Introduction
When the snowpack on sea ice is heavy and thick enough to depress the top surface below sea level, a slush layer or slurry is formed through the mixture of seawater or brine and snow at the base of the snow cover. The snow ice is incorporated into the ice cover after subsequent periods of freezing. In the Antarctic, overall higher snow accumulation on thinner sea ice results in the widespread occurrence of surface flooding and snow-ice formation (Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994; Massom and others, Reference Massom2001). The snow-ice formation on the pack ice plays a critical role in the mass and energy balance of Antarctic sea ice (Arrigo and others, Reference Arrigo, Worthen, Lizotte, Dixon and Dieckmann1997; Maksym and Markus, Reference Maksym and Markus2008). The snow-ice formation is also an essential factor for biological productivity within the ice by controlling brine and nutrient flux from seawater to the flooded layer and ice layer below (Ackley and Sullivan, Reference Ackley SF and Sullivan1994; Fritsen and others, Reference Fritsen, Lytle, Ackley and Sullivan1994). Lack of knowledge about the snow-ice contribution in Antarctic sea ice could hinder accurate estimates of sea-ice thickness and snow cover depth using remote sensing (Drinkwater and Lytle, Reference Drinkwater and Lytle1997; Xie and others, Reference Xie, Tekeli, Ackley, Yi and Zwally2013).
After snow ice is integrated into the sea-ice column, it becomes troublesome to discriminate the snow ice consistently from granular frazil ice. However, the stark contrast in the isotopic signatures of snow ice and seawater originated ice can help determine contributions of snow ice to the total ice thickness. Previous studies of the snow-ice component utilized the standard practice to identify all granular ice layers with negative isotope signature (δ18O < 0) as snow ice (e.g. Lange and others, Reference Lange, Schlosser, Ackley, Wadhams and Dieckmann1990; Eicken and others, Reference Eicken, Lange and Wadhams1994; Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994). However, this procedure to determine the percentages of core length that contain isotopic signatures of meteoric water might be biased due to the normally low-resolution isotopic measurements, with a single isotope value designated for each 10 cm vertical section cut from sea-ice cores. The mixing and diffusion processes during the flooding and refreezing of snow ice might also modify the isotopic signatures of the snow-ice layers (Maksym and Jeffries, Reference Maksym and Jeffries2001). In contrast to previous research deriving percentages of core length that contain snow ice, we applied an updated isotope mixing model to determine snow-ice contribution in the mass balance in an effort to produce more consistent estimates of snow-ice contributions (e.g. Eicken, Reference Eicken1998; Maksym and Jeffries, Reference Maksym and Jeffries2001).
In the Antarctic, sea-ice extent showed an overall gradual increasing trend during the past four decades (Maksym and others, Reference Maksym, Stammerjohn, Ackley and Massom2012); however, this trend may have reversed during recent years (Parkinson, Reference Parkinson2019). The Bellingshausen/Amundsen Sea sector is an anomalous sector of the Southern Ocean, having a 40-year decreasing trend instead, reaching a minimum in 2007, and showing an upward trend since 2007 (Parkinson, Reference Parkinson2019). The snow-ice contributions varied spatially in the Antarctic, comprising only 7% of the ice thickness in the Weddell Sea (Lange and others, Reference Lange, Schlosser, Ackley, Wadhams and Dieckmann1990) to as high as 36% in the northern Bellingshausen/Amundsen Seas (Jeffries and others, Reference Jeffries, Krouse, Hurst-Cushing and Maksym2001). The modeling of snow-ice thickness in the Antarctic produced the thickest snow ice along the coast in the Amundsen Sea (Maksym and Markus, Reference Maksym and Markus2008). The observed spatial variations of snow-ice contributions can have several causes including regional differences in: sea-ice extent/thickness; meteorological conditions, i.e. variations in snow precipitation between different regions; and surface topography affecting snow buildup around roughness elements. There is almost no in situ data available along the southern Amundsen Sea until now, except a few cores obtained by Jefffries and others (Reference Jeffries, Shaw, Morris, Veazey and Krouse1994). Sea-ice cores were collected in the southern Amundsen Sea during the Oden Southern Ocean 2010/11 (OSO1011) expedition from late December 2010 and January 2011 (Fig. 1). In this paper, the measurements of stable isotope variations and ice texture profiles of these ice cores are utilized to delineate snow-ice contribution to ice development in the Amundsen Sea. These results are then compared to previous investigations of snow-ice occurrence around Antarctica.
Fig. 1. Oden Southern Ocean 2010/11 expedition track with locations of eight ice stations (red crosses) in the Amundsen Sea. The detailed transit data of the OSO 1011 expedition are available at https://oden.geo.su.se/oso1011. The station locations at the time of sampling were well inside the ice edge (> 50 km) and in areas of greater than 80% ice concentration.
Materials and methods
Sampling and field measurements
Eight ice stations A2, A5, A6, A7, A8, A9, A10 and A12 were designated during the Amundsen Sea transits A-II, A-V, A-VI, A-VII, A-VIII, A-IX, A-X and A-XII, respectively (Fig. 1). Internal decay features indicate that they are second-year or multiyear ice. While slush layers were observed often at the base of the snowpack, the latest snow-ice formation was unlikely due to the summer conditions for these ice stations at the time of sampling. There was no superimposed ice observed for all ice stations. All ice cores were collected from drifting pack ice, except ice core A12 sampled from fast ice. Sampling information and core description for the eight cores collected are given in Table 1.
Table 1. Summary of the ice stations with sampling information, core information, snow depth and bulk salinity
Most ice stations had significant components of ridged ice; however, ice cores were acquired on level ice within the station area. Snow depth at our sampling sites was measured using meter sticks. The snow depth at our core sites exhibits strong spatial variability, with a range of 12–72 cm except for only trace snow observed at ice core A2 (Table 1). Negative freeboard was observed at ice stations A5, A6, A7, A8 and A9 due to their relatively high snow-to-ice thickness ratio, while other stations (A2, A10, A12) had positive freeboard. We would investigate if the snow depth is an indicator of the snow-ice contributions in these ice cores later.
Analyses of salinity, ice texture and water isotope ratios
The cores were placed in plastic bags and transferred to a cold room (− 40°C) onboard the R/V Oden. Cores were shipped back frozen to the US Army Cold Regions Research and Engineering Laboratory (CRREL) in Hanover, NH. In a cold room there, each core was cut into two halves vertically using a bandsaw. One half core was cut into nominal 10 cm vertical sections, and then the ice sections were melted in separate containers. Salinity was measured with a conductivity probe and meter (Beckman Coulter®) for each melted subsample. A vertical thick section slice (3 mm thickness) was taken from the center of another half core. The ice texture of each thick section (10 cm length) was examined between crossed polaroids on a light table in the cold room. Three ice textures (columnar, granular and mixed columnar/granular) were distinct and distinguishable in these thick sections. Then the thick sections were sealed and melted in plastic ziplock bags at room temperature. Immediately after melting, the meltwater was poured into vials, capped with no headspace, sealed with parafilm and stored in a refrigerator for future water isotope analyses.
A total of 92 melted sea-ice samples from the eight cores (Table 1) were prepared for stable isotope analyses. The δ18O and δD measurements of these samples were performed on a Picarro L2130-i water isotope analyzer (cavity ring-down spectroscopy, CRDS) in the Department of Geological Sciences, University of Texas at San Antonio. Results are reported as relative to the standard Vienna Standard Mean Ocean Water (VSMOW). The measurement precisions for δ18O and δD are 0.1‰ and 0.4‰, respectively. δ18O and δD are given in the following equations:
Isotopic approaches to determine snow-ice contributions
Most previous research has only used δ18O as an index tracer to determine the percentages of core length that contain snow ice (e.g. Lange and others, Reference Lange, Schlosser, Ackley, Wadhams and Dieckmann1990; Eicken and others, Reference Eicken, Lange and Wadhams1994; Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994). Granular ice could originate from either frazil ice or snow ice, but the stable isotope signal of these layers could demonstrate their origin (Lange and others, 1990). Typically, the granular layers with δ18O < 0 were assigned as snow ice; while granular layers with δ18O > 0 were assigned as frazil ice.
However, there are several limitations in such a classification approach that need to be addressed. The brine exchange with ocean must occur through the porous channels when the slush layer is formed at the base of the snow cover; while the flooded layer freezes, the diffusive-convective process would also occur along the ice column during freeze-up of the porous channels (Lange and Hubberten, Reference Lange, Hubberten, Maeno and Hondoh1992; Eicken, Reference Eicken1998; Massom and others, Reference Massom2001). Then the snow-ice layers will tend to lose isotopic signature of snow, while underlying ice layers may become enhanced with snow signature. In addition, the initial freezing process through frazil accumulation could incorporate falling snow particles floating at the sea surface. Thus, the underlying frazil ice sections could then have negative δ18O due to varying degrees of the admixture of snow and seawater isotopic signatures.
The simulated distribution of sea-ice isotopic values, utilizing typical mean and standard deviation of isotope values for seawater and snow ice, should be a bimodal distribution with two peaks (Fig. 2a): one peak is the seawater distribution with higher mean value and smaller variance; one peak is snow-ice distribution with lower mean value and larger variance. The error of snow-ice classification (δ18O < 0) should be negligible. However, the actual distribution of isotopic values for our 92 sea-ice samples lack the peak of snow ice (Fig. 2b), which suggests significant mixing between the snow ice and seawater originated samples, either due to in situ diffusion process or low-resolution sampling. The classification of many slightly negative δ18O samples, falling between − 3‰ and 0‰, as snow-ice layers using the standard practice of snow-ice identification might be problematic.
Fig. 2. (a) Simulated distribution of sea-ice samples with typical isotope values for seawater (1 ± 0.5‰) and snow ice (− 10 ± 3‰); (b) actual distribution of isotopic values for 92 sea-ice samples. The shade in Figure 2a was referred to as ‘false negatives’ which represents the misclassification of snow-ice layers; in Figure 2b, the tick marks on the horizontal axis represent measured isotope values of sea-ice samples, and the blue curve is the kernel density estimation by smoothing the observed density.
Snow-ice layers in ice cores as defined by negative δ18O values suggest that the snow or meteoric-water fractions in snow-ice sections could be as low as 10% (Jeffries and others, Reference Jeffries, Krouse, Hurst-Cushing and Maksym2001). However, the model of snow-ice formation implied snow ought to have occupied 30–50% of snow-ice mass during the freezing of a slush layer (Maksym and Jeffries, Reference Maksym and Jeffries2001) utilizing typically observed snow densities (Sturm and others, Reference Sturm, Morris and Massom1998). The discrepancy between the isotopic approach and model results indicates either that an ice layer with negative δ18O might not be snow ice, or the snow-ice layer became enriched due to the exchange of brine during freezing (Lytle and Ackley, Reference Lytle and Ackley2001).
Many previous studies tried to determine the snow/meteoric-water fraction within the snow-ice sections or ice cores using an isotope mixing model (Lange and others, Reference Lange, Schlosser, Ackley, Wadhams and Dieckmann1990; Eicken and others, Reference Eicken, Lange and Wadhams1994; Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994). The ice sections are not a closed system for water isotopes due to brine drainage or convective transport during the snow-ice formation and freezing (Golden and others, Reference Golden, Ackley and Lytle1998). While in an Arctic study (Tian and others, Reference Tian2018), the hydrochemical characteristics of sea-ice core and seawater depth profiles indicated little snowmelt enters the upper ocean during sea-ice evolution. Since it is not known how much of the meteoric signature in the slush drains into the ocean during freezing in this study, we instead assume that the meteoric signal is ‘diffused’ through the ice column but not lost to the ocean. In that sense, the whole ice column is considered as a closed system with regard to water isotopes but not brine, which is the primary source of uncertainty in our model.
With assumptions of a similar endmember of snow ice for all ice cores and the mixing process during the snow-ice refreezing, we applied an updated isotope mixing model (one-tracer for two-component) to determine ‘snow-ice’ contribution in the mass balance for each ice core based on the following equation:
where F SI is the unknown ‘snow ice’ fraction; δ18OCore is bulk oxygen isotope value for each ice core; δ18OSI and δ18OSW are oxygen isotope values for the ‘snow ice’ and the ‘seawater’ components, respectively. The endmembers for δ18OSI and δ18OSW would be assigned later for water balance calculations.
The meteoric-water fraction in ice cores could be derived utilizing a similar isotope mixing model (Granskog and others, Reference Granskog2017). We could also calculate the meteoric-water fraction in our sea-ice cores based on the following equation:
where F M is the unknown ‘meteoric water’ fraction; δ18OM is oxygen isotope value for the ‘meteoric water’ component; δ18OSW and δ18OCore are the same as those in equation (3). The endmembers for δ18OM and δ18OSW would be assigned later for water balance calculations.
However, there are several caveats in our isotope mixing models: we assume no snowmelt enters the upper ocean during the surface flooding and snow-ice formation; the initial incorporation of snowfall through frazil accumulation is considered limited; the limited sampling of ice cores may not be representative of snow-ice conditions in the Amundsen Sea.
Previous standard classification method identified snow-ice layers for all ice sections, based on the ‘prior’ information that snow-ice layers have formed in the upper layers of ice cores with negative isotopic signals. The total percentages of core length containing snow ice (P SI) would then be derived from this standard classification method. Our updated isotope mixing model estimated instead the ‘posterior’ snow-ice contribution (F SI) for the bulk of ice core, considering the whole ice column as a closed system for water isotopes allowing mixing/diffusion processes between ice layers. In this study, we will compare the performance of these two different methods for snow-ice apportionment and explore underlying reasons for the differences among the methods. The meteoric-water fraction (F M) calculated from equation (4) would be weighed against these two snow-ice apportionment results (P SI and F SI) to check if there is significant initial incorporation of snowfall and/or snowmelt lost to the upper ocean.
Results and discussion
Water isotopes, salinity and texture profiles
The depth profiles of stable oxygen isotope, salinity and ice texture of eight sea ice cores are shown in Figure 3, and the data used in this study can be found in Suppleme ntary Material. δ18O values for all sea ice samples vary from 2.3‰ to − 7.3‰. Cores A2, A7 and A8 all show low δ18O signals at the surface portion (0–10 cm); both cores A5 and A6 appear with low δ18O signal at a relatively greater depth; core A10 shows the low δ18O signals both at the surface portion and bottom portion; both cores A9 and A12 have generally higher δ18O signals with no particular pattern. Lack of low δ18O layers in cores A9 indicates the diffusion and homogenization in the snow-ice refreezing process. There should be no snow-ice formation at fast ice station A12 due to its identical columnar texture and no slush layer observed. Previous isotope measurements (Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994) and modeling results (Maksym and Jeffries, Reference Maksym and Jeffries2001) illustrated that negative δ18O values occurred at the surface then steadily enriched δ18O values with increasing depth. However, in our study, the possible snow-ice layers with low δ18O values also occur at some distance below the surface, primarily in granular layers but occasionally in columnar layers. The snow-ice layers below the surface could be a result of prior deformation through the dynamic rafting process; however, the ice layers with slightly negative values are more likely due to mixing and diffusion processes during flooding and refreezing of snow ice.
Fig. 3. Ice texture, salinity and oxygen isotope profiles for the eight ice cores collected in the Amundsen Sea.
The salinities for all sea-ice samples vary from 1.7 to 9.5 psu (Fig. 3), while the bulk salinities of all ice cores vary between 3.4 and 6.2 psu (Table 1), which are lower than typical salinity of first-year sea ice. The low salinity could happen in thicker first-year ice or typical multiyear ice. In our study, the pack ice stations (except A12) might be first-year ice that has survived through much of the current summer, and they would be second-year ice soon. The salinity profiles show significant negative correlations with δ18O profiles in most ice cores, for the reason that snow-ice layers with low δ18O signals generally have higher salinities. However, salinity is not a good indicator of provenance for sea-ice layers.
There are three different classes of ice textures: columnar (n = 27), granular (n = 48), mixed columnar/granular (n = 17). The eight cores show variable contributions of granular ice from 0 to 100% of the total length of these cores (Fig. 3), with an average of 52%. The granular ice portions are significantly higher than those observations from earlier studies, which may indicate a greater amount of snow-ice accretion in the Amundsen Sea than what have been reported previously for Antarctic sea ice elsewhere. Lower columnar texture fractions might indicate little additional congelation growth for most pack ice stations.
The normally low-resolution (one subsample per 10 cm) isotopic measurements were performed for all ice cores. High-resolution isotopic analyses (one subsample per 1 cm) were performed in the 0–10 cm section of core A5. As shown in Figure 4, there is an enriched trend of water isotopes in the 0–10 cm section, which could be due to the diffusion process or growth-rate dependent fractionation (Eicken, Reference Eicken1998). The 0–6 cm layer with negative δ18O signature should be classified as snow ice, and the 6–10 cm layer with positive δ18O signature should be classified as frazil ice. However, the mean δ18O of the 0–10 cm ice section is − 0.5‰, then the whole 0–10 cm layer would be classified as snow ice. Therefore, the classification approach would provide an upper limit to the contribution of snow ice (Jeffries and others, Reference Jeffries, Morris, Weeks and Worby1997) due to normally low-resolution isotopic measurements.
Fig. 4. High-resolution isotopic analyses performed in the 0–10 cm section of core A5.
Ice textures are also examined for their relationship with salinity and isotope signal (Fig. 5). The boxplot for salinity (Fig. 5a) of the three ice texture classes shows small differences for mean salinity values among the three ice textures. The boxplot for δ18O (Fig. 5b) shows the granular layers with a lower mean δ18O signal and a broad range of δ18O signals. Typically uppermost granular ice layers with negative δ18O have been interpreted as snow ice and underlying granular layers interpreted as frazil ice. However, in our study, the frazil layers might also possess negative δ18O signals from the exchange or diffusion processes discussed earlier.
Fig. 5. Boxplot for salinity (a), δ18O (b) among the three classes of ice textures.
Snow-ice contribution calculations
As shown in Figure 6, δ18O against δD for all sea ice samples indicates a significant linear relationship: δD = 7.91δ18O − 0.01(N = 92; R 2 = 0.996). Thus, δ18O and δD would work equivalently in mass balance calculations using the isotope mixing model for these ice cores. All sea-ice subsamples align on a straight line with a slope very close to the Global Meteoric Water Line δD = 8δ18O + 10 (Craig, Reference Craig1961), which indicates similar growth conditions for all sea-ice stations. The deuterium intercept of the regression equation is close to 0, which indicates evaporation and other dynamic fractionations are negligible during the growth phase (Souchez and others, Reference Souchez2000). Among our sea-ice sections, the most enriched isotopic values for all ice stations are close to δ18O = 2‰, which could be considered as pure seawater origin signals. The most depleted δ18O value (− 7.3‰) found at ice station A10 could be considered as snow-ice endmember; however, there is a lack of pure snow-ice signal in other ice stations due to the mixing and diffusion processes.
Fig. 6. Cross plot of δ18O against δD for the sea-ice samples of the eight ice cores collected during the OSO1011 expedition. The best-fitting regression line for all sea-ice samples is shown in the figure.
Isotopic fractionation occurs during freezing, so the solid phase becomes enriched in the heavy isotopes (O'Neil, Reference O'Neil1968). The isotopic fractionation during the freezing of seawater leads to elevated δ18O in the ice phase, and the equilibrium fractionation was determined as 2.8‰ by Beck and Muennich (Reference Beck and Muennich1988). Seawater had a mean δ18O value of − 0.5‰ in the Amundsen Sea (derived from Global Seawater Oxygen-18 Database: https://data.giss.nasa.gov/o18data/). The most enriched δ18O value of basal ice layers is 2.3‰ (Fig. 3), which agrees with the predicted isotopic fractionation very well. Maksym and Jeffries (Reference Maksym and Jeffries2001) suggested the effective fractionation coefficient could be a shift in δ18O of 1–5‰ depending on freezing conditions, which combined effects of the fractionation coefficient, solid to liquid phase change during grain coarsening of snow in the slush, and convective transport of enriched brine out of the slush during freezing. Thus the fractionation coefficient used here might underestimate the net effect of various isotopic exchange processes. However, the most enriched isotopic values for all ice stations could be considered as pure seawater origin signals. Therefore, we directly use the mean with one standard deviation of the most enriched isotopic values for all ice stations as our seawater endmember (after freezing fractionation): δ18O = 1.9 ± 0.4‰(N = 8).
We did not have isotopic measurements of snow/meteoric water samples in our study. To quantify the meteoric-water endmember, the long-term mean isotopic values of annual precipitation were calculated using the Online Isotopes in Precipitation Calculator (OIPC, http://www.waterisotopes.org). The modeled δ18O and δD of annual precipitation for all ice stations are shown in Figure 7. The modeled δ18O of annual precipitation for all pack ice stations display similar isotopic signatures (around − 22‰) except the fast ice station A12 with more negative isotopic value (− 29‰ ) due to its long distance from the moisture source. The mean δ18O value of meteoric water for all pack ice stations is − 21.9‰. Utilizing the endmembers for seawater (2.3‰ ) and meteoric water (− 21.9‰ ), and mass mixing ratios varying from 7:3 to 5:5 for snow-ice mass derived from the typical snow densities (Sturm and others, Reference Sturm, Morris and Massom1998), we could obtain snow ice with δ18O values from − 5.0‰ to − 9.8‰. The most depleted δ18O value (− 7.3‰) of the surface ice section at A10 is the mixture of 60% seawater and 40% of meteoric water. Thus, we can utilize the δ18O of − 7.3‰ ± 2.5‰ as the range of potential values of snow ice (Fig. 7).
Fig. 7. Cross plot of δ18O against δD for the thickness-weighted means of ice cores, and the endmembers for meteoric water, snow ice and seawater. The modeled isotopes in annual precipitation for the eight ice stations were calculated using the Online Isotopes in Precipitation Calculator (OIPC, http://www.waterisotopes.org). The green dashed mixing lines partition contributions of meteoric water and seawater for snow ice, and the red dashed mixing lines partition contributions of snow ice and seawater for all ice cores. The error bars of δ18O for these endmembers are also shown in the figure.
Identifying all sea-ice layers with negative isotope signature (δ18O < 0) as snow ice, we calculated total percentages of core length containing snow ice (P SI) for all pack ice cores. Utilizing isotope mixing model in mass-balance calculations, we calculated the snow-ice fraction (F SI) for all pack ice cores based on equation (3), and meteoric-water fraction (F M) for all ice cores based on equation (4). There should be no snow-ice formation at fast ice station A12, then only meteoric-water fraction was calculated at A12. The water balance calculations utilized the IsoError single-isotope two-source model described by Phillips and Gregg (Reference Phillips and Gregg2001). This mixing model calculates means and standard errors of source proportional contributions to a mixture using stable isotope analyses.
These apportionment results of snow ice (P SI and F SI) for each ice core are presented in Table 2. The P SI results using the standard classification method show a broad range of change from 5 to 50%, with a thickness-weighted average of 22.9%. In contrast, the F SI results for pack ice calculated from our updated isotope mixing model vary from 10.2 to 29.6%, with a thickness-weighted average of 15.9%. Compared with the previous classification method, we feel our updated mixing model returned more consistent and reasonable snow-ice contribution results. Five ice cores (A2, A5, A7, A8 and A10) yield P SI > F SI, which might be due to the vertical mixing process or low resolution of sampling. Thus, many slightly negative δ18O layers could be mixtures of the snow-ice and seawater signals. Only two ice cores (A6 and A9) yield P SI < F SI, which indicates that the diffusion processes during the refreezing of snow ice might modify the isotopic signatures of the snow ice and underlying ice layers, then the positive δ18O layers could also contain some degree of snow-ice signals. Another possibility is that the isotope mixing model would overestimate snow-ice fraction due to the incorporation of falling snow through frazil accumulation. Therefore, the classification approach might provide not only an upper limit to the amount of the snow ice (Jeffries and others, Reference Jeffries, Morris, Weeks and Worby1997) but also a lower limit of snow-ice contribution for some ice cores.
Table 2. Summary of the ice cores with the mean and range of δ18O, thickness percentages of core length containing snow ice (P SI), snow-ice fractions (F SI), meteoric-water fractions (F M). F SI and F M presented here are the thickness percentages (mean ± standard error) calculated by the isoError single-isotope two-source model (https://www.epa.gov/eco-research/stable-isotope-mixing-models-estimating-source-proportions)
The derived meteoric-water fraction (F M) for each ice core is also presented in Table 2. The F M fractions for pack ice vary from 3.9 to 11.3%, with a thickness-weighted average of 6.2%. The F M fractions are nearly 40% of the F SI fractions for all pack ice stations, which agrees with the assumption that meteoric water occupied from 30 to 50% the snow-ice during the freezing of a slush layer (Maksym and Jeffries, Reference Maksym and Jeffries2001). However, the F M fractions account for 20–100% of the P SI fractions, which contradicts the derived mass mixing ratio based on the typical snow densities (Sturm and others, Reference Sturm, Morris and Massom1998). Cores A6 and A9 have low P SI and almost equal F M, but the snow/meteoric water could not occupy that high fraction for the snow-ice layers. Thus, many positive δ18O layers in these two cores might also have some F M fractions due to the diffusion process. The F M fraction for the fast ice station A12 is 1.0%, which indicated limited incorporation of falling snow through the congelation ice growth.
We found the snow-ice contributions (Table 2) are independent of the measured snow depths at the time of sampling (Table 1). Even though only trace snow was observed at core A2, our isotope mixing model still returned 14.5% of sea-ice mass as snow ice. Thus, the measured snow depth at the time of sampling may not indicate snow-ice formation previously, and snow depth may be a poor indicator of total snow accumulation (Maksym and Markus, Reference Maksym and Markus2008). The possible reasons include thick sea ice would not flood with a larger amount of snow or thin ice flooded even with a small amount of snow; the deep snow with slush layers observed in summer does not necessarily imply substantial snow-ice formation previously or erased any relationship from previous winter.
Comparison with previous investigations
Snow ice has been observed in all regions and seasons in Antarctic pack ice (Lange and others, Reference Lange, Schlosser, Ackley, Wadhams and Dieckmann1990; Allison and Worby, 1994; Eicken and others, Reference Eicken, Lange and Wadhams1994; Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994, Reference Jeffries, Morris, Weeks and Worby1997, Reference Jeffries, Krouse, Hurst-Cushing and Maksym2001). However, these studies reported highly variable contributions of snow ice to the total sea-ice thickness: ranging from only 7% of the ice thickness in the Weddell Sea (Lange and others, Reference Lange, Schlosser, Ackley, Wadhams and Dieckmann1990) to as high as 36% in the northern Bellingshausen/Amundsen Seas (Jeffries and others, Reference Jeffries, Krouse, Hurst-Cushing and Maksym2001). The snow-ice contributions varied seasonally in the same region (Allison and Worby, Reference Allison and Worby1994). The contribution of snow ice in the winter (Jeffries and others, Reference Jeffries, Morris, Weeks and Worby1997) was greater than that in the summer (Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994) for pack ice in the Bellingshausen Sea. There was an elevated snow-ice contribution for second-year ice than first-year ice (Eicken and others, Reference Eicken, Lange and Wadhams1994). However, there is almost no in situ data available along the southern Amundsen Sea until now. Our updated isotope mixing model indicated snow-ice contributions in the southern Amundsen Sea range from 10.2 to 29.6% with an average of 15.9%, which lie within the range of previously reported results around Antarctica.
The modeling of snow-ice thickness in the Antarctic produced the thinnest snow ice in the Weddell Sea, whereas the thickest snow ice was predicted along the coast in the Amundsen Sea (Maksym and Markus, Reference Maksym and Markus2008). However, our mixing model returns moderate snow-ice contribution in the southern Amundsen Sea, which contradict with the modeling results predicted by Maksym and Markus (Reference Maksym and Markus2008). It is possible that the model might not be well-enough defined yet due to the limits of the snow depth product or the uncertainty of processes that control snow accumulation on sea ice. Our results suggest, however, the quantities of snow ice might not be significantly higher than other regions despite more snowfall in the Amundsen Sea. Comparing to previous studies (Jeffries and others, Reference Jeffries, Shaw, Morris, Veazey and Krouse1994, Reference Jeffries, Krouse, Hurst-Cushing and Maksym2001) in the Amundsen Sea, the snow-ice contributions in our study are much lower even though the meteoric-water contributions are comparable with those of the previous studies. Our isotope mixing models returned meteoric-water and snow-ice contributions with reasonable ratios. We speculate the previous studies may have overestimated the snow-ice contributions, due to the possible misclassification of ice layers with slightly negative isotope values utilizing the standard classification method.
These observed spatial or temporal variations of snow-ice contributions might be attributed to different meteorological conditions and surface topography amongst these ice stations. The pack ice with greater snow accumulation and more dynamic environments favors flooding and snow-ice formation (Jeffries and Adolphs, Reference Jeffries and Adolphs1997). However, parts of the inconsistency for snow-ice contributions might be attributed to the significant error of standard practice of snow-ice identification. There is a lack of understanding of controlling factors on the evolution of snow ice, and in-situ measurements and isotopic analyses of ice cores are very limited. We suggest that more ice cores in the Antarctic with differing meteorological conditions and surface topography should be examined using the updated isotopic mixing model in future research.
Conclusions
Depth profiles of stable isotopes, salinity and ice texture were described to serve as illustrations of snow-ice formation and the evolution of pack ice in the Amundsen Sea. We utilized an updated oxygen isotope mixing model to determine the snow-ice contribution in the mass balance. The main conclusions drawn are:
(1) The standard procedure to calculate the percentages of core length containing meteoric water might be biased due to the normally low-resolution isotopic measurements for ice cores. The mixing and diffusion processes during the flooding and refreezing of snow ice may also modify its isotopic signature. The previous snow-ice identification method might provide not only an upper limit to the amount of the snow ice but also a lower limit of snow-ice contribution for some ice cores.
(2) The water balance calculations utilized the IsoError single-isotope two-source model. The most enriched and most depleted δ18O values represent the best estimate of endmembers of ‘seawater’ and ‘snow ice’ respectively. The derived snow-ice contributions for pack ice range from 10.2 to 29.6% with a thickness-weighted average of 15.9% in the Amundsen Sea. The meteoric-water fractions for pack ice range from 3.9 to 11.3% with an average of 6.2%.
(3) The meteoric water has occupied 40% of snow-ice mass for all ice stations using our isotope mixing model, whereas the meteoric water has highly variable fractions in the snow-ice mass utilizing the classification approach. Comparing to previous studies, the more consistent and reasonable snow-ice contribution results in this study verified the validity of our isotope mixing model. Due to only a small sample size of cores obtained in our study, more ice cores in the Antarctic need to be examined using our refined isotope mixing model in future research.
Contribution statement
BW collected the ice cores and provided field observations of sampling conditions. SFA conducted the cold room analyses of ice structure and salinity on returned cores. LT and YG conducted water isotope analyses and analyzed the isotope records for snow-ice contributions. LT wrote the first draft of the paper and all authors contributed to interpretations and final writing and editing.
Supplementary material
Data of stable isotopes, salinity, and ice texture used in this study can be found at U.S. Antarctic Program Data Center (USAP-DC): http://www.usap-dc.org/view/dataset/600106.
Acknowledgements
SFA acknowledges the support of NASA through the NASA Center on Advanced Measurements in Extreme Environments at UTSA (NASA CAMEE #80NSSC19M0194) during the conduct of this research. This project was also funded (in part) by the University of Texas at San Antonio (UTSA), Office of the Vice President for Research, the Amy Shelton and V.H. McNutt endowment, and the Center for Water Research at UTSA.
Open access
