Variability in Southern Hemisphere Ocean Circulation from the …klinck/Reprints/PDF/katsumataJPO2013.pdf · 2013. 9. 23. · Variability in Southern Hemisphere Ocean Circulation

Variability in Southern Hemisphere Ocean Circulation from the 1980s to the 2000s

K. KATSUMATA AND S. MASUDA

RIGC JAMSTEC, Yokosuka, Japan

(Manuscript received 18 October 2012, in final form 9 May 2013)

ABSTRACT

Interannual-to-decadal variability of ocean circulation in the Southern Hemisphere was examined using

data from the 1980s to the 2000s in a box inversemodel to estimate transport across hydrographic sections and

three ocean general circulation models (OGCMs). The westerly wind stress over the OGCMSouthernOcean

showed a steady increase of 5%–8%decade21. The meridional overturning circulation was quantified by the

transport across 308S. The OGCMs suggested a slight strengthening [from 0.26 1.0 to 0.8 6 1.3 Sv decade21

(1 Sv[ 106m3 s21)] of the uppermeridional cell (Deacon cell) and twoOGCMs showed a weakening (20.860.6 and 21.0 6 0.3 Sv decade21) of the lower meridional [Antarctic Bottom Water (AABW)] cell, partly

explained by contraction of the AABW volume. The box inverse estimates did not contradict these two

findings. For Antarctic Circumpolar Current transport, quantified by zonal transport across four key sections,

the box inverse model estimated a decrease of 5–21 Sv. Decomposition of the decrease into baroclinic

transport by the Subantarctic and Polar Fronts, barotropic transport, and others shows that the decrease is

mostly due to barotropic transport and transport carried by the flow north of the Subantarctic Front and south

of the Polar Front. In the OGCMs, the variability of transport across key sections is often correlated with

transport carried by a flow south of the Polar Front and with the southern annular mode index. In all models,

then, the transport of the Antarctic Circumpolar Current, defined as the transport carried by the fronts, has

not decreased significantly over the study period.

1. Introduction

The Southern Hemisphere ocean circulation is char-

acterized by the eastward-flowing Antarctic Circumpolar

Current (ACC) and the meridional overturning circula-

tion (MOC). The MOC appears different depending on

whether it is zonally averaged on surfaces of constant

density or on surfaces of constant pressure (D€o€os and

Webb 1994), but south of 408S, both types of averaging

show two cells. The upper cell consists of the near-surface

northward-flowing branch and the mid-depth southward-

flowing water, while the bottomwater flows northward to

form the lower cell. The upper cell is often referred to as

the Deacon cell (e.g., Speer et al. 2000). In this paper, we

call the lower cell the Antarctic BottomWater (AABW)

cell.

Under the influence of the strong westerly wind over

the Southern Ocean, the deep waters in the Atlantic,

Indian, and Pacific Oceans outcrop in the Southern

Ocean, making subsurface water masses susceptible to

changes in air–sea fluxes. Indeed, significant changes in

subsurface water masses in the Southern Hemisphere

oceans have been observed in recent decades, which in-

clude freshening (Aoki et al. 2005; Rintoul 2007; B€oning

et al. 2008; Durack and Wijffels 2010), warming of mid-

depth waters (700–1100m) (Gille 2002, 2008; B€oning

et al. 2008), and warming of deep waters (.3000m)

(Purkey and Johnson 2010; Kouketsu et al. 2011). Some

but not all of these changes can be explained by the

southward shift of the ACC axis (Cai et al. 2010; Meijers

et al. 2011). The southward shift of the ACC (Sokolov

and Rintoul 2009a), in turn, is likely associated with the

southward shift and intensification of the westerlies as

quantified by the increasing southern annular mode

(SAM) index (Marshall 2003; http://www.nerc-bas.ac.

uk/icd/gjma/sam.html) (Fig. 1). A least squares fit be-

tween 1988 and 2008 shows an increase in the zonal wind

stress at a rate of 5%–8%decade21.

a. Antarctic Circumpolar Current transport

The subsurface hydrographic changes can lead to

changes in the meridional slope of the isopycnals, which

Corresponding author address: K. Katsumata, RIGC JAMSTEC,

2–15 Natsushima, Yokosuka, 2370061 Japan.

E-mail: [email protected]

SEPTEMBER 2013 KAT SUMATA AND MASUDA 1981

DOI: 10.1175/JPO-D-12-0209.1

� 2013 American Meteorological Society

result in changes in the baroclinic transport of the ACC.

However, hydrographic observations have detected no

significant decadal changes in the meridional isopycnal

tilts and concurrent changes inACC transport.Cunningham

et al. (2003) and Renault et al. (2011) reported no signif-

icant changes in the baroclinic transport across the Drake

Passage from 1975 to 2000 and from 1975 to 2006, re-

spectively. Similarly, no significant changes in baroclinic

transport have been reported south of Africa between

08 and 308E (Legeais et al. 2005; Swart et al. 2008) or

across 1408E (Rintoul and Sokolov 2001; Rintoul et al.

2002). Steady baroclinic ACC transport has also been

inferred in the circumpolar average of hydrographic

data along the dynamic height contours from the 1960s

to the 2000s (B€oning et al. 2008) as well as in satellite

altimeter data from 1991 to 2000 (Sokolov and Rintoul

2009b).

An explanation for this limited sensitivity of the baro-

clinic ACC transport to increased wind stress (Fig. 1) is

that mesoscale eddies counteract the increase in meridi-

onal isopycnal tilts caused by anomalous Ekman trans-

port. This ‘‘eddy saturation’’ (Straub 1993) was found

in eddy-permitting ocean-only models (Hallberg and

Gnanadesikan 2006; Yang et al. 2007) and in an eddy-

permitting ocean–atmosphere coupled model (Farneti

et al. 2010) as well as in a satellite observation (Meredith

et al. 2004).

Changes in the total transport are more difficult to

observe because it is necessary to know the depth-

independent barotropic component of the transport,

which does not show up in hydrographic observations

and has a much shorter time scale (days) than the baro-

clinic component (years). Direct velocity measurements

using mooring (Whitworth and Peterson 1985) or low-

ered acoustic Doppler current profilers (Renault et al.

2011) have been used to estimate total transport. The

different time scales of baroclinic and barotropic re-

sponses in ACC transport mean that short-period fluc-

tuations in the bottom pressure or sea surface height

anomaly can be associated with the barotropic compo-

nent of transport fluctuations (Olbers and Lettmann

2007). Satellite gravity observations can detect these

short-term bottom pressure fluctuations (Zlotnicki et al.

2007; Bergmann and Dobslaw 2012). In gravity records

from 2003 to 2005, Zlotnicki et al. (2007) found a de-

creasing trend in bottom pressure records averaged

along the southern edge of the ACC.

At decadal time scales, the ACC transport in coupled

models shows both increasing and decreasing trends

(Wang et al. 2011) although all atmospheric models

FIG. 1. (top) SAM index (Marshall 2003) and (bottom) annual-mean zonal wind stress ap-

plied to three OGCMs averaged between 508 and 608S. Note that both OFES and K7 models

use the same National Centers for Environmental Prediction (NCEP)–National Center for

Atmospheric Research (NCAR) reanalysis product but the K7 model adjusts the wind stress

field and other fluxes to minimize the model error (details in section 2b).

1982 JOURNAL OF PHYS ICAL OCEANOGRAPHY VOLUME 43

show an increase in the zonal westerly jet in the mid-

latitude Southern Hemisphere. This suggests a signifi-

cant role for driving mechanisms other than wind stress,

such as buoyancy (Hogg 2010), mesoscale eddies (Straub

1993), and eddy parameterization in themodel (Kuhlbrodt

et al. 2012). It is interesting that two recent efforts to re-

alistically simulate these effects by using eddy-permitting

resolutions (1/38 in latitude and longitude) in 100 model

year runs (Graham et al. 2012) and by using a four-

dimensional variational (4D-VAR) data-assimilation

technique (Wang et al. 2010) both show a decrease in the

ACC total transport through the Drake Passage under

increasing zonal wind stress. Wang et al. (2011) and

Graham et al. (2012) explained this decrease as a result

of narrowing of the ACC.

Although no significant trends have been detected on

decadal time scales, there are some studies suggesting

a relationship between the wind and the ACC transport

at interannual time scales. Meredith et al. (2004) found

a significant correlation between the SAM index and the

bottom pressure record from approximately 1000-m

depth during the 1980s and 1990s on the south side of the

Drake Passage, a proxy for ACC transport at subseasonal

time scales. In ocean general circulationmodels (OGCMs),

Yang et al. (2007) and Treguier et al. (2010) also found a

statistically significant correlation between ACC trans-

port and the SAM index at interannual time scales.

b. Meridional overturning circulation transport

The relationship between zonal wind changes and

MOC is not intuitively clear. Lacking observations, stud-

ies of the variability of the MOC have used numerical

simulations. Using a coarse-resolution coupled model,

Hall and Visbeck (2002) explained oceanic responses

to the westerly wind increase as anomalous northward

Ekman transport, which leads to increased meridional

isopycnal tilts (i.e., increased ACC transport) and in-

creased MOC south of 408S. In simulated responses of

the Southern Ocean to a southward shift of the subpolar

westerly jet (Oke and England 2004), a 5.48 latitudinalshift over a 100-yr simulation caused an insignificant

change to the Deacon cell strength and a slight (21.1 Sv;

1 Sv [ 106 m3 s21) decrease in the AABW cell. The

reason for the AABW decrease was not clear. In a

model with 1/68 resolution (Hallberg and Gnanadesikan

2006), a 20% increase in wind stress produced about

a 20% increase in the Deacon and AABW cells, al-

though the ACC transport showed only a 3%–5% in-

crease (eddy saturation). An increase in MOC strength

in response to an enhanced westerly jet was also found

in an eddy-permitting ocean–atmosphere coupled model

(Farneti et al. 2010) as well as in an eddy-permitting

ocean-only model (Yang et al. 2007). These simulated

increases in MOC with enhanced westerlies can be ex-

plained by enhanced isopycnal eddy diffusivity in re-

sponse to eddy kinetic energy increase (Meredith et al.

2011; Abernathey et al. 2011). Under the assumption that

increased eddy kinetic energy also enhances the dia-

pycnal diffusivity, an increase in the AABW cell can

similarly be expected (Ito and Marshall 2008; Saenko

et al. 2011).

The paucity of observational data showing the vari-

ability of the Southern Hemisphere MOC reflects the

difficulty of observing the Southern Hemisphere oceans

in full-depth and land-to-land coverage. Nevertheless,

with the internationally coordinated efforts of ship-based

hydrographic programs, it is now possible to discuss the

difference in the Southern Hemisphere circulation at

decadal intervals. Given the internal variability of the

ocean (Wunsch 2008), however, it is imperative that we

combine the observed results with simulation estimates.

In this paper, we report the results of one such effort. The

OGCMs used in this work, as well as the box inverse

model used to combine the hydrographic observations,

are described in section 2. The results are separately shown

anddiscussed for themeridional transports in section 3 and

for the zonal transports in section 4.

2. Models

a. Box inverse model

Eight hydrographic sections across the Southern Hemi-

sphere oceans have been occupied at least twice; once in

the 1980s and 1990s as a part of the World Ocean Cir-

culation Experiment (WOCE) (Fig. 2), which we will

call theWOCEHydrographic Program (WHP), and once

in the 2000s as part of the International Ocean Carbon

Coordination Project and the Climate Variability and

Predictability Program (Fig. 3), which wewill call Revisit.

The details of the two occupations are summarized in

Table 1.

The hydrographic sections and the continental land-

masses define the horizontal extent of the boxes, while

the neutral density surfaces (Jackett and McDougall

1997), the sea surface, and ocean bottom define the

vertical extent. Following Sloyan andRintoul (2001), we

labeled five water masses according to their approxi-

mate neutral density gn; thermocline water (TW) for

gn , 26.0, intermediate water (IW) for 26.0, gn , 27.4,

Upper Circumpolar Deep Water (UCDW) for 27.4 ,gn , 28.0, Lower Circumpolar DeepWater (LCDW) for

28.0, gn, 28.2, and BottomWater (BW) for 28.2, gn.

Our calculations used the measured temperature and

salinity to derive the geostrophic velocity across the

sections with assumed zero-velocity surfaces. The


velocity corrections at these initial zero-velocity sur-

faces and the diapycnal fluxes were then estimated so as

to conserve the volume, temperature, and salinity in

each box. The results are shown in Figs. 2 and 3. Details

of the inverse box model are described in the appendix.

b. Ocean general circulation models

As the eddy saturation hypothesis suggests, it is de-

sirable that a simulation model of the Southern Ocean

has an eddy-resolving grid (Hallberg and Gnanadesikan

2006; Farneti et al. 2010). At the same time, biases and

drifts in the model need to be kept minimal, which re-

quires a long, preferablymillennial, run time.Amillennial

run of an eddy-resolving model is still beyond modern

computers’ capability and compromises have to be found.

One way is to parameterize eddies and use coarser grids

for longer run time. The coupled models in phase 3 of the

Coupled Model Intercomparison Project employed this

approach. The ACC in these models have been studied

by Wang et al. (2011). Another is to assimilate observed

data to minimize biases and drifts. Indeed, an eddy-

resolving assimilation model of the Southern Ocean

provides a realistic description of the Southern Ocean

circulation (Mazloff et al. 2010), but the 4D-VAR

method, which preserves perfect mass and momentum

balances, is computationally expensive and decadal runs

are still difficult at eddy-resolving resolutions. We use

three OGCMs to represent three different approaches to

these requirements. The OGCM for the Earth Simulator

(OFES) has an ‘‘eddy resolving’’ grid but is not con-

strained by observed data. In an attempt tominimize initial

transients, the model had been spun up for 50 years by

climatological forcing (Masumoto et al. 2004) before

a hindcast run from 1950 was started (Sasaki et al. 2008).

The Simple Ocean Data Assimilation (SODA) model

(Carton and Giese 2008; Carton et al. 2012) and the K7

model (Masuda et al. 2010) are constrained by observed

data. The SODA model has an ‘‘eddy permitting’’ reso-

lution but uses less computationally expensive method of

data assimilation than the 4D-VARmethod at the cost of

errors in mass and momentum balance. The K7 model

is a 4D-VAR model but does not resolve eddies. The

details of these model implementations are summarized in

Table 2.

FIG. 2. Transports estimated by box inverse model applied to WHP data between 1987 and 1995. Black numbers beside the hydro-

graphic sections show transports across the hydrographic sections (Sv) with the uncertainty estimated by the box inverse model. The

transport across the zonal (meridional) hydrographic lines were integrated every 58 (28) in lon (lat) and shown by blue (positive) and red

(negative) patches. Positive is northward and eastward. The blue arrows show the diapycnal transport at the top of the box and the red

arrows show those at the bottom of the box. Positive is upwelling. The green arrows show subduction from and upwelling into the surface

mixed layer box. Positive is upwelling. The bottom topography is contoured at different depths roughly corresponding to the watermass in

the subtropical gyre.


c. Comparison between observation and OGCMs

TheOGCM results and the observations are compared

across ameridional section at 308E inFig. 4.All datawere

linearly interpolated onto the coarsest model grid (K7).

All OGCMs captured the approximate distribution of

water masses, with salty surface water around gn 5 26.0

from the Indian Ocean, the Antarctic Intermediate Water

salinity minimum along gn 5 27.4, the high-salinity sig-

nature of North Atlantic DeepWater at 27.4, gn, 28.0,

and Antarctic Bottom Water 28.0 , gn. Reflecting its

low-resolution grid, the K7 model showed the smoothest

water mass distribution, which led to the least-inclined

isopycnal surfaces. In the OGCMs, salty North Atlantic

Deep Water (NADW) extended more broadly than sug-

gested by the observations, even in the OFES model that

had the finest grid, suggesting that the OGCMs over-

estimated the diapycnal diffusion of salinity. The OGCMs

successfully reproduced the Antarctic IntermediateWater

(AAIW) salinity minimum intrusion along gn 5 27.4,

FIG. 3. As in Fig. 2, but for the Revisit data between 2003 and 2009.

TABLE 1. Hydrographic sections. Unless otherwise noted, data were downloaded from the Clivar and Carbon Hydrographic Data Office

(CCHDO) website (http://cchdo.ucsd.edu/). Parentheses around S03 indicate that only a small amount of data was used from this source.

Location WHP cruise Source Revisit cruise Source

P06 32.58S May–Jul 1992 Tsimplis et al. (1998) Aug–Oct 2003 Katsumata and Fukasawa (2011)

I3/4 208S Apr–Jun 1995 CCHDOa Dec 2003–Jan 2004 Katsumata and Fukasawa (2011)

I5 348S Nov–Dec 1987 Toole and Warren (1993) Mar–May 2009 CCHDOa

A10 308S Dec 1992–Jan 1993 Siedler et al. (1996) Nov–Dec 2003 Katsumata and Fukasawa (2011)

I6S 308E Feb–Mar 1996 CCHDOa Feb–Mar 2008 CCHDOa

I9Sb 1158E Jan 1995 McCartney and Donohue (2007) Dec 2004–Jan 2005 CCHDOa

(S03) — Mar 1995 Rintoul and Sokolov (2001) — —

SR1c 688W Jan 1990 Roether et al. (1993) Feb 2009 McDonagh (2009)

SR3 1408E Jan 1994 Rintoul and Sokolov (2001) Mar–Apr 2008 CCHDOa

aWe were unable to locate references other than the cruise reports (http://cchdo.ucsd.edu).b To fill the data gap on the Antarctic Shelf for the WHP cruise, stations 7–10 from the WHP S03 cruise (expo code 09AR9404) were

added.cRevisit data are available on request (http://www.bodc.ac.uk/).


except in the K7 model, where the salinity minimum

barely extended northward below the high-salinity sur-

facewater at 408S.As a result, in theK7model the salinity

around gn 5 27.4 was higher than in the observations or

the other OGCMs (Fig. 4d0), leading to a shallow bias of

the isopycnal. For the K7 model, we therefore used gn 527.6 as the boundary between thermocline water and

intermediate water. In the K7 model, overmixing is also

indicated by the low-salinity bias in the high-salinity wa-

ter from the Indian Ocean around gn 5 26.0 (Fig. 4d0).For the SODA and OFES models (Figs. 4b0, c0), thelargest differences were found around 398S above 2000m,

whose structure suggest that the models did not resolve

the observed eddy (Fig. 4a). The near-surface salinity in

OFES south of 508S was too high, where the data-as-

similated SODA and K7 model performed better, sug-

gesting errors in the surface freshwater fluxes.

Model drift is examined against one of few observed

changes in the Southern Ocean—contraction of AABW

(Purkey and Johnson 2012). Because all OGCMs em-

ployed in this study do not explicitly have a sea ice

component, the AABW production processes are

mimicked in the models either by nudging to monthly

climatological salinity at the surface and the southern

end of the model domain or by assimilating to observed

data (see Table 2). All OGCMs showed decreases in the

AABW volume (Fig. 5) although the OFES model un-

derestimated and the K7 model overestimated the

trend. The agreement does not mean that these models

are free from drifts and biases. Indeed, it was found that

the ACC fronts in the K7 model showed an unexplain-

able northward drift (see Figs. 13–16, described in

greater detail below). The agreement, however, shows

that the parameterized diabatic processes (diapycnal

mixing, AABW production, etc.) in the models work

reasonably well and that the relaxation processes are at

least in the right direction. We therefore regard trends

reproduced in these models as likely to have occurred in

the real oceans.

3. Meridional overturning circulation

Most of the WOCE hydrographic data used in Fig. 2

overlap with the data used by Ganachaud and Wunsch

(2000) and Sloyan and Rintoul (2001). Our results for

meridional transport are mostly consistent with these

previous works within the uncertainty. This section dis-

cusses the MOCs, as quantified by the transport across

the hydrographic sections along approximately 308S.

a. South Pacific Ocean: Section P6

The box inverse model estimated a 12 Sv increase

in the total northward transport across the South

Pacific section P6 (Fig. 6, Total), although the differ-

ence was not statistically significant with overlapping

uncertainties. The OGCMs did not show this increase.

There were nonnegligible differences amongOGCMs as

well as between the OGCMs and the inverse model. A

particularly large one involves the WHP Upper Cir-

cumpolar Deep Water, where the OGCMs estimated

southward transport between 0 and 25Sv (consistent

with theRevisit box inverse estimate), whereas the inverse

model estimated it at220Sv. Comparison of the transport

and isopycnals of two occupations (Fig. 7) shows that the

gn5 28.0 contour for the 1992 (WHP) data had a steep

slope along the western side of the Tonga–Kermadec

TABLE 2. OGCM characteristics where 20CRv2 5 Twentieth-Century Reanalysis, version 2; KPP 5 K-profile parameterization;

SSHA 5 sea surface height anomaly; SSS 5 sea surface salinity; WOA 5 World Ocean Atlas; and T 5 relaxation time. Version 2.2.4 of

SODA is used (SODA 2.2.4).

SODA 2.2.4 K7 OFES

Zonal grid size 0.48* 18 0.18Meridional 0.258* 18 0.18Levels 40 46 54

Period 1890–2008 1957–2009 1950–2010

Vertical mixing KPP Fickian and Noh mixed layer KPP

Lateral mixing Biharmonic Gent–McWilliams Biharmonic

Wind 20CRv2 (Compo et al. 2011) Optimized NCEP–NCAR NCEP–NCAR

Buoyancy 20CRv2 (Compo et al. 2011) Optimized NCEP–NCAR NCEP–NCAR bulk

SSS relaxation T 5 3 months (WOA 2001) None T 5 6 days (WOA 1998)

South boundary

relaxation

None (J. Carton 2011,

personal communication)

708–758S; T 5 30 days month21

(WOA 1998)

728–758S; T 5 1–720 days

month21 (WOA 1998)

Assimilation Incremental Analysis Update 4D-VAR (Adjoint) None

Data Temperature, salinity, and SST Temperature, salinity, SST, and SSHA None

Reference Carton and Giese (2008),

Carton et al. (2012)

Masuda et al. (2010) Sasaki et al. (2008)

*Output is mapped onto a 0.58 3 0.58 grid.


FIG. 4. Salinity (color) and density (contour) along 308E for (a) observed data (practical salinity),

(b) SODA model, (c) OFES model, and (d) K7 model. (b0),(c0),(d0) Difference of model outputs from the

observed salinity is shown. The top color bar applies to the salinity in (a),(b),(c),(d), and the bottom color bar

applies to the differences in (b0),(c0),(d0). The hydrographic data were collected from 21 Feb to 21Mar 1996.

TheOGCMoutputs were averaged forMarch 1996. The neutral density contours are at gn5 26.0, 27.4, 28.0,

and 28.2. For model K7 in (d) and (d0), gn 5 27.6 is shown with a dashed contour.


trench, which explains the southward jet of about 10 Sv

near 1828E (green circle on Fig. 7). There were also

differences in the barotropic correction at the assumed

zero-velocity surface with the 1992 barotropic flow

having a more negative (southward) component. Be-

cause the narrow jet along the Tonga–Kermadec trench

was neither seen in the OGCMs (not shown) nor in the

Revisit observations, it is plausible that this jet was a

transient feature.

It was possible to reduce the negative bias in theWHP

Upper Circumpolar DeepWater by prescribing a greater

covariance to the barotropic adjustments across this

section, but this would add a positive bias to theAntarctic

Intermediate Water transport and the total transport,

that is, the Indonesian Throughflow (see section 3b).

Given the reasonable agreement of the total transport of

the current solution with the OGCMs on Fig. 6, we did

not adopt the adjusted solution.

The box inverse model also estimated a significant

change in the thermocline water, which the OGCMs did

not duplicate. This is probably a result of the box inverse

model’s trying to adjust the drastic increase in the Upper

Circumpolar Deep Water in the least-constrained layers.

In summary, the OGCMs did not show a significant

change in the MOC transport across the South Pacific

section. The increases shown by the box inverse model in

theUpper CircumpolarDeepWater and the thermocline

water are probably due to a transient jet along the deep

western boundary (Fig. 7) and not a decadal trend.

b. South Indian Ocean: Section I5

In all density ranges, the transport estimates for 1987

(WHP) and 2009 (Revisit) given by the box inverse

model did not show significant differences beyond the

overlap of the uncertainties (Fig. 8). The OGCM results

also did not show significant trends (discussed in more

detail in section 3d).

Because transport through the Bering Strait con-

necting the Pacific and Atlantic Oceans is small (,1 Sv;

Roach et al. 1995), the total transport through I5 or P6

is almost equal to the Indonesian Throughflow. A re-

cent estimate based on 3-yr mooring observations of

the transport is 15 Sv (Gordon et al. 2010), which sug-

gests that the K7 and OFES models underestimated and

the SODA model overestimated the transport.

c. South Atlantic Ocean: Section A10

As was the case in the South Indian Ocean, the trans-

port across the SouthAtlantic ocean estimated by the box

inverse model for 1992/93 (WHP) and 2003 (Revisit) did

not show a significant difference (Fig. 9).

d. Discussion

Linear trends were tested in the monthly model out-

put by fitting a straight line to the time series with co-

variances prescribed by the temporal variance assuming

that each monthly transport is statistically independent.

We regarded an estimated trend as significant when

a zero trend was not included within one standard un-

certainty.We regarded a trend as robust when the trends

estimated by all three OGCMs were significant. With

this criterion, robust trends were found in 1) positive/

negative trends in the total transport across P6/I5

(reflecting the Indonesian Throughflow transport), re-

spectively, 2) a positive trend in UCDW across P6, and

3) a negative trend in Lower Circumpolar Deep Water

across I5. Lee et al. (2010) studied the interannual var-

iability of the Indonesian Throughflow in 14 ocean data

assimilation products and found that the models showed

reasonable agreement, particularly after the 1990s, and

the most consistent signal was an increase in transport

from 1993 to 2000 associated with the strengthening of

the tropical Pacific trade winds. This increase in the

Indonesian Throughflow appears in the total panel in

Figs. 6 and 8, which explains the trend 1). The trends 2)

and 3) might be associated with the response of the low-

latitude Pacific and Indian Oceans to this trade winds

anomaly (Wijffels and Meyers 2004).

As described in the introduction, some simulations

have shown an increase in theMOC strength in response

to the increased westerly jets (Fig. 1). We examined the

MOC strength at about 308S by adding the transports

across the Atlantic (A10), Indian (I5), and Pacific (P6)

Oceans (Fig. 10). The northward-flowing branch of the

Deacon cell is represented by the sum of thermocline

and intermediate waters, and the southward-flowing

FIG. 5. Volume of AABW (gn . 28.2) in the OGCMs south

of 308S. The solid lines show the AABW volume in the OFES

model (blue), SODA model (green), and K7 model (red) with

dashed lines indicating trends. The shaded triangle indicates the

observed decrease of AABW (potential temperature u , 08C) of28.2 (6 2.6) Sv (Purkey and Johnson 2012).


FIG. 6. Transport between Australia and South America across 328E. The lines show the monthly OGCM trans-

ports, smoothed by a Hanning low-pass filter of 25-month width from models SODA (green), OFES (blue), and K7

(red). Std dev of the monthly transports is shown by error bars. The thick black horizontal lines show the transport

estimated by box inverse models (Figs. 2 and 3) for the WHP (from November 1987 to October 1998) and Revisit

(from August 2003 to July 2009) periods with uncertainty indicated by the gray boxes.


branch is the sum of Upper and Lower Circumpolar

Deep Waters.

As found in Yang et al. (2007), the Deacon cell in the

SODA model showed an increasing trend (Fig. 10a).

The positive trend in the upper branch of the Deacon

cell was also found in K7 and OFES outputs, but these

trends were not statistically significant given thatmonthly

outputs are statistically independent (with estimated

trends for the SODA, K7, and OFESmodels being 0.861.3, 0.6 6 0.9, and 0.2 6 1.0 Sv decade21, respectively).

The southward-flowing mid-depth branch (UCDW 1LCDW, Fig. 10b) did not show a robust trend with only

theOFES trend (0.76 0.5 Sv decade21) being significant.

As shown in Fig. 7, the box inversemodel estimates of the

UCDW transport suffered from the transient jet.

Interestingly, two of the three OGCMs showed signif-

icant negative trends for the AABW branch (Fig. 10c)

with a large decrease from 1990 to 1992 found in all three

models followed by a trend in theOFESmodel and amore

pronounced negative trend in the K7 model (20.8 60.6 Sv decade21 for OFES, 20.0 6 0.7 Sv decade21 for

SODA, and 21.0 6 0.3 Sv decade21 for K7). The box

inverse results do not resolve the decreasing trends,

but the uncertainties estimated by the box inverse

model were greater than the decrease in the OGCM

transports, which means the estimated trends were not

inconsistent with the observation. The negative trend

is opposite to a theoretical prediction that AABW

export will increase under enhanced mixing resulting

from the wind increase (Ito and Marshall 2008; Saenko

et al. 2011). There are three possible reasons for the

discrepancy; first, the use of monthly average velocity

and monthly average thickness in the OGCM outputs

mightmiss eddy transport owing to the velocity-thickness

correlation [however, an eddy-permitting simulation

showed that the eddy-driven component of the MOC is

small at 308S (Saenko et al. 2011)]; second, a monotonic

relationship in which increased wind stress leads to

increased mixing might not hold; and third, another

mechanism not considered in the theoretical models such

as time dependence, easterlyAntarctic winds (Stewart and

Thompson 2012), or global constraints (Nikurashin and

Vallis 2011; Shakespeare and Hogg 2012) might not be

negligible.

Indeed, the time dependence of the background strati-

fication not considered in the theories offers a partial ex-

planation of the transport reduction. The reduction in

transport can be caused by reduction of the area that

AABWoccupies along 308S or by reduction of the velocityof the deep northward current. Figure 11a shows that the

FIG. 7. Comparison of UCDW transport between 1992 (WHP) and 2003 (Revisit) occupa-

tions. (top) Cumulative transport of UCDW, where the transport is integrated westward from

the easternmost station. (bottom) The contours of the density 27.4, gn, 28.0 with an interval

of 0.1. The green circle shows a transient jet found only in the 1992 profile.


AABW area across 308S had a negative trend but the rate

was much less than that of the transport reduction in Fig.

10c (e.g., for OFES, the area would disappear in about 60

decades whereas the transport would disappear in about

12 decades). TheAABW reduction is therefore due to the

combined effect of area reduction and velocity reduction.

The reason for the area reduction, in turn, might be

related to the reduction in AABW volume. Purkey and

FIG. 8. As in Fig. 6, but for the transport between Africa and Australia across 32.58E.


Johnson (2012) comparedWOCEandRevisit hydrographic

data and found an 8.2 6 2.6 Sv reduction in the volume

of water below u 5 08C. This reduction was partly repro-

ducedby theK7andOFESmodels (Fig. 11b). TheAABW

volume change is the difference between production

by winter convection and removal by northward export

(by the deep currents) and upward export (by diabatic

mixing). If the northward export is declining (Fig. 10c),

FIG. 9. As in Fig. 6, but for the transport between South America and Africa across 308E.


the volume reduction therefore is a result of reduced

production and/or enhanced upward export by diabatic

mixing. Because both the K7 and OFES models use

relaxation boundary conditions at the southern bound-

ary of the model region (Table 2), they cannot accu-

rately estimate AABW production. The strength of

winter convection can be calculated in the models by the

volume of water that flows downward across the bottom

of themixed layer. For simplicity, we fixed the bottom of

the mixed layer at z 5 300m, and estimated the down-

ward volume flux as the horizontal convergence of water

of gn . 28.0 (Fig. 11c). The K7 model showed a slight

increase in AABW production, whereas the OFES

model did not show a clear trend. The reduction in the

AABW volume (Fig. 11a) is thus a result of reduced

production of new AABW at the southern relaxation

boundary and/or enhanced upward export of diabatic

mixing. With the present configuration of the OGCMs,

it is difficult to separate the contributions of these two

factors.

4. Antarctic Circumpolar Current

When examining the ACC transport, we could have

decomposed the transport into water masses defined by

density, but we noted that the observed variability in the

ACC is characterized by movement of water masses in

the meridional rather than vertical direction (e.g.,

Sokolov and Rintoul 2009a; Meijers et al. 2011). Con-

sidering that most of the zonal transport is concentrated

along fronts, with two major fronts being the Sub-

antarctic Front and the Polar Front (e.g., Rintoul and

Sokolov 2001; Swart et al. 2008; Renault et al. 2011), we

divided the sections meridionally into three components:

FIG. 10. MOC through approximately 308S for (a) TW1 IW, (b) UCDW1 LCDW, and (c) BW. The lines show the monthly OGCM

transports, smoothed by a Hanning low-pass filter of 25-month width from models SODA (green), OFES (blue), and K7 (red). Other

details are the same as in Fig. 6.


south, north, and central (Fig. 12). The central component

includes the two major fronts. Sokolov and Rintoul

(2009a) established that certain dynamic height con-

tours follow these fronts. Using the altimetric sea sur-

face height anomaly added on top of the climatological

dynamic height above 2500 dbar, Sokolov and Rintoul

(2009a) identified 12 fronts across the ACC (their Fig. 3).

For the hydrographic data, we used the dynamic height

anomaly calculated with reference to 3000dbar and de-

fined the central part as the region between the 1.11- and

2.40-m dynamic height contours. We also constructed

the sea surface height by using the same climatological

and altimetric data used by Sokolov and Rintoul (2009a).

Our definition of the central part corresponds to the re-

gion between the dynamic height labels of 0.98 and 2.17m

from Sokolov and Rintoul (2009a), which were located

between the northern branch of the southern ACC Front

(N-SACCF) and the southern branch of the Polar Front

(S-PF) and between the subantarctic zone/subtropical

zone (SAZ/STZ) and the northern branch of the Sub-

antarctic Front (N-SAF), respectively. The sea surface

height is available in the OGCM outputs and we defined

FIG. 11. AABWbudget in the K7 and OFESmodels. Time series of (a) the area that AABW

(gn . 28.2) occupied across the 308S latitudinal circle and (b) the volume of AABW south of

308S (as in Fig. 5). (c) AABW produced by surface cooling estimated as water convergence at

depths shallower than z5 300m within the area with surface density gn . 28.0. Negative value

means production of AABW.


the central part using the sea surface heights: 20.41 and

21.42m for OFES (but 20.25 and 21.42m for 308E),20.16 and 21.41m for SODA (but 20.15 and 21.41m

for 308E), and20.36 and21.13m for K7 (but20.55 and

21.00m for 308E).Above a depth of 3000m, the central part is further

separated into the barotropic (vertically uniform) and

baroclinic components. The velocity at 3000m defines

the barotropic component, so that the baroclinic veloc-

ity is zero at 3000m. The transport below 3000m is

called the bottom component. The transport carried by

the region north and south of the central part is called

the north and south component, respectively (Fig. 12).

The corresponding decomposition that we applied to the

box inverse model is summarized in Table 3.

For the box inverse model, the uncertainty introduced

by this decomposition is difficult to estimate, and we

used the uncertainty in the thermocline water box of

5 Sv as a nominal measure of the uncertainty. For the

OGCM results, we plotted the standard deviation of the

monthly output as a measure of the uncertainty.

a. Transport across 1158 and 1428E

The box inverse model estimated a 6-Sv reduction

across 1158E, which is not statistically significant. Indeed,

the OGCMs showed no significant changes in total

transport (Fig. 13a). The decomposition shows that all

components were steady over the period 1986–2008. We

note that the variance was much less for the baroclinic

component (Fig. 13c) than for the total transport, and

the largest variance was in the north component.

The K7 output showed a positive drift in the south

component and a negative drift in the north compo-

nent. This was caused by a northward drift of the fronts

(about 1.58 lat decade21). The cause of the drift is not

known.

Because of the geometric constraints, the transport

across 1428E (Fig. 14) behaved similarly to the transport

across 1158E; a steady baroclinic component, a particu-

larly variable north component, and northward shift of

the fronts in K7. The decomposition of the box inverse

model, however, showed slightly different contributions

(Table 3). Across 1158E, the 6-Sv reduction of the total

transport was a result of 4-Sv increase of the north com-

ponent, and a 9-Sv decrease of the central (5 barotropic1baroclinic1 bottom) component. Across 1428E, the northcomponent increased by about 22 Sv, which was offset

by a 11-Sv decrease of the south component and a 16-Sv

decrease in the central component (Table 3). Across

1158E, the fronts are almost zonal and parallel to themajor

FIG. 12. (top) Dynamic height with reference to 3000 dbar and zonal velocity across 1428E(WHP SR3) observed in April 2008. Using the dynamic height, the section is horizontally

divided into north, south, and center regions; the center region is divided into barotropic (zonal

velocity at 3000 dbar extended to depth between 0 and 3000 dbar), baroclinic (residual of

barotropic above 3000 dbar), and bottom (below 3000 dbar) regions.


topographic ridge (Southeast Indian Ridge), whereas

across 1428E, the fronts meander to cross the Southeast

Indian Ridge (Sokolov and Rintoul 2009a). Although

the deviations of the front positions as observed by the

altimetry are not very different between 1158 and 1428E[Figs. 9 and 10 of Sall�ee et al. (2008)], the fronts are

locally parallel to the hydrographic section across 1428Esuch that the large variability was generated.

b. Transport across 308E

The box inverse model estimated a large (19 Sv) re-

duction of the total transport between the WOCE and

Revisit occupations (Fig. 15a). This was a result of a

17-Sv reduction in the north component, a 7-Sv reduc-

tion in the south component, and a 5-Sv increase in the

central part (Table 3). Again, the baroclinic component

was steady in the OGCMs and the box inverse model.

Because the north part of this section is dominated by

the Agulhas current and its return flow, making it one of

the most variable regions in the Southern Ocean [e.g.,

Fig. 10 of Chelton et al. (2011)], we expected that the

variability from the north part would be the largest here.

The north component has a complicated flow struc-

ture andwe note that all OGCMs failed to reproduce the

strong westward flow (i.e., southward deepening iso-

pycnals) present in the observation (around 408S in Fig. 4).As a result, the north transports of the OGCMs showed

a large positive (eastward) bias (Fig. 15f). This positive

bias was compensated by a negative bias in the baro-

tropic and baroclinic components. The negative bias in

the baroclinic component suggests weaker stratification.

Given that the baroclinic components in OGCMs were

always smaller than the box inverse estimates in other

sections (Figs. 13c and 14c) this negative bias is probably

caused by the use of monthly average output in the

model and excessive diffusion in the OGCMs (or lack of

spatial resolution).

c. Transport across the Drake Passage (68 8W)

In theDrake Passage, the Subantarctic Front is next to

the northernmost hydrographic station and conse-

quently the north region does not exist.

The transport across the Drake Passage is probably

the best constrained of all longitudes [e.g., Table 2 of

Renault et al. (2011)]. The transport (mean plus or mi-

nus one standard deviation) estimated by the OGCMs

were 113 6 6 Sv (K7), 142 6 5 Sv (OFES), 154 6 6 Sv

(SODA). The low bias of the K7 model is probably due

to its excessive diapycnal mixing (Fig. 4). The other

values are slightly larger than previous estimates listed

in Renault et al. (2011), but they are compatible with

recent estimates using an eddy-resolving data-assimilated

estimate of 153 6 5 Sv (Mazloff et al. 2010).

The OGCMs did not show a decreasing trend in the

total transport that can explain the decrease estimated

by the box inverse model (Fig. 16a). The decomposition

of the box inverse model shows that this decrease is

largely explained by the reduction in the barotropic

transport (Fig. 16d). The baroclinic transport was rather

steady except in the K7 model. The reduction in the baro-

clinic transport in K7 was compensated by an increase in

the south component. This was a result of northward

migration of the southern boundary of the central part

(about 28 lat decade21, not shown).

d. Discussion

The box inverse model estimated a decrease in the

total zonal transport from the WHP period (1987–95)

TABLE 3. Zonal transport change in the box inverse model. From left to right, the columns are for lon, date, tot, south, north, barotropic,

baroclinic, bottom, lat between south and center regions, and lat between north and center regions.

Lon Date Tot* (Sv) S (Sv) N (Sv) BT (Sv) BC (Sv) B (Sv)

Lat

(S) (N)

308E Mar 1996 168 36 267 64 121 14 54.18 41.08Feb 2008 149 29 284 64 127 13 55.98 40.58

1158E Jan 1995 183 44 26 19 125 1 59.98 45.78Jan 2005 177 43 22 15 122 21 60.18 44.58

1428E Jan 1994 181 44 218 28 121 6 60.28 50.48Apr 2008 176 33 4 22 112 4 61.98 50.78

688W Jan 1990 169 24 0 44 94 7 62.38 —**

Feb 2009 151 21 5 33 91 5 62.48 —**

* The differences from Figs. 2 and 3 are due to different methods of integrating the transport.

** The northern part does not exist across this section.


FIG. 13. Transport between Australia and Antarctica across 1158E for (a) total, (b) bottom, (c) baroclinic,

(d) barotropic, (e) south, and (f) north. The lines show the monthly OGCM transports, smoothed by a Hanning

low-pass filter of a 25-month width, from models SODA (green), OFES (blue), and K7 (red). Other details are

the same as in Fig. 6.


to the Revisit period (2003–09). The decomposition of

the total transport (Table 3) demonstrates that the

baroclinic transport carried mainly by the Subantarctic

and Polar Fronts (baroclinic component in the central

part) varied much less than the total transport except

across 1428E, where the Polar Front flows almost par-

allel to the hydrographic section probably leading to

the variable transport in the central part. The variability in

FIG. 14. As in Fig. 13, but for across 1428E.


FIG. 15. As in Fig. 13, but for across 308E. Note that the (f) north component has a broader vertical range

(120Sv) than other panels (90 Sv).


the estimated total transport therefore came from other

components—the largest one being the north (across 308and 1428E) and barotropic (across 1158E and 688W)

components.

The steadiness of the baroclinic transport was also

confirmed in the OGCM outputs, where the standard

deviation of the baroclinic transport was always smaller

than the standard deviation of the total transport.

FIG. 16. As in Fig. 13, but for across 688W.


Given these steady baroclinic transport estimates and

the much longer time scale of baroclinic responses than

barotropic one (Hughes et al. 1999; Olbers and Lettmann

2007), we conclude that the decrease in total transport

estimated by the box inverse model reflects contributions

from short-period fluctuations, not a robust trend. This

conclusion is consistent with past studies detecting

no trends in the baroclinic transport across the Drake

Passage (Cunningham et al. 2003; Renault et al. 2011),

south of Tasmania (Rintoul et al. 2002), and south of

Africa (Swart et al. 2008).

In the OGCMs, the variability of the total transport

was often correlated with the south component (Table 4).

In discussing the response of the ACC to atmospheric

disturbances, Hughes et al. (1999) demonstrated the

importance of a ‘‘southern mode,’’ a barotropic mode

propagating mostly, but not always, along the closed

geostrophic contours for the Coriolis parameter f and

water depth H of f/H circumnavigating the Antarctica

shelf break. We speculate that the significant correlation

in the south transport is explained as a result of the

southern mode. The relatively low correlations across the

1158E south section might be related to the relatively

narrow shape of the southern mode there [Fig. 4d of

Hughes and Stepanov (2004)].

The trend in each component was estimated in the

same way as for the meridional transport. The only

significant trend was found in the barotropic compo-

nent through the Drake Passage with 20.45 6 0.27,

20.606 0.43, and20.506 0.28 Sv decade21 trends for

SODA, K7, and OFES, respectively. The box inverse

model also estimated a significant decrease in the baro-

tropic component (Fig. 16d and Table 3). This result

appears at odds with the enhancement of the westerly jet

over the Southern Ocean observed mainly in the 1990s

(Marshall 2003), but a close look at the SAM index (Fig.

17) shows that the decadal increase in the 1990s included

conspicuous periods of decrease around the years 1994

and 2000. A close look at the time series of the OGCMs

(Fig. 17) shows that the significant decreasing barotropic

trend can be attributed to the barotropic response of the

ACC to these wind decrease events rather than decadal

trends. The wind change in 2000 might be a part of

a larger change in wind pattern observed over the Pacific

and Indian Oceans (Lee and McPhaden 2008). For the

shorter period from 2003 to 2005, Zlotnicki et al. (2007)

found a decreasing trend in the bottom pressure aver-

aged along the southern edge of the ACC, as indicated

by satellite gravity data. The OFES and K7 models

showed a slight decrease over the period (Fig. 17), al-

though their estimated decrease rate [1.2 cm (H2O) yr21,

corresponding to 23.7 6 1.5 Sv yr21] was much larger

than ours. The decrease in the barotropic component

estimated in the box inverse model is therefore likely

a short time scale fluctuation caused by the wind.

Indeed, the total transport anomaly at all four sections

iswell correlatedwith SAMwhen smoothedby a 25-month

low-pass filter (Fig. 18) with correlation coefficients

TABLE 4. Correlation between total transport and components. Correlation coefficients between the total transport and decomposed

transport were calculated along with the p value, which is the probability that the estimated correlation occurs by chance. The correlation

coefficients larger than 0.5 with the p value less than 0.05 are emphasized in boldface type. The p value was calculated using the degree of

freedom estimated by decorrelation time scale (Trenberth 1984).

Component OGCM 308E 1158E 1428E 688W

South K7 0.54 (0.0) 0.24 (0.0) 0.54 (0.0) 0.25 (0.1)

SODA 0.42 (0.0) 0.19 (0.0) 0.64 (0.0) 0.79 (0.0)OFES 0.55 (0.0) 0.22 (0.2) 0.46 (0.0) 0.65 (0.0)

North K7 20.19 (0.2) 0.42 (0.2) 0.45 (0.1) —*

SODA 20.05 (0.6) 0.40 (0.4) 0.36 (0.3) —*

OFES 0.04 (0.6) 0.3 (0.2) 0.40 (0.0) —*

Barotropic K7 0.06 (0.4) 0.14 (0.2) 0.24 (0.0) 0.84 (0.0)

SODA 0.30 (0.1) 0.19 (0.0) 0.09 (0.1) 0.54 (0.0)OFES 20.03 (0.6) 0.06 (0.4) 0.13 (0.1) 0.63 (0.0)

Baroclinic K7 20.01 (0.9) 0.15 (0.5) 20.19 (0.1) 20.14 (0.3)

SODA 20.15 (0.1) 20.12 (0.0) 0.01 (0.9) 0.16 (0.3)

OFES 20.01 (0.8) 20.03 (0.6) 20.08 (0.5) 20.08 (0.1)

Bottom K7 20.02 (0.8) 0.04 (0.7) 0.40 (0.0) 0.84 (0.0)SODA 0.20 (0.2) 0.02 (0.9) 0.02 (0.8) 0.65 (0.0)

OFES 0.00 (1.0) 0.01 (0.9) 0.17 (0.0) 0.51 (0.0)

*Not defined.


between 0.37 (p 5 0.0) for SODA through the Drake

Passage and 0.85 (p 5 0.03) for OFES across 1158E,where p is the probability that the estimated correlation

occurs by chance. This extends to other key sections the

finding at the Drake Passage (Meredith et al. 2004;

Treguier et al. 2010) that transport estimated in the

Ocean Circulation and Climate Advanced Modelling

Project (OCCAM) is correlated with the SAM index at

interannual time scales.

Wang et al. (2010) found a decreasing trend of

21.80Svdecade21 in the ACC transport across the three

key sections (Drake Passage, 208E, and 1478E) in a

4D-VAR German contribution to estimating the Cir-

culation and Climate of the Ocean (GECCO) model

from 1960 to 2001. In our OGCMs, that trend was

found in SODA (22.30 6 0.56 Sv decade21) and K7

(20.85 6 0.54 Sv decade21) but not in OFES (0.80 60.40 Sv decade21). Wang et al. (2010) argued that the

decrease in total transport could be attributed to dif-

ferent density layers, indicating a baroclinic structure

in the decreasing transport, accompanied by decadal

changes in diapycnal transport among the layers. Our

4D-VAR model (K7) also showed a decreasing trend in

the baroclinic component (27.95 6 0.60 Sv decade21)

but compensated by an increasing trend in the south

component (7.82 6 0.62 Sv decade21). The decrease

estimated by the box inverse model (Table 3) was24 Sv

for the baroclinic and22 Sv for the sum of the baroclinic,

north, and south components. Taking the median dates

for the WHP (January 1993) and Revisit (January 2004)

data, these signify a trend of20.4 and20.2 Svdecade21,

respectively. Given the a priori uncertainty for the

thermocline water (gn , 26.0) of 5 Sv, the estimates by

the box inverse model are inconclusive. The trend esti-

mated by the OGCMs is too small to detect using repeat

hydrographic measurements revisited at decadal intervals

(Keller et al. 2007; Wunsch 2008). In an eddy-permitting

OGCM experiment, Treguier et al. (2010) found a de-

creasing trend in the ACC transport through the Drake

Passage that they attributed to model drift because

a larger decrease happened during the model spinup.

We found a similar long-term trend with decreasing

amplitude in the K7 results, which also showed merid-

ional drifts in the front positions (Figs. 15 and 16). The

model drifts were not clear in the OFES and SODA

outputs (not shown). Given the discrepancy between the

decadal running time for these models and millennial

time scale required for thermohaline equilibrium, further

data collection and extended model runs are required to

detect the ACC trend.

5. Conclusions

For the MOC strength, we find no significant decadal

changes in our OGCM analysis and in the box inverse

model. All OGCMs suggested strengthening of the

FIG. 17. (top) SAM index and (bottom) anomaly of the barotropic component at 688W(see Fig. 16d).


Deacon cell but the trends were not statistically signifi-

cant. No model showed strengthening in the AABW

cell; indeed the K7 and OFES models showed a statisti-

cally significant decreasing trend in theAABWstrength.

Part of this decrease is explained by the AABW vol-

ume contraction (Purkey and Johnson 2012). In the K7

and OFES models, this AABW contraction is associ-

ated with the reduced production and/or enhanced mix-

ing with the water masses above. The hydrographic data

used in the box inverse model do not have sufficient

temporal resolution to detect the AABW export trend,

but the decreasing trend found in the K7 and OFES

models arewithin the uncertainty of the inverse boxmodel

estimates.

For the ACC transport in our OGCMs and box inverse

model, we confirmed the stability of the baroclinic

transport (Rintoul and Sokolov 2001; Rintoul et al. 2002;

Cunningham et al. 2003; Legeais et al. 2005; Swart et al.

2008; B€oning et al. 2008; Sokolov and Rintoul 2009b;

Renault et al. 2011), in particular, the transport car-

ried by Subantarctic and Polar Fronts. The overall

transports through the four key sections, however, varied

FIG. 18. (top) SAM index and (bottom) total transport anomalies across 688W, 1428E, 1158E,and 308E.


with barotropic components and with meridional migra-

tion of the fronts, both of which explain the apparent de-

crease of 5–21Sv found in the box inversemodel estimates.

We find the overall transport anomalies at interannual

time scales are well correlated to the SAM index.

It is interesting that the stability of the MOC (at least

for water masses above and including Lower Circum-

polar Deep Water) and the ACC transports from the

1980s to the 2000s was found under the increasing west-

erly jet over the Southern Ocean (Fig. 1). The stability

could be a useful condition to constrain theories for the

MOC and ACC. The eddy saturation of the ACC is

model dependent and has not yet been fully formulated.

The zonal wind enhancement can lead to an increase

in the AABW cell if diapycnal diffusivity is enhanced

(Ito and Marshall 2008; Saenko et al. 2011), but the

AABW export may become weaker if the enhanced

wind deepens the thermocline (Nikurashin and Vallis

2011; Shakespeare and Hogg 2012). In simulation mod-

els, the link from large-scale wind and buoyancy input at

the surface to diapycnal mixing through eddy generation

depends on model configuration and parameterization.

Observational constraints would be of great use to study

the relationship.

The stability of the MOC and the ACC found in this

work is limited by the model run time and the data’s

temporal span as well as by the uncertainties in the

estimates. The uncertainties for the box inverse results

were typically65 Sv or more (e.g., Fig. 10). Keller et al.

(2007) argued that with 5-Sv uncertainty in revisit ob-

servation every 10 years of a zonal hydrographic sec-

tion, it would takemore than 100 years to detect a trend

in a simulated Atlantic MOC. A practical strategy, at

least in near future, is to seek trends in well-calibrated,

preferably data-constrained simulations and to use data

as consistency checks.We reemphasize the importance of

further data collection and extended model runs.

Acknowledgments. The hydrographic data were col-

lected as contribution to the World Ocean Circulation

Experiment and International GO-SHIP (www.go-ship.

org). We appreciate the efforts of all scientific and ship

personnel who enabled collection of the data. The Argo

data were collected and made freely available by the

International Argo Project and associated contributing

national programs (http://www.argo.ucsd.edu, http://argo.

jcommops.org). Argo is a pilot program of the Global

OceanObserving System. NCEPReanalysis 2 data were

provided by the National Oceanic and Atmospheric

Administration Earth System Research Laboratory,

Colorado, from its website (http://www.esrl.noaa.gov.psd).

The OFES data were made available by Dr. H. Sasaki,

the Earth Simulator Center, JAMSTEC.

APPENDIX

Box Inverse Model

The boxes are vertically separated by neutral density

surfaces: the sea surface, gn 5 26.0, 26.1, 26.2, 26.3, 26.4,

26.5, 26.6, 26.7, 26.8, 26.9, 27.0, 27.1, 27.2, 27.3, 27.4, 27.5,

27.6, 27.72, 27.8, 27.9, 28.0, 28.11, and 28.2, and the sea

floor, where gn denotes the approximate neutral density

of (10001 gn) kgm23 calculated by themethod of Jackett

and McDougall (1997).

Geostrophic velocity across sections with assumed

zero-velocity surfaces was calculated from hydrographic

data. The initial zero-velocity surfaces were at gn5 28.05,

28.10, and 28.15 for the zonal sections in the Pacific,

Indian, Atlantic Oceans, respectively, and at the bot-

tom for the meridional sections except for gn 5 28.318

for the SR3 section, south of Tasmania (Rintoul and

Sokolov 2001). The velocity correction at these initial

zero-velocity surfaces and the diapycnal fluxes were

then determined such that the volume, temperature,

and salinity in each box were conserved within pre-

scribed uncertainties. Preinversion diapycnal fluxes

were assumed to be zero. Silicate integrated from the

surface to the bottom was also conserved in the South

Indian box between I3/4 (208S) and I5 (348S). The

present boxmodel is similar to previous boxmodels (e.g.,

Ganachaud and Wunsch 2000; Sloyan and Rintoul 2001)

but with addition of separate surface boxes (Katsumata

et al. 2013) and constraints by velocity at 1000 dbar es-

timated by float drift data (Davis 1998, 2005; Katsumata

and Yoshinari 2010).

The prescribed uncertainties for each constraint used

in the row scaling were fixed following Ganachaud et al.

(2000) and Ganachaud (2003).

Only volume conservation was applied to the surface

boxes in the Southern Ocean. While it was technically

possible to add salt and temperature constraints, air–sea

fluxes of freshwater and heat have such large uncer-

tainties that addition of the salt and temperature con-

straints did not improve the uncertainty of the solution.

We note that the freshwater flux into the surface layer is

negligibly small [less than 0.1 Sv based on the NCEP–

Department of Energy (DOE) reanalysis]. The tem-

perature and salt fluxes between the surface and

interior boxes were thus calculated to satisfy the inte-

rior temperature/salt conservation conditions, while the

volume flux satisfy the volume conservation conditions

both in the surface and interior boxes. The South Indian

box did not have surface boxes and outcropping layers

there received a volume flux from heat and freshwater

estimated by using theNCEP–DOE reanalysis (Kanamitsu

et al. 2002).


Several deep valleys that are closed in the north,

forming cul-de-sacs, are found in the subtropical zonal

sections. Meridional transport through these valleys

was constrained to be 06 2Sv. In addition, the following

well-observed deep flows were constrained; bottom

water and Lower Circumpolar Deep Water through

the Tonga–Kermadec Trench, 6 6 1 and 9 6 1 Sv, re-

spectively (Whitworth et al. 1999); and bottom water

through theBrazil Basin, 46 2Sv (Speer andZenk 1993).

In addition to the bottom water constraints, transport

through the shallow Bering Strait was constrained at 0.860.3Sv (Roach et al. 1995). Apart from these, transports

across the hydrographic sections were not constrained.

REFERENCES

Abernathey, R., J. Marshall, andD. Ferreira, 2011: The dependence

of Southern Ocean meridional overturning on wind stress.

J. Phys. Oceanogr., 41, 2261–2278.

Aoki, S., S. R. Rintoul, S. Ushio, S. Watanabe, and N. L. Bindoff,

2005: Freshening of theAd�elie LandBottomWater near 1408E.Geophys. Res. Lett., 32, L23601, doi:10.1029/2005GL024246.

Bergmann, I., and H. Dobslaw, 2012: Short-term transport vari-

ability of the Antarctic Circumpolar Current from satellite

gravity observations. J. Geophys. Res., 117, C05044, doi:10.1029/2012JC007872.

B€oning, C. W., A. Dispert, M. Visbeck, S. R. Rintoul, and F. U.

Schwarzkopf, 2008: The response of the Antarctic Circumpolar

Current to recent climate change. Nat. Geosci., 1, 864–869.Cai,W., T. Cowan, S. Godfrey, and S.Wijffels, 2010: Simulations of

processes associatedwith the fast warming rate of the southern

midlatitude ocean. J. Climate, 23, 197–206.

Carton, J. A., and B. S. Giese, 2008: A reanalysis of ocean climate

using Simple Ocean Data Assimilation (SODA). Mon. Wea.

Rev., 136, 2999–3017.

——,H. F. Seidel, and B. S. Giese, 2012: Detecting historical ocean

climate variability. J. Geophys. Res., 117,C02023, doi:10.1029/

2011JC007401.

Chelton, D. B., M. G. Schlax, and R. M. Samelson, 2011: Global

observations of nonlinear mesoscale eddies. Prog. Oceanogr.,

91, 167–216.

Compo, G. P., and Coauthors, 2011: The twentieth century reanalysis

project. Quart. J. Roy. Meteor. Soc., 137, 1–28.

Cunningham, S. A., S. G. Alderson, B. A. King, and M. A. Brandon,

2003: Transport and variability of the Antarctic Circumpolar

Current in Drake Passage. J. Geophys. Res., 108, 8084,

doi:10.1029/2001JC001147.

Davis, R. E., 1998: Preliminary results from directly measuring

middepth circulation in the tropical and South Pacific. J. Geo-

phys. Res., 103 (C11), 24 619–24 639.

——, 2005: Intermediate-depth circulation of the Indian and South

Pacific Oceans measured by autonomous floats. J. Phys.

Oceanogr., 35, 683–707.

D€o€os, K., and D. J. Webb, 1994: The Deacon cell and the other

meridional cells of the Southern Ocean. J. Phys. Oceanogr.,

24, 429–442.

Durack, P. J., and S. E. Wijffels, 2010: Fifty-year trends in global

ocean salinities and their relationship to broad-scale warming.

J. Climate, 23, 4342–4362.

Farneti, R., T. L. Delworth, A. J. Rosati, S. M. Griffies, and

F. Zeng, 2010: The role of mesoscale eddies in the rectification

of the Southern Ocean response to climate change. J. Phys.

Oceanogr., 40, 1539–1557.

Ganachaud, A., 2003: Error budget of inverse box models: The

North Atlantic. J. Atmos. Oceanic Technol., 20, 1641–1655.

——, and C. Wunsch, 2000: Improved estimates of global ocean

circulation, heat transport andmixing fromhydrographic data.

Nature, 408, 453–457.

——,——, J.Marotzke, and J. Toole, 2000:Meridional overturning

and large-scale circulation of the Indian Ocean. J. Geophys.

Res., 105, 26 117–26 134.Gille, S. T., 2002: Warming of the Southern Ocean since the 1950s.

Science, 295, 1275–1277.

——, 2008: Decadal-scale temperature trends in the Southern

Hemisphere ocean. J. Climate, 21, 4749–4765.

Gordon, A. L., and Coauthors, 2010: The Indonesian throughflow

during 2004–2006 as observed by the INSTANT program.

Dyn. Atmos. Oceans, 50, 115–128.

Graham, A. M., R. M. de Boer, K. J. Heywood, M. R. Chapman,

and D. P. Stevens, 2012: Southern Ocean fronts: Controlled by

wind or topography? J. Geophys. Res., 117, C08018, doi:10.1029/

2012JC007887.

Hall, A., and M. Visbeck, 2002: Synchronous variability in the

Southern Hemisphere atmosphere, sea ice, and ocean result-

ing from the annular mode. J. Climate, 15, 3043–3057.

Hallberg, R., and A. Gnanadesikan, 2006: The role of eddies in

determining the structure and response of the wind-driven

Southern Hemisphere overturning: Results from the Model-

ing Eddies in the Southern Ocean (MESO) project. J. Phys.

Oceanogr., 36, 2232–2252.

Hogg, A. M., 2010: An Antarctic Circumpolar Current driven by

surface buoyancy forcing. Geophys. Res. Lett., 37, L23601,

doi:10.1029/2010GL044777.

Hughes, C. W., and V. N. Stepanov, 2004: Ocean dynamics asso-

ciated with rapid J2 fluctuations: Importance of circumpolar

modes and identification of a coherent Arctic mode. J. Geo-

phys. Res., 109, C06002, doi:10.1029/2003JC002176.

——, M. P. Meredith, and K. J. Heywood, 1999: Wind-driven

transport fluctuations through Drake Passage: A southern

mode. J. Phys. Oceanogr., 29, 1971–1992.

Ito, T., and J. Marshall, 2008: Control of lower-limb overturning

circulation in the Southern Ocean by diapycnal mixing and

mesoscale eddy transfer. J. Phys. Oceanogr., 38, 2832–2845.

Jackett, D. R., and T. J. McDougall, 1997: A neutral density vari-

able for the world’s oceans. J. Phys. Oceanogr., 27, 237–263.

Kanamitsu, M., W. Ebisuzaki, J. Woollen, S. Yang, J. J. Hnilo,

M. Fiorino, and G. L. Potter, 2002: NCEP DOE AMIP-II

Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1643.

Katsumata, K., and H. Yoshinari, 2010: Uncertainties in global

mapping of Argo drift data at the parking level. J. Oceanogr.,

66, 553–569.——, and M. Fukasawa, 2011: Changes in meridional fluxes and

water properties in the Southern Hemisphere subtropical

oceans between 1992/1995 and 2003/2004. Prog. Oceanogr.,

89, 61–91.

——, B. M. Sloyan, and S. Masuda, 2013: Diapycnal and isopycnal

transports in the Southern Ocean estimated by a box inverse

model. J. Phys. Oceanogr., in press.

Keller, K., C. Deutsch, M. G. Hall, and D. F. Bradford, 2007: Early

detection of changes in the North Atlantic Meridional over-

turning circulation: Implications for the design of ocean ob-

servation systems. J. Climate, 20, 145–157.

Kouketsu, S., and Coauthors, 2011: Deep ocean heat content

changes estimated from observation and reanalysis product


and their influence on sea level change. J. Geophys. Res., 116,

C03012, doi:10.1029/2010JC006464.

Kuhlbrodt, T., R. S. Smith, Z. Wang, and J. M. Gregory, 2012: The

influence of eddy parameterizations on the transport of the

Antarctic Circumpolar Current in coupled climate models.

Ocean Modell., 52, 1–8.

Lee, T., and Coauthors, 2010: Consistency and fidelity of Indonesian-

throughflow total volume transport estimated by 14 ocean data

assimilation products. Dyn. Atmos. Oceans, 50, 201–223.—— and M. J. McPhaden, 2008: Decadal phase change in large-

scale sea level and winds in the Indo-Pacific region at the end

of the 20th century.Geophys. Res. Lett., 35, L01605, doi:10.1029/

2007GL032419.

Legeais, J., S. Speich, M. Arhan, I. Ansorge, E. Fahrbach,

S. Garzoli, and A. Klepikov, 2005: The baroclinic transport of

the Antarctic Circumpolar Current south of Africa. Geophys.

Res. Lett., 32, L24602, doi:10.1029/2005GL023271.

Marshall, G. J., 2003: Trends in the southern annular mode from

observations and reanalyses. J. Climate, 16, 4134–4143.

Masuda, S., and Coauthors, 2010: Simulated rapid warming of

abyssal North Pacific waters. Science, 329, 319–322.

Masumoto, Y., and Coauthors, 2004: A fifty-year eddy-resolving

simulation of the World Ocean. J. Earth Simul., 1, 35–56.

Mazloff, M. R., P. Heimbach, and C. Wunsch, 2010: An eddy-

permitting SouthernOcean state estimate. J. Phys. Oceanogr.,

40, 880–899.McCartney, M. S., and K. A. Donohue, 2007: A deep cyclonic gyre

in the Australian–Antarctic Basin. Prog. Oceanogr., 75, 675–

750.

McDonagh, E. L., 2009: RRS James Cook Cruise JC031, 03 Feb-03

Mar 2009. Hydrographic sections of Drake Passage. National

Oceanography Centre Southampton, Tech. Rep. 39, 170 pp.

Meijers, A. J. S., N. L. Bindoff, and S. R. Rintoul, 2011: Frontal

movements and property fluxes: Contributions to heat and

freshwater trends in the Southern Ocean. J. Geophys. Res.,

116, C08024, doi:10.1029/2010JC006832.

Meredith,M. P., P. L.Woodworth, C.W.Hughes, andV. Stepanov,

2004: Changes in the ocean transport through Drake Passage

during the 1980s and 1990s, forced by changes in the Southern

Annular Mode. Geophys. Res. Lett., 31, L21305, doi:10.1029/

2004GL021169.

——, A. C. Naveira Garabato, A. M. Hogg, and R. Farneti, 2011:

Sensitivity of the overturning circulation in the Southern

Ocean to decadal changes in wind forcing. J. Climate, 25, 99–

110.

Nikurashin, M., and G. Vallis, 2011: A theory of deep stratification

and overturning circulation in the ocean. J. Phys. Oceanogr.,

41, 485–502.

Oke, P. R., and M. H. England, 2004: Oceanic response to changes

in the latitude of the Southern Hemisphere subpolar westerly

winds. J. Climate, 17, 1040–1054.Olbers, D., and K. Lettmann, 2007: Barotropic and baroclinic

processes in the transport variability of the Antarctic Cir-

cumpolar Current. Ocean Dyn., 57, 559–578.

Purkey, S. G., and G. C. Johnson, 2010: Warming of global abyssal

and deep Southern Ocean waters between the 1990s and

2000s: Contributions to global heat and sea level rise budgets.

J. Climate, 23, 6336–6351.

——, and ——, 2012: Global contraction of Antarctic Bottom

Water between the 1980s and 2000s. J. Climate, 25, 5830–5840.Renault, A., C. Provost, N. Senn�echael, and A. Kartavtseff, 2011:

Two full-depth velocity sections in the Drake Passage in 2006

Transport estimates. Deep-Sea Res. II, 58, 2572–2591.

Rintoul, S. R., 2007: Rapid freshening of Antarctic Bottom Water

formed in the Indian and Pacific Oceans. Geophys. Res. Lett.,

34, L06606, doi:10.1029/2006GL028550.

——, and S. Sokolov, 2001: Baroclinic transport variability of the

Antarctic Circumpolar Current south of Australia (WOCE

repeat section SR3). J. Geophys. Res., 106 (C2), 2815–2832.

——, ——, and J. Church, 2002: A 6 year record of baroclinic

transport variability of the Antarctic Circumpolar Current at

1408E derived from expendable bathythermograph and altim-

eter measurements. J. Geophys. Res., 107, 3155, doi:10.1029/

2001JC000787.

Roach, A. T., K. Aagaard, C. H. Pease, S. A. Salo, T. Weingartner,

V. Pavlov, and M. Kulakov, 1995: Direct measurements of

transport and water properties through the Bering Strait.

J. Geophys. Res., 100 (C9), 18 453–18 457.

Roether, W., R. Schlitzer, A. Putzka, P. Beining, K. Bulsiewicz,

G. Rohardt, and F. Delahoyde, 1993: A chlorofluoromethane

and hydrographic section across Drake Passage: Deep water

ventilation and meridional property transport. J. Geophys.

Res., 98, 14 423–14 435.Saenko, O. A., A. S. Gupta, and P. Spence, 2011: On challenges in

predicting bottom water transport in the Southern Ocean.

J. Climate, 25, 1349–1356.

Sall�ee, J. B., K. Speer, and R. Morrow, 2008: Response of the

Antarctic Circumpolar Current to atmospheric variability.

J. Climate, 21, 3020–3039.

Sasaki, H., M. Nonaka, Y. Masumoto, Y. Sasai, H. Uehara, and

H. Sakuma, 2008: An eddy-resolving hindcast simulation of

the quasi-global ocean from 1950 to 2003 on the Earth Simu-

lator. High Resolution Numerical Modeling of the Atmosphere

and Ocean, W. Ohfuchi and K. Hamilton, Eds., Springer, 157–

185.

Shakespeare, C., and A. M. Hogg, 2012: An analytical model of

the response of the meridional overturning circulation to

changes in wind and buoyancy forcing. J. Phys. Oceanogr., 42,

1270–1287.

Siedler, G. T., J. M€uller, R. Onken, M. Arhan, H. Mercier, B. A.

King, and P. M. Saunders, 1996: The zonal WOCE sections in

the South Atlantic. The South Atlantic, Present and Past Cir-

culation, G. Wefer et al., Eds., Springer, 83–104.

Sloyan, B.M., and S. R. Rintoul, 2001: The SouthernOcean limb of

the global deep overturning circulation. J. Phys. Oceanogr.,

31, 143–173.

Sokolov, S., and S. R. Rintoul, 2009a: Circumpolar structure and

distribution of the Antarctic Circumpolar Current fronts:

1. Mean circumpolar paths. J. Geophys. Res., 114, C11018,

doi:10.1029/2008JC005108.

——and——, 2009b: Circumpolar structure and distribution of the

Antarctic Circumpolar Current fronts: 2. Variability and re-

lationship to sea surface height. J. Geophys. Res., 114,C11019,

doi:10.1029/2008JC005248.

Speer, K. G., and W. Zenk, 1993: The flow of Antarctic Bottom

Water into the Brazil Basin. J. Phys. Oceanogr., 23, 2667–2682.

——, S. R. Rintoul, and B. Sloyan, 2000: The diabatic Deacon Cell.

J. Phys. Oceanogr., 30, 3212–3222.

Stewart, A. L., andA. F. Thompson, 2012: Sensitivity of the ocean’s

deep overturning circulation to easterly Antarctic winds. Geo-

phys. Res. Lett., 39, L18604, doi:10.1029/2012GL053099.

Straub, D. N., 1993: On the transport and angular momentum

balance of channel models of the Antarctic Circumpolar

Current. J. Phys. Oceanogr., 23, 776–782.Swart, S., S. Speich, I. J. Ansorge, G. J. Goni, S. Gladyshev, and

J. R. E. Lutjeharms, 2008: Transport and variability of the


Antarctic Circumpolar Current south of Africa. J. Geophys.

Res., 113, C09014, doi:10.1029/2007JC004223.

Toole, J. M., and B. A. Warren, 1993: A hydrographic section

across the subtropical South Indian Ocean. J. Deep Sea Res. I,

40, 1973–2019.

Treguier, A. M., J. Le Sommer, B. Molines, and J. M. de Cuevas,

2010: Response of the SouthernOcean to the southern annular

mode: Interannual variability and multidecadal trend. J. Phys.

Oceanogr., 40, 1659–1668.

Trenberth, K. E., 1984: Some effects of finite sample size and

persistence on meteorological statistics. Part I: Autocorrela-

tions. Mon. Wea. Rev., 112, 2359–2368.Tsimplis, M. N., S. Bacon, and H. L. Bryden, 1998: The circulation

of the subtropical South Pacific derived from hydrographic

data. J. Geophys. Res., 103, 21 443–21 468.Wang, W., A. K€ohl, and D. Stammer, 2010: Estimates of global

ocean volume transports during 1960 through 2001. Geophys.

Res. Lett., 37, L15601, doi:10.1029/2010GL043949.

Wang, Z., T. Kuhlbrodt, andM. P.Meredith, 2011:On the response

of the Antarctic Circumpolar Current transport to climate

change in coupled climate models. J. Geophys. Res., 116,

C08011, doi:10.1029/2010JC006757.

Whitworth, T., and R. G. Peterson, 1985: Volume transport of the

Antarctic Circumpolar Current from bottom pressure mea-

surements. J. Phys. Oceanogr., 15, 810–816.

——, B. A. Warren, W. D. Nowlin, S. B. Rutz, R. D. Pillsbury, and

M. I. Moore, 1999: On the deep western-boundary current in

the Southwest Pacific Basin. Prog. Oceanogr., 43, 1–54.Wijffels, S., and G. Meyers, 2004: An intersection of oceanic

waveguides: Variability in the Indonesian Throughflow re-

gion. J. Phys. Oceanogr., 34, 1232–1253.

Wunsch, C., 2008: Mass and volume transport variability in an

eddy-filled ocean. Nat. Geosci., 1, 165–168.

Yang, X., D. Wang, J. Wang, and R. X. Huang, 2007: Connection

between the decadal variability in the Southern Ocean circu-

lation and the southern annular mode.Geophys. Res. Lett., 34,

L16604, doi:10.1029/2007GL030526.

Zlotnicki, V., J.Wahr, I. Fukumori, andY. T. Song, 2007: Antarctic

Circumpolar Current transport variability during 2003–05

from GRACE. J. Phys. Oceanogr., 37, 230–244.


Documents

Variability in Southern Hemisphere Ocean Circulation from the …klinck/Reprints/PDF/katsumataJPO2013.pdf · 2013. 9. 23. · Variability in Southern Hemisphere Ocean Circulation