldrick, 1984) at Carleton University. A small amount (2-3 g) of the homogenized and oven dried sample (< 0.5 mm soil fraction) was placed in a crucible and heated to 550 • C for 6 h to determine the organic-matter content as

145
where M 105 S is the mass of sediment after oven drying at 105 • C, and M 550 S is the mass of sediment after ignition at 550 • C.
To avoid combustion problems, reduced amounts (0.5-1 g) were processed when samples consisted of plant residue with very little visible mineral soil. When no mineral soil component was visible after coarse components were removed, samples were not processed and an LOI of 80 % was estimated. This occurred only in the top metre and almost exclusively in samples from soil pits. The gravimetric percentage (P 0.5 ) of the <0.5 mm soil fraction has been lost from the original analysis. Later, this 150 was determined again based on dry sieving for 183 of 357 samples.
Data quality was assessed in a second analysis on the samples using the same procedures and tools as during the original processing. Based on measured blanks, the accuracy is about 0.03 % LOI, the median accuracy based on doubles is 0.04 % LOI with the highest difference being 0.30 % LOI.

160
following Hossain et al. (2015), who conducted their study in geologic settings similar to the project area. This resulted in an estimated DBD for the fine-grained soil, i.e. excluding the volumes V ei and V c . To account for this, soil organic-carbon density (SOCd, kg C m −3 ) was derived as and finally applied as average values over depth intervals within each terrain type to obtain SOCs. For the samples without Detailed grain-size distribution was measured on selected samples using a Beckman Coulter LS 13320 laser-diffraction particle-size analyzer. Samples were first oven-dried at 105 • C and then crushed and homogenized with a mortar and pestle.
Samples were then passed through a 2 mm sieve to remove the coarse fraction that was then weighed. Organic matter was removed from the fines using hydrogen peroxide. The samples were then mixed with Calgon to prevent flocculation and passed 180 through the particle-size analyzer. Results were classified according to the USDA textural classification system (2 mm > sand > 53 µm > silt > 2 µm > clay).
4 Field observation and sampling

Study sites
Four terrain types ( Figure 2) comprised of upland tills, fluvially reworked till (the Valley), organic terrain and eskers were 185 sampled with drill cores and soil pit at 24 locations (Table A1).
Upland tills: Ten boreholes were sampled to depths of 2.5-9.5 m in smoothly rounded hills comprised of thick till and in till veneer over bedrock. The dominant plant species were dwarf shrubs, Labrador tea and grasses. Thaw depths were about 2 m on hill tops and nearly a meter at the bottom of hills.
The Valley: Eight boreholes were drilled to depths of 1-6 m in a gently sloped valley that contrasts with other terrain types 190 because its silts and sands are well sorted and likely derived from fluvial reworking of local tills. Boreholes located on the more elevated sides of the Valley typically had coarser sediments, whereas those near its axis had mostly fine sediments with high silt contents and organics with ice-wedge polygons. Few water logged sites contained tall shrubs with water channels.
Sites were sparsely to moderately covered with plant species such as dwarf birch, Labrador tea and grasses. Thaw depths were 35-40 cm. Eskers: Four boreholes were drilled to depths of 1.5-12 m at hilltop locations with sparse vegetation or exposed soil.

Soil texture
Most soils consisted of poorly to very poorly sorted silt and sand. The relative proportion of silt was high in samples from mineral soils beneath organic terrain and in valley bottom sites with average values exceeding 40 %. Clay content was low and always below 20 %.

Ground ice 205
Field logged visible-ice content is available for 113 core sections. The average, weighted by the length of core sections, is 24 %.
Laboratory analyses show that water and excess-ice contents increase progressively with depth in till. Zones of high moisture content ( Figures 3A and 4A) were often associated with ice lenses, several centimeters thick (e.g., Figures B1 and B3). Excessice content greater than 50 % in till became increasingly common below 4 m depth. In organic terrain, high moisture content (>80 %) but low excess-ice content in permafrost reflect saturated organic soils with low bulk density ( Figure 3B). The sharp high moisture content near the surface, where organics were present, and deeper down ( Figure 3C) due to 20-50 % excess ice in mineral soil ( Figure 4C). In eskers, water content was mostly below 20 % and pore ice the dominant ground-ice type ( Figure   3D).

Organic carbon
Organic-carbon density in the active layer was typically greater than at depth in permafrost ( Figure 5). Statistics of soil organiccarbon density and storage are given for consistent depth intervals and the four terrain types in Table 1. The spatial average of soil organic-carbon storage in the study area, accounting for the abundance of bedrock and lakes, is estimated as 8 and 14 kg C m −2 for the depth ranges 0-1 m and 0-3 m, respectively.

Total soluble cations
The concentration of soluble cations in organic-rich, shallow soils were mostly higher and more variable than those in mineral permafrost soils at depth ( Figure 6). In organic materials, the concentration of soluble cations near the top of permafrost was relatively high ( Figure 5B and 6B). In till, soluble cation concentrations, as with ice content, increased gradually with depth ( Figure 4A and 6A). Differences between active layer and permafrost, as well as between organic and mineral soils are all significantly (p < 0.01) different from each other based on Kruskal-Wallis tests. Although the dry bulk density of organic soil is lower than that of mineral soil, these patterns persist even when expressed relative to wet soil mass. of the drill barrel caused partial or complete thaw of the core. Depending on the degree of thaw and core composition, this 235 resulted in intervals erroneously shown with reduced or no excess ice content ( Figure 4C). The results are, therefore, likely to be conservative (low biased) estimates of excess ice and gravimetric water content at the locations sampled. The difficulty of drilling though large clasts, on the other hand, may have caused bias towards sampling locations with higher contents of excess ice and of fines. When drilling organics within polygon networks, the drill rig was placed on polygon centres. As a consequence, wedge ice, which is known to be present in the area based on the surface expression of polygon networks, is 240 systematically avoided in sampling and, therefore, largely excluded from the present quantitative data and interpretation.
While this study was not designed to elucidate the origin of ground ice, a number of observations merit discussion. In organic terrain, the excess ice recovered resembles pool ice (clear with small bubbles and embedded peat filaments) and wedge Table 1. Soil organic-carbon density (SOCd, kg C m −3 ) per depth interval and soil organic-carbon storage (SOCs, kg C m −2 ) for the four terrain types investigated. SOCs is based on average SOCd accumulated from the surface down to the specified maximum depth. Ranges in parentheses indicate minimum and maximum values rounded to the nearest integer, the number of samples is indicated in square brackets. ice (foliated with bubbles and some sediment, Figure B2 panels B and C) as previously reported for polygonal ground in organics (Mackay, 2000;Morse and Burn, 2013). In the Valley and in organic terrain, the increase in water and excess-ice  Hu et al., 2003) and show geomorphic evidence of melt-out (Prowse, 2017(Prowse, ). 1988St-Onge and McMartin, 1999). Both, ground ice characteristics and geomorphic features suggests that a large proportion of the excess ice in this hummocky till is Laurentide basal ice preserved beneath ablation till.

Organics
The volumetric contents of excess ice encountered in mineral soil were often 20-60%. As a first-order estimate, this implies that complete thaw of permafrost can cause about 0.2-0.6 m of subsidence for each vertical meter of permafrost in till. The boreholes in hummocky till, which is estimated to be 10-30 m thick in the area (Haiblen et al., 2018) show an increasing trend 270 of excess ice content with depth, based on our limited sampling to 9 m, alone. A potential surface lowering of many meters, up to more than ten meters, is thus to be expected from areas of thick till if this permafrost was to thaw completely. This includes the potential for thermokarst processes to mobilize sediments, solutes and organic carbon at depth more quickly than expected in strictly conductive one-dimensional thaw. A number of geomorphic features reminiscent of retrogressive thaw slumps (Fig.   ??) and the presence of kettle lakes (Prowse, 2017) in the area both exhibit local relief that suggest meltout of massive ice 275 several metres in thickness.

Organic carbon
Terrain variation in organic-carbon density and in organic matter content in fine material occur in association with surficial material and topographic setting. Organic terrain is frequently characterized by peat deposits up to 2.5 m thick and associated with low lying poorly drained portions of the landscape. In other terrain types, organic materials may have become vertically 280 redistributed in the top few metres of soil profiles by cryoturbation (Dredge et al., 1994;Haiblen et al., 2018) and by burial during permafrost aggradation due to colluviation/alluviation (Kokelj et al., 2007). The low organic-carbon density at depth likely indicates the absence of sediment reworking and permafrost preservation in tills during the Holocene.
Mean organic-carbon density in the top 3 m of soil profiles near Lac de Gras is about half that reported in recent circumpolar statistics (Table D1). This is similar to the mean of about 12 (kg C m −3 ) for the top 1 m in the northern Canadian Arctic and 285 its difference to about 30 (kg C m −3 ) in the southern Canadian Arctic reported by Hossain et al. (2015, Fig. 5D). A recent circumpolar compilation of permafrost carbon data (Hugelius et al., 2014) estimated SOCs for the study area to be 5-15 (0-1 m) and 15-30 (0-3 m) kg C m −2 . These values are similar (0-1 m) and slightly higher (0-3 m) than the spatially-averaged results of this study, that include zero SOCs in bedrock and in lakes. The low organic-carbon density in the study area, especially at depth, is interpreted to derive from the short duration of Holocene carbon accumulation following at least partial evacuation of  (Koven et al., 2015), corresponding data (Hugelius et al., 2014;Tarnocai et al., 2009) is rare.

Total soluble cations
In mineral soil, the lower concentration of total soluble cations in the active layer compared with permafrost is interpreted to be caused by leaching of ions from unfrozen soil, and is similar to observations in other regions (Table D2). Additionally, the 295 redistribution of ions along thermal gradients during freezing may have caused solute enrichment during the development of segregated ice ( Figure B1 and B2) in aggrading permafrost (cf. Cary and Mayland, 1972;Qiu et al., 1988) in mineral soils of the Valley and beneath peat in organic terrain. There, zones of increased cation concentrations at depth corresponded with ice-rich intervals in permafrost, especially at sites in till and in organic terrain (Figure 4 and 6). Where high amounts of organic matter are present, also the concentration of total soluble cations is high. As a consequence, mineral-soil permafrost has lower 300 concentrations of soluble cations than organic active-layer soils but higher concentrations than mineral active-layer soils.
The absolute concentrations of soluble cations obtained in the study area near Lac de Gras are low compared to previous studies from northwestern Canada that report higher concentrations in active layer and permafrost across diverse terrain types (Table D2). In the Mackenzie Delta, alluvial materials derived from sedimentary and carbonate rock of the Taiga plain and regular flooding produce solute rich active layer and permafrost deposits. A range of forest-terrain types contained more 305 soluble cations, often several times higher, in the active layer and permafrost than the mineral soils in this study. Also in comparison with undisturbed terrain on Herschel Island, the absolute concentrations of soluble cations in our study are low.
Sediments on Herschel Island are silty-clay tills that include coastal and marine deposits excavated by the Laurentide Ice Sheet (Burn, 2017). These materials are frequently saline below the thaw unconformity indicating permafrost preservation of soluble materials below the maximum depth of early Holocene thaw (Kokelj et al., 2002) or their concentration in colluviated materials 310 (Lacelle et al., 2019). The low concentrations in our study area are associated with the contrasting nature and origins of surficial materials. Tills in our study region are generally coarser grained than many glacial deposits studied in the western Arctic, are regionally sourced from mostly granitic rocks and have been exposed only to minor postglacial landscape modification (Haiblen et al., 2018;Rampton and Sharpe, 2014).

315
The research area near Lac de Gras is characterized by a mosaic of terrain types (48 % till, 38 % lakes, 8 % bedrock, 5 % organic material and 1 % eskers and outwash complexes) with a high degree of fine-scale spatial variability in subsurface conditions.
Permafrost there contains much more ground ice, slightly less organic carbon and fewer soluble cations compared with global compilation products or published research from sites in the western Canadian Arctic. This study provides quantitative data in a region with few previous studies and it supports six specific conclusions: 320 1. Excess-ice contents of 20-60 % are common, especially in till and till-derived sediments, and the average field logged visible-ice content is 24 %. This new regional insight improves upon coarse-scale compilations that rate the area north of Lac de Gras as ice poor (O'Neill et al., 2019;Brown et al., 1997;Heginbottom et al., 1995).
2. Thick occurrences of excess ice found in upland tills are likely remnant Laurentide basal ice, and aggradational ice is found beneath organic terrain and in fluvially-reworked till. Gras is about half that reported in recent circumpolar statistics (Hugelius et al., 2014). Estimated areal means, accounting for the abundance of bedrock and lakes, of soil organic-carbon storage are 8 and 14 kg C m −2 for the depth ranges 0-1 m and 0-3 m, respectively, similar to or slightly lower than in a global estimate (Hugelius et al., 2014). landscapes where dramatic transformations are already observed (e.g., Kokelj et al., 2017;Rudy et al., 2017). The characteristics of thaw-driven landscape change, however, are expected to differ from observations in ice-marginal positions due to differences in topography and climate affecting location and timing, geotechnical properties affecting stability and mobility of sediments, and geochemistry affecting solute and carbon release to surface water, ecosystems and the atmosphere.

345
These findings highlight the importance of geological legacy in determining the characteristics of permafrost and the potential responses of permafrost systems to disturbance and climate change. Continued research on permafrost and landscape response to warming at locations in the interior of the Laurentide Ice Sheet will help to understand and predict changes specific to these landscapes and how they affect ecology, climate, land use and infrastructure.
Code and data availability. Drill logs, visible ice content and core photos from the 2015 campaign are published (Gruber et al., 2018a    Appendix D: Tabulated comparison with previous studies Table D1 compares soil organic-carbon densities and Table D2 soluble cation concentrations between this and previous studies. Table D1. Soil organic-carbon density (SOCd, kg C m −3 ) per depth interval for three terrain types from the Lac de Gras study area compared with similar soils reported in a circumpolar compilation (Hugelius et al., 2014, Table 2). Tills are compared to Turbels (cryoturbated permafrost soils), Eskers with Orthels (mineral permafrost soils unaffected by cryoturbation) and Organics with Histels (organic permafrost soils). Circumpolar values below 1 m are for "Thin sediment". For Orthels, values in "Thick sediment" are more than ten times larger.