Introduction
Over the past decade, the global ocean took up atmospheric carbon dioxide
(CO2) at a rate of about 2.5 Pg C yr-1
(1 Pg = 1015 g), roughly a quarter of all the anthropogenic
CO2 released from fossil fuel burning, cement production, and land-use
change (Le Quéré et al., 2014). The oceans' uptake of atmospheric
CO2 plays an important role in slowing down the increase of atmospheric
CO2 (Sabine et al., 2004; Takahashi et al., 2009) and hence the global
climate change. Therefore, it is important to accurately document changes of
the oceanic CO2 sink in order to accurately project future atmospheric
CO2 levels and global climate change (Takahashi and Sutherland, 2013).
The oceanic CO2 sink is mainly controlled by the gradient of the partial
pressure of CO2 (pCO2) between the atmosphere and the ocean
(ΔpCO2=pCO2water-pCO2air).
Considering that the spatiotemporal variability of atmospheric pCO2
is much smaller than that of the surface water pCO2, one could assume
that the magnitude of ΔpCO2 and hence the net air–sea CO2
flux are governed primarily by oceanic pCO2 (Takahashi and Sutherland,
2013). While atmospheric CO2 levels increased from
∼ 280 ppm
(parts per million by volume) in the preindustrial era to ∼ 393 ppm
in 2012 almost homogeneously across the globe (Tans and Keeling, 2013),
oceanic pCO2 showed different rates of change depending on local
oceanographic processes (e.g., lateral mixing, upwelling, and biological
activity) (Takahashi et al., 2006). For instance, Le Quéré et
al. (2009) found that the pCO2 increase rate in the North Atlantic was
larger than that in the atmosphere during the 1981–2007 period, while the pCO2
increase rate in the North Pacific was smaller than that in the atmosphere.
Even a decrease in sea surface pCO2 was observed in the vicinity of
the Bering and Okhotsk seas between 1970 and 2004 (Takahashi et al., 2006).
Consequently, the oceanic CO2 sink also exhibited different trends. For
example, a decrease in the sink for atmospheric CO2 was observed in the
North Atlantic subpolar gyre (50–70∘ N, 80–10∘ W) from
1982 to 1998 (Lefèvre et al., 2004), whereas a large increase of the
CO2 sink was found in the western tropical North Atlantic
(19–20∘ N, 65–68∘ W) from 2002 to 2009 (Park and
Wanninkhof, 2012). It is therefore crucial to determine changes in sea
surface pCO2 in local regions to better understand changes and
variability in the global oceanic carbon sink.
While effectively alleviating the global climate change, the oceans' uptake
of atmospheric CO2 is taking a toll on the world's oceans (Doney et al.,
2009). It causes decreasing pH, carbonate ion concentrations, and
carbonate mineral saturation states, a process commonly termed “ocean
acidification” (OA) (Caldeira and Wickett 2003; Feely et al. 2004; Orr et
al. 2005). The pH decline can induce speciation shifts of major and minor
elements in seawater, affecting their bioavailability to phytoplankton (Doney
et al., 2009). The decrease of calcium carbonate (CaCO3) saturation
state could affect the ability of marine calcifying organisms to form their
CaCO3 shells and skeletons.
The Indian Ocean, strongly influenced by seasonal monsoonal forcing (Schott
and McCreary, 2001), is a unique basin with highly variable oceanic
circulation and multi-scale air–sea interaction processes (Schott et al.,
2009). It plays an important role in the global biogeochemical cycling of
carbon and nutrients (e.g., Wiggert et al., 2009). Much effort has been
devoted to this region to understand distributions of inorganic carbon
parameters and their controlling processes as well as the region's role in
the global carbon cycle. Activities have been primarily propelled by several
national or international programs, such as the Joint Global Ocean Flux Study
(JGOFS) and the World Ocean Circulation Experiment (WOCE) (e.g., Naqvi, 2002;
Sabine et al., 2002; Bates et al. 2006).
Nevertheless, the oceanic pCO2 change and its influence on the oceanic
CO2 sink in the Indian Ocean are far from well documented when compared
to those in the Pacific and Atlantic oceans (Takahashi et al., 2009; Lenton
et al., 2012; Fay and McKinley, 2013; Sarma et al., 2013). Furthermore, many
of the existing studies in the region are based on model results (e.g., Sarma
et al., 2013; Valsala and Maksyutov, 2013), which may not well reflect the
real situation. For instance, a recent study, based on ocean biogeochemical
models, suggested a small enhancement in the CO2 sink during the
1990–2009 period in the southern Indian Ocean (Sarma et al., 2013), even though Metzl (2009)
observed that oceanic pCO2 increased at a rate of
2.11 ± 0.07 µatm yr-1 during the 1991–2007 period in the
southwestern Indian Ocean and the corresponding Antarctic sector (implying a
reduction in the oceanic CO2 sink). It is worthwhile to conduct more
studies using in situ observations in the Indian Ocean to better understand
the change of surface pCO2 and the oceanic CO2 sink in the Indian
Ocean.
Furthermore, surface water pH in the Indian Ocean is relatively low, when
compared to other oceans (Takahashi and Sutherland, 2013). This, combined
with the fact that corals are widely distributed in the region (Allen and
Adrim, 2003), makes the Indian Ocean one of the most vulnerable regions in
terms of OA. However, relatively little information on OA is available in the
Indian Ocean. It is urgent to examine the changes of pH and CaCO3
saturation state in the Indian Ocean and their potential influence on marine
organisms.
Study site (a) shows the relative location of the study
site (within the dashed frame) in the Indian Ocean, annual air–sea CO2
fluxes (mol C m-2 yr-1, colored) estimated by Takahashi et
al. (2009), and schematic representations of ocean currents during the winter
monsoon, redrawn from Schott et al. (2009). The currents shown in this map
include the South Equatorial Current (SEC), South Equatorial Countercurrent
(SECC), Northeast and Southeast Madagascar Current (NEMC and SEMC), East
African Coastal Current (EACC), Somali Current (SC), Indonesian Throughflow
(ITF), South Java Current (SJC), the Northeast Monsoon Current (NMC) during
the winter monsoon, and so on. The subsurface return flow for the supergyre
is shown in red. (b) shows the study site, the eastern equatorial
Indian Ocean (EIO, 5∘ N–5∘ S, 90–95∘ E). The
lines show all cruise tracks for surface CO2 partial pressure
(pCO2) measurements during the 1962–2012 period in the EIO. The red
triangle presents the location of the atmospheric CO2 observation
station in Bukit Kototabang, Indonesia (BKT, 0.20∘ S,
100.38∘ E).
In this paper, we report an oceanic pCO2 increase in the eastern
equatorial Indian Ocean (EIO, Fig. 1) for the first time, detected using
pCO2 data collected during May 2012, together with historical data
since 1962 as integrated by Takahashi et al. (2013) (i.e.,
LDEO_Database_V2012). We examine temporal changes in air–sea CO2 flux
and OA indicators (pH and aragonite saturation state,
Ωarag), and explore the factors responsible for the
pCO2 increase and OA.
Material and methods
Study site
The study region was located to the west of Sumatra, covering the area from
5∘ N to 5∘ S and 90 to 95∘ E (Fig. 1). The local
climate is characterized by the seasonal monsoon, which exhibits a weak westerly
annual mean wind and has a strong semiannual cycle. Consequently, the
equatorial currents are quite unique and different from those in the other
equatorial oceans (Schott et al., 2009). The ocean currents are mainly in the
zonal direction and the most distinguished are the strong surface eastward
flows, known as Wyrtki jets (Wyrtki, 1973), which occur during two
intermonsoon periods in spring (April–May) and fall (October–November).
There is no climatological equatorial upwelling due to lack of steady
equatorial easterlies. Another important current is the Equatorial
Undercurrent (Knauss and Taft, 1964), which exists at the thermocline depth
and occurs mainly during the later winter to spring. The strong equatorial
zonal currents link the EIO with the western equatorial Indian Ocean, through
which it joins the basin-scale circulation.
Summary of cruise information and mean values of surface
temperature, salinity, and surface pCO2 in the equatorial belt
(2∘ N–2∘ S, 90–95∘ E),
which are reported as mean ± standard deviation. These mean values in
this table are calculated as described in Sect. 2.2.2, and are not
deseasonalized.
Cruise name
Observation perioda
Ship/experiment
Surface
Surface
Surface
temperature
salinity
pCO2
LUSIAD_62
30 Jun, 2–3 Jul 1962
R/V Argo
29.24 ± 0.21
34.21 ± 0.06b
304 ± 2
LUSIAD_63
8–26 Apr 1963
R/V Argo
29.81 ± 0.26
33.84 ± 0.19b
307 ± 4
SAGA_II_Leg_2
30 Jun–2 Jul 1987
R/V A. Korolev SAGA II
29.08 ± 0.19
34.47 ± 0.08
357 ± 9
R. F. Weiss Surface Data Files 42–63
10–17 Feb 1995
R/V Knorr Weiss Data
29.50 ± 0.24
33.74 ± 0.13b
350 ± 2
IO95legc
31 Oct–1 Nov 1995
R/V M. Baldrige IO95
29.19 ± 0.29
34.05 ± 0.31
359 ± 5
JASMINE1999_2
4–6, 28–31 May 1999
R/V Ron Brown 1999
29.34 ± 0.22
34.38 ± 0.19
373 ± 4
I09N_Underway_pCO2
9–16 Apr 2007
CLIVAR repeat sections
29.83 ± 0.25
34.01 ± 0.05
381 ± 4
MOMSEI
1–9 May 2012
R/V Madidihang 03
30.00 ± 0.26
34.30 ± 0.08
385 ± 5
a The observation period in this table refers to the period when the
data used in this study were collected.
b We used salinity data from the simple ocean data assimilation (SODA)
since in situ data are not available during these cruises
(http://coastwatch.pfeg.noaa.gov/erddap/griddap/hawaii_d90f_20ee_c4cb.graph).
cWe did not use data from this cruise to determine the trend due to the
poor coverage latitudinally.
Data sources and processing
Data sources
Data in 2012 were from the Monsoon Onset Monitoring and its Social and
Ecosystem Impact (MOMSEI) project cruise conducted during the period
1–9 May 2012. During this cruise, sea surface pCO2 was continuously
measured every 15 min with a HydroC™ CONTROS
sensor. The CO2 mole fraction in the headspace behind a membrane
equilibrator was measured using a two-wavelength nondispersive infrared
detector (NDIR). The equilibrator consists of a flat silicone composite
membrane, and additional sensors for pressure, temperature, and relative
humidity measurements. Regular zeroings are automatically performed to
correct instrument drift with time by scrubbing CO2 from the internal
gas stream (Saderne et al., 2013). More details on pCO2 measurements
can be found in Saderne et al. (2013). Fietzek et al. (2013) presented the
detailed in situ calibration of the data, and asserted that the average
difference between sensor reading and reference pCO2 was
-0.6 ± 3.0 µatm with a root-mean-square error (RMSE) of
3.7 µatm. Before the cruise in May 2012, a comparison study between
this sensor and the Apollo CO2 instrument (e.g., Jiang et al., 2008)
indicated an accuracy of better than 5 µatm (see Fig. s1 in the
Supplement). During the MOMSEI cruise, sea surface temperature (SST) and
salinity (SSS) data were also collected every 15 min using a SBE 21 Seacat
thermosalinograph.
Sources of data used in this study.
Parameter
Source
Sea surface salinity
SODA – POP 2.2.4 monthly means with a spatial resolution of 0.5∘ × 0.5∘
(http://coastwatch.pfeg.noaa.gov/erddap/index.html)
Mixed layer deptha
IFREMER/LOS mixed layer depth climatology website with a spatial resolution of 2∘ × 2∘
(www.ifremer.fr/cerweb/deboyer/mld)
Chlorophyll a
SeaWiFS with a spatial resolution of 0.1∘ × 0.1∘
(http://las.pfeg.noaa.gov/oceanWatch/oceanwatch.php)
Sea surface pCO2b
LDEO_Database_V2012, Takahashi et al. (2013)
(http://cdiac.ornl.gov/ftp/oceans/LDEO_Database/Version_2012/)
Atmospheric CO2
CO2 monthly mean data from Mauna Loa, Hawaii, and Bukit Kototabang, Indonesia (BKT)
(http://www.esrl.noaa.gov/gmd/ccgg/trends/mlo.html)
Wind speed
NCEP (National Centers for Environmental Prediction) wind speeds with a spatial
resolution of 2.5∘ × 2.5∘ (http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.html)
a This mixed layer depth is in fact a temperature–mixed layer depth, or
isothermal layer depth. Mixed layer depths are computed as the depth with a
0.2 ∘C absolute temperature difference from 10 m temperature
(Keerthi et al., 2013).
b See specific cruise information in Table 1.
Surface water pCO2 and its associated parameters (temperature and
salinity) before 2012 were extracted from the Lamont-Doherty Earth
Observatory (LDEO) Database (version V2012; Takahashi et al., 2013)
(Tables 1 and 2). The pCO2 results of the LDEO database are based on
measurements made using air–seawater equilibration methods (Takahashi et al.,
2013). All data points have been individually quality controlled before they
were integrated into this database. The uncertainty of the pCO2 data
is estimated to be about ± 2.5 µatm on average, given
differences in equilibrator designs, calibration methods, and some
interpolated parameters (Takahashi et al., 2013).
In addition, several ancillary parameters including wind speed, atmospheric
CO2 concentration, mixed layer depth (MLD), SSS and chlorophyll a
(Chl a) (Table 2) were used. Because long-term atmospheric CO2 data
since 1962 are not available in the Indian Ocean, we used atmospheric
CO2 data measured as the mole fraction in dry air at Mauna Loa, Hawaii
(Table 2), as an alternative. Some limited observations collected during the
2004–2010 period at the BKT (Bukit Kototabang, Indonesia; Table 2)
atmospheric CO2 station, close to the study area (Fig. 1), showed that
the atmospheric CO2 level at Mauna Loa was at least 4.3 ppm larger than
that at BKT, Indonesia (see Fig. s2). Therefore, we corrected the
atmospheric CO2 concentration from Mauna Loa by subtracting 4.3 ppm.
The atmospheric CO2 mole fraction was then converted to pCO2 by
correcting to 100 % humidity at the mean SST and SSS during the
investigation period, following Jiang et al. (2008).
Spatial averaging and seasonal correction
We first grouped all the data points (temperature, salinity, and pCO2)
into their individual 0.1∘ latitudinal bands, then calculated the
average for each band, and took the average of all mean values from each band
as cruise mean. Considering that very few of the cruises covered the entire
study area well (Fig. 1), we just used the data in the equatorial belt
(2∘ N–2∘ S, 90–95∘ E) to determine the
trends of SST, SSS, and pCO2. We also corrected the effects of
the seasonal cycle by using the climatological data from Takahashi et al. (2009)
(Fig. s3) before the trend analysis.
Calculation of air–sea CO2 fluxes
We calculated the air–sea CO2 flux based on Eqs. (1) and (2):
F=0.24×k×s×pCO2water-pCO2air,k=0.262± 0.022×U102×Sc/660-0.5,
where F is the air–sea CO2 flux (mmol m-2 d-1), in which
a positive value represents CO2 releasing from the ocean to the
atmosphere; k is the gas transfer velocity (cm h-1), calculated using
Eq. (2) based on the parameterization of gas transfer velocity with wind
speed proposed by Wanninkhof (1992) and recently revised by Ho et al. (2011);
s is the solubility coefficient of CO2 (mol L-1 atm-1)
(Weiss, 1974); and pCO2water and pCO2air
are the pCO2 in the surface ocean and in the atmosphere
(µatm), respectively. In Eq. (2), U10 (m s-1) is the wind
speed at a height of 10 m above the sea surface. We used monthly mean wind
speeds from the National Centers for EnvironmentalPrediction
(NCEP) with a spatial resolution of 2.5∘×2.5∘
(http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.html) for
calculating of CO2 fluxes. Sc is the Schmidt number, which was
calculated based on the formula of Wanninkhof (1992).
Estimate of TA, DIC, pH and CaCO3 saturation
state
The conservative behavior of TA allows us to estimate TA using the salinity
data collected. We used data in the upper 20 m collected in April 2007 in this
region during the CLIVAR/CO2 (Climate Variability and Predictability)
section I9N cruise (http://cchdo.ucsd.edu) to build the relationship between
TA and SSS (Fig. 2) as follows:
TA=75.31(±5.15)×SSS-332.95(±175.35)r=0.94,n=28,p<0.0001.
This relationship produces an RMSE of 1.7 µmol kg-1, which was
less than that produced by the Lee et al. (2006) formula
(±10.6 µmol kg-1).
Relationship of total alkalinity (TA) and salinity determined
from the upper 20 m data collected in April 2007 during the CLIVAR/CO2
section I9N cruise
(http://cchdo.ucsd.edu).
The DIC (dissolved inorganic carbon), pH, and CaCO3 saturation state were calculated from pCO2,
TA, temperature, and salinity using the CO2sys program (Lewis and Wallace,
1998) and adopting the CO2 system coefficients of Mehrbach et al. (1973)
as refitted by Dickson and Millero (1987). Because the saturation state
of calcite is usually about 50 % greater than that of aragonite at
25 ∘C (Mucci, 1983), hereafter we only discuss the results for
Ωarag.
Sea surface CO2 partial pressure (pCO2) observed in
July 1962 (a), April 1963 (b), July 1987 (c), February and October 1995 (d),
May 1999 (e), April 2007 (f), and May 2012 (g) in the EIO. This figure is
plotted with ODV software (Schlitzer, 2014).
Sea surface pCO2 trend observed in various regions of the
ocean.
Region
Period
pCO2 growth
Sink/sourcea
Reference
rate (µatm yr-1)
Eastern subpolar Atlantic (32–10∘ W, 50–64∘ N)
1972–1989
2.3 (±0.8)
sink
Omar and Olsen (2006)
Subtropical North Atlantic near Bermuda
1983–2003
1.7 (±0.3)
sink
Bates (2007)
Western tropical North Atlantic
2002–2009
1.01–1.11
sink
Park and Wanninkhof (2012)
Eastern equatorial Atlantic
1982–1992b
2.5–2.8
source
Oudot et al. (1995)
Subarctic western North Pacific
1995–2003
1.6 (±1.7)
sink
Lenton et al. (2012)
Western subtropical North Pacific
1995–2005
1.8 (±0.6)
sink
Lenton et al. (2012)
Eq. Pacific (Niño 3.4 and warm pool regions included)
1990–2004
1.8–2.3
source
Takahashi et al. (2006)
East eq. Indian Ocean (90–100∘ E, 2∘ S–2∘ N, spring)
1963–2012
1.56 (±0.08)
source
this study
Southwestern Indian Ocean (30–90∘ E, 50–55∘ S, summer)
1991–2007
2.4 (±0.2)
sink
Metzl (2009)
Indian and Pacific sectors of the Southern Ocean
1995–2008
2.2 (±0.2)
sink
Lenton et al. (2012)
a “sink” means that the ocean absorbs atmospheric CO2, while
“source” indicates that the ocean releases CO2 to the atmosphere.
b The growth rate here refers to the mean difference between 1982 and
1992.
Respective contribution of temperature, salinity, TA, and DIC to
the changes of pCO2, pH and Ωarag
We used the method of Wakita et al. (2013) to quantify the contributions of
temperature, salinity, TA, and DIC to changes of pCO2, pH, and
Ωarag. For example, the contributions of these properties to
pCO2 change can be expressed as the sum of the individual
contributions as follows:
ΔpCO2=(αpCO2/αT)ΔT+(αpCO2/αS)ΔS+(αpCO2/αTA)ΔTA+(αpCO2/αDIC)ΔDIC,
where T and S are SST and SSS, respectively; and ΔT, ΔS, ΔDIC, and ΔTA denote the changes in SST, SSS, DIC, and TA,
respectively.
Temporal changes in sea surface pCO2 (black squares) and
atmospheric pCO2 (green triangles) measured at Mauna Loa, Hawaii
(a), and sea surface pCO2 along the Equator during April 1963
and May 2012 (b). The dashed line in (a) shows the linear
regression line based on the mean value of each cruise (red circle), which
has been deseasonalized using the climatological data identified by Takahashi
et al. (2009).
During the calculation, we evaluated the rate of pCO2 change by
allowing one parameter to vary while using mean values for the other
parameters. For example, we estimated the contribution of DIC change to
pCO2 ((αpCO2 / αDIC)ΔDIC) by
calculating pCO2 using the DIC trend from 1962 to 2012 and mean values
for the other parameters, and then we computed the impact of DIC change on
pCO2. The contributions of these properties to pH and
Ωarag changes were calculated similarly.
Temporal changes in the difference of pCO2 between the
atmosphere and the ocean (ΔpCO2=pCO2water-pCO2air) (a), wind speed (b), and air–sea CO2 flux
(c). All data are corrected to the same month (April).
Results and discussion
Temporal change of surface water pCO2
The in situ sea surface pCO2 data in the EIO starting from 1962 are
shown in Figs. 3 and 4. Spatially, pCO2 distributions were relatively
homogeneous in each cruise, with a standard deviation of less than
9 µatm (Fig. 3). A gradual increase in sea surface pCO2 with
time is the most evident feature (Figs. 3, 4). The mean value of sea
surface pCO2 in the equatorial belt (2∘ N–2∘ S,
90–95∘ E) increased from
∼ 307 µatm in April 1963 to ∼ 373 µatm in May 1999, ∼ 381 µatm in April 2007, and
∼ 385 µatm in May 2012 (Fig. 4, Table 1). After seasonal
correction, we find that sea surface pCO2 increased at a mean rate of
1.56 ± 0.08 µatm yr-1 from 1963 to 2012 (Fig. 4a).
For comparison purposes, we also estimated the surface pCO2 trend
along the Equator from 89.5 to 94.5∘ E, and found that
since the International Indian Ocean Expedition (IIOE, 1960–1965),
surface pCO2 increased from 307 ± 4 µatm in April 1963
to 392 ± 6 µatm in May 2012 (Fig. 4b), with a mean rate of
∼ 1.64 µatm yr-1 (after seasonal correction). This rate
of surface pCO2 increase is not significantly different from that
obtained using all data measured during the 1962–2012 period (Fig. 4a). The pCO2
increase rate in the EIO is similar to that in the subtropical North Atlantic
(Bates, 2007), higher than that in the western tropical North Atlantic (Park
and Wanninkhof, 2012), and lower than that in the equatorial Atlantic and
Pacific (Oudot et al., 1995; Takahashi et al., 2006) (Table 3).
Temporal changes in sea surface pH (in the total hydrogen
scale, pHt) (a) and aragonite saturation state
(Ωarag) (b). pHt and Ωarag were calculated from pCO2 and
TA as described in Sect. 2.4. All data are corrected to the same month
(April).
Temporal changes in sea surface temperature (a), salinity (b),
TA (c), and DIC (d). TA was
estimated from SSS, and DIC was calculated from pCO2 and TA as
described in Sect. 2.4. All data are corrected to the same month (April).
Temporal changes of air–sea CO2 flux, pH, and
Ωarag
The increase in surface pCO2 could potentially affect air–sea CO2
flux by changing the gradient of air–sea pCO2 (Eq. 1). The mean rate
of pCO2 increase in the EIO from 1962 to 2012
(∼ 1.56 µatm yr-1) was only slightly greater than that
in the atmosphere (∼ 1.46 µatm yr-1; Fig. 4a), suggesting
a weak trend of increase for the air–sea CO2 gradient (Fig. 5a). Equation (1)
showed that the air–sea CO2 flux could also be affected by the gas transfer
velocity, which is typically expressed as a power function of wind speed
(e.g., Wanninkhof, 1992). Figure 5 showed no obvious changes in wind speed.
All factors considered, air–sea CO2 fluxes showed a weak but
insignificant increasing trend during the 1962–2012 period (Fig. 5).
OA has taken place in this region during the past 50 yr, as indicated by pH
and Ωarag. Surface water pH (in the total hydrogen scale)
decreased significantly at a rate of -0.0016 ± 0.0001 yr-1 from
1962 to 2012 (Fig. 6), similar to the rate observed at the time-series
stations of Bermuda (BATS) and Hawaii (HOT) (Takahashi and Sutherland,
2013). An average decline of 0.08 pH units over the past 50 yr (1963–2012)
in the EIO is surprising, considering that the average surface ocean pH has
just declined by about 0.1 units since the 1700s due to the absorption of
anthropogenic CO2 (Raven et al., 2005). A rapid reduction in
Ωarag with a rate of -0.0095 ± 0.0005 yr-1 in
this region was also observed (Fig. 6). This reduction rate is faster than
that in the subsurface waters in the subarctic western North Pacific Ocean
(-0.004 to -0.005 yr-1) (Wakita et al., 2013), and similar to that
in the southern California Current system (0.009 ± 0.006 yr-1)
(Leinweber and Gruber, 2013) and in the North Atlantic Ocean
(-0.0100 ± 0.0012 yr-1) (Bates et al., 2012).
Contributions of temperature, salinity, TA, and DIC changes to
the increase in pCO2, and decreases in pH and
Ωarag
Using the method of Wakita et al. (2013), we quantified the contribution of
temperature, salinity, TA, and DIC (Fig. 7) to the increase in surface
pCO2 and decreases in pH and Ωarag. The results show
that the DIC increase played the most important role in elevating surface
water pCO2 and decreasing pH and Ωarag. In contrast,
the contributions from temperature, salinity, and TA were insignificant
(Fig. 8). These results (Fig. 8) are in good agreement with the trends of
temperature, salinity, TA, and DIC (Fig. 7). From 1962 to 2012, the changes
of temperature, salinity, and TA were not significant, while DIC increased
significantly during the 1962–2012 period (Fig. 7). In addition, the good consistency
between the sum of the decomposed individual contributions (Tot) and the
observed trend (Obs) verified the robustness of the method of Wakita et
al. (2013) in our study (Fig. 8).
Contribution of SST, SSS, TA and DIC to the change of
pCO2 (a), pHt (b), and aragonite saturation state, Ωarag (c). Tot denotes the sum of the decomposed individual
contributions and Obs the observed trend.
A schematic of DIC increase in the mixed layer induced by
atmospheric CO2 increase in a CO2 source region with respect to the
atmosphere. DIC changes in the mixed layer can be attributed to vertical
entrainment (ΔDICEnt), vertical and horizontal advection and
diffusion (ΔDICAdv), biological activities (ΔDICbio) and air–sea exchange (ΔDICas). From time t0
to t when atmospheric CO2 increases, the driving force of air–sea
exchange (pCO2water–pCO2air) and
correspondingly the CO2 outgassing from the ocean to the atmosphere
would decrease. This may induce DIC increase in the mixed layer due to
reduction in the magnitude of CO2 source (Schneider et al., 2012). F
denotes air–sea CO2 fluxes, k the gas transfer velocity, s the
solubility coefficient of CO2, and pCO2water and
pCO2air are the pCO2 in the surface ocean and in the
atmosphere. Details on air–sea CO2 fluxes can be found in Sect. 2.3. H
and D are mixed layer depth and seawater density, respectively.
Temporal changes in mixed layer depth (a), and
Chl a (b). Squares show the mean value, and bars show the standard
deviation. See Table 2 for more details on data sources. All data are from
the same month (April).
Factors contributing to the DIC increase
Air–sea CO2 exchange
It is obvious that the DIC increase with time was not due to local uptake of
CO2 via air–sea exchange, given that the EIO was almost always a
CO2 source to the atmosphere (Fig. 5, Bates et al., 2006; Takahashi et
al., 2009). However, the rapidly rising atmospheric CO2 since 1962
created the potential to reduce or even reverse the CO2 release in this
region from the ocean to the atmosphere (Fig. 4), and can directly induce DIC
increase in the mixed layer. Figure 9 gives a schematic of DIC increase in
the mixed layer induced by atmospheric CO2 increase in a CO2 source
region with respect to the atmosphere. When atmospheric CO2
concentration increases, CO2 outgassing in this region from the ocean to
the atmosphere would be reduced or even reversed, and would increase DIC
concentration in the mixed layer. Thus, oceanic DIC increase in a CO2
source region to the atmosphere could be caused via reduction in the
magnitude of CO2 source (Schneider et al., 2012), different from the
situation in a CO2 sink region, where more anthropogenic CO2 was
directly absorbed by the ocean from the atmosphere due to higher atmospheric
CO2 concentrations.
Ocean circulation
Transport via basin-scale ocean circulation also contributed to DIC increase.
This can be verified by the high contents of anthropogenic CO2 in the
mixed layer of the EIO (e.g., Sabine et al., 1999; Sabine et al., 2004), which
has no direct uptake of atmospheric CO2 (Fig. 5, Bates et al., 2006;
Takahashi et al., 2009). The observed increase in anthropogenic CO2 in
the EIO is likely due to accumulation of anthropogenic CO2 in CO2
sink regions and subsequent transport to the equatorial belt via basin-scale
ocean circulation. For instance, carbon in the region between 15 and
50∘ S in the Indian Ocean could be finally transported to the EIO.
On the one hand, this region (between 15 and 50∘ S) is a major
subduction zone (Schott et al., 2009), serves as a significant sink of
atmospheric CO2 (Fig. 1a, Takahashi et al., 2009; Valsala et al., 2012),
and hosts the largest inventories of anthropogenic CO2 across the Indian
Ocean (Sabine et al., 2004). Furthermore, it is reported that the oceanic
increase in carbon storage roughly kept pace with atmospheric CO2
increase (Sabine et al., 1999).
On the other hand, the upper ocean horizontal circulation and the meridional
overturning cells are believed to account for carbon transport towards the
EIO (Fig. 1; Schott et al., 2002; Schott et al., 2009; Valsala et al., 2012).
In the upper layer, water masses move westward in the South Equatorial
Current (SEC) and partly merge into the East African Coast Current (EACC);
they can move further to the EIO along with the eastward flows, including the
South Equatorial Countercurrent (SECC), the Wyrtki jets (Wyrtki, 1973), and
the Equatorial Undercurrent (EUC) (Knauss and Taft, 1964). At the thermocline
depth, shallow overturning cells, including the cross-equatorial cell (CEC)
and the Southern Hemisphere subtropical cell (STC), bring water masses from
the subduction zone to the off-equatorial upwelling zone, where they upwell
to the surface. Thus, the southern subtropical Indian Ocean may act as a
window for carbon uptake (Valsala et al., 2012), and the basin-scale
circulation provides the route to transport the absorbed anthropogenic
CO2 ultimately to the EIO. A similar mechanism was also proposed in the
equatorial Pacific by Feely et al. (1999), who pointed out that entrained
subtropical water was injected into upwelled water at the Equator.
Nevertheless, the pathways of ocean circulation are very complicated, and
there must be other ways to increase DIC in the equatorial Indian Ocean. For
instance, the Red Sea–Persian Gulf Intermediate Waters formed in the
northwestern Indian Ocean carry anthropogenic CO2 signals and spread
equatorward (Sabine et al., 2004; Alvarez et al., 2009, and references
therein), which also contributes to the increase of equatorial waters' DIC.
Overall, ocean circulation may play an important role in transporting carbon
accumulated in the CO2 sink region to the equatorial belt on a basin
scale. To a large extent, this could account for the paradox that the
increase in anthropogenic CO2 occurred in the CO2 source region,
where CO2 was emitted to the atmosphere from the ocean.
Vertical mixing and biological activity
Vertical mixing with deep waters rich in CO2 can elevate surface DIC
content. For instance, the enhanced vertical mixing, usually accompanied by
salinity rise and MLD deepening, will bring more CO2-rich waters to the
surface layer, leading to higher DIC levels (e.g., Takahashi et al., 2006;
Dumousseaud et al., 2010). However, there was no significant trend for SSS
and MLD during the study period (Figs. 7, 10), indicating an insignificant
influence of vertical mixing on DIC changes.
Biological activity could also affect DIC (e.g., Zhang et al., 2010). Model
studies indicate that there were no significant changes in net primary
production, particle export and export efficiency from 1960 to 2006 in this
region (Laufkötter et al., 2013). Satellite data also do not show
a significant trend for Chl a (a proxy for biological activity) from 1998
to 2012 (Fig. 10). Therefore, biological activity was not the main factor
leading to DIC rise, either.