Seawater samples were taken at a 2 m depth underneath the MOSAiC ice floe using a CTD rosette in the main oceanography tent (Rabe et al., 2022). Bubble-free seawater was collected in 500-mL glass vials that were sealed with rubber stoppers and crimped with aluminum caps. Lead ice coring followed the ice coring procedures of MOSAiC (Angelopoulos et al., 2022). In short, at least three lead ice cores 20–50 cm apart were taken using a Kovacs Mark II 9 cm ice core drill. At each sampling site and day, the first ice core was used for in situ temperature measurements. Thereafter, two cores for the analyses of methane concentration, the stable carbon isotopic signature of methane (δ¹³C–CH₄) and salinity, were taken. Once transferred onboard, the cores were immediately cut into 10-cm slices. Each slice was melted in a gas-tight Tedlar bag after evacuating the air therein in darkness at 4°C. Pieces of thin new ice and frost flowers were cut using a knife and collected with a spittle, respectively, and directly saved in Tedlar bags. Once melted, the meltwater was collected in 500-mL glass vials sealed with rubber stoppers and crimped with aluminum caps. In the following, seawater and melted lead ice samples were processed using the same procedure. In the sample bottles, a headspace was created by injecting 20 mL of hydrocarbon-free synthetic air while, at the same time, water was pushed out via a second needle to balance the pressure. After the gas and water were allowed to come to equilibrium by shaking, the headspace was subsampled with a gas-tight syringe (SGE, Victoria, Australia), and the sample was immediately injected into the small sample isotope module (SSIM, see Section 2.2).

Methane concentrations and stable carbon isotope ratios (δ¹³C–CH₄) of seawater and bulk ice samples were determined using a Picarro G2132-i cavity ring-down spectrometer coupled to a small sample isotope module (SSIM) (Picarro, Santa Clara, California, USA). The SSIM is a sample inlet system used to analyze discrete samples. The SSIM coordinator software was run in high-precision mode, with a syringe injection setting, and on dilution with hydrocarbon-free air to account for our analysis procedures. The Picarro G2132-i runs with a flow rate of approximately 25 mL/min. For calibration, we used standard gas mixtures with concentrations of 2 ppm, 5 ppm, and 10 ppm (Linde Company) and standard gas mixtures with isotopic ratios of −25‰, −45‰, and −69‰ vs. VPDB (Airgas Company). Considering the complete sampling procedure, the overall total uncertainty was ±5% estimated in duplicate. In addition to this uncertainty, we cannot exclude a concentration-dependent effect on our isotopic ratio measurements (Rella et al., 2015; Uhlig and Loose, 2017; Pohlmann et al., 2021) because many of our measurements were made on samples with mixing ratios <1.2 ppm CH₄ (Supplementary Figure S1). Future studies using a Picarro gas analyzer should use isotopic gas standards spanning the full range of mixing ratios observed in samples in order to isolate the concentration-dependent effect and improve the absolute accuracy of the isotopic ratio measurements.

Surface seawater temperature and salinity at depths of 2 m were continuously measured along the drift track by two thermosalinographs (SBE21, Sea-Bird GmbH) installed in the hull of RV Polarstern (Haas et al., 2021). The temperature and bulk salinity of the lead ice cores were measured according to procedures described in Angelopoulos et al., 2022. In short, ice temperatures were measured in 4.5 cm deep, narrow holes drilled into the core at intervals of 5 cm from the bottom to the top of the ice immediately after the cores were retrieved. Bulk salinity was analyzed on melted subsamples using a WTW 193 Cond 3151 salinometer equipped with a TetraCon 325 four-electrode conductivity cell.

We analyzed the concentration and stable carbon isotope ratio of methane in lead ice and used the methane concentration and δ¹³C–CH₄ ratio in surface seawater immediately measured before lead opening as our values at time zero (t = 0). Surface seawater was in equilibrium or slightly supersaturated (less than 10%) relative to the atmosphere, while the δ¹³C–CH₄ ratios of dissolved methane in seawater deviated by 8‰–10‰ ¹³C-enrichment from the Arctic atmospheric equilibrium δ¹³C–CH₄ signature (≈−47.5‰; White et al., 2018).

Figure 3 shows lead ice measurements during the 2-month sampling period. We divided this period into five different phases to highlight different predominant processes of lead ice evolution in the following subsections.

In addition to ice collected from the same lead during the 2-month observation period, we sampled thin new ice less than a week old and frost flowers at several other places between January and the beginning of May to evaluate the impact of the first freeze phase on methane uptake and discharge, respectively (Figures 1, 2C).

Lead opening in winter exposes the relatively warm underlying seawater to the cold air, resulting in rapid ice formation and growth by heat loss to the atmosphere. Strong brine entrapment affected by fast freeze causes high bulk salinities within the newly formed lead ice (Perovich and Gow, 1996). Indeed, we detected bulk surface salinities higher than those in FYI grown at the same time on this floe (Figure 6; Angelopoulos et al., 2022). A fast freeze favors upward brine expulsion to the surface ice layer (Notz and Worster, 2009). Pronounced half-C-shaped profiles of bulk salinity confirmed this upward-directed brine transport (Figures 3, 6). Exposing the highest-saline layer to the cold ice surface provoked counterintuitive physical properties: despite an ongoing strong freeze, the newly formed lead ice remained permeable for longer than 2 weeks (Figures 3, 4). Considering permeable ice as an open system (Golden et al., 2007), newly formed lead ice favors gas exchange coupled with brine migration between both interfaces, that is, to the atmosphere and ocean during that period.

In addition to affecting the physical properties, strong cooling also influenced the chemical properties, increasing the methane solubility at the ice surface and promoting in-gassing from the atmosphere (Figure 2). An eventual strong freeze induced a pronounced shift in the MSC-II in permeable lead ice compared to the MSC in seawater underneath. More precisely, accounting for the cooling of migrating brine to approximately −6°C causes an increase in MSC-II by approximately 20% related to the MSC of seawater (Figure 2). Therefore, seawater saturated with methane should induce an undersaturation related to MSC-II when the uptake of dissolved marine methane in permeable lead ice occurs. However, we observed methane saturation relative to MSC-II in lead ice, while methane was supersaturated relative to MSC in seawater. Consequently, the increase in solubility in permeable lead ice induced a solubility pump transporting methane downward through lead ice. Uptake at the ice surface might occur directly by dissolution of gaseous atmospheric methane or via an intermediate step, that is, indirectly by re-dissolution of methane previously released from high-saline frost flowers.

Cooling to −12°C further increased MSC-II (Figure 2), promoting continued dissolution of gaseous methane at the lead ice–air interface but also causing squeezing and subsequent gravity brine drainage. The “squeezing” caused a loss of methane by brine rejection, shown by the reduced concentration compared to the former phase at the lead ice surface. The interaction of the two processes initiated an incipient top-down transport of dissolved methane through the more than 30-cm-thick permeable lead ice layer (Figure 4). The ongoing uptake at the surface partly counterbalanced the loss by brine drainage and initiated a brief steady state period reflected in linear concentration and solubility gradients (Figures 3, 4). However, this period ceased before the end of Phase II. The convex-shaped concentration gradient pointed to an interrupted downward transport when the bottom layer grew.

Finally, cooling to −16°C resulted in the formation of a subsurface impermeable layer (SSIL), while the surface layer remained permeable due to the slightly higher bulk salinity. The SSIL separated the formerly complete permeable lead ice into an upper and lower permeable layer, interrupted the top-down brine drainage, and stopped the methane pump (Figures 3, 4). In impermeable ice, brine movement stops, but the brine is stored in isolated pockets (Golden et al., 2007; Crabeck et al., 2019). Hence, the formation of the SSIL caused a temporal offset in terms of methane stored in lead ice. While the SSIL preserves “frozen-in relicts” of earlier phases of ice formation, the permeable top and bottom layers mirror the impact of ongoing exchange at the interfaces between the ice surface and atmosphere on the one side and seawater on the other side. Indeed, methane uptake at the permeable top layer proceeded as cooling continued. Because the SSIL from that point on inhibited the downward transport, a methane excess on top of the SSIL developed (Figures 3, 4). By comparison, underneath the SSIL, methane loss by brine discharge continued from the permeable bottom layer to seawater. As uptake at the air–ice interface and loss at the ice–ocean interface took place contemporaneously, the concentration profiles changed from linear to half-C-shaped. This shift in gradients clearly assigned counter-rotating methane pathways when lead ice started to convert from completely permeable to partly impermeable (Figures 3, 4).

After a month, cooling at the surface ceased, disrupting the increase in solubility, and therefore, methane uptake at the air–ice interface stopped. At the same time, the lead ice surface became impermeable, enclosing the methane excess formed in the first month in disconnected brine pockets (Figures 3, 4). By comparison, cooling slowed but continued in subsurface lead ice. Consequently, the thicknesses of both the SSIL and the permeable bottom layer still increased, but brine release from the permeable layer weakened, as shown by the low Rayleigh numbers (Figure 3). The restricted brine release also reduced the loss of methane. Due to the gradient between MSC in seawater and MSC-II in open brine channels, methane replenishment from seawater into the highly permeable bottom layer started to occur at this point (Figures 2, 4). Accordingly, the former half-C-shaped concentration profiles developed toward rather C-shaped concentration profiles during the fifth and last phase, that is, the second month (Figure 4). Remarkably, this shift reflects how the strong freeze during the first month shaped the concentration profiles compared to the second month when mainly ice growth shaped the profiles, that is, either brine/methane release or ice growth/methane uptake dominated.

While surface seawater was saturated to just slightly supersaturated, the δ¹³C–CH₄ ratios in seawater clearly deviated from the atmospheric equilibrium. Based on the atmospheric background of −47.5‰ (White et al., 2018) and taking into account up to 1‰ ¹³C-enrichment in seawater by kinetic isotopic fractionation (KIF) at the water–air interface (Knox et al., 1992; Happell et al., 1995), the δ¹³C–CH₄ ratio in surface seawater equilibrated with the atmosphere results in −46.5‰ VPDB. However, the δ¹³C–CH₄ ratio in seawater deviated by 8‰–10‰ ¹³C-enrichment from the atmospheric equilibrium value during the winter study. Remarkably, this pronounced shift even exceeded up to 5‰ of the ¹³C-enrichment detected during the melt season in this area (Damm et al., 2015; Verdugo et al., 2021). Seasonally altering δ¹³C–CH₄ ratios in seawater points to impacts of the sea ice life cycle on the marine methane pool by KIF effects. KIF effects are caused by the higher binding energy of the heavier isotopic molecule (¹³C), resulting in a lower mobility than the lighter isotopic molecule (¹²C) (Mook, 1994). Consequently, seasonal variations in δ¹³C–CH₄ ratios reveal a shift in the relative abundance of the ¹²C and ¹³C isotopes obviously related to melting and freezing processes.

Indeed, a ¹³C-depletion by at least 2‰ in newly formed lead ice related to the methane pool in seawater points to a higher abundance of the lighter (¹²C) isotope when marine methane became enclosed in brine (Figure 7). The ¹³C-depletion evidently strengthened when brine discharge occurred, as shown by the shift in δ¹³C–CH₄ ratios between newly formed and older lead ice (Figure 7). As occurring in the opposite direction than uptake, gravity favored the higher abundance of the heavier (¹³C) isotope on its way back to seawater, reinforcing on one side ¹³C-depletion in brine-methane and on the other side ¹³C-enrichment in marine methane.

At the ice–air interface, methane dissolution induced by the solubility pump also contributed to δ¹³C–CH₄ ratio shifts in lead ice. Caused by differences in the molecular transfer rates of the lighter and heavier isotope (Mook, 1994), dissolution initiates a shift toward a ¹³C-enrichment in the relative abundance of the ¹²C and ¹³C isotopes, ultimately weakening the ¹³C-depletion in lead ice formerly induced by the KIF effect during ice formation. Consequently, because brine drainage and dissolution take place simultaneously and both processes partly counterbalance each other, δ¹³C–CH₄ ratios just vary less in aging lead ice.

Ice growth under winter conditions results in highly saline surface skim, often additionally covered by frost flowers. Frost flowers frequently develop at the surface of new sea ice, especially when the temperature gradient between surface ice and the overlaying atmosphere is large, and wind speeds are low. Initially, frost flowers are frozen freshwater, as they grow from the vapor phase, but the ice crystals wick up the brine from the surface of the sea ice, producing salinities of up to 180 PSU (Perovich and Richter-Menge, 1994; Domine et al., 2005). Their crystal-shaped configuration tends to provide a high total surface area of up to 1000 m² of ice surface (Domine et al., 2002), suggesting that frost flowers are sites for enhanced exchanges and chemical reactions (Rankin et al., 2002). These circumstances might also induce fast but pronounced shifts in the methane solubility capacity at the ice surface. Consequently, lead ice covered by frost flowers is a favored environment for phase changes between gaseous and dissolved methane. However, in opposition to the one-way direction of phase changes due to temperature decrease, the high variability in salinity at the ice surface during the life cycle of frost flowers enables phase changes in both directions. While low salinities in freshly formed frost flowers favor the dissolution of gaseous methane, increasing salinity induces a backward phase change, that is, a shift from dissolved to gaseous methane despite strong freezing. Large variations in δ¹³C–CH₄ ratios in frost flowers compared to those in lead ice potentially might refer to pronounced KIF effects by phase changes (Figure 7). However, we cannot exclude that the low concentrations affected the analyses of the δ¹³C–CH₄ ratios, finally inducing the large variations in δ¹³C–CH₄ ratios. To prove the proposed processes, in situ studies on frost flowers are required in the future.

To estimate the Arctic-wide relevance of methane uptake at the ice surface on newly formed lead ice, we upscale our results with satellite observations of leads during MOSAiC. We use lead observations based on divergence previously presented in von Albedyll et al. (2023). With this dataset, we analyze the aerial coverage and lifetime of up to 10-day-old leads in a circle with a 10 km radius around MOSAiC and compare the results to circles with radii of 50 km, 100 km, and 150 km.

The mean aerial fraction (10-km scale) of up to 10-day-old leads was 3.2%, with a maximum of 10% (Figure 8). On 50% of the days, at least 2.5% of the area was covered by leads. Consequently, thin lead ice was an abundant feature of the sea ice cover. As shown above, the physical and chemical properties of up to 10–14-day-old lead ice mostly favored methane uptake at the lead ice surface. Von Albedyll et al. (2023) also showed that most of the leads during MOSAiC were short-lived (Figure 8). The median lifetime of leads was only 3 days, while only 2% of all leads did not close within 11 days. Thus, most of the leads during MOSAiC were an open system between the atmosphere and the ocean until an impermeable subsurface layer formed after this period.

The mean lead fractions from the MOSAiC expeditions on the 10 km scale (3.2%) are representative of the larger (50 km, 100 km, and 150 km) surroundings with a lead fraction of 3.1%–3.2%. However, some spatial lead dynamics differences emerge when the time series data are compared on a daily basis, resulting in a Pearson correlation coefficient of R = 0.6–0.7 (not shown).

Analyzing datasets from different Arctic regions, other studies found similar average lead fractions and maximum lifetimes of leads (Wadhams, 2000; Reiser et al., 2020; Hutter et al., 2023). Thus, we consider our estimates of lead aerial fraction and lifetime representative for a larger region and time beyond the drift of MOSAiC. Consequently, uptake at the ice surface and pumping through thin lead ice is a significant yet unconsidered methane pathway in the Arctic. The relevance of this process may even increase when the Arctic fully transitions into a seasonally ice-covered ocean where leads may be more abundant than in the central ice pack (Kwok et al., 2006).