The seasonal cycle of δ13CDIC in the North Atlantic Subpolar Gyre

This study introduces for the first time the δCDIC seasonality in the North Atlantic subpolar gyre (NASPG) usingδCDIC data obtained in 2005–2006 and 2010–2012 with dissolved inorganic carbon (DIC) and nutrient observations. On the seasonal scale, the NASPG is characterized by higherδCDIC values during summer than during winter, with a seasonal amplitude between 0.70 ± .10 ‰ (August 2010–March 2011) and 0.77 ± 0.07 ‰ (2005–2006). This is mainly attributed to photosynthetic activity in summer and to a deep remineralization process during winter convection, sometimes influenced by ocean dynamics and carbonate pumps. There is also a strong and negative linear relationship between δCDIC and DIC during all seasons. Winter data also showed a large decrease in δCDIC associated with an increase in DIC between 2006 and 2011–2012, but the observed time rates ( −0.04 ‰ yr−1and +1.7 μmol kg−1 yr−1) are much larger than the expected anthropogenic signal.


Introduction
The stable isotopic composition of dissolved inorganic carbon (DIC) in the ocean is a useful semi-conservative tracer for various oceanic analysis (e.g., water masses, paleocirculation) and global carbon studies (constraint for the global carbon budget, interpreting anthropogenic carbon). This is reported relative to the reference V-PDB (Vienna-Pee Dee Belemnite) in terms of δ 13 C in ‰ (Craig, 1957): (1) The mean annual large-scale distribution of this tracer is now relatively well documented at the sea surface and at depth. Values range typically between 0.5 ‰ and 2.5 ‰ in the modern surface ocean (Gruber et al., 1999). Regular observations conducted since the 1970s-1980s at a few time series stations or repetitive tracks were used to describe and interpret its temporal variability. In subtropical oceanic regions, the seasonal δ 13 C amplitude of 0.2-0.3 ‰ at the BATS (Bermuda Atlantic Time series) station and of 0.1 ‰ at the ALOHA station (Pacific time series north of Hawaii) reflect the local balance between air-sea CO 2 exchange, biological activity and vertical mixing (Gruber et al., 1998(Gruber et al., , 1999Gruber and Keeling, 1999;. This is also observed in the Southern Indian Ocean, with a seasonal δ 13 C amplitude of 0.3 ‰ in the western part of the basin (Racapé et al., 2010), whereas it reaches 0.45 ‰ south of Tasmania (McNeil and Tilbrook, 2009). Studies have also highlighted a significant decrease in δ 13 C DIC , attributed to the anthropogenic CO 2 uptake in the ocean. This results from human-induced perturbations that reduce the δ 13 CO 2 in the atmosphere with a time rate of change of −0.024 ‰ yr −1 (estimate for the 1985-2008 period with the Alert station data set; Keeling et al., 2010), as the combustion of the organic carbon pool produces CO 2 gas that is strongly depleted in the heavy carbon isotope ( 13 C). This signal, known as the oceanic 13 C Suess effect, provides an additional constraint to estimate the anthropogenic carbon uptake in the ocean. This has been estimated as less than −0.007 ‰ yr −1 in polar regions (McNeil et al., 2001;Olsen et al., 2006;Sonnerup and Quay, 2012) and as large as −0.025 ‰ yr −1 in subtropical regions (Bacastow et al., 1996;Gruber et al., 1999;Sonnerup and Quay, 2012). The high latitudes of the North Atlantic Ocean are considered to be one of the strongest anthropogenic CO 2 sinks, a consequence of the large heat loss and deep convection during winter as well as strong biological activity during spring and summer. In this region, the temporal variability (seasonal to decadal) of the δ 13 C DIC still needs to be investigated. The winter-summer difference is expected to be as large as 1 ‰ (Gruber et al., 1999), which could potentially mask the oceanic 13 C Suess effect in surface water (McNeil et al., 2001). In this context, this study introduces for the first time the δ 13 C DIC seasonality in the North Atlantic subpolar gyre using δ 13 C DIC surface data obtained in 2005-2006 and 2010-2012 from merchant vessels Skogafoss and Reykjafoss (14 cruises) along with DIC and nutrient measurements. Driving processes will be identified from isotopic data and from the semi-quantitative tracer ( δ 13 C bio ) of Gruber et al. (1999), which enables us to isolate the biological contribution to the seasonal amplitude from the physical contributions (e.g., air-sea gas exchange fractionation, ocean dynamics). The time rate of change in δ 13 C DIC will be investigated over the short period of 5 to 7 years and compared to other regions.

Data collection
Data used in this study were collected in the North Atlantic subpolar gyre, mostly from merchant vessels Skogafoss and Reykjafoss between Iceland and Newfoundland ( Fig. 1) over two periods in 2005-2006 and 2010-2012. This is part of sampling efforts carried from vessels of opportunity within the SURATLANT project initiated in 1993. The sampling periods for each cruise and vessel are summarized in Table 1.
During all cruises, seawater was pumped from a depth of ∼ 3-5 m, depending on the ship. Sea surface salinity (SSS) was from discrete salinity samples. Sea surface temperature (SST) was either from the continuously recording ThermoS-alinoGraph (TSG seabird Electronics, Inc. USA, model 21) corrected for an average bias to sea surface temperature, or from nearly simultaneous XBT temperature profiles (at 3 m). SST and SSS accuracies were estimated at 0.2 • C and 0.01, respectively. δ 13 C DIC , DIC and nitrate concentrations (NO − 3 ) were also obtained from discrete samples collected every 4 h and analyzed later in land laboratories. Over the 2005-2006 period, DIC concentrations were estimated manometrically during the acid CO 2 extraction procedure for δ 13 C DIC measurements from the helium stripping technique. This analytical method has been described previously by Quay et al. (1992. These measurements have an accuracy of  ± 0.02 ‰ for δ 13 C DIC based on a helium stripping technique adapted from the one used by Kroopnick (1974) and ± 5 µmol kg −1 for DIC, based on a comparison to coulometric DIC values and to the Certified Reference Material (CRM) provided by Prof. A. Dickson (Scripps Institution of Oceanography, San Diego, USA). Over the 2010-2012 period during the SURATLANT cruises, δ 13 C DIC were measured via an acid CO 2 extraction method in a vacuum system developed by Kroopnick (1974), whereas DIC were determined at the same time as total alkalinity (Alk) by a potentiometric titration derived from the method developed by Edmond (1970) using a closed cell and a CRM calibration. Further details on the sampling methods and analytical techniques are provided in Racapé et al. (2010Racapé et al. ( , 2013 for δ 13 C DIC and in Corbière et al. (2007) for DIC. Both parameters were analyzed at the LOCEAN-IPSL Laboratory (Paris, France). δ 13 C DIC values have a precision of ± 0.01 ‰ (Vangriesheim et al., 2009) and a reproducibility of ± 0.02 ‰, whereas DIC concentrations have an accuracy of ± 3 µmol kg −1 . Finally, nutrient concentrations were measured for all cruises with standard colorometric methods at the Marine Research Institute (Reykjavik, Iceland). The analytical procedure and the quality control for the nutrient analyses have been described in detail in Olafsson et al. (2010), where the long-term accuracy has been estimated as ± 0.2 µmol L −1 for nitrate and silicate, and ± 0.03 µmol L −1 for phosphate.

2.2
δ 13 C bio and δ 13 C phys estimation To determine the major process driving the seasonal δ 13 C DIC distribution (t 1 → t 2 ), we used two isotopic and semiquantitative tracers, noted δ 13 C bio and δ 13 C phys , which reveal the biological and physical signatures, respectively, as shown by Eqs. (2) and (3).
Described by Gruber et al. (1999), the biological processes contribution ( δ 13 C bio ) results from the combined photosynthesis, respiration and remineralization processes (the soft-tissue pump) as well as the formation-dissolution of the calcium carbonate shells (the carbonate pump). In this study, the δ 13 C bio will be expressed according to observed biogeochemical conditions from one month or season (t 1 ) to another one (t 2 ) by Eq. (4), with the stoichiometric carbon-nitrate ratio (r c:n ) of 117 : 16 from Anderson and Sarmiento (1994), the δ 13 C value of the carbonate (δ 13 C carb ) of 2 ‰ like Gruber et al. (1999), and the δ 13 C value of organic matter (δ 13 C org ) of −24 ‰ from Goericke and Fry (1994). This last value depicts the average for the northern North Atlantic where sea surface temperature is around 10 • C in summer.
As the δ 13 C variation associated with the carbonate pump (< 0.02 ‰ of the measurement error) does not contribute much to the biological signature, this mechanism will be afterwards neglected.
According to Eq. (3), the physical processes contribution (air-sea CO 2 exchange fractionations and ocean dynamics; δ 13 C phys ) corresponds to the difference between the δ 13 C DIC at t 2 and the expected δ 13 C DIC if biological activity was the only process controlling the t 1 to t 2 distribution. This is expressed in Eq. (5).
As t 1 to t 2 DIC change due to physical and biological processes (up to 100 µmol kg −1 ) is lower than the mean DIC concentration observed in the surface water of this region (between 2050 and 2150 µmol kg −1 ), we assume that DIC t1 /DIC t2 equals 1.

Surface δ 13 C DIC distribution
During each transect, surface δ 13 C DIC is distributed according to a north-south gradient, with an amplitude between 0.2 ‰ and 0.4 ‰ depending on the season and with higher values close to the Newfoundland coast (up to 1.8 ‰; (variation up to 100 µmol kg −1 ; Fig. 2, middle panel), which increases northward as observed previously by Corbière et al. (2007). These authors have also shown that the area south of 53 • N was characterized by large spatial variability in SSS (Fig. 2, top panel), SST, DIC (Fig. 2, bottom panel) and nutrient concentrations when crossing the Labrador Current and approaching the Newfoundland coast (Fig. 1). Influence of coastal waters was also identified north of 62 • N near Iceland and on δ 13 C DIC distribution (Fig. 2, bottom panel). Thus, we selected a region where both hydrological and carbon properties (including δ 13 C DIC ) are much more homogeneous between 53 • N and 62 • N, which corresponds to the interior of the North Atlantic subpolar gyre, denoted NASPG, west of the Reykjanes Ridge.
In the NASPG, the δ 13 C DIC meridional gradient is weaker (close to 0.1 ‰). Highest values were observed during summer (up to 1.65 ± 0.  Table 1). Compared to 2005-2006, δ 13 C DIC values collected in 2010-2012 are lower, suggesting a potential impact of anthropogenic carbon uptake. This would then require the correction of δ 13 C DIC observations by this anthropogenic signal to compose the "climatological" seasonal cycle. If this change in δ 13 C DIC (decrease) represents the oceanic 13 C Suess effect only, this should also be observed on DIC, with an expected theoretical increase of 5-7 µmol kg −1 in winter (Corbière et al., 2007;Metzl et al., 2010). Figure 3a highlights a strong and negative link between δ 13 C DIC and DIC concentration in the NASPG over both sampling periods. Lower δ 13 C DIC in winter are associated with higher DIC concentrations, conversely for summer. According to this figure, August 2010 is clearly outside the general trend observed here. This is a linear relationship from winter to winter, nevertheless with a difference in the values   , we notice an increase in DIC by 7 to 10 µmol kg −1 over the 5-to 6-year period. This is much larger than those expected from the anthropogenic effect (∼ 5-7 µmol kg −1 ). To remove natural variation due to a potential salinity anomaly (Fig. 2, top panel), we evaluated the trend for salinity normalized DIC (nDIC reported in Table 1) of 1.3 µmol kg −1 yr −1 during this period (the 5-to 6-year period). This suggests a significant additional contribution of the natural variability, thus masking the impact of the anthropogenic carbon uptake in the NASPG region. This comment is supported by the observed time rates of change in δ 13 C DIC (−0.04 ‰ yr −1 ) that are much larger than those observed by Quay et al. (2007) in the NASPG (−0.017 ‰ yr −1 over the 1993-2007 period), Sonnerup and Quay (2012) in the Labrador Sea (−0.014 ‰ yr −1 over the 1970-1995 period), Olsen et al. (2006) in the Nordic seas (up to −0.024 ‰ yr −1 over the 1981-2003 period) and by Keeling et al. (2010) in the atmosphere (−0.024 ‰ yr −1 over the 1985-2008 period). As the residence time of the surface water (∼ 1 yr) is shorter than the time needed for CO 2 to reach isotopic equilibrium with the atmosphere (∼ 10 yr; Broecker and Peng, 1974;Lynch-Stieglitz et al., 1995), this large decrease in δ 13 C DIC cannot be attributed to the anthropogenic signal only. In addition, the mixed layer depth time series depicted in Fig. 4b reveals two winter episodes, with the deepest layers between 2005 and 2012, which should be associated with a higher DI 12 C input of subsurface water, thus contributing to the δ 13 C DIC decrease during this period (and DIC increase). As we suspect a large impact of the natural variability that we cannot estimate, observations are not corrected for an anthropogenic signal and the mean δ 13 C DIC seasonal cycle will be composed for each detected period. Study of the relationships between δ 13 C DIC and DIC over both sampling periods reveals, in fact, another important signal to compose and interpret the δ 13 C DIC seasonal cycle. According to Fig. 3b

δ 13 C DIC versus DIC in NASPG
Based on a Student's t test and its associated error, there is no significant difference between the slopes of regression 1 (Eq. 6) and 3 (Eq. 8). These are also equivalent to those expected if the soft-tissue pump (photosynthesis/respiration/remineralization) were the only mechanism affecting DIC and δ 13 C DIC seasonal distributions (−0.011 ‰ (µmol kg −1 ) −1 with δ 13 C org of −24 ‰ and r c:n of 117 : 16). As a result, the slope of regression 2 (Eq. 7), estimated from August 2010 to March 2011, is significantly different from the other two (−0.007 ‰ (µmol kg −1 ) −1 ). This is influenced by the August 2010 DIC values that are the lowest concentrations reported in this study (Table 1, as well as for nDIC) but are not associated with the highest δ 13 C DIC as expected. Indeed, high productivity recorded in the whole region during the summer of 2010 (Fig. 4a) would explain the low DIC values (2054.40 ± 9.92 µmol kg −1 , Table 1) but, according to previous results, an enhanced photosynthetic activity would increase δ 13 C DIC linearly with a ratio of −0.011 ‰ (µmol kg −1 ) −1 to reach a maximum value. This suggests that additional processes also contribute to DIC and δ 13 C DIC seasonal distributions during this period, thus changing their relationship. The estimate of the isotopic tracers ( δ 13 C bio and δ 13 C phys ) might be clarified to interpret this peculiar situation.

The period July 2005-June 2006
Based on satellite information (the monthly SST composite of MODIS Aqua 9 km, disc.sci.gsfc.nasa.gov/giovanni), the SURATLANT data set (NO − 3 and DIC) and Eq. (6), we built the climatological seasonal cycle of SST, NO − 3 , DIC and δ 13 C DIC for 2005-2006 (Fig. 5). Because DIC concentrations have slightly changed during this period (by less than the climatological cycle standard deviation; Corbière et al., 2007;Ullman et al., 2009), we chose to consider the available DIC (and NO − 3 ) data set sampled between 2001 and 2006, which is much more complete for composing the seasonality. The observations collected on board M/V Skogafoss during this period fit well within this climatology (Fig. 5b  and c).  in August during the summer bloom (Antoine et al., 2005;Corbière et al., 2007). Based on Eqs. (4) and (5), processes controlling this seasonality could be identified. Table 2 summarizes the results. The standard deviation displays the meridional gradient reported previously. Positive δ 13 C bio is associated with phytoplankton activity, which preferentially uses the light carbon isotope during the photosynthetic mechanism ( NO W→S 3 < 0 µmol kg −1 , nutrient consumption), whereas negative δ 13 C bio highlights the remineralization process, which releases much more DI 12 C ( NO W→S 3 > 0, nutrient input). Based on 2005-2006 values (Fig. 5), we estimate a δ 13 C bio of 0.74 ± 0.12 ‰ for the winter to summer amplitude of 0.77 ± 0.07 ‰. During this period, the δ 13 C DIC seasonality observed in the NASPG surface waters is largely controlled by biological activity, which increased δ 13 C DIC values during summer (photosynthesis) and decreased them during winter (remineralization). These high concentrations of NO − 3 and DIC related to remineralization are attributed to the winter entrainment of subsurface water (Takahashi et al., 1993).
According to the previous section, data from November 2006 do not get into line with other data collected in [2005][2006]. Results obtained from isotopic tracers indicate that physical processes also contribute to controlling the δ 13 C DIC and DIC distributions from summer to November 2006 ( δ 13 C phys = −0.23 ± 0.15 ‰; Table 2). As no anomaly was detected in November 2006 (e.g., SST, MLD), this is difficult to assess further.

The period August 2010-March 2011
There were only three cruises conducted between August 2010 and March 2011 (Table 1), but δ 13 C DIC were measured during winter and summer, when maximal and minimal values are expected. According to this comment, we cannot compose the mean δ 13 C DIC seasonal cycle for this period but, we can evaluate a seasonal amplitude of 0.70 ± 0.10 ‰ for δ 13 C DIC and of 92.9 ± 5.9 µmol kg −1 for DIC concentration. The biological and physical δ 13 C DIC signature estimated from summer 2010 to winter 2011 is much higher (−0.97 ± 0.06 ‰ and 0.27 ± 0.13 ‰, respectively; Table 2) than those obtained from the 2005-2006 data set. This is in agreement with the hypothesis of high productivity that is masked by additional processes.
The extensive winter mixed layer observed in 2009 (Fig. 4b) could have lowered δ 13 C DIC at the sea surface and persisted through to August 2010. The lack of δ 13 C DIC measurements in the winter of 2010 precludes estimating δ 13 C phys directly. If photosynthesis were the only process controlling the seasonal amplitude, as was observed during the 2005-2006 cycle, the δ 13 C DIC values of the winter of 2010 would be around 0.36 ‰ (δ 13 C August 2010 DIC + δ 13 C W→August 2010 bio = 1.36-1.00). However, such low values have not yet been observed in the NASPG (Quay et al., 2007;Racapé et al., 2013) and are unlikely.
t 1 to t 2 δ 13 C DIC ± SD (‰) δ 13 C bio ± SD (‰) δ 13 C phys ± SD (‰) 0.2 ‰ at the isotopic equilibrium conditions, thus masking the biological signal on δ 13 C DIC . In the subpolar region, the surface δ 13 C DIC is never in isotopic equilibrium with the atmospheric δ 13 CO 2 . Nevertheless, if we assume that change in SST affects δ 13 C DIC substantially at the seasonal scale in this region, this is in opposite effect to the uptake of 13 C poor CO 2 in surface water. Data collected in December 2010 are relatively higher in δ 13 C DIC and lower in DIC than those sampled in December 2011. This suggests either that the impact due to the change in SST in 2010 is dominated by the CO 2 uptake effect, or that additional physical driving processes, like horizontal transport, influenced the δ 13 C DIC and DIC seasonal distributions in 2010.
In the subpolar region, phytoplanktonic bloom is regularly dominated by the coccolithophore species, which are promoted by specific environmental conditions like high SST and shallow MLD ( Fig. 4b; Raitsos et al., 2006) observed during this period. Coccolithophores are photosynthetic calcifying algal species. As we mentioned in Sect. 3.2, a small 13 C fractionation occurs during the formation of the calcium carbonate coccoliths so that it is insignificant for the δ 13 C DIC variation. Coccolithophores bloom starts generally in June and would induce by their soft-tissue pump a winter to summer increase in δ 13 C DIC and decrease in DIC according to a ratio of −0.011 ‰ µmol kg −1 and by their carbonate pump, an additional decrease in DIC without change in δ 13 C DIC (conversely from summer to winter), thus impacting the seasonal relationships observed between these both parameters over the period August 2010-March 2011. There is no information on δ 13 C DIC in the spring of 2010, and the cloudy conditions during summer precluded observations with satellite imagery, so that it is difficult to confirm this hypothesis.
According to regression slope 2 (Eq. 7), the lowest DIC values associated with low δ 13 C DIC and the isotopic tracers ( δ 13 C bio and δ 13 C phys ), DIC and δ 13 C DIC seasonal distributions may be controlled by these combined processes.

The period March 2011-September 2012
In 2011-2012, the seasonal minimum was sampled two times but the seasonal maximum was never reached (Table 1). As all DIC data collected in the NASPG region between March 2011 and September 2012 are presented in this study, it is impossible to complete the seasonal DIC cycle like in [2005][2006] and to deduce that of δ 13 C DIC (Fig. 5). Based on 2012 values (from March 2012 to June 2012; Fig. 5), we can nevertheless observe a seasonal δ 13 C DIC amplitude of at least 0.68 ± 0.14 ‰, with a biological contribution of 0.76 ± 0.11 ‰ (Table 2). These results are comparable to those observed in [2005][2006], suggesting that the soft-tissue pump was the major driving process of δ 13 C DIC seasonality in the NASPG region.

Conclusions
In this study, we present new sea surface observations of δ 13 C DIC and associated hydrological and biogeochemical properties (DIC, NO − 3 ) in the NASPG (53 • N-62 • N) based on 14 transects conducted in 2005-2006 and 2010-2012. During all transects, we observe a positive north-south gradient coherent with model studies of Gruber et al. (1999) and Tagliabue and Bopp (2008), and opposite to DIC gradients (high DIC in the north, low in the south). We also detect a very large range of δ 13 C DIC depending on the periods. Coupling  Biogeosciences, 11, 1683-1692, 2014 www.biogeosciences.net/11/1683/2014/ During these both periods, the seasonality of δ 13 C DIC was investigated taking into account the observed changes and was totally computed for the years 2005-2006. Our "climatological" seasonal cycle shows a large δ 13 C DIC variation of 0.77 ± 0.07 ‰ (larger in summer, lower in winter), which represents the largest seasonal amplitude recorded at the sea surface in this region by Gruber et al. (1999) and built by Jonkers et al. (2013). This is coherent with large DIC (70 µmol kg −1 ), nutrients and pCO 2 seasonal signals previously observed in the region (Corbière et al., 2007). A diagnostic process model suggests that the biological processes control the δ 13 C DIC seasonality, thus masking air-sea CO 2 exchange and the ocean dynamics effect. The comparison between the slope obtained from the δ 13 C DIC -DIC relationship and those expected if the soft-tissue pump is the only driving process emphasized the importance of the biological pump in DIC seasonality in the subpolar North Atlantic region.
The δ 13 C DIC -DIC relationship has also helped to identify a third period (from August 2010 to March 2011) characterized by a lower regression slope, significantly different from the other two. This may be due to a combination of processes such as the formation of a carbonate shell associated with a coccolitophore bloom, the temperature effect on the carbon isotopic fractionation, and the ocean dynamics processes. Coccolithophore bloom being common in the NASPG surface water, it will be interesting to estimate the relationship between δ 13 C DIC and DIC in such a bloom in order to validate one of these hypotheses and to better understand the carbon cycle evolution in this region.
The observed δ 13 C DIC seasonal cycle now being well established, a challenge for future analysis will be to investigate this seasonality in ocean models. We strongly recommend maintaining such observations in the future to better separate natural and anthropogenic signals. The δ 13 C DIC is also able to estimate the oceanic δ 13 C in disequilibrium with the atmosphere (δ 13 C des ) used in the atmospheric inversion models to evaluate CO 2 flux. This study has revealed a large seasonal variability of δ 13 C DIC that strongly influenced the δ 13 C des seasonality, although we have observed a winter to summer change in SST of 6 • C. Taking into account this variability would improve the CO 2 flux uncertainty due to the evaluation of the δ 13 C des . This highlights the need to continue measuring both parameters to investigate the anthropogenic carbon signals, as well as to better constrain atmospheric inversions and validate ocean carbon models.