We used high-quality data for dissolved inorganic carbon and related water properties in the Indian Ocean along 20°S (World Ocean Circulation Experiment Hydrographic Program line I3) and 24°S (I4) obtained 8 years apart (1995–2003/2004) to estimate decadal-scale increases of anthropogenic CO2 in the interior of the South Indian Ocean. Significant increases were detected to about 1800 m depth in the longitude range 35–45°E. In the upper thermocline subtropical subsurface water and Indian Central Water, anthropogenic CO2 increased an average of 7.9 ± 1.1 and 7.7 ± 0.5 μmol kg−1, respectively, whereas in the lower thermocline Antarctic Intermediate Water, the increase was 3.8 ± 0.7 μmol kg−1. A significant increase was also detected in Circumpolar Deep Water (2.5 ± 1.0 μmol kg−1). The estimated uptake rate of anthropogenic CO2 along the I3/I4 line over this time interval was 1.0 ± 0.1 mol m−2 a−1. Seasonal variations, which are influential in this ocean because of the Indian monsoon, did not affect detection of the anthropogenic CO2 signals. Comparisons with previous studies showed that increases of anthropogenic CO2 became larger in the most recent decade and that the CO2 uptake rate was similar to that in the South Pacific (1.0 ± 0.4 mol m−2 a−1) but higher than those in the South Atlantic (0.6 ± 0.1 mol m−2 a−1) and North Pacific (0.5 ± 0.1 mol m−2 a−1) Oceans. Deep penetration of anthropogenic CO2 is possibly associated with the higher uptake rate.
 According to the Intergovernmental Panel on Climate Change review [Solomon et al., 2007], the concentration of CO2 in the atmosphere has been increasing at a rate of 1.9 ppmv a−1 in recent decades as a result of CO2 emissions through human activities. Because increasing CO2 is considered to be a main cause of global warming, estimation of how much of the emitted CO2 remains in the atmosphere is an important issue in environmental science. The ocean and terrestrial biospheres currently absorb 50% of the anthropogenic CO2 emitted into the atmosphere, with each absorbing a similar amount [Solomon et al., 2007].
 Absorption of anthropogenic CO2 by the ocean, which has been generally estimated by numerical ocean carbon-cycle models, can now be corroborated mainly by two different approaches based on field measurements. One approach calculates air -sea fluxes of CO2. This calculation requires data for the differences in partial pressures of CO2 (ΔpCO2) between the atmosphere and the surface seawater, and gas exchange coefficients for CO2. Because variations of ΔpCO2 depend more on surface seawater pCO2 than on atmospheric pCO2, surface seawater pCO2 data are more critical. This approach has been widely used to investigate regional and local-scale distributions of atmospheric CO2 sinks and sources [e.g., Feely et al., 2006; Lüger et al., 2006], although the calculated CO2 fluxes include a large uncertainty derived from the calculated gas transfer velocity [Wanninkhof and McGillis, 1999]. Synthesizing data for surface seawater pCO2 collected worldwide and using appropriate gas transfer velocities, Takahashi et al.  reported the global ocean net flux of CO2 to be 2.0 Pg-C a−1 (1 Pg [petagram] = 1015 g). Takahashi et al.  revised this value to 1.6 Pg-C a−1 after examining three times the surface seawater pCO2 data used in the previous study.
 The second approach estimates anthropogenic CO2 stored in the ocean interior from measured dissolved inorganic carbon concentration (CT). Anthropogenic CO2 is basically calculated by subtracting the CT attributable to preindustrial origins and increases owing to biological activities from measured CT. The preformed CT is obtained by assuming seawater to be equilibrated with air with a preindustrial CO2 concentration of 280 ppmv. The increases due to biological activities are associated mostly with decomposition of organic matter and dissolution of CaCO3. The variations of CT related to biological activity are usually corrected using data for related properties, such as dissolved oxygen and total alkalinity for decomposition and dissolution, respectively. This method was originally developed by Chen and Millero  and Brewer . Although the method was criticized [Broecker et al., 1985; Shiller, 1981, 1982], a revised version (the ΔC* technique; see Gruber et al. ) and its variations [Pérez et al., 2002; Lo Monaco et al., 2005] have been widely used to estimate distributions of anthropogenic CO2 in the ocean interior; for example, Sabine et al.  and Peng et al.  for the Pacific Ocean, Gruber  and Lee et al.  for the Atlantic Ocean, and Sabine et al.  for the Indian Ocean. Sabine et al.  synthesized the global storage of anthropogenic CO2 as estimated by the ΔC* technique and concluded that the oceanic sink accounts for 48% of the total emissions from fossil fuels and cement manufacturing.
 Although information as to the mean uptake of CO2 by the ocean is now available, spatial and temporal variations in the uptake are still not understood, especially those variations in response to climate changes such as global warming. To investigate these spatial and temporal variations, global surveys, such as those revisiting observation lines previously occupied during cruises of the World Ocean Circulation Experiment (WOCE) Hydrographic Program (WHP) in the 1990s, are now conducted under the international framework of the Climate Variability and Predictability (CLIVAR)/CO2 repeat hydrography program. Because the revisits are usually made a decade or more after the original observations, decadal-scale variations in anthropogenic CO2 storage can be estimated. Anthropogenic CO2 storage and uptake rates on that time scale have already been estimated along select sections in the Pacific [Murata et al., 2007, 2009; Sabine et al., 2008] and Atlantic [Murata et al., 2008] Oceans. From these studies, it is evident that rates of uptake of anthropogenic CO2 differ from basin to basin.
 The purpose of the present study was to quantify decadal increases of anthropogenic CO2 in the South Indian Ocean by comparing the data collected during our revisit cruise (2003/2004) with those from the WOCE cruise (1995). As in our previous studies [Murata et al., 2007, 2008, 2009], we noticed increases of anthropogenic CO2 in characteristic ocean water masses, such as Indian Central Water (ICW), Sub-Antarctic Mode Water (SAMW), and Antarctic Intermediate Water (AAIW) for thermocline waters and Indian Deep Water (IDW) and Circumpolar Deep Water (CDW) for deep and bottom waters. One feature in the Indian Ocean is a distinct seasonal variation caused by the Indian monsoon. We evaluated the influences of seasonal variations on estimated increases of anthropogenic CO2 by using data collected by different WOCE cruises. Furthermore, we evaluated basin-to-basin differences in anthropogenic CO2 uptake rates. This study is a continuation of our previous work [Murata et al., 2007, 2008, 2009] in the South Indian Ocean.
 Six sections comprise the rest of this paper. In section 2, the data and methods are introduced. In section 3, results of data analyses are shown; the oceanic conditions in the study area in terms of the potential temperature-salinity relationships and the distributions of CT are presented in sections 3.1 and 3.2, respectively. The characteristics of decadal-scale increases in anthropogenic CO2 are examined in section 3.3. Influences of seasonal variations on calculated results are analyzed in section 3.4. Section 4 includes separate discussions comparing estimated increases of anthropogenic CO2 with results from previous studies (section 4.1), a theoretical examination of the increases based on the buffer factor calculation for individual water masses (section 4.2), and comparisons with the results obtained in other basins (section 4.3). Finally, section 5 presents the conclusions of the present study. In Appendix A, we look into some of the assumptions made for the present study. In Appendix B, analyses with data from Geochemical Ocean Section Study (GEOSECS) are introduced briefly, which are made complementally.
2. Data and Methods
 As part of an around-the-world cruise in the Southern Hemisphere conducted by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) and known as the Blue Earth Global Expedition (BEAGLE), we reoccupied WHP lines I4 and I3 in the South Indian Ocean (Figure 1) at nominal latitudes 24°S and 20°S, respectively, from 9 December 2003 through 24 January 2004, using the R/V Mirai of JAMSTEC. During the BEAGLE cruise, dissolved inorganic carbon (CT) and total alkalinity (AT) over the full water column were measured at 64 of the 124 hydrographic stations. Repeatability of CT and AT, which was expressed as the average and standard deviation of the absolute values of differences between replicate samples collected from the same Niskin bottles during the cruise, was estimated to be 0.9 ± 0.7 μmol kg−1 (n = 229) for CT and 2.0 ± 1.8 μmol kg−1 (n = 220) for AT. For calculating increases in anthropogenic CO2 (section 2.2), we also used data for dissolved oxygen (DO); the repeatability (standard deviation) of DO data was estimated to be 0.08 μmol kg−1 (n = 489). For details of the data and cruise, please refer to the cruise data book [Uchida and Fukasawa, 2005]. The data are available from the Carbon Dioxide Information Analysis Center (CDIAC; http://cdiac.ornl.gov/ftp/oceans/CLIVAR/I03I04_2003.data/).
 As part of a data synthesis, Lo Monaco et al.  examined the consistency of data obtained on the BEAGLE cruise. They concluded that no adjustments were necessary for DO or CT, implying that no systematic errors were included, whereas a correction of +10 μmol kg−1 was recommended for AT.
 The I3 and I4 lines were previously occupied by a group from the USA as part of the Indian Ocean CO2 Survey [Johnson et al., 2002] conducted from the R/V Knorr between 23 April and 19 June 1995. The data for CT, AT, DO, and other properties were downloaded from the CDIAC data site (http://cdiac.ornl.gov/oceans/ndp080/). According to Sabine et al. , no corrections are necessary for CT and AT data from the I3/I4 section. Gouretski and Jancke  recommended a DO correction of 1.2 and −0.1 μmol kg−1 to the original values for I3 and I4, respectively. These values were small enough that we neglected them in the present study.
 Our simple subtraction method for calculating increases in anthropogenic CO2 (section 2.2) is sensitive to data quality, especially systematic errors. As mentioned earlier in this section, the data for CT and DO, which are the principal data for the method, include little systematic error for either WOCE or BEAGLE data. Accordingly, the calculated increases of anthropogenic CO2 are believed to be almost free from systematic errors.
 We used the same method for calculating increases of anthropogenic CO2 as Murata et al. [2007, 2008, 2009]. The following is a brief summary of the method.
 Increases in anthropogenic CO2 (ΔnCTCAL) over 8 years were calculated using the following equation:
where nCTCAL(B) and nCTCAL(W) represent CTCAL normalized to a salinity of 35 for the BEAGLE and WOCE cruises, respectively, along the I3/I4 line. CTCAL is defined as:
where CTm (μmol kg−1) is the measured CT, and the parameter γC:O is the C:O Redfield ratio for which we used a value of 0.69 [Anderson and Sarmiento, 1994]. We omitted the terms ΔCTdiseq and ATm from equation (2), which are included in one of the pioneering works in the field of ocean carbon cycles [Gruber et al., 1996] and in related studies [Sabine et al., 1999; Lee et al., 2003]. ΔCTdiseq is the difference between CT in the mixed layer at equilibrium with atmospheric CO2 and actual CT in the mixed layer at the time of water mass formation, and ATm is the measured AT. This omission is valid if it can be assumed that both ΔCTdiseq and ATm remained unchanged between the two sampling periods.
 We interpolated nCTCAL to given isopycnal surfaces of σθ = 25.0, 25.1, …, 27.8 kg m−3 for thermocline waters. We also interpolated nCTCAL in deep waters to the isopycnal surfaces of σ3 = 41.40, 41.45, 41.50, and 41.55 kg m−3, which corresponded approximately to 2215, 2599, 3175, and 3886 m, respectively. We checked the significance of temporal increases in nCTCAL on each isopycnal surface in the longitudinal ranges of 35°E–45°E, 45°E–68°E, 68°E–88°E, and 88°E–120°E; these ranges were determined with consideration for the bathymetry and ocean structure [Tomczak and Godfrey, 2001]. We first applied the F test with a significance level of P = 0.05 to nCTCAL collected from these longitudinal ranges to determine whether the variance of nCTCAL from the BEAGLE cruise was statistically equivalent to that from WOCE. We then used a one-tailed t test with a significance level of P = 0.05 to test whether the average nCTCAL from BEAGLE was higher than that from WOCE, taking into account the equal or unequal variance as determined by the F test.
 The same statistical tests used for nCTCAL were used to judge the significance of changes in AT normalized to a salinity of 35 (nAT) and AOU, except that a two-tailed t test with a significance level of P = 0.05 was used.
 We normalized CT and AT to a salinity of 35 to remove variations (in the longitude ranges 45°E–68°E, 68°E–88°E, and 88°E–120°E, the maximum value of absolute differences of salinity between the BEAGLE and WOCE cruises was 0.06, while in the longitude range 35°E–45°E, it was 0.10) owing to the addition and removal of freshwater by precipitation and evaporation. As discussed in the work of Murata et al. , salinity-dependent errors associated with this normalization [Friis et al., 2003] mostly cancel out.
 The applicability of assumptions made for the calculations (i.e., omission of ATm and ΔCTdiseq) were tested and assessed in general for both assumptions at a decadal time scale. In addition, we have examined and demonstrated the applicability of these assumptions repeatedly in previous studies [Murata et al., 2007, 2008, 2009]. The method for examining the applicability of these assumptions was same as in our previous studies, although some local variations were considered. Thus, to focus on the main theme of the present study, we present a discussion of the applicability not in the main text but in Appendix A.
 The uncertainty of each property considered was expressed as the standard deviation [Murata et al. 2009]. For results from t tests, the pooled estimate of standard deviations or the root of the sum of the squared standard deviations was adopted as a measure of uncertainty.
 The θ-S diagram along the I3/I4 section is shown in Figure 2. To facilitate detection of east -west differences of water masses, we plotted θ and S averaged over each longitudinal range on individual isopycnal surfaces, just as for ΔnCTCAL (see section 2.2). Accordingly, we produced two θ -S diagrams, one from WOCE data and the other from BEAGLE. As there was little difference in the distribution patterns between the two diagrams, indicating that there were only slight water mass changes, the averaged values from the two diagrams are shown in Figure 2.
 In the upper layer (<26.0σθ), differences of salinity among longitudinal ranges were distinct. The most saline water, in excess of 35.76, was found on isopycnal surfaces of 25.7σθ to 25.9σθ at longitudes 68°E–88°E. Salinity maxima were also found on these isopycnal surfaces in other longitudinal ranges. Above these isopycnal surfaces, waters became less saline, probably indicating the influence of Indonesian Throughflow Water (ITW). The waters between 26.4σθ and 26.8σθ, with temperatures and salinities ranging from about 10–5°C and from 35.5 to 34.5, respectively, showed little longitudinal differences, implying the existence of well-mixed water masses. Salinity minima occurred on isopycnal surfaces from 27.1σθ to 27.4σθ in each longitudinal range.
 Water mass characteristics in the South Indian Ocean are also indicated schematically in Figure 2 with reference to Tomczak and Godfrey  and Fine et al. . For the discussions that follow, we defined ICW as the water mass ranging from 26.0σθ to 27.0σθ and AAIW from 27.1σθ to 27.4σθ. As SAMW has almost the same θ and S as ICW [Karstensen and Tomczak, 1997], SAMW also contributes to the water masses in the study area. For the shallower layers of < 26.0σθ, we did not define specific water masses but referred to them simply as subtropical subsurface waters (SSW).
3.2. Distributions of CT Along the I3/I4 Section
 The depth -longitude distributions of CT obtained during BEAGLE are displayed in Figure 3. As these distribution patterns are not very different than those from WOCE, we describe the characteristics of these distributions on the basis of data from BEAGLE.
 From the pattern of isopleths, distributions of CT were largely divided into two parts: the upper layer from the sea surface to about 1000 m and the lower layer from about 1000 m to the bottom. In the upper layer, CT increased gradually from the surface (2000 μmol kg−1) to about 1000 m (2260 μmol kg−1) uniformly from west to east. The concentration isopleths were slightly elevated toward the east. In the lower layer, CT showed small spatial variation of 2290 ± 20 μmol kg−1. There was a steep gradient of CT (50 μmol kg−1 ·100 m−1) from 700 to 1200 m, which was a boundary between the two layers. The maximum CT (2300 μmol kg−1) was found at longitudes 50°E–56°E and 104°E–112°E, but at the different depths of 2600–3400 m and 1800–2600 m, respectively. West of Madagascar (longitude 36°E–42°E), CT was less than 2250 μmol kg−1 below 2000 m.
 The distribution of water masses is also shown schematically in Figure 3, as determined from distributions of potential temperature (θ) and salinity [Uchida and Fukasawa, 2005], with the aid of previous studies [Tomczak and Godfrey, 2001; Fine et al., 2008]. Please refer also to Figure 1 for the bathymetry of the study region. CDW (θ ≤ 1.0°C) occupies the bottom layers (>3800 m) in the Madagascar Basin and the West Australia Basin. CDW is also found in the Central Indian Basin, where it flows over deep saddles in the Ninety East Ridge near 10°S and 5°S and turns south [Tomczak and Godfrey, 2001] below 4500 m at the western edge of the Ninety East Ridge. Above CDW, IDW is widely found from about 3800 m upward to about 1500–2000 m. From these depths to 700 m, AAIW with a salinity minimum is found. In the thermocline, ICW is dominant from about 700 m to 200–300 m.
3.3. Increases of Anthropogenic CO2
3.3.1. Distributions of ΔnCTCAL on Isopycnal Surfaces
 The distributions of nCTCAL on isopycnal surfaces of σθ = 25.8, 26.6, and 27.3 kg m−3 are presented as a function of longitude in Figure 4. The depths of 25.8σθ, 26.6σθ, and 27.3σθ represent SSW (average depth = 230 m), the middle layer of ICW (410 m), and AAIW (950 m), respectively.
 On 25.8σθ, the longitudinal distributions of nCTCAL showed noticeably different patterns for WOCE and BEAGLE. This possibly reflects year-to-year differences of upper-layer oceanic conditions. However, nCTCAL during BEAGLE was generally higher than during WOCE, implying increases of anthropogenic CO2 between the two time periods. The increases were larger in the eastern half of the I3/I4 line, exceeding 10 μmol kg−1. West of 65°E and east of 100°E, longitudinal variations of nCTCAL were larger than those at longitudes in between.
 On 26.6σθ, the distributions of nCTCAL were nearly constant across longitudes during both cruise periods. As a result, increases of nCTCAL were about 8 μmol kg−1 on average. Closer examination, however, shows longitudinal differences such as larger (>10 μmol kg−1) and smaller (6 μmol kg−1) increases at the eastern and western halves of the longitude range, respectively.
 On 27.3σθ, the distributions of nCTCAL revealed increases of about 3–4 μmol kg−1 between the time periods of the two cruises. The increases were indistinct in the western half of the longitude range.
3.3.2. Vertical Distributions of ΔnCTCAL in Thermocline Waters
 Vertical distributions of ΔnCTCAL as a function of depth and σθ are illustrated for each longitudinal range in Figures 5 and 6, respectively. Each panel also includes red lines indicating the penetration depths of CFCs, which were determined from the CFC distributions measured during BEAGLE [Kumamoto and Watanabe, 2007]. As in the work of Murata et al. , we judged that anthropogenic CO2 increases (ΔnCTCAL) were reliably detected to these depths.
 Increases of anthropogenic CO2 could be detected to about 1800 m (27.7σθ) in the westernmost longitudinal range (Figures 5 and 6). The maximum depth of detection became shallower toward the east. In the upper layers of about 700 m (27.0σθ) or less, occupied by SSW and ICW, the ΔnCTCAL values ranged between 5.0 and 10.0 μmol kg−1 in all of the longitudinal ranges, with vertical variations in the distributions. In the longitude range 88°E–120°E, ΔnCTCAL maxima of 12.2 and 12.0 μmol kg−1 were observed at depths of about 200 m (25.7σθ) and 350 m (26.6σθ), respectively. Deeper maxima were found in the other longitudinal ranges at a depth of approximately 700 m (26.9σθ), although the maximum values were lower at 7–9 μmol kg−1. From about 700 m to 1000 m (27.3–27.4σθ), where AAIW was found, ΔnCTCAL showed a steep decrease ranging in magnitude from 2 to 6 μmol kg−1, with similar patterns among the longitudinal ranges.
 Overall, the vertical distributions of ΔnCTCAL revealed distinct longitudinal differences in the upper layers (approximately 700 m and above) but not in the deeper layers. The weighted average ΔnCTCAL for SSW (25.6–25.9σθ), ICW (26.0–27.0σθ), and AAIW (27.1–27.6σθ) along the I3/I4 section was calculated as 7.9 ± 1.1, 7.7 ± 0.5, and 3.8 ± 0.7 μmol kg−1, respectively.
3.3.3. ΔnCTCAL in Deep Waters
 The ΔnCTCAL for σ3 = 41.40, 41.45, 41.50, and 41.55 kg m−3 are listed in Table 1. The upper three of these isopycnal surfaces and 41.55σ3 in longitude range 35°E–45°E were in IDW, whereas 41.55σ3 in the other three longitude ranges was found in upper CDW. On all of these isopycnal surfaces at longitudes 45°E–68°E, a significant ΔnCTCAL was detected. On 41.55σ3, the ΔnCTCAL was significant at all of the longitudinal ranges except for 68°E–88°E, where the 41.55σ3 isopycnal surface did not appear because of the bathymetry.
Table 1. Increases of ΔnCTCAL on Isopycnal Surfaces of σ3 for 1995–2003/2004a
Values are expressed as μmol kg−1 (mean ± standard deviation).
Significant increase (one-tailed t test; P < 0.05).
 To further investigate the ΔnCTCAL in deep waters, water masses with θ ≤ 1.0°C and those of the salinity maximum below 2000 m were selected for further analysis. These selection criteria are associated with CDW and IDW, respectively. Table 2 shows the results of t tests applied to the selected water masses. For the water masses in the deep salinity maximum (IDW), ΔnCTCAL was significant in longitude ranges 35°E–45°E, 45°E–68°E, and 68°E–88°E, whereas for water masses with θ ≤ 1.0°C (CDW), it was significant in longitude ranges 45°E–68°E and 88°E–120°E.
Table 2. Increases of ΔnCTCAL in Selected Water Masses for 1995–2003/2004a
Water Mass Criterion
Values are expressed as μmol kg−1 (mean ± standard deviation).
ND, no data.
Significant increase (one-tailed t test; P < 0.05).
 The weighted average values of ΔnCTCAL for CDW and IDW were 3.2 ± 1.4 and 2.5 ± 1.0 μmol kg−1, respectively. For these calculations, the significant values in Tables 1 and 2 were used.
3.3.4. Water Column Inventory
 The water column inventory of anthropogenic CO2 increases along the I3/I4 section was calculated as in previous studies [Murata et al., 2007, 2008, 2009]. To exclude the seasonal influences dominant in the mixed layer, we used the ΔnCTCAL of 25.6σθ for isopycnal surfaces shallower than 25.6σθ; this surface is generally below the winter mixed-layer depth in the subtropical gyre of the South Indian Ocean [Kara et al., 2003]. Only ΔnCTCAL values from above the depths of penetration (Figures 5 and 6) were used for the calculations. This cutoff probably operates to decrease values of water column inventories. Thus calculated results indicate lower limits.
Table 3 lists the water column inventories along the I3/I4 section separately for each longitudinal range. There is some small variation among the longitudinal ranges. The average uptake rate of anthropogenic CO2 along the I3/I4 section was calculated as 1.0 ± 0.1 mol m−2 a−1.
Table 3. Water Column Inventories of ΔnCTCAL Along the I3/I4 Sectiona
Water Column Inventory (mol m−2)
Maximum Depth (m)
Values are expressed as mean ± standard deviation.
7.1 ± 2.5
1832 ± 54
8.2 ± 2.1
1479 ± 32
8.1 ± 1.5
1230 ± 27
8.6 ± 2.1
1024 ± 23
3.4. Seasonal Influence on ΔnCTCAL
 Because of the Indian monsoon, the Indian Ocean shows distinct seasonal variations in physical and chemical properties. A number of studies have dealt with the large seasonal differences in the CO2 system and related properties [e.g., Bates et al., 2006]. It is possible that changes on this time scale influence the detection of increases in anthropogenic CO2 because this detection depends on snapshot data collected in different seasons; the WOCE and BEAGLE cruises were conducted in April–June and in December–January, respectively.
 To investigate the influence of seasonal variations, we compared the ΔnCTCAL at crossover points along the I3 line. That is, we first defined four crossover points along the I3 line at which the WOCE I5, I8, I9, and I10 lines cross the I3 line (Figure 1 and Table 4). The WOCE observations were also made as part of the Indian Ocean CO2 Survey [Johnson et al., 2002] as well as for the I3/I4 lines. We then selected two WOCE stations close to each crossover point: one from the I3 line and the other from the crossover line. We also selected the BEAGLE stations close to the crossover points along the I3 line. This produced two pairs of ΔnCTCAL values at each crossover point: one from BEAGLE I3 and WOCE I3 stations and the other from BEAGLE I3 and WOCE I5, I8, I9, or I10 stations. From the ΔnCTCAL, we computed differences of ΔnCTCAL (ΔΔnCTCAL) as follows:
where line_x refers to the line that crosses the I3 line (i.e., x = I5, I8, I9, or I10).
Table 4. Averaged Absolute Values of ΔΔnCTCAL at Crossover Points Along the I3/I4 Sectiona
Values are expressed as μmol kg−1 (mean ± standard deviation).
The I5, I8, I9, and I10 lines cross the I3 line at 54.5°E, 80.0°E, 95.0°E, and 111.0°E, respectively (see Figure 1), and the crossover observations were made on 11 June, 25 March, 2 February, and 16 November 1995, respectively.
3.3 ± 3.2
1.4 ± 1.0
2.1 ± 1.3
7.1 ± 7.5
σθ ≤ 26.5
2.0 ± 2.2
1.3 ± 1.1
2.2 ± 1.5
10.6 ± 8.2
σθ > 26.5
5.0 ± 3.6
1.7 ± 1.0
1.9 ± 1.2
2.5 ± 2.2
Table 4 lists the ΔΔnCTCAL at crossover points separately for the full water column (25.0–27.8σθ) and for the upper (≤26.5σθ) and lower (>26.5σθ) layers. The ΔΔnCTCAL was generally less than 5.0 μmol kg−1, and within the error of estimated ΔnCTCAL (Figure 6). However, the values at the line I10 crossover were relatively large. This can be attributed to short-term variability of water masses [Rodgers et al., 2009] because this crossover point is located in an area of mixing with ITW [Fieux et al., 1996]. Considering that the observations at each crossover point were made at different seasons (see a footnote in Table 4) than the I3/I4 observations during the BEAGLE cruise (December–January), we judged the influence of seasonal variations on estimated increases in anthropogenic CO2 to be relatively minor.
4.1. Anthropogenic CO2 Increases in the Indian Ocean
Peng et al.  investigated decadal-scale uptake of anthropogenic CO2 by the ocean on isopycnal surfaces using data for CT and related properties obtained 17 years apart in 1978 and 1995. They found CT increases ranging from 25 to 6 μmol kg−1 on isopycnal surfaces from 26.6σθ down to 27.0σθ between 20°S and 10°S along 80°E. They calculated the inventory as 11.05 mol m−2. Sabine et al.  examined 18 year temporal changes in anthropogenic CO2 storage in the main Indian Ocean basin (north of 35°S), using the multiple linear regression method. The distributions of water column inventory [Sabine et al., 1999, Figure 10] showed that the values along 20°S were 10–13 mol m−2. Their values showing anthropogenic CO2 increases are close to those in the present study (Table 3). However, considering that the intervals between cruises were longer in these previous studies (17 or 18 years) than in the present study (8 years), decadal increases of anthropogenic CO2 are probably somewhat larger in the more recent decade, although direct comparison is difficult because of different methods of calculation.
 To confirm the larger increases of anthropogenic CO2 in the recent decade, we compared the increases between 1978 and 1995 with those between 1995 and 2003/2004 on the basis of the method used in the present study. For the details of the comparison, please refer to Appendix B. Calculated uptake rates of anthropogenic CO2 in respective time periods are listed in Table 5. From Table 5, it is found that the uptake rates show a tendency to be larger in the latter period than in the former period. This result supports the discussion stated above, but again we would like to stress that more detailed investigations are necessary, because the data available were limited.
Table 5. Decadal-Scale Changes of Uptake Rates of ΔnCTCAL at Selected GEOSECS Stationsa
For station 425 by GEOSECS, data collected at stations 545 by WOCE and 545 by BEAGLE were compared. Likewise, for stations 438 and 452 by GEOSECS, data collected at stations 465 and 499 by WOCE and at stations 466 and X08 by BEAGLE were compared.
425 (55.9°E, 17.3°S)
438 (101.3°E, 19.5°S)
452 (80.0°E, 20.1°S)
4.2. Anthropogenic CO2 Increases in Individual Water Masses
 We found that anthropogenic CO2 storage in the ocean increased significantly between the WOCE and BEAGLE time periods (see section 3.3). In addition to the general increasing trend in thermocline waters, we detected some spatial variations in the increases. Here we discuss these variations in terms of the uptake of anthropogenic CO2 at the time of water mass formation.
 To relate the increases of anthropogenic CO2 in the ocean interior to the uptake of anthropogenic CO2 in assumed water formation regions, we made a simple calculation on the basis of the buffer factor (β):
where δpCO2 (μatm or ppmv) implies a change of pCO2. We calculated increases of CT (δCT) by inserting values of CT and pCO2 that are appropriate for the respective water masses (outcrop regions), and values of δpCO2 equal to the increases of atmospheric pCO2.
4.2.1. SSW and ICW/SAMW
 SSW and ICW/SAMW are found from 25.6σθ to 27.0σθ, with estimated water mass ages ranging from 2 to 4 to 20–30 years [Fine et al., 2008]. Thus we calculated δCT using the rates of pCO2 increase (δpCO2) during the period from 1995 to 2005 and in the 1970s.
 For SSW and ICW, we assumed that anthropogenic CO2 was transported into the ocean interior in the area between 20°S and 30°S in the Indian Ocean, which is included in the ventilation region defined by Karstensen and Quadfasel . We then substituted 2050–2100 μmol kg−1 and 300–350 ppmv for CT and pCO2, respectively, and assumed β = 10 [Sabine et al., 2004]. The CT values used are those observed at a depth of about 200 m between 20°S and 30°S along nominal 95°E, which were taken from the WOCE I9N observations (http://cdiac.ornl.gov/ftp/ndp080/). Subsurface values were selected on the assumption that they are representative of winter surface values. The pCO2 values were determined on the basis of the work of Takahashi et al.  and Metzl .
 Using these values for CT and pCO2, and δpCO2 = 1.9 μatm a−1, which is an estimated rate of increase for the period from 1995 to 2005 [Solomon et al., 2007], the δCT for an 8 year interval was calculated to be 9–11 μmol kg−1. If a value of 1.3 μatm a−1 is adopted for δpCO2, which is the rate of increase in the 1970s [Patra et al., 2005], the calculated δCT is 6–7 μmol kg−1.
 To estimate the uptake of anthropogenic CO2 by SAMW, we assumed that the water mass was formed north of the Sub-Antarctic Front and subducted into the ocean interior. We then conducted the same calculation as for SSW and ICW, but setting CT = 2095 μmol kg−1, β = 12, pCO2 = 350 μatm, and δpCO2 = 1.9 μatm a−1. The value used for CT is an average of those observed at a depth of about 200 m between 40°S and 45°S along the WOCE I8S line. The calculations yielded δCT = 8 μmol kg−1. We recalculated δCT with a δpCO2 of 1.3 μatm a−1, which yielded δCT = 5 μmol kg−1.
 These δCT estimates are almost equal to our calculated values for ΔnCTCAL (Figures 5 and 6). Thus it seems that anthropogenic CO2 stored in the upper thermocline water masses generally increases, keeping pace with the atmospheric CO2 increase.
 We detected two ΔnCTCAL maxima with values reaching 12 μmol kg−1 in the longitude range 88°E–120°E: one centered on 25.7σθ and the other centered on 26.6σθ. Because the water masses on the shallower of these isopycnal surfaces have an age of 2–4 year [Fine et al., 2008], a value for δpCO2 of 1.9 μatm a−1 representing the most recent decade is appropriate, and the upper maximum can be accounted for by the uptake of anthropogenic CO2 accompanying SSW and ICW water mass formation. In contrast, the age of the water masses on the lower isopycnal surface are estimated at 8–10 year [Fine et al., 2008]. In this case, using δpCO2 = 1.6 μatm a−1, which is a value for the 1980s through the 1990s [Patra et al., 2005], δCT is 9 μmol kg−1 at maximum.
 In addition to the uptake of anthropogenic CO2 by the water masses, the deeper ΔnCTCAL maximum is probably affected by changes during transport of anthropogenic CO2. Specifically, the vertical distribution of ΔAOU (Figure 7) illustrates that during the BEAGLE cruise, AOU was significantly smaller by 4–7 μmol kg−1 than during WOCE on isopycnal surfaces from 26.8σθ to 27.0σθ at longitudes 88°E–120°E. The smaller AOU might indicate that the water mass was transported faster in the WOCE period than in the BEAGLE period. The intensified northward transport in the thermocline between 1995 and 2002 [Palmer et al., 2004] supports this possibility.
 AAIW extends worldwide, especially in the Southern Hemisphere [Talley, 1996]. In spite of this extensive distribution, the formation area is believed to be restricted to the high latitudes of the South Pacific Ocean [Talley, 1996]. On the basis of this information, we substituted β = 11 or 12 [Sabine et al., 2004], δpCO2 = 0.8–1.3 ppmv a−1, and pCO2 = 330–360 ppmv into equation (4). The values for δpCO2 are the annual rates of increase from the 1960s through the 1970s [Patra et al., 2005] and those for pCO2 are surface seawater values adopted from Murata et al. . These rates of increase were selected because the estimated age of AAIW is 20–30 year [Fine et al., 2008], implying that the water mass was last in contact with the atmosphere in the 1960s through the 1970s. In addition, CT in equation (4) was set to 2110 μmol kg−1, the value observed by the WHP cruises in the high latitudes of the South Pacific Ocean. As a result, δCT = 3–6 μmol kg−1, which is close to the observed increases of anthropogenic CO2 in AAIW (3.8 ± 0.7 μmol kg−1; section 3.3.2). From this we conclude that the accumulation of anthropogenic CO2 in AAIW keeps pace with the increase of atmospheric CO2.
4.2.3. IDW and CDW
 We detected significant increases of anthropogenic CO2 in IDW and CDW (Tables 1 and 2). For CDW, the significant increases were found along the two flow routes of CDW [Tomczak and Godfrey, 2001; Boswell and Wright, 2002; Fine et al., 2008]. Specifically, the increases found at longitudes 45°E–68°E and 88°E–120°E correspond to the western flow route of CDW along the Madagascar continental slope and the eastern flow route leading to the South Australia Basin, respectively. CFC-11 was also detected on the western route at the lowest level (0.02 pmol kg−1) of the BEAGLE cruise [Kumamoto and Watanabe, 2007]. A map of CFC-11 concentrations based on data collected in the WOCE era also shows a slight concentration of 0.01 pmol kg−1 along the western route below 3500 m [Fine et al., 2008].
 CFC-11 is of completely human origin and first appeared in the ocean in the 1950s. Assuming that a water mass formed in the Weddell Sea [Orsi et al., 1999] in the 1950s and was transported [Hoppema, 2004] with little modification, the increase of anthropogenic CO2 can be estimated from equation (4). By substituting pCO2 = 310 ppmv, which is a concentration from the 1940s through the 1950s reconstructed from ice cores [Neftel et al., 1994; Etheridge et al., 1998], β = 14 [Sabine et al., 2004], δpCO2 = 0.4 ppmv a−1 [Neftel et al., 1994; Etheridge et al., 1998], and CT = 2250 μmol kg−1 into equation (4), a value of 2 μmol kg−1 is obtained for δCT. If δpCO2 = 0.8 ppmv a−1 (atmospheric increases in the 1960s; see Patra et al. ) instead of 0.4 ppmv a−1, δCT is calculated at 3 μmol kg−1. These values are close to the values observed in this study (Table 1). Considering that mixing processes probably diluted the concentrations of CFCs in deep water, the higher of these δCT estimates may be more reasonable. That is, the water mass age might be an overestimate, leading to a smaller estimated δpCO2. These results suggest that anthropogenic CO2 is increasing even in the bottom ocean layer in a manner that parallels increases in atmospheric CO2.
 IDW is believed to originate with North Atlantic Deep Water (NADW) [Tomczak and Godfrey, 2001]. Nevertheless, decadal-scale increases of anthropogenic CO2 were not found for NADW in the South Atlantic Ocean [Murata et al., 2008]. In the present study, however, we did find significant increases of anthropogenic CO2 in ICW (Tables 1 and 2), although CFCs were not detected in this water mass [Kumamoto and Watanabe, 2007]. At present, the mechanisms associated with these significant increases in anthropogenic CO2 are not known, but at higher latitudes (32°S), anthropogenic CO2 has already been detected in ICW [Álvarez et al., 2009]. It is possible that water masses that form in the Southern Ocean and are transported northward affect the anthropogenic CO2 signals because waters of Southern Ocean origin are mixed into IDW [van Aken et al., 2004].
4.3. Comparison With Results Obtained in Other Oceans
 Using the same method of calculation as used in the present study, we have estimated increases of anthropogenic CO2 in the South Pacific Ocean [Murata et al., 2007], in the South Atlantic Ocean [Murata et al., 2008], and in the North Pacific Ocean [Murata et al., 2009]. The results are summarized in Table 6. The increases of anthropogenic CO2 in mode waters show a relatively large difference between the oceans, whereas those in intermediate waters have almost equal values in all oceans. The larger difference among the mode waters is probably because mode water formation is strongly affected by interannual variations of upper ocean conditions [Hanawa and Talley, 2001], which also substantially influence uptake and storage of anthropogenic CO2 [Bates et al., 2002].
Table 6. Decadal Increases and Uptake Rates of ΔnCTCAL in Individual Water Massesa
ΔnCTCAL (μmol kg−1)
Uptake Rate (mol m−2 a−1)
Values are expressed as mean ± standard deviation. Abbreviations are as follows: AAIW, Antarctic Intermediate Water; NPIW, North Pacific Intermediate Water; SAMW, Sub-Antarctic Mode Water.
The mode waters are Subtropical Mode Water, South Pacific Subtropical Water, and North Pacific Subtropical Water.
 Note that the increases in AAIW are similar among the oceans. The source region of AAIW is thought to be confined to the high latitudes of the South Pacific Ocean [Talley, 1996]. This implies that AAIW found in regions far from the source region should be older than AAIW found near the source region. Accordingly, the increases of anthropogenic CO2 in AAIW far from the source region should be smaller than those from near the source region. However, the observed increases do not support this expectation. We therefore infer that formation or transport of AAIW has recently become weaker. A quantitative discussion of this inference is beyond the scope of the present study.
 If the uptake of anthropogenic CO2 by the ocean is distributed evenly over the global ocean, the uptake rate should be about 0.55 mol m−2 a−1 [Sabine et al., 2008]. The uptake rates in the North Pacific and South Atlantic Oceans are close to this calculated value, whereas those in the South Pacific and South Indian Oceans are greater than this value (Table 6). In the latter two oceans, significant deep increases of anthropogenic CO2 were detected to water depths of 1500 m and more (see Murata et al. [2007, Table 3] and Table 6), whereas in the other oceans, significant increases were detected only down to 1000 m at the deepest. Thus we speculate that differences in the penetration depths of anthropogenic CO2 result in the basin-to-basin differences in the uptake rates. The mechanisms transporting anthropogenic CO2 into the deeper layers may be more effective in the South Pacific and South Indian Oceans.
5. Concluding Remarks
 Using high-quality data for CT and related properties, we could detect significant decadal-scale increases of anthropogenic CO2 in the subtropical region of the South Indian Ocean. We related the increases mostly to water masses of Southern Ocean origin. We found that the anthropogenic CO2 in individual water masses increased between 1995 and 2003/2004 at a pace paralleling or locally exceeding that of atmospheric CO2 increases (section 3.3). In addition, the uptake rate of anthropogenic CO2 in the South Indian Ocean was larger than rates in other oceans (section 4.3). Together these results suggest an effective sink for anthropogenic CO2 in the Southern Ocean, contradicting Le Quéré et al.  who pointed out a weakening of the CO2 sink in the Southern Ocean between 1981 and 2004.
 In contrast, Mikaloff Fletcher et al.  showed that ITW plays a critical role in determining the transports of anthropogenic CO2 in the Indian and Pacific Oceans. In the present study, there were water masses influenced by ITW in the upper layers, judging from the θ-S diagram (Figure 2). In fact, increases of anthropogenic CO2 in the upper layers became larger in the easternmost longitudinal ranges in this study (section 3.3.2), which is close to a region influenced by ITW. However, the impacts of ITW on anthropogenic CO2 increases seem to be less than that from water masses of Southern Ocean origin. These contradictory results highlight the need for more data to explain anthropogenic CO2 increases in the South Indian Ocean. Furthermore, we have to consider calculation methods used for anthropogenic CO2, differences of which often cause nonnegligible differences of anthropogenic CO2 estimates [Álvarez et al., 2009].
 It is the merit of this work that a consistent approach is used to estimate anthropogenic CO2 increases in world oceans, which makes it possible to find regional differences. However, as pointed in some parts of the present study, there existed uncertainty and bias in the estimated results, which were not considered enough quantitatively. Thus caution should be made when one uses the results to discuss carbon budget. Further investigations are necessary to obtain more accurate estimation.
Appendix A:: Validation of Assumptions
 To estimate increases of anthropogenic CO2 using a simple calculation method (section 2.2), we assumed that both ΔCTdiseq and ATm remained unchanged between the two time periods when data were collected. In this appendix, we justify this assumption with the aid of results from previous studies.
 The assumption of a constant ΔCTdiseq is an assumption that we have to discuss enough [Gruber et al., 1996]. The reality of the assumption can be verified by checking for temporal changes in the air -sea difference in partial pressures of CO2 (ΔpCO2). If the time series of ΔpCO2 shows a decreasing or increasing trend, the assumption does not hold true.
Metzl  detected average rates of increase in ΔfCO2 (≈ΔpCO2) of 1.8 and 2.2 μatm a−1 for winter and summer, respectively, in the South Indian subtropical zone (20°S–35°S) over 16 years from 1991 to 2007. As this area corresponds to a formation area of subtropical mode waters, the changes in ΔpCO2 can cause a local bias in calculated increases of anthropogenic CO2 in the upper layers. If the changes of ΔpCO2 affected the uptake of anthropogenic CO2 between the WOCE and BEAGLE time periods, the changes in ΔnCTCAL would be about 2 μmol kg−1. This value is rather large as a bias. On the other hand, a recent global survey using an increased number of surface seawater pCO2 data shows a rate of oceanic pCO2 increase similar to that for the atmospheric CO2 increase [Takahashi et al., 2009], meaning that there has been little change in ΔpCO2. A separate comprehensive review similarly reported that surface seawater pCO2 had generally followed atmospheric CO2 over more than two decades [Bindoff et al., 2007]. We conclude, therefore, that the assumption of constant ΔCTdiseq holds true, although there is a possibility that local ΔpCO2 changes impact ΔCTdiseq.
 To confirm whether ATm remained constant between the WOCE and BEAGLE time periods, we investigated differences in nATm (ΔnATm). The vertical distributions of ΔATm for each longitudinal range in this study as a function of isopycnal surfaces are shown in Figure A1. It is evident that nATm during BEAGLE was smaller overall (−5.6 ± 2.2 μmol kg−1) than in WOCE. The same tendency was found in ΔnATm on isopycnal surfaces of σ3 (Table A1). The average difference was −7.9 ± 1.7 μmol kg−1. As the differences were almost equal throughout the vertical distributions with little longitudinal variation, we attributed the differences not to the ATm signals but to systematic errors in ATm. A data synthesis study by Lo Monaco et al.  confirmed the probable inclusion of systematic errors of almost the same magnitude (−10 μmol kg−1) in the ATm data.
Table A1. Differences in nATm (ΔnATm) Between the WOCE and BEAGLE Time Periodsa
Time periods are WOCE, 1995, and BEAGLE, 2003/2004. Values are expressed as μmol kg−1 (mean ± standard deviation).
Significant increase (one-tailed t test; P < 0.05).
 Closer inspection of Figure A1 reveals that the magnitude of ΔnATm decreased in the upper layers compared to the deeper layers. This could be interpreted as indicating an increase in AT in the upper layers. According to recent studies on ocean acidification, increases of AT are expected in the upper layers of the ocean because of the reduction of calcification in response to changes of the CaCO3 saturation state. In fact, some studies report an increase of AT [e.g., Sarma et al., 2002], although there are questions as to whether the detected changes of AT are significant [Friis et al., 2006].
 As for decadal changes, a model study showed that increases of AT between WOCE and WOCE revisit cruises are so small (<1 μmol kg−1) that they are not yet detectable with currently available data [Ilyina et al., 2009]. If the decreases of ΔnATm in the upper layers are real and are due to increases of AT, and assuming that the increases are about 5 μmol kg−1 (the approximate difference of ΔnATm between the upper and lower layers), then increases of anthropogenic CO2 in the upper layers should be overestimated by 2–3 μmol kg−1.
 On this basis, we judged that the observed ΔnATm was due to a systematic error, and even accepting an increase of AT resulting from ocean acidification, the effect on the estimated increases of anthropogenic CO2 were negligible. Accordingly, we concluded that the assumption of constant AT is applicable to the present study.
Appendix B:: Decadal-Scale Changes of Rates of Anthropogenic CO2 Storage
 To investigate temporal variations of anthropogenic CO2 storage in the South Indian Ocean, we used data obtained by GEOSECS, which were downloaded from the data site (http://irid.ldeo.columbia.edu/SOURCES/GEOSECS/). We selected three GEOSECS stations (425, 438 and 452) occupied in 1978, considering the proximity to the WOCE I3 line (Figure 1). By deep water (>2000 m) comparison between GEOSECS and WOCE/BEAGLE data, we found an offset of +4 μmol kg−1 and +13 μmol kg−1 for GEOSECS CT at stations 425 and 438, respectively. After adding the offsets to the original data, we calculated uptake rates of anthropogenic CO2 on a station-by-station basis from the ΔnCTCAL between GEOSECS and WOCE, and between WOCE and BEAGLE, following the same method as used in section 3.3.4.
 We thank the officers and crew of the R/V Mirai for 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 and also worked to conduct the highest-level quality control of data obtained from the shipboard observations.