By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 This study determines the temporal changes of wintertime surface ocean partial pressure of CO2 (pCO2SW) in the eastern subpolar North Atlantic (esNA) (50–64°N; 32–10°W) by using data of carbon-system parameters and chlorofluorocarbon-12 acquired in 1993. Wintertime pCO2SW and its temporal trend from early 1970s through to the late 1980s were reconstructed through the application of a back-calculation method that isolates surface variations which have been transmitted to the ocean interior during the formation of Subpolar Mode Water. Our computations suggest a pCO2SW growth rate (3 μatm/yr) which is twice as large as that of atmospheric pCO2, 1.47 μatm/yr. The sensitivity of the estimated pCO2SW growth rate to remineralization ratios as well as to the CFC-12-derived ages is discussed. Cooling and northward advection of surface water equilibrated with the increasing atmospheric CO2 is suggested as the process responsible for the excessive pCO2SW growth rate.
 It is generally accepted that the North Atlantic (NA) is an important sink region for atmospheric carbon dioxide (CO2). Assuming a constant wind field, the strength of the sink is proportional to the CO2 disequilibrium across the air-sea interface ΔpCO2 = pCO2atm − pCO2SW, i.e., the difference between sea surface and atmospheric CO2 partial pressures, pCO2SW and pCO2atm, respectively. In the subpolar NA, extensive heat loss and primary production lower pCO2SW so that a disequilibrium exists throughout most of the year and the flux of CO2 is directed into the ocean. Additionally the pCO2atm is currently increasing so that an increasing air-sea CO2 disequilibrium is expected, and this is often indicated by ocean carbon models [e.g., Wetzel et al., 2005], and has been an underlying assumption employed when compiling the global air-sea CO2 flux climatology [Takahashi et al., 2002]. However, the contrary has been suggested in a number of recent studies. Anderson and Olsen  employed a conceptual advective model to calculate the change of the air-sea CO2 flux in the NA since pre-industrial times. For latitudes north of ∼50°N, their computation suggested a slightly decreasing ΔpCO2. Lefévre et al.  analyzed pCO2SW data acquired in the region and showed that the disequilibrium decreased at an annual mean rate of −0.3 μatm yr−1 over the period 1982–1998. And Friis et al.  provided measurement-based estimates of anthropogenic carbon (Cant) accumulated in the subpolar NA over the period 1981–1999. For the surface water, they found that the increase of the Cant was only reconcilable with a decreasing ΔpCO2.
 In this paper, we reconstruct the time history of wintertime ΔpCO2 in the eastern subpolar NA (50–64°N; 32–10°W) (Figure 1) over the time period 1972–1989.
 Our approach takes advantage by the fact that mode water characteristics in the interior of the ocean reflect variations of wintertime sea surface properties in the formation region [e.g., Hanawa and Talley, 2001]. The winter mixed layer in the eastern subpolar NA (esNA) is dominated by a type of Subpolar Mode Water (SPMW) and during summer this is isolated from the surface by a seasonal pycnocline [McCartney and Talley, 1982]. We assume that climatological wintertime (January–March) surface seawater properties can be used to define the SPMW that forms in the esNA during winter. This water type is henceforth referred to as eastern Subpolar Mode Water (eSPMW) in order to differentiate it from the broader term SPMW.
 By using data acquired during the third leg of the 1993 Ocean Atmosphere Carbon Exchange Study (OACES) cruise of the R/V Malcolm Baldrige (18–29 August 1993) (Figure 1) we identify eSPMW that formed in the esNA at times determined from their apparent CFC-12 ages. We then extend the back-calculation approach of Brewer  to estimate preformed (i.e., wintertime) pCO2SW values for the eSPMW encountered during the cruise. The results are combined with data for the atmospheric mole fraction of CO2 (xCO2) to compute ΔpCO2 and the temporal variability is analyzed.
 The Baldrige data have been described in detail by others [Körtzinger et al., 2003, and references therein] and were made available by the Global Ocean Data Analysis Project (GLODAP) [Key et al., 2004]. Data for salinity (S), temperature (T), total dissolved inorganic carbon (CT), total alkalinity (AT), phosphate (P), silicate (Si), oxygen (O2) and CFC-12 ages (see Key et al.  for the computation of these ages) were retrieved from the database (http://cdiac.esd.ornl.gov/oceans/home.html, accessed on March 4, 2005).
 The accuracy of the CT, AT, and O2 data have been determined to 1.5, 2.5, and 1 μmol kg−1, respectively [Körtzinger et al., 2003, and references therein]. A value of 7.5 μmol kg−1 were substracted from the O2 data because these data have been double corrected (first after Wanninkhof et al.  and then after Gouretski and Jancke ) (R. Key, personal communication, 2005). After this re-correction, apparent oxygen utilization (AOU = O2 saturation − O2 measured) was recomputed.
 Wintertime climatological ranges of surface density (σ), salinity, and temperature for the esNA were determined using data from the World Ocean Atlas 2001 [Conkright et al., 2002]. The resulting ranges (Figure 2) of temperature (5.5–11.2°C), salinity (34.93–35.49), and density (27.06–27.65 kg m−3) agree well with those reported by McCartney and Talley , for the winter mixed layer of the esNA and were used to define eSPMW. All water samples with properties falling within the climatological ranges were identified in the cruise data and assumed to be remnants of eSPMW that formed in the esNA at times given by their apparent CFC-12 ages.
 The preformed pCO2SW values for the eSPMW were determined from preformed AT (AT0) and CT (CT0). The former was computed from the empirical relationship:
which was determined by regressing surface (depth < 200 m) salinity with alkalinity. The coefficient of determination (r2) was 0.79 and the standard error of estimate was ±3 μmol kg−1. CT0 was computed according to Körtzinger et al. :
where CT and AT are the measured concentrations, RC:O and RN:O are the remineralization ratios between CT and O2, and between NO3 and O2. We used the RC:O and RN:O values which were estimated by Körtzinger et al.  for the North Atlantic. Alternative values will affect our results as discussed in section 5.
 Preformed pCO2SW was then computed from CT0, AT0, P, Si, S, and potential temperature (θ), using the constants of Merbach et al.  refitted by Dickson and Millero . These constants have been recommended by Lueker et al.  for the computation of pCO2SW from CT and AT. By propagating the uncertainties in CT0 (= 3.3 μmol kg−1) and AT0(= ±3 μmol kg−1) the maximum uncertainty in the computed pCO2SW values was obtained to be ±7 μatm (or 2 %).
 The Baldridge cruise was carried out during August and, in order to limit seasonal biasing effects, only data from deeper than 200 m were used for the temporal analysis.
 Values of pCO2atm were calculated from xCO2 data measured at Niwot Ridge, Colorado, USA which have been made available by the Carbon Cycle Greenhouse Gases Group at the National Oceanic and Atmospheric Administration (NOAA), USA. Monthly xCO2 values were retrieved (ftp://ftp.cmdl.noaa.gov/ccg/co2/flask/month/), averaged over January to March, and converted to pCO2atm according to:
where SLP is the monthly sea level pressure averaged over January to March for each year, is the wintertime water vapor pressure and was computed from θ according to Cooper et al. . The SLP data originate from the NCEP/NCAR reanalysis project [Kalnay et al., 1996] and were obtained from the IRI/LDEO Climate Data Library (http://ingrid.ldeo.columbia.edu/).
 Finally, ΔpCO2 was computed for each sample as the difference between pCO2atm and pCO2SW.
Figure 3 depicts the spatial distribution of eSPMW during the Baldrige cruise. From 46°N and northward, eSPMW was found at each station throughout the 200–1000 m layer. South of 45°N the presence of eSPMW was sporadic both with respect to station and depth, indicating that eSPMW indeed forms in the subpolar region.
 There were no trends in wind speed (not shown) so over the study period, the waters embodying the winter mixed layer of the esNA (i.e., eSPMW) changed from being strong sink to weak sink of atmospheric CO2. This can be appreciated from Figure 4 which show wintertime pCO2atm and the estimated values of pCO2SW and ΔpCO2 as a function of formation year (= 1993 − CFC-12 age) of eSPMW. Evidently, the eSPMW was undersaturated with respect to atmospheric CO2 throughout the study period (Figure 4a). However, this disequilibrium decreased consistently over the years and nearly vanished at the end of the study period (Figure 4b). Straight lines fitted to the data (Figures 4a and 4b) suggest linear trends of 1.47, 3.00, and −1.53 μatm yr−1 for pCO2atm, pCO2SW and ΔpCO2, respectively.
 Our computations suggest that the wintertime pCO2SW in the eastern subpolar North Atlantic increased at a greater rate than the atmospheric pCO2 so that the disequilibrium decreased between 1972 and 1989. This is in qualitative agreement with Lefévre et al. . Quantitatively, however, the pCO2SW growth rate estimated in this study (3.00 μatm yr−1) is larger than the ∼1.8 μatm yr−1 estimated by Lefévre et al.  for the months January–March. The difference in growth rates can be reconciled as our estimate, being based on reconstructions rather than observations, is sensitive to the choice of RC:O and RN:O values as well as to the CFC-12 ages.
 If we were to use a lower value of RC:O then the estimated pCO2SW growth rate would decrease because the correction for remineralized carbon (= − RC:O × AOU, see equation (2)) would be reduced more for “older” eSPMW than for “younger” eSPMW. Changing RN:O has a similar effect on the estimated pCO2SW growth rate, but the effect is much smaller than that of RC:O because RN:O × AOU is an order of magnitude smaller than RC:O x AOU. There are several alternative remineralization quotients available [e.g., Redfield et al., 1963; Takahashi et al., 1985; Anderson and Sarmiento, 1994], but only those of Redfield et al.  are applicable to our study area and at the same time free from known biases. These latter quotients, however, are higher than those of Körtzinger et al.  and give a pCO2SW growth rate of 3.2 μatm yr−1.
 The use of apparent CFC-12 ages implicitly assumes that the surface-to-interior spreading of water can be estimated from a single transit time. This assumption is strictly only valid for the limiting case of no mixing and, in order to account for mixing, Waugh et al.  applied the concept of transit time distribution (TTD) in the subpolar North Atlantic. They found that CFC-12 ages are smaller than the mean transit times (mean ages, henceforth) for much of the gyre. Being unable to constrain by how much CFC-12 ages underestimate the mean ages they provided several possible relationships between CFC-12 ages and mean ages [Waugh et al., 2004, Figure 6b] by choosing the ratio Δ/Γ (which denotes the width of the TTD divided by the mean age) equal to either 0, 0.75, 1, and 1.25. Thus, the only certain consequence for our results is that since the CFC-12 ages are a lower limit of a range of possible ages the estimated pCO2SW growth rate is an upper limit. This is so because increasing the ages implies that the total pCO2SW increase of ≈50 μatm (Figure 4a) would be distributed over a longer time period and, thus, the growth rate would be reduced.
 In order to evaluate the combined effect of the loosely constrained remineralization quotients and the possible underestimation of water mass ages on the pCO2SW growth rate we (i) utilized the remineralization quotients of both Körtzinger et al.  and Redfield et al.  for the determination of pCO2SW, (ii) replaced the CFC-12 ages with mean ages obtained through employing the relationships of Waugh et al.  and (iii) re-estimated the pCO2SW growth rate. Important to mention is that, when the mean ages obtained by setting Δ/Γ equal to 0.75, 1, or 1.25 were used, the temporal trend of pCO2SW became quadric so that the data best fitted to equations of the form:
where year = 1993 − mean age and A, B, and C are constants. For these cases, therefore, the pCO2SW growth rate was determined as the mean value of the time derivative of equation (4) evaluated for the years 1972 through to 1989. The re-estimated growth rates are depicted in Figure 5 and fall between 1.36 μatm yr−1 and 3.2 μatm yr−1. This suggests that during 1972–1989 pCO2SW in the esNA increased at a rate around or above the atmospheric pCO2 growth rate of 1.47 μatm/yr. Figure 5 also indicate that the 1.8 μatm yr−1 growth rate estimated by Lefévre et al.  would be obtained by using a Δ/Γ value between 0.75 and 1.
6. Conclusions and Further Remarks
 Using an extended back calculation approach we have shown that wintertime pCO2SW in the surface waters of the eastern subpolar North Atlantic increased between 1972 and 1989. Our approach resulted at first hand in a pCO2SW growth rate estimate of 3 μatm yr−1. After taking the caveats of the analysis into account we have shown that this is an upper limit, and a lower limit is close to the atmospheric growth rate of 1.47 μatm yr−1.
 With respect to the cause of the trend of increasing pCO2SW, no single process can be pointed out. The estimated pCO2SW depends mainly on preformed concentrations of total dissolved inorganic carbon (CT0) and alkalinity (AT0), potential temperature (θ), and salinity (S). Of these four parameters only CT0 exhibited a significant temporal trend (not shown) which, in turn, have caused the trend of increasing pCO2SW. There are several processes that may have caused the increase in CT0. Analysis of NCEP/NCAR reanalysis data reveal no increase in the mean wind speed during the winter season so it is unlikely that enhanced air-sea CO2 flux into the study area increased CT0 during the study period. Similarly, a long term decrease in the amount of carbon fixed into organic matter and/or increased remineralization is also unlikely because the preformed phosphate (P0 = P − RC:O × AOU) did not show any significant trend (not shown). We suggest that uptake of anthropogenic carbon is the main cause of the estimated pCO2SW increase. Especially, since this process can produce excessive pCO2SW growth rates in the North Atlantic as shown by Wallace  and Anderson and Olsen . The reason for this is that the surface water feeding this region originates mainly further south where temperatures are higher and buffer capacities larger. As water with any given CT change are moved northward the corresponding pCO2SW change will be amplified since the buffer capacity decreases due to cooling induced CT uptake. Therefore if the pCO2SW in the south changed at the same rate as the atmosphere, greater than atmospheric increases can take place in the north.
 This is a contribution to the EU IP CARBOOCEAN (contract 511176-2) and publication A120 of the Bjerknes Centre for Climate Research. Comments from Karsten Friis and a second anonymous referee helped improve the manuscript.