Variability of sea salts in ice and firn cores from Fimbul Ice Shelf , Dronning Maud Land , Antarctica

Major ions were analysed in firn and ice cores located at Fimbul Ice Shelf (FIS), Dronning Maud Land – DML, Antarctica. FIS is the largest ice shelf in the Haakon VII Sea, with an extent of approximately 36 500 km2. Three shallow firn cores (about 20 m deep) were retrieved in different ice rises, Kupol Ciolkovskogo (KC), Kupol Moskovskij (KM), and Blåskimen Island (BI), while a 100 m long core (S100) was drilled near the FIS edge. These sites are distributed over the entire FIS area so that they provide a variety of elevation (50–400 m a.s.l.) and distance (3–42 km) to the sea. Sea-salt species (mainly Na and Cl) generally dominate the precipitation chemistry in the study region. We associate a significant sixfold increase in median sea-salt concentrations, observed in the S100 core after the 1950s, to an enhanced exposure of the S100 site to primary sea-salt aerosol due to a shorter distance from the S100 site to the ice front, and to enhanced sea-salt aerosol production from blowing salty snow over sea ice, most likely related to the calving of Trolltunga occurred during the 1960s. This increase in seasalt concentrations is synchronous with a shift in non-seasalt sulfate (nssSO2− 4 ) toward negative values, suggesting a possible contribution of fractionated aerosol to the sea-salt load in the S100 core most likely originating from salty snow found on sea ice. In contrast, there is no evidence of a significant contribution of fractionated sea salt to the ice-rises sites, where the signal would be most likely masked by the large inputs of biogenic sulfate estimated for these sites. In summary, these results suggest that the S100 core contains a sea-salt record dominated by the proximity of the site to the ocean, and processes of sea ice formation in the neighbouring waters. In contrast, the ice-rises firn cores register a larger-scale signal of atmospheric flow conditions and a less efficient transport of sea-salt aerosols to these sites. These findings are a contribution to the understanding of the mechanisms behind sea-salt aerosol production, transport and deposition at coastal Antarctic sites, and the improvement of the current Antarctic sea ice reconstructions based on seasalt chemical proxies obtained from ice cores.


Introduction
Antarctic ice and firn cores contain valuable information about the climate and atmospheric chemical composition of the past and provide evidence for the important role of Antarctica in the global climate system.Numerous ice and C. P. Vega et al.: Variability of sea salts in ice and firn cores from Fimbul Ice Shelf firn cores have been drilled in Antarctica during the past decades (Stenni et al., 2017).However, relatively few cores were drilled in coastal regions, which are more sensitive to changes in climate than the dry and cold interior of Antarctica.In fact, two recent review papers point out the lack of ice core data from low elevation coastal areas when discussing Antarctic climate variability (Stenni et al., 2017;Thomas et al., 2017).In an effort to understand the role of ice shelves in stabilizing the Antarctic ice sheet, particular focus has been laid on the investigation of ice rises and ice rumples as buttressing elements within the ice sheet-ice shelf complex (Paterson, 1994;Matsuoka et al., 2015).Furthermore, due to their radial ice flow regime, generally low ice velocities, and relatively high surface mass balance (SMB), ice rises are potentially useful sites for ice core retrieval (Philippe et al., 2016;Vega et al., 2016).Firn and ice cores drilled at ice rises allow obtaining high-resolution climate records to investigate sub-annual and long-term temporal changes in the loads of different chemical compounds found in the snow, providing information about their sources and transport, particularly of sea-salt ions, such as sodium (Na + ) and chloride (Cl − ), which are strongly modulated by sea ice extent and meteorological conditions.Recent modelling efforts to study the use of sea salts as proxies for past sea ice extent have shown that, under present climate conditions and on interannual timescales, meteorological conditions rather than sea ice extent are the dominant factor modulating atmospheric sea-salt concentrations that are deposited at the interior and coastal sites in Antarctica (Levine et al., 2014).However, sea salts have the potential as proxy for sea ice extent at glacialinterglacial scales when large changes in sea ice extent took place (Levine et al., 2014).
At most Antarctic sites, atmospheric sea-salt concentrations present maxima during austral winter (Wagenbach et al., 1998;Weller and Wagenbach, 2007;Jourdain et al., 2008;Udisti et al., 2012), with the exception of Dumont D'Urville where maxima occur during summer (Wagenbach et al., 1998).Similarly, sea-salt fluxes obtained from Antarctic ice cores also show winter maxima (Abram et al., 2013 and references therein).However, in some recent core records from coastal sites, no clear seasonality is observed, e.g. at Mill Island during the period 1934-2000 (Inoue et al., 2017).Abram et al. (2013) conclude that despite the seasonal signal registered in different Antarctic ice cores, sea-salt fluxes do not show a consistent relationship with sea ice extent on inter-annual timescales, and on the contrary, are highly dependent on atmospheric transport, and/or the presence of polynyas.
Hitherto, two main sources of increased winter sea-salt aerosols have been proposed: (i) increased storminess, leading to an enhancement of sea-salt aerosols above the open ocean with possibly faster meridional transport (Petit et al., 1999;Fischer et al., 2007), and (ii) a direct input of sea-salts associated to increases in sea ice, overcoming source (i), e.g.due to frost flowers (Rankin and Wolff, 2002;Rankin et al., 2004;Roscoe et al., 2011), brine (Rankin et al., 2000), and the contribution of snow transported over sea ice by wind (Yang et al., 2008(Yang et al., , 2010;;Huang and Jaeglé, 2017;Rhodes et al., 2017).
In the review by Abram et al. (2013), the authors suggest that the brine-frost flower system is a plausible source of sea-salt aerosols to coastal Antarctic sites.This hypothesis is supported by the experimental evidence that the original seawater SO 2− 4 / Na + ratio cannot be used in the non-sea-salt sulfate (nssSO 2− 4 ) calculations, leading to negative nssSO 2− 4 values both in winter aerosol and fresh snow sampled at coastal sites (Hall and Wolff, 1998;Wagenbach et al., 1998;Curran et al., 1998;Rankin andWolff, 2002, 2003), and also in ice cores from both inland (Wagenbach et al., 1994, Kreutz et al., 1998) and coastal sites (Inoue et al., 2017).These negative values indicate that a lower SO 2− 4 / Na + ratio has to be used in nssSO 2− 4 calculations, i.e. a depletion of SO 2− 4 with respect to seawater composition occurred in wet and dry deposition.
During the process of sea ice formation, ions present in the water are not incorporated in the ice crystal matrix, but remain as highly concentrated brine in brine pockets or channels.The brine can be transported by capillary effects through brine channels to the newly formed ice surface, resulting in a thin layer of highly saline surface brine.This fractionated brine is unlikely to be a direct source of sea salts because it usually quickly gets covered by snow, and no clear mechanism has been found to explain how this brine could become airborne (Abram et al., 2013).With further cooling of the ice, the volume of brine decreases and consequently, its salinity increases, leading to the precipitation of different saline compounds.This depends on temperature, e.g.sodium sulfate or mirabilite (Na 2 SO 4 × 10 H 2 O) starts to precipitate at temperatures below −8 • C, while sodium chloride (NaCl) at temperatures below −26 • C. Consequently, the remaining brine is depleted in sodium and sulfate ions via precipitation of mirabilite at relatively mild polar temperatures.Frost flowers can form from this brine when meteorological conditions are adequate, i.e. at low intensity winds, which allows these delicate structures to grow without breaking apart, and on very thin ice where a strong temperature gradient is present between the ice surface and the overlying air (Rankin et al., 2000;Rankin and Wolff, 2002, and references therein).Thus, frost flowers formed at temperatures below −8 • C will be depleted in sodium and sulfate relative to other ions present in seawater (Rankin et al., 2000;Rankin and Wolff, 2002), evidenced by negative nssSO 2− 4 values measured in aerosols and snow (see Sect. 2.3 for more details on the calculation of the nss-fractions).
For most of the last decade, frost flower formation, transport, and deposition, has been considered the most plausible mechanism behind the fractionated aerosol detected at coastal areas.However, Yang et al. (2008Yang et al. ( , 2010) ) and Huang and Jaeglé (2017) proposed an alternative mechanism: the origin of sea-salt aerosol could be due to the sublimation of blowing salty snow.This salty snow could be a result of frost flower formation, upward migration of brine within the snow (Massom et al., 2001), or by the input of sea spray from the open ocean or nearby leads or polynyas (Dominé et al., 2004).Flooding of sea ice under the weight of accumulated snow can also induce increased salinity of snow (Massom et al., 2001).As the snow can be contaminated or wetted with fractionated brine or frost flowers, it could be expected that this salty snow also shows such fractionation.As pointed by Yang et al. (2008 and2010), this salty snow can be transported by wind and if the air is not saturated, the snow particles may lose water by sublimation and become sea-salt aerosols.These aerosols could then be transported and deposited either by dry or wet deposition, depending on local meteorology.
According to Abram et al. (2013), the idea proposed by Yang et al. (2008) is plausible for coastal sites, along with the frost flower mechanism.Consequently, snow present on new sea ice and frost flowers are features that, combined with wind transport, need to be taken into account when interpreting the sea-salt record of coastal ice and firn cores.
This study discusses sub-annual and long-term temporal changes in sea salt and major ion concentration measured in three recently drilled firn cores from different ice rises located at Fimbul Ice Shelf (FIS): Kupol Ciolkovskogo, Kupol Moskovskij, and Blåskimen Island, a 100 m long core drilled near the FIS edge (S100) (Fig. 1).The main goals of the present study are to investigate possible mechanisms behind deposition, sub-annual, and spatial variability of sea salts in this coastal region.The results presented here contribute to bridging the data gap existent at coastal Antarctic sites, and to the improvement of current Antarctic sea ice reconstructions based on sea-salt chemical proxies.

Study area
With an extent of approximately 36 500 km 2 , FIS is the largest ice shelf in the Haakon VII Sea (Fig. 1).Fed by Jutulstraumen, the largest outlet glacier in DML, FIS is divided into a fast moving ice tongue, Trolltunga, directly feeding the central part of the ice stream, and slower surrounding parts.Several ice rises (250-400 m a.s.l.; 10-42 km from the coast) are found at FIS, varying in size from 15 to 1200 km 2 , and located approximately 200 km apart.
SMB obtained from the S100 core (Fig. 1) retrieved at FIS shows a mean long-term accumulation rate of 0.3 m water equivalent per year (m w.e.yr −1 ) for the period 1737-2000 with a significant negative trend in SMB for the period 1920-2000 (Kaczmarska et al., 2004).This negative trend in SMB has been reported in several shorter firn cores from the region (Isaksson and Melvold, 2002;Divine et al., 2009;Schlosser et al, 2014), including one record from the Kupol Ciolkovskogo ice rise (Vega et al., 2016).
More detailed information on previous campaigns, glaciological and meteorological conditions at FIS, and the core sites at the ice rises, can be found in Vega et al. (2016) and Goel et al. (2017), and references therein, whereas an overview on Antarctic ice rises is given in Matsuoka et al. (2015).
core was drilled from the bottom of a 2 m snow pit (not sampled for major ions).The firn density was determined as bulk density of each sub-core piece (average length of 45 cm).The samples were collected following clean protocols (Twickler and Whitlow, 1997), shipped frozen to NPI and later to the Paul Scherrer Institute (PSI), Switzerland, for cutting and chemical analysis.Sample resolution varied between 4 and 8 cm depending on sample depth and density.Thickness of ice lenses, water stable isotope ratios, and SMB for the three ice rises are reported in Vega et al. (2016).Additionally, unpublished major ion concentrations measured in the 100 m deep S100 core drilled in austral summer 2000/2001(Kaczmarska et al., 2004) ) were included in this study (Fig. 1, Table 1).The S100 core was sampled at 5 cm resolution between the top and 6 m deep, and then at 25 cm resolution between 6 to 100 m deep.
Major ion concentrations were separated into sea-salt (ss) and non-sea-salt (nss) fractions calculated from the mean seawater composition using ssNa + as standard ion, as follows: where with k seawater values listed in Table S1 in the Supplement.
To calculate the ssNa + fraction, we used the following equations system, in which nssCa 2+ was employed as reference ion to obtain the nssNa + associated with Earth's crustal material: where ] crust = 1.40 (using molar concentrations) (Lutgens and Tarbuck, 2012), and ] seawater [Na + ] seawater = 0.02, (using molar concentrations, Table S1) (Summerhayes and Thorpe, 1996).Due to the low concentrations of NO − In addition, water stable isotopes analyses of the KC, KM and BI cores are described in Vega et al. (2016); while analysis of the S100 core is described in (Kaczmarska et al., 2004).

Firn and ice core timescales
The timescales of the KM and BI cores were obtained based on annual layer counting of water stable isotope ratios (δ 18 O), and found to cover the periods between austral winter 1995/96 and summer 2014, respectively.The error in the dating was estimated as ±1 year for both of these cores (Vega et al., 2016).Both KC and the S100 cores were dated using a combination of annual layer counting of δ 18 O and identification of volcanic horizons, i.e. by using the SO 2− 4 , dielectric profiling (DEP) and electrical conductivity measurements (ECM), with timescales covering the time period 1958-2012 (±3 years) at KC (Vega et al., 2016) and 1737-2000 (±3 years) at S100 (Kaczmarska et al., 2004).

Ion concentrations and sources
Table 2 shows median concentrations for all ions measured in the cores.In addition, mean, maximum, minimum, and standard deviation (σ ) are shown in Table S2, and boxplots of raw ion concentrations in the different cores are shown in Fig. S1.
In general, median concentrations in the KM core are higher than in the other ice-rise cores, e.g. six to eightfold higher concentrations of Na + , K + , Mg 2+ and Cl − in the KM core than in the KC core (which is further inland than the other sites) are found for the period 1995-2012.The relatively high Na + and Cl − concentrations observed in the KM core are also detected in the upper metres of the S100 core (in the periods 1995-2000 and 1950-2000, respectively, see Table 2).Similarly high values have been reported in several snow and firn samples from other western DML coastal sites (Kärkäs et al., 2005), and in Mill Island, Wilkes Land (Inoue et al., 2017).We found no significant relationship between median annual ion concentration and latitude, site elevation, and distance from the sea for most of the species with the exception of annual SO 2− 4 and annual MSA concentrations which show a significant decrease (at the 95 % confidence level) with latitude, and east longitude, respectively.However, additional ice cores from Fimbul are needed to obtain a robust conclusion on spatial distribution of major ions.
In order to assess the most important sources explaining the total variance in the glacio-chemical records from FIS, a principal component analysis (PCA) was applied to the different ion series measured in the KC, KM, BI, and S100 cores.Years in which no sub-annual concentrations were available in the S100 (1793, 1841, 1866, 1918, and 1944) due to low resolution, were filled in by linearly interpolating between the annual means of the previous and following year.For the PCA analysis, the logarithms of the raw concentrations were used, at sub-annual (using the raw values as input) and annual resolutions, and standardized by subtracting the mean of the data series from each data point and then dividing the result by the standard deviation of the data series.Due to the sampling resolution, only the KM and BI cores were comparable at a sub-annual level.PCA analyses were performed for three different periods of the S100 core: for the entire time interval spanning 1737-2000, for the subsection between 1737-1949, and between 1950-2000.The sum of the variances of the first three principal components (PC1, PC2 and PC3) was ≥ 80 % of the total variance of the original sub-annual and annual data in all cores.Since the results of the sub-annual and annual PCA analysis are similar only the annual results are considered.The loadings of the first three (KC) and two (KM, BI, and S100) principal components are shown in Table 3. PCA results are consistent between the different cores.Consequently, the ions can be separated in two main groups: sea-salts species (Na + , Cl − , K + , Mg 2+ , and Ca 2+ ) and marine biogenic-mixed (MSA, SO 2− 4 , including NO − 3 ) (Table 3).Generally, our results indicate that the major sources of ions at the different sites are the same, independent of the core site and mean concentrations of ions in the cores.Only at the KC site the PCA results imply an additional input of Ca 2+ from other sources than sea salt, as for instance mineral dust.Table 3 shows high loadings of NO − 3 and MSA in PC2, and thus, coherence between both species.This correspondence has been previously observed in an ice core from Lomonosovfonna, Svalbard, and a fertilizing effect was proposed as explanation for those findings (Wendl et al., 2015).Wendl et al. (2015) suggest that enhanced atmospheric NO − 3 concentrations and the corresponding nitrogen input to the ocean can trigger the growth of dimethyl-sulfide (DMS)producing phytoplankton.However, there is a variety of possible NO − 3 sources for polar sites, and the relative importance of these sources at certain locations and time is still in discussion (Mulvaney and Wolff, 1993;Savarino et al., 2007;Wolff et al., 2008;Weller et al., 2011;Pasteris et al., 2014;Sofen et al., 2014).

Long-term variability of ion concentrations
We use the two longest available records for FIS (KC and S100) to explore the long-term temporal variability of major ions, with special focus on sea salts, represented by Cl − and Na + (Fig. 2).In the S100 core, Na + , Cl − , K + , and Mg 2+ median concentrations show a marked sixfold increase after the 1950s.However, there is no significant increase of the concentration of these species in the KC core.Due to its limited time coverage, it cannot be determined if there was a substantial relative increase in concentrations at this site after the 1950s.MSA and NO − 3 concentrations do not show such a marked increase in the S100 core and values agree between both cores after the 1950s (Fig. S2).Consequently, www.the-cryosphere.net/12/1681/2018/The Cryosphere, 12, 1681-1697, 2018  2 and S2).
With the exception of MSA, all ions show a positive trend (significant at the 95 % confidence level) during the period 1950-2000, although the slope for NO − 3 is three orders of magnitude smaller than for the other ions.Such significant linear trend was not observed in the KC ion record over the same period (slope and error of the linear regression are shown in Table S3).Ions, with the exception of MSA, also show a positive and significant trend between 1737 and 1949 (Table S3); however, the increase is less marked than during the 1950-2000 period.

Sub-annual variability of ion concentrations
The lack of extensive precipitation measurements at subannual resolution near the sampling sites at FIS, makes a precise reconstruction of the precipitation regime at the area difficult.To obtain a time scale for the KC, KM, and BI icerise cores, Vega et al. (2016) employed δ 18 O winter minima and summer maxima, and assumed uniform precipitation throughout the year at the core sites.The assumption was made on the basis of precipitation data for DML reported by Schlosser et al. (2008), which showed high temporal variability in the monthly sums, due to the influence of cyclone activity affecting both coastal and inland regions.In addition, at Neumayer station (70 • 39 S, 8 • 15 W), the closest to the ice-rise core sites, two precipitation maxima (April and October) are identifiable for the period 2001-2006, a manifestation of the semi-annual oscillation of the circumpolar trough (Schlosser el al., 2008).Considering the above, to investigate the sub-annual variability of the different ion groups in the KM, BI, and S100 cores, we associated the winter minima and summer maxima in δ 18 O determined in the KC, KM, and BI cores (Vega et al., 2016), and in the S100 core (Kaczmarska et al., 2004), with the months of July and January, respectively.The values for April and October were derived  S3.
by interpolation between January-July, and July-January, respectively, in each core time scale.We defined summer samples, as samples within November and April (NDJFMA), and winter samples, as samples within May and October (MJ-JASO).Summer and winter mean concentrations were then calculated based on logarithms of raw ion concentrations expressed in µmol L −1 .Ion concentrations were not available at the top 2 m (removed before drilling); therefore, the composite year consisted of 16 (1996-2011) and 15 (1997-2011) complete years for the KM and BI cores, respectively.In the S100 core, sub-annual variability were investigated only during the period 1995-2000, where the concentrations have sufficient temporal resolution.The resulting summer and winter mean concentrations in the cores are presented in Fig. 3.
Sea-salt species (Na + and Cl − , Fig. 3a) show lower concentrations during summer in the BI, and S100 core, whereas in the KM core summer and winter show similar means.Both Mg 2+ and Ca 2+ (Fig. 3b) show similar means in both summer and winter.MSA concentrations (Fig. 3c) show summer maxima in all three cores, with a higher summer to winter difference in the BI core, compared with the KM, and S100 cores.These summer maxima are in agreement with the main source of MSA (marine-biogenic), most active during the warmer months.The MSA winter minimum is not as pronounced in the KM core as in the BI core, while the lowest MSA minimum is reached in the S100 core.NO − 3 and SO 2− 4 concentrations (Fig. 3d) show a distinct increase toward the summer in the BI core, which is also observed in the KM core, although less marked.KM and BI SO 2− 4 concentrations are higher in the summer, while both NO − 3 and SO 2− 4 summer and winter means are similar in the S100 core.

Sea-salt and non-sea-salt fractions
PCA results presented in Sect.3.1 show two main groups in which ions can be separated: sea salts (ss-fraction), and marine biogenic-mixed (nss-fraction).In order to confirm the common sea-salt source for Na + and Cl − , we calculated the Cl − / Na + ratio and ion sea-salt and non-sea-salt fractions.Table 4 shows median Cl − / Na + ratios (using concentrations in µmol L −1 ) in the KC, KM, BI, and S100 cores.Medians of the Cl − / Na + ratio in the ice-rise cores are equal (KC, and BI) or slightly higher (KM) than the expected ratio in seawater (i.e.Cl − / Na + = 1.2), while Cl − / Na + medians in the S100 core are lower than the expected ratio in seawater, both before and after 1950.Maxima in the Cl − / Na + ratio vary between 1. 5-3.8, and   Table 4. Median Cl − / Na + ratio (expressed in µmol L −1 ), ssNa + , and nssNa + , and percentage of mean nssNa + to mean total Na + in the KC, KM, BI, and S100 cores.Since some of the calculated ssNa + values in the KC core were negative, ssNa + statistics are shown considering all data points and only positive ssNa + values (sample rejection percentage is shown in parenthesis).0.1-0.9(Table S4).These results show a clear difference in the Cl − / Na + ratio between the ice-rise cores and the S100 core, i.e. a Cl − to Na + imbalance in the S100 core associated to an excess of Na + .This excess of Na + can be due to the recombination of biogenic SO 2− 4 with ssNa + and/or to additional nssNa + sources (Legrand and Delmas, 1988).This imbalance can further be enhanced by a depletion of Cl − , due to shorter sea-salt atmospheric residence time and HCl loss from snow (Legrand and Delmas, 1988;Wagnon et al., 1999).HCl loss becomes significant at relatively low snow accumulation rates (Röthlisberger et al., 2003;Benassai et al., 2005) below the accumulation rate reported for the S100 site.Therefore, it is unlikely that HCl loss is a dominant factor that could account for the low Cl − / Na + ratios at this site.Cl − depletion by recombination of ssCl − with atmospheric acids is dependent on the acidic condition of the atmosphere, especially sulfuric acid (H 2 SO 4 ), linked to marine biogenic emissions.Due to the seasonality of biogenic sulfur, the Cl − / Na + ratio would present lower values during the summer months compared to the winter season (Jourdain and Legrand, 2002).Sub-annual Cl − / Na + ratios (estimated as explained in Sect.3.3) in the S100 core show values of 1.4 ± 0.5 for the winter period, and 1.2 ± 0.1 for the summer period.Since the temporal resolution of the S100 core only allows sub-annual values for the period 1995-2000, is not possible to assess a sub-annual pattern on the Cl − / Na + ratio, and Cl − depletion by acidification cannot be ruled out as mechanism to explain the low ratios registered in the S100 core during the last centuries.In addition to Cl − loss, low Cl − / Na + ratios can also be a product of excess Na + from non-sea-salt sources (nssNa + ), as with, for example, crustal material from snow-free coastal areas, nunataks, or dust transported from other continents.Table 4 shows the ssNa + , nssNa + (calculated as explained in Sect.2.3) and percentage of mean nssNa + to mean total Na + in the different cores.Since some of the calculated ssNa + values in the KC core were negative (5 % of the values), ssNa + statistics in Table 4 and S4 are shown, considering all data points and only positive ssNa + values.The KC core presents the largest contribution of nssNa + to total Na + with a 19 % in comparison to the KM, BI, and S100 cores (0.6, 1.4, and 0.5 %, respectively), which is in agreement with PC3 in Table 3 pointing to a strong source of Ca 2+ to the KC site.

Site
As mentioned in Sect.2.3, we used the ssNa + fraction to calculate nss-and ss-fractions for Cl − , SO 2− 4 , K + , and Mg 2+ , while ssCa 2+ and nssCa 2+ were obtained using the equations system described in Sect.2.3.Table 5 shows median concentrations of ss-and nss-fractions for Cl − , SO 2− 4 , K + , and Mg 2+ , and Ca 2+ , while ssNa + and nssNa + median concentrations are shown in Table 4. Complementary statistics are shown in Table S4 and S5.The sea-salt fraction clearly dominates in all ions, with the exception of SO 2− 4 in the KC core, which shows almost three times more nssSO 2− 4 than ssSO 2− 4 .Nss-fractions often have negative values which can be associated to an ssNa + enrichment or to a depletion of major ions found in snow in comparison to bulk seawater, i.e. ion fractionation.Negative nss-fractions represent a higher percentage of total values at the S100 core compared to the ice-rise cores, with values up to 93 % for the S100  (for nssCl − ) and up to 51 % for the BI core (for nssMg 2+ ).

Evidence for increased fractionated nss-SO 2− 4 after 1950s
The nssSO 2− 4 fraction contains all SO 2− 4 sources besides sea salts, e.g.marine biogenic emissions and volcanic emissions.In coastal regions, most of the nssSO 2− 4 can be attributed to marine biogenic activity via DMS oxidation (Legrand et al., 1992) with maxima in concentrations during the summer (Minikin et al., 1998).To evaluate if ion fractionation is evidenced in the core records, i.e. the snow is strongly depleted in ssSO 2− 4 relative to ssNa + (Rankin and Wolff, 2002), leading to an underestimation of nssSO 2− 4 , it is necessary to account for the biogenic contribution to total nssSO 2− the values for k of 0.02 and 0.03 (Table 7, and Fig. 4a, b, red and blue lines, respectively).

Discussion
From the spatial and temporal variability of sea-salt concentrations in the different FIS cores discussed here, it was found that several mechanisms are contributing to the load of sea salts at FIS, in agreement with the findings by Abram et al. (2013).The ice core data from S100 also suggest that there was a change in sea-salt deposition regime after the 1950s evidenced by an increase, up to sixfold, of median sea-salt concentrations after the 1950s in comparison with the previous 200 years.Although a negative trend in SMB has been observed in the S100 and KC cores for the second half of the 20th century (Fig. 2e and f) (Vega et al., 2016), the 0.2 % m w.e.y −1 decrease in accumulation registered in the S100 core after 1950 (Table S3) cannot account for the increase observed in sea-salt concentrations after 1950s.This increase in concentration is accompanied by a clear shift in 2− 4 toward negative values, indicative of ssSO 2− 4 depletion in sea salts measured in the core in comparison to bulk seawater, with ssSO 2− 4 depletion factors of two for the period 1737-1949, and three for the period 1950-2000.
The negative nssSO 2− 4 values found in the FIS records could be explained by an enhanced input of sea salts from (i) windblown frost flowers and/or (ii) aerosol formed after fractionated salty-snow sublimation, with both (i) and (ii) being formed in the neighbouring waters at the eastern flank of FIS.Yang et al. (2008) have reported that aerosol production via (ii) can be more than onefold larger per unit area than sea-salt production from the open ocean.There is no (or a very limited) amount of multi-annual sea ice near FIS and young sea ice formed during winter in the vicinity of the S100 site is quickly covered by snow due to cyclonic activity.Trajectory studies of air with high sea-salt concentrations and low SO 2− 4 / Na + ratios arriving at Halley station, showed that these air masses mainly originate in regions where young sea ice and frost flowers are formed (Hall and Wolff, 1998;Rankin and Wolff, 2002).However, conditions at Halley are not comparable to FIS, as the main easterly or northnortheasterly wind direction prevailing at Halley means an off-land air flow, thus creation of polynyas with open water and consecutive new ice formation, whereas at FIS, and most of the Dronning Maud Land coast, the wind is mainly parallel to the coast or even slightly towards the coast.In particular, a quantification of the areas covered by frost flowers is still missing.It is possible that those areas are comparatively small due to the generally high wind speeds prevailing above the Southern Ocean, resulting in a high percentage of frazil ice, and synoptic conditions lead to the quick development of a snow cover on the young sea ice.Although it is not possible to apportion the contribution of fractionated sea salts via (i) or (ii) with the current data, it is plausible that a larger contribution of fractionated aerosol formed from salty-snow than by frost flowers, based on recent experimental evidence that frost flowers would not be a direct source of sea-salt aerosols (Yang et al., 2017).In addition, frequent stormy conditions in the area are detrimental for the formation of frost flowers, which form under quiet, undisturbed conditions, usually only in leads or small polynyas under the influence of anticyclonic weather.This also means low wind speeds and thus not much transport of frost flowers to the sampling sites at FIS.Thus, mechanism (ii), blowing salty snow formed on thin sea ice that sublimates during transport to form sea-salt aerosols, appears to be a much more probable explanation considering the local meteorological conditions in the study area.
Considering that we found no correlation between ion concentrations and site elevation, a decrease in wind transport efficiency of frost flowers (size of 10-20 mm) and aerosol formed via (ii) (size >0.95 µm) (Seguin et al., 2014) due to increased elevation cannot be addressed to explain the lower sea-salt values observed at the ice rises compared to the S100 site.Local effects on annual SMB due to topography and meteorology at the KM and BI sites reported by Vega et al. (2016) are most likely involved in the different load of sea salt to these sites.
The dramatic increase in fractionated sea salt in the S100 core after the 1950s could be associated with a greater exposure of the S100 site to primary aerosol, in addition to an enhanced production of fractionated aerosol, evidenced by a dominance of negative nssSO 2− 4 values after 1950. Figure 2a  and c show that sea salts started to increase after 1950 with a marked peak corresponding to the year 1966 (±3 years).According to Rignot et al. (2011), ice velocities near S100 were in the order of 100s m y −1 for the period 2007-2009; therefore, the S100 site has moved at least 5 km closer to the ice front between 1950 and 2000.In addition, as can be seen in the FIS contours shown in Fig. 1, by 1963, the ice front in the vicinity of the S100 site was located about 17 km further north than its present position.As a consequence, the distance from the S100 site to open water would have changed through time and this would likely have caused a greater inwww.the-cryosphere.net/12/1681/2018/The Cryosphere, 12, 1681-1697, 2018 put of sea spray to the S100 site by shortening the distance to the open sea.However, sea spray enhancing alone cannot account for the increase of sea-salt concentration and the negative nssSO 2− 4 found in the S100 samples.We hypothesize that the calving of Trolltunga, which occurred in 1967 (Vinje, 1975) (Fig. 1), modified the sea ice conditions north of the S100 coring site, enhancing the input of fractionated sea salts, which then caused the marked sea-salt peak appearing in 1966 (±3 years).Furthermore, additional support for an effect of the Trolltunga calving event is the fact that negative nssSO 2− 4 slowly decreased between 1950-1966, showing a marked minimum around 1966 (±3 years) (Fig. 4b).The much extended Trolltunga, which was present before the calving event would have formed a larger bay to the east, favourable for a higher concentration of sea ice caused by the prevailing easterly winds, resulting in both thicker and longer-lasting sea ice which would limit sea-spray formation.Such thick sea ice does not seem to form under postcalving event conditions, e.g. with a shorter tongue.In order to explain the fractionated sea-salt values detected in the S100 core, there must be an enhanced source of fractionated sea salts after the calving event.This would be the case if young sea ice (where fractionation of sea salts can take place) formed near the S100 site as a result of the greater area of open sea available after the calving event.Thicker, longlasting sea ice present before the calving event would have been a more stable substrate, prone to less flooding through cracks and leads, and would also likely have reduced salinity content in the snow compared to young sea ice (Massom et al., 2001).Considering the conjecture by Rhodes et 4 vs. ssNa + concentrations in the S100 core.nssSO 2− 4 was calculated using the seawater ratio as described in Sect.2.3 and using a k seawater = 0.06 (in µmol L −1 ).Positive nssSO 2− 4 values are denoted with black dots, while negative values are denoted with blue dots.A linear regression was calculated using all nssSO 2− 4 data points to infer corrected k value (k ), following the approach by Wagenbach et al. (1998).
al. (2017), i.e. that young sea ice is more saline than multiyear ice, it can be expected that sea-salt aerosols produced by blowing snow over sea ice would have higher sea-salt concentrations when young ice is formed than when multi-year sea ice is formed, in coherence with the proposed hypothesis.The higher sea-salt concentrations and the negative nssSO 2− 4 values found in S100 after the Trolltunga calving, could thus be explained by a combination of shortening the distance from the S100 site to the ice shelf front, and an enhanced contribution of sea-salt aerosols entrained by blowing salty snow found over young sea ice formed near the S100 site.If the air is unsaturated, water in these snow particles will sublimate, producing fractionated sea-salt aerosol.As schematized in Fig. 2 in Rhodes et al. (2017), the sea-salt aerosol can be transported inland and be deposited either by dry or wet deposition.Since sea-salt concentrations are much higher at the S100 core than in the ice-rise cores, it is plausible that most of the flux of sea salts at the S100 site is due to dry deposition, due to the short distance from the coast and low elevation, while deposition at the ice rises would be balanced between the wet and dry regimes.Rhodes et al. (2017) found a marked gradient in the ratio between sea-ice sea salts and oceanic sea salts (produced by bubble bursting) with distance to the source, in Arctic sites, with higher ratios closer to the sea ice source and when the sample location is in the path between sea ice and prevailing winds.To test the hypothesis presented here, a closer analysis of satellite and historical sea ice data and a model-based study to estimate the spatial and elevation gradient of sea-ice sea salts to FIS can be done; however, this is beyond the scope of the present study.
Other possible mechanisms, such as deposition of sea salts with rime or windblown snow present over multi-annual sea ice, can explain neither the increase in sea-salt concentration nor the fractionation observed in S100 after the 1950s.Additionally, annual averages of monthly zonal and meridional wind speeds (ERA40, Uppala et al., 2005) for the area (69-71 • S, 3.5 • W-5 • E) between 1955-2001 (Fig. 6) (Fig. 6) show no significant positive trends, thus evidencing that the S100 seasalt increase after the 1950s cannot be related to enhanced transport by wind.
Due to the limited time coverage of the KC, KM, and BI cores, we do not know whether there was a relative increase in sea-salt concentrations in the ice rise cores after the 1950s influenced by the Trolltunga calving.Due to the large input of bio-nssSO 2− 4 to the ice-rise sites, any possible signal of fractionated sea salts in any of the ice-rise cores could be easily masked by the biogenic fraction (e.g.no significant negative nssSO 2− 4 values would be observed).Relatively higher sea-salt concentrations measured in the KM core in comparison to the other ice-rise cores could be explained by a combination of distance to the sea and the prevailing precipitation and wind conditions in the area: precipitation on FIS is mainly caused by frontal systems of cyclones in the circumpolar trough that move eastwards north of the coast, thus leading to easterly or east-north-easterly surface winds on FIS (Schlosser et al., 2008).This means that even though BI is equally close to the sea as KM (Fig. 1), KM has by far the shortest distance to the source of marine aerosols of all three ice-rise cores, which could explain the comparatively high sea-salt concentrations (Table 2).

Conclusions
This study reports sub-annual and long-term temporal sea salt and major ion concentration changes measured in three recently drilled firn cores from different ice rises located at Fimbul Ice Shelf (FIS): Kupol Ciolkovskogo, Kupol Moskovskij, and Blåskimen Island, and a 100 m long core drilled near the FIS edge (S100).A significant increase in sea salts is observed in the S100 core after the 1950s, which is associated with an enhanced exposure of the S100 site to primary sea-salt aerosol associated with a shorter distance from the S100 site to the ice front during the last decades, and an enhanced input of fractionated sea salts.This increase in seasalt concentrations was accompanied by a shift in nssSO 2− component associated to the concentration gradient between the S100 site and the FIS edge.In contrast, the ice-rise cores record the signal of larger-scale conditions of atmospheric flow, large inputs of bio-nssSO 2− 4 , and less efficient transport of sea salts evidenced by lower median concentrations than at the S100 site.These findings are of vital importance for the understanding of the mechanisms of sea-salt aerosol production, transport, and deposition at coastal Antarctic sites, and for the improvement of the current Antarctic sea ice reconstructions based on sea-salt chemical proxies.

Figure 1 .
Figure 1.Satellite image of Fimbul Ice Shelf (FIS) showing the KC, KM, BI, and S100 core sites, Jutulstraumen and Trolltunga.In addition, 50 m contours are shown at each ice rise, as derived from GPS profiles (Vikram Goel, personal communication, 2016).In addition, the dashed line shows the extent of Trolltunga according to Corona Satellite data from 1963 (Jelte van Oostveen, personal communication, 2017).Map image is from the MODIS Mosaic of Antarctica (MOA).Additional information regarding the sampling sites and traverses in FIS can be found in Schlosser et al. (2014) and Vega et al. (2016).

Figure 2 .
Figure 2. Annual sea-salt (Cl − and Na + ) concentrations and surface mass balance (SMB) in the two longest records retrieved at Fimbul Ice shelf, S100, (a, c, e), and KC, (b, d, f).Linear trends in Cl − and Na + concentrations, and SMB measured in the S100 core are shown for two different periods: 1737-1949 (blue dashed line) and 1950-2000 (black dashed line) in (a, c, e), respectively.Linear trends in Cl − and Na + concentrations, and SMB measured in the KC core are shown for the period 1958-2007 (black dashed line) in (b, d, f), respectively.Significance, slope, and standard error of the linear regressions are given in TableS3.

Figure 3 .
Figure 3. Sub-annual variability of selected ions: Cl − and Na + (a), Mg 2+ and Ca 2+ (b), MSA, (c) and NO − 3 and SO 2− 4 (d) in cores KM, BI, and S100.Mean summer and winter concentrations were calculated for the months NDJFMA and MJJASO, for a period of 16, 15, and 5 years in the KM, BI, and S100 cores, respectively.

Figure 5 .
Figure 5. Scatter plot of annual nssSO 2−4 vs. ssNa + concentrations in the S100 core.nssSO 2− 4 was calculated using the seawater ratio as described in Sect.2.3 and using a k seawater = 0.06 (in µmol L −1 ).Positive nssSO 2− 4 values are denoted with black dots, while negative values are denoted with blue dots.A linear regression was calculated using all nssSO 2− 4 data points to infer corrected k value (k ), following the approach byWagenbach et al. (1998).

Figure 6 .
Figure 6.Annual averages of monthly (a) zonal and (b) meridional wind speeds (ERA40) for the area (69 • S-71 • S, 3.5 • W-5 • E) between 1958-2001.Slope, standard deviation, and p value of the linear regression are shown in the figure.

Table 1 .
Cores (KC, KM, BI, S100) locations and sampling details.Distances of the core locations to the ice shelf side were obtained using the GIS package Quantarctica (www.quantarctica.org, last access: 9 May 2018).

Table 2 .
Vega et al. (2016)rations (in µmol L −1 ) in the KC, KM, BI, and S100 firn and ice cores.Ion concentrations at the top 2 m of the KC, KM, and BI cores were not measured.Non-detected concentrations were set as half the detection limit of each ion.Values of water stable isotopes and deuterium excess for the KC, KM, and BI are reported byVega et al. (2016).

Table 3 .
PCA loadings of the first three (KC) and two (KM, BI, and S100) principal components calculated at an annual resolution in a set of eight different ions measured in the ice rises and S100 cores.PCA loadings were obtained at three different time intervals in the S100 core:1737-2000, 1737-1949, and 1950-2000.Sources related to the different components are displayed in the bottom row.