Ion geochemistry of a coastal ice wedge in northwestern Canada: Contributions from marine aerosols and implications for ice‐wedge paleoclimate interpretations

Ice wedges are a characteristic ground ice feature in permafrost regions that form primarily from the meltwater of the seasonal snowpack. Ice‐wedge oxygen and hydrogen stable isotopes have been used in winter paleotemperature reconstructions; however, until recently, the ion geochemistry of ice wedges has rarely been analyzed as a potential paleoclimate proxy. This potential is greatest for ice wedges located in coastal regions, where marine aerosols are the dominant contributor to snowpack impurities. Here, we evaluate the source and integrity of ionic concentrations of a coastal ice wedge in the northwestern Canadian Arctic (Beaufort Sea coast) to evaluate the use of ice wedges as a marine aerosol archive. Comparison to a regionally comparable snowpack reveals remarkably similar ionic concentrations for Cl−, Na+, Br−, SO42−, Ca2+, and Mg2+, with a Cl−/Na+ ratio similar to bulk seawater (1.80 vs. 1.79 in seawater), suggesting that marine aerosols, probably from sea salt aerosol production during blowing snow events over sea ice as indicated by depleted SO42− values relative to Na+, are probably the dominant contributor to ion concentrations. A previously established linear age model for the ice wedge is used to develop a continuous ion record spanning ~4,600 to ~700 yr b2k. Cl− and Na+ concentrations reveal a strong and continuous increase in concentrations over the late Holocene, thought to be driven by reduced distance‐to‐coast of up to 1 km as a result of coastal erosion. This study presents a novel interpretation of ice‐wedge geochemical data and represents the first Holocene ice‐wedge ion record.


| INTRODUCTION
The preservation of the chemical composition of the seasonal or annual snowpack is a powerful tool in paleoclimatology and forms the basis of interpretation for glacier ice core studies. The chemical composition of the polar snowpack is the cumulative result of the chemical composition of initial snow crystal formation, snowpack water volume fluctuations, contributions from dry deposition and blowing snow, and snow photochemistry. Processes regulating impurity deposition and postdepositional emission (volatility) are specific to each chemical species, and the relationship between species may reflect the source and history of snowpack impurities. 1 Long-term changes in snowpack composition may also be observed due to changes in the moisture source region, regional climate, long-distance transport, atmospheric composition, and sea ice processes, among others.
One of the major contributors to snowpack geochemistry in coastal polar regions are marine aerosols. Marine aerosols, including sea-salts (e.g., Na + , Cl À , Br À , Mg 2+ , Ca 2+ ) and biogenic aerosols (e.g., methanesulfonic acid [MSA À ], I À ), are released from open water regions and sea ice surfaces, with production moderated by sea ice and open water extents (e.g., [2][3][4] ) and first-year sea ice formation (e.g., 5,6 ). In glaciated regions, these aerosols are preserved in glacial firn and ice and can be used with modern climatological data to reconstruct high-resolution records of sea ice variability, marine productivity, and regional atmospheric dynamics, though these studies are geographically limited to a few polar and alpine regions. Recently, studies in nonglaciated regions of the Arctic have utilized the geochemical composition of ice wedges in permafrost regions for winterspecific paleoclimate reconstructions using techniques traditionally applied in ice core analysis (e.g., [7][8][9][10][11][12][13]. Ice wedges are a vertically foliated ground-ice feature characteristic of permafrost regions, formed from the thermal contraction of the ground under rapid and sustained drops in air temperature. 14 Following thermal contraction, meltwater from the seasonal snowpack fills the contraction crack and rapidly freezes on contact, preserving a vertical ice vein of winter precipitation, with minor contributions from hoarfrost and frozen snow. [15][16][17] Repeated thermal contraction cracking and ice vein formation over hundreds to thousands of years results in the development of large ice wedges, preserving the geochemical composition of the winter snowpack (defined for this study as the meteorological winter and spring, DJFMAM) at varying resolutions. 10 Ice wedges have previously been used to reconstruct relative changes in winter temperatures over the Pleistocene and Holocene from the stable isotopes of oxygen and hydrogen in water, capturing a seasonality that is rarely represented in biogenically mediated climate proxies. Since ice wedges are fed primarily by the winter snowpack (e.g., 16,18 ), ice-wedge ion geochemistry may preserve impurity composition of the original snow. While ice wedges may be subject to other impurity sources from the ground surface and encompassing sediments, ice wedges in coastal regions, where ion concentrations in the snowpack are dominantly influenced by sea salt deposition, may offer great potential as a marine aerosol archive.
A recent study from Utqiaġvik (formerly Barrow), Alaska, utilized the ion geochemistry of a coastal ice wedge in conjunction with the isotope composition to investigate the potential of Pleistocene-aged ice wedges to preserve the marine aerosol composition of the winter snowpack. 19 The study found that the integrity of the marine aerosols was preserved, and a sea ice signal may be distinguished in the ion geochemical record. This is significant as regional sea ice reconstructions are limited to a few marine sediment records (e.g., [20][21][22], thus offering new insight into Pleistocene sea ice conditions in the Beaufort Sea. However, multiple considerations are required in the paleoclimate interpretations of individual ice-wedge geochemical records, and few other studies have reported on the ion content in ice wedges (e.g., 7,[23][24][25][26]. Ice wedges are subject to large terrestrial contributions from the ground surface and active layer during the infilling process, with snowmelt potentially washing-in annually accumulated sea salt from the tundra surface as well as soil-derived minerals. 27 Further, the origin of impurities in the snowpack must be considered due to regional differences in aerosol formation, transport, and deposition, lithogenic dust inputs, and postdepositional alteration of marine aerosols due to the physical (e.g., sublimation) and chemical (e.g., volatized compounds) evolution of the late-season snowpack. Thus, more studies are needed to characterize the source of ice-wedge ion content and to better understand regional variabilities.
This study presents an ion geochemical record of an ice wedge from the Tuktoyaktuk Coastlands along the Beaufort Sea coast. An age-distance model based on the radiocarbon dating of dissolved organic carbon (DOC 14 C) and an oxygen isotope (δ 18 O) record for this ice wedge have been described in Holland et al., 9 and used in this study. The marine and lithogenic origin of ions characteristic of major sea-salt components (Cl À , Na + , Br À , SO 4 2À , Mg 2+ , Ca 2+ , and K + ) are investigated. A marine aerosol record is developed based on an updated radiocarbon age model, and the influence of ice-wedge growth processes and changes in geography are considered in the interpretation of the record. This study represents the first Holocene marine aerosol record for the western Canadian Arctic and will aid in future interpretations of ice-wedge geochemical records in coastal regions.

| Field methods
The study site is located in a high-centered ice-wedge polygon near the Pingo Canadian Landmark region in the Tuktoyaktuk Coastlands, The frozen samples were transported in a cooler to the Aurora Research Institute (ARI) and placed in temporary storage at À20 C.
The samples were shipped to the University of Alberta for physical processing and laboratory analysis.
Subsampling was conducted using a Laguna LT14BX bandsaw in a À25 C freezer lab at the Canadian Ice Core Lab (University of Alberta). The outer 2.0 mm of ice on all sides of each field sample were removed to minimize possible contamination from meltwater during transport and storage. These "clean" field samples were then sliced along the plane of foliation into 1-to 1.5-cm subsamples, representing approximately one to three ice veins (i.e., infill events). Subsamples were then transferred to Whirl-Pak™ bags and stored at À20 C until analysis.

| Radiocarbon dating
A chronology of ice formation for the PNM-01 ice wedge was established based on 10 dissolved organic radiocarbon (DOC 14 C) ages from the center of the ice wedge (youngest ice) to the right-side sediment-ice interface (oldest ice), the details of which are described at length in Holland et al. 9 Calibrated age ranges and age probability distributions were calculated using OxCal v4.4 29 based on the IntCal20 calibration curve (Table S3). 30 A point estimate for the ages is given by the median of the highest posterity density region and is reported in calendar years before CE 2000 (yr b2k) (see 9 ). Calibrated ages for the PNM-01 ice wedge indicate a consistent increase in age towards the outer edge of the wedge (i.e., theoretically older veins), which is consistent with a central cracking pattern during ice-wedge growth. Thus, a linear age-distance model was developed from the median ages using CLAM, 31 assuming sequential and linear growth from the center of the wedge following the classic ice-wedge growth model ( Figure 2).

| Stable isotopes of water
The subsamples were divided for separate stable isotope and ion geochemical analysis. The apportioned samples for stable isotope analysis were thawed in clean Whirl-Pak bags at room temperature and immediately filtered with 0. were monitored for contamination. The limit of detection (LOD) and limit of quantification (LOQ) were calculated using the standard error and slope of the calibration curve and are reported in Table S1. Na + is used as a conservative tracer for sea salt, as Cl À can be subject to multiple fractionation processes before and after deposition. [34][35][36] To differentiate between sea-salt and lithogenic sources, the non-sea salt (nss) fraction of K + , Ca 2+ , Mg 2+ , and Br À (i.e., bromine excess) is calculated using: where [X] Total is the total concentration of the ionic species of interest and R is the bulk seawater mass ratio of the species relative to sodium (Br À /Na + = 0.006; Mg 2+ /Na + = 0.121; Ca 2+ /Na + = 0.03791; K + / Na + = 0.03595). Further, Br À is susceptible to postdepositional F I G U R E 2 Age model for PNM-01 from Holland et al 9 using CLAM. 31 1σ error is given in gray. The age probability distribution of each dated sample is indicated in red, and median of the highest posterity density region indicated by the black diamonds.
photochemical release in the snowpack leading to depleted concentrations relative to seawater, as well as enrichments due to contributions from Br explosions in first-year sea ice regions. To better quantify this, percentage bromine enrichment is calculated using: where positive values indicate enrichment and negative values indicate depletion. 37

| Radiocarbon ages
Calibrated radiocarbon ages for PNM-01 indicate ice-wedge growth spanning the mid-to late Holocene between $4,700 and $640 yr b2k (see 9 ). Samples taken from the approximate center of the wedge (i.e., left edge of sampling cut) reveal a clustering of ages between $800 and $640 yr b2k. The youngest ages are associated with samples taken from the center of the wedge, with a consistent increase in ages toward the outer right edge (i.e., wedge ice-sediment lateral contact). The linear age model indicates slow growth from $4,700 to $1,600 yr b2k, followed by rapid growth until $640 yr b2k ( Figure 2).

| Isotopic record
The mean with a range of À25.5 to À22.6‰ (À196.8 to À172.2‰ for δD) and a mean d-excess of 8.7‰ (range of 6.5 to 11.2‰). This agrees with winter (DJFMAM) precipitation (snow and snow/rain events) collected in the Inuvik region (À24.2 ± 3.9‰ for δ 18 O, À186.1 ± 29.8‰ for δD, 7.5 ± 4.3‰ for d-excess 38 ). Further, the δD-δ 18  year-round precipitation data described by Fritz et al. 39 The δ 18 O values for the PNM-01 ice wedge are comparable to modern winter precipitation, which should reflect modern regional climate warming with more positive isotopic values than characteristic of the mid-to late Holocene. 10  It should be noted that ice wedges are subject to large internal isotopic variations, often with low signal-to-noise ratios within horizontal profiles sampled at the same depth within a single ice wedge. 28 Further, sampling depth dependencies due to Rayleigh fractionation during snowmelt freezing and ice plugging, noncentric contraction cracking, and variable infill timing 28 can result in both high inter-and intra-wedge isotopic variability. As such, over-interpretation of the PNM-01 isotope and ion time-series is cautioned, and more icewedge records are needed in this region to enhance the signal-tonoise ratio and support paleoclimatic interpretations.

| Geochemical record
Of the 275 samples analyzed for δ 18  All species are positively and significantly correlated ( p < 0.01) to each other (Table 2). A strong correlation (r = 0.99) is observed between Cl À and Na + (Figure 4a), and Br À is strongly correlated to Cl À (r = 0.88), Na + (r = 0.85), and Ca 2+ (r = 0.86). Results of the PCA are displayed in the biplot in Figure 5, and the loadings for the geochemical variables are given in  (Figure 3c,d).
Cl À and Na + are predominantly sourced from sea salt, and thus their concentrations reflect the strength of marine contribution. As such, a continuous record of Cl À and Na + was developed from the age model from $4,600 to $700 yr b2k. The concentrations of both Cl À and Na + were binned at 100-year intervals to reduce the influence of data-rich periods and minor age uncertainties on the trend, and the resultant data show a strong positive trend in Cl À and Na + concentrations from $4,600 to $700 yr b2k (Figures 6 and S2).

| Sea-salt aerosol source
The ion geochemistry of wedge ice should reflect the geochemical composition of the initial meltwater of the late-season winter snowpack. Thus, ice-wedge geochemical composition is expected to reflect the accumulation of snow impurities over the period of seasonal snowpack formation. 19,26 Though the ion concentrations are comparable between the PNM-01 record and the Utqiaġvik snowpack (Table 1), the sources of these ionic species are complex and subject to great regional variability, and some influence from the ground surface or active layer must be assumed in ice wedges due to the high sediment loads and plant remains commonly found in wedge ice. Nevertheless, the concentrations of major marine tracers (Na + , Cl À ) and lithogenic tracers (Ca 2+ , Mg 2+ ) in PNM-01 are similar to that of the T A B L E 1 Average major ion mass concentrations of PNM-01 and reported values from Jacobi et al, 1 from the spring snowpack in Utqiaġvik (Barrow), Alaska. Snowpack values represent the summary statistics of deposited snow types, including blown snow, wind-packed snow, wind-packed surface snow, snow with ice layers, and depth hoar.  Figure S4).
Results from the PCA ( Figure 5) indicate a similar driver of variability in the ion geochemistry, potentially describing a similar sourcing and/or transport processes, probably related to SSAs. This is further supported by the strong correlation between Cl À and Na + (r = 0.99) and a Cl À /Na + mass ratio of 1.80, which is effectively identical to the expected value of 1.79 for bulk seawater (Figures 4a and   5). Variability in Cl À and Na + concentrations in glacial ice cores has been associated with open water extent in marginal sea-ice zones and leads (e.g., 43,44 ). In open water, bursting bubbles produced by breaking Frost flowers, however, may influence the local salinity of the sea ice snowpack, and thus indirectly contribute to SSA production.
Freshly deposited snow on high-salinity first-year sea-ice surfaces is subject to capillary wicking, with the upward migration of unfractionated sea salt (relative to the brine) occurring up to 17 cm in height in the snowpack. 50 Further, polynyas and leads can contribute to marine snowpack salinity through the production of sea-spray, and thus SSAs, resulting in Na + concentrations up to 10 times greater than frost flowers. 50 Under strong winds, snow particles from the marine snowpack are released into the boundary layer, and sublimation of the particles results in the formation of SSAs. 57 Studies have shown that SSA production during blowing snow events over the marine snowpack is the dominant source of SSAs to coastal sites during the fall and winter, explaining the winter SSA maxima in atmospheric concentrations and glacial records (e.g., 5,6,46,53,57,58 ). Blowing snow contributions from the marine snowpack can be identified by strong deficits in SO 4 2À relative to bulk seawater in the terrestrial snowpack. This is due to SO 4 2À uptake during the precipitation of mirabilite (Na 2 SO 4 .10H 2 O) on new sea ice surfaces below À8 C, resulting in a F I G U R E 5 PCA biplot of the ice-wedge geochemistry dataset for the first two principal components. The age of the sample is denoted by shading, with younger samples in lighter shades and older samples in darker shades F I G U R E 6 One-hundred-year binned concentration averages of (a) Cl À and (b) Na + from PNM-01 shown on a logarithmic scale. Regression is indicated by the red line.

| Br À enrichment and nonmarine sources
The sea-ice snowpack is a source of bromide sea salts, though the processes governing production and release differs from other seasalt species. Br À production occurs predominantly in the interstitial air of the first-year sea-ice snowpack. 60  Bromide enrichment is only observed before 2,800 yr b2k in the PNM-01 record. %Br enr is consistently negative between $2,800 and $700 yr b2k, displaying weak Br À depletion relative to seawater over this period (average Br À /Na + of 0.005 vs. 0.006 in seawater). This may be due to Br À loss during the aerosol phase 67 or postdepositional photochemical reactions in the snowpack resulting in Br À recycling.
Following the deposition of Br À with snow, bromine activation may resume, resulting in the re-release of Br radicals and subsequent depletion of bromide in the snowpack. 1,68 Though the earliest part of the record may reflect the original sea-salt signal, postdepositional removal of Br À in the source snowpack is evident and is probably reflective of the age of the snowpack, and thus the extent of ion relocation.
Notably, before $2,800 yr b2k, %Br enr is positive and up to 82% enriched in Br À relative to seawater (Br À /Na + = 0.008), showing a consistent linear decrease from $4,600 to $2,800 yr b2k. While Br À concentrations fall within the observed range of the Utqiaġvik snowpack ( Figure S3), this distinct trend in Br À /Na + suggests the presence of some systematic driver resulting in a strong linear decrease in enrichment relative to other parts of the record. Over this period, Br À is weakly correlated with Na + (r = 0.53) compared to after $2,800 yr b2k (r = 0.98). Though ratio values are in agreement with the Utqiaġvik snowpack ( Figure S3), Mg 2+ /Na + , Ca 2+ /Na + , and Cl À /Na + , and to a lesser degree SO 4 2À /Na + and K + /Na + , show a similar trend to % Br enr (Figure 3b). These observations suggest Br À enrichment is not likely to be driven by bromine explosions as this process would not influence other ionic species. Mg 2+ , Ca 2+ , and K + , while present in SSAs, may exhibit significant contributions from terrestrial sources, 43 including the deposition of enriched dust from both local and longdistance origins. 1,50 Nonsea-salt accumulation may be further compounded in wedge ice due to snowpack meltwater interaction with surface materials and active layer soils (e.g., soil dust), particularly during initial ice-wedge growth where contraction cracks are more likely to have direct contact with host sediments. 19 This may contribute to ratios greater than bulk seawater before 2,800 yr b2k for Mg 2+ /Na + , Ca 2+ /Na + , and K + /Na + , though a similar enrichment of cation species is likewise observed in the Utqiaġvik snowpack (except K + ); thus, enrichment in the PNM-01 ice wedge is probably inherited from the snowpack rather than other terrestrial sources during infill. Regarding the strong enrichment in K + relative to Na + , snowpack or meltwater interaction with illite, a common clay component in soils, may result in locally higher K + concentrations and weak correlations with other ionic species, though this is speculative.
Regardless, lithogenic contribution is unlikely to yield higher con- coupled with isotopic analysis, may provide further insight into ice-wedge formation processes that have thus far been difficult to identify, though postdepositional changes to snowpack chemistry can be attributed to many processes which cannot be distinguished.

| Paleogeographic implications
The PNM-01 record reveals an increasing trend over the mid-to late Holocene in the Cl À and Na + records and the sea-salt components of Mg 2+ , Ca 2+ , and Br À (Figure 6). Enhanced SSAs in glacial records have been correlated to increased sea ice extent (i.e., blown snow) (e.g., 49 ), increased open water extent (i.e., sea salt spray) (e.g., 43 ), and sea ice production and turnover in polynyas in the moisture source region (e.g., 74 ). Nevertheless, snowpack ion concentrations are strongly controlled by distance from the coast due to early deposition of SSAs, particularly in the coarse mode (1-10 μm), during air mass transport inland. 36,43,75 In glacial ice core studies, the distance-to-coast of coring sites impacts the proportion of SSAs reaching the site relative to more coastal sites, influencing the magnitude of ion concentrations at sites with the same moisture source. At glacier sites suitable for ice coring, variations in geochemistry driven by changes in the continentality of the site are marginal relative to climatically driven changes in geochemical records, and thus consideration of paleogeography is often not considered. Comparatively, coastal ice wedges in the Arctic can experience relatively large changes in distance-to-coast over short timescales due to the susceptibility of permafrost to coastal erosion driven by changes in relative sea level (RSL) and sea surface temperature, in addition to climatically driven changes in permafrost stability.
Due to PNM-01's proximity to the Beaufort Sea coast, the increase in sea salt concentrations is probably dominantly driven by coastal recession rather than sea ice processes.
Specific to the PNM-01 study site, a rapid reduction in distanceto-coast over the Holocene was probably driven by high rates of coastal erosion through thermal and mechanical processes driven partially by a rise in RSL. The Tuktoyaktuk Coastlands are composed of ice-rich glacial till and outwash overlying unconsolidated sand, making the coast exceedingly susceptible to high rates of erosion, 76 as evidenced by the current rates of erosion up to 12 m yr À1 on the highly exposed coastline of Pullen Island. 77  The RSL curve is supported by geophysical imagery that shows an increase in the depth of shoreface erosion from the mid-to late Holocene, suggesting continual but decreasing RSL rise. 79 Further, geothermal modelling of drill hole temperatures and permafrost observations within the limits of offshore ice-bonded permafrost (i.e., formerly subaerially exposed areas) off the Tuktoyaktuk Peninsula shoreline show evidence of a marine transgression occurring 3,500-4,000 years ago at sites up to 45 km from the present shoreline. 80,81 While local RSL rise over the Holocene is not fully compatible with reconstructed eustatic sea-level changes 82 and modelled isostatic rebound effects 83 over the last 4.6 kyr, local processes such as consolidation, forebulge collapse, and basin subsidence are likely to have played a role. 78 Apart from RSL rise, marine transgressions in the Tuktoyaktuk Coastlands can be forced by permafrost degradation and thaw subsidence as well as accelerated erosion during storm events during the ice-free season. 77 Coastal erosion, and thus decreasing distance-to-coast of the study site, offers a plausible explanation for the long-term increase in ion concentrations in the PNM record. A study conducted in Utqiaġvik found that the Cl À and Br À concentrations in January snow profiles and blowing snow above the snowpack decreased substantially 4-10 km from the Chukchi Sea coast, with a decrease in Cl À in deposited snow up to $9.69 mg L À1 km À1 . 84 A similar study conducted in Utqiaġvik between March and May found a similar relationship with Na + and Br À . 68 This decay in sea salt concentrations has been observed in the geochemical differences between ice wedges in coastal and inland environments across the Arctic, 23,85 as well as in undisturbed lakes in the Mackenzie Delta region. 86 Assuming the rate of decay in SSA deposition in Utqiaġvik found by Duce et al. 84 is a reasonable estimate for the PNM-01 site, and was constant over the late Holocene, the increase in sea salt species (2.5 mg L À1 kyr À1 for Cl À , 0.05 mg L À1 kyr À1 for ssCa 2+ ) in the PNM-01 record may indicate a reduced distance-to-coast of up to $1 km over the late Holocene. Further studies are needed to better constrain local decay rates in sea-salt deposition on the seasonal snowpack to provide a more precise estimate of past coastal recession. Nevertheless, these findings are noteworthy as they suggest that ice-wedge ion geochemistry may provide a new proxy of paleogeographic change that has not yet been considered, with the potential of estimating local long-term shoreline change.

| CONCLUSION
Ice wedges offer a unique paleoclimatic record of the seasonal snowpack in nonglaciated Arctic regions, with potential as archives for marine aerosols. This study analyzed the geochemical composition (Cl À , Na + , Br À , SO 4 2À , Mg 2+ , Ca 2+ , K + ) of an ice wedge from the Tuktoyaktuk Coastlands. SO 4 2À depletion indicates that ice-wedge ion geochemistry is dominantly of marine origin, primarily sourced from SSAs from blown snow from sea ice surfaces. A linear decrease in bromine enrichment, and to a lesser extent concentrations of Mg 2+ , Ca 2+ , and K + relative to Na + and d-excess and δ 18 O (δD) values, before $2,800 kyr b2k suggests the presence of some systematic driver resulting in this strong trend, potentially due to contributions from depth hoar during early ice-wedge growth, though this is speculative. Over the entirety of the record, a long-term increase in sea-salt concentrations from $4,600 to $700 yr b2k is observed and may suggest a reduced distance-to-coast of the study site over the late Holocene, driven by coastal recession due to high rates of coastal erosion and minor changes in RSL. This study is a first step towards developing a better understanding of the controls on the ion geochemistry of coastal ice wedges over the Holocene and supports the potential of ice wedges to provide valuable insight into local climatic and environmental changes. While this paleogeography application is limited to coastal ice wedges with a well-constrained age model, this study represents a promising new approach to ice-wedge geochemical interpretation.