Introduction

Sea ice has previously been considered to be an impermeable barrier to gas exchange, and global climate models do not included CO2 exchange with the oceans during periods of closed ice cover (Tison et al. 2002; Loose et al. 2011; Rysgaard et al. 2011). However, recent observations of gas exchange using both tower-based micrometeorological approaches and chamber sampling indicate that uptake and degassing of CO2 does occur over sea ice (Semiletov et al. 2004; Zemmelink et al. 2006; Delille et al. 2007; Nomura et al. 2010; Miller et al. 2011; Papakyriakou and Miller 2011; Geilfus et al. 2012). In addition, the recognition that there are extensive and active microorganism communities within sea ice, and the finding of the sea-ice-driven carbon pump (Rysgaard et al. 2007), has changed this general perception. Sea ice is now considered to be an active component in the biogeochemical cycling of carbon in ice-covered waters.

During winter, as sea ice grows, salts are partly rejected from the sea ice and partly trapped within the sea ice structure, concentrated into brine pockets, tubes, and channels (Weeks and Ackley 1986; Petrich and Eicken 2010). As the temperature decreases, brine salinity and concentration of solutes and gases increase in brine, and the resulting brine pCO2 is furthermore, strongly influenced by brine drainage and CaCO3 precipitation (Papadimitriou et al. 2007, 2012; Rysgaard et al. 2011, 2012).

Brine drainage from sea ice causes the formation of highly saline dense cold water that sinks to deeper layers and contributes to the global ocean circulation. Furthermore, brine drainage may also change the pCO2 concentrations below sea ice by releasing dissolved gasses and solutes into the water column (Gibson and Trull 1999; Semiletov et al. 2004, 2007; Delille 2006). CaCO3 precipitation can occur in sea ice due to the changes in the mineral-liquid thermodynamic equilibrium (Marion 2001). Ikaite, a hexahydrate polymorph of calcium carbonate (CaCO3·6H2O), begins to precipitate at −2.2 °C and has been found in both Antarctic and Arctic sea ice (Dieckmann et al. 2008, 2010; Rysgaard et al. 2012, 2013; Fischer et al. 2013; Geilfus et al. 2013). CO2 is expelled more efficiently from the ice than total alkalinity (TA), because alkalinity is trapped in ikaite crystals within the interstices between the ice crystals and sea ice becomes enriched in TA (Rysgaard et al. 2013). Studies in the Arctic suggest that, due to brine drainage, CO2 and TCO2 can be transported below the pycnocline and, subsequently, incorporated into intermediate and deep-water masses (Rysgaard et al. 2007, 2011). A combination of the transport of CO2, TCO2, and meltwater from sea ice will lead to a decrease in surface pCO2 with a corresponding increase in the air–sea CO2 flux.

During spring, when the sea ice melts, dissolution of CaCO3 (Nedashkovsky et al. 2009), autotrophic assimilation of CO2 (Søgaard et al. 2010), and dilution of brine by melting sea ice are all processes that can decrease the pCO2 of the brines and, ultimately, of the surface waters (Geilfus et al. 2012). This will result in a lowering of the surface seawater CO2, thereby causing an increase in the air–sea flux of CO2 (Rysgaard et al. 2011, 2012). However, the significance of CaCO3 precipitation/dissolution in sea ice on the air–sea flux of CO2 depends on the sea ice permeability, the timing and location of CaCO3 precipitation, the rate of CaCO3 precipitation, and the fate of the CaCO3 (Delille 2006).

Another potential process in sea ice that could counteract an atmospheric CO2 draw-down is heterotrophic respiration releasing CO2 (Deming 2010; Søgaard et al. 2010).

The understanding of sea ice CO2 dynamics and whether or not the polar regions are, or will be, a source or a sink for CO2 exchange is of fundamental importance for understanding global air–ocean CO2 dynamics. Several studies have focused on the spatial heterogeneity of sea ice algae and/or bacteria, as well as key biogeochemical parameters (Gosselin et al. 1986; Eicken et al. 1991; Rysgaard et al. 2001; Granskog et al. 2005; Steffens et al. 2006; Mikkelsen et al. 2008; Søgaard et al. 2010; Fischer et al. 2013). However, to date, few attempts have been made to investigate the spatial and temporal variability of CaCO3 and the relationship between CaCO3 precipitation and other biogeochemical parameters.

Factors affecting the precipitation, growth, and dissolution of CaCO3 crystals in natural sea ice are still poorly understood (Rysgaard et al. 2012). However, it has been shown that orthophosphate can inhibit the crystallization of the anhydrous forms of CaCO3 (Bischoff et al. 1993; Lin and Singer 2006) but does not interact with ikaite (Bischoff et al. 1993; Buchardt et al. 2001). Furthermore, high concentrations of dissolved organic matter (DOM) can inhibit CaCO3 precipitation in natural environments (Berner et al. 1978; Zullig and Morse 1988). Since sea ice is known to have high concentrations of both DOM and phosphate, especially associated with high biological activity (Thomas et al. 2010), it is pertinent to study these together with CaCO3 dynamics in ice.

The objectives of this study were (1) to investigate the factors that control the spatial and temporal distribution of CaCO3, dissolved organic carbon and nitrogen (DOC, DON), TCO2, TA, inorganic nutrients, bulk salinity, and temperature as well as primary and bacterial productivity in subarctic first-year sea ice, (2) to discuss the potentially synergy between these parameters, and (3) to understand how they can affect the sea ice CO2 system.

Materials and methods

Sampling was conducted from February 17 to May 1, 2010, on first-year fast ice in Kapisigdlit Bight (60 km to the SW of Nuuk), SW Greenland (64°26′N 50°13′W; Fig. 1). During this period, sea ice thickness gradually reduced from the maximum thickness of 81 cm until it completely melted. A maximum snow thickness of 7 cm was measured in March. The water column in the Kapisigdlit Bight was fully mixed and the salinity under the sea ice varied between 33 and 33.5. The water depth under the sea ice was 30–40 meter. More details on the oceanographic conditions in the Godthåbsfjord are provided by Mortensen et al. (2011).

Fig. 1
figure 1

Map showing the sea ice area in Kapisigdlit in SW Greenland and the temporal development and spatial variability sampling stations

Two sampling designs were used:

  1. 1.

    Spatial variability was investigated over a perpendicular xy transect covering 0.07 km2 that was investigated on March 11 and April 8, 2010 (See Fig. 1 for location of these). The parameters sampled in this part of the work were CaCO3, dissolved organic carbon and nitrogen (DOC, DON), inorganic nutrients (PO4 3−, Si(OH)4, NO2 , NO3 , and NH4 +), temperature, salinity, and snow and sea ice thickness.

  2. 2.

    The temporal development of CaCO3, TCO2, TA, inorganic nutrients, DOC, DON, primary production, bacteria production, bulk salinity, snow and sea ice thickness, and temperature was also investigated at a single location in the study area (Fig. 1) every 2–3 weeks from February to May 2010 (i.e. 17 February, 10, 11, 12 and 15 March, 8 April and 1 May).

Spatial variability

In both of the transect surveys (i.e., March 11 and April 8, 2010), 25 sea ice cores were taken along a 266-m-long perpendicular xy transect. The cores were collected at distances of 0, 0.2, 0.4, 0.6, 0.8, 1, 3, 9, 20, 42, 64, 128, and 266 m in both x and y directions. At each sampling point, a sea ice core was collected using a MARK II coring system (Kovacs Enterprises Ltd) and an overlying snow sample was collected using a small shovel. The air temperature was measured 2 m above the snow, and vertical profiles of temperatures within the ice were measured using a calibrated thermometer (Testo®). Sea ice temperature measurements were complemented by a custom-built string of thermistors that were frozen into the sea ice from 11 March until 8 April. The thermistor string data were recorded every 6 h at a spatial resolution of 4 cm.

The retrieved ice cores were cut into 12 cm sections with a stainless steel saw and placed in plastic containers and transported back to the laboratory in dark thermo-insulated boxes. Sea ice and snow samples were slowly melted in the dark at 3 ± 1 °C, which took between 2 and 3 days. We measured all the parameters in all sea ice sections. However, we only report data from the top and bottom section as they are the most important.

To determine the amount of CaCO3 within the ice and snow, between 300 and 500 ml of melted ice or snow was divided into three subsamples and filtered (3 ± 1 °C) through pre-combusted (450 °C, 6 h) Whatman® GF/F filters. The exact volume of the filtered meltwater was measured. The filters were transferred to tubes (12 ml Exetainer®) containing 20 μl HgCl2 (5 % w/v, saturated solution) to avoid microbial activity during storage and 12 ml deionized water with a known TCO2 concentration. The tubes were then spiked with 300 μl of 8.5 % phosphoric acid to convert CaCO3 on the filters to CO2, and after coulometric analysis (Johnson et al. 1993) of CO2 the CaCO3 concentration was calculated.

At each sampling occasion, samples from different ice depth horizons were inspected under the microscope to check for the presence of microorganisms with calcium carbonate external structures such as coccolithophores and foraminifers. These inspections showed that no microorganisms with calcium carbonate external structures were present in any of the samples.

The remaining meltwater was filtered through 25-mm Whatman® GD/X disposable syringe filters (pore size 0.45 μm). A subsample of the filtrate was transferred to pre-combusted glass vials; 100 μl of 85 % phosphoric acid was added; and the vials were frozen for later analyses of DOC. The remaining filtrate was transferred to pre-combusted (450 °C, 6 h) and alkali-washed glass vials and frozen for later analysis of DON, PO4 3−, NO3 , NO2 , Si(OH)4, and NH4 +. The DOC, DON, and nutrient samples were frozen at −19 °C until analysis. DOC was measured by high-temperature catalytic oxidation, using a MQ 1001 TOC Analyzer (Qian and Mopper 1996). The concentrations of NO3, NO2 , PO4 3−, and Si(OH)4 were determined by standard colorimetric methods (Grasshoff et al. 1983) as adapted for flow injection analysis (FIA) on a LACHAT Instruments Quick-Chem 8000 autoanalyzer (Hales et al. 2004). The PO4 3− samples from the first sampling was contaminated, and therefore we did not include them. The concentration of NH4 + was determined with the fluorometric method of Holmes et al. (1999) using a HITACHI F2000 fluorescence spectrophotometer. The concentration of DON was determined by subtraction of the concentration of DIN (DIN = [NO3 ] + [NO2 ] + [NH4 +]) from that of the total dissolved nitrogen determined by FIA on the LACHAT autoanalyzer, using online peroxodisulphate oxidation coupled with UV radiation at pH 9.0 and 100 °C (Kroon 1993).

The conductivity of the melted sea ice sections was measured (Thermo Orion 3-star with an Orion 013610MD conductivity cell), and values were converted to bulk salinity (Grasshoff et al. 1983). The brine volumes of the original sea ice samples were calculated from the measured bulk salinity and temperature and a fixed density of 0.917 g cm−3 according to Leppäranta and Manninen (1988) for temperatures >−2 °C and according to Cox and Weeks (1983) for temperatures <−2 °C.

Spatial autocorrelation (Legendre and Legendre 1998) was used to analyze the correlation of the horizontal and vertical distribution of CaCO3 concentration, DOC, DON, inorganic nutrients, temperature, and salinity as well as the snow and sea ice thickness. Autocorrelation was estimated by Moran’s I coefficients (Moran 1950; Legendre and Legendre 1998). This coefficient was calculated for each of the following intervals along the transect (classes of distance): 0–0.25, 0.25–0.50, 0.50–1.5, 1.5–2.5, 2.5–5.0, 5.0–10, 10–50, 100–150, 150–200, 200–250, and >250 m. The autocorrelation coefficients estimated by the Moran’s I coefficient were tested for significance according to the method described in Legendre and Legendre (1998). A 2-tailed test of significance was used. Positive (+) indicates positive autocorrelation (correlation) and negative (−) indicates negative autocorrelation. A zero (0) value indicates a random spatial pattern. We applied a significance level of P < 0.05. Pearson’s correlation was used to find the correlation between the parameters. Furthermore, a full factorial generalized linear model (GLM) including time, depth, and position as explanatory variables, which was reduced based on Akaike’s Information Criterion (AIC), was applied. The same model was applied for several dependent variables: position (horizontal), depth (vertical), and time on CaCO3 concentration, bulk salinity, DOC, and DON.

Temporal development

On each of the 7 sampling occasions for the temporal study, triplicate ice cores and environmental parameters were collected from a defined area (5 m2), and samples were processed as described above.

Primary production was measured (Søgaard et al. 2010) on 4 occasions (i.e.17 February, 11 March, 8 April, and 1 May). In short, primary production was determined on melted sea ice samples (melted within 48 h in the dark at 3 ± 1 °C). The potential primary production in the sea ice at different sea ice depths (i.e. 12 cm sections) was measured in the laboratory cold room at 3 irradiances (72, 50,14 μmol photons m−2 s−1) and corrected with one dark incubation, using the H14CO3 incubation technique (incubation time was 5 h). The potential primary production measured in the laboratory at different sea ice depths was plotted against the three laboratory light intensities 42, 21, and 9 μmol photon m−2 s−1 and fitted to the following function described by Platt et al. (1980)

$${\text{PP}}(\upmu {\text{g}}\;{\text{C}}\;{\text{l}}^{ - 1} \;{\text{h}}^{ - 1} ) = P_{\rm m} \left[ {1 - \exp \left( {\frac{{ - \upalpha E_{\rm PAR} }}{{P_{\rm m} }}} \right)} \right]$$
(1)

where PP is the primary production, P m (μg C l−1 h−1) is the maximum photosynthetic rate at light saturation, α (μg C m2 s μmol photons−1 l−1 h−1) is the initial slope of the light curve, and E PAR(μmol photons m−2 s−1) is the laboratory irradiance. The photoadaptation index, E k (μmol photons m−2 s−1), was calculated as P m/α.

In situ down-welling irradiance was measured at ground level (Kipp & Zonen pyrometer, CM21, spectrum range of 305–2,800 nm) once every 5 min, and hourly averages were provided by Asiaq (Greenland Survey). Hourly down-welling irradiance was converted into hourly photosynthetically active radiation (PAR; light spectrum 300–700 nm) after intercalibration (R 2 = 0.99, P < 0.001, n = 133) with a Li-Cor quantum 2 pi sensor connected to a LI-1400 data logger (Li-Cor Biosciences®). The in situ hourly PAR irradiance was calculated at different depths, depending on sea ice and snow thickness, using the attenuation coefficients measured during the sea ice season.

In situ primary production was calculated for each hour at different sea ice depths using hourly in situ PAR irradiance (Eq. 1). Total daily (24 h) in situ primary production was calculated as the sum of hourly in situ primary production for each depth. The depth-integrated net primary production was calculated using trapezoid integration.

Light attenuation of the sea ice samples was measured with a Li-Cor quantum 2 pi sensor connected to a LI-1400 data logger (Li-Cor Biosciences®) in a dark, temperature-regulated room (at in situ temperatures to avoid brine loss) using a fiber lamp with a spectrum close to natural sunlight (15 V, 150 W, fiber-optic tungsten–halogen bulb). The sensor was placed under the sea ice section and the fiber lamp was placed above. Light attenuation was measured in this way for each sea ice section. We assume depth-independent attenuation in the sea ice. To measure light attenuation of the snow cover, we gently removed the snow and placed the sensor on the ice surface and then we placed the snow on top of the sensor. Thus, down-welling irradiance was measured directly above and below the snow (Søgaard et al. 2010).

The bacterial production procedures employed (i.e., 17 February, 11 March, 8 April, and 1 May) have been described by Søgaard et al. (2010), except those between 13 and 16 March, when the measurements were made using an ice-crushing method described by Kaartokallio (2004) and Kaartokallio et al. (2007). The two methods used for bacterial production measurements yield comparable results (the mean values from March using the ice-crushing method were 2.4 μg C l−1 day−1 and the mean values from April using the melting sea ice approach were 2.5 μg C l−1 day−1).

Bacterial production in melted sea ice samples was determined by measuring the incorporation of [3H] thymidine into DNA. Triplicate samples (volume = 0.01 L) were incubated in darkness at 3 ± 1 °C with 10 nmol l−1 of labeled [3H] thymidine (New England Nuclear®, specific activity 10.1 Ci mmol−1). Trichloroacetic acid (TCA)-treated controls were made to measure the abiotic adsorption. At the end of incubation period (T = 6 h), 1 ml of 50 % cold TCA was added to all the samples. The samples were filtered and counted using a liquid scintillation analyzer (TricCarb 2800, PerkinElmer®).

For the ice-crushing method, samples were prepared by crushing each intact 5 to 10 cm ice core section, first using a spike tool, and then grinding ice chunks in an electrical ice cube crusher. Approximately 10 ml of crushed ice was placed in a scintillation vial and weighed. To ensure even distribution of labeled substrate, 2–4 ml of sterile-filtered (0.2 μm minisart filters, Sartorius®) seawater was added to the scintillation vials. All ice-processing work was done in a cold on-deck laboratory at near-zero temperature. Two aliquots and a formaldehyde-killed absorption blank were amended with [methyl-3H] thymidine (New England Nuclear®; specific activity 20 Ci mmol−1). Concentrations of 20 nmol l−1 for thymidine were used for all samples. Samples were incubated in the dark at −0.2 °C in a seawater/ice-crush bath for 17–18 h and incubation stopped with the addition of 200 μl of 37 % sterile-filtered formaldehyde. Samples were processed using standard cold-TCA extraction and filtration procedure (using Advantec® MFS 0.2 μm MCE filters). A Wallac Win-Spectral 1414 counter (PerkinElmer®) and InstaGel (PerkinElmer®) cocktail were used for scintillation counting.

For both methods, the bacterial carbon production was calculated, using the conversion factors presented in Smith and Clement (1990).

Bacterial carbon demand (BCD) for growth was calculated as:

$${\text{BCD}}(\upmu {\text{g}}\;{\text{C}}\;{\text{l}}^{ - 1} \;{\text{h}}^{ - 1} ) = \frac{\text{BP}}{\text{BGE}}$$
(2)

where BP is the bacteria production and BGE is a bacterial growth efficiency estimate of 0.5 measured in polar oceans (Rivkin and Legendre 2001).

To investigate the temporal distribution of TCO2 and TA, an additional sea ice core was collected on each sampling occasion. The core was cut into 12-cm sections and placed in laminated transparent NEN/PE plastic bags (Hansen et al. 2000) fitted with a gas-tight Tygon tube and a valve for sampling. These sections were brought back to the laboratory cold room (3 ± 1 °C). Cold (1 °C) deionized water of known weight and TA and TCO2 concentration was added (10–30 ml) to each NEN/PE bag (Hansen et al. 2000). The bags were closed, and excess air quickly extracted through the valve. Then, the ice was melted (<48 h) in the dark. Gas bubbles released from the melting sea ice were transferred to tubes (12 ml Exetainer ®). Sea ice meltwater was likewise transferred to similar tubes containing 20 μl HgCl2 (5 % w/v saturated solution; Rysgaard and Glud 2004). Standard methods of analysis were used: TCO2 concentrations were measured on a coulometer, TA by potentiometric titration (Haraldsson et al. 1997), and gaseous O2, CO2, N2 by gas chromatography (SRI 8610C; FID/TCD detector; Lee et al. 2005).

A Wilcoxon rank sum test was used to test whether CaCO3 concentration was significantly differently distributed within the sea ice. We applied a significance level of 95 %.

Following the bulk determination of TCO2 and TA, the bulk pCO2 and pH (on the total scale) were computed using the temperature and salinity conditions in the field and a standard set of carbonate system equations (See Rysgaard et al. 2013), excluding nutrients, with the CO2SYS program of Lewis and Wallace (2012). We used the equilibrium constants of Mehrbach et al. (1973), refitted by Dickson and Millero (1987, 1989). We assumed a conservative behavior of CO2 dissociation constants at subzero temperatures since Marion (2001) and Delille et al. (2007) suggested that a thermodynamic constant relevant for the carbonate system can be assumed to be valid at subzero temperatures.

Results

Figure 1 shows a map of the investigated sea ice area with the different sampling stations in Kapisigdlit, SW Greenland (Fig. 1).

The air temperature during the study period ranged from −17 °C in February to +16 °C in May just before the sea ice break-up (Fig. 2). Temperatures within the snow and sea ice varied from −6.0 ± 0.1 to 0 ± 0.02 °C, with minimum temperatures measured in February and March, followed by a gradual increase to maximum values in late April (Fig. 3a).

Fig. 2
figure 2

Meteorological data on air temperature (°C) at the Asiaq meteorological station in Kapisigdlit

Fig. 3
figure 3

Temporal development in (a) sea ice and snow temperature [°C] N.B. The temperature data collected from retrieved ice cores are supplemented by thermistor string data from 11 March until 8 April, (b) bulk salinity, (c) relative brine volume fraction [%]. The black dots represent triplicate measurements

The bulk salinity of the sea ice samples varied from 1.7 to 6 (Fig. 3b). The brine volume varied from 5 to 32 % at the top of the sea ice and from 12 to 40 % at the bottom of the sea ice (Fig. 3c), an indication that the ice was permeable for most of the study (Golden et al. 1998). In April, when air temperatures varied between −2 and +16 °C, the ice began to melt, which resulted in high relative brine volumes and low bulk salinities (Fig. 3b, c).

Spatial variability

Moran’s I (Table 1) was used to estimate the spatial autocorrelation within the datasets collected for the two transect samplings in March and April. All parameters had a random spatial distribution pattern with no apparent patches, indicating that the distribution of the parameters investigated was highly heterogeneous on the scale of meters to hundreds of meters (Table 1).

Table 1 Moran’s I as a function of distance class (m) between sites for CaCO3, bulk salinity, DOC, DON, NH4 +, NO2 , NO3 , Si(OH)4 and PO4 3− in snow (S), top sea ice (T) and bottom sea ice (B). First sampling period was on March 11 and second sampling period was on April 8, 2010. Positive (+) indicates positive autocorrelation (correlation) or negative (−) negative autocorrelation (dispersion). A zero value indicates a random spatial pattern

Despite this heterogeneity, there was evidence of a correlation between several parameters: CaCO3 had significant correlation with PO4 3−, bulk salinity, Si(OH)4 and NO3 . CaCO3 and DON were significantly correlated (negative) only during the second sampling. There was no correlation between CaCO3 and DOC, NH4 +, NO2 , and temperature (Table 2).

Table 2 Pairwise comparison of CaCO3, bulk salinity, temperature, DOC, DON, NH4 +, NO2 , NO3 , Si(OH)4 and PO4 3−. (−) indicates significantly negative (−) and (+) indicates significantly positive (+) correlated variables (Pearsons) at a 5 % significance level. First sampling period was in March and second sampling period in April 2010

A GLM where depth (vertical), time, and position (horizontal) as explanatory variables is used to test whether there was a significant effect on several dependent variables: CaCO3, bulk salinity, DOC, and DON. The GLM test was applied on the datasets collected for the two transect samplings in March and April. There was a significant effect of depth for all the parameters. The significant effect of depth was largely dependent on the time of sampling for CaCO3 (F 5,261 = 15.9784, P < 0.001), bulk salinity (F 5,248 = 5.6322 P < 0.001), and DON (F 5,239 = 3.6861 P < 0.001) (data not shown). Furthermore, when the two studies in March and April were compared, a significant effect of time was found for bulk salinity, DOC, and DON (Table 3). No significant effect of position (horizontal) was found for CaCO3, bulk salinity, and DON (Table 3).

Table 3 Results from a GLM model where depth, time, and position as explanatory variables were used to test whether there was a significant effect on several dependent variables: CaCO3, bulk salinity, DOC, and DON

Temporal distribution

In the temporal survey, there were no vertical differences in TCO2 and TA (Fig. 4), although the concentrations of TCO2 and TA decreased with time (Fig. 4). The highest TCO2 and TA concentrations were measured in February (293 ± 6.1 and 410 ± 40 μmol kg−1 in melted sea ice), which decreased to 181 ± 1.3 and 190 ± 1.2 μmol kg−1 in melted sea ice in the beginning of May, respectively (Fig. 4). However, high concentrations of TA and TCO2 were also measured in April (Fig. 4). No small-scale variability was observed for TCO2 and TA concentrations.

Fig. 4
figure 4

Temporal development of the vertical concentration profiles of TCO2 (black bars) and TA (gray bars) and the TA:TCO2 ratio (circles) in bulk melted sea ice during the 2010 sea ice season. Note that TCO2 and TA below 60 cm are water column values. Horizontal dotted line represents the sea ice–water column interface. Data points represent treatment mean ± SE (n = 3)

In the water column, the highest TCO2 concentration, of 2,101 ± 7.7 μmol kg−1, was measured in February, which decreased to 2,085 ± 27 μmol kg−1 in May (Fig. 4). The highest TA concentration of 2,253 ± 2.5 μmol kg−1 was measured in May and the lowest of 2,220 ± 2.2 μmol kg−1 in March (Fig. 4).

The average TA:TCO2 ratio within the sea ice was >1 during February (average 1.25) and March (average 1.20), and higher than that in the water column (Fig. 4). The highest TA:TCO2 ratio (1.75) was calculated for the uppermost horizons of the sea ice in April. In April, the average ratio was 1.32, while the value in the beginning of May was 1.1.

The CaCO3 concentrations varied vertically within the ice cores: in February the highest concentration of 2.4 ± 0.4 μmol CaCO3 l−1 was measured in the upper most layer of the sea ice (Wilcoxon rank sum test; P < 0.05; Fig. 5). Conversely, just before ice break-up in early May, the highest CaCO3 concentration of 4.4 ± 0.1 μmol CaCO3 l−1 was measured in the lowermost ice horizon (Wilcoxon rank sum test; P < 0.05; Fig. 5). In March and April, CaCO3 was evenly distributed within the sea ice with an average concentration of 1.8 ± 0.40 μmol CaCO3 l−1 in March and 2.0 ± 0.30 μmol CaCO3 l−1 in April (Wilcoxon rank sum test; P > 0.05). No small-scale variability was observed for CaCO3 concentration in the sea ice (Fig. 5). CaCO3 concentrations in the snow decreased throughout the sea ice season: from 5.2 ± 1.5 μmol CaCO3 l−1 in February to 3.0 ± 1.5 μmol CaCO3 l−1 in April (Fig. 5). Sediment traps were deployed under the sea ice during the study period but no CaCO3 crystals were found (data not shown).

Fig. 5
figure 5

Temporal development of the CaCO3 concentration [μmol CaCO3 l−1] in bulk sea ice and snow in February, March, April, and May 2010

The DOC concentrations increased over time with a maximum concentration of 160 μmol l−1 being measured in May in the bottommost layer of the ice (Fig. 6). DON concentrations did not vary vertically within the sea ice during winter. However, in April and May, when the sea ice began to melt, the DON concentrations increased, and maximum DON values of 15 μmol l−1 were measured in the bottommost part of the sea ice (Fig. 6). The DOC:DON ratios ranged from 5 to 20 (average 12).

Fig. 6
figure 6

Vertical profiles of DOC, DON, primary production, and bacterial carbon demand in bulk sea ice in February to May. N.B. there were no data available on the vertical profiles of DOC and DON in February

The highest volume-specific primary production (bulk) of 25 μg C l−1 day−1 was measured in March in the bottom of the sea ice (Fig. 6). Bacterial carbon demand varied vertically within the sea ice, with maximum values measured in the top and bottom section of the sea ice in March and May (Fig. 6). In February and April, the maximum BCD was measured in the internal sea ice layers. The highest BCD of 5.8 μg C l−1 day−1 was estimated in the bottommost sections of the sea ice in March.

Bulk nutrient concentrations for each sampling date were plotted as a function of bulk salinity and compared with the expected dilution line (Clarke and Ackley 1984). To calculate the dilution line, we used the average nutrient concentration and salinity measured at 17 February, 11 March, 8 April, and 1 May in the water column (i.e. 0–10 m; PO4 3− = 0.94 μmol l−1, Si(OH)4 = 7.0 μmol l−1, NO2  + NO3  = 8.6 μmol l−1, NH4 + = 0.28 μmol l−1, DOC = 62.6 μmol l−1, DON = 1.2 μmol l−1 and a average salinity of 33). If values are below the line, depletion of nutrients has taken place, and if above the dilution line production, or net-deposition, of the solute has occurred. Plots of salinity versus PO4 3−, Si(OH)4, NO2  + NO3 , NH4 +, DOC, and DON in sea ice were generally all above the dilution line implying accumulation of nutrients and organic matter within the ice (Fig. 7a–f). However, depletion of PO4 3− was observed in February and March. Furthermore, NO2  + NO3 was depleted in April and May.

Fig. 7
figure 7

Temporal development of (a) PO4 3−, (b) Si(OH)4, (c) NO2  + NO3 , (d) NH4 +, (e) DOC, and (f) DON concentrations versus bulk salinity in sea ice from February to May. The solid line indicates the expected dilution line predicted from salinity and nutrient concentrations in seawater (0–10 m, average salinity of 33 under the sea ice)

There was a negative correlation between PO4 3− and CaCO3 and Si(OH)4 and CaCO3, while a positive correlation was observed between NO3 and CaCO3 (Table 2).

Figure 8 shows nTA and nTCO2 (TA and TCO2 value being normalized to a salinity of 33 to remove correlation to salinity) relationships in seawater samples and bulk ice samples. The different lines represent the theoretical effects of precipitation–dissolution of CaCO3, CO2 release–uptake and photosynthesis–respiration on the ratio nTCO2:nTA. The precipitation of CaCO3 decreases both TCO2 and TA in a ratio of 2:1. An exchange of CO2 has no impact on TA, while TCO2 will be affected. Biological activity has an almost negligible effect on TA, with a ratio TA:TCO2 = −0.16 (Zeebe and Wolf-Gladrow 2001). Considering that sea ice was formed from seawater with a known TA:TCO2 ratio (Fig. 4), we are able to decipher which process took place in the ice: in February and March, the sea ice samples were aligned on the theoretical line for CaCO3 precipitation (Fig. 8). In April and May, the ice samples were well aligned (slope 0.75; R 2 = 0.86) between the theoretical trend for CaCO3 precipitation/dissolution and the one for CO2 release/uptake (Fig. 8). In April two sea ice samples were aligned on the theoretical line for CaCO3 dissolution. This implies that both CaCO3 precipitation/dissolution and CO2 release/uptake had occurred in the ice.

Fig. 8
figure 8

nTA:nTCO2 (values normalized to a salinity of 33) relationship in seawater samples and bulk ice samples from February to May. The different lines represent the theoretical evolution of TCO2:TA following precipitation/dissolution of calcium carbonate (dashed line), a release or uptake of CO2(g) (dotted line) and impact of biology (solid line)

Discussion

Biological activity

A maximum BCD of 5.8 μg C l−1 day−1 (Fig. 6) was estimated, which is low compared with maximum rates of 27 μg C l−1 day−1 estimated in the neighboring fjord Malene Bight in April 2008 (60 km to the SW from the present study site; Søgaard et al. 2010).

Pairwise correlations between primary production and BCD revealed significant positive relationships. Furthermore, accumulation of DOC was observed (Fig. 6e–f) as well as DOC/DON ratios ranging from 5 to 20 indicating a probable production of carbon-rich extracellular polymeric substances (EPS) by the sea ice algae and bacteria (Underwood et al. 2010; Krembs et al. 2011; Juhl et al. 2011). Correlation between primary production and BCD, high DOC/DON ratio and accumulation of DOC points to there being a low-quality substrate resulting in a low bacteria production. Previous studies have shown that EPS is a low-quality substrate for heterotrophic bacteria (Pomeroy and William 2001), which might explain the low BCD and the observed DOC accumulation (Fig. 6). However, several studies suggest the opposite that EPS serve as high-quality substrate for bacteria (e.g. Junge et al. 2004; Meiners et al. 2008).

Another explanation for the low BCD might be the value for bacterial growth efficiency used. We used a bacterial growth efficiency of 0.50 (Rivkin and Legendre 2001). However, growth efficiency is an inverse function of temperature, and small changes in temperature would influence growth efficiency (~2.5 % decrease in growth efficiency per 1 °C increase) and thereby the BCD (Rivkin and Legendre 2001). Using a lower growth efficiency (<0.15; e.g. Middelboe et al. 2012), the seasonal net autotrophic sea ice would change to a net heterotrophic sea ice, which compares to result found by Long et al. (2011) in Kapisigdlit Bight in March 2010. Thereby the biological activity would not contribute to the atmospheric CO2 uptake at all. However, we believe the use of a growth efficiency of 0.50 to be most valid since it also agrees with the growth efficiency of 0.41 measured by Kuparinen et al. (2011) and Del Giorgio and Cole (1998).

The highest estimates of primary production (Fig. 6) were measured in the bottom ice horizons in March. Average rates of sea ice algal primary production during March (4.03 mg C m−2 day−1) are at the lower end of the scale for values reported from the Arctic (0.2–463 mg C m−2 day−1; Arrigo et al. 2010 and references therein). A decrease in TCO2 was measured in the bottommost ice at the time of the algae growth, suggesting that primary production was responsible for the decrease in TCO2 (Figs. 4, 6). However, the average net biological production in the bottom of the sea ice in March was only 1.6 μmol C day−1 (Fig. 6) and the average TCO2 loss in the bottom of the sea ice in March was 6.4 μmol day−1, indicating that processes other than primary production influence the inorganic carbon dynamics in the sea ice.

TCO2 in sea ice is controlled by primary production and respiration by both autotrophic and heterotrophic organisms, CaCO3 precipitation/dissolution, and CO2 release/uptake. The magnitude of the primary production and thus the potential role of the production in controlling the sea ice inorganic carbon cycle depend primarily on light availability and the size of the inorganic nutrient pool (Papadimitriou et al. 2012). The highest light attenuation coefficient of 12 m−1 was measured in March in the snow, corresponding to coefficients reported in snow cover in both the Arctic and Antarctic (Thomas 1963; Weller and Schwerdtfeger 1967; Søgaard et al. 2010). High snow reflection causes low light conditions in the sea ice. However, the highest primary production was found in the bottom of the sea ice in March (Fig. 6), where the light availability was low compared to spring, indicating that light was not the main factor controlling the primary production. However, the low primary production in the bottom of the sea ice after March suggests that the sea ice algae were nutrient-limited late in the sea ice season. Assuming that nutrient uptake by the ice algae follows the Redfield-Brzezinski ratio of 106C:16N:15Si:1P (from Redfield et al. 1963; Brzezinski 1985), then the nitrogen (N:P ratio <2) and silicate (Si:P ratio <6) appear to have limited the sea ice algal primary production in April and May, while phosphate was found at relatively higher concentrations. This is supported by the nutrient-salinity plot for NO2 +NO3 in Fig. 7, which indicates depletion of NO2 +NO3 at the end of the sea ice season. A further factor known to influence the sea ice algal communities is grazing (Gradinger et al. 1999; Bluhm et al. 2010); however, grazing was not measured during the present study.

CaCO3 precipitation

The observed CaCO3 formation was expected (Anderson and Jones 1985; Marion 2001; Papadimitriou et al. 2004) and is consistent with the elevated TA:TCO2 ratios (Fig. 4). We measured much lower TCO2 and TA concentrations within the sea ice compared to concentrations found in the underlying seawater (average TA:TCO2 ratio = 1.06 in seawater; Fig. 4). Likewise, we found elevated TA:TCO2 ratios in the sea ice with a maximum of 1.75, as compared to the underlying seawater (Fig. 4). However, the TA:TCO2 ratios found in the sea ice in the present study are low compared to values found in other sea ice studies (Rysgaard et al. 2007, 2012; Papadimitriou et al. 2012; Geilfus et al. 2012; Rysgaard et al. 2013).

Ikaite precipitation is favored by near-freezing temperature, alkaline condition, and elevated phosphate concentrations (>5 μmol l−1; Bischoff et al. 1993; Buchardt et al. 2001; Selleck et al. 2007).

In this study, bulk phosphate concentrations between 0.2 and 3.1 μmol l−1 were measured in the sea ice, and therefore, the brine phosphate concentrations were occasionally above 5 μmol l−1. Furthermore, the phosphate concentrations observed in present study were 5–20 times higher than concentrations found in previous studies in the Arctic (e.g. Krembs et al. 2002; Mikkelsen et al. 2008; Søgaard et al. 2010). Sea ice temperatures during our study ranged from −6 to 0 °C, which is the temperature where ikaite will form. Alkalinity condition was also satisfied as a C-shaped pH profile with high pH (>9) in surface, and bottom sea ice layers, and slightly lower pH conditions (8.5) in the internal sea ice layers were calculated for February, using temperature and bulk salinity (Fig. 3), TA and TCO2 concentrations (see “Materials and methods” section; Fig. 4). A C-shaped pH profile was also observed in a recent study on experimental sea ice (Hare et al. 2013).

In the early part of the study, we measured the highest CaCO3 concentrations in the surface of the sea ice, and the concentrations decreased with depth (Fig. 5). This is similar to that described by Geilfus et al. (2013) and Rysgaard et al. (2013).

However, just before the ice break-up, the highest CaCO3 concentration was measured in bottom of the sea ice (Fig. 5). This high CaCO3 concentration might be due to a migration of the crystals through the brine channels network or could result from the effects of surface flooding and/or gravity drainage. The average brine salinity was 67 in February, 39 in March, and 27 in April; thus, gravity drainage was only possible in February and March. A negative correlation between CaCO3 concentration and bulk salinity was evident (Table 2), which also suggests that the highest CaCO3 concentrations were found in the sea ice with low bulk salinities, i.e., in the bottom of the sea ice and/or in summer sea ice.

The concentration of CaCO3 in the sea ice was 2.13 g m−2 (Fig. 5), which compares with values from Antarctic sea ice of 0.3 to 3.0 g m−2 (Dieckmann et al. 2008) and 0.1 to 6.5 g m−2 found in the top 10 cm of Antarctic sea ice (Fischer et al. 2013). However, these values are one order of magnitude lower than measurements in Arctic ice by Rysgaard et al. (2012, 2013).

Assuming that sea ice is formed from surface seawater with a TA:TCO2 ratio of 1, the TA:TCO2 ratio of 1.1–1.75 observed in the sea ice could be generated by a CaCO3 concentration of 25–120 μmol CaCO3 l−1. The calculated CaCO3 concentrations are between 6 and 25 times higher than our measured CaCO3 concentrations. However, the highest calculated CaCO3 concentration compares with surface concentrations of 160–240 μmol CaCO3 l−1 melted sea ice reported in the Fram Strait (Rysgaard et al. 2012). The reason for this difference is not clear but could be associated with the melting procedures used. We assume that ikaite does not dissolve if the temperature is maintained below 4 °C, but ikaite could dissolve according to: \({\text{CaCO}}_{3} \cdot 6{\text{H}}_{2} {\text{O }} + {\text{CO}}_{2} \leftrightarrow {\text{Ca}}^{2 + } + 2{\text{HCO}}_{3}^{ - } + 5{\text{H}}_{2} {\text{O}}\) in the meltwater being in contact with atmospheric CO2 during the melting procedure. Another explanation could be that not all the changes in TA originate from CaCO3 precipitation, and so the TA concentration is higher than CaCO3 concentration.

The snow contained relatively high amounts of CaCO3, which has also been found in Antarctic snow overlying sea ice (Fischer et al. 2013). The mass of CaCO3 found in the snow ranged from 0.12 to 0.30 g m−2 (Fig. 5), which is 7 times lower than the values measured in the sea ice. An explanation for the occurrence of CaCO3 in the snow could be the upward brine expulsion, which builds up a high salinity layer on top of the ice. Subsequently, when temperature decreases, some salts may reach their solubility threshold and begin to precipitate (Geilfus et al. 2013).

The strong horizontal, vertical, and temporal variability in CaCO3 concentration suggests that CaCO3 concentration is influenced by variability in several inherent sea ice properties. We found a strong correlation between CaCO3 concentrations and bulk salinity, and inorganic nutrients (Table 2). Pairwise correlations between all measured parameters revealed a negative relationship between CaCO3 concentration and DON in the second sampling period (Table 2), which may be due to the adverse effect of high DOM concentrations on the CaCO3 precipitation (Berner et al. 1978; Zullig and Morse 1988).

There was no clear correlation between CaCO3 and temperature (Table 2), which was unexpected as it is a key variable controlling CaCO3 release/uptake and the precipitation of CaCO3 (Papadimitriou et al. 2004). However, a similar lack of correlation was also observed in a study on Antarctic sea ice (Papadimitriou et al. 2012). This is probably because the measured temperature was not in equilibrium with the temperature conditions that led to the precipitation in the first place.

There was no effect of position for CaCO3 concentrations, DON, and bulk salinity, indicating that the parameters were homogenously distributed in the sea ice (Table 3). However, Moran’s I indicates that the parameters were highly heterogeneous on the scale of meters to hundreds of meters (Table 1). Fischer et al. (2013) also found high spatial variability for CaCO3 concentration, and high spatial variability has also been observed for other sea ice biochemical and biological properties (Rysgaard et al. 2001; Granskog et al. 2005; Steffens et al. 2006; Søgaard et al. 2010). This clearly emphasizes the importance of seasonal and spatial studies and replicate sampling when collecting sea ice biogeochemical samples.

Effects of CaCO3 precipitation/dissolution and biological activity on CO2 dynamics

The changes in the TA:TCO2 ratio were to a large extent explained by CaCO3 precipitation and CO2 release/uptake from the ice (Fig. 8). This observation compares with studies by Munro et al. (2010), Fransson et al. (2011), Rysgaard et al. (2012), Papadimitriou et al. (2012), and Geilfus et al. (2012), in which most of the depletion in TCO2 was explained by processes other than biological activity. Furthermore, there was no correlation between CaCO3 and primary production or BCD, which suggests that CaCO3 precipitation and/or dissolution is not significantly influenced by microbial activity.

The relative effects of biological activity and precipitation/dissolution of CaCO3 in influencing the air–sea CO2 exchange in sea ice can be estimated from measurements during the entire sea-ice-covered season. Using a carbon-to-Chl a ratio of 20–40 for sea ice algae (Arrigo et al. 2010), an estimate of primary production of 220–540 mg C m−2 (data not shown) was calculated, which was low compared to the integrated CaCO3 concentration of 2,130 mg m−2 (Fig. 5). Furthermore, the integrated gross primary production of 153 mg C m−2 recorded was also relatively low compared to the integrated CaCO3 concentration. Thus, contribution of primary production to TCO2 depletion was minor (7–25 %) compared to the contribution of CaCO3 precipitation (Fig. 8).

The primary production measured in this study might be a conservative estimate. Assuming that between 10 and 61 % of the carbon fixed by photosynthesis was released as DOC (Passow et al. 1994; Chen and Wangersky 1996; Malinsky-Rushansky and Legrand 1996; Gosselin et al. 1997; Riedel et al. 2008), the average primary production in this study of 0.19 μmol C l−1 day−1 is equivalent to a DOC accumulation of 0.02–0.11 μmol l−1 day−1. The average DOC concentration in the sea ice was 75 μmol l−1 (Fig. 6), which is much higher than the expected DOC accumulation calculated from the primary production. If the primary production is a conservative estimate, the contribution to the changes in TCO2 and TA by the net biological production might be higher.

However, a recent study with high primary production showed that only 10–20 % of the changes in TCO2 could be explained by net biological production (Fransson et al. 2011). In another study by Rysgaard et al. (2012), abiotic processes including CaCO3 dissolution accounted for between 88 and 98 % of the air–sea CO2 uptake, indicating that the contribution of primary production was relatively low.

Assuming that all CaCO3 dissolved within the sea ice or in the mixed layer, melting a 0.4-m-thick sea ice cover (temperature: −1.1 °C, bulk salinity: 4.4, TA: 291 μmol kg−1, and TCO2: 238 μmol kg−1) into a 20-m-thick mixed water layer (temperature: 0 °C, salinity: 33, TA: 2,239 μmol kg−1, TCO2: 2,109 μmol kg−1) from Figs. 3 and 4 would result in a 1.5 ppm decrease in pCO2 per week. Assuming that no CaCO3 precipitates (e.g. TA and TCO2 are both 291 μmol kg−1), the resultant pCO2 decrease would be 0.9 ppm per week. Using average conditions during the field campaign (Fig. 4), this corresponds to an air–sea CO2 uptake of 1.5 mmol m−2 sea ice day−1 (with CaCO3) and 1.1 mmol m−2 sea ice day−1 (without CaCO3). Therefore, dissolution of CaCO3 increased the air-sea CO2 uptake by 36 % as compared with CaCO3-free sea ice. Furthermore, the abiotic sea-ice-driven CO2 is higher than the net biological production of 0.07 mmol m−2 day−1 (which accounts for 4 % of CO2 uptake, Fig. 6); and, therefore we conclude that the abiotic processes, including CaCO3 dissolution and CO2 uptake, played a more important role than the net biological production in the sea ice for the air-sea CO2 dynamics during this period.

Comparing the sea-ice-driven CO2 uptake with a recent study by Rysgaard et al. (2012), the uptake found in the present study is sevenfold lower. Furthermore, the amount of CaCO3 found in the present study is low compared to other studies (Dieckmann et al. 2008; Rysgaard et al. 2012; Fischer et al. 2013). Our study was performed in subarctic sea ice exposed to higher seasonal air temperatures than the aforementioned studies, which might explain the lower CaCO3 concentrations and, hence, the lower sea-ice-driven CO2 uptake.

Conclusion

Abiotic processes contributed to an air–sea CO2 uptake of 1.5 mmol m−2 sea ice day−1 and dissolution of CaCO3 increased the air–sea CO2 uptake by 36 % as compared to a theoretical assessment of melting CaCO3-free sea ice. Furthermore, primary production only contributed marginally to TCO2 depletion of the sea ice (7–25 %), which was mainly controlled by physical export via brine drainage and CaCO3 precipitation/dissolution. The net biological production could only explain 4 % of the sea-ice-driven CO2 uptake. These estimates must be considered with the caveat that for all of the parameters investigated, their distribution in the ice was highly heterogeneous. Furthermore, there was considerable temporal variability for all parameters.