Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 The changes in anthropogenic CO2 are evaluated in the South Pacific, along the meridional line P18 (110°W) and the zonal line P06 (32°S), using the extended multiple linear regression (eMLR) method. The structure of the column inventory of anthropogenic CO2 on P18 is similar to the southern section of P16 in the central South Pacific (150°W), but the overall increase is greater by approximately 5–10 μmol kg−1. The value of the anthropogenic CO2 inventory on P18 is in agreement at the crossover point of an earlier evaluation of P06. Subsequent changes in pH due to the increase in anthropogenic CO2 are also evaluated. The change in pH is determined from the changes in anthropogenic CO2 and do not reflect variability in other decadal signals. For both cruise tracks, the average annual change in pH is −0.0016 mol kg−1 yr−1. This value is in good agreement with the average decrease in pH in the North Pacific, at the Hawaii Times Series and the subtropical North Atlantic. The uptake rates of anthropogenic CO2 are within reasonable agreement with similar studies in the South Pacific. There is evidence for greater uptake of anthropogenic CO2 in the western South Pacific and is attributed to the formation of subtropical Mode Water in the region.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The average rates of anthropogenic CO2 intrusion and acidification over the previous two centuries are relatively well constrained, but our ability to understand how ocean chemistry will change under further anthropogenic stress is limited. Model and geochemical estimates indicate the storage of anthropogenic CO2 in the ocean is expected to decrease due to a decreased carbonate buffer capacity and climate change [Wetzel et al., 2005; Egleston et al., 2010; Le Quéré et al., 2010]. However, others have suggested that the anthropogenic CO2 uptake and storage in the North Atlantic might be less limited by changes in the carbonate system, i.e., a decrease in buffer capacity, than previously thought [Tanhua et al., 2008]. The Climate Variability and Predictability (CLIVAR)/CO2 Repeat Hydrography Program was launched to investigate the response of the atmosphere–ocean interactions to anthropogenic forcing on decadal time scales by reoccupying sections of the WOCE/JGOFS Global CO2 Ocean Survey. This paper evaluates the changes in the anthropogenic carbon storage, as well as the decreases in oceanic pH, on two cruises in the South Pacific and compares the findings with similar studies in the region.
2. Data Quality
 The changes in ocean carbon chemistry from the increase in anthropogenic CO2 are examined on WHP sections P18 (∼110°W) and P06 (∼32°S) (Figure 1). Prior to the CLIVAR expedition in 2007/2008, P18 was last occupied during the WOCE expedition in 1994. The P06 line has been occupied by three separate hydrographic programs: WOCE (1992), Blue Earth Global Expedition (BEAGLE 2003) and CLIVAR (2009/2010). All bottle data from these programs is available from the Carbon Dioxide Information Analysis Center (http://cdiac.esd.ornl.gov/oceans/). At least 2 carbon parameters were measured on all cruises, with TCO2 measured by coulometric titration on every cruise [Dickson et al., 2007; Johnson et al., 1987]. Total alkalinity (TA) was measured by potentiometric titration [Millero et al., 1993; Dickson et al., 2007], and pH by spectrophotometry using the methods outlined by [Clayton and Byrne, 1993] on all cruises except for WOCE P06. Fugacity of CO2 (fCO2) was measured on WOCE P18, P06 and CLIVAR P18. fCO2was measured by non-dispersive infrared analysis [Dickson et al., 2007; Wanninkhof and Thoning, 1993] for WOCE and CLIVAR P18. On WOCE P06, fCO2 was measured by gas chromatography [Neill et al., 1997].
 The accuracy of the WOCE/JGOFS data, including P18 and P06, was evaluated by the Global Ocean Data Analysis Project (GLODAP) [Sabine et al., 2005]. The BEAGLE cruise was examined together with the original WOCE/JGOFS cruises as part of the CARINA data synthesis [Sabine et al., 2010]. The P18 and P06 CLIVAR cruises were also compared to the GLODAP and CARINA data sets using the same tools but were not released in time to be included in the publication. No adjustments were recommended for any of the cruises used here. The overall uncertainty in the TCO2 and TA data is estimated to be ± 3 μmol kg−1 and ± 5 μmol kg−1, respectively [Lamb et al., 2001]. Murata et al.  also evaluated the consistency of the WOCE and BEAGLE P06 TCO2 on 41.4σ3 and TA on 27.2σθ, and found no significant variability between the cruises.
3. Data Analysis
 To make direct comparisons of the difference in the anthropogenic carbon storage, it is necessary to remove any natural mesoscale variability between the cruises. The multiple linear regression (MLR) approach [Wallace, 1995] determines the short-term increase in anthropogenic CO2 between two measured cruises, by fitting the TCO2as a function of multiple physical and biological parameters. This method has the advantage of being independent of many assumptions required for the back-calculation techniques [Sabine et al., 2008] and any differences between CFC and CO2 solubility and transport inherent in the CFC proxy approaches [Álvarez and Gourcuff, 2010]. The MLR is an empirical method based on a comparison of partial correlations with physical and biological parameters to remove the natural variability in TCO2 [Wallace, 1995]. The method is not dependent on a Redfield ratio to correct for changes in biological production/respiration and, assuming the correct parameters are applied in the regression, it inherently corrects for water mass mixing and changes in circulation. The independent variables used in MLR calculations are not universal and are selected based on data availability, quality and statistical assessment of the fit [Friis et al., 2005]. The independent variables that provide the most robust fit do not appear to be equivalent for all regions of the oceans. In the North Atlantic, Friis et al.  found apparent oxygen utilization (AOU) is not necessary in the determination of anthropogenic CO2, whereas [Sabine et al., 2008] found AOU is a necessary parameter in the Pacific Ocean.
 This method assumes the MLR completely accounts for the natural variation and any differences between the two fits are from anthropogenic forcing. The MLR method is highly dependent on the ability of the partial correlations to parameterize the natural system, and Friis et al.  recommended a modified version of the MLR, the ‘extended’ multiple linear regression (eMLR), to reduce error propagation and variability. In the eMLR approach both cruises are independently fit with respect to physical and biological parameters, and the anthropogenic change is determined by taking the difference in the coefficients of the independent parameters. By applying the differences in the coefficients, any random variability and propagated error in an independent parameter is partially canceled [Wanninkhof et al., 2010]. There are three major sources of uncertainty associated with this method. First, there is the assumption that a multiple parameter regression can fully constrain the measured parameter. As the standard errors of the regressions are comparable to the overall measurement uncertainties, we assume this uncertainty is similar to that of the measurement. Second, any biases in the measured parameters between cruises will increase the overall uncertainty. This uncertainty is minimized by careful evaluation of data quality and is previously addressed. Last, the method is dependent on the assumption that the differences in the regression coefficients are only related to the change in the anthropogenic carbon storage. The fact that there are no anthropogenic CO2 changes in the deep waters, where anthropogenic influences should be negligible, indicates this uncertainty is less than the uncertainty introduced by the regression. The regression is therefore the largest source of uncertainty and the detection limit is defined as the maximum standard error in the regressions.
 In this study, the eMLR technique is applied to determine the increase in anthropogenic CO2 below the seasonal mixed layer (150 m) on both cruise tracks. On P18 the increase in anthropogenic CO2 is examined between 1994 and 2008. The increase in anthropogenic CO2 on P06 is examined from 1992 to 2010, 1992 to 2003, and 2003 to 2010. The lower limit of data used for the fits is 1500 m as there is no significant increase in anthropogenic CO2 (less than 5 μmol kg−1) found below 1200 m in the South Pacific. The measured TCO2 is fit to a linear function of physical and biological hydrographic data, and an Ftest was used to ensure the model is robust. A t-test is used to evaluate the significance of each parameter in the model. The validity of the model is further evaluated by randomly selecting half the measured data and fitting it to the regression model. The data, not used in evaluation of the regression, is then used to calculate the model residuals. In all cases, the full data fit standard error is within ± 0.15 of the model verification standard error. To ensure the model is robust for all regions and water masses, the residuals are plotted as a function of longitude/latitude and depth (Figure 2). After iterating this routine with different combinations of hydrographic data, the measured TCO2 is found to be best fit to a linear function of salinity, potential temperature, potential density, oxygen, silicate and phosphate. Where the eMLRCO2 is calculated using:
Where the subscript 2 indicates the data and coefficients are from the most recent occupation, and the 1 represents coefficients derived from a prior occupation. The coefficients and standard errors for the fits are given in Table 1. The standard errors for all the regressions are below 5 μmol kg−1, and only increases of anthropogenic CO2 above this value are regarded as significant. With the exception of oxygen, these parameters are the same as those used by Sabine et al.  on P16. Because of the substantial changes in North Pacific dissolved oxygen concentrations, Sabine et al.  independently fit AOU with an eMLR to provide an estimate of the change in apparent decomposition rate for organic matter in the water column. Separating the AOU term out allowed them to independently evaluate the changes in carbon resulting from atmospheric uptake and from changes in circulation driven apparent decomposition rates. Since AOU changes were very small in the South Pacific, this separation was not necessary. Sabine et al.  found that including AOU in the TCO2 fit gave very similar results to fitting AOU separately then subtracting the AOU eMLR from the TCO2 eMLR.
Table 1. Model Coefficients for P18 and P06 eMLR Regressionsa
The root square error (RSE) of the regression is also provided.
 To ensure that none of the variation in TCO2 is due to the dissolution or precipitation of carbonate minerals, the salinity normalized TA is compared on isopycnal surfaces. There are no significant TA changes between cruises and no need to correct the TCO2 for carbonate precipitation or dissolution. These results are consistent with observations by Murata et al. .
 Due to the differences in seasonal processes both physical and biological, the eMLR is not used to predict the carbon system above 150 m. Applying the eMLR to the full water column results in highly scattered residuals in the mixed layer and increases the standard error of the regression. In order to avoid the complications of calculating anthropogenic CO2 in the mixed layer, Sabine et al.  estimated the anthropogenic CO2 in the surface waters of line P16 using atmospheric pCO2 with TCO2 and TA measurements from the mixed layer. This method is based upon observations that indicate the pCO2 of Pacific surface waters are increasing at rates approximately equal to the atmosphere [Feely et al., 2006; Takahashi et al., 2006], and the observed consistency of the TA data between cruises. This same approach is applied for the upper 150 m of P06 and P18. Data is taken from the upper 50 m from CLIVAR P18/P06 and the CO2 fugacity (fCO2) is calculated using the CO2SYS program [van Heuven et al., 2009]. The calculated fCO2 is then corrected for changes in atmospheric CO2 between the cruises using the global atmospheric annual mean CO2. The TCO2 is then calculated using the TA and fCO2 values, and the surface anthropogenic CO2 is derived from the difference in TCO2 between cruises.
 When at least 2 of the carbon parameters are known, it is possible to calculate the remaining unknown parameter(s) using the CO2 acid dissociation constants [Millero, 2007, and references therein]. The change in pH due to anthropogenic forcing is calculated using the CO2SYS program [van Heuven et al., 2009], with the carbonate dissociation constants of Millero et al.  The pH is calculated, at 25°C, using inputs of TCO2, TA, salinity, silica, and phosphate. Where the nutrients are used to remove the minor effect they have on the TA. The anthropogenic CO2 is then subtracted from the TCO2, and the pH is recalculated providing an estimate of the pH without anthropogenic influences. Anthropogenic acidification is then assumed to be equal to the difference in the calculated pH values.
 The increase in anthropogenic CO2 on P06 (2003–2009/10) is within ∼2–3 μmol kg−1 of the eMLR detection limit, and is restricted to the upper 400 m. This low storage is consistent with the anticipated average annual increase in anthropogenic CO2. Due to the minor increase in measurable anthropogenic CO2 between these sections, further analysis was not carried out and the results are not presented. The maximum increase in anthropogenic CO2 on P06 (1992–2003) and P06 (1992–2010) occurs west of 170°W in the upper 550 m (Figures 3a and 3b). Slightly lower increases are evident in the upper 200 m of P06 (1992–2010) between 120 and 90°W. The largest increase on P18 occurs between 15 and 40°S, the region where the subtropical gyre is found (Figure 3c). On all lines, the deepest infiltration of anthropogenic CO2 occurs at the maximum penetration (∼1000 m) of Antarctic Intermediate Water (AAIW). On P18, equatorial upwelling, north of 5°S, suppresses the storage of anthropogenic CO2 below 300 m, and the shoaling of anthropogenic CO2 on the eastern boundary of P06 is consistent with coastal upwelling. These trends are consistent with isopycnal surfaces. Table 2 gives an overview of the total and average annual anthropogenic CO2 storage rates for the mode/central water masses and intermediate water masses.
Table 2. Total (μmol kg−1) and Annual Mean Uptakes (μmol kg−1 yr−1) of Anthropogenic CO2 for the Central and Mode and Intermediate Water Masses
Mode and Central Water
Annual Mean Uptake
Annual Mean Uptake
16.9 ± 3.2
0.94 ± 0.2
9.2 ± 2.1
0.51 ± 0.1
10.2 ± 2.6
0.93 ± 0.2
6.3 ± 1.9
0.57 ± 0.1
12.4 ± 4.1
0.89 ± 0.4
9 ± 2.6
0.64 ± 0.2
 The P06 and P18 lines cross at 32°S and 110°W, and the anthropogenic CO2 profiles can be directly compared at this point. Figure 4 shows the P18 and P06 crossover profile for the average annual increase in anthropogenic CO2. The annual average increases are plotted so the data is normalized to a consistent time scale and direct comparisons can be made. The discrepancies in the profiles are within the estimated standard error of the calculations.
 The intrusion of anthropogenic CO2 has led to an overall decrease in the pH in the surface and intermediate water on P06 and P18 (Figures 5a and 5b). For the upper 250 m the annual average pH change is −0.0018 yr−1 on P06 and −0.0014 on P18. This change in pH is directly calculated using the TCO2 and anthropogenic CO2 concentrations, and therefore has been corrected for any natural variations in physical and biological factors that would affect the pH. The decrease in pH is thus directly attributed to increases in anthropogenic CO2. The maximum penetration of the calculated decrease in pH coincides with the maximum penetration of measurable increases in anthropogenic CO2 in AAIW, and the maximum changes in pH correspond to the greatest increases of anthropogenic CO2. As the pH change is calculated using the anthropogenic CO2 signal, it is expected that the maximum changes coincide. Attempts are made to verify the calculated change in pH reflects the differences in measured pH, which with the exception of WOCE P06 was directly determined on each cruise. However, it is difficult to derive the pH change, from anthropogenic forcing, through direct comparison of the measured data due to the natural background variability is similar to the calculated pH change.
Murata et al.  presented anthropogenic CO2 increases on P06, from 1994 to 2003, using the isopycnal technique [Peng et al., 1998]. Here we present a new evaluation of the 1992–2010 anthropogenic CO2 and a comparison of the 1992–2003 anthropogenic CO2 using the eMLR technique. The methods of this study differ from those of Murata et al.  in both the calculation technique and the region over which we apply our calculation. Murata et al.  applied the isopycnal calculation over the entire water column, whereas the eMLR is only applied over the region below the seasonal mixed layer and the mixed layer anthropogenic CO2 is derived from atmospheric pCO2. Due to large seasonal TCO2 variations in the mixed layer, calculations of anthropogenic CO2 in the mixed layer using the isopycnal technique are likely influenced by seasonal processes [Murata et al., 2007], where as the method applied here should not be as highly influenced by the seasonality of the mixed layer. Due to the discrepancies in the treatment of the mixed layer, a direct comparison of the column inventories is a poor evaluation of the consistency between the studies. Consistency is instead evaluated by comparing the average annual anthropogenic CO2 increase in the mode and central water masses (MW) and intermediate (IW) water masses between the studies. Between the studies there are significant discrepancies in the location of the anthropogenic CO2 storage, but when the concentrations of anthropogenic CO2 in the mode and intermediate water masses are averaged over the length of the P06 line and compared they are in good agreement (Table 3). Murata et al.  observed higher concentrations of anthropogenic CO2 in the MW east of 160°W, where as this study finds higher concentrations of MW anthropogenic CO2 west of 160°W. The high concentrations of anthropogenic CO2 observed by Murata et al.  west of 160°W are greater than the anticipated increases based on changes in atmospheric pCO2 and are compensated by a decrease in AOU from 1992 to 2003. When the data are corrected for observed changes in AOU the observed increases in anthropogenic CO2 are more consistent with the anticipated increase calculated from the increase in atmospheric CO2 [Murata et al., 2007] and our results.
Table 3. Comparison of Mean Anthropogenic CO2 Uptake (μmol kg−1) for P06 (1992–2003)
 There are no previous published studies of decadal changes on the P18 line, but Sabine et al.  have conducted a decadal study on the P16 line (150°W). The structure and magnitude of annual change in the anthropogenic CO2 column inventory for the southern region of P16 is similar to that of P18. The maximum increase in anthropogenic CO2 is found within the subtropical gyre on each of the lines and the concentrations of anthropogenic CO2 decrease in the equatorial region and south of approximately 55°S. Spatial differences in the storage of anthropogenic CO2 on P18 and P16 [Sabine et al., 2008] are due to differences in equatorial upwelling and variability in the locations of convergence zones.
Table 4 provides a comparison of annual average anthropogenic CO2 uptake rates for the South Pacific from this and other studies. There is an discrepancy in the average inventories for meridional cruises extending south of 50°S (P16 of Sabine et al. ; P18 of this study) and cruises occurring north of 50°S (P14 and P15 of Peng et al. ; P06 of Murata et al. ; P06 of this study). The low inventories of anthropogenic CO2 poleward of 55°S on P16 and P18 appear to be diluting the average annual uptakes from ∼ 0.9 mol m−2 yr−1 to ∼ 0.4 mol m−2 yr−1. However, calculations indicate the low anthropogenic CO2 inventories south of 50°S are not small enough to decrease the average uptake by the observed amount. The remaining difference in the uptake rates is instead attributed to the formation of mode water in the western South Pacific. Murata et al.  noted higher anthropogenic CO2in the thermocline of eastern South Pacific basin, and attribute it to differences in Sub-Antarctic Mode Water (SAMW). A SAMW salinity maximum occurs west of 160°W, and a salinity minimum occurs off the southern tip of South America [Piola and Georgi, 1982]. However, our results indicate there is no significant difference (greater than 5 μmol kg−1) in SAMW uptake of anthropogenic CO2 for either the eastern or western region of the South Pacific, and the higher inventory of anthropogenic CO2 in the western basin is due to the subduction of subtropical mode water (STMW) (Figure 6). STMW shows a greater increase in anthropogenic CO2 because the Revelle factor of STMW is lower than that of the SAMW. Where the Revelle factor is a buffer factor related to the change in TCO2 for a given change in pCO2 and is inversely related to the anthropogenic CO2 uptake capacity of a water mass [Sabine et al., 2004]. The large increase in anthropogenic CO2 in the subtropical gyre, evident in P06 (1992–2010) and P18 (1994–2008), is due to subduction within the gyre leading to the formation of subtropical underwater (STUW).
Table 4. Comparisons of Annual Average Anthropogenic CO2 Uptake Rates in the South Pacific
Uptake Rate (mol m−2 yr−1)
Sabine et al.  used the eMLR method to fit both TCO2 and AOU. The TCO2eMLR was corrected for changes in AOU using Redfield ratios to calculate a change in TCO2 from the AOUeMLR. The results from this approach are qualitatively similar to using AOU as a regressor.
 We attempt to resolve the discrepancies between our results and those of Murata et al.  by using the transit time distributions (TTD) method for anthropogenic CO2. A TTD can be used to calculate the concentration of a passive tracer using the time-dependent surface history of the tracer, and has been used to give estimates of oceanic anthropogenic CO2 that are largely independent of measured carbon [Hall et al., 2002; Waugh et al., 2004, 2006; Tanhua et al., 2008; Khatiwala et al., 2009]. The TTD is determined on CLIVAR P06, using the methods of Waugh et al. , from in-situ measurements of CFC-12. The atmospheric history of CFC-12 [Bullister, 2011], a constant surface saturation [Waugh et al., 2004] of 95% for CFC-12 [Hartin et al., 2011], and the measured salinity and temperature are used to create the time-dependent CFC-12 surface history. In order to apply the calculated TTD to determine the total anthropogenic CO2 it is necessary to derive the surface history of anthropogenic CO2. The surface history of anthropogenic CO2 is determined using similar methods to those outlined for the determination of anthropogenic CO2 in the mixed layer of the eMLR. As the TTD determines the total anthropogenic CO2, and not the temporal increase, the anthropogenic CO2 is calculated for the WOCE and CLIVAR cruises and the difference between the two cruises is temporal increase in anthropogenic CO2 using the TTD method. The overall average annual uptake calculated from this method is ∼0.79 mol m−2 yr−1. This value is consistent with the results of both this study and Murata et al. . However, when the average uptakes are calculated for western (∼ 0.72 mol m−2 yr−1) and eastern (∼0.93 mol m−2 yr−1) regions of the South Pacific, the results from the TTD are consistent with the eMLR method but not the isopycnal method (Table 4).
 The anthropogenic pH change on the P06 and P18 lines shows pH decreasing over all regions, with the smallest increases south of 55°S and in the coastal and equatorial upwelling regions. The average annual change on P06 (1992–2003 and 1992–2010) and P18 (1994–2008) is −0.0016 ± 0.0005 yr−1, and is consistent with the expected rate of change of −0.0018 yr−1 found when calculating the change in pH as a function of increase in atmospheric pCO2 for this period (1992–2010) using the CO2SYS program with a constant TA. This value is also consistent with the decreased pH reported in the mixed layer of the North Pacific [Byrne et al., 2010], and from time series annual averages in the surface waters off of Hawaii of (−0.0019 ± 0.0003 yr−1) and the subtropical North Atlantic (−0.0015 yr−1) [González-Dávila et al., 2007]. As it appears acidification rates are closely coupled with atmospheric pCO2 growth rates over the Pacific, it is probable that acidification will become an even greater problem as atmospheric pCO2 growth rates continue to rise.
 The majority of our present understanding of the effects of ocean acidification come from laboratory and short-term field studies and it is difficult to evaluate all the possible impacts on decadal time scales. However, it is evident that carbonate mineral saturation states are decreasing in response to acidification and biogeochemical cycles are likely impacted [Doney et al., 2009]. Many calcifying organisms, such as coral, are likely to be adversely affected by decreases in carbonate saturation [Gattuso et al., 1998; Kleypas et al., 1999; Langdon and Atkinson, 2005; Langdon et al., 2000], and calcification in the Great Barrier Reef of Australia may have already been negatively impacted by ocean acidification [Wei et al., 2009]. The speciation of trace metals and inorganic nutrients will also be affected [Millero et al., 2009; Doney et al., 2009]. Pascal et al.  have found increases in free Cu concentrations and antagonistic toxicity in coastal copepods when the pH of seawater is decreased. Increases free Cu could have dramatic impacts on Pacific fisheries by increasing toxicity in both low and high trophic level species. Even though many of the potential effects of ocean acidification are not well constrained, these results demonstrate the pH of surface waters in the South Pacific is declining and organisms within these systems are likely to be impacted.
 The results presented are estimates of the minimum decadal changes in anthropogenic CO2 and pH on P06 and P18 in the South Pacific. There are significant regional differences in the uptake of anthropogenic CO2 related to water mass subduction and upwelling. Our results indicate the regions of STMW and STUW formation represent the greatest increase in anthropogenic CO2, but SAMW and AAIW also show significant decadal increases. It appears the average annual increase in anthropogenic CO2 in the western South Pacific is greater than in the eastern region due in part to the subduction of STMW. It is also evident the pH of surface waters are decreasing at rates equivalent to the increases in atmospheric CO2, but many of the potential impacts this will have on marine organisms are not well constrained on decadal time scales. These results indicate the importance of the repeat hydrography programs in the understanding of the decadal and spatial variability in the anthropogenic carbon system.
 The authors wish to acknowledge the support of the National Science Foundation Chemical Oceanography section and the National Oceanic and Atmospheric Administration Office of Climate Observations for supporting our CO2 studies. We would also like to thank Samar Khatiwala for making the code available for calculation of the TTD and providing assistance with the calculations. This is PMEL publication number 3709.