During freeze-up and consolidation, sea ice rejects to its surface brine of marine origin that is incorporated into overlying snow. To evaluate the transport of biological components in brines from ice to snow, vertical profiles of temperature, salinity, bacterial abundance, and extracellular polysaccharide substances (EPS) were obtained through snow and first-year sea ice (Barrow, AK) in consecutive winters (2010, 2011). Snow profiles showed strong interannual variation, with 2010 presenting higher values and wider ranges in salinity (0.3–30.9, practical salinity), bacterial abundance (2.8 × 102–1.5 × 104 cells mL− 1), and particulate EPS (pEPS, 0.04–0.23 glucose equivalents (glu-eq) mg L− 1) than 2011 (0–11.9, 2.7 × 103–4.2 × 103 cells mL− 1 and 0.04–0.09 glu-eq mg L− 1, respectively). Surface ice also differed interannually, with 2010 presenting again higher salinity (19.4, n = 1), bacterial abundance (5.4 × 104–9.6 × 104 cells mL− 1) and pEPS (0.13–0.51 glu-eq mg L− 1) than 2011 (7.7–11.9, 1.7 × 104–2.2 × 104 cells mL− 1, and 0.01–0.09 glu-eq mg L− 1, respectively). Transport of bacteria and pEPS from sea-ice brines into snow was evident in 2010 but not 2011, a year with more extreme winter conditions of colder temperature, thinner snow, and stronger wind. By size fraction, the smallest EPS (< 0.1 µm) dominated (> 80%) total EPS in both ice and snow; the > 3 µm fraction of EPS in snow appeared to have an atmospheric source. Evaluation of membrane integrity by Live/Dead stain revealed a high percentage (85%) of live bacteria in saline snow, identifying this vast environment as a previously unrecognized microbial habitat.
 During its consolidation and cooling, Arctic first-year sea ice (FYI) expels to its surface a thin layer of brine [Perovich and Richter-Menge, 1994; Ehn et al., 2007]. This surface brine can be transported by capillary action into frost flowers [Drinkwater and Crocker, 1988] or snow [Barber et al., 1995]. Bacteria (referring collectively to Bacteria and Archaea) and organic material present in sea-ice brines [Junge et al., 2001, 2004; Krembs et al., 2002, 2011] are also available for transport. The incorporation of sea-ice bacteria and extracellular polysaccharide substances (EPS) into short-lived saline frost flowers has been shown in both field and laboratory settings [Bowman and Deming, 2010; Aslam et al., 2012], but their transport and long-term (seasonal) persistence have not been demonstrated for snow over sea ice.
 The majority of the Arctic sea-ice surface, with an estimated extent of 7× 106 km2 in winter [Nghiem et al., 2007], is covered with snow, mostly accumulated in early fall with some additional accumulation in late winter and early spring [Warren et al., 1999; Sturm et al., 2002]. Snow deposited over new ice incorporates sea-ice brines from the ice surface or from collapsed frost flowers formed prior to snowfall [Drinkwater and Crocker, 1988; Barber et al., 1995; Massom et al., 2001]. The brine-wetted basal snow, referred to as saline snow, has bulk salinities > 10 ppt and up to 40 ppt in the fall, with salinity decreasing through winter due to gravity drainage [Langlois et al., 2007]. Salinity decreases upward from the basal stratum, reaching < 5 ppt at 10 cm above the ice surface [Langlois et al., 2007; Barber et al., 1995].
 Although the physical and chemical properties of saline snow have been examined in some detail [Langlois et al., 2007; Domine et al., 2004; Poulain et al., 2007; Geldsetzer et al., 2009; Barber et al., 2003], microbial studies targeting brine-wetted snow are not available. The few reported measurements of bacterial abundance in snow over Arctic sea ice range from 9 × 102 to 5.2× 105 cells mL− 1 melted snow, with lowest values detected in upper spring layers [Møller et al., 2011] and highest values in lower summer layers [Poulain et al., 2007]. Bacterial deposition with snow grains or atmospheric particles has been documented [e.g., Segawa et al., 2005], but the potential of sea-ice brines as a source of bacteria (or EPS) to overlying snow has not been considered. Yet, substantial numbers of bacteria are present in the brines of upper sea ice and available for upward transport. A typical bacterial abundance (scaled to brine volume) in upper sea-ice brines of late fall FYI is 7×105 cells mL− 1 [Collins and Deming, 2011], with midwinter values as high as 2× 108 cells mL− 1 [Collins et al., 2008].
 Upper FYI also contains high levels of EPS, with bulk winter concentrations of 1.5 mg C L− 1 for particulate EPS (pEPS; >0.45 µm) [Krembs et al., 2002]. Even higher values of pEPS (2–570 mg C L− 1) were measured in brines expelled onto the surface of new ice formed in spring [Bowman and Deming, 2010]. According to mesocosm work, the proportion of EPS in dissolved form (dEPS) increases in surface ice as it matures [Aslam et al., 2012]. We can find no reports of EPS concentration in snow over sea ice (or over ground).
 The objectives of this study were to examine the potential upward transport of bacteria and EPS in sea-ice brines, based on vertical profiles of biological and physical parameters for snow and underlying FYI. Results, interpreted in the context of salinity, ion composition, and meteorological information, underscore the influence of atmospheric conditions on the transport and persistence of brine, sea-ice bacteria, and EPS in snow throughout winter. Because the saline snow layer represents an areally vast habitat that persists until spring melt [Barber et al., 1995], it offers a much longer timeframe and greater spatial scale than the frost-flower habitat for interaction between marine and atmospheric components. This interaction is of particular importance in view of current changes in the Arctic sea-ice cover, where the expected increase in FYI will also lead to an increase in the extent of saline snow.
2.1 Study Site and Sampling Approach
 Samples of snow and underlying ice were collected from landfast FYI near Barrow, AK, during 10–15 February 2010 (BW'10) and 7–11 March 2011 (BW'11). Sampling sites were selected by their proximity (300 m) to the UAF Barrow Sea Ice Mass Balance Observatory Site (MBS; 156.5°W, 71.4°N [Druckenmiller et al., 2009]), 13 km from the NOAA weather station (156.8°W, 71.3°N; [National Operational Hydrologic Remote Sensing Center, 2011]; Figure 1). Sites were flat extensions of undeformed ice, although the BW'11 site was surrounded by rubbled ice. Samples were collected from random, undisturbed snow patches 20 m apart. BW'10 snow patches (n = 4) had different depths of 6, 9, 10, and 19 cm. BW'11 snow patches (n = 5) were uniform in depth at 8–9 cm, the average snow depth in a 100 m snow-depth transect (see Figure S1 in the supporting information). Additional samples (n = 17) were collected from two snow patches 3 cm deep.
2.2 Sample Collection and Physical Measurements
 Snow layers, hardness, and vertical profiles of temperature and salinity (3 cm intervals) were measured in snow pits 60 cm wide according to Sturm  with zero-depth defined at the ice-snow interface. Density and grain size were also determined during BW'11. Temperature was measured by a handheld temperature probe (precision 0.1°C); density was measured with a 100 cm3 cutter, and spring scale (Taylor-LaChapelle snow density kit, Model ST-2, Hydro-Tech). Snow grain size was visually estimated using a handheld magnifying glass and a gridded card. Samples of 0.1 L were collected for salinity.
 For biological profiles, 3–6 L of snow were collected from a new wall, following either layers of different hardness (BW'10, 0–2, 2–3, 3–6, 6 cm–surface) or uniform 3 cm intervals (BW'11). Samples were collected starting with the uppermost interval, removing snow until the required volume was achieved. To ensure that similar volumes of snow were collected from intervals of different thickness (BW'10), snow from the upper layers was carefully removed and set aside until enough material from the lower layers was exposed. Horizontal variability in salinity was measured in a 3 cm deep snow patch by selecting random quadrats (n = 13) from a 61×61 cm square frame divided into 10×10 cm quadrats (BW'11). Samples were collected by removing the full depth of snow from each quadrat. Another 3 cm deep snow patch was sampled to determine bacterial viability in thin snow fully exposed to the atmosphere (2 L snow, n = 4; 1 L surface ice, n = 4). All volumes collected refer to unmelted samples.
 The ice surface was cleaned of snow and, immediately upon exposure, punctured with a chisel, opening a 1 cm deep slot for the temperature probe measurement. Adjacent ice surface was then fragmented with a chisel or ice pick to a depth of 2–3 cm. Ice fragments were collected for salinity (0.1 L) and biological (1 L) measurements. The full thickness of sea ice (1.1 m) was sampled with a CRREL-style core barrel of 8.5 cm inner diameter (BW'10). Following Pringle and Ingham , temperature was measured in two ice cores at 10 cm intervals, and salinity on melted 10 cm sections. Four additional ice cores were cut into 10 cm sections for pEPS and bacterial abundance. All biological samples from snow and ice were collected using ethanol-rinsed tools, placed in sterile Whirl-Pack bags at in situ air temperature and transported in an insulated cooler for immediate processing at the Barrow Arctic Research Center (BARC).
2.3 Salinity and Ion Measurements
 Bulk salinity was measured on melted samples using a handheld YSI-30 conductivity meter (accuracy ± 0.1) and reported as practical salinity (Sp) according to the practical salinity scale (PSS-78; UNESCO, 1981). Samples may have composition anomalies, hence salinity values serve only as a reference of brine content [Millero et al., 2008]. Brine salinity and volume fraction were calculated using the phase equations for ice, as applied to snow by Drinkwater and Crocker  and others [e.g., Barber et al., 2003; Langlois et al., 2007]. The specific phase equations from Cox and Weeks  were used, and brine volume fraction was expressed as fraction of total volume of ice plus brine (i.e., not including volume of air in the sample). Aliquots of one snow sample (BW'10) were stored at 2°C to determine the concentration of major ions (Ca2 +, K+, Mg2 +, Na+, Cl−, SO ) at the University of Washington (UW) with an ion chromatography Dionex 120 and inductively coupled plasma-atomic spectrometry (ICAP 61E Model, Thermo Jarrell Ash Co.).
 Enrichment factors (Ef) were calculated according to Toom-Sauntry and Barrie  (equation (1)) and using molar concentrations from Pilson :
where [X] and [Mg2 +] are the molar concentrations of the ions X and Mg2 + (the latter used as reference ion because Na+ precipitates at low temperature and Cl− is scavenged by snow).
2.4. Biological Measurements
2.4.1. Melting Protocols
 To determine bacteria losses due to osmotic shock during sample melting, two types of melts (direct and saline) were compared for all samples, except BW'10 snow (direct only), BW'10 surface ice (saline), and BW'11 viability assay (saline). Concentrations were scaled to original volume of sample (bulk-scaling). Saline solutions (milli-Q water and Sigma Sea Salts or NaCl) were prepared to a final concentration of 265–270 ppt and filtered (0.22 µm) at BARC (BW'10) or filtered and autoclaved in advance at UW (BW'11). Chilled saline solution was added to the sample to obtain a final meltwater salinity of 100 ppt (BW'11) or similar to in situ brine salinity (BW'10). The choice of 100 ppt was informed by laboratory experiments, where bacterial isolates from frost flowers did not experience significant cell loss when exposed to a salinity change from 220 to 100 ppt (data not shown). Samples were melted at room temperature (shaking often, processing immediately) or in a 0°C water bath. Because bacteria and EPS partition into the brine phase of sea ice [Junge et al., 2001; Krembs et al., 2011], bulk concentrations were also scaled to brine volume (brine-scaling), first subtracting background levels from nonsaline snow (if background exceeded sample level, no brine scaling was made).
2.4.2 Bacterial Abundance
 Total bacterial abundance was determined by epifluorescence microscopy as in Collins et al. , fixing 15–45 mL of sample with 0.22 µm filtered formaldehyde (final concentration of 2%) and processing at UW (BW'10). During BW'11, slides were prepared at BARC by filtering 70–300 mL onto 0.2 µm polycarbonate filters, then fixing with formaldehyde and staining with 4′,6-diamidino-2-phenylindole. Samples were kept in the dark at 4°C until counting at UW within 1 month of collection.
2.4.3 Extracellular Polysaccharide Substances
 To determine pEPS, 150–300 mL was gently filtered (< 350 mm Hg) onto 0.4 µm pore-size polycarbonate filters. Additional BW'11 samples were passed through consecutive polycarbonate filters: 800 mL onto a 3 µm pore-size filter (pEPS > 3 µm), 400 mL of filtrate onto a 0.4 µm pore-size filter (pEPS < 3 µm), and 50 mL of the resulting filtrate onto a 0.1 µm pore-size filter (dissolved EPS (dEPS) > 0.1 µm). The 0.1 µm filtrate was stored in acid-washed plastic bottles (dEPS < 0.1 µm). Samples were kept frozen at –20°C until processed at UW within 1 (pEPS) and 5 (dEPS) months. Filters were resuspended in 0.25 mL of artificial seawater. Filtrates for dEPS < 0.1 µm were first dialyzed (Thermo Scientific SnakeSkin dialysis tubing, 3500 MWCO) for 24 h at 2°C until reaching a salinity of < 2 ppt; dialysates were held at 80°C for 3–4 days until volume evaporated to 1–2 mL. The phenol-sulfuric assay of Dubois et al.  was performed on the resuspensions and concentrated dialysates to quantify EPS as glucose equivalents (glu-eq) as in Krembs et al. .
2.4.4 Bacterial Viability
 The Live/Dead assay (BacLight kit L-13152, Invitrogen) was applied to saline snow and surface ice samples kept frozen at –20°C and processed at UW within 2 months. The protocol of Tam et al.  was modified to avoid thermal or osmotic shock of sea-ice bacteria as follows: melted samples were filtered onto a 0.22 µm polycarbonate filter at 2°C; the filter was resuspended into 0.5 mL of chilled, filtered (0.22 µm), 100 ppt NaCl brine and incubated (30 min, 2°C) with 0.5 mL 2:1 mixture of SYTO 9 and propidium iodide. A minimum of 20 fields and 200 “live” cells was counted with a Zeiss Universal epifluorescence microscope. The modified protocol was tested against cultures (2°C, 35 ppt) of the marine psychrophilic bacterium Colwellia psychrerythraea strain 34H, confirming detection of “live” cells (preserved membrane) in exponential growth phase and “dead” cells in death phase.
2.5 Graphical and Statistical Analysis
 To evaluate if sea-ice bacteria, EPS, and salts were incorporated in the same proportion in snow, the expected contents of bacteria (Be) and pEPS (pEPSe) were calculated assuming a 1:1 transport with the salts: Be = Ba + Bb, where Ba is bacteria from the atmosphere (average bacterial number in snow samples with the lowest salinity, < 0.5) and Bb is bacteria from the brine (calculated by multiplying bulk snow salinity with the ratio of bacteria-to-salts in surface sea ice). Similar calculations were made for pEPS. The movement of bacteria or pEPS was assumed not to exceed that of salts. Graphical and statistical analyses were performed with R v.2.13.1 [R Development Core Team, 2011]. Statistical difference between groups was determined using independent two-sample t-test, paired t-test, and analysis of variance with repeated measures. A Mann-Whitney test was used in cases where the normality assumption did not hold. Nonparametric Spearman's correlation analysis was used to evaluate relationships between variables.
3.1 Meteorological and Environmental Conditions
 Interannual comparison of parameters expected to influence the sampled microbial communities revealed more extreme conditions for BW'11 relative to BW'10 (Table 1): thinner snow depth, higher wind speed and, for comparable days in January, lower temperature and higher brine salinity at the ice surface. Seasonal and daily fluctuations in temperature, brine salinity, and brine volume fraction during the 40 days leading to BW'11 were calculated for four environments: air (indicating conditions experienced by snow directly exposed to the atmosphere), ice-snow interface, and 10 cm and 20 cm depths into the sea-ice column. Daily fluctuations in all parameters were strongest, on average, for snow exposed to the atmosphere (Figure 2). Compared to the ice column, temperature fluctuations were 16 times greater at the ice surface and 80 times greater in snow exposed to the atmosphere.
Table 1. Meteorological Parameters
Average (minimum to maximum) from NOAA-PABR Station for the period from ice formation (mid-November) to sampling week (10–15 February 2010; 7–11 March 2011)
Daily averages from the UAF Mass Balance Site for 23 January 2010 and 26 January 2011.
 Average snow density (BW'11) ranged from 286 ± 11 kg m− 3 in the basal stratum to 320 ± 29 in the upper stratum (n = 5). Many profiles had an icy layer at 2 cm above the ice surface, with a hardness value of blade (see additional measurements of snow layers, hardness, and grain size in Table S1 of the supporting information). Melted samples of saline snow contained a white mineral crystal metastable at room temperature, possibly a polymorph of CaCO3 as found in Fischer et al. .
 For snow profiles of comparable depth, both years showed similar temperature and brine salinity, but bulk salinities and brine volume fractions were lower during BW'11 (Figure 3). Temperature decreased upward at a rate of –0.41 to –0.51°C cm− 1 in the snow and –0.12°C cm− 1 in the ice column (Figure 3a; see full ice-core data in Figure S2 of the supporting information). Brine salinity, determined by temperature, had an opposing trend, reaching highest values in the snow (Figure 3b). Bulk salinity and brine volume fraction were highest at the ice-snow interface and lowest in the upper layers of snow (Figures 3c and 3d). Saline snow showed limited horizontal variability in bulk salinity (11.3 ± 0.67; mean ± SD, n = 13; Figure 3c).
 Enrichment of major ions relative to Mg2 + (Figure 4; see bulk concentrations in Table S2 of the supporting information) revealed strong sulfate depletion in the snow. Na+ showed moderate depletion in the bottom 6 cm, whereas Ca2 + and K+ showed slight depletion at all depths. Cl− was enriched up to 6 cm above the ice.
3.3 Bacterial Measurements
 For both years, bacterial abundance in snow (2.8× 102–1.5× 104 cells mL− 1) was always lower than in the uppermost 3 cm of ice (“surface ice”, 1.3× 104–9.6× 104 cells mL− 1) which, in turn, was lower than in the upper 10 cm section of the BW'10 ice cores (“ice column”, Figure 5). During BW'10, bacterial abundance in snow declined upward from the ice surface, correlating significantly with depth and salinity (Figure 5a and Table 2), with the minimum value of the study recorded at 4.5 cm. In contrast, all snow depths during BW'11 had similar bacterial abundances, with a higher minimum value (5.8× 103 cells mL− 1) and no correlation with depth or salinity (Figure 5b and Table 2).
Table 2. Correlation Coefficients (Spearman's) for Parameters in Snowa
Bold highlights significant relationships at p < 0.05 (*) and p < 0.005 (**).
Direct melts for 2010, saline melts for 2011.
 Bacterial abundance in snow was lower than expected for a 1:1 transport with the salts (Figure 5), with ratios of bacterial abundance to salts decreasing 7- to 10-fold from ice column to surface ice, and surface ice to saline snow (Figure 5). A high percentage of “live” bacteria, determined by membrane integrity (Live/Dead stain), was observed for both saline snow (85% ± 4.6, n = 4) and surface ice (78% ± 6.9, n = 4) during BW'11.
 Abundances in the BW'10 sea-ice column were distributed in a C-shaped profile (Figure S3 of the supporting information). No differences due to melting procedure were observed, except for the upper section where a 55% cell loss was observed for direct melts (n = 1, Figure 5a). Significant cell loss was observed during BW'11 in direct melts of surface ice (30% lower, p = 0.009, t-test) and the medium (3–6 cm) snow horizon (20% lower, p = 0.041, t-test).
 The sea-ice column had a relatively constant concentration of pEPS (but for the maximum value in the upper section, Figure S3 in the supporting information), with a mean pEPS value (0.740 ± 0.73 mg glu-eq L− 1, n = 22) significantly higher than in snow (0.098 ± 0.07 mg glu-eq L− 1, n = 11; p < 0.001, Mann-Whitney). During BW'10, pEPS concentration in the snow decreased upward from the surface ice, with a significant negative correlation with depth (Table 2). pEPS appeared to follow the expected 1:1 transport with salts (Figure 6a). Although the correlation was not significant (p = 0.096), the low p-value provides evidence against the null hypothesis that pEPS does not correlate with salinity. During BW'11, pEPS correlated significantly with salinity but not with snow depth (Table 2). Unlike ratios of bacterial abundance to salts, the ratio of pEPS to salts decreased only by half between surface ice and saline snow during BW'10 and doubled during BW'11 (Figure 6).
 The mean total EPS (carbon equivalents) in BW'11 snow was 0.27 mg C L− 1 ± 0.05 (n = 15). The smallest dEPS size fraction (< 0.1 µm) accounted for 85–90% of the total EPS in both surface ice and snow (Figure 7). Concentrations of the remaining EPS size fractions (dEPS > 0.1 µm, pEPS < 3 µm, and pEPS > 3 µm) differed significantly from each other at each snow depth (p < 0.01, analysis of variance with repeated measures), with the largest pEPS fraction having the highest concentrations at all snow depths (p < 0.001, paired t-test, Figure 7). Concentrations of these same three EPS size fractions in the underlying ice did not differ significantly from each other.
3.5 Additional Scalings for Bacteria and EPS
 Brine-scaling of bacterial abundance and EPS concentration (Table S3 of the supporting information) indicated similar patterns to bulk-scaling, with both parameters lower in saline snow than in surface sea ice, except during BW'11 when pEPS was slightly higher in the saline snow layer. Scaling to brine, which reflects the in situ conditions experienced by bacteria, also allowed comparisons of the amount of presumably protective pEPS available on a per cell basis. Highest values characterized saline snow (15 and 100 pg glu-eq per cell during BW'10 and BW'11, respectively), and lower values, surface sea ice (4.3 and 2.3 pg glu-eq per cell during BW'10 and BW'11, respectively). Likewise, total values of EPS per cell (BW'11) were two orders of magnitude higher in saline snow (3360 pg glu-eq per cell) than in surface ice (30 pg glu-eq per cell).
4.1 Physical Properties
 Ionic composition (Figure 4) confirmed that the source of salt in the snow we sampled was sea-ice brine and not seawater flooding (sulfate and sodium depletion is associated to mirabilite precipitation in brines exposed to temperatures < −8°C [Rankin et al., 2000]). Furthermore, brine content decreased away from the ice-snow interface, as expected for fluids incorporated by capillary rise (Figure 3c, Coléou et al. ). Low salinity snow samples had higher ion concentrations by 1–3 orders of magnitude, and a stronger depletion signal, than reported for snowfall in the high Arctic [Toom-Sauntry and Barrie, 2002]. Background salinity may include salt from blown saline snow or other atmospheric depositions [e.g., Barrie et al., 1985; Kumai, 1985]. Cl− enrichment was likely due to the scavenging of surface reactive chloride compounds by the snow [Toom-Sauntry and Barrie, 2002].
 Compared to literature values, bulk snow salinities were generally high in BW'10 and low in BW'11 [Barber et al., 2003; Langlois et al., 2007]. BW'11 had a longer lead time for brine gravity drainage [Langlois et al., 2007], but other factors may have contributed to the interannual differences. Air temperature during freeezup, lower in BW'10 (–18.6°C) than BW'11 (–9.9°C), may have resulted in faster freezing rates and a greater expulsion of brine then available for incorporation into snow. Snow porosity and grain size [Coléou et al., 1999] also affect fluid incorporation into snow, but measurements for the time of snow deposition are lacking.
 Snow had lower bulk salinity (0–39) than reported for frost flowers (10–120 ppt) [Perovich and Richter-Menge, 1994; Bowman and Deming, 2010], yet held a greater amount of salt per unit area of sea ice covered. The volume of snow covering 1 cm2 would hold 37–64 mg salt in BW'10 (assuming snow density of 300 kg m− 3) and 9–12 mg in BW'11. Frost flowers would hold up to 6 mg salt per cm–2 (assuming weight per unit area of 25–50 mg cm–2 and occupation of one-third of the available surface [Perovich and Richter-Menge, 1994]). The higher salt content in snow per unit area supports the hypothesis that saline snow is a more important source of bromide salts and other salt aerosols to the atmosphere than frost flowers [Simpson et al., 2007; Yang et al., 2008; Roscoe et al., 2011].
4.2 Bacterial and EPS Measurements
 Total bacterial abundances in the snow were low to intermediate compared to the few reported values for snow on sea ice from milder seasons [Poulain et al., 2007; Møller et al., 2011]. Influence of melting method on bacterial loss for the upper section of the ice was consistent with initial work by Deming . Previous work by Helmke and Weyland  showed no effect of melting protocols on the culturability of bacteria from winter sea-ice cores, but culturable bacteria are only a fraction of the total population in sea ice [Junge et al., 2002].
 Although estimates of EPS by the phenol-sulfuric assay method (assayed at absorbance wavelength of 490 nm) can underestimate acidic sugars [Dubois et al., 1956] known to be present in EPS [Underwood et al., 2010], EPS concentrations in the sea-ice column were consistent with the winter pEPS profile determined by Collins et al.  using the Alcian Blue staining method. We found no measurements of EPS content in saline snow for direct comparison, but total dissolved organic carbon values reported for snow above the saline layer (1.3–4 mg C L− 1) were 5 to 15 times higher [Møller et al., 2011]. Atmospheric sources of dissolved organic carbon in snow, including EPS from aerosols, have been previously considered [Leck and Bigg, 2008]. On average (mean ± SD), total dEPS accounted for 87% ± 6 in upper low-salinity snow, with saline snow at 94% ± 2 and surface ice at 93% ± 2. These percentages agree with dominance of the dEPS fraction in new surface sea ice and frost flowers grown in mesocosms by Aslam et al. , and observed in late spring sea ice (bottom 10 cm) in Barrow (72% ± 9.5) by Krembs et al. . Differences in the relative abundance of the two evaluated pEPS size fractions, with the largest fraction (> 3 µm) being dominant in the snow but not in the ice, suggest atmospheric contributions of pEPS to the snow. Correlation with salinity, however, which was negative in previous studies of sea ice [Collins et al., 2008] and sea-ice brines [Aslam et al., 2012], was significantly positive in BW'11 snow, indicating possible marine origin of the pEPS.
4.3 Bacterial and pEPS Transport in the Saline Snow Layer
 In the first year of this study, bacteria and pEPS (> 0.45 µm) in the snow followed a gradient similar to salinity and had significant relationships with snow depth (Table 2) suggesting upward transport and incorporation with sea-ice brines. Upward transport of marine bacteria and pEPS was not clear in the second year, when higher background levels of bacteria and pEPS may have masked a comparatively smaller contribution from the brines. Higher pEPS values in surface ice during BW'10 (Figure 6) and higher background levels in snow during BW'11 (0.09 mg glu-eq L− 1 vs. 0.04 mg glu-eq L− 1) likely contributed to these differences in gradients.
 In both years, fewer bacteria were detected in the snow than expected from the concentration of salts (Figure 5). One cause could be their selective retention in the ice due to ice affinity of the bacterial coating (implied by the work of Ewert and Deming ) or by physical blockage of the brine veins by pEPS [Krembs et al., 2011], affecting the passive transport of bacteria and salts in different ways. Ratios of bacteria to salts in frost flowers (data from Bowman and Deming ) are lower than in surface ice, in agreement with selective retention occurring before brine leaves the ice. Retention in the basal layer of snow (by ice affinity of cells to snow crystals) could also contribute to higher bacterial abundance in the lowermost snow layer.
4.4 Loss of Marine Bacteria and Extracellular Polysaccharide Substances in the Snow
 Selective loss of bacteria in the surface environment would also account for the lower bacterial numbers observed in the surface environment. Bacterial loss over the course of winter (up to 49%) has been observed in upper sea-ice horizons (upper 25 cm [Collins et al., 2008]), where it was attributed to osmotic lysis, viral lysis and/or cell impingement by ice or salt crystals. Viral lysis in particular may be triggered by fluctuations in salinity and temperature [Ghosh et al., 2009; Shkilnyj and Koudelka, 2007] as occur in sea ice.
 Fluctuations in temperature, brine volume, and brine salinity were progressively less strong the deeper into the ice. This dampening effect of sea ice [Collins et al., 2008; Petrich and Eicken, 2010] may explain higher bacterial abundance in the 10 cm ice horizon (BW'10) compared to surface ice, as well as the reduction in bacteria-to-salt ratios in samples near or above the sea ice surface (Figure 5). The strongest fluctuations were registered at the air interface, with daily changes in brine salinity up to 180 (Figure 2). Actual fluctuations in brine salinity may depend on available water content of the environment and time of exposure; for instance, smaller fluctuations should occur in the more ephemeral frost flowers, which have a freshwater (pure ice) content lower than saline snow.
 Brine skim and frost flowers are young environments that, even if directly exposed to the air, still contain high numbers of bacteria and EPS per unit area (0.3× 104–4.8× 104 cells cm− 2 and 0.2–36.2 µg glu-eq cm− 2 respectively; data from Bowman and Deming ). Within hours to days, however, brine from both environments will be incorporated into deposited snow. The snow sampled in BW'10 had incorporated high amounts of salt per unit area (see above), but the numbers of bacteria (2.1× 103–9.6× 103 cells cm− 2) and pEPS (0.03–0.3 µg glu-eq cm− 2) per unit area were low. This apparent loss of bacteria in the snow may be due to a longer time of exposure, weeks to months, to more extreme environmental conditions of temperature and salinity (despite the higher relative proportion of protective EPS per cell in snow compared to sea ice). Because we did not observe a higher proportion of “dead” cells in saline snow compared to surface ice, the mechanism accounting for the lower cell numbers in the snow must involve complete cell lysis.
 Another microbial stressor at the ice surface is UV-B irradiation. pEPS absorption spectra peaked in the UV range (data not shown)—a characteristic of polysaccharides susceptible to UV-B photolysis [Ortega-Retuerta et al., 2009]. Penetration of UV light into snow, however, is limited by absorption and scattering, with attenuation of 1 order of magnitude in 4 cm deep snow patches [King and Simpson, 2001; Cockell and Córdoba-Jabonero, 2004]. Also, UV-B driven photolysis may be irrelevant during periods of reduced daylight, but could occur in the uppermost layers of snow and thinner (< 4 cm) snow packs during daylight periods.
 Annual differences in the snow cover may have contributed to the observed differences in bacterial and EPS content. The deeper BW'10 snow cover provided milder conditions in the saline snow layer, with warmer temperatures (by 18.2°C) and fresher salinities (by 72) than exposed brines. BW'11 had a thinner snow pack, with average snow depth of 4.7 cm (MBS, 40 days leading to sampling), 28 days showing minimum snow depth < 4 cm and only one day having maximum snow depth > 20 cm. The thinner snow cover and lower air temperatures in BW'11 led to more extreme conditions in the saline snow layer, consistent with a greater loss of bacteria and pEPS in the snow. Furthermore, increasing daylight during BW'11 (up to 10 h of light per day) augmented the potential for UV-B photolysis.
 Testable hypotheses emerge from this work. The insulating role of the snow may translate to milder daily and seasonal fluctuations for thick snow packs, making them more habitable than thinner snow. Changing snow-depth differences could thus generate heterogeneity in the bacterial community of the ice-snow interface. Likewise, because snow accumulated over collapsed frost flowers shows higher salinity than snow accumulated directly over the ice [Massom et al., 2001], it may also contain higher bacterial and pEPS abundances.
4.5 Potential Aerial Transport of Saline Snow
 Wind speeds capable of initiating blowing snow events (> 7 m s− 1 [Savelyev et al., 2006]) were common during the weeks leading to sampling, and more frequent prior to BW'11 (68 days) than BW'10 (40 days). Bare ice was not evident, but MBS snow was thinner than 3 cm on 10 out of 40 days leading to sampling, indicating blowing of saline snow in BW'11. Blowing snow could redistribute marine components and contribute to higher background levels of bacteria and EPS, masking existing gradients. Wind dispersal of the bottom 2 cm of snow, however, would have been limited by observed hard icy layers. Aerosol production and aerial dispersal from frost flowers remain open questions [Obbard et al., 2009; Roscoe et al., 2011], but frost flowers may play an indirect role by contributing brine, bacteria, and EPS to the snow that collapses them.
 Bacteria and EPS of marine origin are present in saline snow overlying Arctic first-year sea ice where they are susceptible to dispersal by snow blowing events. The observed bacteria and EPS did not occur in the same proportion as the salts in this newly identified and potentially vast habitat, suggesting either selective retention in the source ice or loss after transport into snow. Losses may be due to the more extreme conditions encountered near the atmospheric interface, documented as wide daily fluctuations in temperature, brine salinity, and brine volume fraction.
 This research was supported by NSF-IGERT support to M.E. and J.C.L. and by NSF-OPP award 0908724 to J.W.D. We thank BASC, BARC, and UMIAQ personnel for logistical support in the field, Jeff Bowman, Hajo Eicken, Steve Warren, Bonnie Light, and Regina Carns for helpful discussion, and anonymous reviewers for significant input and insight.