Observation sites

Field observations were carried out in the semi-closed system of Saroma-ko, a lagoon on the Okhotsk Sea coast of Hokkaido, Japan, from 18 February to 9 March 2006 and from 22 February to 4 March 2008 (Fig. 1; Table 1). The surface area of the lagoon is 149.2 km² and the mean depth is 14.5 m. Sea ice usually covers almost the entire surface of the lagoon from early January through to early April, with large year-to-year variability (Reference Shirasawa and LeppärantaShirasawa and Leppäranta, 2003). A river-water plume develops under the sea ice in the lagoon as a result of increased discharge of river water during the ice-melt period (Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009). For this study, we chose six ice stations: one (station main) located ∼1.5 km offshore and five others (stations 1–5) forming a transect between the mouth of the Saromabetsu river and the second (east) inlet (Fig. 1; Table 1). For each station, we carried out air–sea-ice CO₂ flux measurements, sampling of sea-ice cores, brine and under-ice water, and ice surface observations. Details of the sampling methods for sea-ice cores, brine and under-ice water, and analytical procedures for physical and chemical properties are given in the Appendix.

Fig. 1. Locations of sampling stations in Saroma-ko, Hokkaido, Japan: stations 1–5 (44°05.30′ −44°07.49′ N, 143°56.59′ −143°56.61′ E) and station main (44°07.18′ N, 143°57.34′ E).

Air–sea-ice CO₂ flux measurements

Air–sea-ice CO₂ flux was measured by a closed dynamic chamber method (Fig. 2). This method was used to directly determine CO₂ flux in situ to allow for high spatial resolution. We used a stainless-steel chamber 50 cm in diameter and 30 cm high with a serrated bottom edge to prevent gas from leaking through the chamber–ice interface. The chamber was pushed down firmly into the sea-ice surface to prevent lateral diffusion of CO₂. Sample air from the chamber passed through Teflon tubes connected to a non-dispersive infrared (NDIR) analyzer (GMP343 carbon dioxide probe; Vaisala Ojy, Vantaa, Finland) that was connected to a data logger (RVR 52 data logger; T&D Corp., Nagano, Japan). Temperatures in the chamber were measured during each flux measurement with a temperature sensor (RVR 52 temperature recorder; T&D Corp., Nagano, Japan). The air in the chamber was circulated with a pump at a flow rate of 0.7 L min⁻¹ through a mass flow controller and chemical desiccant column containing Mg(ClO₄)₂ (Fig. 2). These devices were installed in a portable adiabatic box (40 cm × 25 cm × 30 cm). Electricity was provided by an electric generator. The CO₂ concentration in the chamber was measured every 10 s during experiments lasting 40 min. Four standards (typically 324, 341, 363 and 406 ppm CO₂ in natural air) traceable to the World Meteorological Organization (WMO) mole fraction scale (Reference Yoshikawa-Inoue and IshiiYoshikawa-Inoue and Ishii, 2005) were used to calibrate the CO₂-measuring system in the laboratory before the experiment in 2006 and both before and after the experiment in 2008. The instrument drift (output voltage) before and after the experiment for each calibration gas agreed to well within 1%. The CO₂ flux was calculated from the changes in CO₂ concentration in the chamber and the elapsed time (Fig. 3). The precision of the air–sea-ice CO₂ flux from duplicate determination was within ±0.13 mg C m⁻² h⁻¹.

Fig. 2. Schematic diagram of the CO₂-measuring system and stainless-steel chamber. Solid thick lines indicate the closed loop of flowing air for measuring the concentration of CO₂ in the air in the chamber. CDC: chemical desiccant column; MFC: mass flow controller; NDIR: non-dispersive infrared analyser; DL: data logger; EG: electric.

Fig. 3. Example of the temporal variation in CO₂ concentration in the chamber at station main on 29 February 2008 with the condition depicted in Figure 4b (neither snow nor slush). Open circles and crosses reflect the values obtained before and after the measurement, respectively. The regression line is based on the data obtained after the measurement. The vertical dashed line indicates an elapsed time of 0 min.

During the observation period in 2008, the CO₂ flux was measured under two different conditions: (1) a chamber was installed above the sea ice through the snow cover and slush layer deposited over the sea ice (Fig. 4a); and (2) a chamber was installed directly above the sea ice after first removing the snow cover and slush layer with a scoop (Fig. 4b).

Fig. 4. Schematic diagrams showing two situations for measuring the air–sea-ice CO₂ flux during the 2008 study period in Saromako. (a) The chamber was installed above the sea ice through the snow cover and slush layer deposited over the sea ice. (b) As in (a) but with snow and slush layer removed artificially.

Observations of snow cover, slush layer and sea-ice surface conditions

We examined the properties of the snow cover and slush layer deposited over sea ice and determined the condition of the ice surface by measuring the snow thickness and the thickness of the slush layer. We then classified the state of the sea-ice surface into three categories: (1) snow+slush; (2) superimposed ice; and (3) snow + superimposed ice. The temperatures of the air, the air–snow interface, the snow–slush-layer interface and the slush– or snow–sea-ice interface were measured using needle-type temperature sensors (SK-250WP; Sato Keiryoki Mfg. Co., Ltd, Tokyo, Japan).

Properties of snow cover, slush layer, sea ice and sea-ice surface

The properties of the snow cover, slush layer, sea ice and sea-ice surface conditions during our experiments are shown in Table 1. During the observation period in 2006, there was considerable variation in snow thickness and slush-layer thickness. On 18 and 19 February 2006, there were both snow and slush layers on top of the sea-ice surface (Table 1). However, as air temperature increased, the state of the sea-ice surface changed suddenly. From 23 February to 1 March, because of the melting of the snow and the refreezing of the meltwater, the surface of the sea ice was frequently covered by an icy glaze (herein referred to as ‘superimposed ice’) with neither a snow nor a slush layer (Fig. 5). On 3 March, there was >16 cm of snow on top of the superimposed ice. This loading of snow resulted in an upward flushing of water through brine channels in the ice (Reference Hudier, Ingram and ShirasawaHudier and others, 1995), and a slush layer had formed again at station 3 and station main on 3 March. From 4 to 9 March, the snow melted and the snow thickness decreased with time. During this period, there was no slush layer over the sea ice.

Fig. 5. Photograph of the superimposed ice at station 3 in Saromako on 28 February 2006. The photograph was taken looking west. The length of the standing cores is ∼40 cm.

During 2008 at station main, snow and slush layers were observed throughout the observation period (Table 1). Snow thickness and slush-layer thickness varied according to the amount of snow deposited over the sea ice and the rate of melting.

The sea-ice thickness was almost constant during each experimental period (Table 1). The ice thickness was 42 ± 5 cm (mean ± standard deviation) in 2006 and 45 ± 5 cm in 2008. The sea-ice temperature ranged from −3.1°C to −0.1°C in 2006 and from −2.8°C to −0.9°C in 2008 (Table 1). The temperature of sea ice, snow and slush layer increased overall through the experimental period each year (Table 1). These results suggest a transition from the cold period to the warm or melting period throughout the observation period. The brine volume fraction of sea ice was >5% at all stations (Table 1), indicating that the CO₂ in brine can diffuse through the brine channel (Reference Gosink, Pearson and KelleyGosink and others, 1976) due to the high permeability of sea ice (Reference Golden, Ackley and LytleGolden and others, 1998).

Brine pCO₂

The brine pCO₂ is one of the important factors determining the magnitude of the air–sea-ice CO₂ flux (Reference DelilleDelille, 2006; Reference Nomura, Yoshikawa-Inoue and ToyotaNomura and others, 2006). Here we focus on the distributions of the brine pCO₂ and discuss the mechanisms controlling it.

The brine pCO₂ increased with increasing brine salinity, with a high degree of correlation (r ² = 0.86; P < 0.05) (Fig. 6). During the observation period in 2006, both brine salinity and brine pCO₂ spanned wide ranges: 2.3–29.3 for brine salinity and 2.7–195.1 μatm for brine pCO₂. Plots of low brine salinity and low brine pCO₂ corresponded to values observed near the river mouth (stations 1–3) during the entire period and through the transect during the warm period 1–9 March 2006. At station main during the 2006 observations, both brine pCO₂ and brine salinity decreased drastically (Fig. 7). Brine salinity decreased from 38.2 on 19 February to 8.5 on 7 March. Brine pCO₂ decreased from 195.1 μatm on 23 February to 8.6 μatm on 7 March. Near the river mouth (stations 1 and 3) during the 2006 observations, salinity and pCO₂ of under-ice water at 1 m depth below the surface of the sea ice ranged from 3.3 to 18.8 and from 26.1 to 187.2 μatm, respectively. These values are low compared with those of under-ice water at 1 m depth below the surface of the sea ice at station 5 on 5 March 2006 (20.2 for salinity and 320.5 μatm for pCO₂).

Fig. 6. Relationship between brine salinity and brine pCO₂ during observations in 2006 and 2008. The dashed curve indicates the best fit to describe the relationship (y = 0.22x ^1.88; r ² = 0.86; P < 0.05).

Fig. 7. Changes in brine pCO₂ and brine salinity at station main during observations in 2006.

During observations in 2008, brine salinity and brine pCO₂ at station main stayed within relatively narrow ranges: 23.3–33.2 for brine salinity and 54.9–175.2 μatm for brine pCO₂ (Fig. 6). In all experiments, brine pCO₂ was under-saturated with respect to the overlying air, with an average value of 388.5 ± 6.1 μatm (mean ± standard deviation) for all observations.

In order to elucidate the process controlling the changes of brine pCO₂, we examine the relationships between the salinity-normalized brine dissolved inorganic carbon (n-DIC) and salinity-normalized total alkalinity (n-TA) (Fig. 8). The samples with low salinity and low pCO₂ for brine and under-ice water exhibit extremely high n-DIC and n-TA when dissolved inorganic carbon (DIC) and total alkalinity (TA) are normalized to a salinity of 35 (Fig. 8). The relationships between n-DIC and n-TA for brine and under-ice water show good agreement (r ² = 0.94). The slope of the regression line for brine and under-ice water is 1.3, and plots of brine and under-ice water are slightly deviated above and below the regression line, respectively.

Fig. 8. Relationships between the n-DIC and n-TA (normalized to a constant salinity of 35) for brine and under-ice water during observations in 2006 and 2008. The regression line is based on the brine and under-ice water with slope of 1.3 (r ² = 0.94). The inset indicates the theoretical slope of the different processes affecting n-DIC and n-TA.

In Saroma-ko, a salinity and chemical substances (nutrients) gradient has been observed moving offshore from the river mouth (Reference Shirasawa and LeppärantaShirasawa and Leppäranta, 2003; Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009). Based on the salinity–oxygen-isotopic ratio, the fraction of river water just below the sea ice was estimated to be >80% at stations 2–4 during the warm period of the experiment from 1 to 8 March 2006, whereas the fraction of ice meltwater just below the sea ice was almost zero through the transect during the entire period (Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009). These results suggest that the changes of n-DIC and n-TA for under-ice water shown in Figure 8 were mainly driven by the effect of the river-water plume developed under the sea ice.

Our study area was located in the southernmost part of the seasonal sea-ice zone in the Northern Hemisphere, where the sea ice is relatively thinner, ice temperature is higher and sea-ice brine volume fraction is higher than in polar seas (Table 1). Water exchange between the well-developed brine channels and under-ice water readily occurred through upward flushing caused by loading of snow deposited over the sea ice (Reference Hudier, Ingram and ShirasawaHudier and others, 1995; Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009). Therefore, it can be considered that the changes of n-DIC and n-TA for brine were caused by the exchange with under-ice water characterized by the river-water plume under the ice (Fig. 8). At first glance, the slope of the regression line for brine and under-ice water is also close to the theoretical slope for CaCO₃ dissolution (slope = 2.0) (Fig. 8). However, because the high concentration of dissolved inorganic phosphorus in brine (Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009) and the high ice temperature (Table 1) (compared with the polar seas) impedes the formation of CaCO₃ (Reference Killawee, Fairchild, Tison, Janssens and LorrainKillawee and others, 1998; Reference PapadimitriouPapadimitriou and others, 2005), we suggest that CaCO₃ formation is minimal in sea ice. In addition, in order to demonstrate the wide distributions of n-DIC and n-TA for brine (>10 000 μmol kg⁻¹) (Fig. 8), large quantities of CaCO₃ are needed (Reference DelilleDelille, 2006). The slight difference between the plots for brine (above the regression line in Fig. 8) and under-ice water (below the regression line) might be attributed to other processes such as photosynthesis. Chlorophyll-a concentrations of brine were higher than in under-ice water during our observation period (Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009). This can lead to the photosynthetic activity of ice algae in brine and depletion of the brine pCO₂ relative to the under-ice water.

Reference DelilleDelille (2006) reported that brine pCO₂ changes dramatically from the season of sea-ice formation to the season of melting conditions in the Antarctic pack ice. During sea-ice formation, brine pCO₂ increases with increasing DIC and with changes in CO₂ solubility and the dissociation constants of carbonic acid (Reference Papadimitriou, Kennedy, Dieckmann and ThomasPapadimitriou and others, 2003; Reference Nomura, Yoshikawa-Inoue and ToyotaNomura and others, 2006). Therefore, brine pCO₂ is supersaturated with respect to the air. On the Antarctic pack ice, the brine pCO₂ increases up to 900 μatm in late winter (Reference DelilleDelille, 2006). In contrast, during the ice-melting season, brine pCO₂ becomes undersaturated with respect to the air as a result of the increased net primary production, dilution by melting ice and dissolution of carbonate minerals (Reference Gleitz, van der Loeff, Thomas, Dieckmann and MilleroGleitz and others, 1995; Reference DelilleDelille, 2006). In Arctic fast ice in spring, brine pCO₂ is 220–280 μatm (Reference Semiletov, Makshtas, Akasofu and AndreasSemiletov and others, 2004), and in Antarctic pack ice in summer, brine pCO₂ is about 50 μatm (Reference DelilleDelille, 2006).

Air–sea-ice CO₂ flux

The values of the air–sea-ice CO₂ flux obtained in this study ranged from −1.8 to +0.5 mg C m⁻² h⁻¹ (where negative values indicate a sink for atmospheric CO₂) (Fig. 9). The values agree well with the results obtained over bare Antarctic pack ice (−2.6 to +1.0 mg C m⁻² h⁻¹; Reference DelilleDelille, 2006). However, more negative flux values (i.e. increased CO₂ flux from air to sea or water) were observed over melt ponds and open brine channels in the Arctic fast ice (−19.3 to −9.8 mg C m⁻² h⁻¹; Reference Semiletov, Makshtas, Akasofu and AndreasSemiletov and others, 2004) and over snow and slush layers on Antarctic pack ice (−9.1 to −3.3 mg C m⁻² h⁻¹; Reference Zemmelink, Delille, Tison, Hintsa, Houghton and DaceyZemmelink and others, 2006), where there is a highly productive ice algae community.

Fig. 9. The air–sea-ice CO₂ flux vs brine pCO₂ for all experiments. The horizontal dotted line indicates an air–sea-ice CO₂ flux of 0 mg C m⁻² h⁻¹.

Previous studies report that the air–sea-ice CO₂ flux is mainly influenced by the difference between pCO₂ of the brine in sea ice and that of the overlying air (Reference DelilleDelille, 2006; Reference Nomura, Yoshikawa-Inoue and ToyotaNomura and others, 2006). We saw during our period of observations that the pCO₂ of the overlying air was almost constant compared with that of the brine. Therefore, the air–sea-ice CO₂ flux will depend primarily on the brine pCO₂. However, in our data there was no clear relationship between the air–sea-ice CO₂ flux and brine pCO₂ when we plotted all the data corrected during the observation period (Fig. 9). These results suggest that there are other processes, in addition to the difference in pCO₂ between the brine and overlying air, that influence the air–sea-ice CO₂ flux. We now examine these other processes.

Relationships between air–sea-ice CO₂ flux and ice surface properties

In order to clarify the process controlling the air–sea-ice CO₂ flux, in addition to the difference of pCO₂ between the brine and overlying air, we examine the relationships between the air–sea-ice CO₂ flux and ice surface properties.

The air–sea-ice CO₂ flux for each ice surface type as a function of the snow thickness is shown in Figure 10 and Table 2. The air–sea-ice CO₂ flux approaches zero with increasing snow thickness for sea-ice surface conditions of snow + slush (Fig. 10). When the snow is deeper than about 9 cm, most of the air–sea-ice CO₂ flux values are near zero (Fig. 10; Table 2). In contrast, when a chamber was installed above the sea ice after first removing the snow and slush layer (Fig. 4b), the negative air–sea-ice CO₂ flux ranged from −1.8 to −0.6 mg C m⁻² h⁻¹ (Fig. 10; Table 2). Even when the snow thickness was less than about 9 cm, the air–sea-ice CO₂ flux values were near zero under conditions where the sea-ice surface was classified as snow + superimposed ice or superimposed ice (Fig. 10; Table 2). These results suggest that a superimposed ice surface or a melted snow deposit deeper than 9 cm acts as a physical barrier to CO₂ diffusion from air to sea ice in this area during the ice-melting season, even though the brine in the sea ice has the potential (i.e. relatively low pCO₂) to take up CO₂ from the atmosphere under all ice surface conditions (Fig. 6). These results were coincident with the limited measurement of air–sea-ice CO₂ flux over the superimposed ice in the Antarctic pack ice (Reference DelilleDelille, 2006).

Fig. 10. Relationship between snow thickness and air–sea-ice CO₂ flux for various ice surface conditions. Crosses reflect the conditions depicted in Figure 4b. The horizontal dotted line indicates an air–sea-ice CO₂ flux of 0 mg C m⁻² h⁻¹.

We also examined the relationships between the air–seaice CO₂ flux and the difference in pCO₂ between the air and brine in sea ice (pCO_{2_air} − pCO_{2 brine} = Δ pCO_{2_air–brine}) for the two situations depicted in Figure 4 (Fig. 11). We found a notable difference between the two situations. When the snow and slush layer were removed artificially (Fig. 4b), there was a clear relationship between the air–sea-ice CO₂ flux and ΔpCO_{2_air–brine} (r ² = 0.59; P < 0.05), showing that the air–sea-ice CO₂ flux was driven mainly by the difference in pCO₂ between the brine and overlying air (Reference DelilleDelille, 2006; Reference Nomura, Yoshikawa-Inoue and ToyotaNomura and others, 2006). However, when the air–sea-ice CO₂ flux was measured over the snow cover (>9 cm) and slush layer (Fig. 4a) the flux was near zero.

Fig. 11. Relationships between air–sea-ice CO₂ flux and the difference in pCO₂ between the air and the brine in sea ice (ΔpCO₂__air–brine = pCO_{2_air} − pCO_{2_brine}) for the two situations depicted in Figure 4 during observations in 2008. Open circles reflect the conditions depicted in Figure 4b. The solid line indicates the linear regression of CO₂ flux vs ΔpCO_{2_air–brine} for open circles (r ² = 0.59; P < 0.05). The cross indicates the result obtained on 23 February with a snow thickness of 5.5 cm (Table 1). The horizontal dotted line indicates an air–sea-ice CO₂ flux of 0 mg C m⁻² h⁻¹.

There are some related studies concerning the air–soil CO₂ flux through snow deposited over soils (Reference Sommerfeld, Mosier and MusselmanSommerfeld and others, 1993, Reference Sommerfeld, Massman, Musselman and Mosier1996; Reference TakagiTakagi and others, 2005; Reference Schindlbacher, Zechmeister-Boltenstern, Glatzel and JandlSchindlbacher and others, 2007). These studies found that the snow cover deposited over soils is a porous medium for gases before snowmelting begins. While CO₂ concentrations in soils drive the overall CO₂ flux through the snow cover, temporal and spatial variability of the CO₂ flux may be caused by specific snowpack properties (Reference Schindlbacher, Zechmeister-Boltenstern, Glatzel and JandlSchindlbacher and others, 2007). In particular, during the snowmelting season, because of the meltwater input to the pores in snow (Reference DenothDenoth, 1982), refreezing of meltwater and the formation of a specific ice layer in the snowpack (Reference Schindlbacher, Zechmeister-Boltenstern, Glatzel and JandlSchindlbacher and others, 2007), the air–soil CO₂ flux through the snow cover is restricted (Reference Schindlbacher, Zechmeister-Boltenstern, Glatzel and JandlSchindlbacher and others, 2007). These same conditions could have occurred in the snow in our experiments because of the relatively high temperatures during our experimental periods (Table 1). Although we have no data regarding the detailed properties of the snow, our results provide direct evidence for the effect of the snow deposited over sea ice on the air–sea-ice CO₂ flux during the ice-melting season. To understand the effects of snow properties on the air–sea-ice CO₂ flux, further investigations are highly desirable to state the snow density, grain shape composition and moisture in more detail.