Decadal changes in dissolved inorganic carbon in the Pacific Ocean


  • Shinya Kouketsu,

    Corresponding author
    1. Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Yokosuka, Japan
    • Corresponding author: S. Kouketsu, Japan Agency for Marine-Earth Science and Technology 2-15 Natsushima-cho, Yokosuka 237-0061, Japan. (

    Search for more papers by this author
  • Akihiko Murata,

    1. Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Yokosuka, Japan
    Search for more papers by this author
  • Toshimasa Doi

    1. Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology (JAMSTEC), Yokosuka, Japan
    Search for more papers by this author


[1] Using high-quality data sets obtained about a decade apart, we examined the changes of dissolved inorganic carbon in the Pacific Ocean, separating anthropogenic and natural CO2. Observations along three transoceanic sections along 47°N, 179°E, and 17°S showed both decadal increases (>20 µmol kg–1) and decreases (<−20 µmol kg–1) of anthropogenic CO2 in thermocline waters. As a result, the decadal-scale storage of anthropogenic CO2 north of 40°N in the North Pacific was close to ±0 mol m–2 a–1, except in the western subarctic Pacific. In contrast, in subtropical regions of both hemispheres, we found an increasing trend of >10 µmol kg–1 in oceanic uptake of anthropogenic CO2, reflecting accumulation in mode waters. Along 17°S, increases of anthropogenic CO2 were >20 µmol kg–1, larger than expected from increases of anthropogenic CO2 in the atmosphere. The annual water-column inventories of anthropogenic CO2 changes calculated in 20° longitudinal or 10° latitudinal bands throughout the Pacific Ocean revealed relatively high values (>0.7 mol m–2 a–1) in the subtropical regions of both hemispheres and low values in the tropical Pacific. This distribution pattern is similar to previous estimates for the Anthropocene, implying that the redistribution processes of anthropogenic CO2 have not changed on a basin scale over the last decade. We estimated the total anthropogenic and natural CO2 storage in the Pacific Ocean to be 8.4 ± 0.5 and 0.6 ± 0.4 Pg carbon decade–1, respectively.

1 Introduction

[2] The ocean acts as a major sink for anthropogenic CO2. Solomon et al. [2007] estimated that the ocean has absorbed up to about 30% of the planet's anthropogenic CO2 in recent decades. This estimate is based on data from international observation programs, such as the World Ocean Circulation Experiment (WOCE) and the Joint Global Flux Study (JGOFS), which were conducted mostly in the 1990s, and a subsequent synthesis [Key et al., 2004]. Of interest now is to determine whether the ocean is continuing to absorb anthropogenic CO2 at the same rate as it has in the past. During the last decade, we conducted repeat observation cruises (hereinafter referred to as WOCE revisit cruises) along some of the original WOCE lines under the international framework of the Climate Variability and Predictability/CO2 Repeat Hydrography program. These repeat cruises enabled us to detect recent decadal changes of anthropogenic CO2 storage in the ocean interior.

[3] To detect increases in anthropogenic CO2, high-quality data about the CO2 system and related properties are necessary. Dissolved inorganic carbon (CT, µmol kg–1) is a useful indicator of CO2 levels because CT is increased directly by absorption of anthropogenic CO2. However, the decadal change in CT is only about 0.5%. Thus, decadal changes in anthropogenic CO2 in the ocean have been detected mainly along the WOCE lines, where high-quality data have been collected repeatedly. For example, by reoccupying WOCE lines and collecting high-quality data, Murata et al. [2007, 2008, 2009, 2010] detected significant increases of anthropogenic CO2 (0.5–1.0 mol m–2 a–1) in the subtropical gyres of the South Pacific, the South Atlantic, the North Pacific, and the South Indian oceans. In particular, Murata et al. [2010] highlighted large temporal and spatial variabilities in anthropogenic CO2 storage in the Indian Ocean, especially in mode waters. Brown et al. [2010] examined data collected during three time periods along 24.5°N in the North Atlantic and reported that anthropogenic CO2 does not accumulate at a steady rate. Waters et al. [2011], who recently investigated changes in anthropogenic CO2 in the South Pacific, found that, compared with the eastern South Pacific, the uptake rate was greater in the western South Pacific, which they attributed to the formation of subtropical mode water there.

[4] These results demonstrate that there are large spatial and temporal variations in oceanic absorption of anthropogenic CO2 and call into question the assumption, often adopted in model calculations, of a constant uptake rate of anthropogenic CO2 by the ocean. In addition, Sabine et al. [2008] revealed that changes in the natural CO2 concentration in the North Pacific Ocean have a large impact on decadal changes in the water column CO2 inventory. This result implies that we have to consider not only anthropogenic, but also natural CO2 changes because both play important roles in the carbon cycle and, accordingly, in global environmental change.

[5] The purpose of this study was to examine changes in CT, which we expected to be small but significant, and to quantify the uptake rates of anthropogenic CO2 in the 1990s and the 2000s in the Pacific Ocean. We outline the methods used to detect the decadal changes in CT and anthropogenic CO2 in section 2. In section 3, we present the results of our investigation of the changes in CT along three selected vertical WOCE sections crossing the Pacific Ocean (Figure 1) and report the differences in carbon levels between the WOCE revisit sections (reported here for the first time) and the same WOCE sections visited about a decade earlier. In examining the changes in CT, we distinguish between the contributions of anthropogenic and natural CO2. Next, we present the calculated changes in water column inventories of anthropogenic and natural CO2 (section 4). In the discussion (section 5), we first assess the uncertainty of our calculated results and the shortcomings of our method. Then, we discuss the characteristics of the decadal changes in CT, in comparison with the findings of previous studies, and summarize the uptake rates of anthropogenic CO2 in the Pacific Ocean during recent decades.

Figure 1.

Hydrographic sampling sections (solid lines) used in this study (Table 1). The total anthropogenic CO2 uptake in the Pacific Ocean was calculated over the gray area (section 5.6).

Table 1. List of WOCE Sections Used in This Study With Dates of Occupation
Section NameInitial SurveyRevisit
  1. The superscripts aj correspond to those in Figures 9-11.

P01aMay and Sep 1999Aug and Oct 2007
P02bJan. 1994Jul 2004
P06cMay 1992Sep 2003
P06dSep. 2003Dec 2009
P10eOct. 1993Jun 2005
P14fSep 1992 and Aug 1993Oct and Dec 2007
P16gMar and Sep 1991Jan 2005 and and Mar 2006
P17hJun 1993Aug 2001
P18iMar 1994Jan 2008
P21jMay 1994May 2009

2 Methods

2.1 Changes in CT and Anthropogenic CO2 in Vertical Sections

[6] All data about the CO2 system and related properties used in this study are from the Carbon Dioxide Information and Analysis Center (, unless stated otherwise. For the vertical sections, we used data from three transoceanic transects, namely, WOCE lines P01, P14, and P21 (Figure 1) obtained in 1992–1999 and revisited in 2007 and 2009 (Table 1).

[7] We calculated the decadal changes of CT (∆CT) along the three vertical sections by subtracting the CT values obtained in the original WOCE cruises from those obtained in the WOCE revisit cruises. Along both latitudinal and longitudinal sections, we interpolated the observed CT data on the 0.05 kg m–3 × 0.2° neutral density (γn) surfaces.

[8] We estimated the changes in anthropogenic CO2 in the same way as Murata et al. [2007] by applying the ∆C* method [e.g., Gruber et al., 1996], which we simplified by making certain assumptions, to the two data-collection periods (WOCE and WOCE revisit). We defined the anthropogenic CO2 concentration in the ocean interior (CTANT, µmol kg–1) as shown in equation (1) [e.g., Gruber et al., 1996; Sabine et al., 2002a, 2002b]:

display math(1)

where γC:O is the Redfield ratio for carbon and oxygen; following Murata et al. [2007], we used the value of 0.69 [Anderson and Sarmiento, 1994]. AOU is apparent oxygen utilization, which is defined as the difference between the observed concentration of dissolved oxygen (DO; µmol kg–1) and the saturated DO concentration at each potential temperature and salinity. CTm (µmol kg–1) and ATm (µmol kg–1) are measured CT and measured total alkalinity (AT, µmol kg–1), respectively. AT0 (µmol kg-1) is the preformed AT, which we assumed to be constant from preindustrial times to the present. CT0 (µmol kg-1) is the theoretical CT of the water in equilibrium with the atmosphere without anthropogenic CO2. ∆CTdiseq (µmol kg–1) is the difference between CT in the mixed layer in equilibrium with atmospheric CO2 and CT at the time of the water-mass formation.

[9] When the changes in anthropogenic CO2 between the two data-collection periods are calculated, both CT0 and AT0 cancel out. Furthermore, we assumed that on a decadal time scale, changes in ATm and ∆CTdiseq are negligible and that these values could thus be ignored. Therefore, we calculated the change of anthropogenic CO2 (∆nCTCAL, µmol kg–1) over the decade as shown in equation (2):

display math(2)

where nCTCAL(tr) and nCTCAL (tw) are the normalized preformed CT values (= CTm − γC:O × AOU) for the two sets of observations, WOCE revisit (tr) and WOCE (tw), respectively. The values were normalized to a salinity of 35 to remove the influence of changes caused by the addition or removal of fresh water; the normalization did not greatly affect the results.

[10] We defined preformed natural CT (CAOU, µmol kg–1) as equal to γC:O × AOU. Thus, we computed the changes in natural CO2 (∆nCAOU, µmol kg–1) as shown in equation (3):

display math(3)

[11] Both ∆nCTCAL and ∆nCAOU were calculated on given neutral density surfaces after interpolating the individual properties (CT, AOU, and salinity) on the surfaces. Note that ∆nCAOU reflects changes due to both circulation and biological activity, because below the mixed layer, AOU is related to remineralization processes by the Redfield ratio. Thus, it is considered to represent how long and by how much water masses have been affected by remineralization processes.

2.2 Water Column Inventories of ∆nCTCAL and ∆nCAOU

[12] First, we calculated 20° longitudinal or 10° latitudinal averages of ∆nCTCAL and ∆nCAOU on neutral density surfaces along the sections. At each station, we interpolated nCTCAL and nCAOU on neutral density surfaces at 0.1 kg m–3 intervals, and then we calculated the longitudinal or latitudinal interval averages of nCTCAL and nCAOU and their standard deviations (SDs) on the neutral density surfaces. Because the station locations were different between the original WOCE cruises and the revisits, we calculated the interval averages and SDs of ∆nCTCAL and ∆nCAOU by using equations (2) and (3) with weights based on the station intervals. Then, we estimated annual ∆nCTCAL and ∆nCAOU rates (µmol kg–1 a–1) by dividing the averaged ∆nCTCAL and ∆nCAOU by the number of elapsed years.

[13] We estimated the specific water column inventories of ∆nCTCAL and ∆nCAOU (mol m–2 a–1) by integrating ∆nCTCAL and ∆nCAOU rates from the neutral density surface at the bottom of the wintertime mixed layer (WML) to the neutral density surface γn = 27.6 kg m–3. The WML, defined as the depth where the density is 0.03 kg m–3 greater than that at the surface, was calculated by using data from the World Ocean Atlas 2005 [Locarnini et al., 2006; Antonov et al., 2006]. Then, following Murata et al. [2007], we calculated ∆nCTCAL and ∆nCAOU values within the WML, assuming ∆nCTCAL and ∆nCAOU rates were the same as those just below the WML and the calculated ∆nCTCAL and ∆nCAOU values to the values below the WML to obtain the final water column inventory values. We did not compute ∆nCTCAL and ∆nCAOU within the WML from actual measurements because the measurements were made in different seasons and were therefore influenced by seasonal variations.

[14] We used the same methods to calculate the decadal water column inventories along other available observation sections in the Pacific Ocean (Figure 1 and Table 1) in order to produce horizontal distribution maps of the ∆nCTCAL and ∆nCAOU water column inventories.

3 Dissolved Inorganic Carbon Changes in Vertical Sections

3.1 Section P01 (Along Latitude 47°N)

[15] The ∆CT distribution (Figure 2a) varied little (0 ± 5 µmol kg–1) in the layers below γn = 27.5 kg m–3, and this feature was also characteristic of other sections (shown later), which implies constant conditions in deep waters. CT decreased strongly (by >20 µmol kg–1) east of 180° at γn = 26.5–27.0 kg m–3. The negative ∆CT values east of 180°, which were due mainly to a decrease in ∆nCAOU (Figure 2c), are consistent with previously reported decadal-scale decreases in AOU [Kouketsu et al., 2010]. The large increases of CT detected above γn = 26.0 kg m–3 were due mainly to seasonal differences because the observations were made at different times of the year (Table 1).

Figure 2.

Distributions of (a) ∆CT, (b) ∆nCTCAL, and (c) ∆nCAOU (µmol kg–1) along 47°N (section P01) from 1999 to 2007. In each panel, the black line indicates the WML density calculated from World Ocean Atlas 2005 data.

[16] Below γn = 27.5 kg m–3, ∆nCTCAL was small (0 ± 5 µmol kg–1), indicating that the anthropogenic CO2 did not reach the deep layers during the interval between the original WOCE and revisit cruises (Figure 2b). The area of large negative ∆nCTCAL extending from 165°E to 150°W within γn = 26.5–27.0 kg m–3 shows that anthropogenic CO2 did not increase over the last decade in thermocline waters along the P01 section, except west of 165°E. ∆nCTCAL was more than 5 µmol kg–1 at γn = 26.7–27.3 kg m–3 throughout the section, except around 165°E. The reason for the positive ∆CT and ∆nCTCAL values detected east of 165°W in deep layers extending down to γn = 28.0 kg m–3 is unknown.

[17] At some locations, ∆nCTCAL was smaller than ∆CT, or even opposite in sign. For example, at γn = 26.8–27.5 kg m–3 and 160°E–165°E, ∆nCTCAL was positive (Figure 2b), whereas ∆CT was negative (Figure 2a). The trends on these density surfaces were also opposite at 165°E–172°E. These contrasts between ∆nCTCAL and ∆CT are probably related to physical and biogeochemical changes in this region, as discussed in section 5.2.

[18] We calculated the rate of change of anthropogenic CO2 (µmol kg–1 a–1) as ∆nCTCAL averaged over each 20° longitudinal interval and divided by the number of elapsed years (Figure 3). Because of local positive and negative patterns (Figure 2b), the standard errors (SEs) are relatively large, compared with those in the other sections shown later. On the density surfaces γn = 26.0 to 27.0 kg m–3, the rate of change was negative east of 180°. A significant increase in anthropogenic CO2 was detected around γn = 27.0 kg m–3 in all longitudinal bands.

Figure 3.

∆nCTCAL rates (µmol kg–1 a–1) along 47°N, section P01. The rates were calculated by dividing the averaged ∆nCTCAL in each 20° longitudinal band by the number of elapsed years. Error bars indicate SEs.

3.2 Section P14 (Along the 179°E Meridian)

[19] At around 40°N along section P14 (Figure 1), ∆CT of more than 5 µmol kg–1 extended down to γn = 27.4 kg m–3 (Figure 4a), which is below the lower boundary of North Pacific Intermediate Water (NPIW). At around 20°N, ∆CT of more than 5 µmol kg–1 reached γn = 27.0 kg m–3, which corresponds to the NPIW salinity minimum. Large positive ∆CT values of more than 10 µmol kg–1 were detected at densities of 26.0–26.5 kg m–3 north of 30°N and 25.2–26.0 kg m–3 south of 30°N. These density ranges correspond to North Pacific Subtropical and Central Mode Waters (NPSTMW and NPCMW), respectively. These large ∆CT values were associated with positive ∆nCAOU around these density ranges (Figure 4c), as previously reported [Kouketsu et al., 2010].

Figure 4.

Distributions of (a) ∆CT, (b) ∆nCTCAL, and (c) ∆nCAOU (µmol kg–1) along 179°E (section P14) from 1992/1993 to 2007. In each panel, the black line indicates the WML density calculated from World Ocean Atlas 2005 data.

[20] Below the WML, the ∆nCTCAL (Figure 4b) and ∆CT distribution patterns were generally similar, although the magnitude of ∆CT was smaller in most areas. As with ∆CT, positive ∆nCTCAL values extended down to γn = 27.4 kg m–3 around 40°N. Moreover, we found fewer discrepancies between ∆CT and ∆nCTCAL below the WML along this section than along section P01. Nevertheless, we note a few contrasts, albeit weak, along section P14: South of 25°N, we observed negative ∆CT values of <−5 µmol kg–1 and positive values of >5 µmol kg–1 in a banded pattern below the γn = 27.0 kg m–3 density surface (Figure 4a), whereas ∆nCTCAL values in this density range are close to zero and the bands are indistinct (Figure 4b).

[21] Except north of 40°N, we detected significant positive rates of change in anthropogenic CO2 from the surface down to γn = 27.0 kg m–3 in all latitudinal bands (Figure 5), implying that the mid-latitude regions along section P14 have robustly accumulated anthropogenic CO2 over the past decade. Near where sections P14 and P01 intersect, the vertical profile of section P14 along 179°E between 40°N and 50°N (Figure 5g) resembles the vertical profiles of section P01 along 47°N in the 160°E–180° and 180°–160°W bands (Figures 3b and 3c, respectively). As the observations along the two sections were carried out at different times of the year, this similarity suggests that the changes near their intersection are due not to seasonal variations, but to decadal-scale changes.

Figure 5.

∆nCTCAL rates (µmol kg–1 a–1) along 179°E, section P14. The rates were calculated by dividing the averaged ∆nCTCAL in each 10° latitudinal band by the number of elapsed years. Error bars indicate SEs.

3.3 Section P21 (Along Latitude 17°S)

[22] ∆CT values were large (>20 µmol kg–1) on the γn = 26.0–27.0 kg m–3 density surfaces in the eastern part of section P21 (Figure 1) and above γn = 26.5 kg m–3 in the western part of the section (Figure 6a). The large increase in the eastern part is in the Subantarctic Mode Water (SAMW) region. ∆CT was <−5 µmol kg–1 and >5 µmol kg–1 at γn = 27.0–27.8 kg m–3 around 120°W–130°W and 115°W–120°W, respectively, and in these two zones ∆nCAOU was also negative and positive, respectively (Figure 6c), indicating increases and decreases of DO. Thus, these changes were probably caused by water-mass shifts. Consistent with this inference, we observed no negative values in the corresponding parts of the ∆nCTCAL distribution (Figure 6b).

Figure 6.

Distribution of (a) ∆CT, (b) ∆nCTCAL, and (c) ∆nCAOU (µmol kg–1) along 17°S (section P21) from 1994 to 2009. In each panel, the black line indicates the WML density calculated from World Ocean Atlas 2005 data.

[23] ∆nCTCAL values of more than 5 µmol kg–1 extended to the γn = 27.5 kg m–3 density surface, more prominently in the western part of this section. The anthropogenic CO2 may have been transported by deep circulation associated with Antarctic Intermediate Water (AAIW).

[24] We calculated the annual rate of change in anthropogenic CO2 averaged over 20° longitudinal zones along section P21 (Figure 7). The vertical profiles in the five longitudinal bands from 140°E to 120°W (Figures 7a–7e) are similar, with relatively large increase rates in the density range corresponding to SAMW and in the deep layers down to γn = 27.0 kg m–3. Moreover, the SEs are relatively small (<1.0 µmol kg–1 a–1), compared with those calculated for other sections (Figures 3 and 5), indicating that few factors affected the storage of anthropogenic CO2.

Figure 7.

∆nCTCAL rates (µmol kg–1 a–1) along 17°S, section P21. The rates were calculated by dividing the averaged ∆nCTCAL in each 20° longitudinal band by the number of elapsed years. Error bars indicate SEs.

[25] As changes in natural and anthropogenic CO2 in the vertical sections other than P01, P14, and P21 have been previously reported [e.g., Murata et al., 2007, 2009; Sabine et al., 2008; Waters et al., 2011], we do not describe them here (but we show them in the supporting information).

4 Specific Water Column Inventories

4.1 Sections P01, P14, and P21

[26] We calculated specific water column inventories of ∆nCTCAL and ∆nCAOU in longitudinal or latitudinal bands along each section (Figure 8).

Figure 8.

Specific water column inventories of ∆nCTCAL (gray) and ∆nCAOU (white) in mol m–2 a–1 averaged over 20° longitudinal or 10° latitudinal bands along (a) P01, (b) P14, and (c) P21. The specific water column inventories were calculated by vertically integrating the ∆nCTCAL (Figures 3, 5, and 7) and ∆nCAOU rates.

[27] Along section P01 (Figure 8a), the longitudinally averaged water column inventories of ∆nCTCAL were close to zero, except in the 140°E–160°E band, where the water column inventory exceeded 0.5 mol m–2 a–1. The natural CO2 (∆nCAOU) inventories were generally negative and larger at the western and eastern ends of the section.

[28] Along section P14 (Figure 8b), relatively large latitudinally averaged ∆nCTCAL values of more than 0.5 mol m–2 a–1 were detected between 10°N and 40°N, where increases of anthropogenic CO2 in NPSTMW and NPCMW contributed to the carbon inventory. The large increase in the ∆nCAOU water column inventory was found between 20°N and 30°N.

[29] Along section P21 (Figure 8c), the longitudinally averaged ∆nCTCAL water column inventories were larger in the western (0.6–0.8 mol m–2 a–1) than in the eastern part (0.4–0.6 mol m–2 a–1), caused by smaller ∆nCTCAL values in the deep layers of the eastern part (Figure 6b).

4.2 Specific Water Column Inventories in the Whole Pacific Ocean

[30] We calculated specific water column inventories along the other available Pacific Ocean observation sections in the same way as described in section 4.1 and found that ∆nCTCAL and ∆nCAOU ranged from 0.0 to 1.0 mol m–2 a–1 and from −0.4 to 0.6 mol m–2 a–1, respectively. Along all sections (Figure 9), specific water column inventories of ∆nCTCAL were relatively high in the subtropical regions of both hemispheres (>0.5 and >0.7 mol m–2 a–1 in the Northern and Southern hemispheres, respectively). In contrast, specific water column inventories were lower (0.0–0.5 mol m–2 a–1) in tropical regions. This distribution pattern is generally in accordance with that obtained for long-term anthropogenic CO2 storage since the Industrial Revolution (IR) [Sabine et al., 2002a].

Figure 9.

Distributions of specific water column inventories of ∆nCTCAL (mol m–2 a–1), averaged over 20° longitudinal or 10° latitudinal bands, over the whole Pacific. Black lines denote the observation sections. Superscript letters correspond to the section names in Table 1. Subscripts show the SDs of the averages. Values greater than 1 or 2 SDs of the average for the entire Pacific Ocean are colored yellow and pink, respectively.

[31] Nevertheless, we found some regional differences in the patterns between decadal-scale and long-term storage values. In the western Pacific between the equator and 20°N, the specific water column inventories calculated here are relatively large, comparable to those north of 20°N, whereas Sabine et al. [2002a] reported their smallest water column inventories in that region. In contrast, in the subarctic North Pacific, the inventories calculated here are smaller than the long-term estimates of Sabine et al. [2002a] and some are even negative.

[32] Distributions of anthropogenic CO2 are often accounted for by water-mass distributions in the thermocline [Sabine et al., 2004; Murata et al., 2007], because individual water masses take up anthropogenic CO2 at the sea surface at the time of their formation. To relate the spatial variations in the specific water column inventories of ∆nCTCAL (Figure 9) to the water masses, we calculated the inventories separately for mode (∆nCTCAL (MW)) and intermediate (∆nCTCAL (IW)) waters (Figure 10). We defined MW as those in the density range γn = 25.6–26.5 kg m–3 in the North Pacific and γn = 25.6–26.8 kg m–3 in the South Pacific and IW as those in the density ranges γn = 26.6–27.6 kg m–3and γn = 26.9–27.6 kg m–3 in the North Pacific and South Pacific, respectively. In the North Pacific, ∆nCTCAL (MW) of >0.2 mol m–2 a–1 contributed to the large ∆nCTCAL values of >0.5 mol m–2 a–1 at 10°N–40°N (Figure 9), accounting for more than 50% of the inventory values. In the South Pacific, large ∆nCTCAL values of >0.7 mol m–2 a–1 were detected across the bands south of 20°S, generally recognized as the formation region and circulation area of AAIW and SAMW. Consistent with this, both ∆nCTCAL (MW) and ∆nCTCAL (IW) were relatively large in this region (Figure 10).

Figure 10.

Distributions of specific water column inventories of ∆nCTCAL (mol m–2 a–1), averaged over 20° longitudinal or 10° latitudinal bands, in (a) MW and (b) IW density ranges. Black lines denote the observation sections. In the North Pacific, the mode and intermediate waters are defined by the density ranges γn = 25.6–26.5 kg m–3 and γn = 26.6–27.6 kg m–3, respectively, and in the South Pacific by the ranges γn = 25.6–26.8 kg m–3 and γn = 26.9–27.6 kg m–3, respectively. Encoding as in Figure 9.

[33] Large specific water column inventories of ∆nCTCAL were found in the subtropical gyres, where contributions of ∆nCTCAL (MW) were large. However, large inventories of ∆nCTCAL were also detected in the easternmost part of the North Pacific (Figure 9), where MW are not usually observed. These large inventories might be due to decadal water-mass shifts [e.g., Deutsch et al., 2006; Di Lorenzo et al., 2009] as well as to the effects of noisy measurements, which are reflected in the large SEs found there.

[34] The specific water column inventories of ∆nCAOU in the Pacific Ocean (Figure 11) were generally smaller than those of ∆nCTCAL (Figure 9), indicating that the contribution of ∆nCAOU to the total changes in CT was small. This result implies that the natural carbon cycle has not been altered substantially during the time covered by the observations, at least on a basin scale. Nevertheless, we detected large ∆nCAOU values, irrespective of sign, in areas close to land. These large changes may be related to processes specific to land-ocean systems. Furthermore, ∆nCAOU was also large at around 40°N in the eastern North Pacific, around the boundary zone between the subtropical and subarctic gyres. The characteristic oceanic conditions in this boundary region may have contributed to the large ∆nCAOU. Note that because the changes detected in this region were based on data from a limited number of sampling sites, the data are probably insufficient as a basis for a discussion on high spatial variability in boundary areas. Along 30°N, ∆nCAOU was low in the western part and high in the eastern part, perhaps reflecting changes in the thickness of the surface layer due to the large-scale baroclinic Rossby wave propagation previously reported in this area [e.g., Qiu, 2003; Qiu et al., 2007], because Rodgers et al. [2009] pointed out that surface-layer thickness changes can affect the estimation of water column inventories. In this study, the estimation was affected only by thickness changes and not by high CT concentration water in the deep layer, because the water column inventories were integrated from isopycnal surfaces.

Figure 11.

Distributions of specific water column inventories of ∆nCAOU (mol m–2 a–1) averaged over 20° longitudinal or 10° latitudinal bands. Black lines denote the observation sections. Encoding as in Figure 9.

5 Discussion

5.1 Biases

[35] We used high-quality data collected by WOCE and WOCE revisit cruises to examine decadal changes in CT. However, even though we used the most accurate data currently available, we could not avoid systematic measurement biases. To estimate the systematic biases in our results, we compared ∆CT and ∆nCAOU values from the deepest layers, where we can safely assume that decadal changes in the measured water properties are very small. Thus, these values should be a good indication of any systematic biases of our measurements. For this comparison, we selected the neutral density surface γn = 27.8 kg m–3.

[36] The section-averaged ∆CT values between the WOCE and WOCE revisit data at γn = 27.8 kg m–3 (Table 2) were mostly within ±0.3 µmol kg–1 a–1, that is, about 3 µmol kg–1 decade–1, which is small, compared with the observed decadal changes (see Figures 2, 4, and 6). Moreover, these measurement biases can be ignored in our estimates of specific water column inventories because they usually canceled out. However, the relatively large biases in sections P02 (−0.49 µmol kg–1 a–1) and P06 (−0.54 µmol kg–1 a–1) may have influenced the small-scale ∆CT and ∆nCTCAL estimates. For example, the negative bias along section P02 probably contributed to the negative ∆nCTCAL (IW) along section P02 (Figure 10b). Furthermore, correcting for the large negative bias between 2003 and 2009 along section P06 would cause the increases of anthropogenic CO2 in the intermediate layer to become larger, thus increasing the differences in ∆nCTCAL between 1992–2003 and 2003–2009.

Table 2. Section-Averaged Differences of CT and nCAOU at the Neutral Density Surface γn = 27.8 kg m–3
Line NameΔCT µmol kg-1 a-1ΔnCAOU µmol kg-1 a-1
  1. The superscripts aj correspond to those in Figures 9-11.

P01a0.21 ± 0.130.10 ± 0.04
P02b−0.49 ± 0.15−0.26 ± 0.05
P06c0.19 ± 0.140.00 ± 0.04
P06d−0.54 ± 0.24−0.08 ± 0.08
P10e0.07 ± 0.120.00 ± 0.04
P14f−0.27 ± 0.10−0.01 ± 0.03
P16g−0.01 ± 0.150.02 ± 0.11
P17h0.20 ± 0.270.01 ± 0.05
P18i−0.01 ± 0.12−0.02 ± 0.03
P21j0.14 ± 0.08−0.02 ± 0.03

[37] The data biases for ∆nCAOU were all very small (Table 2) and thus negligible.

5.2 Shortcomings of the Method Used

[38] As we have discussed so far, the accumulation of anthropogenic CO2 is the main cause of decadal changes in CT in the ocean interior. Unfortunately, conclusive quantification of the true spatial and temporal variations of anthropogenic CO2 is not possible. Therefore, several approaches have been proposed and their validity and merits assessed [Wanninkhof et al., 1999; Álvarez et al., 2009]. Brewer [1978] and Chen and Millero [1979] presented a method for estimating anthropogenic CO2 throughout the Anthropocene, which was subsequently improved by Gruber et al. [1996], who reduced the uncertainties of the method [Shiller, 1981, 1982; Broecker et al., 1985]. Sabine et al. [2004] used the method with some minor changes to estimate the global anthropogenic CO2 budget. Other variations of the method that reduce its inaccuracies have been proposed and tested [e.g., Pérez et al., 2002; Lo Monaco et al., 2005]. Our method is also based on that of Gruber et al. [1996] (see section 2), but we neglected some terms included in the original formulas and used the modified equations to detect recent decadal increases in anthropogenic CO2, rather than long-term changes. Because of its simplicity, we expect that our method is probably quite sensitive to data precision and accuracy. However, because of the high quality of the data obtained during the WOCE and WOCE revisit cruises, uncertainties due to low precision should be minimal. For example, the SD of DO measured at crossover points was about 2 µmol kg–1 [Johnson et al., 2001], and that of CT was about 3 µmol kg–1 after the correction of Sabine et al. [2005] was applied. Furthermore, as discussed in section 5.2, biases due to low accuracy cancel out in basin-scale estimates of anthropogenic CO2 storage.

[39] In our method, we assumed no significant changes in AT and constant ∆CTdiseq from the mid-1990s to the 2000s. The first assumption is probably valid at basin scale, because no such changes have been reported [Ilyina et al., 2009], and Murata et al. [2007, 2009] have shown that the second assumption is also valid.

[40] In the present study, the Redfield ratio for carbon and oxygen was used to remove the contributions of remineralization from the CT changes. However, ventilation also alters DO content, and such DO changes are not necessarily linked to the CT changes through the Redfield ratio. This decoupling of CT and DO changes is probably predominant in regions with high ventilation variability, such as the subarctic North Pacific. The negative ∆nCTCAL values observed along section P01 (Figure 2b) may be due, in part, to this decoupling.

[41] Multilinear regression analysis (MLR) [Wallance, 1995; Friis et al., 2005], as well as an improved approach called extended MLR (eMLR) [Friis et al., 2005; Sabine et al., 2008; Waters et al., 2011], are frequently used for decadal-scale surveys of anthropogenic CO2. It is difficult to compare our method with MLR and eMLR approaches, but Peng and Wanninkhof [2010] reported that increases of anthropogenic CO2 in the upper thermocline of the Atlantic Ocean estimated from salinity-normalized CT after correction for AOU on isopycnal surfaces are consistent with those obtained by both MLR and eMLR. In contrast, Waters et al. [2011] reported that on a regional scale increasing trends of anthropogenic CO2 along section P06 differed from those obtained by Murata et al. [2007], although the mean anthropogenic CO2 uptake was consistent between the two studies.

[42] Anthropogenic CO2 is transported into the ocean interior mainly along isopycnal surfaces [Sabine et al., 2004]. Our method effectively detected changes in anthropogenic CO2 due to transport by analyzing only two dissolved water properties (CT and AOU) on the density surfaces.

5.3 Decadal Changes of CT and Anthropogenic CO2 Along the Sections

[43] CO2 concentrations in the ocean are currently increasing as the oceans absorb anthropogenic CO2 from the atmosphere. If no other factors affect oceanic CO2, then concentrations of CT observed at the same location, but a decade apart, should also increase. Furthermore, the increase in anthropogenic CO2 levels (positive ∆nCTCAL) should equal the increase of CT (positive ∆CT). The following discussion examines this point by focusing on decadal changes in CT and anthropogenic CO2.

[44] Along section P01, distinctly negative values of both ∆CT and ∆nCTCAL were found in thermocline waters east of 180° (Figure 2). As stated in section 3.1, the negative values were observed where AOU was decreased (Figures 2c and S1 in the supporting information). In this region, AOU increases due to weakening of the ocean ventilation were observed in the 1990s [Emerson et al., 2004; Deutsch et al., 2006; Mecking et al., 2008], and AOU decreases were reported after 1999 [Kouketsu et al., 2010]. According to Whitney et al. [2007], observed decadal changes in AOU can be attributed to changes in the strength of ocean circulation and ventilation. As anthropogenic CO2 distributions are controlled mainly by isopycnal transport [Sabine et al., 2004], the negative ∆nCTCAL values are consistent with the decadal changes. Between 165°E and 180° along section P01 in the density range γn = 26.5–27.0 kg m–3, ∆CT was positive whereas ∆nCTCAL was negative. This discrepancy is probably due to dominant changes of CAOU, which are included in ∆CT, often masking ∆nCTCAL in subarctic regions of the North Pacific [Wakita et al., 2010]. In addition to the decadal-scale oceanic changes, both differences due to the observations being made at different times of the year and the small number of observations may have influenced ∆CT and ∆nCTCAL along section P01.

[45] Along section P14, a distinct pattern of negative and positive ∆CT bands was apparent below γn = 27.0 kg m–3 south of 20°N, whereas this pattern was indistinct in the ∆nCTCAL profile (Figure 4). The disappearance of these distinct bands is related to the cancellation of ∆CT by AOU changes (compare Figures 4a and 4c), which affect ∆nCTCAL through the term γC:O × AOU in equation (1); this result suggests the occurrence of water-mass shifts, probably due to eddies in the ocean interior below the WML.

[46] As stated in section 3.3, we attributed negative ∆CT and positive ∆nCTCAL from 125°W to 118°W along section P21 (Figure 6) to water-mass shifts. Areas of positive ∆nCTCAL greater than 20 µmol kg–1 are also clearly seen along this section; this increase exceeds that expected from increases in anthropogenic CO2 in the atmosphere; if we assume a CT increase of 1.0 µmol kg–1 a–1 in response to the recent rise in atmospheric CO2 [Murata et al., 2007, 2009], then a simple estimation of the CT increase from 1994 to 2009 would result in a maximum ∆nCTCAL value of 15 µmol kg–1. Hence, we suggest that this high CT increase reflects changes in the oceanic conditions. According to Lovenduski et al. [2008], upwelling of subsurface water rich in CO2 in the Southern Ocean and subsequent equatorward transport of the water are enhanced by a southward shift of the winds over the Southern Ocean. This series of changes leads to a decrease in the net CO2 uptake by the Southern Ocean, but it also speeds up the transport of the anthropogenic CO2 absorbed by water masses. Consistent with this, Roemmich et al. [2007] reported a speeding-up of the flow of the subtropical gyre in the South Pacific from the 1990s to the early 2000s. The large increases of anthropogenic CO2 detected along section P21 were probably produced by this process, although Lovenduski et al. [2008] reported that anthropogenic CO2 uptake is largely unaffected. We found a general tendency for uptakes of anthropogenic CO2 to be large in the subtropical South Pacific (Figure 9), as reported by Murata et al. [2007, 2010].

5.4 Short-Term Variability of CT

[47] In the present study, we focused on decadal-scale changes of CT. For this purpose, we did not use data from the WML, where short-term variability was clearly detected. However, the aliasing effect due to short-term variability may not have been removed completely, and this effect would influence detection of decadal changes of CT and natural CO2.

[48] From the repeated observations along 137°E, Takatani et al. [2012] reported no distinct seasonal changes of DO on isopycnal surfaces below the WML, but they detected significant long-term DO decreases on isopycnal surfaces after 1990. The year-to-year DO variability in their data seems to be ±3–5 µmol kg–1 (Figure 4 of Takatani et al. [2012]). Considering the Redfield ratio, we estimate that the CT variability corresponding to this short-term DO variability is, at most, about 3.5 µmol kg–1. As the changes in CT and AOU discussed here greatly exceed this value, we infer that the CT changes reflect slowly evolving decadal changes. For more precise evaluation, however, spatially and temporally denser data coverage is necessary.

5.5 Anthropogenic CO2 in IW

[49] In the South Pacific, we detected large increases of anthropogenic CO2 in the intermediate layer corresponding to AAIW south of 25°S (Figure 10), and the contribution of IW to the total specific water column inventories of ∆nCTCAL was large (compare Figures 9 and 10). Large increases of ∆nCTCAL in the density range of AAIW were also prominent along section P06. We found large specific water column inventories of ∆nCTCAL along section P21 as well, although these were confined to the western part of the section. These results imply that anthropogenic CO2 taken up in the Southern Ocean is steadily spreading northward via AAIW.

[50] Although increases of anthropogenic CO2 in IW of the North Pacific (NPIW) were smaller than those in the South Pacific, in the region north of 25°N these increases account for 10–20% of the total specific water column inventories. A large uptake of anthropogenic CO2 in the western subarctic Pacific, which is demonstrated by the high specific water column inventory values obtained for the westernmost longitudinal band along section P01 (Figure 10a), is consistent with the uptake reported by Wakita et al. [2010] (0.4 mol m–2 a–1). These changes might reflect the contribution of NPIW to the redistribution of anthropogenic CO2 in the North Pacific.

5.6 Storage of Anthropogenic and Natural CO2

[51] To assess the basin-scale storage of anthropogenic and natural CO2 in the Pacific between 1992 and 2009, we calculated the specific storage of anthropogenic and natural CO2 in four latitudinal bands from 65°N to 50°S (Table 3). To estimate the values in the four bands, we computed the area-average values at the centroids of each band by linearly fitting the specific water column inventories in each latitudinal band (Figure 9). The specific storage for the whole Pacific (Figure 1) was obtained by summing the four area averages.

Table 3. Decadal Storage of Anthropogenic and Natural CO2 by Latitudinal Band and for the Whole Pacific (in Boldface; Gray Area in Figure 1)
Areainline image (SE) PgC [10 a]-1ΔnCAOU PgC [10 a]-1
  1. Values are average ± SE.

40°N–65°N0.3 ± 0.2−0.1 ± 0.1
20°N–40°N1.5 ± 0.20.5 ± 0.1
20°S–20°N2.7 ± 0.40.3 ± 0.3
50°S–20°S3.9 ± 0.3−0.1 ± 0.3
50°S–65°N8.4 ± 0.50.6 ± 0.4

[52] The largest and smallest anthropogenic CO2 storage values were 3.9 ± 0.3 and 0.3 ± 0.2 PgC decade–1 in 50°S–20°S and 40°N–65°N, respectively, reflecting the distributions of the specific water column inventories of ∆nCTCAL (Figure 9). The total storage of anthropogenic CO2 over the whole Pacific (gray area in Figure 1) was 8.4 ± 0.5 PgC decade–1. Note that in our calculation, we did not include the regions south of 50°S because of insufficient data. However, in the high-latitude Southern Ocean, storage of anthropogenic CO2 has been estimated to be small even though the air-sea fluxes of CO2 are high [Caldeira and Duffy, 2000]. Thus, we believe that our value for the whole Pacific is not substantially underestimated in spite of the lack of data from the extreme south. The low estimate of long-term storage of anthropogenic CO2 in the ocean since the IR reported by Sabine et al. [2002b] supports this inference. The Pacific storage is larger than that estimated for the Atlantic (5 ± 1 PgC decade–1; Wanninkhof et al. [2010]) and accounts for about 40% of the estimate for the global ocean (2.2 PgC ± 0.25 PgC a–1; Fletcher et al. [2006]).

[53] The natural CO2 storage in the latitudinal bands, other than the 20°N–40°N band, was almost zero, judging from the SEs (Table 3). The total natural storage was 0.6 ± 0.4 PgC decade–1, which is substantial despite the relatively large SE. From these results, we conclude that in the whole of the Pacific, sinks and sources of natural CO2 did not change over the period from 1992 to 2009. Nevertheless, we found that very few of the natural CO2 storage values in the 40°N–65°N and 50°S–20°S bands were caused by the canceling out of negative and positive water column inventories (Figure 11). In the 40°N–65°N band, the spatial variation of the specific water column inventories was large, indicating an effect of small-scale processes on the natural CO2 storage. In the 50°S–20°S band, large changes in areas close to land were possibly associated with oceanic conditions specific to these areas.

[54] In the 20°N–40°N band, the natural CO2 storage was large: 0.5 ± 0.1 PgC decade–1. This large sink may have been caused by weakening of the ocean circulation and ventilation in the 1990s, as reported by Deutsch et al. [2006].

[55] In our approach, we assumed that both the original WOCE cruises and the revisits were carried out synoptically in each campaign period. The actual observations, however, were carried out in different years during the observation period. Although the influence of the different occupation years on the estimation might be included in the uncertainties (Table 3), we could not quantify that influence exactly. To evaluate the uncertainties more exactly, uptake rate estimations obtained by many kinds of analyses, especially ones taking account of temporal variability (e.g., based on assimilation results), should be compared.


[56] The authors thank all of the officers and crew of R/V Mirai for their exceptional support during the P01, P14, and P21 revisit cruises. The anonymous reviewers’ comments helped us to improve the manuscript. The authors express our special thanks to the physical and chemical oceanography marine technicians of Marine Works Japan, who worked on board the R/V Mirai. Our profound thanks go to all of the people engaged in national and international research programs on the global carbon cycle; without their endeavors and the resulting high-quality data obtained, we could not have completed this study.