Gas dynamics within landfast sea ice of an Arctic fjord (NE Greenland) during the spring–summer transition

The Arctic Ocean is an important sink for atmospheric carbon dioxide (CO₂), with current estimates of net uptake of CO₂ from −66 to −199 Tg C yr⁻¹ (Bates and Mathis, 2009; Takahashi et al., 2009; Yasunaka et al., 2016). The presence of sea ice at high latitudes has been suggested to affect the CO₂ cycle strongly, through the seasonal cycles of sea ice growth and melt (Rysgaard et al., 2011), although the net role of sea ice in terms of air–sea exchanges of CO₂ remains uncertain (Lannuzel et al., 2020). Recent studies have highlighted the complex dynamics of sea ice and its role in CO₂ exchange with the atmosphere, from a strong release of CO₂ towards the atmosphere during initial sea ice growth (9.9 mmol m⁻² d⁻¹; Geilfus et al., 2013) to a net sink for atmospheric CO₂ as sea ice melts (−9.8 mmol m⁻² d⁻¹; Van der Linden et al., 2020). While most of the fluxes have been reported in spring and summer (Semiletov et al., 2004; Geilfus et al., 2012; Nomura et al., 2013; Delille et al., 2014; Geilfus et al., 2014; Geilfus et al., 2015), the impacts of melt ponds as contributors to the sea ice CO₂ exchange with the atmosphere and their role in gas dynamics within sea ice remain poorly known, despite the fact that melt ponds are a major and increasing feature of Arctic sea ice (Rösel and Kaleschke, 2012).

Melt ponds have been studied intensively due to their strong impact on the summer energy and mass budgets over sea ice through the sea ice–albedo feedback mechanism (Perovich, 1996; Fetterer and Untersteiner, 1998; Taylor and Feltham, 2004; Perovich et al., 2011; Polashenshi et al., 2012). Melt ponds result from the accumulation of meltwater (initially from snow, then sea ice) on the surface of relatively impermeable ice. The volume of meltwater is determined by the balance of inflow and outflow, accumulated in the lowest point of local topography. Inflow is based on melting rate, possible precipitation, and the size of the catchment basin of the pond. Outflow depends on the hydraulic pressure gradient in the brine system (Darcy’s law) and the presence of outflow pathways such as cracks, leads, or other openings, including seal breathing holes (Polashenski et al., 2012). As sea ice warms, its brine volume increases and meltwater ponds, located above the freeboard, may drain through the ice (i.e., brine flushing; Fetterer and Untersteiner, 1998) to the underlying seawater, freshening the upper layer of the water column. This process can lead to the formation of under-ice melt ponds, where freshwater lenses are trapped under thinner ice areas or in depressions at the bottom of thicker ice (Weeks, 2010). Therefore, melt pond formation not only impacts the physical and optical properties of sea ice (Perovich, 1996; Perovich et al., 2002), but also influences the temporal evolution of sea ice salinity, the carbon and nutrient distributions within the ice matrix, and the salt and heat budget of the ocean mixed layer below (Eicken et al., 2002; Perovich et al., 2002).

Several studies have reported sea ice uptake of atmospheric CO₂ during active surface melt (Semiletov et al., 2004; Geilfus et al., 2012; Nomura et al., 2013; Geilfus et al., 2015). Increasing sea ice temperature, subsequent brine freshening, and dissolution of ikaite (metastable calcium carbonate precipitate, CaCO₃•6H₂O; Dieckmann et al., 2008) result in a strong decrease of total alkalinity (TA), dissolved inorganic carbon (DIC), and brine partial pressure of CO₂ (pCO₂), causing sea ice to act as a sink for atmospheric CO₂ (Semiletov et al., 2004; Geilfus et al., 2012; Nomura et al., 2013; Delille et al., 2014; Geilfus et al., 2014; Geilfus et al., 2015). This drawdown of pCO₂ is also enhanced by the development of an autotrophic (photosynthesizing) community at the bottom of the sea ice, where more CO₂ is fixed into organic matter than released via respiration (Van der Linden et al., 2020).

However, during sea ice melt, gas inclusions can be created within the ice structure, as water vapor can form inside the brine pockets above the freeboard line or as drained brine pockets and channels are being replaced by atmospheric gases (Tsurikov, 1979). Changes in sea ice physical properties and processes of air inclusion within sea ice can potentially affect the sea ice exchange of gases with the atmosphere. Oxygen (O₂), nitrogen (N₂), and argon (Ar) mainly reside within sea ice in the gaseous form (i.e., bubbles) rather than dissolved within the brine (Zhou et al., 2013; Crabeck et al., 2014; Moreau et al., 2014; Crabeck et al., 2019). The formation of gas bubbles from entrapped brine has been suggested mainly due to the decreasing gas solubility as sea ice temperature and salinity change during sea ice growth (Zhou et al., 2013; Crabeck et al., 2014). Alternatively, gas diffusion has been suggested as a key process controlling gas fluxes across sea ice, even at low brine volume fraction (Gosink et al., 1976; Loose et al., 2009).

Gases can also be produced and/or consumed by organisms living in or under the ice and can be taken up or released during the natural cycle of sea ice formation and melt. The cycles of gases are therefore closely linked to biological processes. Ice algae, phytoplankton and bacterial communities will adapt to changes in sea ice (Lannuzel et al., 2020) with direct consequences for the uptake and release of climate-active gases. Heterotrophic activity and algal production affect the O₂ and CO₂ dynamics within sea ice by releasing or consuming O₂ and CO₂ as well as impacting the concentrations of DIC and TA, both key parameters driving the exchange of CO₂ across the ocean-sea ice-atmosphere interface. During oxic conditions (calcification apart), the contribution of microbial processes to DIC changes is best described by the net community production (NCP). NCP is defined as the difference between gross primary production and ecosystem respiration. The succession of distinct heterotrophic and autotrophic phases has recently been reported in both Arctic and Antarctic sea ice, based on the traditional methods of melted ice core incubations (Søgaard et al., 2010; Campbell et al., 2017) or using the ratio O₂:Ar (Zhou et al., 2014; Van der Linden et al., 2020). Depending on the autotrophic or heterotrophic state of the sea ice, one could expect opposite effects on CO₂ uptake or release between the atmosphere and the sea ice cover.

A common way to estimate the relative contribution of meteoric water (including snow melt) in the marine environment (including sea ice) is to use the stable isotope composition of oxygen in water, δ¹⁸O-H₂O (herein referred to as δ¹⁸O; Souchez et al., 1988; Macdonald et al., 1995; Eicken, 1998). The δ¹⁸O is expressed in ‰, relative to the Vienna Mean Standard Ocean Water (VSMOW) as:

where R = ¹⁸O/¹⁶O. Both equilibrium thermodynamic and kinetic effects favor the accumulation of lighter isotopes of water in the vapor phase and heavier isotopes in the liquid phase (Craig and Gordon, 1965). Repeated cycles of evaporation and precipitation, as water molecules travel towards the poles, cause polar latitude meteoric water to be depleted in ¹⁸O relative to ¹⁶O, resulting in more negative δ¹⁸O values in meteoric water compared to seawater. As such, reported values of δ¹⁸O for both glacial ice and snow cover in polar latitudes have ranged from −20‰ to −30‰, with some values as low as −40‰ (Dansgaard and Tauber, 1969; Johnsen et al., 2001). In contrast, sea ice formation slightly favours heavier isotope incorporation into the solid phase (Souchez et al., 1988; Eicken, 1998). Reported δ¹⁸O values typically range from −8‰ to 2‰ (Carnat et al., 2013; Alkire et al., 2015), with values as low as −22‰ (Geilfus et al., 2021), based on the isotopic signature of the surface water signature from which sea ice grew (Macdonald et al., 1995; Geilfus et al., 2021). Therefore, glacial ice and snow will have distinguishable δ¹⁸O values from that of sea ice and of seawater (δ¹⁸O close to 0‰).

This study presents the changes of physical properties occurring within landfast sea ice during the spring–summer transition in Young Sound, Northeast Greenland. We discuss the relative contribution of physical and biological processes associated with melt pond formation and meltwater drainage, and how these contributions affect the gas composition and dynamics within sea ice and the exchange of CO₂ with the atmosphere.

This study was based at the Daneborg field station (74°18 N, 20°13 W), a marine research facility located 25 km southeast of the Zackenberg research station, in Northeast Greenland (Figure 1). This fjord system is characterized by an outer part, Young Sound, and an inner part, Tyrolerfjord. The fjord is approximately 90 km long from its mouth to the inner part of Tyrolerfjord, 2 to 7 km wide, and covers about 390 km² (Rysgaard et al., 2003). The fjord system is usually covered by landfast sea ice for approximately 9 months of the year, extending up to 10–30 km from the fjord mouth before transitioning to pack ice where a wind-driven shelf polynya has been observed, though it varies in occurrence, location, and size (Pederson et al., 2010). During the year of our observations, sea ice started to form in mid-October 2013 (Kirillov et al., 2015). Several northerly storm events maintained a coastal polynya until early March 2014 after which landfast sea ice covered the polynya area and remained stable until breakup (July 15, 2014; Dmitrenko et al., 2015; Kirillov et al., 2015). During our survey, the middle layer of the water column (40–140 m depth) was filled with cool, saline, and oxygen-enriched polynya water (Dmintrenko et al., 2015). During summer, the fjord is under strong influence of the surrounding glaciers (i.e., Greenland Ice Sheet and local glaciers) as summer runoff from the surrounding catchment areas decreases surface seawater salinity from approximately 33 to 5 in the inner part of the Tyrolerfjord (Bendtsen et al., 2014).

Sea ice cores, snow, and seawater samples were collected during the spring–summer transition, from May 24 to June 24, 2014 (Table 1). The sampling site was located across from the research station, a few km within the fjord (74.322°N, 20.269°W; Figure 1). At the site, adjacent 5 m × 5 m plots were chosen for snow, sea ice and under-ice seawater sampling. The sea ice in this defined area was assumed to be homogeneous, with six sampling events over time designated as S1–S6. First, snow cover and slush layer (i.e., a mixture of snow and water), when present, were collected using Ziplock bags, with samples stored in a cooler until transported back to the station. Four sea ice cores were collected using a MARK II coring system (internal diameter of 9 cm, Kovacs Enterprises^®, Roseburg, OR, USA) for: i) in situ temperature; ii) bulk ice salinity, total alkalinity (TA), dissolved inorganic carbon (DIC), and water stable isotopes (δ¹⁸O); iii) chlorophyll-a (Chl-a) and nutrient (NH₄⁺, NO₂⁻, NO₃⁻, and PO₄³⁻) concentrations; and iv) argon (Ar), oxygen (O₂) and nitrogen (N₂) gas concentrations. Sea ice cores were wrapped in polyethylene bags immediately after extraction and stored horizontally, to minimize brine drainage, in a cooler containing −20°C eutectic bags. Snow and sea ice in situ temperature was measured using a calibrated temperature probe (Testo 720^®, Lenzkirch, Germany, ±0.1°C precision). For sea ice, temperature was measured directly after the core extraction by inserting the temperature probe into pre-drilled holes at 5 to 10 cm vertical resolution.

Surface seawater was collected through a core hole using a peristaltic pump (Cole Palmer^®, Masterflex Environmental Sampler) with Tygon^® tubing from which the length corresponds to the observed sea ice thickness (plus a few centimeters). Water column samples were also collected at 1, 2.5, 10, 20, 30, 50, and 100 m using 5 L Niskin bottles. Seawater samples were collected using sealed glass 100 mL BOD bottles. A headspace of 1% of the bottle volume was introduced to accommodate thermal expansion. Vertical profiles of water temperature and salinity were measured with a newly factory-calibrated Sea-Bird SBE 19plus V2 (Sea-Bird Scientific, Bellevue, WA, USA) conductivity-temperature-depth (CTD) probe, accurate to ±0.005°C and ±0.0005 S m^-1 (Dmitrenko et al., 2015).

A few meters away from the sampling site, a soil CO₂ flux system was deployed with four long-term deployment automatic chambers (LI-COR^® 8100-104, Lincoln, NE, USA). Chambers for CO₂ flux measurements were deployed on snow-covered sea ice surfaces. Each chamber closed and measured fluxes for 15 min every hour, while remaining open the rest of the time to avoid surface disturbance as much as possible. Chamber maintenance was performed every day, while chamber repositioning to unaltered surface was performed when snow drift was accumulating and preventing chamber closure or when sea ice was too altered by melting to ensure proper sealing of the collar between the ice and the chamber. The chambers consist of an automated moveable metal dome (internal diameter = 20 cm, height = 9.7 cm), closed at the top by a pressure vent designed to maintain an ambient pressure inside under calm and windy conditions. A rubber seal surrounds the soil collar and ensures an airtight connection at the chamber-ice interface. The metallic soil collar has been designed for sea ice measurements with serrated circumference to allow its penetration into the ice. For measurements over snow, a stainless-steel tube was mounted at the base of the chamber and pushed down to the ice to enclose snow and prevent lateral diffusion of air in the snowpack.

The chambers were temporarily closed above the snow or ice interface and connected using a multiplexer (LI-COR^® LI-8150, Lincoln, NE, USA) to an infrared gas analyzer (LI-COR^® LI-8100A, Lincoln, NE, USA) recording the air pCO₂ during a designated time interval (30 min). The flux was determined by measuring the change of air pCO₂ within the chamber. Flux calculations were performed routinely using the LI-COR^® SoilFluxPro software, with volumes adjusted to include collar offsets (i.e., rim height above the surface) and applying water vapor corrections for dilution and band broadening. The fluxes were determined by the slope of the linear or exponential regression of pCO₂ over time, depending on the best fit. All fluxes were visually checked to discard measurements biased by contamination or poor sealing of the chamber. The detection limit was ±0.004 μmol m⁻² s⁻¹. Fluxes below the detection limit were considered as null. The uncertainty of the flux computation due to the standard error of the regression slope was ±8%.

Back at the station, seawater samples were transferred to gas-tight vials (12 mL Exetainer, Labco High Wycombe, UK), poisoned with 12 µL solution of saturated HgCl₂, and stored in the dark at room temperature until analysis for TA and DIC. Water aliquots of 2 mL were also collected in fully filled gas-tight glass vials for δ¹⁸O measurements, with salinity measured on the remaining sample volume. A sea ice core dedicated to bulk ice salinity and the carbonate system was cut into 5–10 cm sections. Each section was immediately placed into a gas-tight laminated (Nylon, ethylene vinyl alcohol and polyethylene) plastic bag (Hansen et al., 2000) fitted with a 20 cm gas-tight Tygon tube and valve. The plastic bag was sealed, and excess air was gently removed through the valve using a vacuum pump. Snow and slush samples were processed following the same procedure. Bagged sea ice, slush, and snow samples were then melted at room temperature. As soon as samples were completely melted, meltwater was transferred to gas-tight vials (12 mL Exetainer, Labco High Wycombe, UK), poisoned with 12 µL solution of saturated HgCl₂, and stored in the dark at room temperature until analysis for TA and DIC. Water aliquots of 2 mL were also collected in fully-filled gas-tight glass vials for δ¹⁸O measurements, with salinity measured on the remaining volume of meltwater.

For Chl-a and nutrient (NO₃⁻, NO₂⁻, PO₄³⁻, and NH₄⁺) analyses, the dedicated ice core was cut into 10 cm sections and melted in the dark, at room temperature. Snow and slush samples were melted following the same procedure, without adding pre-filtered seawater, as suggested by Rintala et al. (2014). Soon after complete melting, samples for Chl-a determination were filtered using Whatman GF/C glass-fibre filters (which may underestimate Chl-a content compared to GF/F filters of smaller effective pore size). Filters were stored frozen in aluminium foil until measurement at the home laboratory; 50 mL of the filtrate was kept frozen for nutrient analyses. Filters were extracted in 10 mL acetone at 4°C for 16 h. Chl-a concentrations were determined using a Turner Design TD700 fluorometer. Nitrate plus nitrite ([NO₃⁻] + [NO₂⁻], later noted as NO_x⁻) was measured using the sulfanilamide colorimetric method; ammonium ([NH₄⁺]), with the dichloroisocyanuratesalicylate-nitroprussiate colorimetric method, and phosphate ([PO₄⁻]), with the ammonium molybdate-potassium antimonyl tartrate method (Westwood, 1981; American Public Health Association, 1998). Analyses were made colorimetrically using a 5-cm optical path, with a Genesys 10 vis spectrophotometer (Thermo Fisher Scientific, Waltham, Massachusetts, USA). The detection limits of the methods were 0.15, 0.03, 0.3, and 0.06 μmol L⁻¹ for NO₃⁻, NO₂⁻, NH₄⁺, and PO₄⁻ respectively.

Salinity was calculated based on conductivity (Grasshoff, 1983) measured using a meter (Orion 3-star, Thermo Scientific, Walham, MA, USA) coupled with a conductivity cell (Orion 013610MD, Thermo Scientific) with a precision of ±0.1. Brine volume was estimated from measurements of bulk ice salinity, temperature, and density according to Cox and Weeks (1983) for temperatures below −2°C and according to Leppäranta and Manninen (1988) for ice temperatures between −2°C and 0°C. δ¹⁸O was measured on an isotope analyzer (Picarro L2130-i, Santa Clara, CA, USA) with a precision of 0.025‰, calculated as the standard deviation from 9 repeated measurements of standard materials.

TA was determined by Gran titration (Gran, 1952) using a TIM 840 titration system (Radiometer Analytical, ATS Scientific, Burlington Ontario, Canada), consisting of a Ross sure-flow combination pH glass electrode (Orion 8172BNWP, Thermo Scientific) and a temperature probe (Radiometer Analytical, Lyon, France). A 12 mL sample was titrated with a standard 0.05 M HCl solution (Alfa Aesar, Haverhill, MA, USA). DIC was measured on a DIC analyzer (Apollo SciTech, Newark, DE, USA) by acidification of a 0.75 mL subsample with 1 mL 10% H₃PO₄ (Sigma-Aldrich, Saint-Louis, MO, USA), and quantification of the released CO₂ with a nondispersive infrared CO₂ analyzer (LI-COR, LI-7000, Lincoln, NE, USA). Results were then converted from µmol L⁻¹ to µmol kg⁻¹ based on sample density, which was estimated from salinity and temperature at the time of the analysis. Accuracies of ±3 and ±2 µmol kg⁻¹ were determined for TA and DIC, respectively, from routine analysis of certified reference materials (A.G. Dickson, Scripps Institution of Oceanography, San Diego, CA, USA). From TA and DIC measurements, pCO₂ is calculated by the CO2sys program (Pierrot et al., 2006) using the dissociation constants from Mehrbach et al. (1973) refitted by Dickson and Millero (1987), assuming these constants are valid for the range of temperatures and salinities encountered within sea ice (Papadimitriou et al., 2004). Those constants were selected as they have shown good agreement between measured and calculated values in Arctic waters (Chen et al., 2015; Wooseley et al., 2017).

Analysis for O₂, N₂, and Ar concentrations in sea ice (denoted [O₂], [N₂], and [Ar]) were performed by gas chromatography. The gas phase was extracted from the ice using the dry-crushing technique, initially developed for gas measurements in continental ice (Raynaud et al., 1982). The sea ice core was cut into 5 cm sections, and 60 g of each section was placed into a vessel together with stainless steel beads. The sample was then crushed in the vessel at a temperature of −25°C and under a vacuum of 10⁻³ Torr, allowing the extraction of both gas bubbles and dissolved gas within sea ice brines (Stefels et al., 2012). The vessel, kept at −25°C in a cold ethanol bath, was then connected to the gas chromatograph equipped with a thermal conductivity detector for concentration analyses. AlphagazTM2 He (Air liquid^®—P0252) was used as the carrier gas and a 22 mL packed column (Mole Sieve 5A 80/100; 5 m × 1/8”). Precision of analysis was 0.7% for O₂ and N₂, and 2.2% for Ar (Zhou et al., 2014).

The amount of gas collected (i.e., [O₂], [N₂], and [Ar], in µmol L_ice⁻¹) includes both the gas bubbles from the ice and the dissolved phase within sea ice brine. Therefore, measured gas concentrations are compared to their respective maximum concentrations in bulk ice when sea ice brines are at atmospheric saturation (denoted [O₂]_sat, [N₂]_sat, and [Ar]_sat). The latter represents the maximum concentration of O₂, N₂, or Ar in the dissolved phase, if no supersaturation exists in the brine (Zhou et al., 2013; Crabeck et al., 2014), and is calculated based on respective temperature and salinity-dependent solubility coefficients (Garcia and Gorden, 1992; Hamme and Emerson, 2004), using brine temperature and salinity, brine volume fraction, and sea ice density to express the value in micromoles per liter of bulk ice (Crabeck et al., 2014). These estimates have been suggested to remain valid for the range of temperature and salinity observed in sea ice (Zhou et al., 2013).

According to Crabeck et al. (2014), the difference between observed gas concentration measured in bulk ice (i.e., [O₂], [N₂], and [Ar], denoted [X]) and their respective maximum concentrations when brines are at saturation (i.e., [O₂]_sat, [N₂]_sat, and [Ar]_sat) provides an estimate of gas concentration in the ice that resides in the gas phase (i.e., the gas content of bubbles, denoted [O₂]_bub, [N₂]_bub, and [Ar]_bub), assuming no supersaturation within sea ice brines and that the sum of gas partial pressure is above 1 atm. If the latter condition is correct from a thermodynamical standpoint, then bubble formation could not be entirely related to the thermodynamical conditions, as interfacial bubble nucleation processes and supersaturation of gases in the liquid phase have been reported previously (Jones et al.,1999; Tison et al., 2016) and pressures higher than atmospheric have also been suggested to exist within sea ice brines (Crabeck et al., 2019). Therefore,

with the percent gas content in the bubbles (f_bub) estimated as

and the supersaturation factor (Sat_f, in %) estimated by

where X corresponds to either O₂, N₂, or Ar.

To estimate the net community production (NCP) from the [O₂]:[Ar] ratio, we used the method developed by Zhou et al. (2014) and Tison et al. (2017), modified by Van der Linden et al. (2020). We first calculated the [O₂]:[Ar] deviation from saturation as:

Equilibrium concentrations of O₂ and Ar can differ from their solubility due to physical processes. To account for potential biases using O₂ and Ar solubilities for [O₂] and [Ar], we estimated the impact of physical processes on O₂ concentration (i.e., bubble formation and migration, brine movements, [O₂]_phys), using [Ar] as a witness of physical processes only:

From there, the O₂ concentration associated with in situ biological activity is obtained by:

[O₂]_bio is expressed in µmol L_ice⁻¹.

NCP is then determined from the change of oxygen concentration associated with in situ biological activity ([O₂]_bio) as:

where Δt represents the elapsed time between two sampling events (see Table 1), and t_s and t_s+1 correspond to the sampling event s and the next sampling event (s+1), respectively.

Thin sections are used to show ice crystal texture and bubble content. Thin section analysis provides information regarding the conditions during sea ice formation (Eicken, 2003). Within a cold room set at −25°C, vertical ice core sections were first thinned to about 1.5 cm thickness and stuck onto a glass plate by freezing a small amount of distilled water along its perimeter. Ice sections were then thinned to a few millimeters using a microtome (Leica SM2400, Concord, Ontario, Canada). Thin sections were detached from the glass plate by controlled warming from below, turned over and reattached through melting and refreezing of the thin melted ice interface, a procedure that has been shown not to alter the original ice crystallographic properties (Omstedt, 1985). Thin sections were then thinned further to about 600–700 µm using the microtome and examined on a light table equipped with cross-polarized sheets. The textural types were identified visually based on the size, shape, and orientation of the ice crystals and compared to descriptions found in the literature (e.g., Eicken and Lange, 1991; Eicken, 2003).

Over the survey, the snow thickness decreased from 60 cm (S1) to 14 cm (S6; Table 1 and Figure 2). During the first sampling events, a strong temperature gradient was observed within the snow cover, with temperature increasing from −4.5°C at the snow-ice interface to −0.7°C at the interface with the atmosphere. As the atmospheric temperature increased (presented later), the snow cover thinned, and the temperature gradient progressively disappeared (Figure 2a). Maximum salinity, observed at the snow-ice interface, decreased from 14.5 (S2) to 0.4 (S6; Figure 2b). The presence of a slush layer (dotted line in Figure 2) was first observed at S2 with a salinity of 30.6 and temperature of −1.9°C. As the snow cover melted, slush salinity decreased to a salinity of 0.1 at S6. Melt ponds were first sampled at S5 (dashed line in Figure 2) and exhibited positive temperature (0.1°C) and low salinity.

Highly depleted δ¹⁸O values (minima = −21.4‰) were observed at the surface layer of the snow cover (Figure 2c). Throughout the snowpack, δ¹⁸O increased with maxima (up to −5.2‰) observed at the snow-ice interface. Slush δ¹⁸O values decreased from −3.0‰ (S2) to −12.8‰ (S6) while in melt ponds, δ¹⁸O averaged −13.6‰ (SD = 0.2, n = 2).

Low TA and DIC, averaging 59 (SD = 11, n = 6) and 45 (SD = 15, n = 6) µmol kg⁻¹, respectively, were observed in the upper 30 cm of the snow cover (Figure 2d and e). Snow TA and DIC increased towards the snow-ice interface, with maxima of 1178 and 1120 µmol kg⁻¹, respectively. Over time, slush TA and DIC decreased from 2141 to 92 µmol kg⁻¹ and from 2028 to 107 µmol kg⁻¹, respectively. TA and DIC concentration in melt ponds closely followed the overall trend observed in the slush, with TA and DIC decreasing from 252 to 76 µmol kg⁻¹ and from 235 to 82 µmol kg⁻¹, respectively. Both slush and melt ponds were undersaturated in CO₂ compared to the atmosphere (401 µatm). Slush pCO₂ ranged from 49 (S5) to 284 (S2 and S3) µatm, while melt ponds pCO₂ ranged from 19 (S5) to 83 (S6) µatm (Figure 2j).

Within the snow cover, Chl-a concentrations ranged from 0.0 to 0.22 µg L⁻¹, while within slush Chl-a ranged from 0.0 to 0.81 µg L⁻¹ (Figure 2f). No measurements were made on melt pond samples. Nutrient concentrations <10 µM were observed within snow, slush, and melt ponds. NO_x⁻ concentration ranged from 0.77 to 9.48 µM within the snow cover (Figure 2g) and increased in the slush, from 0.05 at S2 to 5.33 µM at S6. PO₄⁻ concentration ranged from 0.0 to 0.25 µM (Figure 2h) in the snow cover, while slush concentration decreased from 0.55 at S2 to 0 at S6. NH₄⁺ concentrations ranged from 0.0 to 5.0 µM (Figure 2i) in the snow cover and from 0 to 3.0 in the slush.

Sea ice thickness remained relatively stable over the survey, ranging from 135 to 145 cm thick (Table 1 and Figure 3). Relatively warm temperatures were observed at the ice interface with both the snow cover and the water column. Sea ice temperature increased consistently (Figure 3a) from −1.8°C to −0.1°C in the upper section of the sea ice, and from −1.8°C to −0.2°C in the bottom sea ice section. Overall, bulk ice salinity initially exhibited a typical C-shaped profile (Eicken, 1992), with maximum salinity of 8.2 observed in the upper section of the sea ice (S1; Figure 3b). In addition, a local minimum can be observed at 25–30 cm depth in S1 to S5. Over the survey, bulk ice salinity gradually decreased to form an inversed C-shape, with lower salinity observed in the upper and lower sea ice section at S5 and S6. Based on the measured temperature and salinity, the calculated brine volume fraction was consistently above the permeability threshold (i.e., 5% from Golden et al., 2007, or 7% from Zhou et al., 2013), with higher brine volume content in both the upper (up to 51%) and the lower (up to 49%) sections of the sea ice (Figure 3c). Sea ice δ¹⁸O values ranged from −9.6‰ to 1.7‰ (Figure 3d). The upper 25 cm of the sea ice was strongly depleted in δ¹⁸O (<−5‰), followed by steeply increasing δ¹⁸O values in the 25–55 cm section of the ice column (grey area in Figure 3d), after which a moderated increase in δ¹⁸O values was observed throughout the ice thickness in most cases. Unlike the other sampling events, bottom values of δ¹⁸O at S5 and S6 became more depleted at the sea ice interface with the water column, down to −5.3‰ and −8.7‰, respectively.

Concentrations of Chl-a and nutrients observed within the sea ice (Figure 4) were in the range of concentrations reported previously in the area for the same time period (Rysgaard and Glud, 2004). Chl-a concentrations ranged from 0.0 to 1.39 µg L⁻¹ (Figure 4a). Highest concentrations were observed at S5, with a local maximum at the sea ice interface with the snow cover (0.9 µg L⁻¹) and an increase towards the bottom of the ice column (1.35 µg L⁻¹). NH₄⁺ concentrations averaged 1.13 µM (SD = 0.8, n = 91) with values ranging from 0.02 µM to 2.89 µM (Figure 4b). PO₄³⁻ concentrations averaged 0.59 µM (SD = 1.32, n = 92; Figure 4c). From S1 to S5, the PO₄³⁻ concentration remained stable, with an average value of 0.18 µM (SD = 0.41, n = 81). At S6, the PO₄³⁻ concentration increased, averaging 2.8 µM (SD = 2.2, n = 15) and roughly mimicking S5 profiles of Chl-a, with a maximum value of 7.55 µM in the upper sea ice section. NO_x⁻ concentrations ranged from 0.0 µM to 6.15 µM (Figure 4c). S5 exhibited higher concentrations (averaging 2.34 µM, SD = 1.96, n = 17) compared to the other stations (averaging 0.29 µM, SD = 0.38, n = 75).

TA and DIC concentrations measured in bulk melted sea ice ranged from 64 to 657 µmol kg⁻¹ and from 52 to 622 µmol kg⁻¹, respectively (Figure 5a and d). TA and DIC exhibited similar profiles as for bulk ice salinity, with C-shaped profiles from S1 to S4 and inverse C-shaped profiles for S5 and S6, where lower TA and DIC were observed in both upper and bottom sea ice sections. Both TA and DIC correlated strongly with salinity changes (Figure 5c and f), except for S1, where TA and DIC were higher and lower, respectively, than expected from salinity changes. To discard concentration and dilution effects, sea ice TA and DIC were normalized to the average bulk salinity of 4.5 (denoted as nTA and nDIC, respectively; Figure 5b and e): nTA averaged 390 µmol kg⁻¹ (SD = 103, n = 48) and nDIC averaged 350 µmol kg⁻¹ (SD = 83, n = 48). Based on sea ice temperature and salinity, we estimated the brine volume content (Figure 3c), as well as brine salinity according to the equations from Eicken (2003). Assuming that TA and DIC are contained mainly within the brine, we used the in-situ sea ice temperature, calculated brine salinity, TA, and DIC to calculate the in-situ brine pCO₂ (as further described in Section 3.3). Brine pCO₂ was undersaturated in CO₂ compared to the atmosphere, with values ranging from 363 (S1) to 3 (S6) µatm, except in the uppermost ice at S1, where a pCO₂ of 523 µatm was estimated (Figure 5g). In Figure 5h, the ratio between nTA and nDIC from both sea ice and seawater (also normalized to a salinity of 4.5) is used to investigate the main processes affecting the carbonate system. Dashed lines in Figure 5h represent the theoretical response of TA and DIC to different processes. In the case of CO_2(g) exchanges, only DIC is affected, while precipitation and dissolution of calcium carbonate (CaCO₃) will affect TA and DIC at a ratio of 2:1. Biological activity (photosynthesis/respiration) will affect both TA and DIC at a ratio of −0.16 (Lazar and Loya, 1991). While most of the bulk ice data aligned with the theoretical trend of calcium carbonate dissolution, a few data points from S1 aligned with the theoretical trend of calcium carbonate precipitation.

The evolution of [O₂], [Ar], and [N₂] in bulk sea ice (Figure 6) is compared to their respective maximum concentrations in bulk ice when brines are at atmospheric saturation (i.e., [O₂]_sat, [N₂]_sat, and [Ar]_sat; black dotted line in Figure 6). Overall, [O₂] ranged from 1 to 237 µmol L_ice⁻¹ (averaging 73 µmol L_ice⁻¹, SD = 54, n = 73), [Ar] ranged from 0 to 10 µmol L_ice⁻¹ (averaging 3 µmol L_ice⁻¹, SD = 2, n = 73), and [N₂] ranged from 5 to 796 µmol L_ice⁻¹ (averaging 208 µmol L_ice⁻¹, SD = 192, n = 73). At S2, gas concentrations were generally higher than [X]_sat, with the maximum concentration observed at the ice-snow interface. Gas concentrations in the bottom 20 cm were similar to [X]_sat or slightly lower. At S3, gas concentration at the ice-snow interface decreased, reaching [X]_sat. Maximum concentrations were observed at about 20 cm depth. In the upper 75 cm, gas concentrations were higher than their respective [X]_sat, while below 75 cm they aligned with [X]_sat (except for the bottom value). At S4 and S5, gas concentrations are mostly aligned with [X]_sat, except in the upper 50 cm (S4) and at 25 cm depth (S5) where maximum gas concentrations were observed. For both S4 and S5, minimum gas concentrations were observed at the sea ice surface (with lowest values at S4). At S6, gas concentrations were slightly higher than [X]_sat for most of the ice thickness, with a strong increase at the sea ice surface where maximum concentrations were observed. For all gases, the bottom 5–15 cm at S5 and S6 are strongly undersaturated.

The saturation factor (Sat_f; Figure 7) ranged from 9% to 564% for O₂, from 4% to 513% for Ar, and from 5% to 1003% for N₂. The maximum Sat_f was systematically observed at 25 cm below the ice surface, while the minimum was observed in the permeable bottom 15 cm of the sea ice and at the ice-snow interface (S4 and S5). N₂ was the most supersaturated gas species, averaging 259% (SD = 2.38, n = 80), compared to O₂ averaging 162% (SD = 1.22, n = 80) and Ar averaging 139% (SD = 1.06, n = 80). Early in the survey (S2) most of the gas content was in bubbles (Figure 8), with maximum bubble content estimated in the upper part of the sea ice. From S3 to S5, the ice-snow interface had minimal bubble content, while most of the bubbles were centred at 25 cm depth. Almost no bubbles were present in the bottom 75 cm sea ice section. At S6, bubble content increased throughout the entire ice thickness.

Seawater temperatures were only recorded during the first half of the survey (S1 to S3) and ranged from −1.76°C to −1.62°C, with warmer temperatures in the upper 20 m of the water column (Figure 9a). Seawater salinity ranged from 28.2 to 30.7. From S1 to S4, the water column was well mixed, as no salinity gradient was observed throughout the water column. Starting on June 14 (between S4 and S5; Table 1), a halocline started to develop with salinity lower than 30 observed in the upper 20 m (Figure 9b). Seawater δ¹⁸O values were slightly depleted, ranging from −2.90‰ to −0.73‰ (Figure 9e). The upper layer of the water column exhibited more depleted δ¹⁸O values, averaging −2.1‰ (SD = 0.35, n = 21). Over the survey, surface water δ¹⁸O decreased, associated with the lower sea surface salinity. TA and DIC concentrations ranged from 1965 to 2294 µmol kg⁻¹ and from 1942 to 2248 µmol kg⁻¹, respectively (Figure 9c and d). Starting on June 14, a slight decrease was observed in the upper layer of the water column. Lower concentrations were observed throughout the water column on June 24.

During the survey, the air–ice CO₂ fluxes averaged −0.31 mmol m⁻² d⁻¹ (SD = 0.66, n = 712) and ranged from −4.26 to 1.05 mmol m⁻² d⁻¹ (Figure 10c). Up to June 11 (S4), strong uptake of atmospheric CO₂, averaging −0.59 mmol m⁻² d⁻¹ (SD = 0.83, n = 337), was observed. After June 11, the overall uptake of atmospheric CO₂ decreased, averaging −0.04 mmol m⁻² d⁻¹ (SD = 0.16, n = 352). This difference is also noticeable when fluxes are averaged for each day (Figure 10d). Daily averaged air–ice CO₂ fluxes ranged from −26.6 to 1.2 mmol m⁻². Before June 11 (S4), a large daily uptake of atmospheric CO₂ was estimated. After that, the magnitude of the air–ice CO₂ exchange decreased with a moderated release of CO₂ from the ice to the atmosphere on some occasions (June 19, 21, 24, and 25).

The increase in atmospheric temperature (Figure 10a) led to thinning snow cover (Figure 2), increasing sea ice temperature (Figure 3a), and decreasing bulk ice salinity (Figure 3b). As atmospheric temperature remained above 0°C, melt ponds were observed at the ice surface (S5 and S6; Table 1) associated with surface ice temperature around 0°C and surface ice salinity lower than 2. Strongly depleted δ¹⁸O values observed in the top 25 cm section of the ice column (Figure 3d) could be due to sea ice formation from already depleted surface seawater or to the contribution of meteoric ice (i.e., precipitation transformed into ice, including snow ice and/or superimposed ice formation) to the total ice thickness.

In summertime, the area is under the strong influence of surrounding glaciers, as inputs of freshwater runoff, depleted in δ¹⁸O, are responsible for the large decrease in surface seawater salinity and δ¹⁸O (Bendtsen et al., 2014), as observed during our survey starting on June 14 (Figure 9e). Sea ice formation from surface seawater already depleted in δ¹⁸O would result in depleted δ¹⁸O within sea ice (Geilfus et al., 2021). However, δ¹⁸O values at or above −5‰ have been observed for surface seawater in early fall (Rysgaard, unpublished data). Sea ice formation from such fall surface seawater would not likely be responsible for the δ¹⁸O values lower than −5‰ observed in the upper sea ice section of this study, especially as the freezing process potentially enriches the ice in heavy isotopes (maximum fractionation of +3‰ at low freezing rate; Souchez et al., 1988).

During our survey, sea ice was covered with 35–55 cm of snow (Table 1, Figure 2), insulating the ice from changes in atmospheric temperatures (Figure 10a) and thereby muting the thermal fluctuation within sea ice (Massom et al., 2001). Melting snow and meltwater refreezing at the ice surface is a well-known mechanism of ice formation (i.e., superimposed ice) during the warm season (Haas et al., 2001; Freitag and Eicken, 2003; Kawamura et al., 2004). Prior to the survey, the succession of short periods with warm atmospheric temperatures (Figure 10a) could have initiated successive melting and refreezing cycles of snow, leading to the formation of superimposed ice. However, a layer of slush was not observed initially (Table 1, Figure 2), and the relatively warm temperature at the snow-ice interface (Figures 2a and 3a) combined with the high salinity of the slush layer (Figure 2b) would make the formation of superimposed ice unlikely. In addition, the presence of large polygonal crystals, typical of superimposed ice (Kawamura et al., 2004), was not observed during the analysis of sea ice microstructure (Figure 11).

The thick snow cover also had the potential to depress the ice surface below sea level due to its own weight, as evidenced by the development of a negative freeboard (Table 1). This negative freeboard could result in seawater (laterally from the edge of the floe or through cracks and ridging) and/or infiltration of brine onto the ice surface, producing a layer composed of a mixture of snow and seawater and/or brine (i.e., slush). Under cold atmospheric conditions, this saline slushy layer can freeze, forming snow ice (Maksym and Jeffries, 2000). Snow ice is typically characterized by polygonal granular ice texture, low bulk ice salinity and depleted δ¹⁸O (Eicken and Lange, 1989; Lange et al., 1990). The sea ice microstructure (Figure 11), analyzed only for S6, highlights the presence of granular ice texture in the upper 35 cm of the ice column, followed by columnar ice towards the bottom of the ice where a thin layer of granular ice was also observed. The top 35 cm section of granular ice is associated with strongly depleted δ¹⁸O values (Figure 3d), followed by a layer of transition between granular and columnar ice texture (35–45 cm), supporting the presence of snow ice in the upper 35 cm at S6. The transition from granular to columnar ice (from 35 cm to 50 cm depth), characterized by a steep increase in δ¹⁸O values (from −5‰ to −2‰) and the presence of a local minimum in bulk salinity (Figure 3b), could correspond to the initial sea ice formation in the fjord from surface seawater of low salinity. Assuming the transition between granular and columnar ice and the steep increase in δ¹⁸O values observed at S6 are representative of the earlier sampling events, we estimated that the snow ice contribution to total ice thickness ranged from 25% (S6) to 40% (S4; Table 1).

However, granular ice texture, associated with depleted δ¹⁸O (−5.3‰; Figure 3d), was also observed in the bottom 4 cm of S6 (Figure 11). While sea ice microstructure analysis was not performed for S5, a depleted δ¹⁸O value (−9.6‰) was also observed in the bottom sea ice layer (Figure 3d), suggesting that a similar process may have occurred at both S5 and S6. Depleted δ¹⁸O values observed at the sea ice interface with the water column were associated with low bulk ice salinity (Figure 3b). Therefore, freezing of depleted δ¹⁸O and low salinity water at the ice-water column interface could have resulted in the observed depleted δ¹⁸O granular sea ice bottom.

Meltwater from snow or sea ice melt can drain through the sea ice matrix and accumulate below the sea ice. At the interface between this fresh water and the underlying seawater, double-diffusive convection of heat and salt occurs, leading to the formation of underwater ice, characterized by low bulk ice salinity and depleted δ¹⁸O values (Martin and Kauffman, 1974; Eicken et al., 2002; Notz et al., 2003; Polashenski et al., 2012), and a granular (frazil origin) texture. This process has been suggested to be the only process by which new ice can be formed during summer (Notz et al., 2003). Alternatively, refreezing of δ¹⁸O-depleted river water at the ice-water column interface could also be responsible for the presence of granular ice texture associated with depleted δ¹⁸O at the sea ice bottom. Starting on June 14, a decrease of both sea surface salinity and δ¹⁸O was observed (Figure 9) associated with increasing river discharges into Tyrolerfjord. As river waters in the area have a major influence from snow and ice (from the Greenland ice sheet) melt, their δ¹⁸O values must be close to the observed snow δ¹⁸O (Figure 2c). If a thin layer of undisturbed river water (i.e., depleted in δ¹⁸O freshwater) was present at the water column surface, its freezing at the contact with the sea ice bottom could be responsible for the presence of δ¹⁸O-depleted ice bottom layer. However, the presence of undisturbed river water below the sea ice was not observed. In addition, averaged surface seawater δ¹⁸O (−2.1‰, SD = 0.3, n = 5; Figure 9e) suggests that a seawater contribution to the bottom granular ice layer (with δ¹⁸O from −5.3‰ to −9.6‰) was unlikely. Therefore, we propose that a mixture of melt pond water (average δ¹⁸O of −13.6‰, SD = 0.2, n = 2; Figure 2c) and sea ice meltwater, draining through the sea ice and freezing at the sea ice-water column interface was responsible for the depleted δ¹⁸O, low bulk salinity, and granular ice texture observed at the sea ice bottom of S5 and S6.

The observed ranges of gas concentrations are consistent with previous studies from both Arctic (Zhou et al., 2013; Crabeck et al., 2014) and Antarctic (Matsuo and Miyake, 1966; Van der Linden et al., 2020) sea ice. While O₂ and N₂ are both biogenic gases, potentially affected by both biological activity and physical processes, Ar does not have biological sources or sinks and is only affected by physical processes. Therefore, Ar can be considered an inert gas and used as a tracer of physical processes impacting gas concentration and transport within sea ice (Zhou et al., 2013; Crabeck et al., 2014). The ratio [O₂]:[Ar] has been used in seawater to remove the physical contribution to O₂ concentration variability and determine the biological production of O₂ (Cassar et al., 2009; Eveleth et al., 2014). It has also been used to estimate the net community production within Arctic (Zhou et al., 2014) and Antarctic sea ice (Van der Linden et al., 2020). Sea ice gas ratios were compared with the gas ratio estimated for seawater (based on averaged surface seawater temperature, −1.7°C, and salinity, 30; blue dashed line in Figure 12) and the atmosphere (National Oceanic and Atmospheric Administration, 1976; back dotted line in Figure 12). Regarding the [O₂]:[Ar] ratio, the abiotic range is restrained by the seawater and atmospheric [O₂]:[Ar] value (20.5 and 22.5, respectively). Thus, for [O₂]:[Ar] values falling within the abiotic range, differentiating biological from physical and chemical processes is not possible, while [O₂]:[Ar] outside the abiotic range could be attributed to biological activity and considered for computation of net community production.

The relative proportions of O₂, N₂, and Ar (Figure 12b, c, and d) reflect a mixed contribution of the dissolved and gaseous fractions (i.e., bubbles) of each gas within the sea ice, with a dominance of the gaseous fraction, as our observations (regression line in Figure 12) were closer to the atmospheric ratio (dotted line in Figure 12) compared to the seawater ratio (dashed line in Figure 12). In addition, the estimated bubble fraction (Equation 3, Figure 8) suggests that up to 80% of the gas species were in the gas phase, while less than 20% were dissolved in the brine. This difference between gas and dissolved phase is consistent with gas observations previously reported in sea ice from Antarctica (Matsuo and Miyake, 1966), an Arctic fjord (Crabeck et al., 2014), and an artificial sea ice mesocosm (Tison et al., 2002). In this study, [O₂] and [N₂] correlated strongly with [Ar] (R² = 0.99 and 0.98, respectively; Figure 12), suggesting that physical processes mainly controlled the gas composition within sea ice, while biological processes had only a minor effect. However, literature record values of [O₂]:[Ar] as high as 31, well above the abiotic range, were observed at S6 (Figure 12a), suggesting a significant contribution of biological processes, as discussed later.

Maximum gas concentration (Figure 6) and gas content as bubbles (Figure 8) appear to be centred at 25 cm depth in the sea ice. This depth coincides with the granular–columnar ice transition and the presence of snow ice in the upper sea ice layer (Figure 11). At that depth, about 80% of the gas species were in the gas phase while less than 20% were dissolved in the brine. Snow ice and granular ice have been reported to contain more bubbles compared to columnar ice (Crabeck et al., 2016). The gas peak decreased over time, with minimum concentrations observed at the sea ice surface, suggesting a release of gas towards the atmosphere. Gas bubbles released to the atmosphere result in gas concentrations reaching the atmospheric saturation (as at S3 in Figure 7). However, surface sea ice was far below atmospheric saturation in gases in S4 and S5 (Figure 7) when atmospheric temperatures increased and remained above 0°C (Figure 10a), leading to the formation of melt ponds. The refreezing of meltwater, already depleted in gases (Glud et al., 2002; Rysgaard and Glud, 2004), will release gases towards the atmosphere, resulting in an ice layer largely depleted in gas. However, the presence of superimposed ice, formed from refrozen snow meltwater at the sea ice surface, was not observed during the thin section analysis at S6 (Figure 11). Alternatively, meltwater has been suggested to refreeze as it intrudes into brine-filled pore structures within the ice to form interposed ice plugs (Freitag and Eicken, 2003; Polashenski et al., 2012), resulting in a surface ice layer strongly depleted in gas (Figures 6 and 7). At S6, gas concentrations in the surface ice layer increased above the atmospheric saturation. At the time, melt ponds were completely developed and sea ice cores were extracted from the “dry ice,” usually referred to as “white ice” due to its high reflectance (Greenfell and Maykut, 1977; Perovich, 1996), located next to the melt ponds. As a result of melt pond formation, meltwater is drained from the upper layer of the white ice, leaving the drained ice surface with a high content of air inclusions (Perovich, 1996). Therefore, the upper sea ice section exhibited similar gas ratios (i.e., O₂:Ar) as the atmosphere (Figure 12) and concentrations above gas saturation due to the fact that the latter refers to a maximum dissolved concentration in the brines, lacking the impact of “air spaces” in the “rotten” surface ice.

In the sea ice interior, gas concentrations were close to, or below, atmospheric saturation (Figure 6), suggesting that gas incorporation at the ice-seawater interface occurred close to the atmospheric solubility (Zhou et al., 2013; Crabeck et al., 2014), while brine drainage could also have contributed to lowering the gas concentration. However, gas concentrations at the ice bottom of S5 and S6 were much lower than atmospheric saturation. These ice sections, characterized by depleted δ¹⁸O (Figure 3) and granular ice texture (Figure 11), are suggested to have formed by refreezing meltwater. Sea ice meltwater is undersaturated in O₂ (Glud et al., 2002); assuming that O₂, Ar, and N₂ behave in the same way (Rysgaard and Glud, 2004), the refreezing of meltwater, depleted in gases, will contribute to the undersaturation observed at the sea ice bottom. However, we cannot rule out partial gas losses during core extraction.

In order to estimate the sea ice net community production, we first estimated the impact of physical processes on the O₂ concentration ([O₂]_phys; Equation 6, Figure 13b), then the O₂ concentration associated with biological activity ([O₂]_bio; Equation 7, Figure 13c). While the observed [O₂] ranged from 1 to 237 µmol L_ice⁻¹ (averaging 73 µmol L_ice⁻¹, SD = 54, n = 73; Figures 7a and 13a), [O₂]_phys and [O₂]_bio ranged from 2.7 to 208 (average: 62, SD = 45, n = 80) and from 6.5 to 44 µmol L_ice⁻¹ (average: 10.5, SD = 10.3, n = 80, Figure 13), respectively, with maximum O₂ observed in the upper 25 cm of the ice column. Note that [O₂]_bio values are well above the measurement uncertainties for both O₂ and Ar (<1%; see Methods), revealing a clear signal. However, overall, the relative contribution of [O₂]_bio to the O₂ pool appears to be minor compared to [O₂]_phys, with estimated values ranging from −0.2 to 0.3 (Figure 13d). Nevertheless, increases in [O₂]_bio contributions were observed at S6, with maximum [O₂]_bio:[O₂] values in accordance with the high [O₂]:[Ar] ratio in S6 (Figure 12a).

Sea ice NCP assessments (in µmol L_ice⁻¹ d⁻¹; Figure 13e) are given for the upper 25 cm of the ice (noted surface), bottom 15 cm layer (noted bottom), and the sea ice interior (noted interior). Positive NCP values indicate an O₂ production (equivalent to an inorganic carbon uptake, i.e. autotrophy), while negative values correspond to O₂ consumption (equivalent to inorganic carbon production, i.e., heterotrophy). The range of NCP values estimated in this study (from −2.25 to 2.6 µmol L_ice⁻¹d⁻¹) fall within previous estimates for sea ice derived from [O₂]:[Ar] ratios (Zhou et al., 2014; Van der Linden et al., 2020) or from oxygen incubations (Søgaard et al., 2010; Campbell et al., 2017). Early in the survey (S2 and S3), NCP exhibited positive values, indicating net autotrophy (Figure 13). From S3 to S5, NCP values became negative, indicating that heterotrophy dominated. At the end of the survey (S5 and S6), sea ice became autotrophic again. Heterotrophy within sea ice has been previously reported in Young Sound, where an O₂ consumption up to 13 µmol L_ice⁻¹ d⁻¹ was estimated (Rysgaard and Glud, 2004), as well as in landfast sea ice in Franklin Bay (Canadian Arctic), with an estimated O₂ consumption of 3 µmol L_ice⁻¹ d⁻¹ (Rysgaard et al., 2008). The increase in the [O₂]_bio contribution to the overall pool of O₂ (Figure 13d) associated with positive estimates of NCP (Figure 13e) indicates that sea ice was net autotrophic from S5 to S6. If both Chl-a and nutrient concentrations were low (Figure 4), suggesting a low sea ice algal productivity during the study period, a slight (though up to 4-fold in the areas of high O₂:Ar) increase of Chl-a was observed in S5 associated with an increase of NO₃⁻, potentially increasing [O₂]_bio. Its rapid degradation would have led to the increase of PO₄³⁻ observed in S6. Therefore, the strong correlations between [O₂], [N₂], and [Ar] (Figure 12), associated with the minor contribution of [O₂]_bio to the overall pool of O₂ (Figure 13d), suggest that gas composition within sea ice is mainly controlled by physical processes, but with a minor and clearly discernible contribution of biological processes.

Most of the changes in TA and DIC observed within sea ice appeared to be driven mainly by salinity changes (Figure 5c and f). However, a fair number of data points from S1 exhibited lower/higher TA and DIC than expected from salinity changes. Data points with higher-than-expected TA and DIC (Figure 5c and f) correspond to data with nTA and nDIC >400 µmol kg⁻¹ (Figure 5b and e), aligned with the theoretical trend of calcium carbonate dissolution (Figure 5h). Conversely, data points with lower-than-expected TA and DIC (Figure 5c and f) correspond to data with nTA and nDIC <300 µmol kg⁻¹ (Figure 5b and e), aligned with the theoretical trend of calcium carbonate precipitation (Figure 5h). These patterns suggest that precipitates of calcium carbonate were present within the sea ice at S1 and started to dissolve in the upper and bottom layers of the ice as the ice warmed. While the presence of calcium carbonate precipitates (such as ikaite, CaCO₃•6H₂O) within sea ice was not observed during this survey, the area is known to support ikaite precipitation (Rysgaard et al., 2007), with estimates of ikaite precipitation in surface sea ice up to 900 µmol kg⁻¹ (Rysgaard et al., 2013). From S2 onwards, most of the sea ice data fall along the calcium carbonate dissolution line, with a slight offset toward CO₂ uptake and/or respiration (Figure 5h). Where the sea ice uptake of atmospheric CO₂ has been quantified (Figure 10c and d), the impact of biology may be minor overall, but is noticeable for S6 (Figures 12 and 13).

Air–ice CO₂ fluxes mainly depend on sea ice permeability and the CO₂ concentration gradient between the atmosphere and the ice surface (i.e., brine and meltwater, including slush and melt ponds). While brine volume calculations (Figure 3c) confirmed that sea ice was permeable to gas exchanges (Golden et al., 2007; Zhou et al., 2013), slush, brine, and melt ponds were all undersaturated in CO₂ compared to the atmosphere (Figures 2j and 6g), supporting the overall uptake of atmospheric CO₂ by the snow-covered sea ice (Figure 10c and d).

The air temperature and its impacts on the surface ice layer strongly affected the air–ice CO₂ exchanges. During most of the survey, a diurnal pattern of CO₂ fluxes can be observed, with stronger uptake of atmospheric CO₂ occurring when the chamber temperature rose to about 0°C as the air temperature increased during daylight. At nighttime, as atmospheric and chamber temperatures decreased, the air-ice exchanges of CO₂ weakened, with a slight release of CO₂ from the ice to the atmosphere on a few occasions (Figure 10c). Increasing air temperature initiated the snow melt and the formation of a slush layer at the snow-ice interface with the atmosphere (Figure 2). The input of snow meltwater into the slush resulted in decreasing slush salinity and pCO₂ (Figure 2j), supporting the uptake of atmospheric CO₂ by the snow-covered sea ice. The nocturnal refreezing of the slush layer and surface ice reduced the exchanges of CO₂ with the atmosphere and episodically induced a CO₂ release to the atmosphere, like the release of CO₂ to the atmosphere by young growing sea ice (Geilfus et al., 2013; Fransson et al., 2015). Similar patterns have been reported over landfast sea ice using both the chambers (Van der Linden et al., 2020) and eddy covariance technique (Papakyriakou and Miller, 2011).

Due to slush and sea ice brine undersaturation in CO₂ (Figures 2j and 5g), compared to the atmosphere, the snow-covered sea ice acted as a sink for atmospheric CO₂, with uptake up to −4.26 mmol m⁻² d⁻¹ observed from S1 to S4 (up to June 11; Figure 10c). However, as the air temperature remained above 0°C (June 11) and slush and brine pCO₂ reached their minima, the overall uptake of atmospheric CO₂ weakened (averaging −0.04 mmol m⁻² d⁻¹; Figure 10c). On June 12, meltwater began to accumulate at the sea ice surface, forming melt ponds. Therefore, to endure the air–ice CO₂ flux measurements, chambers were installed on the remaining sea ice from which the meltwater had drained from the surface elevation into the melt ponds, the so-called white ice (Perovich, 1996). As the upper layer of the white ice is depleted of liquid (brine and/or meltwater) and has a high content of air inclusions (Perovich, 1996), its ability to exchange CO₂ is reduced, leading to a weakening of the measured uptake of atmospheric CO₂ by the chambers. Therefore, the values obtained on white ice may not be representative for the actual uptake of atmospheric CO₂ by both melt ponds and sea ice.

Overall, the CO₂ fluxes reported in this study are in the same range as in previous studies, both Arctic and Antarctic, using a similar system (Geilfus et al., 2012; Nomura et al., 2013; Delille et al., 2014; Brown et al., 2015; Geilfus et al., 2015; Van der Linden et al., 2020). The decreasing uptake of CO₂ as melt ponds developed was also reported by Geilfus et al. (2015). However, this decrease was observed mainly over melt ponds and was attributed to the decreasing pCO₂ gradient between melt ponds and the atmosphere as melt ponds tend to reach equilibrium with the atmosphere over time (Geilfus et al., 2015). In this study, the sea ice uptake of atmosphere CO₂ weakened before melt pond formation, probably due to the decreasing amount of brine and meltwater contained in the ice as it is drained from the upper ice layer, either vertically through the ice thickness or horizontally to form the melt ponds.

During the spring–summer transition at our North Greenland Arctic fjord location, the increase in air temperature led to the formation of melt ponds, resulting in significant changes in sea ice physical properties, gas composition, and the ability of sea ice to take up atmospheric CO₂. Physical processes mainly controlled the gas composition, while the impact of biological processes was noticeable at the end of the survey (S5 and S6; Figures 12 and 13). Maximum gas concentrations and bubble content were observed in the upper sea ice sections, with the surface sea ice concentration decreasing over time due to the release of gas to the atmosphere. At the sea ice interface with the atmosphere, the accumulation of meltwater into melt ponds was associated with the formation of interposed ice, where meltwater refreezes as it intrudes into the brine-filled pore structure, forming ice plugs (Freitag and Eicken, 2003; Polashenski et al., 2012). As meltwater drained throughout the ice column and accumulated at the sea ice bottom, it refroze forming a layer of underwater ice (Martin and Kauffman, 1974; Eicken et al., 2002; Notz et al., 2003; Polashenski et al., 2012). In both cases, ice formation from meltwater already depleted in gas resulted in ice strongly depleted in gas, first at the ice surface during early melt pond formation, then at the sea ice bottom (Figures 6 and 7).

Due to the increasing sea ice temperature, brine freshening, and ikaite dissolution, the sea ice was mainly undersaturated in CO₂ compared to the atmosphere (Figure 5), resulting in the uptake of atmospheric CO₂ (Figure 10). However, as melt ponds formed, the remaining ice surface (the “white ice”) became depleted in water as brine and meltwater drained and accumulated into melt ponds. Therefore, even though the sea ice was strongly undersaturated in CO₂ compared to the atmosphere, it could not exchange CO₂ with the atmosphere, resulting in a weakening uptake of atmospheric CO₂. This study illustrates the important role of melt pond formation and dynamics in controlling the physical properties and gas composition of sea ice, with strong consequences for sea ice uptake of atmospheric CO₂.

Data are available at Geilfus, N. (2022). Daneborg—Young Sound 2014, HydroShare, https://doi.org/10.4211/hs.a3c0d38322fc46ea96ecea2438b29283. Air temperature data (Figure 10a) are accessible at https://data.g-e-m.dk/, doi:10.17897/XV96-HC57.

We gratefully acknowledge the contributions of the Canada Excellence Research Chair (CERC) and Canada Research Chair (CRC) programs. Support was also provided by the Natural Sciences and Engineering Research (NSERC) Council, the Canada Foundation for Innovation and the University of Manitoba. We wish to thank the entire team behind the Daneborg Research Station, managed by Aarhus University. This work is a contribution to the ArcticNet Networks of Centres of Excellence and the Arctic Science Partnership (ASP) asp-net.org. BD is a research associate of the F.R.S-FNR. The authors are in debt to Marie Kotovitch, Odile Crabeck, and Egon Frandsen for their technical assistance in the field experiment and subsequent analysis.

Financial support for this study was provided by the Canada Excellence Research Chairs program, the Canada Research Chairs program, the Natural Sciences and Engineering Research Council, and the Canada Foundation for Innovation, the Belgian Science Policy (project Bigsouth, SD/CA/05A) and the Fonds National de la Recherche Scientifique (F.R.S-FNRS). BD is a Research Associate of the FRS-FNRS.

The authors declare no competing interests.

Field sampling was performed by N-XG and BD, laboratory and data analyses were performed by N-XG, ML, BD, and J-LT, and the manuscript was written and edited by N-XG, BD, ML, J-LT, and SR. All authors approved the submission of this manuscript.

How to cite this article: Geilfus, N-X, Delille, B, Tison, J-L, Lemes, M, Rysgaard, S. 2023. Gas dynamics within landfast sea ice of an Arctic fjord (NE Greenland) during the spring–summer transition. Elementa: Science of the Anthropocene 11(1). DOI: https://doi.org/10.1525/elementa.2022.00056

Domain Editor-in-Chief: Jody W. Deming, University of Washington, Seattle, WA, USA

Associate Editor: Christine Michel, Department of Fisheries and Oceans, University of Manitoba, Winnipeg, Canada

Knowledge Domain: Ocean Science

Part of an Elementa Special Feature: Insights into Biogeochemical Exchange Processes at Sea Ice Interfaces (BEPSII-2)