The meridional WOCE line A14, just east of the South Atlantic Mid-Atlantic Ridge, was surveyed during the austral summer of 1995 from 4°N to 45°S. Full-depth profiles of pH, total alkalinity (TA), and total inorganic carbon (CT) were measured, allowing a test of the internal consistency of the CO2 system parameters. The correlation between CT measured and calculated from pH and TA was very good (r2 = 0.998), with an insignificant average difference of 0.1 ± 3.0 μmol kg−1 (n = 964 data). CO2 certified reference materials (CRMs) and a collection of selected samples subsequently analyzed at the Scripps Institution of Oceanography were used to assess the accuracy of our measurements at sea with satisfactory results. The three measured CO2 system variables were then used to identify the characteristic array of zonal flows throughout the South Atlantic intersected by A14. Equatorial, subequatorial, subtropical, and subantarctic domains were identified at the depth range of the surface water, South Atlantic Central Water (SACW), Antarctic Intermediate Water (AAIW), Upper Circumpolar Water (UCPW), North Atlantic Deep Water (NADW) and Antarctic Bottom Water (AABW). The nonconservative CO2 system parameters (pH, TA, CT) have been useful in identifying the transition from aged subequatorial to ventilated subtropical surface, central and intermediate waters. They have been identified as good tracers of the zonal circulation of NADW, with marked flows at the equator, 13°S, and 22°S (the “Namib Col Current”) and the sharp transition from UNADW to UCPW at 23°S. The anthropogenic CO2 inventory (CANT) was estimated and compared with CFC-derived apparent ages for different water masses along A14. The anthropogenic entry reached maximum in the relatively young and ventilated subantarctic and subtropical domains where AAIW was the most efficient CO2 trap. The calculated annual rate of CANT entry by AAIW was 0.82 μmol kg−1 y−1, in agreement with the annual rate estimated from the equilibrium between the atmospheric pCO2 increase and the upper mixed layer.
 During the last decade, many oceanographers focused their efforts on global ocean circulation in relation to climate. They began the cooperative “World Ocean Circulation Experiment” (WOCE) to obtain high-quality hydrographic data complemented with tracer measurements. Since ∼30% of the anthropogenic emissions of the greenhouse gas CO2, seem to have been trapped by the oceans [Watson et al., 1995], knowledge of the CO2 system parameters, partial pressure (pCO2), pH, total alkalinity (TA), and total inorganic carbon (CT), is essential for understanding the regulating action of the ocean on climate. Using the appropriate dissociation constants for carbonic acids [Millero, 1995], only two of the four parameters are needed to determine the CO2 system. Simultaneous measurements of at least one of the remaining two parameters allow a test of the internal consistency. However, field studies including measurements of more than two parameters are scarce [Millero et al., 1993; Clayton et al., 1995; Lee et al., 1997; McElligott et al., 1998; Johnson et al., 1999].
 Potentiometric pH and TA and coulometric CT were measured at sea along WOCE line A14, in the Eastern Equatorial and South Atlantic (Figure 1). CO2 “certified reference materials” (CRM) were used to test the accuracy of our measurements. In addition, duplicate analyses were performed to examine the precision of these methods. High internal consistency, accuracy, and precision of CO2 parameters during the cruise allow us to characterize (1) the equatorial, subequatorial, subtropical, and subantarctic regimes imposed by the climatologically driven circulation in the upper ocean [Peterson and Stramma, 1991] and (2) the complex advection and mixing of intermediate, deep, and bottom waters of Antarctic origin with deep waters from the north Atlantic that exhibit opposing flow at different depths [Reid, 1989; Peterson and Whitworth, 1989].
 The initial concentrations of the CO2 parameters in source regions of different water masses are modified by the oxidation of organic materials and the dissolution of carbonaceous skeletons [Broecker and Peng, 1982; Takahashi et al., 1985; Anderson and Sarmiento, 1994]. In addition, the initial pCO2, pH, and CT of any water mass depend on the CO2 levels in the overlying atmosphere, which have increased from preindustrial reference value of 278 ppm to 359 ppm in 1995 [Keeling et al., 1995]. The total amount of anthropogenic CO2 dissolved in the ocean can be estimated assuming that (1) dissolved oxygen and CO2 are close to equilibrium with the atmosphere during water mass formation and (2) alkalinity is not significantly affected by the CO2 increase [e.g., Chen and Millero, 1979; Poisson and Chen, 1987; Gruber et al., 1996]. On this basis we have examined the potential of the eastern South Atlantic as a sink for the anthropogenic CO2 emitted throughout the industrial era. Finally, surface pCO2 is compared with atmospheric concentration to establish the CO2 sources and sinks at the time of the cruise in relation to the different zones observed in the upper ocean.
 The meridional line A14 was occupied during the first leg of cruise CITHER-3 (R/V L'Atalante, 13 January to 16 February 1995) from 4°N to 45°S along ∼9°W (Figure 1). Data from a Neil-Brown Mark IIB CTD probe were acquired from the surface to 15 m above the bottom at 107 sampling stations. Water samples for chemical analyses were collected in 8 liter Niskin bottles at 30 depths throughout the water column. A comprehensive description of the cruise, sampling procedures, analytical techniques, and raw data are given by Groupe CITHER-3 [1996, 1998]. Data are also available from the WOCE Hydrographic Program Office (http://whpo.ucsd.edu/data/onetime/atlantic/a14).
2.1. Sample Analyses
 Salinity was determined from conductivity measurements with a PORTASAL salinometer using the PSS78 equation [UNESCO, 1981] and dissolved oxygen was analyzed by the Winkler method following the recommendations of Joyce and Corry . Oxygen saturation for computing “apparent oxygen utilization” (AOU) was calculated using the equation of Benson and Krause [UNESCO, 1986]. Nutrients (nitrate + nitrite, phosphate, and silicate) were analyzed by segmented flow analysis following Hansen and Grasshoff , with some modifications [Mouriño and Fraga, 1985; Álvarez-Salgado et al., 1992]. Salinity, dissolved oxygen (in μmol kg−1), and nutrients (in μmol kg−1) were determined in all samples collected. Chlorophyll a (Chl a) concentration in surface samples (in mg m−3) was estimated fluorometrically with a 10,000 R Turner fluorometer after 90% acetone extraction [Yentsch and Menzel, 1963]. Chlorofluoromethane CFC11 was measured using an extraction-trapping method in combination with gas chromatography using electron capture detection, as described by Bullister and Weiss .
 The pH was measured with a Metrohm E-654 pH meter equipped with a Metrohm 6.0233.100 combination glass electrode and a Pt-100 probe for temperature control. The system was standardized with a National Bureau of Standards (NBS) phosphate buffer (pH = 7.413 at 25°C). The Nernstian response of the electrode was tested as described by Pérez and Fraga [1987a]. Total alkalinity (TA) was determined by automatic potentiometric titration with HCl ([HCl] = 0.1310) to a final pH of 4.44 [Pérez and Fraga, 1987b], using a “Titrino” Metrohm automatic potentiometric titrator with a Metrohm 6.0233.100 combined glass electrode and a Pt-100 probe. The [HCl] was potentiometrically titrated versus borax (B4O7Na · 10H2O) to a final pH of 5.3 ± 0.05 [Pérez and Fraga, 1987b]. The pHs were measured at lab temperature and referred to 15°C (pH15) using the parameterisation of Pérez and Fraga [1987a], and TA results are expressed in μmol kg−1. Full water column profiles of pH15 were produced at each of 107 stations occupied, whereas TA was determined at 39 stations. Additionally, TA of the surface sample was measured at each station.
 CT samples were collected following the procedure described by U.S. Department of Energy  and preserved immediately with a saturated mercuric chloride solution to prevent biological production or consumption of CO2. Samples were analyzed within 14 h after collection; in the meantime they were kept cool and in the dark. The CT concentration was measured using a “Single Operator Multiparameter Metabolic Analyzer” (SOMMA) [Johnson et al., 1987, 1993] coupled with coulometric detection. The seawater sample was drawn into a calibrated pipette and dispensed into a stripping chamber, where it was acidified with 8.5% phosphoric acid. The resultant CO2 was carried into a coulometric cell with high purity N2 gas (99.95%) where it was absorbed by and reacted with ethanolamine in dimethylsulfoxide. Seawater samples were collected at 42 stations to provide full CT profiles. Samples from the upper 1000 m were analyzed at 11 additional stations, to provide CT data in the main thermocline. CT and TA have been normalized to salinity 35.0 to produce the corresponding NCT and NTA parameters.
2.2. Precision, Accuracy, and Internal Consistency of Measured CO2 Parameters
 At least one pair of Niskin bottles were closed at the same depth at each sampling station (field duplicate), providing a check of shipboard methodology precision. The average absolute difference between duplicate analyses was 0.003 for pH15 and 1.1 μmol kg−1 for TA, based on 174 and 61 field duplicates, respectively. In addition, each of 30 bottles of the rosette sampler were closed at ∼3000 m depth at station 45. The standard deviation of the 30 analyses of pH15 and TA were 0.002 and 0.7 μmol kg−1, respectively. Analyses of CT field duplicates were performed ten times, with an average absolute difference of just 0.3 μmol kg−1. The average absolute difference of 80 laboratory duplicate CT analyses was 0.8 μmol kg−1. These laboratory analyses were performed by R. F. Keeling at Scripps Institution of Oceanography (SIO) immediately following the cruise.
 The accuracy of the measured CO2 parameters was tested by means of “Certified Reference Materials” (CRMs) supplied by A. G. Dickson (University of California). CRMs (batch 24) were analyzed routinely every twenty samples for a total of 114 CT measurements. The average difference from the certified CT value (1987.55 μmol kg−1) was <0.94 μmol kg−1. The pH15 and TA were also measured routinely using CRMs (Table 1). The average of 76 CRM analyses for TA (2217.2 ± 1.4 μmol kg−1) was in very good agreement with the certified concentration (2216.0 ± 1.0 μmol kg−1). In addition, ten selected CT samples were collected at seven stations along the cruise to be analyzed by P. Guenther (SIO) by the manometric method (Table 2). The results reinforce the accuracy of shipboard CT measurements, with 0.9 ± 2.4 μmol kg−1 as the average difference between samples analyzed at SIO and those analyzed on the ship.
Table 1. The pH15, TA, Measured CT (CT)M, Calculated CT (CT)C and Calculated pCO2 on CRMs (Batch 24)a
TA, μmol kg−1
(CT)M, μmol kg−1
(CT)C, μmol kg−1
Standard deviation of the CRMs analyzed per day (2nd column), in brackets. The pCO2 was calculated using pH15 and TA.
The pH(NBS) calculated from CT and TA of CRM at 15°C. Independent measurements of pH(SWS) at 25°C were performed on CRMs and converted to the NBS scale at 15°C (pH15 = 8.203) using the activity coefficient of Mehrbach et al. .
The pCO2 was calculated from certified CT and TA at 15°C.
Table 2. Comparison of CT Measurements Performed by the Coulometric Method on Board and the Manometric Method at the Scripps Institution of Oceanography (SIO) at Ten Selected Samplesa
CT calculated from pH and TA measurements on board are included as well. (CT)M and CT are measured coulometrically, (CT)C and CT are calculated from pH15-TA, and (CT)SIO and CT are measured manometrically at SIO. Concentrations in μmol kg−1.
 CT was calculated from pH15 and TA using the thermodynamic equations of the carbonate system [Dickson, 1981] with the dissociation constants of Mehrbach et al. , the pKB of Lyman  recalculated by Edmond and Gieskes , and KW of Dickson and Riley . Analytical errors of ±0.003 in pH and ±1.1 μmol kg−1 in TA produce errors of ±1.2 and ±1.1 μmol kg−1 in CT, estimated from the thermodynamic equations and the average values of pH15, TA, salinity, and temperature during the cruise. Consequently, the total error from the calculation of CT is ±2.3 μmol kg−1. The regression between measured and calculated CT was very high (Figure 2). The average difference between calculated and measured CT during the cruise was 0.1 μmol kg−1, with an average error of the (CT)M estimate of ±3 μmol kg−1. Finally, calculated CT for 76 CRMs (Table 1) agrees well with the certified concentration and the average difference of the 10 CT samples measured at SIO and calculated from pH15 and TA (Table 2) was only 0.9 ± 2.6 μmol kg−1. The results of these comparisons prove the high internal consistency of our data. Consequently, good precision, accuracy, and internal consistency of shipboard CO2 measurements along line A14 make our data reliable for carbon studies in the South Atlantic.
 Surface pCO2 can be calculated from pH15–TA, pH15–CT, and TA–CT using the thermodynamic equations of the carbonate system and the acid constants described above, together with the solubility coefficient of CO2 in seawater [Weiss, 1974]. Analytical errors of ±0.003 in pH15, ±1.1 μmol kg−1 in TA, and ±0.8 μmol kg−1 in CT produced errors of ±3.5, ±0.2 and ±2 μatm in pCO2, respectively. The resultant total errors for the pH15–TA, CT–TA, and pH15–CT pairs were ±3.7, ±5.5, and ±2.2 μatm, respectively. Correlation between these three sets of calculated pCO2 values was very high (r2 > 0.98) and the average difference was within the error of the pCO2 estimate. For this study then, we have used the average of three estimates of surface pCO2. Calculated pCO2 of CRMs from pH15–TA (Table 1) had an average standard deviation of ±4.0 μatm, in agreement with the expected error of the pCO2 estimate.
3. Results and Discussion
3.1. Imprint of the Eastern South Atlantic Circulation on Carbon System Variables
 A14 is the westernmost WOCE line in the eastern South Atlantic. It is situated just east of the Mid Atlantic Ridge (Figure 1) and affected by the well-defined zonal circulation of the South Atlantic Ocean [Peterson and Stramma, 1991]. The objective of this part is the identification of the main zonal regimes (equatorial, subequatorial, subtropical, and subantarctic) at the level of surface, central, intermediate, and deep waters along line A14, by means of measured and derived CO2 system variables. The effect of this zonation on the entry of anthropogenic carbon will be assessed in section 3.2.
3.1.1. Upper Ocean
 The climatologically driven upper ocean circulation originates a marked zonal array on thermocline waters, which produces a clear impact on the thermohaline and CO2 system distributions (Figures 3, 4, 5, and 6). The 13°C isotherm, superimposed on the salinity distribution (Figure 3), was used to delimit the different zonal regimes observed along line A14 [Brea, 1998]. This isotherm is well within the South Atlantic Central Water (SACW) domain, which ranges from 10°C to 18°C [Gordon and Bosley, 1991].
 From north to south, line A14 began at 4°N (Figure 3). In the Equatorial Current System (ECS), between 4°N and 3°S, the 13°C isotherm was situated at ∼250 m depth. Poleward rising to 100 m was observed near 3–4°S, which marks the eastward flowing South Equatorial Under-Current (SEUC) at the boundary between the ECS and the Subequatorial Gyre (SEG). The 13°C isotherm was present at 125–150 m in the SEG domain (4–9°S). Poleward deepening was observed from 9°S to 23°S, tracing the northwest flowing South Equatorial Current (SEC). The 13° isotherm reached 300 m depth at 23°S. In the Subtropical Gyre (STG), between 23°S and 36°S, the isotherm was found at ∼325 m. Two V-shaped structures deepening about 200 m were evident at 26°S (stations 68–69) and 31.5°S (stations 77–79) that correspond with two eddies detached from the Agulhas current [Arhan et al., 1999]. At 36–37°S the 13°C isotherm rose abruptly from 250 to 100 m, denoting the Subtropical Front (STF). Further south, within the Subantarctic Zone, the isotherm rose to shallower depths (∼40 m) and disappeared south of 41°S.
126.96.36.199. Surface Waters
 Surface distributions of temperature, salinity, chlorophyll, and pCO2 (Figure 7) are very sensitive to the observed zonal regimes. Therefore it is the objective of this section to study how the contemporary uptake of atmospheric CO2 in the Eastern South Atlantic was affected by such zonal arrangement. The average mole fraction in dry air of CO2 (xCO2) for 1995, 359 ppm according to Keeling et al. , is represented by the dotted line in Figure 7b.
 Surface waters were pCO2 undersaturated north of the Equator, reaching a minimum of 313 μatm at 4°16′N, where maximum surface chlorophyll a was recorded (3 mg m−3). An area of constant surface pCO2 (∼330 μatm) and very low chlorophyll concentrations (∼0.08 mg m−3) appeared just south of the pCO2 minimum (Figure 7b) and extended up to 2°N. In this zone, nutrients were depleted at the surface and the nutricline was located at ∼20 m (not shown). Maximum surface temperature were recorded (>28°C) when salinity was <35 (Figure 7a). The salinity minimum and temperature maximum in this region is known as the Low-Salinity Region (LSR) and is expected to be pCO2 undersaturated [Bakker et al., 1999]. The position of the LSR corresponds to the Intertropical Convergence Zone (ITCZ), where the wind systems of the two hemispheres converge producing high precipitation, lower salinity, and higher sea surface temperature. The position of this zone varies seasonally, being close to the equator in winter and moving northward in summer [Siedler et al., 1992]. Similar results of low salinity accompanied by low pCO2 were found during several cruises executed in different seasons, suggesting that the LSR is a recurring or possibly a permanent feature [Oudot and Andrié, 1989; Lefèvre et al., 1998; Bakker et al., 1999]. Surface salinity and temperature at the LSR along A14 were 1 unit lower and 6°C higher than at 20 m depth, where pCO2 levels rose to 485 μatm. Therefore the contribution of CO2-free rainwater to the composition of surface waters in the LSR was about 3% compared with seawater at 20 m depth. This dilution produces a decrease of 65.8 μmol kg−1 in TA and a corresponding pCO2 decrease of 14 μatm (from 485 to 471 μatm). Since surface pCO2 in the LSR decreases to 330 μatm, intense net primary production must be the cause of the observed low-surface pCO2. This seems contradictory because chlorophyll concentrations in the LSR were quite low. However, it must be noted that (1) surface pCO2 does not depend only on surface chlorophyll but on its vertical distribution and (2) chlorophyll is not an index of primary production but an index of the biomass retained in the water column after grazing and sedimentation of primary producers. The fact that pH in surface waters was 0.24 units higher than at the nutricline suggests intense net primary production in the LSR. This pH increase resulted in a pCO2 decrease of 294 μatm (from 485 to 191 μatm). The difference between the observed (330 μatm) and the expected (485 − 14 − 294 = 177 μatm) surface pCO2 levels after consideration of freshwater dilution and net primary production (330 − 177 = 153 μatm) was due to the surface temperature increase of 6°C. Therefore the combination of thermohaline characteristics and biological activity can explain the low pCO2 values in the LSR at the time of A14 (January–February 1995).
 In the 0–4°S zone, salinity increased abruptly and temperature decreased slowly (Figure 7a), due to equatorial upwelling. Surface pCO2 also increased, reaching values near 400 μatm, whereas chlorophyll levels were ∼0.15 mg m−3 (Figure 7b). Despite low surface pCO2 levels in the LSR, the whole equatorial zone (4°N–4°S) was pCO2 oversaturated acting as a CO2 source to the atmosphere. The average ± std oversaturation (ΔpCO2) for the whole equatorial zone was 9 ± 29 μatm at the time of A14. Andrié et al.  and Smethie et al.  obtained similar values (8 ± 19 μatm and 11 ± 17 μatm, respectively) at the same time of the year (January–February) farther east. The extremely high ΔpCO2 (45 ± 25 μatm) recorded by Oudot et al.  was attributed to a climate condition anomaly in 1984.
 Surface salinity increased while temperature decreased along the SEG zone (4–9°S), where pCO2 values were constant (402 ± 5 μatm) and chlorophyll decreased from 0.15 to 0.03 mg m−3. The highest salinity values (36.5 ± 0.1) were observed in the SEC zone (9–23°S) in combination with relatively low temperatures (24.6 ± 0.6 °C). The conspicuous V-shaped pCO2 distribution centered at ∼13°S traced the limit between the subtropical and subequatorial regimes. A local chlorophyll maximum (0.36 mg m−3) was observed at that position, coinciding with local pCO2 minimum (383 μatm). Temperature decrease and net primary production could be the causes of these relatively low pCO2 values. It should be noted that the depth of the nutricline at the center of the SEC was located at 137 m, while in the equatorial zone it was found at 20 m depth.
 Within the STG salinity, temperature, and pCO2 decreased by 1.3, 1.3°C and 75 μatm, respectively, from 23°S to 36°S. Chlorophyll concentrations were constant below the detection limit (∼0.03 mg m−3). South of the STF (36–37°S) a sharp decrease of surface pCO2 levels was observed, reaching values below saturation. The pCO2 undersaturation was caused by low temperatures associated with Subantarctic Surface Water and increased biological activity in the Subantarctic Zone, which produced high surface chlorophyll concentrations of ∼0.4 mg m−3. Net primary production was favored in the Subantarctic Zone by a marked increase of surface nutrient concentrations; nitrate reached >15 μmol kg−1 in the southern end of A14 at the SAF (not shown). Although surface salinity in the Subantarctic and Equatorial zones was quite similar, TA clearly indicates the difference in origin of surface waters in these two regions. The TA-S plot shows two well-defined (r2 > 0.99) linear relationships north and south of 33°S (Figure 8). For any salinity <36, TA was greater to the south (Subantarctic Surface Water influence) than to the north (Equatorial LSR influence).
 More than 80% of the total variability of surface pCO2 along A14 can be explained by a multiple linear regression with surface salinity and temperature (r2 = 0.81, n = 107; p < 0.001). The remaining 20% have to be due to the local effect of net primary production. In fact, a significant negative correlation was observed between surface pCO2 and chlorophyll (r2 = 0.21, n = 106, p < 0.001). However, the multiple regression of surface pCO2 with salinity, temperature, and chlorophyll did not significantly improve the regression with the thermohaline properties (r2 = 0.82, n = 106, p < 0.001) due to the inadequacy of chlorophyll measurements to represent net primary production.
188.8.131.52. Central Waters
 The evolution of salinity and CO2 system variables (pH15, NCT, and NTA) along the 13°C isotherm (Figure 9) clearly show the effect of different zonal regimes in the SACW domain. Salinity remained constant at 35.29 ± 0.01 north of 15°S where a sharp decrease of 0.13 was observed, denoting the abrupt transition from the subequatorial to the subtropical SACW branches. The observed salinity excess of subequatorial SACW on the eastern South Atlantic could be due to regional excess evaporation in the SEG [Gordon and Bosley, 1991] or transport of the Salinity Maximum Water formed in the tropical western South Atlantic by the SECC [Mémery et al., 2000]. Salinity along the 13°C isotherm was constant at ∼35.16 from 15°S to the STF (36–37°S). South of the STF, salinity decreased abruptly coinciding with a sudden rise of the 13°C isotherm in the subantarctic zone.
 The pH15 levels along the 13°C isotherm increased southwards. The most conspicuous features were a local maximum (excess of ∼0.05 units) between 3°N and 3°S and successive steep increases of ∼0.10 units at 15°S (subequatorial-subtropical transition) and ∼0.05 units at 21–22°S (northern limb of the STG) and at 36–37°S (STF). Maximum pH15 values (8.23) were recorded within the Subantarctic Zone, coinciding with an observed salinity minimum. The evolution of NCT was opposite that of pH15 (Figure 9b), with an excess of ∼60 μmol kg−1 in the SEG compared to the STG as a result of a sudden NCT change between 15°S and 23°S. On the other hand, NTA concentrations showed a brief excess of ∼4 μmol kg−1 in the subtropical SACW domain. The observed ΔNCT/ΔNTA molar ratio of −15 between SACW at 13°C in the SEG and the STG indicates that >96% of the NCT excess in the SEG results from organic carbon oxidation rather than from CaCO3 dissolution, as expected at these shallow levels (140 m). It should be noted that the NCT excess in the SEG is maximum at 400–700 m depth, where extreme pH15 and NCT values of <7.80 and >2265 μmol kg−1 were recorded (Figures 4 and 6). The low pH15 and high NCT concentrations observed were due to the sedimentation associated with high net primary production rates in the surface water [Reid, 1989] and prolonged residence time (4.4 and 8.5 years) of thermocline waters in the SEG [Gordon and Bosley, 1991].
 The Agulhas ring centered at 26°S clearly affected salinity and pH15 distributions along the 13°C isotherm (Figure 9a). The pH15 values were 0.03 units higher than in the surrounding waters. Expected NCT and NTA minima were not observed (Figure 9b) because these variables were not measured at stations where the 26°S ring was located. On the contrary, although the 31.5°S ring was sampled, no indications of its presence appeared in any of the plotted distributions. According to Arhan et al. , the 26°S ring remained at the higher latitudes of the Agulhas retroflection region during a winter, where it experienced a large heat loss (it is 3.5°C cooler than the second ring) and ventilation, causing an increase of oxygen (13 μmol kg−1 higher than the ring centered at 31.5°S). This resulted in a net increase of pH15 and therefore an expected decrease of NCT.
3.1.2. Intermediate and Deep Waters
 Below the thermocline, carbon-rich waters of Antarctic origin mix with carbon-poor waters of North Atlantic origin. The northward spread of Antarctic Intermediate Water (AAIW) in the South Atlantic has been traditionally marked by a conspicuous salinity minimum clearly observed in Figure 3 [e.g., Wüst, 1935; Reid, 1989]. The meridional evolution of the AAIW salinity minimum denotes the sequence of zonal regimes observed in thermocline waters, as previously described in great detail by other authors [Warner and Weiss, 1992; Boebel et al., 1997]. We note here a sudden pH15, NTA, and NCT changes at the level of the AAIW salinity minimum (not shown) denoting the transition from the ventilated subtropical to the aged subequatorial AAIW domains that occurred at 23°S, i.e., 7° south of the transition in the overlying SACW domain.
 The northward spread of the aged Upper Circumpolar Water (UCPW) can be traced by the meridional evolution of the NTA and NCT maxima (Figures 4, 5, and 6) observed just below the AAIW salinity minimum at the southern end of A14. UCPW was present at ∼1400 m from the SAF to the STF with NTA and NCT concentrations of 2364 and 2265 μmol kg−1, respectively. Equatorward rising of these extrema occurred throughout the STG up to ∼1000 m at 23°S, where NTA and NCT concentrations were 2350 and 2250 μmol kg−1, respectively. North of 23°S the NTA and NCT extrema did not occur at the same depth. The NCT maximum coincided with the AAIW salinity minimum centered at 800 m. The NTA maximum (∼2342 μmol kg−1) remained at 1000 m, coinciding with the silicate maximum (∼34 μmol kg−1) and the CFCs minimum (∼0.01 pmol kg−1) in the subequatorial and equatorial domains (not shown). The presence of appreciable volumes of UCPW north of the subtropical-subequatorial transition is still a matter of open discussion. McCartney , Andrié et al. , and Oudot et al.  identified the property extrema at 1000 m depth as UCPW. Conversely, Tsuchiya et al.  considered that they are really part of the AAIW domain, and Larqué et al.  suggested that the UCPW progresses northward to about 24°S in the western basin and most of it returns southwards in the eastern basin. As a consequence of this controversy, Stramma and England  give no circulation scheme in the most recent review of the South Atlantic circulation.
 The Deep Western Boundary Current (DWBC) transports North Atlantic Deep Water (NADW) from the Northern Hemisphere into the South Atlantic. Frequently, the NADW is separated into three layers: upper, middle, and lower NADW [e.g., Wüst, 1935; Friedrichs et al., 1994; Stramma and England, 1999]. UNADW along A14 can be clearly distinguished by extreme values of all parameters (S > 35, pH15 > 7.99, NTA < 2325 μmol kg−1, NCT < 2190 μmol kg−1) at 1500–2000 m depth in the Equatorial domain (Figures 3, 4, 5, and 6). Eastward deflection of the southward flowing DWBC along and just south of the equator has been observed in the NADW domain [Weiss et al., 1985; Richardson and Schmitz, 1994; Rhein et al., 1995; Andrié, 1996; Mercier and Arhan, 1997; Arhan et al., 1998]. A uniform layer of S > 34.9, pH > 7.97, NTA < 2350 μmol kg−1 and NCT < 2200 μmol kg−1 was observed below the core of UNADW, which correspond to the deeper branches of NADW. A bend in the θ-S at ∼2°C (not shown), coined as the Two-Degree Discontinuity (TDD) by Broecker et al. , constitutes the lower limit of depth of the NADW. This limit is at ∼4000 m on the northern end of A14, indicating that the deepest branches of NADW cross the Mid-Atlantic Ridge (MAR) through the Romanche Fracture Zone to enter the Eastern South Atlantic [Friedrichs et al., 1994; Mercier and Morin, 1997].
 An isolate NTA minimum (<2330 μmol kg−1) at 1400–1700 m near 13°S (Figure 5) depicts the presence of a zonal jet transporting UNADW. According to Stramma and England , NADW flows westward at ∼15°S as a result of recirculation at the eastern flank of the MAR UNADW and UCPW encountered at 23°S, where they form a marked front that is easily observed in the distributions of all the study properties (Figures 3, 4, 5, 6, and 10). As a consequence, NADW flows to the eastern basin through the Rio de Janeiro (22°S) and the Río Grande (26°S) Fracture Zones of the MAR [Mercier et al., 2000], and is deflected eastwards to meet a zonal jet that Warren and Speer  called the “Namib Col Current” near 22°S in the Angola Basin. The existence of an eastward flow near 25°S centered at 2000 m, which transports NADW from the western South Atlantic, was first suggested by Reid  and corroborated by Speer et al. . Local maxima of salinity and pH and minima of NTA and NCT at ∼22°S in the UNADW domain (∼1800 m) traced the “Namib Col Current” along A14. To the south of 23°S, the deeper branches of NADW can be observed, centered at 2400 m, by the distributions of the measured variables (Figures 3, 4, 5, 6, and 10).
 Antarctic Bottom Water (AABW) along A14 consisted exclusively of Lower Circumpolar Water (LCPW) since the >46.04 σ4 characteristic density of Weddell Sea Deep Water (WDSW) was not observed. AABW can be distinguished in the southernmost side of the section by low pH15 (7.881–7.915) and high NTA (2370–2380 μmol kg−1) and NCT (2260–2270μmol kg−1) values in the bottom layer, spreading north to 36°S. At this point, the Walvis ridge limits the northward extension of AABW. Minor volumes of this water mass were observed in abyssal depths (>4000 m, <2°C) at the equator, which enter from the western basins through the Romanche Fracture zone and can be traced by high NTA (2355–2360 μmol kg−1) and NCT (2210–2215 μmol kg−1) concentrations.
3.2. Anthropogenic CO2
 The method used for calculating anthropogenic carbon (CANT) is based on a modification of the back-calculation technique proposed by Brewer  and Chen and Millero , further modified by Gruber et al.  and improved by Pérez et al. . In this technique, correction of the effect of organic carbon oxidation and CaCO3 dissolution processes on the distribution of CT allows the determination of the initial CO2 (CT0) of different water masses in source regions. In addition, assuming that dissolved oxygen and CO2 are close to equilibrium (or with the same disequilibrium) with the atmosphere during waters mass formation, CT0 has progressively increased since the preindustrial era. Conversely, TA0 is not significantly affected by the entry of anthropogenic CO2 because in surface waters that are CaCO3 supersaturated, CaCO3 dissolution due to progressive acidification is negligible [Chen and Millero, 1979; Poisson and Chen, 1987; Gruber et al., 1996]. The total amount of anthropogenic CO2 trapped in any parcel of water (CANT) according to Ríos et al.  and Pérez et al.  is calculated from CT, TA and dissolved oxygen measured at the sea:
where AOU/RC is the CT increase by organic matter oxidation. Apparent Oxygen Utilization (AOU) is calculated with the oxygen saturation equation of Benson and Krause [UNESCO, 1986]. RC is a stoichiometric coefficient (=−ΔO2/ΔC). Additionally, 1/2·(ΔTA + AOU/RN) is the CT change due to CaCO3 dissolution in deep ocean waters [Broecker and Peng, 1982], where ΔTA is the total alkalinity change from the initial value during water mass formation (TA0), and RN is a second stoichiometric coefficient (=−ΔO2/ΔN). For open ocean waters below 400 m RC = 1.45 and RN = 10.6 [Anderson and Sarmiento, 1994]. TA0 stands for preformed alkalinity and was calculated using the A14 database, with an error of the estimation of 4.4 μmol kg−1, from salinity and “PO” for the layer between 50 and 200 m, as the best approximation to the winter mixed layer, using the empirical equation obtained: TA0 = 336.86 + 56.23·S + 0.0200·‘PO’ (r2 = 0.97, n = 116, p < 0.001) where “PO” = O2 + RP·PO4. RP is a stoichiometric coefficient (= −ΔO2/ΔP) with a constant value of 170 molO2/molP [Anderson and Sarmiento, 1994]. CT2780 is the initial CT of any water mass in the preindustrial era, which can be calculated from time-independent TA0 and the atmospheric pCO2 level at that time and water vapor pressure [Pérez et al., 2002], where pCO2 = xCO20·(Patm-vapor pressure) according to U.S. Department of Energy . The mole fraction of CO2 in the preindustrial atmosphere (xCO20) was 278.2 ppm [Neftel et al., 1994; Sarmiento et al., 1995]. The method to estimate CANT is subject to a number of uncertainties. Using an error propagation analysis for CANT as given by Gruber et al.  and Sabine et al. , the maximum error associated with the CANT calculation was ±5.6 μmol·kg−1.
 The distribution of CANT along A14 is shown in Figure 10. The 0.1 pmol kg−1 CFC11 isoline, corresponding roughly to an apparent age of 37–39 years, is traced along the CANT isoline of 15 μmol kg−1. In general, CANT was a maximum in the upper 200 m, with the highest values (45–75 μmol kg−1) to the north of 23°S (Figure 10). Minimum penetration of CANT was observed in the equatorial zone, where the CANT isoline of 15 μmol kg−1 was situated at ∼400 m due to the upwelling of old waters. The reduced CANT penetration extended to 15°S, at the subequatorial-subtropical transition. Further south, within the STG, large-scale downwelling produced a relative sink of the CANT rich upper waters that contribute to the penetration of new water volumes which were ventilated to the south of the STF. Concentrations >10 μmol kg−1 were found above a depth of ∼2000 m in the STG, whereas south of the STF the isoline of 10 μmol kg−1 extended to ∼2200 m. The invasion of CANT is remarkable south of 23°S, especially south of the STF where the isoline of 5 μmol kg−1 reaches the bottom, corresponding with the penetration of AABW. It is interesting to note that the penetration of CANT at the NADW domain in the Eastern South Atlantic had already begun at the core of eastward flowing UNADW centered at ∼1400 m at the equator.
 The distribution of CANT along A14 (Figure 10), was similar to that described by Chen  (GEOSECS 1973), Gruber  (SAVE 1989) east of A14, and Wanninkhof et al.  (NOAA/OACES 1991) west of A14, although with some remarkable differences. In general, the CANT penetration was deeper during A14 than found by these authors, mainly due to the time elapsed among different database used in the CANT estimates. In addition, the CANT estimates can vary depending on the method used [Wanninkhof et al., 1999].
 AAIW in the Subantarctic Zone (south of 36–37°S) appears as a thick layer of salinity <34.2 (Figure 3), which transports the high CANT concentrations (>30 μmol kg−1) acquired south of the SAF to the interior of the South Atlantic. Figure 11 shows the meridional evolution of CANT along the AAIW salinity minimum during A14. Maximum CANT values of 36 μmol kg−1 were observed at the southern end of the section (45°30′S) and decreased northward 15°S. North of 15°S, CANT became constant at 14 ± 2 μmol kg−1. CFC11 exhibited the same trend as CANT along the AAIW salinity minimum, as suggested by the corresponding linear regression (r2 = 0.88, n = 35, p < 0.0001):
Warner and Weiss  found similar CFC11 and CFC12 evolutions at the AAIW 27.2 isopycnal surface along the prime meridian during the Ajax expeditions (1983–1984). In their study the concentrations of both CFCs became constant at 20°S instead of at 15°S, which we observed 10 years later.
 CFC concentrations supply information on time scales of oceanic processes such as advection and mixing. Equilibration with the overlying atmosphere at the time of water mass formation is usually invoked during the estimation of water mass ages by means of CFC concentrations [Warner and Weiss, 1992]. The “apparent age” (AgeCFC) of AAIW at the salinity minimum can be calculated converting the measured CFC11concentrations into CFC-11 partial pressures (pCFC11) [Doney and Bullister, 1992]:
where F is the solubility of CFC11, a function of potential temperature and salinity given by Warner and Weiss . AgeCFC is then obtained matching the calculated pCFC11 with the atmospheric pCFC11 record. Reconstructed CFC11 annual mean dry air mole fractions in the Southern Hemisphere were used [Walker et al., 2000]. The course of AgeCFC at the AAIW salinity minimum along A14 referred to 1995 (Figure 11) is a mirror image of CANT. AAIW at 45°30'S was ∼15 years old and became older as it evolved northwards. North of 15°S, AgeCFC was constant at 42 ± 1 year. There was a very good correlation between CANT and AgeCFC (r2 = 0.96, n = 35, p < 0.001).
 According to this regression, the annual rate of CANT entry in the ocean via AAIW is 0.82 μmol kg−1 y−1. This rate is in very good agreement with the annual rate of 0.80 μmol kg−1 y−1 estimated from the equilibrium between the upper mixed layer and the average atmospheric xCO2 increase, according to Keeling et al. . In order to check the age of the AAIW salinity minimum at 45°S (CANT = 36 μmol kg−1 and AgeCFC = 15 years) a back calculation was performed. Calculated CT0 in the preindustrial time (xCO2 = 278 ppm) and in 1980 (xCO2 = 336 ppm) [Keeling et al., 1995], 15 years before A14, were 2069 μmol kg−1 and 2100 μmol kg−1, respectively. Therefore the net anthropogenic carbon uptake was 31 μmol kg−1 indicating that AgeCFC is comparable with the proper age of the water mass in this case. North of 15°S the CFC11 signal was below the detection limit. Consequently, we can only suggest that the AAIW was >42 years old in this zone.
 Similar CANT and AgeCFC calculations were performed for the overlying SACW 13°C isotherm along A14 (not shown). The average CANT entry along this isotherm was 33 ± 3 μmol kg−1. The transition at 15°S separates the 22 ± 1 years old (AgeCFC) subequatorial SACW to the north from the 10 ± 1 years old subtropical SACW to the south, with CANT entries of 32 ± 4 and 34 ± 2 μmol kg−1, respectively. The apparent age difference between the subtropical and subequatorial SACW branches (12 ± 2 years) exceeds the residence time of SACW in the SEG estimated by Gordon and Bosley . This indicates the time that subtropical SACW spends to arrive at the SEG through the SEC and then the ECS.
 Regarding the old and poorly ventilated UCPW just below the AAIW domain, the meridional evolution of CANT and AgeCFC have been examined along the core of maximum NTA (Figure 5). Constant CANT of 17 ± 2 and 13 ± 1.6 μmol kg−1 were obtained south and north of 15°S, respectively. The corresponding AgeCFC were 38 ± 5 and >45 ± 2 years. The youngest UCPW (33 ± 1.4 years) appeared south of the STF. The slight CANT penetration into the domain of the UNADW although coincident with the uncertainty (5.9 ± 2.0 μmol kg−1), showed a systematic signal. This signal was also visible on the transect NOAA/OACES [Wanninkhof et al., 1999]. The penetration below 2500 m, south of 23°S, with AABW influence, showed an average CANT of 5.7 ± 2.0 μmol kg−1, while north of 23°S in the domain of the aged middle and deep NADW, the CANT signal (2.6 ± 2.4 μmol kg−1) is not significant. Since the standard deviation of CANT for this old water is 2.4 μmol kg−1, the uncertainty of 5.6 μmol kg−1 obtained by calculation of error propagation is coherent.
 In order to estimate the inventory of anthropogenic carbon along A14, calculated CANT values were depth integrated from the surface (except the shallowest waters with AOU < 0) to entire water column, erasing negative values. The average CANT inventory for the entire line A14 was 540 gC m−2. Figure 12 provides the CANT specific inventory for the different zones intersected by A14. South of 15°S, the specific inventory is twice that north of 15°S so that nearly 70% of the CANT penetration occurred in the subantarctic and subtropical domains. A comparison of the CANT inventory by zonal bands between A14, Chen , Gruber  estimate for 1989, and Sarmiento et al.  calculations with the Princeton Ocean Biogeochemistry Model (POBM) has been made (Figure 12). The CANT specific inventory for A14 (black dots) was binned (open dots) in the same latitude bands as Gruber , represented by open triangles down, to facilitate the comparison. Inventory from Gruber  was estimated from the zonal section of CANT of his Figure 10 and represented by black triangles down (Figure 12). The average inventory for 1989 obtained using a segment with the same range of latitude as the A14 line, was 440 gC m−2 that is in agreement with the average specific inventory (36 mol m−2) taken from Table 4 of Gruber . In the same way we have estimated the inventory from Chen , using the CANT vertical distribution of his Figure 4, obtaining an average inventory of 342 gC m−2 for 1973.
 The zonal distribution of our CANT specific inventory is higher than Gruber's  and Chen's  inventories, showing differences of 20 and 29 mol m−2, respectively, south of 15°S. The differences are lower north of 15°S (5 and 12 mol m−2, respectively). In the equatorial zone our CANT inventories are slightly lower (5 mol m−2) than Gruber's and slightly higher (2 mol m−2) than Chen's estimations. In the equatorial zone the ocean could be acting as a source due to the upwelling in this zone, favoring that water loaded with CANT reaches the surface water. Also, this decrease in the CANT inventories in the equatorial zone could be affected by the differences in station locations with regard to the coast.
 The average CANT specific inventory (45 mol m−2) for A14 is higher than the estimate by Gruber and Chen in 9 and 16 mol m−2, respectively. Our inventory is also higher in 15 mol m−2 than the POBM model for the South Atlantic Ocean for 1989. The zonal distribution of the POBM corresponds very well with the Chen  estimate for 1973. Taking into account these inventory differences from our data, increase rates of 0.73 mol m−2 y−1 with regard to Chen  and 1.5 mol m−2 y−1 to Gruber  were obtained. The inventory difference between Gruber  and Chen  of 7 mol m−2, results in an increase rate of 0.44 mol m−2 y−1.
 Considering the annual rate of CANT entry in the ocean via AAIW is 0.82 μmol m−2 y−1 and the maximum depth penetration of this water reaches about 800 m, the atmospheric input of CANT is 0.67 mol m−2 y−1 that agrees with the increase rate estimated between A14 and Chen  inventories for the period 1973–1995. Conversely, the increase rates calculated from Gruber  inventory with regard to Chen  and W14, for the periods 1973−1989 and 1989−1995, respectively, does not show this agreement. Although it can be expected that the penetration would increase in recent years, it seems to be very high for the period 1989−1995 and rather low between 1973 and 1989.
 Accurate estimation of the CO2 anthropogenic entry in the oceans requires an accurate, precise and consistent database. The accuracy was tested by means of CRMs and the precision by duplicate analyses. The consistency was assessed by in situ measurements of three CO2 system parameters. Knowledge of the dependence of CO2 system parameter distributions on water masses ventilation, circulation, and mixing patterns is needed to better understand how the anthropogenic CO2 penetration is produced. On the basis of these considerations we can summarize the following conclusions regarding the meridional evolution of measured and derived CO2 system parameters along line A14:
 1. Simultaneous measurements of three CO2 system variables, the systematic use of CO2 certified reference materials, and the collection of samples for comparison with manometric CT measurements at SIO provided a means to test the accuracy and consistency of our shipboard measurements. CO2 data produced during A14 were quite reliable to use for estimation of the potential of the Eastern South Atlantic as an anthropogenic CO2 trap.
 2. The power of CO2 system parameters as tracers of the ventilation, circulation, and mixing patterns of water masses in the Eastern South Atlantic has been explored. The subtropical and subequatorial branches of SACW have been characterized by the pH15, NTA, and NCT meridional evolution along the 13°C isotherm. The CPW can be traced by the conspicuous NTA maximum. Maximum pH15 and minimum NTA and NCT values allows characterization of NADW, marking important zonal flows in the UNADW domains at the equator, 13°S and 22°S. CO2 system variables also characterized AABW on the northern (Romanche Fracture Zone) and southern (south of the Walvis Ridge) sides of line A14.
 3. About 70% of the anthropogenic CO2 accumulated by the Eastern South Atlantic along A14 during the industrial era occurred in the subantarctic and subtropical domains, where penetration depths were maximum. Extreme values of CANT are observed in the AAIW domain, being maximum in the relatively young AAIW of the subantarctic zone and minimum in the very old AAIW of the subequatorial gyre. The annual rate of CANT incorporation by AAIW is 0.82 μmol kg−1 y−1, in agreement with the annual rate (0.8 μmol kg−1 y−1) estimated from the equilibrium between the upper mixed layer and the atmospheric pCO2 increase in the section.
 We would like to thank the master, officers, and crew of R/V L'Atalante and all the participants on the cruise CITHER-3. We are very grateful to T. Rellán for pH and alkalinity measurements and preparation of figures, A. Billant and J.P. Gouillou in charge of the CTD, M.J. Messias for CFC measurements, M.J. Pazó for nutrients and chlorophyll determinations, and Melchor González-Dávila for providing pHSWS measured on CRM batch 24. Special thanks to M. Arhan (coordinator of the WOCE-France program CITHER) and H. Mercier (chief scientist of cruise CITHER 3). This work was supported by the Spanish CICYT (grant ANT94-1168-E) and IFREMER (University contract 94 1430 087).