Enhancement of the North Atlantic CO₂ sink by Arctic Waters

The North Atlantic north of 50^∘ N is one of the most intense ocean sink areas for atmospheric CO₂ considering the flux per unit area, 0.27 Pg-C yr⁻¹, equivalent to −2.5 mol C m⁻² yr⁻¹. The northwest Atlantic Ocean is a region with high anthropogenic carbon inventories. This is on account of processes which sustain CO₂ air–sea fluxes, in particular strong seasonal winds, ocean heat loss, deep convective mixing, and CO₂ drawdown by primary production. The region is in the northern limb of the global thermohaline circulation, a path for the long-term deep-sea sequestration of carbon dioxide. The surface water masses in the North Atlantic are of contrasting origins and character, with the northward-flowing North Atlantic Drift, a Gulf Stream offspring, on the one hand and on the other hand the cold southward-moving low-salinity Polar and Arctic waters with signatures from Arctic freshwater sources. We have studied by observation the CO₂ air–sea flux of the relevant water masses in the vicinity of Iceland in all seasons and in different years. Here we show that the highest ocean CO₂ influx is to the Arctic and Polar waters, respectively, $-\mathrm{3.8}\pm \mathrm{0.4}$ and $-\mathrm{4.4}\pm \mathrm{0.3}$ mol C m⁻² yr⁻¹. These waters are CO₂ undersaturated in all seasons. The Atlantic Water is a weak or neutral sink, near CO₂ saturation, after poleward drift from subtropical latitudes. These characteristics of the three water masses are confirmed by data from observations covering 30 years. We relate the Polar Water and Arctic Water persistent undersaturation and CO₂ influx to the excess alkalinity derived from Arctic sources. Carbonate chemistry equilibrium calculations clearly indicate that the excess alkalinity may support at least 0.058 Pg-C yr⁻¹, a significant portion of the North Atlantic CO₂ sink. The Arctic contribution to the North Atlantic CO₂ sink which we reveal was previously unrecognized. However, we point out that there are gaps and conflicts in the knowledge about the Arctic alkalinity and carbonate budgets and that future trends in the North Atlantic CO₂ sink are connected to developments in the rapidly warming and changing Arctic. The results we present need to be taken into consideration for the following question: will the North Atlantic continue to absorb CO₂ in the future as it has in the past?

1 Introduction

The oceans take up about a quarter of the annual anthropogenic CO₂ emissions (Friedlingstein et al., 2019). This may even be an underestimate (Watson et al., 2020). The North Atlantic north of 50^∘ N is one of the most intense ocean sink areas for atmospheric CO₂ considering the flux per unit area (Takahashi et al., 2009). The reasons are strong winds and large natural partial pressure differences, ΔpCO₂ = (pCO_2sw − pCO_2a), between the atmosphere and the surface ocean. The ΔpCO₂ in seawater is a measure of the escaping tendency of CO₂ from seawater to the overlying air. The ΔpCO₂ is proportional to the concentration of undissociated CO₂ molecules, [CO₂]aq, which constitutes about 1 % of the total CO₂ dissolved in seawater (the remainder being about 90 %–95 % as [HCO${}_{\mathrm{3}}^{-}$] and 4 %–9 % as [CO${}_{\mathrm{3}}^{\mathrm{2}-}$]). The seawater pCO₂ depends sensitively on temperature and the TCO₂ $/$ Alk ratio, the relative concentrations of total CO₂ species dissolved in seawater (TCO₂ = [CO₂]aq + [HCO${}_{\mathrm{3}}^{-}$] + [CO${}_{\mathrm{3}}^{\mathrm{2}-}$]), and the alkalinity, Alk, which reflects the ionic balance in seawater. Large ΔpCO₂ has been attributed to (a) a cooling effect on the CO₂ solubility in the poleward-flowing Atlantic Water, (b) an efficient biological drawdown of pCO₂ in nutrient-rich subpolar waters, and (c) high wind speeds over these low-pCO₂ waters (Takahashi et al., 2002). Evaluations of ΔpCO₂ based on observation and models have indicated that the Atlantic north of 50^∘ N and northward into the Arctic takes up as much as 0.27 Pg-C yr⁻¹, equivalent to −2.5 mol C m⁻² yr⁻¹ (Takahashi et al., 2009; Schuster et al., 2013; Landschützer et al., 2013; Mikaloff Fletcher et al., 2006). The North Atlantic is a relatively well-observed region of the ocean (Takahashi et al., 2009; Bakker et al., 2016; Reverdin et al., 2018). Nevertheless, estimates of long-term trends for the North Atlantic CO₂ sink due to changes in either ΔpCO₂ or wind strength are conflicting, particularly the Atlantic Water-dominated regions (Schuster et al., 2013; Landschützer et al., 2013; Wanninkhof et al., 2013). The drivers of seasonal flux variations are considered to be inadequately understood (Schuster et al., 2013), and a mechanistic understanding of high-latitude CO₂ sinks is regarded as incomplete (McKinley et al., 2017). It has been common to many large-scale flux evaluations, modelled or from observations, that they are based on regions defined by geographical borders, latitude, and longitude, e.g. between 49 and 76^∘ N for the high-latitude subpolar North Atlantic (Takahashi et al., 2009; Schuster et al., 2013). A more realistic approach is to define biogeographical regions or biomes (Fay and McKinley, 2014). The influence of oceanographic property differences within this region on CO₂ fluxes has generally not been apparent, primarily due to Arctic latitude data limitations. The ability of current-generation Earth system models to predict trends in North Atlantic CO₂ has recently been questioned, and it as been suggested that their inadequacies may be caused by biased alkalinity in the simulated background biogeochemical state (Lebehot et al., 2019).

Figure 1 Mean July to September surface salinity in the vicinity of Iceland. The S=35 isohaline marks the boundary between northward-flowing Atlantic Water and southward-flowing cold Arctic Water and low-salinity Polar Water. Stations in the Irminger Sea are marked with ×, and stations in the Iceland Sea are marked with • for 1994–1995 and ▴ for 1995–1996 observations. Collection of Polar Water stations in 1983–2012 are marked with ▪. IRM and IS mark the location of time series stations. NAC: North Atlantic Current; IC: Irminger Current; NIIC: North Iceland Irminger Current; EIC: East Icelandic Current; EGC: East Greenland Current. Map based on the NISE dataset (Nilsen et al., 2008) and drawn using the Ocean Data View programme (Schlitzer, 2018).

The high-latitude North Atlantic Ocean in the vicinity of Iceland is a region of contrasting surface properties (Fig. 1). The northward-flowing North Atlantic Current carries relatively warm and saline Atlantic Water, derived from the Gulf Stream, as far as the Nordic Seas and the Arctic Ocean north of Svalbard. The Irminger Current branch carries Atlantic Water to south and west Iceland, and a small branch, the North Icelandic Irminger Current that transports 1 Sv (1 Sv = 10⁶ m³ s⁻¹), reaches the Iceland Sea (Stefánsson, 1962; Våge et al., 2011). The temperature and salinity properties of the Atlantic Water are known to change with atmospheric forcing and with freshening events (Dickson et al., 1988; Hátún et al., 2005; Holliday et al., 2020). The rapid East Greenland Current (EGC) (Håvik et al., 2017) flows southward from the Arctic to the North Atlantic, carrying cold Polar Water with low salinity, S<34.4, due to ice melt and a portion of the large freshwater input to the Arctic from rivers that contribute about 11 % of the global riverine discharge (Sutherland et al., 2009; McClelland et al., 2012). In between these extremes there are large areas of the Greenland and Iceland seas that contain predominantly intermediate Arctic Water which is a product of heat loss and freshwater export from the EGC (Fig. 1) (Våge et al., 2015). The north- and southward-flowing currents are separated by the Arctic Front outlined in Fig. 1 by the salinity = 35 contour generally oriented SW–NE. Deep water formation in the high-latitude North Atlantic produces cold, dense waters which, together with a similar product in the Labrador Sea, are source waters for the global thermohaline circulation linking the regional air–sea CO₂ flux to a route for the long-term deep ocean sequestration of anthropogenic CO₂ (Broecker, 1991). Downstream from the southward Polar Water and Arctic Water flows is the subpolar North Atlantic with high water column inventories of anthropogenic carbon (Khatiwala et al., 2013; Gruber et al., 2019). The high-anthropogenic-CO₂ regions have been attributed to the combined effects of the solubility and biology gas exchange pumps on the CO₂ fluxes (Takahashi et al., 2002). The region of our study affects large-scale ocean–atmosphere CO₂ exchange processes in the North Atlantic.

Here we evaluate regional, seasonal, and interannual air–sea carbon dioxide fluxes for the main surface waters characteristic of this region (Fig. 1). We base this work on extensive observations which cover regional water masses and all seasons and include different states of the North Atlantic Ocillation, NAO (Flatau et al., 2003). We employ two different observation approaches for flux estimates. Firstly, repeat station hydrography with emphasis on the seasonal flux patterns in Atlantic Water and in Arctic Water (Fig. 1). Secondly, we use underway ship records of surface pCO₂ where the emphasis was on the different surface water masses (Fig. 2).

Figure 2 Cruise tracks where surface layer salinity and pCO₂ were recorded underway. The left column shows sea surface salinity, and the right column shows pCO_2sw (µatm) along the cruise tracks. Maps were drawn using the Ocean Data View programme (Schlitzer, 2018).

We describe long-term carbon chemistry characteristics of water masses in mid-winter when physical forces prevail over biological processes. For the Irminger Sea and Iceland Sea we use time series observations (Olafsson et al., 2010), and for the EGC Polar Water we use a collection of pCO₂ data assembled in the period 1983 to 2012.

2 Methods

2.1 Data acquisition

2.1.1 Seasonal studies 1993–1996

Seasonal carbon chemistry variations in the relatively warm and saline (S>35) Atlantic Water were studied in 1993–1994 on 15 cruises from February 1993 to January 1994 with five stations on a 167 km long transect over the core of the Irminger Current and into the northern Irminger Sea (Fig. 1 and Tables S1 and S2 in the Supplement). In order to close the full annual cycle, until 23 February 1994 we use data from the previous year and date. In 1994–1996 the study centred on the colder and less saline Arctic Water of the Iceland Sea and was conducted on 22 cruises with sampling dates from 11 February 1994 to 12 February 1996.

In 1994 measurements were made at four stations on a 168 km long transect into the Iceland Sea Gyre and in 1995 at three stations across the East Icelandic Current (Fig. 1 and Tables S1 and S3). On each cruise the station work was completed in 1–2 d. For both regions, the timing of cruises was with the period of the phytoplankton spring bloom in mind (Takahashi et al., 1993). The work was conducted on vessels operated by the Marine Research Institute (MRI) in Reykjavik, Iceland, R/V Bjarni Sæmundsson and R/V Árni Friðriksson. Three times in 1994 the fishing vessel M/V Solrun was hired. In August 1994 the stations were completed on the Norwegian vessel R/V Johan Hjort.

Discrete surface layer, 1, 5, and 10 m pCO₂ samples were collected in 500 mL volumetric flasks, and total dissolved inorganic carbon samples, TCO₂, were collected in 250 mL flasks from water bottles on Rosette and Sea-Bird 911 conductivity–temperature–depth (CTD) instruments. ThepCO₂ samples were preserved with mercuric chloride and analysed ashore by equilibration at 4 ^∘C with a gas of known CO₂ concentration followed by gas chromatography with a flame ionization detector. The instrument was calibrated with N₂ reference gas and three standards, 197.85, 362.6, and 811.08 ppm, calibrated against standards certified by NOAA-CMDL at Boulder, CO, USA. The standards used for the underway measurements were similarly calibrated (Chipman et al., 1993). Samples for total dissolved inorganic carbon, TCO₂, were similarly preserved with mercuric chloride and analysed by coulometry ashore. Quality assurance and sample storage experiments indicated an overall precision of the discrete sample pCO₂ determinations better than ±2 µatm and for the TCO₂ determinations about ±2 µmol kg⁻¹ after 1990 but ±4 µmol kg⁻¹ earlier (Olafsson et al., 2010).

2.1.2 Underway pCO₂ records 2006–2007

The underway pCO₂ determinations in 2006–2007 covered areas of the East Greenland Current in and northwards from the Denmark Strait, in addition to Atlantic and Arctic waters. The six cruises (Table S4) covered all seasons and all three water masses but with variable areal extensions (Fig. 2). Seawater was pumped continuously from an intake at 5 m depth at 10 L min⁻¹ into a shower-head equilibrator with a total volume of 30 L and a headspace of 15 L. Temperature at the inlet and salinity were measured with a Sea-Bird model SBE-21 thermosalinograph (Sea-Bird Electronics, Seattle, WA, USA). Underway pCO₂ determinations were carried out with a system similar to the one described by Bates et al. (1998). The mole fraction of CO₂ ($V_{{\mathrm{CO}}_{\mathrm{2}}}$) in the headspace was determined with a Li-Cor infrared analyser model 6251 (Li-Cor Biosciences, Lincoln, NB, USA). The instrument was calibrated against four standards of CO₂ in air certified by NOAA-CMDL at Boulder, CO, USA, and a N₂ reference gas. The standards had CO₂ dry-air mole fractions of 122.19, 253.76, 358.41, and 476.81 ppm. The pCO₂ seawater (sw) determinations were corrected to in situ seawater temperatures using the equation (Takahashi et al., 1993)

$$\begin{array}{}\text{(1)}& p{\mathrm{CO}}_{\mathrm{2}\mathrm{sw}}\left(\text{in situ}\right)=p{\mathrm{CO}}_{\mathrm{2}\mathrm{sw}}\left(\text{eq}\right)e^{\mathrm{0.0423}(T_{\mathrm{in}\mathrm{situ}}-T_{\mathrm{eq}})}.\end{array}$$

The precision of the underway pCO₂ determinations is estimated by SOCAT to be better than ±5 µatm (Bakker et al., 2016).

2.1.3 Time series data

We use discrete sample pCO₂ and TCO₂ data to calculate total alkalinity from the Irminger Sea and the Iceland Sea time series stations (Ólafsson, 2012, 2016).

2.1.4 Polar Water data collection

Discrete samples for carbon chemistry studies were taken at stations (N=146) in the East Greenland Current when opportunity permitted on cruises in the period from 1983 to 2012. The 25 m surface layer data include >400 TCO₂ samples and >300 pairs of pCO₂ and TCO₂ for calculation of carbonate system parameters. The seasonal cycle by month is evaluated from the composite data.

2.1.5 Carbonate chemistry calculations

The most desirable way to compute carbonate chemistry parameters is to use pCO₂ and TCO₂ (Takahashi et al., 2014). We calculate total alkalinity from discrete sample pCO₂ and TCO₂ data pairs using the CO2SYS.xls v2.1 software (Lewis and Wallace, 1998; Pierrot et al., 2006) and select carbonic acid dissociation constants (Lueker et al., 2000), the constant for HSO${}_{\mathrm{4}}^{-}$ (Dickson, 1990), and boron concentrations (Lee et al., 2010).

2.2 CO₂ air–sea flux calculations

In this study, the partial pressure of carbon dioxide in seawater samples has been measured by gas–seawater equilibration methods (Olafsson et al., 2010). The results are expressed as pCO₂. The bulk flux of the carbon dioxide across the air–sea interface is often estimated from its relationship with wind speed and sea-air partial pressure difference, ΔpCO₂. We determine the flux (F) from ΔpCO₂ and use Eqs. (2) and (3) for estimating the bulk air–sea fluxes of CO₂ (Takahashi et al., 2009).

$$\begin{array}{}\text{(2)}& {\displaystyle }F=k\cdot \mathit{\alpha }\cdot \mathrm{\Delta }p{\mathrm{CO}}_{\mathrm{2}}\text{(3)}& {\displaystyle }F=\mathrm{0.251}{U}^{\mathrm{2}}(\mathit{Sc}/\mathrm{660}{)}^{-\mathrm{0.5}}\mathit{\alpha }(p{\mathrm{CO}}_{\mathrm{2}\mathrm{w}}-p{\mathrm{CO}}_{\mathrm{2}\mathrm{a}})\end{array}$$

There $k=\mathrm{0.251}{U}^{\mathrm{2}}(\mathit{Sc}/\mathrm{660}{)}^{-\mathrm{0.5}}$ is the gas transfer velocity or kinetic component of the expression (Wanninkhof, 2014), α is the solubility of CO₂ gas in seawater (Weiss, 1974), and ΔpCO₂ = (pCO_2sw − pCO_2a) is the partial pressure difference or thermodynamic component of the expression (Takahashi et al., 2009). For the wind speed, U, we use the CCMP-2 reanalysis wind product (Wanninkhof, 2014; Atlas et al., 2011; Wanninkhof and Triñanes, 2017).

The atmospheric partial pressure values, pCO_2a, used in the ΔpCO₂ calculations are weekly averages from the GlobalView-CO2 database for the CO2-ICE location which is at Vestmannaeyjar islands, off the coast of south Iceland (GlobalView-CO2, 2013). Mauna Loa values were used for periods where CO2-ICE data were missing, 1983–1992 and 2010–2012 (Tans and Keeling, 2019). The dry-air VCO₂ mole fraction values were converted to microatmospheres using pCO2 (µatm) = $V_{{\mathrm{CO}}_{\mathrm{2}}}$ (P_a−P_w), where P_a is the barometric pressure and P_w is the equilibrium water vapour pressure (Weiss and Price, 1980).

For the Irminger Sea seasonal study we use 30 d running means of the squared daily wind speed for the region 63.5 to 64.5^∘ N and 27 to 32^∘ W, and for the Iceland Sea seasonal study a similar wind product for the region 66.5 to 68.5^∘ N and 12 to 19 ^∘ W was used. Fluxes were calculated for the periods between cruises from interpolatedpCO₂ data and period mean 30 d squared wind running mean data. There are thus 14 flux periods covering a year for the Irminger Sea and 21 flux periods covering 2 years in the Iceland Sea (Tables S1 and S2). The annual fluxes were found by summation of the period fluxes (Table 1).

For the underway cruises from 2006 to 2007 we used CCMP-2 daily wind fields at 1^∘ ×1^∘ for the region 62 to 72^∘ N and 5 to 40^∘ W. This region was further divided into four sub-regions by latitude 64.9^∘ N and longitude 20^∘ W. Daily 30 d running means of the squared wind speed from two locations in each sub-region were extracted and their means used for flux calculations when the vessel sailed in the area. Fluxes were calculated for all pCO₂ data from the six cruises, in total 42 938 measurements.

The flux data from each of the six cruises were categorized into the three seawater types using the following criteria:

1. Atlantic Water S>35, Arctic Water S is 34.4–34.9, and Polar Water S<34.4.

2. Seasonal salinity and temperature variations were taken into account.

3. Waters with runoff influences from Iceland were excluded using salinity and ship position data.

Thus a total of 33 352 measurements were used, or 78 % of the flux data points. The CO₂ fluxes in the realm of each water mass were assessed for the duration of each cruise by numerical integration. Fluxes in the five periods between cruises were assessed by interpolation of temperature, salinity, and pCO₂ for each water mass and by using period regional 30 d running means of squared wind speed data. The annual flux for each water mass was assessed by summation.

3 Results

3.1 Seasonal variations and annual CO₂ fluxes at regional water masses

The wind gas transfer coefficient reveals seasonal variations reflecting strong winds in winter when they may be stronger over the Irminger Sea than the Iceland Sea as in 1993–1994 and in 1994–1995 (Figs. 3a and S1 in the Supplement). Both the Irminger Sea and the Iceland Sea seasonal studies reveal the strongest CO₂ undersaturation, with negative ΔpCO₂ of about 100 µatm in May at the time of the phytoplankton spring bloom (Fig. 3b). The undersaturation diminishes through the summer and autumn followed by a gradual return to winter conditions (Takahashi et al., 1985; Peng et al., 1987; Takahashi et al., 1993).

Figure 3 Seasonal variations in the Atlantic Water of the Irminger Sea (red) and in the Arctic Water of the Iceland Sea (blue). The gas transfer velocity (a) reflects the seasonal wind strength and the error bars its variations during intervals between cruises. Delta pCO₂ (b) records the tendency for CO₂ to be transferred to the atmosphere (positive) or from the atmosphere to the ocean (negative). The CO₂ flux rate (c) reveals that the Arctic Water is a CO₂ sink in all seasons, whereas the Atlantic Water is a source in winter and a weak sink at other times of the year. The error bars indicate ±1 standard deviation from the mean and reflect the variations between the stations observed each cruise.

The CO₂ influx in the spring is, however, relatively small as the wind gas transfer coefficient is then moderate (Fig. 3a). In the autumn the winds strengthen with heat loss and vertical mixing while CO₂ undersaturation still persists. In mid-winter, February–March, vertical mixing brings richer CO₂ water to the surface of the Irminger Sea, leading to supersaturation (Ólafsson, 2003); the flux reverses and the region becomes a weak source for atmospheric CO₂ (Fig. 3c).

Table 1 Annual sea–air CO₂ fluxes (mol C m⁻² yr⁻¹) in the three water masses.

The integrated annual CO₂ flux shows that the Atlantic Water in the Irminger Sea was a weak sink, $-\mathrm{0.69}\pm \mathrm{0.16}$ mol C m⁻² yr⁻¹, in 1993 (Table 1). The more extensive underway area coverage of the Atlantic Water in 2006–2007 in essence confirmed the seasonal pattern and indicated that the Atlantic Water was a neutral sink, 0.07±0.15 mol C m⁻² yr⁻¹ for this year (Table 1). The winter gas transfer coefficient was again significantly larger over the Atlantic Water regions than the Arctic and Polar waters, facilitating air–sea equilibration (Fig. 4b).

The years of the Iceland Sea observations, 1994–1996, coincided with a large transition in the North Atlantic Oscillation (NAO) from a positive state 1994/1995 to a negative state in 1995/1996 and large-scale shifts in ocean fronts (Flatau et al., 2003). Vertical density distribution in the Iceland Sea indicated an enhanced convective activity in 1995 (Våge et al., 2015). Cold northeasterly winds were persistent in the spring of 1995, resulting in record low temperature anomalies for the north Iceland shelf (Ólafsson, 1999). In 1995 the spring bloom associated with undersaturation, ΔpCO₂, was only half of that in 1994, possibly due to a weaker stratification in May, and continued over the summer season (Fig. S2) (Våge et al., 2015). As in the Irminger Sea the spring-bloom-associated CO₂ influx is small. The largest CO₂ influx was in the autumn and early winters of 1995 and 1996; as temperature dropped, winds gathered strength and vertical mixing was enhanced. This compensated for the small spring bloom in 1995, and the annual bulk fluxes in 1994 and 1995 are similar and high despite very different physical conditions (Table 1). The UWpCO₂ surveys had less temporal resolution but confirmed year-round undersaturation of the Arctic Water. However, the integrated annual influx, −2.84 mol C m⁻² yr⁻¹, was significantly less than evaluated with repeat station data even though the strength of the gas transfer coefficient was similar in both studies (Table 1, Fig. 4a and b). This may reflect the large underway area coverage compared with the repeated fixed stations.

Figure 4 Seasonal air–sea CO₂ flux variations from UWpCO₂ observations. (a) Atlantic Water (red) is a weak sink in summer and neutral over the year, n=7068. Both Arctic Water (blue) n = 16 874 and Polar Water (green) n=9410 are strong sinks throughout the year. The error bars indicate ±1 standard deviation from the mean. (b) The gas transfer coefficient for the Atlantic Water regime is significantly stronger in winter than for the Arctic and Polar Water.

Ice cover in the East Greenland Current is variable and the ice edge at the seasonal minimum has moved northward and from the Denmark Strait with decreasing Arctic sea ice (Serreze and Meier, 2019). The Polar Water salinity ranges from 34.4 to less than 30 in summer. The lowest-salinity water freezes, leading to salinity around 34 in winter. We covered the Polar Water in all six UWpCO₂ surveys 2006–2007 (Fig. 2), and undersaturation characterized this water mass in all cruises. The integrated annual influx, −4.44 mol C m⁻² yr⁻¹ (Table 1, Fig. 4), shows the Polar Water to be the strongest CO₂ sink, 80 % above the estimated mean for the Atlantic north of 50^∘ N, −2.5 mol C m⁻² yr⁻¹ (Takahashi et al., 2009). Further comparison with the Takahashi climatology indicates a broad agreement with the Arctic Water region NE of Iceland with −3.5 to −4.5 mol C m⁻² yr⁻¹ and with the Atlantic Water region S and SW of Iceland with about −1 mol C m⁻² yr⁻¹ (Takahashi et al., 2009).

3.2 Long-term ΔpCO₂ characteristics of the regional water masses

We evaluate the long-term pCO₂ characteristics of the three water masses from three other data assembled over about 30 years. We use the Polar Water data collection and draft a composite picture of seasonal ΔpCO₂ variations in Polar Water in and north of the Denmark Strait (Fig. 1), which confirms year-round undersaturation, deep in summer and in mid-winter when salinity increases to ∼34, with ΔpCO₂ levels at about −50 µatm (Fig. 5a). Long-term winter ΔpCO₂ in the Irminger Sea and Iceland Sea (Figs. 1 and 5b) when biological activity is minimal (Olafsson et al., 2009) shows the Atlantic Water to be slightly supersaturated and follows the atmospheric pCO₂ increase of 1.80 µatm yr⁻¹, whereas the Arctic Water is undersaturated to about −35 µatm. The Gulf Stream-derived Atlantic Water which reaches the northern Irminger Sea and the Nordic Seas has had a long contact time with the atmosphere to lose heat and almost reach CO₂ saturation (Takahashi et al., 2002; Olsen et al., 2006).

Figure 5 Water mass decadal surface water pCO₂ characteristics. (a) A composite picture of delta pCO₂ from 146 stations with Polar Water pCO₂ observations (n=312) from the 25 m surface layer from 1983 to 2012 shows undersaturation at all times of the year. The error bars indicate ±1 standard deviation from the monthly means. (b) Atlantic Water at the Irminger Sea time series station (red) is generally a weak CO₂ source in winter (24 winters, 52 samples), January–March, whereas winter (25 winters, 61 samples) CO₂ undersaturation persists at the Iceland Sea time series site (blue). The error bars indicate ±1 standard deviation from the surface layer station means.

The Polar Water in the East Greenland Current which is advected southward from the Arctic is in general characterized by low temperature and large seasonal salinity and carbonate chemistry variations. Both physical and biogeochemical processes generate the large seasonal variability but the winter observations represent the state of lowest biological activity (Fig. 6) (Table 2). The TCO₂ data in Table 2 are uncorrected for hydrographic variations or anthropogenic trends, but the Atlantic Water is based on a short period of 10 years, and the Polar Water atmospheric contact history is poorly known.

Table 2 Mean IRM-TS Atlantic Water surface layer conditions in winter in 2001–2010 and in the Polar Water 25 m surface layer from November to April 1984–2012.

The winter conditions in the northward-flowing Atlantic Water at the Irminger Sea time series station in 2001–2010 (Table 2) are in stark contrast and have notably higher pCO₂ and lower TCO₂ $/$ S and Alk $/$ S ratios than the Polar Water in winter.

Figure 6 Polar Water seasonal carbonate chemistry variations. Composite Polar Water data from the 25 m surface layer. The seasonal variations in the (a) total inorganic carbon $/$ salinity ratio, µmol kg⁻¹ psu⁻¹; (b) alkalinity $/$ salinity ratio, µmol kg⁻¹ psu⁻¹; and (c) total inorganic carbon $/$ alkalinity ratio reflect biological carbon assimilation and inorganic processes associated with freshwater inputs which lower the salinity (d) to the annual minimum in late summer. The error bars indicate ±1 standard deviation from the monthly means. The red horizontal lines mark the Atlantic Water benchmarks (Table 2).

We take the Atlantic Water in winter (Table 2) as a proxy (benchmark) for the relatively warm and saline water advected from the North Atlantic to the Nordic Seas and the Arctic and compare with it the carbonate chemistry seasonal variations in the southward-flowing Polar Water (Fig. 6). The Alk $/$ S ratio for the Polar Water is higher than that for the Atlantic Water in winter and throughout the year (Fig. 6b). The TCO₂ $/$ S ratio of the Polar Water is larger than that of the Atlantic Water except in early summer when biological assimilation, photosynthesis, decreases the TCO₂ concentration. The TCO₂ $/$ Alk ratio falls as a consequence (Fig. 6c), which leads to strong pCO₂ undersaturation and large delta pCO₂ (Figs. 6c and 5a). The high TCO₂ $/$ S and Alk $/$ S ratios indicate alkalinity and carbonate inputs as fresh water lower the Polar Water salinity to a minimum in late summer (Fig. 6d).

4 Discussion

The Polar Water TCO₂ $/$ S and Alk $/$ S ratios (Table 2 and Fig. 6) indicate both alkalinity and dissolved carbonate additions. The choice of winter ratios (Table 2) as benchmarks is solely for the evaluation of seasonal changes in the Polar Water. Representative annual long-term TCO₂ $/$ S and Alk $/$ S means would be more realistic but are not available. Still, such a TCO₂ $/$ S ratio would expectedly be lower than the winter one. An assessment of the effects of the relative TCO₂ and Alk additions to Polar Water depends on the benchmarks chosen (Table 2).

The carbonate chemistry of Polar Water differs from that of open ocean waters, e.g. Atlantic Water, in having an increasingly higher alkalinity $/$ salinity ratio as the salinity decreases from about S=34.4. The excess alkalinity has been attributed to the high riverine input from continents to the Arctic (Anderson et al., 2004). The flow-weighted average alkalinity of six major Arctic rivers, discharging 2.245×10³ km³ yr⁻¹, is 1048 µmol kg⁻¹, however, without assessed uncertainty (Cooper et al., 2008). The river runoff into the Arctic is estimated to be about 4.2×10³ km³ yr⁻¹, or 0.133×10⁶ m³ s⁻¹ (0.133 Sv). This is about 11 % of the global freshwater input to the oceans (Carmack et al., 2016). Taking the average alkalinity 1048 µmol kg⁻¹, the amount of alkalinity added by rivers to the Arctic and transported to the North Atlantic via the Canadian Arctic Archipelago and via the Fram Strait and further south with the Labrador and East Greenland Currents would be 4.4×10¹² mol yr⁻¹ (Supplement). Cooper et al. (2008) reported on riverine alkalinity but not on associated inorganic carbonate. A recent assessment of Polar Water boron concentrations indicates insignificant borate contribution with Arctic rivers (Olafsson et al., 2020). The riverine alkalinity may primarily be attributed to carbonate alkalinity, CA = [HCO${}_{\mathrm{3}}^{-}$] + 2[CO${}_{\mathrm{3}}^{\mathrm{2}-}$]. The potential of the added alkalinity to reduce pCO₂ of seawater would depend on its excess over TCO₂.

Linear alkalinity–salinity relationships observed in the Arctic Ocean and the Nordic Seas and their extrapolated intercepts to S=0 have indicated freshwater sources with alkalinity 1412 µmol kg⁻¹ (Anderson et al., 2004) and 1752 µmol kg⁻¹ (Nondal et al., 2009). Climatological data from the West Greenland, Iceland, and Norwegian seas show a high S=0 intercept of 1796 µmol kg⁻¹ but a lower one for the High Arctic north of 80^∘ N, 1341 µmol kg⁻¹ (Takahashi et al., 2014). The climatological relationships were for potential alkalinity, A_P = A_T + NO${}_{\mathrm{3}}^{-}$, which has little influence since the nitrate concentrations are low. The intercepts may be interpreted as the mean alkalinity of freshwaters added to the Arctic by rivers and melting ice and snow. However, the above intercepts indicate considerable variability, and they are also higher than the average alkalinity of Arctic rivers, 1048 µmol kg⁻¹. The excess alkalinity would lower the pCO₂ in seawater (and increase the pH) and thus give it an increased capacity to take up CO₂ from the air. The thermodynamic driving force for seawater CO₂ uptake, (pCO_2sw − pCO_2a), would be enhanced.

How large is the potential effect of excess Arctic alkalinity on the CO₂ uptake by the Nordic Seas and the North Atlantic? We consider an estimate. Suppose that the pCO₂ in seawater was restored to the original value by absorbing CO₂ from the atmosphere. The carbonate equilibrium relations in seawater give that pCO₂ is unchanged if ΔTCO₂ $/$ ΔAlk = 0.85 (Fig. S3 and Supplement). This ratio of additions is nearly constant in the temperature and salinity range of the subarctic North Atlantic surface waters (t=5 ^∘C, S=35). The volume transport of Polar Water, density < 1027.8 kg m⁻³, by the EGC has recently been estimated as 3.9 Sv (Våge et al., 2013). Taking S=33.0 for the mean Polar Water salinity and using Eqs. (6) and (7) in Nondal et al. (2009), the mean Polar Water alkalinity is 2256 µmol kg⁻¹, which is 46 µmol kg⁻¹ more alkalinity than for Atlantic Water calculated at the same salinity (Nondal et al., 2009). This much excess alkalinity would lower the pCO₂ of Atlantic Water by 88 µatm and increase the pH by 0.10. Thus, the excess alkalinity advected to the North Atlantic by the EGC is 5.7×10¹² mol yr⁻¹. Using 0.85 for the ΔTCO₂ $/$ ΔAlk additions at a constant pCO₂, we obtain that the contribution of the excess EGC alkalinity to the uptake of CO₂ from the atmosphere would be 4.8×10¹² mol CO₂ yr⁻¹, or 0.058 Pg-C yr⁻¹. The estimate corresponds to 21 % of the net CO₂ uptake of 0.27 Pg-C yr⁻¹ for the subarctic oceans north of 50^∘ N (Takahashi et al., 2009). We did not include in the estimate any alkalinity contribution with the considerable Canadian Arctic Archipelago Polar Water transport (Haine et al., 2015). The effect of excess alkalinity on the North Atlantic CO₂ uptake flux may therefore be substantially greater than our estimate. We note that the winter undersaturation levels, of −50 and −35 µatm observed in the Polar and Arctic Waters, respectively (Fig. 5), translate to excess alkalinity of 19 and 21 µmol kg⁻¹ for further CO₂ influx downstream.

The difference between the average measured Arctic river alkalinity and the regression-based estimates of alkalinity sources suggests that other origins and processes than the rivers contribute to the Polar Water alkalinity exported with currents from the Arctic to the Atlantic Ocean. Photic layer primary production in the absence of calcification may lower the TCO₂ $/$ Alk ratio and seawater pCO₂ in marginal seas (Bates, 2006), while acidification is increasing in other regions (Anderson et al., 2017; Qi et al., 2017) and projected to become extensive at the end of the century (Terhaar et al., 2020). Furthermore, the sea-ice seasonal formation and melting may affect the TCO₂ $/$ Alk ratio (Grimm et al., 2016; Rysgaard et al., 2007). Efforts to reconstruct alkalinity fields and alkalinity climatology for the Arctic have however been difficult (Broullón et al., 2019).

The Arctic is complex, and complex climate-warming-related changes are observed in the western Arctic Ocean (Ouyang et al., 2020) and expected in marine freshwater systems of the warming Arctic (Carmack et al., 2016). Not least of all, the ice cover and areas of multi-year ice are decreasing (Serreze and Meier, 2019). River water alkalinity increases with an addition of cations derived from the chemical weathering of silicate and carbonate rocks (Berner and Berner, 1987). Accordingly, an increase in Arctic weathering rates, in response to warmer climate and increasing atmospheric CO₂, could increase the river water alkalinity transported into the oceans. Such an increase would lower the pCO₂ in seawater and enhance the oceanic uptake of atmospheric CO₂, providing a negative feedback mechanism to the climatic warming resulting from increased atmospheric CO₂.

5 Conclusions

The North Atlantic region we describe has Atlantic Waters advected from southern temperate latitudes and cold lower-salinity Arctic and Polar waters carried with the East Greenland Current from the Arctic. The Atlantic Water seasonal pCO₂ variations are primarily driven by regional thermal and biological cycles but without much net annual influx of CO₂. The southward-flowing Arctic and Polar waters are on the contrary strong and persistent year-round CO₂ sinks. These waters are advected towards the sub-polar North Atlantic with high inventories of anthropogenic carbon. The TCO₂ $/$ S and Alk $/$ S Polar Water ratios are higher than those for the Atlantic Water, indicating carbonate and alkalinity sources. We point here to the Polar Water and Arctic Water CO₂ influx and excess alkalinity as an additional unrecognized source contributing to the North Atlantic CO₂ sink. We also see that there are gaps and conflicts in the knowledge about the Arctic alkalinity and carbonate budgets and that future trends in the North Atlantic CO₂ sink are connected to developments in the rapidly warming and changing Arctic.