Decadal increases of anthropogenic CO2 in the South Pacific subtropical ocean along 32°S



[1] To estimate decadal increases of anthropogenic CO2 in the ocean, distributions of dissolved inorganic carbon (CT) corrected by apparent oxygen utilization and salinity (nCTANT) were investigated along the World Ocean Circulation Experiment (WOCE Hydrographic Programme (WHP)) P6 section based on data obtained 10 years apart. Significant increases of nCTCAL were detected down to 1500 m (≅27.5σθ) water depth, above which the Sub-Antarctic Mode Water (SAMW) and the Antarctic Intermediate Water (AAIW) are found. The decadal increases of nCTCAL on the isopycnal surfaces (26.6–26.9σθ) of SAMW were higher (5–8 μmol kg−1) to the east of 160°W than to the west of it, while the increases in AAIW were almost constant on the isopycnal surfaces (27.0–27.5σθ). The averaged increases of nCTCAL in SAMW and AAIW were 10 ± 3.1 and 4.1 ± 2.0 μmol kg−1, respectively. Small but significant increases of nCTCAL and salinity-normalized CT (nCT) were also found (approximately 3.0 and 5.0 μmol kg−1, respectively) in abyssal waters occupying depths greater than 3500 m at longitude 180°–160°W, which correspond to Circumpolar Deep Water. Spatial differences of anthropogenic CO2 accumulation are discussed in terms of water mass distributions. The water column inventory of increases of anthropogenic CO2 in the South Pacific subtropical ocean was estimated to be 1.0 ± 0.4 mol m−2 yr−1, which is almost the same as that previously reported.

1. Introduction

[2] It is important to investigate whether the anthropogenic CO2 uptake rate remains unchanged in response to global warming and the associated climate changes. Since a number of studies report oceanic environmental changes on a decadal timescale [e.g., Bindoff and McDougall, 2000; Garcia et al., 2005; Keller et al., 2002; Matear et al., 2000], it is likely that the uptake rate of anthropogenic CO2 is changing on that timescale.

[3] To detect decadal changes in anthropogenic CO2 uptake by the ocean, high-quality data are needed because such signals are usually small, and analytical errors inevitably included in measurements often mask small signals. We now have excellent quality-controlled data covering the world’s oceans that were acquired following common protocols [e.g., WOCE Operation Manuals, 1994; Department of Energy, 1994]. These data were collected in the 1990s by the global survey efforts of the World Ocean Circulation Experiment (WOCE) and the Joint Global Ocean Flux Study. By repeating hydrographic observations like those of the WOCE Hydrographic Programme (WHP), we can revise the inventories and refer to spatial and temporal changes of anthropogenic CO2 uptake by the ocean in response to global warming and the associated climate changes.

[4] Using the R/V Mirai of the Japan Agency of Marine-Earth Science and Technology (JAMSTEC) from July 2003 through February 2004, we conducted worldwide hydrographic observations in the Southern Hemisphere during the Blue Earth Global Expedition 2003 (BEAGLE) and investigated decadal changes of the Antarctic overturn system. We revisited the WHP P6 section in the Pacific Ocean, at about 32°S (Figure 1) latitude (hereafter abbreviated as lat), from 3 August to 19 October 2003.

Figure 1.

Station map for the R/V Mirai (BEAGLE) cruise along the P6 section from 3 August to 19 October 2003. The numbers in the figure indicate station numbers of the WHP P6 section. For details, refer to the data book [Fukasawa et al., 2005].

[5] In the present study, we aimed to detect decadal increases of anthropogenic CO2 in the South Pacific subtropical ocean and to investigate spatial differences among the increases. Sub-Antarctic Mode Water (SAMW) and Antarctic Intermediate Water (AAIW), among the most widespread water masses of mode and intermediate waters, respectively, in the world ocean, are found in the study area in water shallower than 1500 m. Therefore spatial and temporal variations in the distributions of anthropogenic CO2 in these waters are critical for evaluating CO2 uptake by the ocean and for assessing future increases of CO2 in the atmosphere. Furthermore, we attempted to detect anthropogenic CO2 in Circumpolar Deep Water (CDW), in which signals have been difficult to find in data of previous studies because there were no high-quality data on a decadal interval sensitive enough to detect weak anthropogenic CO2 signals in abyssal water.

[6] Following the study of Schmitz [1996], we use the term CDW instead of “Antarctic Bottom Water (AABW)” to designate the Southern Ocean-origin bottom water usually found in the bottom layers in the western half of the South Pacific subtropical ocean.

2. Data Used

2.1. BEAGLE Observations

[7] For BEAGLE, carbonate system properties such as dissolved inorganic carbon (CT) and total alkalinity (AT) were measured by coulometry and potentiometry, respectively, at nearly every other station (3272 samples at 120 stations). The precisions of CT and AT were estimated to be 1.5 and 2.3 μmol kg−1, respectively (Table 1). All values of CT and AT were set based on the certified reference material (batch 60) provided by Prof. A. G. Dickson of the Scripps Institution of Oceanography.

Table 1. Precision of CT and AT During the BEAGLE Cruise
 CT, μmol kg−1AT, μmol kg−1
  • a

    Weighted average.

Leg 11.5, n = 2032.2, n = 188
Leg 21.5, n = 1882.5, n = 168
Averagea1.5, n = 3912.3, n = 356

[8] Dissolved oxygen (DO) was measured with colorimetry following the WHP Operations and Methods (

[9] All data obtained in BEAGLE are publicly available from the data book [Fukasawa et al., 2005; Kumamoto et al., 2005; Murata et al., 2005]. For the details of the measurements, refer to the data book.

2.2. WHP P6 Data

[10] The WHP observations of carbonate system properties along the P6 section were conducted by a University of Keil group during 4–24 May, 1 June to 14 July, and 14–27 July 1992 as the P6e, P6c, and P6w cruises, respectively [Johnson et al., 2001]. The historical data for the P6 section were downloaded from the Climate Variability and Predictability (CLIVAR) and Carbon Hydrographic Data Office ( We used data qualified as being acceptable, that is, flag = 2. The data used were temperature, salinity, CT, partial pressure of CO2 (pCO2), and DO. From the data for CT and pCO2, we calculated AT from the following equation:

equation image

where [ ] indicates the concentration in units of micromole per kilogram. Dissociation constants for the solubility coefficient of CO2 in seawater (K0) were taken from the study of Weiss [1974], the data of Mehrbach et al. [1973] as reformulated by Lueker et al. [2000] for the first (K1) and second (K2) dissociation constants, and the work of Dickson [1990] and Millero [1995] for the dissociation constants of boric acid (KB) and water (KH), respectively.

3. Approach

[11] A method for estimating the amount of anthropogenic CO2 (CTANT in micromole per kilogram) accumulated in the ocean was first proposed independently by Brewer [1978] and Chen and Millero [1979]. Since then, through improvement in uncertainties [Shiller, 1981, 1982; Broecker et al., 1985] by Gruber et al. [1996], the method and variations of it have been applied repeatedly in the world’s oceans to estimate the total amount of anthropogenic CO2 accumulation since the industrial revolution: Sabine et al. [2002b] and Peng et al. [2003] for the Pacific Ocean, Gruber [1998] and Lee et al. [2003] for the Atlantic Ocean, Sabine et al. [1999, 2002a] for the Indian Ocean, and Lo Monaco et al. [2005a, 2005b] for the Southern Ocean, to name a few. These methods are observation-based, not model-based approaches. Accordingly, CTANT inventories estimated by the approaches have become a basis of assessing future estimates of oceanic CO2 uptake [Maier-Remer et al., 1996; Sarmiento and Le Quéré, 1996; Sabine et al., 2004]. Although there is no doubt that observation-based approaches are useful in assessing CTANT uptake by the ocean, for reliable assessment of oceanic CO2 uptake, uncertainties of CTANT are still large. One reason is that anthropogenic CO2 is calculated from a number of observed properties, which inevitably include analytical errors.

[12] Our intention was to detect increases in CTANT over a decade, not to measure the total amount of accumulation since the industrial revolution. An increase is detected by comparing the carbonate system and its related properties obtained at the same locations a decade apart. This relatively short-term comparison is advantageous for calculating CTANT: some properties do not change over a short-term period, which allows the exclusion of some observed properties and thus a smaller uncertainty in determining CTANT. In the present comparison, two different methods have been used: multiparametric [Wallace, 1995; Sabine et al., 1999] and isopycnal [Peng et al., 1998]. We applied the latter method to high-quality data from the WOCE and BEAGLE cruises.

3.1. Calculation of Anthropogenic CO2 Increases

[13] CTANT is typically defined as follows [Gruber et al., 1996; Sabine et al., 2002a, 2002b]:

equation image

where CTm and ATm (micromole per kilogram) are the measured CT and AT, respectively. The parameter γC:O is the C:O “Redfield ratio,” for which we used the value 0.69 [Anderson and Sarmiento, 1994]. The Redfield ratio can be replaced with the ratio determined between carbon and nutrients. We used the C:O ratio because DO is often used as an index to reveal decadal changes in the ocean [Matear et al., 2000; Garcia et al., 2005], which is helpful in interpreting spatial and temporal changes of anthropogenic CO2. AOU indicates the apparent oxygen utilization, which is defined as the difference between the observed concentration of DO (micromole per kilogram) and the saturated DO concentration at the potential temperature and salinity of the sample. AT0 is defined as the AT of a water mass before anthropogenic CO2 was released into the atmosphere. CT0 (micromole per kilogram) is defined as the theoretical CT the water would have if it were in complete equilibrium with the atmosphere including no anthropogenic CO2. ΔCTdiseq (micromole per kilogram) is the difference between the CT of the mixed layer in equilibrium with atmospheric CO2 and the CT actually contained in the mixed layer at the time of water mass formation.

[14] The terms in curly brackets can be omitted if CTANT at time t is subtracted from CTANT at time (t + Δt). That is, both the AT0 and CT0 terms disappear by the subtraction because they conceptually have the same values at times t and (t + Δt). ΔCTdiseq also cancels out if one accepts assumptions such as the following: (1) the air-sea differences of pCO2 in a water mass formation area do not change over time, and (2) the formation areas for water masses found in the present study are fixed over time. ATm cancels out under the assumption that there exists no change in ATm over the observation periods. The assumption of no change in ATm is discussed in section 7.1.

[15] It follows from the above discussion that equation 2 can be simplified as CTCAL:

equation image

Furthermore, we normalized CTCAL to a salinity of 35 (nCTCAL). Increases of nCTCAL (ΔnCTCAL) were calculated by subtracting nCTCAL obtained in WOCE (nCTCAL(W)) from nCTCAL obtained in BEAGLE (nCTCAL(B)):

equation image

For the actual calculation of ΔnCTCAL, we interpolated nCTCAL to given isopycnal surfaces to reduce the influence of water mixing, and we subtracted averages of nCTCAL(W) from those of nCTCAL(B) at intervals of 20° longitude on each isopycnal surface. This interval was arbitrarily determined so that ∼10–20 data could be used for a statistical analysis. For abyssal waters, however, additional data selection rules were applied for the detection (section 6).

[16] To check the statistical significance of differences in the properties between the two observation periods, we used the Student’s t test. Since we expected nCTCAL to increase in the ocean’s interior in response to increases in atmospheric CO2, for ΔnCTCAL, we adopted the one-tailed test with P = 0.05. The number of samples for the t test, that is, the number of values of nCTCAL used in the average, ranged mostly from 10 to 20.

3.2. Evaluation of Errors

[17] The values of ΔnCTCAL inevitably include random and systematic errors propagated from individually measured properties. We judged the statistical significance of increases of anthropogenic CO2 (ΔnCTCAL) using the t test, which takes into consideration the variance of ΔnCTCAL, including random errors. Thus a significant result for ΔnCTCAL implies that ΔnCTCAL is significantly in excess of random errors. Here we evaluate the systematic errors of ΔnCTCAL, together with nCT and AOU.

[18] In the Pacific Ocean, the oldest water mass labeled Pacific Deep Water (PDW) is found between 2500 and 3000 m at longitude (hereafter abbreviated as long) 170°–150°W [Schlosser et al., 2001; Figure 3 in the present study]. Thus it is reasonable to assume that no significant differences of nCTCAL, nCT, and AOU between the WOCE and BEAGLE periods exist in PDW. If significant differences are detected in PDW, the differences of nCTCAL, nCT, and AOU can be regarded as the systematic errors of the respective properties. The systematic errors, if present, are probably derived from BEAGLE observations because it is reported that no adjustments should be made to the WHP P6 data [Lamb et al., 2002; Sabine et al., 2005].

[19] We calculated the differences for nCTCAL, nCT, and AOU at 41.40σ3 that corresponds approximately with 2500–2800 m at long 170°–150°W. Results of the t test applied to the properties are listed in Table 2 (see also Figure 6). The differences of the averaged properties are all negligibly small, and none of them is significant. The results mean that the systematic errors of nCTCAL, nCT, and AOU can be neglected.

Table 2. Results of t Test for nCTCAL, nCT, and AOU on 41.40σ3a
 nCTCAL, μmol kg−1nCT, μmol kg−1AOU, μmol kg−1
  • a

    The t test was judged at the one-tailed 95% significance level except for AOU that was judged at the two-tailed 95% significance level.

WOCE2205.1, n =122335.8, n = 12187.5, n = 12
BEAGLE2205.8, n = 142336.3, n = 14187.2, n = 14
t TestNot significantNot significantNot significant

4. Water Masses and CT Along the P6 Section

[20] To outline water masses along the P6 section, potential temperature-salinity (θ-S) plots were created (Figure 2). We found that water masses shallower than 26.0σθ revealed a large dispersion of θ and S and that the ranges of the dispersion were different between the WOCE and BEAGLE cruises. This result is probably related to seasonal and interannual variabilities of the surface ocean. Water masses deeper than 26.0σθ did not show distinct differences in the distribution patterns between the cruises, implying that there were no substantial changes of water mass distributions.

Figure 2.

Plots of θ-S along the P6 section for (a) WOCE and (b) BEAGLE cruises. The lines indicate densities calculated for given θ and S. The numbers attached to the lines show density as 24 = 1024 kg m−3.

[21] The salinity minimum at about 34.25 is consistent with the core of AAIW, which nominally ranges from 27.0 to 27.5σθ (700–1500 m). Above AAIW, SAMW distributes in the range from 26.6 to 26.9σθ (350–700 m). In the density range, salinity presents a large variation. Because of this, SAMW, including the shallower layers, is often divided into western and eastern water masses, which are called the Western and Eastern South Pacific Central Waters, respectively (WSPCW and ESPCW; [Tomczak and Godfrey, 2001]). For high-density waters, water masses of θ < 1.0°C and S > 34.7 are consistent with CDW, while those aggregated in a small range of 1.0 < θ < 2.0°C and 34.6 < S < 34.7 are consistent with PDW.

[22] Distributions of CT along the P6 section based on data obtained in BEAGLE are displayed in Figure 3. Concentrations of CT generally increase with increasing depths down to 3000 m. The vertical increases of CT are usually observed also in other oceans and can be accounted for by remineralization of organic matters in subsurface layers.

Figure 3.

Distributions of CT (micromole per kilogram) along the P6 section based on the data obtained in BEAGLE. The names of characteristic water masses along the section are shown schematically.

[23] Concentrations of CT for SAMW and AAIW are 2100–2150 and 2150–2250 μmol kg−1, respectively. In the range from SAMW to AAIW (350–1500 m), concentrations of CT show a relatively high vertical gradient of 14 μmol kg−1 per 100 m. There exist relative maxima of >2300 μmol kg−1 at 2200–3400 m at long 180°–130°W, which are consistent with Δ14C minima in the Pacific Ocean [Schlosser et al., 2001]. The water mass of the CT maximum is in accord with PDW. At depths below PDW, concentrations of CT become lower than CT in PDW by ∼30 μmol kg−1, that is, ∼2270 μmol kg−1. The water mass corresponds to CDW. East of 120°W, CT is vertically almost constant from 2000 m to the bottom because the shallow depths in the region prevent CDW of lower CT intruding from the south. In the eastern end of the section, there is a maximum of CT as high as the maximum found at long 180°–130°W.

[24] In the east-west direction, no clear structure can be found. That is, the isolines are as a whole parallel to longitude, neglecting local variations. This fact implies that along the section, no major sink or source of CT exists, which suggests that north-south transports are predominant.

5. Decadal Increases of nCTCAL on Isopycnal Surfaces

[25] Distributions of nCTCAL on isopycnal surfaces with σθ = 26.1, 26.2, 26.3, … , 27.8 along the P6 section were created as a function of longitude. Here we show distributions of nCTCAL(W) and nCTCAL(B) for 26.7, 27.1, and 27.5σθ, which represent SAMW and upper and lower AAIW, respectively (Figure 4).

Figure 4.

Distributions of nCTCAL on isopycnal surfaces of (a) 26.7σθ, (b) 27.1σθ, and (c) 27.5σθ as a function of longitude along the P6 section. The green squares and blue crosses show nCTCAL(W) and nCTCAL(B), respectively. The lines indicate the overall distribution patterns. The numbers along the lines indicate increases of nCTCAL significant at the one-tailed 95% significance level, while “ns” indicates “not significant” at the same level.

[26] Both nCTCAL(W) and nCTCAL(B) show a distinct eastward increase on isopycnal surfaces shallower than 26.7σθ (inclusive). Since preformed CT is the main contributor to the magnitude of nCTCAL, the eastward increases seem to be due to a basin-scale upwelling of waters rich in preformed CT. On isopycnal surfaces heavier than 27.1σθ, nCTCAL in both cruises reveals constant values in the longitudinal direction, suggesting that the nCTCAL spreads uniformly through isopycnal transport.

[27] The value of nCTCAL(B) is generally higher than nCTCAL(W) down to 27.5σθ, while the averaged ΔnCTCAL is significantly greater than zero. The magnitude of ΔnCTCAL decreases with deepening isopycnal surfaces. On the 27.5σθ surface, ΔnCTCAL becomes smallest, and some of the increases are statistically insignificant. On isopycnal surfaces shallower than 26.9σθ, ΔnCTCAL becomes larger east of 160°W than west of it, with a minimum at 160°W.

[28] To examine ΔnCTCAL in detail, a density-longitude cross section of ΔnCTCAL was constructed (Figure 5). On isopycnal surfaces shallower than 27.0σθ, east and west contrasts of ΔnCTCAL can be distinguished: west of 160°W, ΔnCTCAL displays small variations vertically, while east of 160°W, it shows a relatively large vertical gradient. Furthermore, east of 160°W, ΔnCTCAL is higher (5–8 μmol kg−1) than that in the west of 160°W. On the 27.0–27.5σθ surfaces, no such east-west difference can be found although in the west of 160°W, a ΔnCTCAL value of less than 4 μmol kg−1 is found on deeper isopycnal surfaces than east of 160°W. In terms of water masses, the east-west contrast can be summarized as follows: ΔnCTCAL in SAMW (26.6–26.9σθ) has east-west differences, whereas ΔnCTCAL in AAIW (27.0–27.5σθ) does not have.

Figure 5.

Density-longitude cross section of ΔnCTCAL along the P6 section.

[29] In SAMW, the averaged ΔnCTCAL was 10.3 ± 3.1 μmol kg−1, while in AAIW, it was 4.1 ± 2.0 μmol kg−1.

6. Increases of nCTCAL and nCT in Abyssal Waters

[30] Since we had high-quality data for CT and DO collected a decade apart, we attempted to detect anthropogenic CO2 signals in abyssal waters where, even if they exist, the signals are weak and are easily concealed in artificial variations derived from low-quality measurements. We expected to detect significant increases of nCTCAL (ΔnCTCAL) in CDW. Thus, in addition to the statistical analyses on isopycnal surfaces, we surveyed ΔnCTCAL in selected waters. Furthermore, we investigated differences of nCT (ΔnCT) under the assumption that properties of deep water do not have significant differences except for CT, which is expected to increase because of anthropogenic CO2 input.

[31] To investigate the ΔnCTCAL and ΔnCT in CDW, we selected waters based on two different criteria: waters with salinity maximum below 1000 m in a water column, and waters with potential temperature (θ) lower than 1.0°C. The former represents the core of CDW at depths greater than 3500 m, while the latter collects CDW broadly.

[32] Distributions of nCTCAL(W) and nCTCAL(B) on isopycnal surfaces of 41.40, 41.45, 41.50, and 41.55σ3 are illustrated in Figure 6a. In the western part of the P6 section, significant ΔnCTCAL is found for the four isopycnal surfaces. In particular, at long 170°E–170°W, the increases exceed 3 μmol kg−1. The same tendency as observed for nCTCAL is seen also for distributions of nCT, which show larger increases as a result of their being no subtraction of AOU (Figure 6b). The significant ΔnCTCAL and ΔnCT at long 170°E–170°W are probably related to deep western boundary currents [Whitworth et al., 1999].

Figure 6.

Same as Figure 4 but for distributions of (a) nCTCAL and (b) nCT on selected isopycnal surfaces of σ3.

[33] Distributions of nCTCAL(W) and nCTCAL(B) selected by the two criteria are shown in Figure 7a. For waters of salinity maximum <1000 m, those at long 180°–140°W correspond to CDW. The averaged nCTCAL(B) is significantly higher than nCTCAL(W) at long 180–160°W. For waters of θ < 1.0°C, nCTCAL(B) tends to be higher than nCTCAL(W) although the nCTCAL distributions for each cruise overlap considerably. In fact, nCTCAL in BEAGLE, which was averaged over an arbitrary longitudinal range, was significantly higher by 1.0–1.7 μmol kg−1 than nCT in WOCE. For distributions of nCT (Figure 7b), waters of salinity maximum <1000 m show a significant increase of nCT averaged for long 180°–160°W between the WOCE and BEAGLE cruises, as well as for those of nCTCAL. For waters of θ < 1.0°C, the averaged nCT in BEAGLE was significantly higher by 1.0–2.4 μmol kg−1 than that in WOCE.

Figure 7.

Distributions of (a) nCTCAL and (b) nCT selected by θ < 1.0°C (upper panels) and salinity maximum < 1000 m (lower panels) along the P6 section. Solid green circles and solid blue squares in the lower panels indicate averages of each property at an interval of 20° longitude. The letters “ns” indicate that the increases are not significant at the one-tailed 95% significance level.

7. Discussion

[34] Using high-quality data, we estimated decadal increases of anthropogenic CO2 in the South Pacific subtropical ocean. As a result, we found spatial variations of decadal anthropogenic CO2 storage (ΔnCTCAL). In particular, an east-west contrast of the ΔnCTCAL on isopycnal surfaces lighter than 26.9σθ was distinguished (section 5). Furthermore, a small but significant increase of anthropogenic CO2 was also detected in CDW (section 6). In this section, after examining the assumptions made for the calculations of ΔnCTCAL, we discuss the observed increases from the viewpoints of water mass formation and air-sea exchange of CO2, and of their spatial and temporal variations. In addition, to evaluate decadal storage of anthropogenic CO2 in the ocean, water column inventories are calculated.

7.1. Validity of ΔnCTCAL

[35] In calculating ΔnCTCAL, we neglected differences in AT, that is, the effects of dissolution/precipitation of calcium carbonate on decadal changes of anthropogenic CO2. AT decreases due to calcification of CaCO3 shell-forming species, and it increases due to dissolution of CaCO3. AT also changes, albeit by a small amount, by photosynthesis and remineralization [Brewer and Goldman, 1976]. However, these changes are usually found in the upper (mixed) layers of the ocean as seasonality and year-to-year variability of oceanic carbonate systems. It is important to know whether AT changes occur below the upper layers, where SAMW and AAIW are found. Recently, increases of AT (5 to 10 μmol kg−1) over two decades were found in intermediate layers close to the aragonite saturation horizon of the Pacific Ocean [Sarma et al., 2002].

[36] To confirm the possibility of increases or decreases of AT between the WOCE and BEAGLE periods, we compared AT normalized by a salinity of 35 (nAT) in BEAGLE with that in WOCE on an isopycnal surface of 27.2σθ (Figure 8), where significant changes of AT can be expected [Sarma et al., 2002]. From this figure, it can be seen that nAT did not change significantly between the WOCE and BEAGLE observation periods although certain increasing trends are seen at 150°–180°E and 110°–80°W. We judged that systematic errors in the AT measurements caused false increasing trends because the same increasing trends were also found on other isopycnal surfaces of 26.0–27.8σθ.

Figure 8.

Comparison of nAT at 27.2σθ between the WOCE (green squares) and BEAGLE (blue crosses) cruises.

[37] We assumed that the air-sea disequilibrium of CO2 (ΔCTdiseq) did not change over the decade. Inoue et al. [1995] report that increases in surface seawater pCO2 in the subtropical gyre of the North Pacific Ocean kept pace with the increase of atmospheric CO2, implying that ΔCTdiseq does not change over time. The tendency of no decadal changes of ΔCTdiseq is also reported in the Southern Ocean south of Australia [Inoue and Ishii, 2005]. Furthermore, a global survey of surface seawater pCO2 did not present any evidence of change in air-sea differences of pCO2 over time [Takahashi et al., 2002]. These results support our assumption.

7.2. ΔnCTCAL in SAMW

[38] Decadal increases of anthropogenic CO2 (ΔnCTCAL) on isopycnal surfaces lighter than 26.9σθ, which includes SAMW, showed marked differences east and west of 160°W (Figure 5). However, the extremely large ΔnCTCAL on a few top isopycnal surfaces should be due to the seasonality of nCTCAL. We calculated nCTCAL from data obtained during the WOCE and BEAGLE cruises, which took place during early May to late July and during early August to mid-October, respectively. Moreover, the WOCE and BEAGLE cruises started from the east and west ends of the P6 section, respectively. As a result, there was about a maximum 4-month seasonal lag in the observations from 130° to 100°W. From September to October (Austral winter), when the BEAGLE cruise was conducted, the mixed layers at long 130–100°W develop down to 250 m, while from May to July (Austral autumn), when the WOCE cruise was made, the mixed layers develop down to a maximum of 150 m [Kara et al., 2003]. Because of this, CT in the upper layer in BEAGLE became higher than CT in WOCE owing to the entrainment of waters with higher CT from the subsurface layers, leading to extremely large ΔnCTCAL. The depth of 250 m corresponds approximately to 26.4σθ at long 130°–100°W. Thus for the decadal signals, one should neglect ΔnCTCAL at 26.0–26.4σθ around long 130°–100°W. Likewise, the seasonal mixed layer variations caused the extremely low ΔnCTCAL at 160°W. Overall, seasonal influences are strong on isopycnal surfaces lighter than 26.4σθ (inclusive).

[39] Even if one considers seasonal influences, ΔnCTCAL at 26.6–26.9σθ is still higher east of 160°W than west of it. To aid the interpretation of different magnitude of ΔnCTCAL, we calculated ΔnCTCAL (= δCT) at equilibrium with the atmospheric CO2 increase from the so-called buffer factor β:

equation image

where δpCO2 and δCT indicate increments of pCO2 and CT, respectively. Since both mode and intermediate waters along the P6 section are considered to be formed at or close to the subtropical convergence zone [Sabine et al., 2004], we adopted 2110 μmol kg−1 for CT, which was observed by WHP cruises in the subsurface layers of the high latitudes of the South Pacific Ocean. In addition, for the high-latitude case, a value of 11 or 12 was given to β [Sabine et al., 2004]. If atmospheric CO2 increased from 357 ppmv in 1993 at a rate of 1.6–2.0 ppmv yr−1, the corresponding decadal increases of CT (ΔnCTCAL) are calculated from equation 5 to be 8–10 μmol kg−1.

[40] The calculated increases are consistent with ΔnCTCAL at 26.4–26.9σθ west of 160°W (Figure 5). This probably implies that a recently formed water mass that had nearly equilibrated with atmospheric CO2 was transported into the layers. In contrast, ΔnCTCAL on the isopycnal surfaces east of 160°W is higher by approximately 5 μmol kg−1 than the expected increases of CT.

[41] The surplus ΔnCTCAL east of 160°W is compensated by differences of AOU between WOCE and BEAGLE (Figure 9). That is, AOU in BEAGLE is smaller by 6–9 μmol kg−1 than in WOCE, leading to an increase in nCTCAL of 4–6 μmol kg−1. The AOU changes are possibly linked to decadal changes in thermocline, as illustrated in the studies of Bryden et al. [2003] and McDonagh et al. [2005]. That is, the smaller AOU possibly implies that the mode water reached the P6 section faster in BEAGLE than in WOCE.

Figure 9.

Comparison of AOU at 26.7σθ between the WOCE (green squares) and BEAGLE (blue crosses) cruises.

[42] The east-west differences of ΔnCTCAL can be attributed to differences in the formation regions of water masses. The range of isopycnal surfaces is in accordance with SAMW. According to Tomczak and Godfrey [2001], the western and eastern water masses of the thermocline between 20°S and the subtropical front are consistent with WSPCW and ESPCW, respectively. The former is formed and subducted in the subtropical convergence between Tasmania and New Zealand, while the latter is formed between 180° and 150°W. The boundary of the water masses is located at about 150°W, where the distributions of ΔnCTCAL also showed a boundary between western and eastern patterns (Figure 5).

[43] In summary, anthropogenic CO2 storage in SAMW shows regional differences related to differences of the formation areas and is probably affected by changes in the thermocline of the South Pacific.

7.3. ΔnCTCAL in AAIW

[44] ΔnCTCAL showed a slight difference among distributions on a density-longitude cross section (Figure 5): west of 160°W, significant ΔnCTCAL (4–6 μmol kg−1) was found in deeper layers than east of 160°W. This difference corresponds with differences in the flow patterns of AAIW. That is, according to Tomczak and Godfrey [2001], minimum salinity depth, which is indicative of AAIW movement, showed strong contrast between the east and west in the South Pacific Ocean. They suggest that the eastern “new” AAIW originates from a winter convection region west of southern South America, while the western “old” AAIW is advected from the west, which can be traced back to the winter convection region in the Atlantic Ocean. Thus differences in source waters possibly cause the differences of ΔnCTCAL in AAIW.

[45] To examine the uptake rate of anthropogenic CO2 in AAIW, we performed a similar calculation, as presented in section 7.2, but for calculating δpCO2 instead of δCT. We inserted β = 11 or 12, CT = 2110 μmol kg−1, δCT (= ΔnCTCAL) = 4–6 μmol kg−1, and pCO2 = 330–360 ppmv into equation 5. As a result, δpCO2 = 7–12 ppmv was obtained. From these results, we can estimate a growth rate of 0.7–1.2 ppmv yr−1 in the atmospheric CO2 if one assumes oceanic CO2 uptakes in equilibrium with atmospheric CO2. The estimation is close to the observed annual growth rates of atmospheric CO2 from the 1960s through the 1970s [Keeling and Whorf, 2004]. The accordance between the observed and estimated growth rates is reasonable if the age of AAIW is considered to be 10–20 years.

[46] From the above discussion, it can be concluded that the accumulation of anthropogenic CO2 in AAIW has been keeping pace with the increase of atmospheric CO2.

7.4. ΔnCTCAL in CDW

[47] Small but significant ΔnCTCAL was obtained for abyssal waters, that is, CDW (section 6). Here, to clarify the implications of the significant ΔnCTCAL, we discuss deep-water circulation with the aid of data from previous studies.

[48] During the WHP P6 observations, relatively high Δ14C distributions were found near the dateline along the section. The Δ14C distributions signify that CDW flows northward at the site [Schlosser et al., 2001] and that it has been ventilated in a relatively recent year compared with ambient waters of low Δ14C. These observations imply that waters affected by anthropogenic CO2, when in contact with the atmosphere and ventilated into the deep ocean, appear first along the CDW flow route. In reality, significant ΔnCTCAL was confined to long 180°–140°W, which is consistent with the northward pathway of CDW.

[49] The significant ΔnCTCAL in CDW may imply that the water mass has an age younger than 250 years because atmospheric CO2 in the atmosphere has increased since the industrial revolution, which started in the 1750s.

7.5. Water Column Inventory

[50] To quantify storage of anthropogenic CO2 in the South Pacific subtropical ocean, we calculated water column inventories of ΔnCTCAL along the P6 section. First, ΔnCTCAL, which is an average at a 20° longitude range as shown in Figure 4, was converted to units of mole per cubic meter. The depths of the corresponding isopycnal surfaces σθ = 26.1, 26.2, 26.3, …, 27.8 were calculated for both the WOCE and BEAGLE cruises, and the averaged depths were used for vertical integration. From 26.4σθ to the sea surface, we used the ΔnCTCAL of 26.5σθ because we judged that the upper layers had seasonal influences (section 7.2).

[51] Water column inventories of ΔnCTCAL along the P6 section are listed in Table 3 separately for each longitudinal range. The calculated inventories reveal regional differences, reflecting distributions of ΔnCTCAL: the water column inventory for long 170°E–170°W has the maximum (17.3 mol m−2) because in the longitudinal range, deep-water layers also showed significant ΔnCTCAL down to 3433 m (section 7.3). The value reduces to 10.8 mol m−2 if ΔnCTCAL in the deep layers is excluded from the calculation. Overall, the western side along the section reveals large water column inventories due to deep penetration of anthropogenic CO2. Relatively large values are found for long 150°–130°W and 130°–110°W, where high ΔnCTCAL existed in SAMW (section 7.2). In contrast, minima are found at long 170°–150°W and 90°–70°W. The water mass boundary at 160°W (section 7.2) is probably associated with the low inventory at long 170°–150°W.

Table 3. Water Column Inventories of ΔnCTCAL Along the P6 Section
Longitudinal RangeWater Column Inventory, mol m−2Maximum Depth, m

[52] The anthropogenic CO2 uptake rate along the P6 section was estimated to be 1.0 ± 0.4 mol m−2 yr−1. The estimated uptake rate is almost the same as the rate of 0.9 ± 0.3 mol m−2 yr−1 for the South Pacific reported by Peng et al. [2003], which was based on the data for 1974–1996. Thus it is fair to say that the anthropogenic CO2 uptake rate in the South Pacific subtropical ocean did not change considerably compared with that of the previous two decades.

8. Summary and Concluding Remarks

[53] To investigate decadal increases of anthropogenic CO2, we corrected data for CT obtained during the WOCE and BEAGLE cruises, and we compared the two observation periods. The findings obtained in the present study are summarized as follows:

[54] (1) Over the decade from 1992 to 2003, anthropogenic CO2 along the P6 section increased significantly by 10.3 ± 3.1 and 4.1 ± 2.0 μmol kg−1 for SAMW and AAIW, respectively;

[55] (2) Increases of anthropogenic CO2 in SAMW were larger in the eastern part of the P6 section than in the western part. The regional difference could be attributed to the differences between water masses;

[56] (3) The increases of anthropogenic CO2 in SAMW in the eastern part of the P6 section were in excess of the values expected from increases in atmospheric CO2. The surplus increases of anthropogenic CO2 were compensated by AOU changes, which suggest decadal changes of thermocline;

[57] (4) Increases of anthropogenic CO2 in AAIW displayed a slight east-west difference of distribution that was consistent with flow patterns of AAIW;

[58] (5) The CDW also showed a small but significant anthropogenic CO2 increase of approximately 3.0 μmol kg−1 at long 180°–160°W;

[59] (6) The anthropogenic CO2 uptake rate in the South Pacific subtropical ocean was estimated to be 1.0 ± 0.4 mol m−2 yr−1.

[60] We found regional differences in anthropogenic CO2 storage. Such regional differences should be taken into consideration for extrapolating a regional estimation of anthropogenic CO2 inventory to a basin or global scale.

[61] When conducting a repeat hydrography, the season of the cruise should be as similar as possible to the previous cruise because seasonal variation of upper-ocean conditions has considerable influence on the estimation of increases of anthropogenic CO2.


[62] We thank the officers and crew of the R/V Mirai for their exceptional support during the cruises. We also give special thanks to the staff of Marine Works Japan, who worked as physical and chemical oceanography marine technicians onboard the R/V Mirai.