CO₂-system development in young sea ice and CO₂ gas exchange at the ice/air interface mediated by brine and frost flowers in Kongsfjorden, Spitsbergen

Abstract

In March and April 2010, we investigated the development of young landfast sea ice in Kongsfjorden, Spitsbergen, Svalbard. We sampled the vertical column, including sea ice, brine, frost flowers and sea water, to determine the CO₂ system, nutrients, salinity and bacterial and ice algae production during a 13 day interval of ice growth. Apart from the changes due to salinity and brine rejection, the sea-ice concentrations of total inorganic carbon (C_T), total alkalinity (A_T), CO₂ and carbonate ions (CO₃^2–) in melted ice were influenced by dissolution of calcium carbonate (CaCO₃) precipitates (25–55 μmol kg^-1) and played the largest role in the changes to the CO₂ system. The C_T values were also influenced by CO₂ gas flux, bacterial carbon production and primary production, which had a small impact on the C_T. The only exception was the uppermost ice layer. In the top 0.05 m of the ice, there was a CO₂ loss of ∼20 μmol kg^-1 melted ice (1 mmol m^-2) from the ice to the atmosphere. Frost flowers on newly formed sea ice were important in promoting ice-air CO₂ gas flux, causing a CO₂ loss to the atmosphere of 140-800 μmol kg-^-1 d^-1 melted frost flowers (7-40 mmol m^-2 d–¹).

Introduction

Changes in the Arctic Earth System (e.g. decreased ice extent and thickness) have been reported, and in September 2012 the lowest ice extent on record was observed (Parkinson and Comiso, 2013). As sea-ice extent decreases and freshwater volume increases, the biogeochemical processes will change. As a consequence, the marine environment, vertical carbon transport and carbon dioxide (CO₂) exchange between the atmosphere and ocean will all be altered.

Processes in the sea ice affect the CO₂ system (or carbonate system) within the sea ice and at the interfaces with the air and the underlying water (e.g. DeLille and others, 2007; Rysgaard and others, 2007; Nomura and others, 2009; Fransson and others, 2011,2013; Miller and others, 2011a,b). Brine rejection transporting chemical substances between the atmosphere and water is important for CO₂ transport from the atmosphere to deeper layers of water (e.g. Rysgaard and others, 2007; Else and others, 2011). Brine is transported within the sea ice. It is transported downwards due to gravity and upwards at ice formation due to a freezing ice layer near the bottom. This reduces the permeable channels and hence squeezes brines upward. Frost flowers are sometimes formed in this process, as the brine produces a brine skim on the ice surface (Perovich and Richter-Menge, 1994; Alvarez-Aviles and others, 2008). Frost flowers are crystalline structures that form on new sea ice when the air, which is supersaturated with water vapour, condenses at the ice surface. They carry high concentrations of chemical substances, gases, bacteria and salts (e.g. Deming, 2010; Douglas and others, 2012; Fransson and others, 2013; Geilfus and others, 2013; Granfors and others, 2013) and have a 40% larger surface area than sea ice (Domine and others, 2005), promoting gas exchange with the atmosphere. Frost flowers are easily destroyed by wind or blowing snow and consequently only have a brief life span of a few days (e.g. Deming, 2010).

The importance of sea ice for the CO₂ system characteristics in the underlying water has been shown by Fransson and others (2013). Changes in the underlying water will occur during the sea-ice season. This is due to the increased CO₂ in brine and depletion in ∆ _T compared to C _T due to CaCO₃ precipitation during ice freezing (Rysgaard and others, 2007, 2013; Fransson and others, 2011, 2013) and the excess ∆ _T in relation to C _T and salinity during ice melt. Solid CaCO₃ in the form of ikaite has previously been found in Arctic sea ice (Dieckmann and others, 2010; Rysgaard and others, 2012, 2013; Nomura and others, 2013). When solid CaCO₃ precipitates in the sea ice, CO₂ is released to the aqueous phase, and when it dissolves, CO₂ is consumed:

(1)

The solid CaCO₃ may be left in the ice until the onset of melt, during melt or when the ice becomes porous enough to release the CaCO₃ to the underlying water (e.g. Lyakhin, 1970). This (intermittent) storage changes the ratios between the CO₂ system parameters and salinity, and changes the vertical carbon fluxes and ice–air CO₂ gas flux (e.g. Fransson and others, 2013). Ice–air CO₂ gas fluxes were previously estimated in Arctic sea ice and showed that CO₂ escaped from the ice to the atmosphere during sea-ice formation (Papadimitriou and others, 2004; Nomura and others, 2006; Else and others, 2011, 2012; Miller and others, 2011a,b; Papakyriakou and Miller, 2011; Rysgaard and others, 2012; Fransson and others, 2013).

It has been demonstrated that biological processes (e.g. primary production and bacterial respiration) are responsible for the variability of the sea-ice CO₂ system (e.g. Fransson and others, 2011, 2013). In winter, bacterial respiration is an important process for the production of CO₂ at the bottom of the ice, causing an increase in C _T and CO₂ gas flux from the ice (Fransson and others, 2013). In spring, primary production in the bottom ice consumes CO₂, causing a decrease in C _T, which results in CO₂ uptake by the ice from the surrounding environment (Fransson and others, 2013).

In this study, we investigate the development of the CO₂ system in young sea ice, focusing on the effect of sea-ice processes in the role of CO₂ exchange at the ice/air interface during a 13 day interval in Kongsfjorden, West Spitsbergen, Svalbard. To our knowledge, this is the first CO₂ system study performed in a fjord of Spitsbergen. We present: (1) the evolution of the CO₂ system of young sea ice; (2) an estimate of the effect of the biogeochemical processes (e.g. bacterial and primary production and CaCO₃ precipitation/dissolution); (3) estimates of CO₂ gas fluxes at the ice/air interface; and (4) a discussion of the importance of frost flowers for CO₂ transport from the ice to the atmosphere.

Study Area

The ice study was carried out between 21 March and 2 April 2010, in the Thiisbukta Bay area in Kongsfjorden, ∼500m from Ny-lesund, West Spitsbergen, at 78.927° N, 11.901° E (Fig. 1). The total surface area of the bay is ∼10 000 m² (Fig. 2a). The thickness of the young landfast sea ice (referred to as young ice) was 0.27 m at the start of the study and was formed by freezing of pancake ice (Fig. 2a). Throughout the study, the young ice was void of snow, but frost flowers occasionally formed near the study area (Fig. 2b). On a few occasions, we sampled thin pancake ice (PCI; newly formed sea ice, 0.04 m) near Thiisbukta (TB). In addition to TB sea ice, on 28 March we sampled thicker landfast sea ice (0.42 m thick) in Engelskbukta (EB; 78.832° N, 11.895° E; Fig. 1), situated 10 km (flying distance) from Ny-lesund on the other side (southern part) of the Brøgger Peninsula (Fig. 1).

The average daily air temperature varied between -9°C and -16°C. The lowest temperature was recorded on 27 March and the highest on 31 March (Table 1). The average daily wind speed varied between 1 and 6 ms–¹. The highest wind speed was recorded on 31 March (data courtesy of Norwegian Meteorological Institute). The measuring site is located 8ma.s.l. and 570 m from the TB study area. The light conditions, obtained from our own measurements of photosynthetically active radiation (PAR; 400-700 nm), varied between zero (during all the nights) and the maximum value of 800 μmol photons m^–2s–¹ on 26 March, At the end of the study, due to cloudy conditions this dropped to 350-400 μmol photons m^–2s–¹ (Table 1). In the sea ice, the radiation measurements of incident PAR showed that 8% remained at a depth of 0.35 m (at midday).

Fig. 1. Map of the study site at Thiisbukta (78.927° N, 11.901° E) and location of sampling for young landfast sea ice, brine, under-ice water, frost flowers and newly formed pancake ice. Landfast sea ice and brine sampling at Engelskbukta (78.832°N, 11.895°E) is denoted as EB on the map.

Fig. 2. Photographs showing (a) the Thiisbukta site, and (b) the sampling of newly formed frost flowers (magnified frost flowers picture included).

Table 1. Sampling dates of young landfast sea ice, brine and under-ice water, ice thickness of young ice, and meteorological data (mean air temperature, mean wind speed and mean relative humidity) from Ny-lesund weather station (8 m a.s.l.) at the study site in Thiisbukta (78.927° N, 11.901° E), and date for sampling thicker landfast sea ice at Engelskbukta (EB; 78.832° N, 11.895° E). EB is located on the other side of the Brøgger Peninsula 10 km from Ny-lesund (flight distance). See Figure 1 for map location. Note that under-ice water (0–2 m) was collected parallel to sea-ice sampling. There was no snow cover at the site and no precipitation during the study

Methods

Sampling of bulk sea ice, brine, frost flowers and underlying water

During the 13 day study, we collected young ice cores for carbonate chemistry on 11 occasions (21, 22, 23, 24, 26, 27, 29, 30 and 31 March and 1 and 2 April 2010; Table 2 ). Samples of microalgal and heterotrophic bacterial abundances and activities in the young ice were collected on five occasions (21, 23, 26, 29 and 31 March 2010). Additional ice-core sampling of ice in Engelskbukta and pancake thin ice was performed during the study. Sea-ice cores were sampled using a stainless-steel ice-core drill (diameter 0.12 m) with a power head. The ice cores were divided into 0.05 m sections, which were individually placed into Tedlar© gas-sampling bags. After carefully removing surrounding air from the bags using a hand-operated vacuum pump (Nalgene^©), the bulk sea-ice samples (hereafter referred to as sea ice) were slowly thawed in darkness for ∼24 hours to reach a temperature varying between 4°C and 10°C. No solid calcium carbonate (CaCO₃) was observed in our samples. This is likely because on some occasions the sample temperature exceeded 4°C. The resulting volume of the melted sea ice was ∼0.5 L. Microbiological abundance and activity were sampled from separate ice cores collected within a 0.2 m radius of the cores sampled for chemical analysis. The ice cores were immediately wrapped in black plastic to protect the algae from light stress, and were cut into 0.05 m sections for further processing at 0°C.

Table 2. Physical and carbonate system properties of young landfast sea ice in Thiisbukta

In parallel to ice-core sampling, brine samples were collected (Table 3) in 0.1 L borosilicate glass bottles (airtight) from partially drilled holes (known as sackholes) in young sea ice (hereafter referred to as YI). We refer to these samples as brine. The brine, which had seeped into the sackhole at different depths in the ice (0.1-0.2 m), was collected with a plastic syringe and PVC tubing after ∼30-40min and transferred to the bottles. Due to the cold sea-ice conditions, the volumes of brine were usually insufficient for analysis and studies of the full CO₂ system. Further, we could only collect brine samples on three occasions during the study. The sackholes were covered with a plastic isolated lid to minimize gas exchange with the atmosphere and prevent freezing of the brine. During our study, the seeping time for brine was sometimes >30 min to allow for sufficient sample volume (0.5-0.1 L). Consequently, the occurrence of some gas exchange cannot be excluded.

Table 3. Physical and CO₂-system properties in pancake ice (PCI), landfast sea ice in Engelskbukta (EB, thicker ice), brine (Brine), frost flowers (FF_new, FF_old), and under-ice water under young ice in Thiisbukta (UIW) and at Engelskbukta (EB). N denotes number of samples and n.a. stands for not applicable

Under-ice water (UIW) was collected in 0.25 L borosilicate glass bottles on every occasion when sea-ice cores were sampled, using a glass bottle on a (custom-built) shaft immersed through the borehole. Although debris from the ice was removed before sampling, some debris may be left in the water collected. The water and brine samples were immediately placed in an insulated cooling box to prevent them from freezing.

Frost flowers were sampled ∼200m from our study site on 27 and 28 March using a Teflon ladle from a surface area of 1 m² for each sample. The frost flowers were sampled (duplicate) as soon as they formed, and again 20 hours later (one sample). The samples were packed in Tedlar© sampling bags, and the air was gently removed using the hand-operated vacuum pump. After removing the surrounding air from the bags, the samples were thawed in darkness. The melted volume was ∼1 L.

Sea-ice temperature was measured on site immediately after the ice core was recovered, at 0.05 m intervals using a digital thermistor (Amadigit) with an accuracy of ±0.1°C. The holes for the temperature measurement were carefully drilled manually with a clean stainless-steel hand drill to avoid additional heating from the drill. The temperature of brine was measured in the sackhole before it was transferred to a sample bottle. UIW temperature was measured in the sample bottle immediately after sampling using the same handheld thermistor probe (Amadigit).

Data and analysis

After melt of the bulk sea-ice and frost flower samples, we measured the CO₂ system parameters, nutrients, bacterial and algal activity and salinity in the melted samples. All sea-ice and frost-flower measurements are hereafter referred to as concentrations and activities in melted samples.

CO₂ system

We measured the CO₂ system (also referred to as the carbonate system) parameters of total alkalinity (∆ _T) and pH in melted sea ice, brine, frost-flower melt and UIW, using the same method as Fransson and others (2011). ∆_T and pH in sea ice and frost flowers referred to in this study were always measured in liquid phase (sea ice and frost-flower melt). AT was determined by potentiometric titration in an open cell (Dickson and others, 2007) with ∆_T measurement precision of ± 3 μmol kg–¹. The pH was determined spectrophotometrically using a 2 mM solution of the sulphonaphtalein dye m-cresol purple (Sigma-Aldrich®), as an indicator valid for water 3 0 >S>37 (Clayton and Byrne, 1993). The accuracy of the pH measurements is determined by the accuracy of the determination of the equilibrium constants of the dye, which is ∼±0.002 (Dickson, 1993). We used the same indicator batch for the whole study, and studied pH changes between days, so the systematic error due to indicator impurities (Yao and others, 2007) would induce the same error for all measurements. For cold-ice brines, the salinity usually exceeds 40, reducing the accuracy of the measurements. However, Hare and others (2013) investigated the deviation regarding high salinities and low temperatures using m-cresol purple, and found that the natural variability in sea ice and brine is larger than the analytical uncertainty. The samples were thermostatted to 15°C prior to analysis. Samples were measured in a 0.01 m cell. The analytical precision was estimated to ±0.001 pH units, which was determined by a series of ten analyses of a sample. The pH of the indicator solution was measured daily using a 0.0002 m cell. The precision was checked by triplicate analysis of one sample every second day. The perturbation of sea-water pH caused by the addition of the indicator solution was calculated and corrected for using the method described by Chierici and others (1999).

The A_T and pH values, measured at 15°C (pH₁₅), salinity and temperature along with the CO₂ calculation program CO2SYS (Pierrot and others, 2006) were used to calculate total inorganic carbon (C _T (μmol kg^-1), sometimes referred to as DIC or TIC), carbonate ion concentration ([CO₃ ²–]), CO₂ concentrations ([CO₂]; μmolkg–¹), partial pressure of CO₂ (p ^CO2) and fugacity of CO₂ (fCO₂). All calculated CO₂ system parameters mentioned in this study are derived from AT and pH measured in liquid (melted) phase. The CO₂ system dissociation constants (K1 and K2) estimated by Roy and others (1993, 1994) were used, since an internal consistency study showed these were the most suitable constants for cold waters (Chierici and Fransson, 2009). The calculations were performed on the total hydrogen ion scale (pH_T) using the HSO₄ ^- dissociation constant of Dickson (1990). In order to investigate the error and variability introduced by using pH measured in melted bulk sea ice, we performed an internal consistency exercise using a comparison between the calculated C _T (C _Tcal_c) from measured ∆_T and pH_{1 5} and the measured C _T (C _Tmeas). The data were obtained from analyses of melted ice, sampled in Kongsfjorden in April 2013. The CO₂ system consistency check was performed using four different sets of equilibrium constants of K1 and K2: (1) Roy and others (1993, 1994; hereafter referred to as Roy); (2) Mehrbach and others (1973) refit by Dickson and Millero (1987) (hereafter referred to as Mehr); (3) Millero (1979; hereafter referred to as Millero); and (4) on the GEOSECS NBS scale (from Mehrbach and others, 1973; hereafter referred to as NBS). Figure 3 shows a box-and-whisker plot with the resulting difference between C _Tmeas and the calculated C _T based on the four sets of constants (C _Tmeas-C _Tcal_c). We found that the C _T values based on the Roy and Mehr constants showed the lowest median deviation from the measured C _T of ∼2.7μmol kg–¹ (Table 4). They also had the lowest standard error (SE) between the measured and calculated values and the best coefficient of determination (R2). K1 and K2 determined by Millero at zero salinity had the largest median deviation from measured CT of ∼6.3 μmol kg–¹ (Table 4). In addition, the results from the NBS scale showed a larger median deviation from the measured than the constants estimated based on the sea-water salinity range. This means that by using the Roy constants, the median deviation in the calculated C _T was ∼3 μmol kg–¹ and the SE was ±11 μmol kg–¹ (Table 4). This standard error of the calculated C _T was considered insignificant in relation to the natural variability in sea ice.

Fig. 3. The residual (μatm) between the measured C _T (C _Tmeas) in bulk sea ice and the calculated _CT (C _Tcalc) from the four sets of constants: Roy denotes the constants by Roy and others (1993); Mehr denotes the constants by Mehrbach and others (1973), refit by Dickson and Millero (1987); C _T S=0 denotes the constants by Millero (1979) at zero salinity (S = 0); and NBS denotes the constants by the GEOSECS NBS scale from Mehrbach and others (1973). The 25th and 75th percentiles show the limits where 2 5% and 75% of the residuals fall below. The non-outlier range isthe range of values that fall below the upper outer limit of ± 1 . 5 x the height of the box (H).

Table 4. Summary of the resulting median, values at the 25th and 75th percentiles, the standard error (SE) and the coefficient of determination from the regression analysis (R2) of the residual (μatm) between the measured C _T (C _Tmeas) in bulk sea ice and the calculated C _T (C _Tcalc) from each of the four sets of constants: Roy denotes the constants by Roy and others (1993); Mehr denotes Mehrbach and others (1973), refit by Dickson and Millero (1987); at zero salinity (C _T S=0) from Millero (1979); and on the GEOSECS NBS (NBS) scale from Mehrbach and others (1973). The 25th and 75th percentiles show the limits where 25% and 75% of the residuals fall below

Dissolved inorganic nutrients

Dissolved inorganic nutrients of nitrate ([NO₃–]), phosphate ([PO₄ ³–]) and silicic acid ([Si(OH)₄]) were measured on the melted samples, brine, frost-flower melt and UIW. Acid-washed plastic vials were used for sampling. Samples were filtered (0.45 μm) and frozen at-20°Cfor post-analysis at the Swedish Meteorological and Hydrological Institute, Gothenburg. The nutrients were determined by colorimetric detection and analysed on an auto-analyser using the standard method described by Grasshoff and others (2009).

Microbiological analysis

In general, all ice processing for microbiological analysis was performed at 0°C. Bacterial carbon production (BCP) was measured in centrifuged brine from bulk ice (and in sackholes). Bacterial production (BP) in the drained brine was determined by (³H)-leucine incorporation (Smith and Azam, 1992). The detection limit was 0.5 μg C L^–1 d ^{– 1} in the samples. Details on the separation of brine and the methodology for microbiological analysis are thoroughly described by Granfors and others (2013). Sea-ice algae activity and photosynthetic efficiency in the brine were investigated from the maximum quantum yield of photosynthesis (F_v/F_m) using a WATER-PAM chlorophyll fluorometer (Walz Mess- und Regeltechnik, Effeltrich, Germany) and a pulse-amplitude modulated (PAM) fluorometry, respectively. The remaining brine sample was fixed with glutaraldehyde (final concentration 0.1% and 2.5%, respectively) to determine bacterial and microalgal abundance. Samples for bacterial abundance were stored at –80°C for 6 months until analysis, described in Granfors and others (2013).

Salinity

Salinity and conductivity of the melted sea ice, brine, frost flowers and UIW were measured using a conductivity meter (WTW Cond 330i, Germany) with a resolution and accuracy of ±0.05.

We calculated brine volume fractions (BV) from young-ice salinity (S) and ice temperature (T; °C), according to Frankenstein and Garner (1967) derived from Assur (1958):

(2)

The ice becomes less permeable as BV decreases (e.g. Golden and others, 2007; Loose and others, 2009, 2010) and both gas and liquid transport decrease. Ice temperature fundamentally controls the ice porosity (Petrich and Eicken, 2010).

Solar radiation

For radiation measurements, a PMA2100 radiometer equipped with a 2-K PMA2132 sensor (Solar Light, Philadelphia, PA, USA) was used to record photosynthetic photon flux (PPF) at 400-700 nm (corresponding to photosynthetic active radiation (PAR)) on the pier at Ny-lesund, i.e. 500 m from the study site. In addition, PPF was measured in ice with a spherical sensor (QSL-2100, internal diameter 1.25 cm, Biospherical Instruments Inc., San Diego, CA, USA). To avoid shading from the instrument, the sensor was deployed at a 45° angle through the ice.

Results

Physical properties in young sea ice, brine and underlying water

The vertical distribution of salinity in the young-ice cores showed a C-shaped pattern, with higher salinity in the top and bottom ice than in the middle parts, which is typical for first-year ice (e.g. Malmgren, 1927; Thomas and others, 2010), ranging between 14 (22 March) and 10 (31 March) in the top 0.05 m (ice/air interface; Fig. 4a) in young ice. In the interior young ice, the salinity decreased from 9 to 6, whereas in the bottom ice (ice/water interface) little change was observed during the 13 day study. The sea-ice temperature increased linearly with depth from top to bottom and was lowest (-10.9°C) in the top ice, in contact with cold air (Fig. 4b), increasing towards the bottom ice, to reach temperatures close to sea-water temperatures (Fig. 4b; Table 2 and Table 3). The coldest ice was observed between 27 and 30 March, when we also recorded the lowest air temperatures.

Fig. 4. Vertical and daily distribution of (a) bulk sea-ice salinity, (b) temperature (T; °C) and (c) brine volume fraction (BV) in young landfast sea ice, from 21 March to 2 April 2010; obs and calc denote observed and calculated values, respectively.

The thickness of the young ice varied between 0.27 and 0.41 m, with an increase of ∼0.011 m d–¹ (0.14 m increase during the 13 day study) (Table 1). Freeboard was positive throughout the study, suggesting no flooding of the ice, and there was no superimposed ice from melt and refreeze.

BV in the young ice was mostly >0.05, except between 28 March and 1 April, coinciding with the lowest young ice temperatures at depths of 0.05-0.2 m (Fig. 4b and c). These low BV indicated that the mobility of chemical substances (and gases) dissolved in brine was less in the interior ice (BV<0.05; Cox and Weeks, 1983; Loose and others, 2009, 2010) than in the top and bottom ice (BV>0.05).

As the thin pancake ice (<0.04m) formed, ∼45% of the ice salinity was lost (from 34 down to 19), showing a substantial and rapid salinity rejection during the formation (Table 3).

Chemical properties in young sea ice, brine and underlying water

All concentrations for bulk sea ice and frost flowers shown in this and following subsections refer to values in melted samples. Young ice (YI) refers to thinner ice from TB, and EB ice refers to thicker ice from Engelskbukta.

CO2-system evolution

Large variability was observed in all CO₂ system parameters (Fig. 5; Table 2 and Table 3). The ∆_T and C _T of melted samples of young ice (Fig. 5a and b) showed a C-shaped trend each day, similar to salinity, indicating that most of the changes in young-ice C _T and ∆_T were due to salinity changes. ∆_T was highest at the ice/air interface compared to the rest of the ice during the 13 day study. The highest value was measured on 22 March, and ∆_T decreased towards the end of the period (Fig. 5b). Between 24 and 26 March, ∆_T increased by 120 μmol kg–¹from 790 to 910 μmol kg–¹. The calculated concentration of CO₂ ([CO₂]) in melted ice was generally lower in the top ice and higher in the bottom ice, but varied locally on a few occasions (Fig. 5c). [CO₃ ^2–] was highest in the top 0.05 m of the ice and decreased towards the bottom ice (Fig. 5d). In the interior ice, higher [CO₂] land lower [CO₃ ^2–] were found between 27 and 30 March (Fig. 5c and d).

Fig. 5. Vertical and daily distribution of the properties in melted ice of (a) total alkalinity (A _T; μmol kg^–1), (b) total inorganic carbon (C _T; μmol kg^–1), (c) carbon dioxide concentration ([CO₂]; μmol kg^–1) and (d) carbonate ion concentration ([CO₃ ^2–]; μmol kg^–1), in young landfast sea ice from 21 March to 2 April 2010; obs and calc denote observed and calculated values in melted sea ice, respectively.

The values of C _T and ∆ _T in melted EB ice were similar to those in melted young ice, but [CO₃ ^2–] and [CO₂] were almost 200 and 10 times larger, respectively, than in young ice (Table 2). The values of C _T, ∆ _T and salinity in melted pancake ice (PCI; newly formed thin ice <0.04m) were twice as high (Table 3) as in young ice and EB ice. [CO₃ ^2–] and [CO₂] were slightly higher in pancake ice than in young ice, but lower than in EB ice (Table 2). Pancake ice (thin ice) had higher ∆ _T, C _T and salinity than any of the other melted sea-ice samples in the study (Table 3).

Brine concentrations provide an integrated signal of changes in salinity and chemical substances that occurred over time. In our brine samples, [CO₂] was >300 times higher than in melted young ice, whereas [CO₃ ^2–] in brine was of the same order of magnitude as in melted young ice (Table 2 and Table 3). The fugacity of CO₂ (fCO₂) values in our brine samples was up to 18 000 μatm (1824 Pa) (Table 5).

Table 5. Inorganic nutrients (phosphate (PO₄ ^2–), nitrate (NO₃ ^–), silicic acid (Si(OH)₄) in young landfast sea ice (YI), landfast sea ice from Engelskbukta (EB), brine, and frost flower (FF). * denotes that brine values are mean values from four sampling occasions. Minimum and maximum values are denoted min and max, respectively

The salinity, ∆_T and CT values of UIW at the EB site were, on average, higher than the UIW values in TB.

Nutrients evolution

In young ice, [NO₃–] and [Si(OH)₄] did not change significantly during the study, but [PO₄3–] decreased from 0.2 to 0.04 mmol m– 3 in melted ice during the growth of the young ice (Table 5). The nutrient concentrations in the EB ice were, on average, similar to those in the young ice at the end of the study (Table 5). In brine, the nutrient concentrations were higher than in the melted young ice, with [PO₄ ³–] 10 times higher and [NO₃–] and [Si(OH)₄] ∼15 times higher (Table 5). In UIW, the concentrations of all three measured nutrients decreased during the study, with [PO₄ ³–] decreasing from 1.1 to 0.66, [NO₃–] from 14.7 to 8.5 and [Si(OH)₄] from 7.0 to 3.9 μM.

Biological properties in sea-ice brines, frost flowers and underlying water

Abundance and activity of heterotrophic bacteria

In all the ice samples, there was a standing stock of heterotrophic bacteria in young ice with bacteria present. The BCP ranged between 0.5 and 9.1 μgCL^–1 d ^–1 during the study. The highest value of BCP (9.1 μgC ^–1 d^–1) was detected on 29 March in the interior of young ice, corresponding to an increase of inorganic carbon by 0.8 μmol Ckg^–1 d^–1. In the brine from centrifuged bulk ice, the bacterial abundance varied between 7.1 x 1 0 ⁷ and 6.4 x 10⁸ cells L^–1. This was of the same order of magnitude as other measurements made in winter Arctic sea ice (Junge and others, 2004) of 107–10⁹ cells L–¹.

In the EB ice, the BCP was higher (up to 14μgCL^–1 d^–1) than in young ice, with higher values in the top ice and bottom ice. In the frost flowers, the BCP was2.8 μgCL^–1 d^–1. The BCP in UIW was 1.3 μgC L^–1 d^–1 on 23 March, with a maximum value of ∼5.0μgCL^–1 d^–1 on 24 March. It was 1.4 μgCL^–1d^–1 on 26 March, and decreased towards the end of the study.

Abundance and activity of microalgae

Microalgae were quantifiable in the bottom layer of the young-ice core. The abundances varied between 7.1 x 10⁴ and 8.8 x 1 0 ⁵ cells L–¹ in centrifuged brine. The most frequently occurring microalgal genera were pennate diatoms Synedropsis sp., and Fragilariopsis sp., which is typical for ice algal communities. During the study, we measured photosynthetically active algal cells in the young ice. Our estimates of F_v/F_m varied between 0.06 and 0.5. The maximum quantum yield of photosynthesis (F_v/F_m, a measure of the health of the algae) could be measured in the bottom 0.05 m (ice/water interface) of the young ice. The F_v/ F_m was 0.06 at the start on 23 March (bottom 0.05 m) and increased to a maximum of 0.5 on 26 March, likely due to the maximum solar radiation on 26 March.

There were fewer ice algal cells and lower F_v/F_m (0.07-0.25) in the EB bottom ice than the TB young ice. Moreover, the ice algae in the EB thick ice were distributed in the bottom 0.10 m of ice instead of being restricted to the bottom 0.05 m.

Processes driving the CO₂ system

To eliminate the effect of salinity on the changes in young-ice C _T and ∆_T, we used the mean young-ice salinity (S _ice) of 8.8 (following a dilution line; nC = 8.8/S _ice C, where C represents young-ice C _T, ∆_T and nitrate) to obtain salinity-normalized values, hereafter referred to as nδ5C _T and nδA _T. The nδC _T and nδA_T values showed a net increase at all ice horizons except in the top 0.05 m of the ice, where both parameters showed a net decrease over the whole study period (Fig. 6). This estimate was based on the difference between the values on 21 March and 2 April (Fig. 6). The main processes driving the changes in nδC _T are biological processes such as primary production (5C_PP), bacterial carbon production (5C_BCP), CaCO₃ precipitation and dissolution (δC_CaCO3) and CO₂ gas flux (δCFCO2)

Fig. 6. The change (μmol kg–¹) at five different ice horizons in the young ice from top to bottom in salinity-normalized total alkalinity (nAT, black) and total inorganic carbon (nδCT, white). The change in nC _T due to processes other than salinity changes is the combined effect of the biological processes of primary production and bacterial-carbon production (SC_PP + SC_BCP, green) estimated from salinity-normalized [NO₃–] change, the effect of calcium carbonate precipitation and dissolution (δCCaCO3, blue) based on nδA _T, and CO₂-gas flux (δCFCO2, orange). The change was estimated from the start (21 March 2010, day 1) to the end of the study period (2 April 2010, day 13). Positive change denotes a gain in carbon, and negative change a loss of carbon from the ice horizon.

(3)

while nδA _T is mainly affected by CaCO₃ precipitation and dissolution.

To calculate the net effect of biological processes, we used the change in salinity-normalized nitrate concentrations (nδ[NO_3-]) converted to carbon equivalents using the stoichiometric carbon and nitrogen ratios of 106.16 after Redfield and others (1963). We found significant changes in nδ[NO₃–] in the top 0.05 m and in the top 0.15 m young-ice horizons during the study period. In the interior ice (at 0.15 m), we found a net nδ[NO₃–] decrease of 0.9 mmol m^–3corresponding to a carbon uptake of 6 μmolkg–¹ due to primary production (δC_PP; Fig. 6). The fact that we observed photosynthetic activity with significant algae abundances (7 x 10⁴ and 8.8 x 10⁵ cells mL–¹) implies that such organisms were partly responsible for the variations found in nδ[NO₃–] and young ice [CO₂] after 25 March (Fig. 5c). The yield was highest in the bottom ice. At the same time as we found a net carbon uptake, we also found the highest measured BCP of 9.1 μgCL–¹ d–¹ in the interior of the ice (maximum on 29 March 2010), corresponding to an increase in inorganic carbon of 0.8 μmol L–¹ d–¹. Denitrification by bacteria could also explain the n[NO₃ ^–] decrease (Rysgaard and Glud, 2004). In the top 0.05 m we found a small but significant net increase in nδ[NO₃–] of 1 mmol m^–3corresponding to a C _T increase of 7μmolkg–¹likely due to BCP (δC_BCP; Fig. 6).

The sum of the effects of CaCO₃ formation and dissolution and CO₂ gas flux was obtained by calculating the difference between nδCT and the values from biological processes:

(4)

Based on the correction of nδC_T due to biological processes (nδCbio), there was nonetheless a net C _T loss at the top 0.05 m ice horizon and a larger gain at the 0.15 m ice horizon and at all other horizons (Fig. 6). A similar pattern was found for nδAT (Fig. 6). Between the interior ice and the basal ice, nδA _T increased by 40–55 μmol kg–¹mostly explained by the effect of CaCO₃. We estimated the net CO₂ gas flux

(5)

resulting in a negative CO₂ gas flux (loss of 20 μmolkg–¹ melted ice) at the top 0.05 m ice horizon and a positive CO₂ gas flux (gain of maximum 10μmolkg–¹) in the interior ice and in the bottom ice (Fig. 6). The CO₂ loss from the 0.05 m of top ice to the atmosphere was estimated as ∼1 mmol m^–2 (or 12 mg C m^–2) or 0.08 mmol m^{– 2} d–¹ melted ice.

Frost flowers and ice-air CO₂ gas flux

On 26 March the weather conditions were cold with little wind (wind speed of 0.9 ms–¹) and an air temperature of -15.6°C. The relative humidity was the highest during the study (73%; Table 3). New thin sea ice and new frost flowers were formed further out in the bay, 200 m from our study site. Elevated concentrations of ∆_T (3421 μmol kg–¹) and C _T (3005 μmol kg–¹) were observed in the new frost-flower melt, which were substantially higher than the UIW concentrations but less than in the brine (Table 2). The lower C _T in frost-flower melt compared to brine C _T means that there was a loss of C _T (i.e. CO₂) probably before our first sampling. When the frost flowers from the same location were sampled 20 hours later, ∼25% (from 12 to 9μmolkg–¹) of the CO₂ content in the frost-flower melt had disappeared. The salinity in the frost flowers had changed by 1 after 20 hours, supporting the fact that processes other than salinity changes were affecting the CO₂ concentrations and C _T (e.g. air-ice CO₂ flux). Similar to salinity, ∆_T did not change significantly, but [CO₃ ²–] in the melted frost flowers increased (Table 3), which suggests influence of CO₂ loss. Additionally, C _T in the frost-flower melt decreased from 3005 μmol kg–¹ to 2921 μmolkg–¹resulting in a C _T loss of 84 μmolkg–¹ d–¹likely due to CO₂ loss to the atmosphere (Table 3). The BCP was 2.8 μgCL^–1 d–¹ (i.e. 0.23 μmol kg–¹ d–¹) in the 1 day old frost-flower melt, which was insignificant relative to the uncertainty in the C _T analysis. Using salinity-normalized (S = 52) C _T for both days during the frost-flower melt, the resulting CO₂ loss was 142 μmol kg–¹ d–¹ (7mmol m^{– 2} d–¹). After frost-flower formation, the life span of the frost flowers at the sampling site was ∼48 hours.

When [CO₂] in brine was high, the CO₂ loss through frost flowers could have an impact on the atmospheric CO₂ concentrations (mainly near the ice surface). Since CO₂ easily escapes from the ice when the ice is porous enough (BV>0.05), the frost flowers form an efficient transport path for CO₂ from the ice to the atmosphere (e.g. Fransson and others, 2013; Geilfus and others, 2013). We used the formulation by Fransson and others (2013) to estimate the C _T loss and CO₂ outflow from sea ice to atmosphere through frost flowers and brine of Arctic sea ice:

(6)

(7)

where S _FF and Sb_ri_ne are the salinity of frost-flower melt (average S = 51) and brine (S=121), respectively. C _TFF and C _Tbrine are the C _T in frost-flower melt (C _T = 3000 μmol kg–¹) and brine (C _T = 9054 μmol kg^-1). This gives a C _Tcal_c of 3816 μmol kg-1), resulting in a maximum CO₂ loss (C_Tout) of 816 μmol kg^–1 d–¹ (∼40 mmol m^–2 d–¹) from the ice to the atmosphere, assuming no biological production. The salinity-normalized (S=121) nutrient ([PO₄ ³–], [NO₃–], [Si(OH)₄)]) concentrations in new frost flowers were the same as those in brine (Table 5), which supports our assumption that, in our study, no CO₂ was consumed associated with primary production. Moreover, since the fCO2 values in both brine and frost-flower melt (Table 2 and 3) were much larger than the atmospheric fCO2 value of ∼394 μatm (39.9 Pa; Ny-lesund 2010), frost flowers show a great potential to act as a CO₂ source to the atmosphere.

Discussion

We found that the C _T loss from frost flowers was due to a CO₂ loss of 40 mmol melted sample mediated by upward-transported brine and by frost-flower formation. Our estimates were similar to the findings by Geilfus and others (2013), which estimated a CO₂ loss from the ice to the atmosphere of 20-40 mmol m^{– 2} d–¹ melted sample. These estimates indicate the efficient transfer of CO₂ by frost flowers and a large potential for CO₂ loss during a short period. Moreover, the mean calculated fCO₂ in brine was 10 560 μatm (1070 Pa), which was several times higher than the atmospheric values of ∼400 μatm (40.5 Pa), supporting a large potential for a CO₂ loss from the ice due to ice dynamics. Using our estimates of CO₂ loss by the frost flowers of 0.8 molm^{– 3} melted sample (800 mmol m^–3) and assuming that the frost flowers formed at least on one day (24 hour life cycle) on top of all newly formed spring ice in the Arctic Ocean (area of 14 x 10⁶km^– 2 (US National Snow and Ice Data Center, mean sea-ice covered area in April 2010)), we obtain a total CO₂ gas flux to the atmosphere of 0.2 x101 to 1.1 x 1013 mol C (0.02-0.13 PgC) in one day. This estimate can be compared to the summer net CO₂ uptake of 0.1 Pg C a^–1 in the Arctic Ocean based on Schuster and others (2013). However, although it is likely that this is an overestimation, it strongly supports the fact that frost flowers play a large role in the CO₂ outgassing from the sea water, and that the role of frost flowers in CO₂ air–ice flux needs further investigation.

To further investigate the variability of nA _T (and nC _T) in the young ice, we used the ratios between ∆ _T and salinity (∆ _T .S) as proxies for CaCO₃ precipitation in sea ice (e.g. Fransson and others, 2011, 2013; Rysgaard and others, 2012). Increasing ∆ _T .S means fractionation of ∆ _T and salinity, likely indicating the presence and dissolution of solid CaCO₃. Except for in the top 0.05 m, the ∆ _T : S ratio in melted young ice increased (Table 2). Based on the ∆ _T .S-ratio development during the study (Fig. 7a), we distinguished a pattern in the variability of the ice horizons, which may be due to changes in ∆ _T relative to salinity or transport of CaCO₃ to underlying ice horizons during ice formation (Fig. 7a). Moreover, we found that the ∆ _T : S variability in the top 0.05 m ice corresponded to changes in air temperature. The lowest air temperature was recorded on 27, 29 and 30 March, shifting to warmer air on 30–31 March (Table 1). The increased air and young-ice temperatures caused an increase in BV, allowing chemical substances to move through the ice in brine channels.

Fig. 7. The variability of (a) the AT.salinity ratio (∆T : S) and (b) the CT.salinity ratio (CT: S) in the top 0.05 m (filled diamonds), top 0.10 m (open circles) and median sea ice (crosses) for the whole study.

Conclusion

In this wintertime study of growing young ice, we found changes in the sea-ice CO₂ system at all ice horizons. At the ice/air interface (top 0.05 m), we found that nC _T decreased in the ice due to CO₂ loss to the atmosphere during our study and that nC _T and nA _T increased at underlying horizons. The ice–air CO₂ gas flux was partly driven by the process of CaCO₃ precipitation within the ice, as was also concluded by Geilfus and others (2013). The amount of solid CaCO₃ estimated for the 13 day study was smaller or in the same order of magnitude as found by Fransson and others (2013) and Rysgaard and others (2013). Biological processes were active only to a limited extent in the young ice during the study but imposed local effects on the sea-ice CO₂ system.

Our study confirms previous results that frost flowers are important conveyors for CO₂ gas exchange from ice to atmosphere during ice formation. Assuming that frost flowers will form on newly formed sea ice, this source of CO₂ to the atmosphere would be similar to the net CO₂ uptake in the Arctic Ocean (Schuster and others, 2013). Our study implies that CO₂ loss from frost flowers may be more important for the air–ice–sea CO₂ transfer for the Arctic Ocean than previously reported.