The Antarctic Circumpolar Current System - CCPOklinck/Reprints/PDF/rintoulOcnCirClim2001.pdfWe first describe recent observations of the structure and transport of the ACC (Section

4.6The Antarctic Circumpolar Current System

Stephen R. Rintoul, Chris Hughes and Dirk Olbers

271

OCEAN CIRCULATION AND CLIMATE Copyright © 2001 Academic PressISBN 0-12-641351-7 All rights of reproduction in any form reserved

CHAPTER

4.6.1 Flow in the zonally unbounded ocean

The absence of land barriers in the latitude bandof Drake Passage has a profound influence on thedynamics of currents in the Southern Ocean and,more generally, on the earth’s climate. Within thisband, the strong eastward flow of the AntarcticCircumpolar Current (ACC) connects each of theocean basins. Sverdrup dynamics in their usualform cannot be applied to flows within a zonallyunbounded ocean, and as a consequence thedynamics of the ACC have long been a topic ofdebate. Eddy fluxes are believed to play a morecentral role in both the dynamical and thermody-namical balances of the Southern Ocean than inother areas of the world ocean. The interbasinconnection providerchd by the ACC permits aglobal overturning circulation to exist; the over-turning circulation, in turn, dominates the globaltransport of heat, fresh water and other propertiesthat influence climate (see Gordon, Chapter 4.7; Bryden, Chapter 6.1; and Wijffels, Chapter 6.2).The vigorous interbasin exchange accomplished bythe ACC also admits the possibility of oceanic tele-connections, where anomalies formed in one basinmay be carried around the globe to influence cli-mate at remote locations (e.g. White and Peterson,1996). The fact that no net meridional geostrophicflow can exist across the unblocked latitudes iso-lates the Antarctic continent from the warmerwaters at lower latitudes to some extent, con-tributing to the glacial climate of Antarctica; whatheat does get carried poleward to balance the heatlost to the atmosphere must be carried by eddies.

Energetic interactions between the atmosphere,ocean and sea ice result in the formation of watermasses that play an important role in the globaloverturning circulation, and ventilate a substantialfraction of the volume of the global ocean. As aresult of these unique aspects, many characteristicsof the present-day ocean circulation and climatereflect the influence of the Southern Ocean.

The major currents of the southern hemisphereoceans are shown in Fig. 4.6.1. The ACC is thedominant feature in terms of transport, carrying amean transport of 134^13 Sv through Drake Pas-sage (Whitworth, 1983; Whitworth and Peterson,1985). The ACC consists of a number of circumpo-lar fronts, which correspond to water mass bound-aries as well as deep-reaching jets of eastward flow(Orsi et al., 1995). The two main fronts, the Sub-antarctic and Polar Fronts, are shown in Fig. 4.6.1.Poleward of the ACC, cyclonic gyres are found inthe embayments of the Weddell and Ross seas.Westward flow near the continental margin ofAntarctica (the Antarctic Slope Front) is found atmany locations around the continent (Jacobs,1991; Whitworth et al., 1998). Equatorward of theACC, the circulation consists of westward-intensi-fied subtropical gyres in each basin. Strong pole-ward flow in the western boundary currents isbalanced by weaker equatorward flows in the inte-rior of the basins. The main focus of this chapter ison the ACC itself, but exchanges between the ACCand the subtropical and subpolar regimes are alsoan important part of the story.

While the horizontal flows shown in Fig. 4.6.1are the dominant circulation features, the weaker

flows in the meridional-vertical plane are also significant. By following property extrema such asan oxygen minimum or salinity maximum, earlyinvestigators (e.g. Sverdrup, 1933) inferred theoverturning circulation shown schematically in Fig.4.6.2. Deep water spreads poleward and upwardacross the ACC and is balanced by equatorwardflow in lighter and denser layers. This pattern is dri-ven at least in part by the wind stress acting on thesea surface: south of the westerly wind stress maxi-mum (which generally lies near the axis of theACC), the Ekman transport is divergent and deepwater upwells into the surface layer; north of thewesterly wind maximum, the Ekman transport isconvergent and surface waters are downwelled intothe ocean interior. The water masses exported fromthe Southern Ocean to lower latitudes as part of

this overturning circulation are responsible forrenewing the intermediate and abyssal depths of thesouthern hemisphere oceans. However, althoughthe general pattern and significance of the merid-ional circulation of the Southern Ocean has beenrecognized for many decades, until recently noattempt had been made to quantify the flow pathsand water mass conversions implied by Fig. 4.6.2.

The ACC is also unique in the extent to whicheddy fluxes contribute to the dynamical and ther-modynamical balances. For example, in theabsence of mean meridional geostrophic flowacross the Drake Passage gap, eddies must carry asignificant poleward heat flux to balance the heatlost to the atmosphere at high latitudes and theheat carried equatorward in the Ekman layer (de Szoeke and Levine, 1981). A significant eddy

SECTION 4 THE GLOBAL FLOW FIELD272

150°E

120°E

90°E

60°E

30°

E

0°

30°

E

60°E

90°E

120°E

150°E

180°E

Benguela CAg

ulha

sC.

ACC

Subantarctic F.Polar F.

Leeu

win

C.

East

Aust.

C.

Wed

dellG

.

Ross

G.

ACC

MalvinasC.

Brazil C.

Fig. 4.6.1 Schematic map of major currents in the southern hemisphere oceans south of 20°S. Depths shallowerthan 3500 m are shaded.The two major cores of the Antarctic Circumpolar Current (ACC) are shown, theSubantarctic Front and Polar Front. Other abbreviations used are F for front, C for Current and G for gyre.

heat flux is consistent with observations in DrakePassage that the fronts satisfy the linear criteria forbaroclinic instability (Bryden, 1979; Wright,1981). Extrapolation of eddy heat flux measure-ments in Drake Passage and southeast of NewZealand to the entire circumpolar belt suggestedthe eddy fluxes were large enough to close the heatbalance (Bryden, 1979; Bryden and Heath, 1985),but concern remained that these observations werenot representative of the ACC as a whole.

In terms of theory and modelling of the ACC,the state of the art prior to the WOCE (WorldOcean Circulation Experiment) period can bebriefly summarized as follows. Early wind-drivenmodels of the ACC gave huge transport valuesunless very large friction was included. Munk andPalmén (1951) proposed that bottom form stressbalanced the wind stress. High-resolution quasi-geostrophic models tended to confirm this balance(e.g. McWilliams et al., 1978). At the same time,Sverdrup theory appeared to do a reasonable jobof predicting the transport and path of the ACC(Stommel, 1957; Baker, 1982). However, prior tothe WOCE era no high-resolution primitive equa-tion models with realistic geometry, stratification,

and full thermodynamics had been run, and manydynamical questions remained open.

Here we review what has been learnt about theAntarctic Circumpolar Current system over thelast decade, largely as a result of the WOCE pro-gramme. By the ‘ACC system’ we mean not onlythe zonal flow of the ACC itself, but also themeridional overturning circulation and watermasses of the Southern Ocean. Nowlin and Klinck(1986) provide a comprehensive review of ACCphysics as understood at that time, primarily onthe basis of measurements in Drake Passage andearlier circumpolar hydrographic surveys. Weassume the reader has some familiarity with thedynamics of ocean or atmosphere circulation.Readers interested in more background on topicssuch as Sverdrup balance, Ekman layers, andpotential vorticity are referred to textbooks suchas Gill (1982).

We first describe recent observations of thestructure and transport of the ACC (Section 4.6.2).Numerical and analytical models have led to sub-stantial advances with regard to the theory anddynamics of the ACC, and its links to the merid-ional circulation (Section 4.6.3). New observations

4.6 The Antarctic Circumpolar Current System 273

Rintoul, Hughes and Olbers

70 60 50 40 30

ContinentalShelf

Mid-Ocean Ridge

AABWNADW

UCDW

AAIW

SAMW

PF SAF STF

Antarctica

80°S

buoyancyloss

buoyancygain

LCDW

Dep

th (

m)

1000

2000

3000

4000

Fig. 4.6.2 A schematic view of the meridional overturning circulation in the Southern Ocean (from Speer et al.,2000).An upper cell is formed primarily by northward Ekman transport beneath the strong westerly winds andsouthward eddy transport in the UCDW layer.A lower cell is driven primarily by formation of dense AABW near theAntarctic continent. PF, Polar Front; SAF, Subantarctic Front; STF, Subtropical Front;AAIW,Antarctic IntermediateWater; UCDW, Upper Circumpolar Deep Water; NADW, North Atlantic Deep Water; LCDW, Lower CircumpolarDeep Water;AABW,Antarctic Bottom Water.

and analysis techniques have allowed the circula-tion, formation and modification of SouthernOcean water masses to be quantified for the firsttime, as discussed in Section 4.6.4. Observations,theory and modelling over the last decade have ledto new insights into the meridional circulation ofthe Southern Ocean and its link to the global over-turning circulation (Section 4.6.5). Finally, weidentify a number of important questions thatremain open, despite the substantial progress madeduring the WOCE era (Section 4.6.6). These openquestions will provide the challenges for the nextgeneration of field programmes and modellingefforts to follow WOCE.

4.6.2 Observations of the AntarcticCircumpolar Current

A decade ago our knowledge of the structure of theACC was largely built on detailed measurements inDrake Passage and coarse-resolution hydrographicsections at other longitudes. The increase in thenumber of high-quality, high-resolution sectionsand advances in remote sensing have revealed anumber of new features of the ACC.

4.6.2.1 Fronts of the ACCAn example of the high-quality sections spanningthe ACC collected during WOCE is shown in Fig. 4.6.3. This winter section between Australiaand Antarctica near 140°E (WOCE repeat sectionSR3) illustrates a number of features that are char-acteristic of the ACC system. As noted first byDeacon (1937), isopleths of all properties gener-ally slope upward to the south across the ACC in aseries of steps, or fronts. The locations of themajor fronts are indicated above the plots in Fig.4.6.3. (For the criteria used to define the fronts atSR3, see Rintoul and Bullister (1999).) A numberof prominent property extrema in Fig. 4.6.3 definethe well-known water masses of the SouthernOcean, as discussed in Section 4.6.4.

Orsi et al. (1995) and Belkin and Gordon(1996) have described the circumpolar path andcharacteristics of the major fronts of the ACCbased on careful analysis of a large number ofhydrographic sections across the ACC (e.g. Fig.4.6.4, see Plate XX). In addition to the two mainfronts of the ACC, the Subantarctic Front (SAF)and Polar Front (PF), Orsi et al. (1995) identifiedtwo other fronts that were circumpolar in extent,

the ‘southern front’ and ‘southern boundary’ ofthe ACC (Fig. 4.6.4, see Plate XX). They alsoshowed that while in Drake Passage the fronts arealmost always distinct features separating zones ofquieter flow and uniform water properties, atother longitudes the fronts of the ACC merge orsplit. The merging of fronts is particularly dra-matic in the southwest Indian Ocean (Fig. 4.6.1),where the Subantarctic and Polar Fronts of theACC and the Subtropical Front and AgulhasReturn Current are all in close proximity andtogether produce some of the largest temperatureand salinity gradients in the world ocean (Olberset al., 1992; Park et al., 1993; Read and Pollard,1993; Belkin and Gordon, 1996; Sparrow et al.,1996). The SR3 section shown in Fig. 4.6.3 pro-vides another example of a frontal structure morecomplex than the classical description based onDrake Passage experience: both the SAF and PFare split into two branches, and the southern SAFand northern PF have merged near 53°S.

While studies like those above have shown thatat most longitudes it is possible to identify particu-lar fronts in hydrographic sections, and so verifytheir circumpolar extent, advances in remote sens-ing and modelling of the ACC have provided a newview of the rich structure of the current. By repre-senting the SAF and PF as meandering Gaussian-shaped jets, Gille (1994) was able to map the fullcircumpolar path of the fronts from GEOSATaltimeter data and illustrate the extent to whichthe fronts were steered by topography. Stream-function maps from eddy-resolving numericalmodels reveal a more complex, filamented struc-ture to the ACC than that generally inferred fromhydrographic climatologies (e.g. Maltrud et al.,1998). Mean gradients of sea-surface temperature(Fig. 4.6.4, see Plate XX) also show a complex, fil-amented structure similar to that seen in the mod-els (Hughes and Ash, 2000).

4.6.2.2 ACC transportThe need to measure the transport of the ACC wasa key motivation for the International SouthernOcean Studies (ISOS) experiment in the 1970s.Early attempts to use a small number of currentmeters to provide a reference for geostrophic cal-culations were frustrated by the banded nature ofthe flow: the resulting transport estimates varieddramatically depending on whether a particularinstrument was in a front or not. During ISOS, the




0 500 1000 1500 2000 2500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

DISTANCE (km)

PR

ES

SU

RE

(db

ar)

1.5

0.4

0.2

0

0

0.2

0.2

0.4

0.4

0.4

0.6

0.6

0.6

0.8

0.8

0.8

1

1

1

1.2

1.2

1.2

1.2

1.4

1.4

1.4

1.4

1.6

1.6

1.6

1.6

1.8

1.8

1.8 1.8

2

2

2

2.2

2.2

2.2

2.4

2.6

3

5

6

454647484950515253545556575859606162636465lllllllllllllllllllll

LATITUDE (°S)

l lll l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l lll

Θ

SAMW

ST

F

SA

F

SA

FP

F

PF

SF

SB

AS

F

a

0 500 1000 1500 2000 2500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

DISTANCE (km)

PR

ES

SU

RE

(db

ar)

34.234.334.4

34.5

34.6

34.6

34.7

34.7

34.7234.74

34.68


LATITUDE (°S)


SALINITY

AAIW

LCDW

AABW

ST

F

SA

F

SA

FP

F

PF

SF

SB

AS

F

b

Fig. 4.6.3 Properties versus pressure along the WOCE SR3 repeat section between Australia and Antarctica(<140°E): (a) potential temperature (°C; contour interval is 1° for s[3°C, and 0.2° for s\2.6°); (b) salinity (on thepractical salinity scale, contour interval is 0.1 for solid contours, and 0.02 for dashed contours); (c) oxygen (mmolkg[1, contour interval is 20); and (d) neutral density gn (kg m[3, contour interval is 0.1).The section was occupied inwinter (September) 1996.The major fronts are indicated above the plots: STF, Subtropical Front; SAF, SubantarcticFront; PF, Polar Front; SF, southern ACC front; SB, southern boundary of the ACC;ASF,Antarctic Slope Front.The majorSouthern Ocean water masses are also indicated: SAMW, Subantarctic Mode Water;AAIW,Antarctic IntermediateWater; L(U)CDW, Lower (Upper) Circumpolar Deep Water;AABW,Antarctic Bottom Water (see also Section 4.6.4).Bold contours in (d) denote the isopycnals used to define the layers in Fig. 4.6.14.

mean transport of the ACC was estimated to be134 Sv, with an uncertainty in the mean of 13 Sv,and a range of 98 to 154 Sv (Whitworth, 1983;

Whitworth and Peterson, 1985). The ISOS trans-port time series was based on pressure gauges ateither side of the passage, with an average speed


0 500 1000 1500 2000 2500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

DISTANCE (km)

PR

ES

SU

RE

(db

ar)

180

180180

180180

180

200

200

200

200200

200

220

220

220

220 220

240240

260260

280280 300320 320


LATITUDE (°S)

l lll l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l l

OXYGEN

UCDW

ST

F

SA

F

SA

FP

F

PF

SF

SB

AS

F

c

0 500 1000 1500 2000 2500

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

DISTANCE (km)

PR

ES

SU

RE

(db

ar)

27.127.327.7

27.8

27.8

27.9

28.1

28.3

27.4

28

28

28

28.2

28.2


LATITUDE (°S)


γn

ST

F

SA

F

SA

FP

F

PF

SF

SB

AS

F

d

Fig 4.6.3. Continued

calculated from three hydrographic sections refer-enced to current meters used to ‘level’ the gaugesand so determine the absolute transport (Whitworthet al., 1982). By resolving the narrow fronts of the ACC, the ISOS measurements significantlyreduced the uncertainty in transport estimates ofthe ACC. Nevertheless, the loss of two mooringsin the vicinity of the Subantarctic and PolarFronts, the discrepancy between direct and hydro-graphic estimates of the vertical shear, and the fact

that the current meter moorings were not coher-ent, mean that the transports may be subject tosomewhat more uncertainty than the error barquoted above would indicate.

Whitworth et al. (1982) and Whitworth (1983)showed from ISOS measurements that thebarotropic variability was higher in frequency andlarger in magnitude than the baroclinic variability.This observation, together with the similaritybetween repeat sections across Drake Passage

(Reid and Nowlin, 1971), has sometimes beeninterpreted to mean that the variability at DrakePassage is entirely barotropic. This is not sup-ported by the ISOS measurements: the range ofobserved baroclinic transport in the upper 2500 m(70–100 Sv) is comparable to the range of absolutetransport (105–140 Sv), and the respective standard deviations are 5.5 Sv and 8.5 Sv. Thebarotropic variability is larger and of higher frequency, but the baroclinic variability is also significant.

Prior to WOCE, there were no measurements ofeither baroclinic or barotropic variability of theACC at other locations. Repeats of WOCE sectionSR3 south of Australia (140°E) show that thebaroclinic variability there is similar in magnitudeto that measured by dynamic height mooringsdeployed for 1 year in Drake Passage during ISOS.The variability is dominated by changes indynamic height at the northern end, as also foundduring ISOS. The SR3 section south of Australiaextends both further south (into colder water, withlower dynamic height) and north (warmer water,higher dynamic height) than the Drake Passagesection, and so the dynamic height difference andbaroclinic transport is larger at SR3. For example,the mean transport in Drake Passage above andrelative to 2500 m is 87 Sv (Nowlin and Clifford,1982; Whitworth, 1983); at SR3, the mean of sixCTD (Conductivity-Temperature-Depth) sectionsis 107 Sv (Rintoul and Sokolov, 2000), while themean based on 36 summer XBT (eXpendableBathy Thermograph) sections is 109 Sv (Rintoul et al., 2000). The mean baroclinic transport southof Australia (relative to a ‘best guess’ referencelevel: at the bottom except near the Antarctic mar-gin, where a shallower level is used consistent withwestward flow over the continental slope and rise;see Rintoul and Sokolov, 2000) is 147^10 Sv(mean^1 standard deviation), about 13 Sv largerthan the ISOS estimate of absolute transportthrough Drake Passage. The transport south of Australia must be larger than that at Drake Passage to balance the Indonesian Throughflow,which is believed to be of O(10 Sv) (Gordon,Chapter 4.7; Cresswell et al., 1993; Meyers et al.,1995). However, given the large remaining uncer-tainty in the barotropic flow at both chokepoints,the agreement is likely to be fortuitous.

Rintoul and Sokolov’s (2000) estimates of thebaroclinic transport variability south of Tasmania

show that the ACC itself is surprisingly steadywith time (Fig. 4.6.5). Variations in net transportlargely reflect variations in westward flow acrossthe northern end of the section, rather thanchanges in transport of the ACC fronts. Becausethe water flowing to the west south of Tasmania iswarm relative to the rest of the section, thechanges in this current branch have a relativelylarge impact on the net interbasin exchange ofheat. In other words, while the ACC is undoubt-edly the primary means of interbasin exchange,variations in the transport between the Indian andPacific basins are dominated by changes in theflow north of the ACC.

While our picture of the circumpolar baroclinicstructure of the ACC has become more complete inrecent years, progress in determining the barotropicflow, and hence improving our estimates of theabsolute transport, has been slower. The WOCEstrategy to determine the transport of the ACCrelied on several elements: repeat hydrographic sec-tions across each of the Southern Ocean ‘choke-points’, pairs of deep pressure gauges spanningeach chokepoint, and direct velocity measurementsfrom shipboard and lowered ADCPs (AcousticDoppler Current Profilers). Because of the width ofthe sections, directly monitoring the absolute trans-port with a coherent array of traditional mooredinstruments is not feasible.

Shipboard and lowered ADCPs combined withmuch more accurate navigation and heading mea-surements are likely to provide valuable con-straints on the barotropic component of the ACC.For example, Donohue et al. (2000) use ADCPobservations in the Pacific to infer that the flow atthe bottom beneath the SAF is significant, and inthe same direction as the near-surface flow, thusenhancing the transport of the SAF over that esti-mated from the thermal wind alone. At least at thepresent time, however, both SADCP (ShipboardADCP) and LADCP (Lowered ADCP) measure-ments are subject to uncertainties that are largeenough to prevent their direct use as a referencevelocity for estimating transports across long sec-tions (Donohue et al., 2000).

However, it is feasible to monitor a portion ofthe current directly, and such a strategy wasadopted at the Australian chokepoint, where a425-km-long coherent array of current meters,Inverted Echo Sounders (IESs), and seafloor elec-trometers (HEMs) was deployed across the main



axis of the ACC to measure both absolute trans-port and dynamics of the current (the SubantarcticFlux and Dynamics Experiment, SAFDE; Luther et al., 1997). The array spanned the SAF, the maincore of the ACC at this longitude, along the line ofWOCE repeat section SR3 (between roughly 49and 53°S in Fig. 4.6.3). Preliminary analysis of thebaroclinic (from the IESs) and absolute (from theHEMs) transport time series shows that while sub-stantial barotropic flows occur in some parts ofthe array at some times, the 701-day mean baro-clinic (relative to the bottom) and absolute trans-ports through the central 200 km portion of thearray are similar (54.4^3.1 Sv and 50.4^4.2 Sv,respectively) (Luther et al., 1998).

Pairs of deep (1000 m) pressure gauges weredeployed across a number of passages to monitortransport variability during WOCE. The idea,based on ISOS experience, is that pressure differ-ence across the passage is proportional to changesin absolute transport. Meredith et al. (1996) foundthat the variability of pressure difference measuredacross Drake Passage for 4 years during WOCEwas somewhat smaller than measured during theISOS experiment (corresponding to a transportstandard deviation of 8 Sv in WOCE comparedwith 10 Sv during ISOS). The greatest differencebetween the two records was the absence duringWOCE of any events like two observed duringISOS where transport changed by 40% in a few


EAC

SAF 105 ±7

5 ±5

22 ±2PF

SF/SB 32 ±3

ASF 1.5 ±1

SAZ 22 ±8

TAS 8 ±13

130oE 140oE 150oE 160oE 170 o

E 70oS

64oS

56oS

48oS

40oS

32oS

30oS

Fig. 4.6.5 Schematic summary of the main circulation features south of Tasmania, based on six repeats of WOCErepeat line SR3.The numbers represent top-to-bottom transports (mean^1 standard deviation).TAS, ouflow ofTasman Sea water; SAZ, anticyclonic recirculation in the Subantarctic Zone; SAF, Subantarctic Front; PF, two branchesof the Polar Front; SF/SB, southern front and southern boundary of the ACC;ASF,Antarctic Slope Front. See Rintouland Sokolov (2000) for details.

weeks. Hughes et al. (1999) used results from twoeddy-permitting numerical models to show thattransport correlated better with pressure measuredon the south side of Drake Passage than with pres-sure difference across the passage. Pressure to thesouth was also highly coherent around the coast ofAntarctica. The model transport variations werewell correlated with zonally averaged wind stress(with a lag of less than 3 days) near the south ofDrake Passage, and occurred in currents that arestrongly steered by f/H contours, rather than fol-lowing the path of the ACC. The circumpolarcoherence of pressure at the Antarctic continentalmargin is also observed in the WOCE pressurerecords, as is the relationship (also noted from theISOS measurements) between bottom pressure andwind stress, for semiannual and shorter periods.The pressure record at the northern side of thepassage is dominated by local effects, resulting inthe relatively weak correlation between pressuredifference and transport.

While much has been learnt about the circum-polar structure of the ACC in the last decade, wehave not yet made much progress in refining ourestimate of the mean absolute transport of theACC. Improved estimates of absolute transportare likely to come from inverse models capable ofsynthesizing the complete suite of WOCE obser-vations (hydrography, Eulerian and Lagrangianvelocity measurements, and remote sensing) withdynamical constraints. Development of such mod-els is an active research area. Several recent modelsgive absolute transport estimates that are simi-lar to geostrophic estimates relative to the bot-tom (e.g. Macdonald, 1998; Sloyan and Rintoul,2000a; Yaremchuk et al., 2000). However, giventhat each of these models start with a first guess of zero barotropic flow, and no such cal-culations have yet included the full WOCE dataset including direct velocity measurements, it isinappropriate to conclude that the barotropic con-tribution to the mean absolute transport of theACC is small.

4.6.2.3 Antarctic Circumpolar WaveThe ACC is of interest in part because it allowscommunication between the ocean basins. Onephenomenon that depends on the oceanic telecon-nection provided by the ACC is the Antarctic Cir-cumpolar Wave (ACW) identified by White andPeterson (1996). The ACW consists of anomalies

in sea-surface temperature, sea-level pressure, andsea-ice extent that propagate eastward around theSouthern Ocean. The patterns have zonalwavenumber 2 and circle the globe in about 8–9years, so the apparent period at any location isabout 4 years.

The discovery of the ACW has sparked consid-erable interest. Part of this interest lies in thepotential predictability offered by the ACW. Tworecent studies suggest that the ACW has a substan-tial impact on rainfall in Australia and NewZealand, and may provide some predictive skill(White and Cherry, 1998; White, 2000). Thephysics of the ACW, in particular the extent towhich it represents a coupled mode of the ocean–atmosphere system, has also been a topic of activedebate. The initiation of the ACW may be theresult of atmospheric teleconnections related to theEl Niño-Southern Oscillation (ENSO) (Petersonand White, 1998; Baines and Cai, 2000). Otherstudies suggest the ACW arises from, or is at leastmaintained by, atmosphere–ocean coupling withinthe Southern Ocean (Qiu and Jin, 1997; White et al., 1998; Goodman and Marshall, 1999; Talley, 1999; Baines and Cai, 2000). Severalrecent model experiments suggest, on the otherhand, that the ACW is a passive ocean response toatmospheric forcing, and not a true coupled mode.These studies themselves differ as to the nature ofthe atmospheric forcing that drives the ACW, withACW-like oscillations resulting from stochasticforcing (Weisse et al., 1999), standing patterns in the atmosphere (Christoph et al., 1998; Cai et al., 1999), or ECMWF (European Centre forMedium Range Weather Forecasts) re-analysisfluxes (Bonekamp et al., 1999). In summary, avariety of dynamical hypotheses have been pro-posed to explain the ACW, each of which succeedsin explaining at least some of the characteristics ofthe ACW. Longer time series of observations(including subsurface ocean measurements) andfurther modelling studies will likely be required toimprove our understanding of the mechanism ofthe ACW.

4.6.2.4 Eddy fluxes of heat and momentumThe large-scale heat budget of the area south ofthe ACC implies a significant poleward eddy heatflux across the current (de Szoeke and Levine,1981), and observed fluxes in Drake Passage (Bryden, 1979; Nowlin et al., 1985) and southeast



of New Zealand (Bryden and Heath, 1985) are ofthe right sign and sufficient magnitude to close theheat budget if extrapolated around the circumpo-lar belt. But the reliability of such an extrapolationis obviously open to question, given the length andheterogeneity of the ACC. With regard to eddymomentum fluxes, both the Drake Passage andNew Zealand measurements suggest the momen-tum flux carried by the lateral Reynolds stresses is small relative to the wind stress. The primarysignificance of eddies in the momentum budget ofthe Southern Ocean lies in their ability to transfermomentum downward across density surfaces,rather than horizontally, as described below.

Over the last decade, our understanding of theeddy field and its influence on the ACC hasimproved as a result of several advances: satellitealtimeter observations of the ACC as a whole, alimited number of additional current meter mea-surements, and numerical models capable ofresolving (or at least ‘permitting’) eddies.

Measurements of sea-surface height variabilityfrom satellite altimeters has permitted the eddyenergy distribution around the entire ACC to bemapped for the first time (Wunsch and Stammer,1995). High eddy energy is found where the ACCinteracts with topography or with poleward exten-sions of the subtropical western boundary currents(e.g. the Malvinas–Brazil Current Confluence).Morrow et al. (1994) showed that the lateralReynolds stresses were generally small, but onaverage tended to transfer momentum into the jetsof the ACC, accelerating the mean flow, althoughmore recent results suggest that the eddies act todecelerate some of the strongest jets (Hughes andAsh, 2000).

There have been only a few in-situ measure-ments of eddy fluxes in the ACC during WOCE.South of Australia, an array of four tall currentmeter moorings was maintained for 2 years (Phillipsand Rintoul, 2000). The array was deployed at theSubantarctic Front along the WOCE SR3 line (cen-tred on < 50.7° S in Fig. 4.6.3), in a region thataltimetry suggests is one of moderate eddy activity,with the eddy energy increasing rapidly down-stream. Although the eddy heat flux varies acrossthe array, the mean values show poleward eddyheat fluxes at all depths between 300 and 2500 mthat are significant at the 95% level. The eddyheat fluxes are larger in magnitude than the twoprevious such measurements, in Drake Passage

and southeast of New Zealand. If extrapolated tothe circumpolar belt, the eddy heat flux south ofAustralia would carry 0.9 PW of heat poleward(40-h to 90-day band-passed data, ‘poleward’defined as normal to the direction of daily shear),more than sufficient to balance the heat loss to theatmosphere and the export of heat in the Ekmanlayer. (Note that this estimate of the eddy heatflux contains both the divergent, dynamicallyactive part of the eddy heat flux and the non-divergent part, see Marshall and Shutts (1981).)The eddy heat flux scaled by the mean verticaltemperature gradient gives the vertical momentumflux (e.g. Johnson and Bryden, 1989). South ofAustralia, fluctuations in the ‘eddy band’ (40-h to90-day periods) carry momentum downward at a rate of about 0.2 N m[2 (2 dyne cm[2) at alldepths (i.e. at about the same rate as momentum issupplied by the wind stress).

4.6.3 Dynamics of the ACC

The absence of continental barriers in the latitudeband of Drake Passage makes the dynamics of theACC distinctly different in character from those ofcurrents at other latitudes. At levels where notopography exists to support zonal pressure gradi-ents, there can be no mean meridional geostrophicflow. The vertically integrated vorticity balance inthe Sverdrup approximation, which at least quali-tatively succeeds in describing the wind-driven cir-culation in the interior of closed basins, cannot beused to infer zonal flows in the zonally unboundedSouthern Ocean. Even the concept of a wind-drivencirculation in the Southern Ocean is inappropriate,as the wind- and buoyancy-forced circulations areinextricably linked. The unique dynamics of thezonal and meridional circulation of the SouthernOcean have attracted the attention of theoreticiansfor many years. Recent work has led to substantialprogress in understanding the heat, momentumand vorticity budgets of the ACC, although somequestions remain a source of controversy.

4.6.3.1 Sverdrup balance arguments applied to the ACCSverdrup balance holds in the interior of the sub-tropical gyres because the wind-driven circulationdoes not penetrate deep enough to interact withbottom topography. In spin-up calculations, thedeep circulation is effectively cut off (in the


absence of thermohaline forcing) by the westwardpropagation of baroclinic Rossby waves generatedat the eastern boundary and on topographic fea-tures (Anderson and Gill, 1975; Young, 1981). Infact, as long as the mean flow is slow enough thatRossby wave propagation is minimally affected, asuccession of Rossby waves of increasing verticalmode number acts to confine the circulation to anever shallower depth. This only stops when theflow speed becomes comparable to the Rossbywave speed, for the wave mode with the samedepth scale as the current.

In the Southern Ocean, observation confirmsthat the flow at all depths is strongly influenced bybottom topography. From the above argument, thiswould imply a flow speed comparable to theRossby wave mode with a vertical scale of 2000 m:the first baroclinic mode. This is consistent with theobservation that mesoscale features seen in temper-ature and sea-surface height propagate eastwards inthe ACC, compared with westward propagationelsewhere (Hughes et al., 1998). More importantly,it implies that the Sverdrup balance must be upsetby interactions with bottom topography.

Nevertheless, several attempts have been madeto apply Sverdrup theory to the ACC by assumingthat various topographic features act as ‘effectivecontinents’, blocking the flow. Stommel (1957),for example, suggested that the Scotia Island arc,east of Drake Passage, effectively extended theAntarctic Peninsula across the Drake Passage gap.Southward interior flow in Sverdrup balance withthe wind stress curl was returned in an unusualarrangement of boundary currents: a westernboundary current against South America, and aneastern boundary current along the coast of theAntarctic Peninsula and Scotia Island arc, the twocurrents being joined in some unspecified way byflow through Drake Passage. The transport is thengiven by the zonally integrated wind stress curl atthe southernmost latitude of South America(which is also the northernmost latitude of theScotia Island arc – there is no overlap; any overlapwould complicate this, since it would not then beclear at which latitude the wind stress curl was rel-evant). Baker (1982) found some support for thisargument in a comparison of wind stress curl at55°S with baroclinic transport through Drake Pas-sage from hydrography.

Webb (1993) suggested the Kerguelen Plateauwas a sufficient barrier effectively to block the

flow. Webb’s model is a highly idealized source–sink flow in a homogeneous, flat-bottomed ocean,but it can also be recognized as an application ofGodfrey’s (1989) ‘island rule’ to the geometry ofAntarctica, which immediately generalizes it to the case of a stratified ocean obeying Sverdrupdynamics except in specified western boundaryregions. The non-Sverdrup flow all occurs in west-ern boundary currents off the eastern coasts ofSouth America and Kerguelan Plateau (and possi-bly the Antarctic Peninsula), resulting in a flowaround Antarctica that is proportional to an inte-gral of the wind stress along a line encircling thecontinent. This flow becomes infinite in the limitwhere the northernmost latitude of KerguelanPlateau is equal to the southernmost latitude of South America, with no overlap. With onlySverdrup balance and western boundary currentsinvolved, Webb’s model is the most natural exten-sion of wind-driven gyre dynamics to the SouthernOcean.

While the Sverdrup models of Stommel (1957)and Webb (1993) give reasonable values for theACC transport (about 120 Sv) when combined withclimatological wind stress estimates, the theories areincomplete. There are latitudes at which neither ofthe proposed ‘effective’ continental boundaries isshallower than 2000 m, so the assumption thatthese block the flow must at least be contingentupon some assumption of the weakness of stratifi-cation at these latitudes. The dynamics allowing theeastern and western boundary currents to join inStommel’s model are not clear. Perhaps mostimportantly, these flat-bottomed Sverdrup modelsignore interaction between the flow and the bottomtopography.

The observation that the ACC penetrates togreat depth suggests this assumption is not justi-fied. Topographic interactions link the horizontaland meridional circulations, as can be seen mostclearly from the barotropic vorticity equation:

q0bCx\k?=pbe=H]k?=ett]k?=eF (4.6.1)

where C is the barotropic streamfunction, pb isbottom pressure, H is ocean depth, tt is windstress, and F represents frictional and non-linearterms and k is the unit vector in the local vertical(upwards). Integrating (4.6.1) over a zonal bandenclosed by two latitude lines f1 and f2, there isno net northward transport so the left-hand-side



integrates to zero, and Stokes’ theorem can beapplied to turn the right-hand-side into two lineintegrals along the bounding latitude lines, giving

$f1[pbHx]sx]Fx] dx[$f2[pbHx]dx]Fx] dx\0

(4.6.2)

In fact, the zonal momentum balance tells us morethan this, since the integral over depth and longi-tude of this balance is precisely

$[pbHx]sx]Fx] dx\0 (4.6.3)

at any latitude.It has been clearly established (Gille, 1997;

Stevens and Ivchenko, 1997) that this balance isdominated by the first two terms: the northwardEkman flux is balanced by a geostrophic south-ward return flow at depth, with a very small con-tribution from Fx. This means that Fx can also beneglected in equation (4.6.2). There are some com-plications due to the different meridional scales ofterms in (4.6.2), but in practice this is true when/1 and /2 are separated by more than about 3–5degrees of latitude.

The implication of the above is that, for thearea integral between these latitudes, =ett isalmost entirely balanced by =pbe=H. Returningto the barotropic vorticity balance, this means thatthe southward flow driven by the wind stress curl(as in a flat-bottomed Sverdrup balance) returnsnorth in a flow balanced not by viscous terms as ina Munk or Stommel boundary current, but by thebottom pressure torques. (If the two latitudes areseparated by less than about 3°, the dominant bal-ance is between bottom pressure torques and non-linear terms, as found by Wells and de Cuevas(1995).)

The role of topographic torques is graphicallyillustrated in Fig. 4.6.6 (see Plate XX). This showsthe barotropic streamfunction from the SouthernOcean of the global eddy-permitting model OCCAM(Ocean Circulation and Climate Advanced Model-ling Project) (Coward, 1996), superimposed on the bottom pressure torque (the first term on theright-hand-side of (4.6.1). Both quantities havebeen smoothed by 4.25° longitude by 3.25° latitudeaveraging to reduce the effect of non-linear terms. Itis clear from Fig. 4.6.6 that northward flows areassociated with positive torques, and southwardflows with negative torques, as in equation (4.6.1).

Physically, the curvature of the earth means that asmall circle drawn around a point on the earth’ssurface has its poleward extremity closer to theearth’s axis than its centre, and its equatorwardextremity further away, but not by the sameamount. Water flowing into the circle from thepolar side carries with it more azimuthal velocityrelative to the circle’s centre, as a result of planetaryrotation, than water leaving the circle at the equa-torial side. Hence (since f changes sign at the equator), a northward flow represents a removal ofanticlockwise angular momentum from the circle,which must be balanced by a positive torque, suchas bottom pressure torque or wind stress curl. Forcomparison with Fig. 4.6.6, a typical wind stresscurl over this region gives a torque of 10[7 N m[3.The fact that the bottom pressure torques are largerelative to the wind stress curl suggests that theagreement with Sverdrup balance found by Baker(1982) was fortuitous.

4.6.3.2 Meridional circulationMuch discussion of ACC dynamics has centred onthe meridional overturning circulation. In particu-lar, with an eastward wind driving a northwardEkman flux, how does the southward return flowcross the ACC? Integrating the zonal momentumequation over depth and longitude at a particularlatitude, the fact that there must be a return flowmeans that the integrated Coriolis force is zero,and the question can be rephrased as, what zonalforce balances the zonal wind stress? This is thequestion addressed almost 50 years ago by Munkand Palmén (1951), who concluded that Reynoldsstresses and viscous terms were probably toosmall, and that the wind stress is likely balancedby a bottom form stress due to pressure differencesacross major topographic features.

This viewpoint is now well established, with in-situ and satellite altimeter measurements (Brydenand Heath, 1985; Morrow et al., 1994; Phillipsand Rintoul, 2000) confirming the smallness oflateral Reynolds stresses, and eddy-resolving prim-itive equation models (Gille, 1997; Stevens andIvchenko, 1997) clearly demonstrating a balancebetween wind stress and bottom form stress.Eddy-resolving Quasi-Geostrophic (QG) modelsmust (by construction) also show such a balance(McWilliams et al., 1978; Treguier and McWilliams,1990; Wolff et al., 1991), but are useful for illus-trating how the balance is established. A summary


of the momentum balance in the different modeltypes is given in Olbers (1998).

In terms of the meridional overturning (the zon-ally integrated flow at fixed depths), this balance isequivalent to saying that the northward Ekmanflux returns in a geostrophic southward flow, sup-ported by zonal pressure differences across topo-graphy. The only difference from the situation in other oceans is that, at the latitudes of DrakePassage, the topography does not reach the sur-face, so the return flow must occur at depthsbelow about 2000 m, as in Fig. 4.6.7.

At first sight, such a deep overturning cellwould seem to require large diapycnal fluxes;10–15 Sv sinking below 1000 m at around 40°Swould be comparable to the effect of deep convec-tion in the North Atlantic, but in a region wheredeep convection does not occur. However, there isan important difference between flows averaged atconstant depth and at constant density. In the FineResolution Antarctic Model (FRAM), Döös andWebb (1994) have shown that a large fraction ofthis overturning, labelled the Deacon Cell, occurswith very little associated change in density, as inFig. 4.6.8. Fluid at a given density flows north,just east of Drake Passage, at one depth, and

southwards elsewhere at a slightly deeper level.Seen at a given depth, denser fluid flows north,and lighter fluid flows south, with no integratedflow except at the top (the Ekman layer) and atdepths blocked by topography. The flow inte-grated at constant depth then shows a cell pene-trating to great depth, but a fluid particle will passnorthward and southward across a given latitudeat depths separated by only a few hundred metres,and with no appreciable change of density. Thepossibility of such circulations decouples themeridional overturning integrated at constantdepth from that integrated at constant density.Analogous circulations are also seen in the tropos-pheric Ferrel cells (McIntosh and McDougall,1996; Karoly et al., 1997).

This observation means that the meridionaloverturning averaged on potential (or neutral)density surfaces is worth looking at in more detail.For purposes of discussion, it is useful to considertwo extreme possibilities.

¥ Case I. The northward Ekman flux at some lati-tude is all returned to the south in density layersthat do not intersect topography. The mostextreme (and least realistic) version of this



70°S 60° 50° 40° 30°

0

1000

2000

3000

4000

5000

Fig. 4.6.7 The overturning streamfunction from a 6-year mean of the Fine Resolution Antarctic Model, calculated byintegrating meridional velocity at constant depth, and then integrating vertically. Flow is anticlockwise around highs(pale shading) and clockwise around lows (dark), with a contour interval of 2.5 Sv.The unshaded region shows therange of latitudes and depths (Drake Passage latitudes), which are unblocked by topography at any longitude. SeeFRAM Group (1991) and Döös and Webb (1994) for details of the FRAM model.

scenario is the QG picture with only a few den-sity layers, as in the models referred to above.With no flux between layers, all the Ekman fluxreturns to the south in the top layer, and isopyc-nal averaging shows no overturning, whereaslevel averaging shows an overturning cell reach-ing the bottom layer, which is the only layercontaining topography.

¥ Case II. The southward return flow is all in den-sity surfaces that intersect topography. This sce-nario requires a feedback mechanism betweenthe wind stress and thermohaline forcing, so thatthe diapycnal flux into and out of these deep lay-ers to the north and south of the chosen latitudecan balance the northward Ekman flux.

Consider the balance of zonal momentum, inte-grated zonally and over two layers (which may bestratified), separated by an isopycnal. The upperlayer of thickness h includes the Ekman layer, the lower one reaches from z\[h to the oceanbottom. Writing the depth-integrated northward

volume flux in each layer as Vi, i\1, 2, the steady-state balances read

[q0f Vww1w\[hw9wpw9wxw]swxw (4.6.4)

[q0f Vwww2w\[hw9wpw9wxw[Hwwpwwwbwxw (4.6.5)

where the overbar denotes time and zonal mean(see Fig. 4.6.9 for a schematic showing the relation-ship between geostrophic meridional flow andinterfacial form stresses on an arbitrary layer). Incase I, with the Ekman flux all returning above theisopycnal at z\[h, we have Vw1w\Vw2w\0. Frictionand Reynolds stress are generally negligible and theonly remaining terms are wind stress and the pres-sure forces on the boundaries: interfacial form stressand bottom stress. Thus, for case I, the wind stressand bottom stress are in balance, and are equal tothe interfacial form stress at any isopyc-nal belowthe Ekman return flow and above the bottom. Somedynamical mechanism, such as stationary wavesexcited by topography in an eastward current, or


=

N

E

Down

Pressure

LowHigh

Density surface

N NS S W E

Fig. 4.6.8 Schematic showing an idealized trajectory of a water particle in the ACC moving on a density surface.Thetrajectory is shown in three dimensions, and projected onto the horizontal plane (top), a constant longitude plane(left), and a constant latitude plane (lower right).The resulting circulation integrated at constant latitude and depth, forthis density surface, is an overturning cell with a vertical extent of a few hundred metres. Deeper density surfacesshow similar overturning cells, with northward branches at the same depth as the southward branch of the cellrelated to lighter water, so the zonally integrated cell including all density classes represents a meridional overturningpenetrating to great depth, without a need for any water particles to traverse such a large depth range (lower left).Note that this circulation implies higher pressure where the density surface is rising to the east compared with whereit is deepening to the east.This results in an eastward pressure force (interfacial form stress) on the water below.Thisis related to the fact that the northward flow occurs where the vertical thickness of water above the density surface issmall, and southward flow where the thickness is large, so there is a net southward mass flux at lighter densities dueto the geostrophic flow.This partly balances the northward surface Ekman flux, since the interfacial form stress partlybalances the eastward surface wind stress.The same kind of pressure force acting on the sloping bottom topographyleads to the bottom form stress, which closely balances the zonally integrated zonal wind stress, since the zonal anddepth integral of northward transport is very small.

baroclinic instability (again requiring zonal flowscomparable to the baroclinic Rossby wave speed), isnecessary to maintain the structure of correlatedpressure gradients and isopycnal heights that pro-duces interfacial form stresses at depth (see Section4.6.3.3).

In case II, if we identify z\[h with the positionof any isopycnal that does not intersect the bottomor the Ekman layer at the latitude under considera-tion, then the northward Ekman flow all returns tothe south beneath this level, giving Vw1w\[Vw2w\swxw/q0f. The Coriolis force in the lower layerthen exactly balances the form stress, and there isno interfacial form stress on density layers abovetopography. The difficulty here is in supplying thelower layer with the sources and sinks of waternecessary to maintain this flow, without inducing

flows in the intermediate layers. A change in windstress, for example, would upset the balance, andcause some layers to start filling and others empty-ing. If this change in configuration could thenchange the buoyancy forcing by some mechanismsuch as that proposed by Gnanadesikan and Hallberg (2000) (a feedback between buoyancyforcing and interface height), then an equilibriummight be attainable. Understanding the feedbackmechanism could then lead to a prediction for thedensity structure and therefore the baroclinic flow.

Taken together, these two cases make plain theintimate relationship between wind forcing andbuoyancy forcing. Models can produce circumpolarcurrents when forced by wind alone or by buoyancyalone. The steady state requires a balance for both,and cannot be said to be driven by one or the other.Whether the real Southern Ocean is closer to case Ior case II is discussed in detail in Section 4.6.5.

4.6.3.3 Theoretical predictions of ACCtransportA complete theory capable of predicting theabsolute transport of the ACC is a formidablechallenge. Such a theory would need to accountfor both wind and buoyancy forcing, stratification,the effect of eddy fluxes in the momentum andbuoyancy budget, and for interactions between thestrong deep currents and bottom topography.While a complete theory requires elaborate mathe-matics, some insight can be gained into the factorscontrolling the transport of the ACC by appealingto a variety of simpler models.

Estimates from eddy flux parameterizationsSimple estimates of the baroclinic transport can bederived from the above considerations of momen-tum transfer in the ACC. These estimates rest onthe assumption that the transfer is mainly down-ward and carried by the interfacial form stress ofthe transient eddies. In the extreme case when thestress is transferred undiminished through the watercolumn down to depth (and taken up there by topo-graphic form stress) its magnitude is set by the sur-face wind stress sx. With h9\q9/qz and p9x\q0fv9

the interfacial stress hw9wpw9wxw turns into the lateralbuoyancy flux and the momentum balance in thewater column below the Ekman layer becomes

sx\hw9wpw9wxw <[ vw9wqw9w (4.6.6)fg}N2



z =z1 (x )

z =z2(x )

Geostrophic current ρ fvg=–px

pw pe

Longitude

Hei

ght

δz

δx w δxe

Fig. 4.6.9 Schematic demonstrating the meaning ofinterfacial form stress for an arbitrary (not necessarilyconstant density) layer of water (shaded).The neteastward force on the layer is given by[: px dx dz, whichis related to the net northward geostrophic masstransport in the layer by f :qvg dxdz\[: px dx dz.Thecontribution to this area integral from the verticalportion dz is (pw[pe) dz, where pw and pe are pressuresat the upper boundary of the layer.This can be written aspw z1xdxw]pe z1x dxe. Performing the vertical integralthen gives[: px dx dz\$(p1z1x[p2z2x)dx, where p1, 2 ispressure at z\z1, 2.This is the difference between theeastward pressure force on the top interface fromabove, and that on the lower interface.The layerconsidered may be bounded by isopycnals, in which casethese boundary forces are interfacial form stresses, orthe lower interface may be the ocean floor, in which casethe corresponding boundary stress is the bottom formstress. In the limit of a small density difference drbetween upper and lower surface, the difference inboundary stresses becomes the interfacial form stressdivergence (times dq).

Following Green (1970) and Stone (1972) the lat-eral buoyancy transport of eddies, growing in theinstability process, can be parameterized in term ofthe gradient of the mean flow, vw9wqw9w\[jry\

[jr0fuz/g. The idea to combine (4.6.6) with para-meterizations of the buoyancy flux for inferringthe transport in the form

j uz\sx/q0 (4.6.7)

was first pursued by Johnson and Bryden (1989).They used Green’s form of the diffusivityj\aZf Zl2/ÏRwiw obtained for a baroclinically unstableflow, where Ri\N2/(uz)

2 is the local Richardsonnumber, l is a measure of the eddy transfer scaleand the constant a measures the level of correla-tion between v9 and q9 in the buoyancy flux(a\0.015^0.005 according to Visbeck et al.(1997)). The shear of the zonal flow and windstress are then related by

a l2u2z\sx/q0 (4.6.8)

Johnson and Bryden’s results are obtained byequating the turbulence scale l with the baroclinicRossby radius k. For l\p2k, with k\NH/(Zf Zp),we get their estimate of the shear

uz\ 1 21/2\1 . 21/2 (4.6.9)

The first relation was used by Johnson and Bryden (1989), with k taken to be a measure of thebulk Rossby radius, and shows the shear is propor-tional to the local Brunt–Väisälä frequency N(z).More importantly, the shear is proportional to theroot of the wind stress amplitude sx. In the follow-ing we use a local Rossby radius and an exponen-tial Brunt–Väisälä frequency profile, N(z)\N0 exp(z/2d) . With sx\0.2Nm[2, H\3500m,N0\ 1.4e10[3 s[1, d\2500 m, and a width B\

600 km of the ACC, integration of (4.6.9) yields atransport of 82 Sv relative to the bottom.

Visbeck et al. (1997) suggest that in the pres-ence of differential rotation the eddy transfer maybe restricted by the Rhines scale Ïuw/bw rather thanthe Rossby radius. With l\Ïuw/bw we find a cubicrelation between sx and the velocity,

uu2z\ (4.6.10)

For the exponential N(z) this is easily integrated. Atransport of 67 Sv relative to the bottom and atotal transport of 124 Sv is obtained for the aboveset of parameters. In this model the transportwould only mildly increase with the magnitude ofthe wind stress, as (sx)1/3.

The action of eddies is not only manifested inthe interfacial form stress, it also implies an eddytransport of potential vorticity. A formulation of the momentum balance which is more precisethan (4.6.6) is expressed as a balance between theeddy Potential Vorticity (PV) flux and the verticaldivergence of the frictional stress (Marshall et al.,1993),

[ uw9wvw9w]f \vw9wqw9w\[(sx/q0)z (4.6.11)

This balance holds above the depth level wheretopographic blocking sets in. The eddy PV fluxconsists of the lateral Reynolds stress divergenceand the vertical divergence of the interfacial formstress. Equation (4.6.6) is in fact the consequenceof (4.6.11) if the Reynolds stress divergence issmall and significant frictional effects are absentbelow the Ekman layer. If eddy mixing of PV isdown the mean PV gradient, vvw9wqw9w\[kqy, vanish-ing of the eddy PV flux implies homogeneousmean PV. Observations indeed show that isopyc-nal vorticity gradients are small in and north ofthe Antarctic Current regime (Marshall et al.,1993). Furthermore, a linear relation was found toexist between the large-scale PV and density,fqz\a]bq, with d\f /b, the e-folding scale of thedensity field. This implies an exponential N(z), asassumed before, and it also imposes a constrainton the current shear,

uzz[ \b (4.6.12)

obtained by taking the meridional derivative offqz\a]bq. Vertical integration leads immediatelyto the velocity profile and the transport, expressedin term of the shear at some level z0, or the corre-sponding density gradient, or the parameters ofGreen’s parameterization green at the level z0. Amore meaningful interpretation is found if (4.6.12)is reformulated as constraint on the vertical profileof the diffusivity k by inserting (4.6.7),

[ \b (4.6.13)q0N2}

sxN2}jd

N2}j

}z

N2}f2

uz}d

vw9wpw9w}qz

}z

}y

N3(z)}

|f |3sxb}r0a

N(z)}

|f |sx/q0}

p3aH2sx/q0}

p3akHN}|f |

| f |3}N3

f 2}N2


Apparently, the assumption of a homogeneousPV state sets the vertical profile of the lateral diffu-sivity of buoyancy. Since N2 decays exponentiallywith scale d in this model, we find (1/j)z\

q0b/sx\constant, and thus

j(z)\ (4.6.14)

where j0\j(z0). In this model the shear consistsof two parts,

uz\ 3 ]b(z[z0)4 (4.6.15)

The first contribution is directly wind-driven. Thesecond contribution is driven by the eddies thathomogenize the associated PV. The transport (relative to the bottom) of this latter is fairly smalland westward <[2 Sv) whereas the first part contributes 39 Sv for our standard values and a dif-fusivity j0\1000 m2 s[1 at z0\[1000 m. Follow-ing kap then increases to 1200 m2 s[1 at depth3500 m, and (4.6.7) then implies qy(z0)/q0 <[2.1e10[10 m[1, in good agreement with observations.

Wind-driven flow in a two-layer QG channelyields very sluggish flow in the deep layer when itstopography is arranged such that the geostrophiccontours are blocked by the walls (see e.g. Wolff et al., 1991). Straub (1993) found a regime wherethe baroclinic instability arrests the shear at itscritical level, and with the assumption that thedeep flow vanishes, the transport becomes BHbl2

(notice that this correspond to the second term inequation (4.6.15)). This is only a few Sv, andStraub argues that this contribution would add asa ‘channel component of the Southern Ocean’ tothe values obtained from Sverdrup-type estimates.Though neat as a concept, the baroclinicallyarrested state seems not to occur in more realisticmodels like FRAM, nor in the real ocean: here thecurrent is highly supercritical with respect to thebaroclinic Rossby wave propagation (Hughes et al., 1998).

All these concepts determine the transport rela-tive to the bottom velocity. Evidently, with a bot-tom velocity of only 1 cm s[1 (this is the typicalsize of bottom velocities obtained with inversemodels, see e.g. Olbers and Wenzel, 1989) and adepth of 3500 m we gain a contribution of 21 Svfor a current width of 600 km. How good is the

assumption of zero bottom velocity? The compo-nent of the bottom velocity that is normal to theheight contours is constrained by the kinematiccondition of no flow through the bottom,w]u?=H\0 at z\[H. An estimate of the verti-cal velocity at the bottom may be obtained by integration of the planetary vorticity equation,fwz\bv , from below the surface Ekman layer(with depth D) to the bottom. One finds

k?=ett/(r0f )[w([H)\ E D

[Hv dz (4.6.16)

In the zonal mean the transport below the Ekmanlayer is returned in the Ekman layer, thenw([H)\[u([H)?=H<[(sx/y)/(q0f ) and thusu([H) < sx/(r0fdH) where dH is the height of thetopography. Values of the order of a few mm s[1

are obtained. In view of the fact that the cancella-tion between the geostrophic flow and the Ekmantransport certainly does not occur locally, and thefact that this constraint applies only to the compo-nent of u normal to the bathymetry, the estimateof the bottom velocity must be considered as alower bound.

As is evident from (4.6.9), (4.6.10) and(4.6.15), the dependence of the baroclinic trans-port on the amplitude of the wind stress and theBrunt–Väisälä frequency is generally governed bythe degree of non-linearity of the eddy flux para-meterization. It should be kept in mind that inthese parameterizations only transient eddy effectsare taken into account. As shown below (and in allanalyses of the zonally averaged momentum bal-ance of numerical models), vertical transfer ofmomentum is also established by standing eddies.

The barotropic formstress mechanismEstimates of the transport from a more completetheory, which includes the barotropic componentof the flow, are difficult to obtain without elabo-rate mathematics and extreme simplifications. Theflat-bottom case, with the usual frictional parame-terizations of the bottom or Reynolds stress, is cer-tainly an unrealistic oversimplification. For aflat-bottomed channel with constant wind stress,the total transport is Bsx/(q0R) or B3sx/(12q0A),where B is the channel width, R the coefficient oflinear bottom friction and A the lateral eddy vis-cosity. But this model leads to extremely largetransports for reasonable choices of the frictionalparameters (Hidaka’s dilemma). This dilemma is

b}f

sx}r0j0

N2}f2

j0}}}1](r0bj0/sx)(z[z0)



somewhat relieved when partial barriers are intro-duced representing continents and leaving smallergaps (Drake Passage) for the current to passthrough (Gill, 1968).

The flat-bottom case gives unrealistic resultsbecause it does not allow for the bottom formstress to work in the overall momentum balance(4.6.3), repeated here as

sx[sxb[Hwpwbwxw\0 (4.6.17)

This balance has been shown to hold in all moreor less realistic numerical models (sx

b is the fric-tional bottom stress, put to the linear from q0RHubelow). The relevance of equation (4.6.17) to thetransport becomes clear when the relation of thebottom form stress to the physical mechanismsresponsible for establishment of the bottom pres-sure field are considered.

The simplest of such models are barotropic withsimple topography. For Charney and DeVore’s(1979) barotropic model of QG flow over sinu-soidal terrain (with wavelength 2p/k), the bottomform stress is evaluated as

Hwpwbwxw/wqw0w\ (fd)2 (4.6.18)

where cR\b/d2 is the speed of barotropic Rossbywaves and d is the amplitude of the topographyrelative to the mean depth. From the form of(4.6.18) it is obvious that the form stress is mosteffective if the current speed equals the speed ofthe Rossby wave, a situation termed ‘topographicresonance’. Adapted to ACC conditions, (4.6.17)only yields the subcritical solution (R2! (jcR)2

and u ! cR) and the transport per unit widthbecomes

Hu\ (4.6.19)

with a\|f |/b. If the flow is constricted in a channelthis relation still applies (Olbers and Wübber,1991), but if the topography gets sufficiently highso that blocking of the geostrophic contours by thewalls occurs, i.e. d[dc,B/a, the flow switches to adifferent regime with transport

Hu\ (4.6.20)

as shown by Krupitsky and Cane (1994) forR/|f |OO(d3), d[dc, and in similar form by Wangand Huang (1995). The barotropic pressure formstress reflected in these expressions is seen to act as a drag on the flow that considerably reduces the transport compared to the flat-bottom valueBsx(r0R), so that transports of only 10–20 Sv areeasily achieved. In the blocked state with transport(4.6.20), the current runs through the channelentirely in boundary layers at the southern andnorthern walls, connected by an internal bound-ary-layer current following the blocked geostrophiccontours. Krupitsky et al. (1996) use an heuristicequivalent barotropic model (see also Ivchenko et al., 1999) to show that stratification can relievethis unrealistic behaviour by modifying thegeostrophic contours. In an unblocked channel – aCharney–DeVore model with topographic pertur-bations approaching zero at the walls – the currentis allowed to cross the geostrophic contour by fric-tional processes at all values of topography heightand only friction processes allow for a componentof the pressure that is out-of-phase with respect tothe topography.

Baroclinic mechanismsThe reaction of the zonal barotropic pressure forceon the topography leads to a strong reduction inthe transport in wind-driven barotropic models. Innumerical General Circulation Models (GCMs), itis found that baroclinicity increases the transportfrom the small values of the barotropic topo-graphic state to realistic values in the range of theobserved transport of the ACC. This appears bothin coarse-resolution models, for example the earlyexperiments by Bryan and Cox (1972), Cox(1975), and more recently by Olbers and Wübber(1991) and Cai and Baines (1996), and in modelswith eddy resolution, for example the FRAMexperiment (FRAM Group, 1991) and Gille(1997). Analysis of the momentum balance(4.6.17) in FRAM shows that the barotropic andbaroclinic bottom form stress components exceedthe wind stress by two orders of magnitude(Stevens and Ivchenko, 1997), with eastwardacceleration by the barotropic pressure field and acorresponding deceleration by the baroclinic pres-sure field largely cancelling, such that the windstress is almost balanced by the residual and themomentum balance (4.6.17) works essentiallywithout friction. Notice that in these baroclinic

dc}d(d[dc)

Lsx}r0Bpf

sx/(r0R)}}1](1/2)(dak)2

RHu}}R2]j2(u[cR)2

1}2


conditions the barotropic pressure does not act asa drag as in homogeneous models. Coarse modelshave an equally strong effect of the baroclinicpressure field but an unrealistically large contribu-tion from lateral friction.

Baroclinic pressure gradients are establishedeither by thermohaline forcing changing the strati-fication by water mass conversion, or simply byadiabatic rearrangement of a prescribed layeringof stratified mass being lifted over the topography.The latter mechanism is operating in baroclinicadiabatic models (case I in the terminology of Section 4.6.3.2). As shown in Olbers and Völker(1996) and Völker (1999), the waves that producethe topographic resonance of Charney and DeVore(1979) are now baroclinic. These are generated inresonance with the topography and become sta-tionary when the barotropic current speed equalsthe baroclinic Rossby wave speed. The transportdecreases strongly with increasing topographyheight, starting from a frictionally controlled stateat low heights with a transition to a complex reso-nant regime with multiple equilibria at intermedi-ate heights, and further to a state controlled bybarotropic and baroclinic bottom form stresses athigh topography. The dependence of the transportand shear of this model on the topography heightand other system parameters (friction and forcing)are displayed in Fig. 4.6.10. Stable solutions existwhere the curves are bold. Solutions with dotted

curves are unstable (in the window of unstablesolutions homoclinic orbits and chaotic behaviouris found; this disappears when increasing the num-ber of resolved modes). Though the momentumbalance (4.6.17) in this latter solution seems tooperate without friction, it should be pointed outthat the barotropic and baroclinic bottom stressesare due to phase shifts of the topographicallyinduced pressure gradients with respect to thetopographic undulations, which in turn are pro-portional to the coefficients of bottom and interfa-cial friction of the model, again in correspondenceto the barotropic model. It is particularly interest-ing that eddy effects do not appear explicitly in(4.6.17), but transient eddies or bottom frictionare needed to produce the phase shift in standingeddies, which is necessary to produce bottom formstress. The baroclinic topographic resonance the-ory determines the transport in adiabatic modelsin a manner similar to the barotropic Charney–DeVore mechanism: the bottom form stress is acomplicated resonance function of the barotropicand baroclinic velocities and the transport followsfrom (4.6.17) and a corresponding balance for thebaroclinic momentum. The structural properties of this low-order model are preserved when thedegrees of freedom are increased from the simplestnon-trivial model with 11 modes to a number representing a moderately resolved coarse model(with 75 modes).



0 0.05 0.1 0.15–1

0

1

2

3

4

5x 10–3

Wind stress (N m–2)

Tran

spor

t and

she

ar

0 2 4 6 8

Ratio interfacial/bottom friction parameter

0 200 400 600 800

Topography height (m)(a) (b) (c)

Fig. 4.6.10 Sensitivity of the zonal barotropic transport and shear in the low-order two-layer QG model of Völker(1999) to (a) wind stress amplitude, (b) interfacial and bottom friction, and (c) height of the topography.The flow isforced by sinusoidal wind stress sx in a zonal b-plane channel with sinusoidal topography elevation (periodic in the zonaldirection, one half sine in meridional direction and vanishing on the walls), friction between the layers and at the bottomis linear.The lower curve in each panel is the shear, the upper curve the transport (both are scaled). Bold lines indicatestable solutions, dotted lines indicate unstable solutions. Notice that there is a small window in the parameter spacewhere only unstable solutions exist.Values of the parameters (if not varied): interfacial friction parameter 2.9e10[7

bottom friction parameter 1.1e10[7 s[1, topography height 500 m, wind stress amplitude 0.1 N m[2.

The crucial role of transient eddies or non-idealfluid behaviour beneath the mixed layer is alsoclear in realistic conditions. Consider a contour ofconstant time-averaged PV on some density sur-face beneath the mixed layer but above topogra-phy. Since PV and potential density are bothmaterially conserved for an ideal fluid, any steady-state flow can have no component across the PVcontour. Thus if there is a return flow at this den-sity, since it must cross this PV contour (assumedto close around Antarctica), either friction or tran-sient eddies are required to permit this.

Analytical theories of the ACC in which thebaroclinic pressure field, and thus the bottomstress, is established by thermohaline forcing arestill lacking. Several numerical studies (e.g. Olbersand Wübber, 1991; Cai and Baines, 1996; Gent et al., 2000; Gnanadesikan and Hallberg, 2000)with coarse-resolution models have recently inves-tigated the dependence of transport on the buoy-ancy forcing at the surface. The state of the ACCin these models is generally intermediate betweenthe extreme cases I and II described in Section4.6.3.2. There is conversion of deep to lighterwater masses to allow for a deeper (than Ekmanlayer) reaching meridional cell but there is also aparameterized interfacial stress of some kind. It islikely that the barotropic and baroclinic formstresses are not solely created by thermohalineprocesses because the topographic resonancemechanism should be operating as well.

An increase of the ACC transport by anincrease of the buoyancy loss by increased brinerelease off the Antarctic shelf was documented in PE (Potential Energy) models by Olbers andWübber (1991) and clearly described in Gent et al.(2000). Using restoring boundary conditions forheat and salt and different wind fields, Gnanade-sikan and Hallberg (2000) found a similar strongincrease of the ACC transport with strengtheningof the overturning circulation (linked to a deeperthermocline and increased water mass transforma-tion in the northern hemisphere). As shown by Caiand Baines (1996) and Gent et al. (2000), parame-terizations of sub-grid mixing play an essentialrole: the ACC transport and the overturning trans-port are larger in the presence of a larger verticaldiffusivity and a smaller isopycnal diffusivity. Thelatter is in qualitative agreement with the baro-clinic transport models (4.6.7) and (4.6.15). It also

agrees with the baroclinic Charney–DeVore modeldescribed above, where the transport stronglydecreases with friction between the layers (see Fig.4.6.10b).

Topographic steeringThe important role of submarine topography inthe dynamics of the ACC was discussed in the pre-ceding sections. The topography also acts to steerthe current, as noted very early on by Sverdrup et al. (1942, pp. 468; 606–7) and described byGordon et al. (1978) (see also the pressure maps inWebb et al., 1991, and the steric height maps inOlbers et al., 1992). Steering by bathymetry hasalso been detected in the sea surface topogra-phy obtained from altimeter data, as reported byChelton et al. (1990) and Gille (1994). A labora-tory model of homogeneous and linearly stratifiedflow over realistic topography is reported by Boyeret al. (1993). The resulting flow is in fair agreementwith observations regarding steering by the majorridges and troughs, but it also shows significantdiscrepancies that are traced back to inadequaterepresentation of small-scale passages and fracturezones, unrealistic forcing (which was simulated bysources and sinks of mass) and neglect of the plane-tary b effect. Early models of topographic effectson the ACC were homogeneous (Kamenkovich,1962; Johnson and Hill, 1975), so that f/H con-tours inevitably dominated the flow pattern. Thebreaking of f/H or bathymetry control by stratifi-cation is demonstrated in many simple numericalmodels (e.g. Klinck, 1993) and the coarse- andhigh-resolution models of the circumpolar circula-tion discussed above.

Theory suggests the flow may be steered along bathymetry contours, latitude circles or thegeostrophic contours f/H. The conditions underwhich one of these effects will dominate can be clar-ified by use of the balance (4.6.1) of integrated vorticity. If the deep ocean is motionless,q0f keub\[(=p)b\0, the bottom torque term in(4.6.1) vanishes because (=p)b\=pb[gqb=H. Thisallows a ‘free mode’ in the transport streamfunc-tion following f contours; in case of locally weakstress curl the flow would follow such a path.However, a motionless abyss requires strong strat-ification to shield the flow from the influence oftopography. The ACC is far from such a state butthe converse condition of a homogeneous water


column is inappropriate as well. The influence ofstratification is visible in other forms of the vortic-ity balance

(ke=C) ? = ] Ug ? =H\k? =e(tt/(q0H))

(4.6.21)

or

H2ub?= ]Ug?=f\f k? =e(s/q0f)\fwE (4.6.22)

obtained from (4.6.1) using ke=C\Hub]

Ug[ke(s/f ) (ignoring lateral stresses and bottomfriction for simplicity). Here, Ug is the baroclinic(thermal wind) transport relative to the bottom.Obviously, in the case of weak stratification whenthe baroclinic transport term in the above balancescould be ignored, the transport streamfunctionand the bottom velocity both would follow f/Hcontours where the corresponding stress curls areweak. If, in addition, the variation of the planetaryvorticity f along the path of the flow is small thebathymetry contours act as characteristics (thetopographic bT\fDH/(HDL) is in fact generallylarger than the planetary b).

A more detailed consideration of stratificationeffects in models would obviously be required todistinguish between these different possibilities. Anintelligent shortcut has been pursued by Marshall(1995a,b) using the homogeneous potential vortic-ity model of Marshall et al. (1993). With a func-tional dependence fqz\Q(q), the density field qand the baroclinic transport Ug is determined by aboundary value, say q(x, z\0)\qs(x). Also thebottom density qb is determined by qs. Further-more, the bottom is a material surface and – sinceMontgomery potential, M\p]gqz, and densityare conserved along the three-dimensional flow inadiabatic conditions – we have a functional depen-dence Mb\Mb(qb). Assuming no friction in theabyss, the bottom velocity is geostrophic, i.e.fub\ke(=Mb]gH=rb). Inserting these relationsinto (4.6.22) we find (after some manipulation)

A(ke=qs)?= ]B(ke=qs) ? =f\ QswE

(4.6.23)

with

A\H2(H[Href)Qb, B\E 0

[HQz dz (4.6.24)

This vorticity equation determines the surface den-sity from the Ekman pumping wE, provided thefunctional relations Href(qb)\(1/g)Mb/qb andQ(q) are specified. The extreme cases with theircharacteristics f/H\constant and f\constantappear again in (4.6.23) – corresponding to ahomogeneous ocean with large A (because hereHref@H) and motionless abyss with large B(because here Q@Qb). For constant Href and linearQ(q)\a]bq (as found by Marshall et al., 1993)the characteristic problem (4.6.23) is linear andcharacteristics are easily computed. Marshall(1995b) presents a solution for the transportstreamfunction and the interior circulation for anunforced case (wE\0), prescribing a surface den-sity section between Antarctica and Australia (Fig.4.6.11). The solution transports 160 Sv, andstreamlines are largely zonal and quite indifferentto topography in large portions of the domain.Nevertheless, topography clearly steers the cur-rent over the major topographic features such asthe Atlantic Ridges, the South East Indian andMacquarie Ridges, the Pacific–Antarctic Ridge andthe East Pacific Rise. These patterns are enhancedfor larger values of the reference depth Href, whichleads to an increase of the deep currents. Outstand-ing in the solution is an exaggerated northwarddeflection over the Kerguelen Plateau, whereas theobserved equatorward displacement of the currentbehind Drake Passage and the gradual polewardmigration in the rest of the Southern Ocean is notreproduced, presumably due to neglect of the windforcing.

4.6.4 Water mass formation andconversion

Isopycnals shoal steeply to the south across theSouthern Ocean, reflecting the baroclinicity of theACC (Fig. 4.6.3). Water spanning a wide range ofdensity is thus directly forced by exchange ofmomentum, heat and fresh water with the overly-ing atmosphere and sea ice. The air–sea–ice inter-actions strongly modify the physical and chemicalproperties of outcropping layers and transformwater from one density class to another. The watermasses formed in this way ventilate a substan-tial fraction of the world ocean volume and are a key link in the global overturning circulation(Section 4.6.5).

f 2}g

f}H

f}H

f}H2

f}H



4.6.4.1 Subantarctic Mode Water and Antarctic Intermediate WaterSubantarctic Mode Water (SAMW) is formed bydeep winter convection on the equatorward side ofthe ACC (Hanawa and Talley, Chapter 5.4;McCartney, 1977). The deep convection imprintsthe SAMW with its characteristic properties: a ver-tically well-mixed (hence low potential vorticity)layer that is rich in oxygen (Fig. 4.6.3). These trac-ers allow the SAMWs to be tracked from their for-mation regions to the southern hemisphere

subtropical gyres, where they renew the waters ofthe lower thermocline (McCartney, 1982).

The Antarctic Intermediate Water (AAIW) isidentified by a salinity minimum layer thatdescends near the Subantarctic Front (Fig. 4.6.3b;Hanawa and Talley, Chapter 5.4). Early authorstraced the salinity minimum core layer polewardto where it outcropped in the cold, fresh AntarcticSurface Water, and inferred that circumpolar sink-ing along this layer renewed the AAIW tongue.The apparent spreading was thought to be driven


Fig. 4.6.11 Mass streamfunction in Sverdrups from the analytical model of Marshall (1995b), in which fluid parcelsnegotiate a variable bottom topography while conserving density and potential vorticity.The inviscid, adiabaticcirculation follows characteristics which lie between the f/H contours found in a homogeneous ocean and the f contours found in a strongly stratified ocean.

by either wind, buoyancy loss, or mixing alongisopycnals, depending on the author. McCartney(1982) presented an alternative view, that AAIW ismade up of the densest SAMW produced in thePacific, the end-product of gradual cooling, fresh-ening and loss of buoyancy by air–sea fluxes alongthe long SAMW circulation path across the Indianand Pacific Oceans. England et al. (1993) showeda similar mechanism operating in a GCM. Fromthe southeast Pacific, AAIW spreads north andwest in the Pacific and through Drake Passage.Further modification by air–sea fluxes and mixingin the southwest Atlantic produces a cooler andfresher variety of AAIW in the southwest Atlantic(Molinelli, 1981; Piola and Georgi, 1982; Piolaand Gordon, 1989; Talley, 1996).

High-quality hydrographic and tracer sectionscollected during WOCE have focused attention onthe distribution and circulation of SAMW andAAIW. The SAMW becomes progressively cooler,fresher and denser across the Indian and Pacificbasins (McCartney, 1982). Within each basin, thelighter SAMW varieties are injected further west,and are restricted to the southwest corner of thesubtropical gyre. Dense varieties entering the sub-tropical gyres on their eastern sides travel aroundthe gyres and extend to lower latitudes. While thecirculation of the denser AAIW shares some simi-larities with that of the overlying SAMW, there areimportant differences as well. In contrast to thenear-circumpolar formation of SAMW, the forma-tion of ‘new’, well-ventilated AAIW is limited tothe southeast Pacific and southwest Atlantic, asindicated by the gradual decrease in AAIW oxygenand potential vorticity across the Atlantic andIndian basins (Gordon and Molinelli, 1982; Talley, 1996). For example, at the WOCE SR3section south of Tasmania (the eastern limit of theAtlantic–Indian variety of AAIW), the oxygen andchlorofluorocarbon (CFC) saturations of AAIWare only 65% and 10–20%, respectively; the ‘ven-tilation age’ of AAIW based on CFC-11 is 22–26years (Rintoul and Bullister, 1999). AAIW entersthe lower thermocline of the Atlantic and Pacificsubtropical gyres mainly in the southeast quad-rant; in contrast, oxygen-rich AAIW enters theIndian basin preferentially in the southwest (Fine,1993; Toole and Warren, 1993). In the westernboundary currents of the subtropical gyres, modi-fied ‘older’ AAIW returns to the south, resulting inmultiple varieties of AAIW with similar densities

at some longitudes (e.g. 40°E, Read and Pollard,1993; 140°E, Rintoul and Bullister, 1999), and ashift in AAIW properties between basins (Piolaand Georgi, 1982).

Few estimates have been made of the formationrate of SAMW and AAIW. Even less attention hasbeen paid to the rest of the circulation loop associ-ated with export of SAMW/AAIW to lower lati-tudes: what is the fate of SAMW/AAIW exportedfrom the Southern Ocean? What water masses aremodified to supply the SAMW/AAIW, and whereand how does this occur?

Sloyan and Rintoul (2000a,b) have used a boxinverse model, which explicitly includes air–seabuoyancy fluxes and the water mass transforma-tions they drive (e.g. Walin, 1982; Speer andTziperman, 1992), to examine the formation andcirculation of SAMW and AAIW. Comparing thetransport in density layers at each Southern Oceanchokepoint shows that 18 Sv more SAMW/AAIWleaves the Indian Ocean sector south of Australiathan enters the basin south of Africa (Fig. 4.6.12).This production rate represents the net effect ofair–sea buoyancy fluxes, diapycnal mixing andmeridional exchange with the subtropical gyre.The net production in the Indian sector is balancedby net consumption in the other basins. SAMW isformed in the Indian sector both by cooling andfreshening of subtropical water carried south inthe Agulhas Current and its extension, and bywater flowing northward across the ACC (andgaining buoyancy) in the Ekman layer. SAMW is carried from the Indian to the Pacific by theACC, where it is modified by mixing and air–seaexchange, and ultimately returns to the Indianbasin via the Indonesian passages (Gordon, Chapter4.7). The SAMW therefore participates in anIndian–Pacific ‘throughflow gyre’ in which warmthroughflow water is converted to cool SAMW inthe Indian Ocean, and SAMW entering the Pacificbasin is converted back to water warm enough tosupply the throughflow.

In the zonal integral, 34 Sv of Upper Circumpo-lar Deep Water (UCDW) is upwelled south of theACC and converted to IW densities by air–seabuoyancy flux (Fig. 4.6.12; see also the discusion inSection 4.6.5 and Fig. 4.6.14). This transformationby air–sea fluxes of heat and fresh water is largelycompensated by diapycnal mixing south of 40°S.However, the compensation is not immediate.While the zonally integrated export of SAMW/



AAIW from the Southern Ocean is small (\5 Sv),the gross exchange is large (about ^80 Sv). Theinflow of ‘new’ SAMW/AAIW to the subtrop-ical gyres is roughly balanced by an outflow of‘old’ SAMW/AAIW, whose properties have beenmodified by mixing during their transit of thegyres. Part of the returning SAMW/AAIW is then converted to denser UCDW by diapycnalmixing.

Because the SAMW and AAIW are renewed byair–sea interaction on decadal time scales, theyprovide a good place to look for evidence ofchanges in forcing over such time scales. Severalrecent studies have identified decadal changes inthe SAMW and AAIW layers in the south Indianand Pacific oceans (e.g. Bindoff and Church, 1992;Johnson and Orsi, 1997; Bindoff and McDougall,2000). Wong et al. (1999) have shown that thesepatterns are generally coherent throughout thePacific. These changes are consistent with warm-ing and freshening at the surface outcrops of theselayers, as predicted to occur in coupled climatemodels forced by increasing greenhouse gas con-centrations (Bindoff and McDougall, 1994).

4.6.4.2 Circumpolar Deep WaterTwo prominent core layers underly the salinityminimum of the Antarctic Intermediate Waterthroughout the Southern Ocean. An oxygen mini-mum layer (Fig. 4.6.3c) is used to define the UpperCircumpolar Deep Water (UCDW) (Callahan,1972). At slightly greater depth (and density) liesthe salinity maximum of the Lower Circumpolar

Deep Water (LCDW) (Fig. 4.6.3b). The LCDWlayer is supplied by saline North Atlantic DeepWater (NADW) exported from the Atlantic. Theaddition of ‘new’ NADW in the South Atlanticproduces an oxygen and salinity maximumbetween the UCDW and LCDW entering the basinthrough Drake Passage (Reid et al., 1977). Thehigh-salinity signature of the NADW/LCDW canbe traced to low latitudes of the abyssal Indianand Pacific basins (Reid and Lynn, 1971). LCDWenters the low-latitude basins primarily in a seriesof deep western boundary currents (see Hogg,Chapter 4.5). Within the subtropical basins, theLCDW is slowly modified by mixing with sur-rounding fresher water, and the oxygen concentra-tions are reduced by biological consumption. Thereturn of the slightly less dense, low-oxygen waterfrom the Indian and Pacific basins supplies theoxygen minimum of the UCDW (Callahan, 1972).Both types of CDW spread poleward and upwardacross the Southern Ocean, ultimately outcroppingat the sea surface south of the ACC. As describedin the previous section and in more detail in Section 4.6.5, the outcropping of CDW and theresulting water mass transformation by air–seafluxes provides the main connection between thelower and upper limbs of the global overturningcirculation.

4.6.4.3 Antarctic Bottom WaterThe WOCE decade has seen substantial progress inour understanding of where, how and at what rateAABW is formed in the Southern Ocean. In the


0°

0°

60°S

30°

60°S 60°S

60°W 60°E

30°S31

{ 25

10212

39

28 4214

228

22

4

3 11

26

818

8 8➪ ➪ ➪

➪

➪

10

➪

➪

➪

➪

3

10

➪ Diapycnal fluxes

-4.3 +18.1 -13.8

120° 180° 120°W

Fig. 4.6.12 A summary of the circulation and formation of SAMW and AAIW, from the inverse model of Sloyan andRintoul (2000b). Numbers give volume fluxes in Sv of thermocline water (solid line, neutral density \26.0 kg m[3) andintermediate water (SAMW/AAIW, dashed line, neutral density 26.0\gn\27.4 kg m3). Open arrows representdiapycnal fluxes driven by air–sea exchange and interior mixing. Circled numbers in each of the Southern Oceansectors represent conversion of Upper Circumpolar Deep Water to SAMW and AAIW. Numbers in boxes are the netconvergence (]ve) or divergence ([ve) of SAMW/AAIW in each sector of the Southern Ocean due to meridionaland diapycnal fluxes; mass is conserved by a compensating divergence in zonal transport of the ACC.

Weddell Sea several major programmes have con-tributed to this advance (see Fahrbach et al., 1998,for a review). The new results are broadly consis-tent with the picture of circulation and water massformation in the Weddell developed by earlierinvestigators. Relatively warm Circumpolar DeepWater (sometimes called Warm Deep Water in the Weddell) enters the basin from the east in the Weddell gyre and is converted to Weddell Sea Deepand Bottom Water through ice–ocean–atmosphereinteractions along the southern and western mar-gin. Two mixing scenarios have been proposed for the formation of Weddell Sea Bottom Water:(1) mixing of Winter surface water, Warm DeepWater, and Western Shelf Water, whose salinityhas been enriched by brine rejected during sea ice formation (Foster and Carmack, 1976); and (2) mixing of Ice Shelf Water (formed by cool-ing and freshening of Western Shelf Water beneaththe vast ice shelves in the southern Weddell Sea)with Weddell Sea Deep Water and Warm DeepWater (Carmack and Foster, 1975; Foldvik et al.,1985).

Fahrbach et al. (1994) maintained a line of current meter moorings and repeat hydrographicsections across the central Weddell Sea between1989 and 1993. They estimate a Weddell gyretransport of 30 Sv, almost all of which is carried innarrow currents along the continental slope.Recent model results of Beckmann et al. (1999)indicate that Fahrbach’s section does not cutthrough the centre of the Weddell gyre, which hasa maximum transport exceeding 50 Sv, in agree-ment with observations along the Greenwichmeridian (Schröder and Fahrbach, 1999). By con-sidering inflow and outflow in density layersacross this section, Fahrbach et al. (1991) infer anet conversion of 3–4 Sv of Winter and WarmDeep Water to Weddell Sea Deep and BottomWater. Direct measurements of the outflow of bot-tom water (potential temperature\[0.7°C) frommoorings in the western boundary current show amean of 1.7 Sv, and a range from 0.8 to 3.9 Sv;additional bottom water export occurs offshore ofthe narrow boundary current (Fahrbach et al.,1994). Dense shelf water also escapes from theWeddell Sea to the Scotia Sea through gaps in theisland chain separating the basins (Whitworth et al., 1994).

The 700-km-long drift of Ice Station Weddell in1992 further refined our understanding of

ice–ocean–atmosphere interactions in the westernWeddell Sea (Gordon et al., 1993; Gordon, 1998).Taken together, the results of recent programmesidentify a number of distinct AABW sources alongthe southern and western rim of the Weddell Sea,each with a characteristic temperature, salinity andstable isotope signature. The relative importanceof the two mixing scenarios described above variesalong the rim of the Weddell. Entrainment ofwarmer and saltier deep water found over theslope largely determines the ultimate properties ofdeep and bottom water leaving the Weddell Sea.Gordon (1998) estimates the formation of WeddellSea Bottom Water (WSBW) (potential temperature\[0.7°C) during the period of the Ice Stationdrift to be 4.0–4.8 Sv. Mensch et al. (1998) esti-mate a similar formation rate (about 5 Sv of deepand bottom water) from tracer measurementsobtained during the Ice Station.

Satellite observations of a large polynya in theWeddell Sea in the late 1970s, and the large changesin temperature and salinity of Weddell Sea DeepWater that resulted (Gordon, 1982), first sparkedinterest in the variability of Weddell waters. A num-ber of recent studies have documented variations ofdeep and bottom water properties. Gordon (1998)concludes that the average salinity of the WSBWformed during the Ice Station is too low to providethe end-member required to account for the deepwater of the Weddell gyre, and suggests that thebottom water forming at the present time containsmore Ice Shelf Water. Nöst and Österhus (1998)describe the impact of several large grounded ice-bergs north of the Filchner depression, which havecaused the cessation of high-salinity shelf water for-mation there and led to a cooling and freshening inthe depression. Further afield, Coles et al. (1996)and Hogg and Zenk (1997) have observed coolingand freshening (on isopycnals) of AABW spreadingnorth in the South Atlantic, which they attribute tochanges in open ocean convective events in theWeddell Sea.

Recent studies have also identified or confirmeda number of important sources of AABW outsidethe Weddell Sea. Rintoul (1998) shows that theAdélie coast (140–150°E) is likely to be a more sig-nificant source of bottom water than previouslyappreciated. The Adélie Land Bottom Water is evi-dent in Fig. 4.6.3 as a thin layer of cold, fresh,dense, high-oxygen water found over the continen-tal slope and rise of Antarctica. He argues on the



basis of Worthington’s (1981) volumetric census,and T–S properties at the sills bounding the Australian–Antarctic Basin, that the Adélie LandBottom Water accounts for up to 25% of the globalvolume of AABW. The concentration of CFC-11 inplumes of Adélie Land Bottom Water is high([2.9 pmol kg[1 in 1991; Rintoul and Bullister,1999), and the CFC saturation (35%) is similar tothat in plumes observed in the southwest WeddellSea. High CFC concentrations observed over thecontinental slope at 30°E in the eastern Weddell Sea(Mantisi et al., 1991) indicate that a source existsalong the Enderby coast or in Prydz Bay as well, assuggested earlier by Jacobs and Georgi (1977).

Recent studies have also identified regions of theAntarctic coastline where AABW is not formed.For example, Fahrbach et al. (1994) show thatAABW is not formed in the eastern Weddell Sea,primarily because of the inability to confine shelfwater (and increase its salinity) over the narrowshelf there. In the Amundsen and BellingshausenSeas, nearly undiluted warm Circumpolar DeepWater intrudes onto the shelf and melts largeamounts of continental ice, preventing formationof dense saline water (Hellmer et al., 1998).

Foster (1995) confirmed the observation of Carmack and Killworth (1978) that the WilkesLand coast east of 150°E forms water that is notquite dense enough to sink to the seafloor. Sinkingof Antarctic shelf waters to deep, rather thanabyssal, layers is potentially of significant impor-tance to ventilation of the deep ocean at lower lati-tudes, since these lighter layers are less confined bybathymetry and are free to spread northward. Theformation of ‘not quite bottom’ water aroundAntarctica has so far received little attention, andthe transport is unknown. Weppernig et al. (1996)conclude from the distribution of stable isotopesthat a large fraction ([80%) of the dense Ice ShelfWater leaving the continental shelf in the WeddellSea supplies the WSDW at intermediate depth,rather than sinking to the bottom to supply WSBW.

A powerful demonstration of the utility of tran-sient tracer measurements is the use of CFC inven-tories to estimate the formation rate of AABW(Orsi et al., 1999). They find that about 8 Sv(4.9 Sv in the Atlantic and 3.2 Sv in theIndian–Pacific) of new bottom water (a 50 : 50 mixof dense shelf water and entrained deep water)must sink across the 2500 m isobath to explain theobserved CFC-11 inventory in the AABW layer

(water with neutral density greater than28.27 kg m[3). Much of the 3.2 Sv formed in theIndian–Pacific sectors is likely produced by sourcesin the Australian Antarctic Basin (e.g. AdélieLand), since the average CFC-11 concentrationthere is 2.5 times as large as that of the Atlanticand Pacific basins (Fig. 4.6.13).

The CFC-11 inventory of Orsi et al. (1999) pro-vides a strong integral constraint on the formationof AABW. Broecker et al. (1997) have noted thatPO4* (PO4*\PO4]O2/175[1.95mmolkg[1, where175 is the average molar Redfield ratio of O2 con-sumption to remineralization in the deep sea and1.95 is an arbitrary constant.) also provides a con-straint on the relative production of dense venti-lated water in the Southern Ocean and the NorthAtlantic: roughly equal contributions from the twosource regions are required to explain the PO4*value observed in the deep Indian and PacificOceans. The PO4* budget thus requires sinking of about 15 Sv of ventilated shelf water aroundAntarctica, a factor of three to four higher thanimplied by the CFC-11 inventory or by trans-port measurements near the sources. Part of thisdiscrepancy may be explained by exchange of ven-tilated shelf water with deep water lighter thanneutral density of 28.27 kg m[3. However, giventhat the CFC content of the lighter deep water is somuch lower than that of the dense abyssal layer itis difficult to see how this process can account for the discrepancy. Another possibility is thatAABW production in recent decades is substan-tially smaller than the average during the last millennium (Broecker et al., 1997). The conflictbetween estimates of ventilated deep water pro-duction based on CFC and PO4* remains unresolved.

WOCE sections, moored arrays, and otherrecent measurements have helped to map out thesystem of deep western boundary currents andthrough-passage flows that carry AABW andLCDW to lower latitudes (e.g. Mantyla and Reid,1995; Rhein et al., 1998; Whitworth et al., 1999;Zenk et al., 1999; see also Hogg, Chapter 4.5).

4.6.5 The Southern Ocean and the globaloverturning circulations

The global overturning circulation is often takento be synonymous with the circulation loopformed by sinking and export of NADW from the


North Atlantic, upwelling elsewhere in the ocean,and a return flow of lighter water. The ACC car-ries NADW from the Atlantic to the other oceanbasins and so forms an important link in this cir-culation path. But the significance of the SouthernOcean to the global overturning goes beyond thepassive role of redistributing NADW. Upwellingand buoyancy forcing in the Southern Ocean drivethe water mass transformations required to closethe NADW cell, as well as several additional over-turning circulations, both shallow and deep. Thenature of the overturning circulation is also inti-mately related to the dynamics of the ACC, asdescribed above.

To balance the sinking of NADW in the North Atlantic, somewhere deep water must beconverted to lighter water. Given that no zones of

concentrated upwelling from the abyss to the ther-mocline have been observed, the traditional assump-tion is that the upwelling is more or less broadlydistributed in the ocean interior. This assumption,together with conservation of potential vorticity,has profound implications for the deep flow, asillustrated in the simple and elegant theory of theabyssal circulation of Stommel and Arons (1960).

Direct measurements of mixing in the oceanpycnocline, however, typically find diffusivities anorder of magnitude too low to support theupwelling required to balance the sources of deepwater (Ledwell et al., 1993; Toole et al., 1994).Radiocarbon distributions also argue against wide-spread upwelling of deep water through the pycn-ocline (e.g. Toggweiler and Samuels, 1993). Onthe other hand, recent measurements suggest



00oo0o

180o

90o W

40oS

50oS

90oE

0.05

>0.1

0.1

0.15

0.15

>0.4

0.4

0.4

0.3

0.2

0.2

0.4

0.2

<0.15

0.15

0.1

>0.2

0.2

0.5

0.3

0.3

>3.5

0.15

0.1

0.05

0.6

0.4

0.5

0.35

2

0.8

0.3

0.2

2

0.60.6

0.8

0.4

0.3

<0.05

0.2

0.15

0.1

0.05

0.03

0.02

3

2

<1

1

>0.5

1

>1

1

0.5

3 >2>2

>0.7

1>0.5

0.1

0.05

1

AABWCFC-11 (pmol kg–1)

Fig. 4.6.13 Inventory of CFC-11 in the AABW layer (between neutral density 28.27 kg m3 and the seafloor), fromOrsi et al. (1999).The layer-mean CFC-11 concentrations in the Atlantic (Weddell–Enderby), Indian (AustralianAntarctic), and Pacific basins are 0.17, 0.47 and 0.18 pmol kg[1, respectively.

diapycnal mixing throughout much of the watercolumn is enhanced over rough topography(Polzin et al., 1997; Ledwell et al., 2000). Diapyc-nal mixing may also be enhanced near the bound-aries of the ocean (Wunsch, 1970; Armi, 1978).To date there are few direct measurements ofdiapycnal mixing, and so it is not yet clear if (non-uniform) interior mixing will prove to be sufficientto balance the sinking at high latitude.

An alternative view is that the required conver-sion of dense to light water occurs primarily in theSouthern Ocean, where deep isopycnals outcropand are exposed to air–sea fluxes of heat and freshwater (Döös and Coward, 1997; Toggweiler andSamuels, 1998). Here we reconsider the merid-ional overturning in the Southern Ocean, this timefrom the perspective of its connection to the rest ofthe world ocean.

As explained in the introduction, an active over-turning circulation in the Southern Ocean wasinferred by early investigators on the basis of theprominent core layers (Figs 4.6.2 and 4.6.3) whichextend across the ACC (e.g. Sverdrup, 1933).Deep water spreads south and upward across theACC, and is balanced by sinking and northwardflow of both lighter intermediate water and denserbottom water. This picture is broadly consistentwith the requirements of closure of the NADWoverturning – import of NADW to the SouthernOcean is balanced by export of IW and BW, asobserved in the South Atlantic (Rintoul, 1991;Saunders and King, 1995) – and suggests that thewater mass conversions taking place in the South-ern Ocean are a key element of the overall cell.The ‘intermediate’ cell (conversion of DW to IW)is consistent with northward Ekman transport dri-ven by the strong westerly winds: divergent Ekmantransport drives upwelling of deep water, which insteady state must be converted to lighter water bybuoyancy input from the atmosphere as it is drivennorth across mean density contours (Fig. 4.6.1;Toole, 1981; Speer et al., 2000).

However, in the ensuing 60 years, few attemptshave been made to quantify the flow paths in Sver-drup’s diagram. One recent exception is Schmitz’s(1995, 1996a,b) attempts to synthesize global viewfrom a large number of published estimates forindividual branches of the overturning circulation.Summing his estimates for the three individualbasins at about 40°S, the overturning in the South-ern Ocean consists of 53 Sv of deep water

(NADW/UCDW) flowing south, balanced by 48 Svof bottom water (AABW/LCDW) and 5 Sv of inter-mediate water (SAMW/AAIW) flowing north.Schmitz’s summary of ‘best guess’ values from theliterature suggests that in the zonal integral the deepcell (DW to BW) is much stronger than the interme-diate cell (DW to IW) of Sverdrup’s diagram.

Schmitz’s circulation scheme is derived from anumber of individual estimates, which may not beinternally consistent. His results are, however, verysimilar to estimates from a recent Southern Oceaninverse model, which provides an internally consis-tent solution that explicitly includes air–sea buoy-ancy forcing and diapycnal mixing (Sloyan andRintoul, 2000c). In the zonal integral acrossroughly 30°S they find 52 Sv of deep water flowingsouth, balanced by 46 Sv of lower deep and bottomwater and 6 Sv of intermediate water flowing north(Fig. 4.6.14; see caption for layer definitions).

These results suggest the deep overturning cell,in which dense AABW/LCDW exported to theIndian and Pacific is balanced by import of slightlyless dense DW, dominates the overall SouthernOcean overturning. Note that the contribution ofNADW to the zonally integrated overturning issmall by comparison: for example, the 10 Sv pole-ward flow of lower NADW (neutral densitybetween 28.0 and 28.2) is more than compensatedby strong equatorward flow in this density class inthe Indian and Pacific. The intermediate cell (con-version of DW to IW) is weak in the zonal integralat 40°S (although the gross exchanges in this den-sity class are large, as described in Section 4.6.4.1).

The observations of flow entering and leavingthe Southern Ocean across 30–40°S imply signifi-cant poleward transport in density layers shal-lower than the Drake Passage sill (neutral densityof 27.4[28.0 kg m[3). In addition, the net north-ward transport of light water is much smaller thanthe Ekman transport, suggesting that much of theEkman transport returns poleward at similar den-sity south of 40°S. Substantial poleward transportin density layers not blocked by topographyimplies divergence of the interfacial form stress, orequivalently, the eddy buoyancy flux (Section4.6.3). Speer et al. (2000) show that there arestrong gradients in isopycnal thickness (or poten-tial vorticity) across the ACC in the UCDW layer;eddy mixing will therefore tend to smooth out thegradient, resulting in a volume flux to the south.Meridional gradients of isopycnal thickness are




Indian

Pacific

Total

BW

BW

BW

BW

LDW

LDW

LDW

LDW

UDW

UDW

UDW

IW

IW

IW

TW

TW

IW

UDW

TW

SWWedSea SAVE 4 SAVE 2

IND32 IND18

PAC32

≈ 30° - 40° S

1

4

4

4

5

1

8

7

10

6

9

9

3

4

4

3

+2 -5

+12

-7

11

18 4

13

7

6

8

5

4 2

1

420

3

1

9

10

13

20

1225

17

+18

-10

-6

+7

8 12

83

2

25

18

8

1524

11

-14

+9

-6

+1

4

2

52

21

25

834

3

3151

27

TW

8

2

6

8 3Atlantic

γn 28.2 =

γn 28.0 =

γn 27.4 =

γn 26.0 =

γn 28.2 =

γn 28.0 =

γn 27.4 =

γn 26.0 =

γn 28.2 =

γn 28.0 =

γn 27.4 =

γn 26.0 =γn 24.0 =

γn 28.2 =

γn 28.0 =

γn 27.4 =

γn 26.0 =

7

+2

-2

Fig. 4.6.14 A schematic five-layer view of the overturning circulation (units of 106 m3 s[1 in each southernhemisphere basin, and the zonal sum, from the inverse model of Sloyan and Rintoul (2000c). SW, Surface Water;TW,Thermocline Water; IW, Intermediate Water; UDW, Upper Deep Water; LDW, Lower Deep Water; and BW, BottomWater.The neutral surfaces used to define each layer are shown.Air–sea flux-driven diapycnal fluxes are shown bybold, dashed arrows at the sea surface. Net diapycnal fluxes due to interior mixing are indicated by thin dashedarrows; their location along the isopycnal is schematic only. Numbers in boxes represent the net convergence (]ve)or divergence ([ve) of a particular layer in that sector of the Southern Ocean.Two-headed arrows in intermediatewater highlight that the net flux is the difference between nearly balancing northward and southward fluxes. Labelsbeneath each plot indicate the hydrographic sections used in the inverse model (e.g. PAC32 is a section at about 32°Sin the Pacific).

small in the LCDW layer, which lies below thetopography where mean geostrophic meridionalflow is possible. In the terminology of Section4.6.3.2, these observations suggest the real oceanis closer to case I.

Because the overturning circulation in numeri-cal models depends on how the model is forced(e.g. restoring or surface flux boundary condi-tions), how eddies are parameterized, and whetherthe model is in steady state, they do not provideconclusive evidence for or against deep versusshallow compensation of the Ekman transport.Nevertheless, a number of recent simulations, bothcoarse and fine resolution, support the idea that asignificant fraction of the Ekman transport is bal-anced by a return flow shallower than the topog-raphy (case I). For example, Hirst and McDougall(1998) show that a coarse-resolution level modelwith the Gent and McWilliams (1990) parameteri-zation of eddy-induced advection is consistentwith case I: in density coordinates, there is very lit-tle meridional overturning associated with theEkman transport between 40°S and 60°S. Whilethe degree of cancellation between the eddy-induced advection and the Ekman transport(nearly complete in their model) depends on thevalue chosen for the diffusivity, the modelimprovements that result when such a parameteri-zation is used support the notion that eddy-driventransport likely plays an important role in theoverturning circulation of the Southern Ocean.(The simulation is also improved by the decreasein horizontal mixing permitted when the eddyparameterization is used.) Killworth and Nanneh’s(1994) analysis of the zonal momentum budget inisopycnal layers in FRAM supports this conclu-sion: at the latitudes of Drake Passage, almost all(at the southern side) to about half (at the north-ern side) of the northward transport of light waterreturns south at densities that do not intersecttopography.

In summary, observations suggest significantpoleward flow in layers above topography, whichmust be driven by divergence of the eddy (standingand/or transient) interfacial form stress, in addi-tion to a poleward flow in density layers blockedby topography. A variety of numerical simulationsalso suggest that the presence of eddies permits asouthward flux at densities above topography,which compensates a large fraction of the north-ward Ekman flux. In QG simulations this compen-

sation must be complete, since no diabatic trans-port is permitted. In PE models that either resolveor adequately parameterize the effect of eddies,there is still a large degree of compensation. Thetime and zonal mean flow at constant depth, inwhich the Ekman mass transport returns at depthsblocked by topography, is thus decoupled from thetime and zonal mean flow at constant density, inwhich most of the Ekman transport returns at den-sities above topography.

4.6.6 Conclusions

Substantial progress has been made in understand-ing the circulation of the Southern Ocean duringthe ‘WOCE decade’. This progress has relied onadvances in observations, theory and modelling.Observations collected during WOCE represent asignificant achievement, given the challenges posedby the remote and often hostile nature of theSouthern Ocean. Highlights include a circumpolarsurvey of high-quality hydrographic, tracer andADCP data (including some of the first repeat sec-tions obtained in the region), a number of mooringarrays, float and drifter deployments, and satellitemeasurements of sea surface height and tempera-ture. Analytical models have provided insight intothe mechanisms responsible for setting the trans-port of the ACC. The last decade has also seenrapid development of numerical models of theSouthern Ocean, including the first GCMs to incor-porate stratification, realistic topography, and toresolve (or ‘permit’) eddies. The ability of coarse-resolution models to simulate the Southern Oceanhas also improved significantly, in part due to moreeffective parameterizations of the effect of eddies.Fine-resolution models have achieved sufficientrealism that we can use them to estimate the mag-nitude of individual terms in the momentum or vorticity budgets, to identify sites of strongtopographic influence, and to describe qualitativelythe circumpolar structure of the complex, fila-mented ACC, although it is still not clear what resolution is necessary for a truly realisticsimulation.

The momentum, vorticity and buoyancy bud-gets – and as a consequence, the zonal and merid-ional circulations – are intimately linked. Thedynamics of the ACC differ in character fromthose of strong currents in other ocean basins thatare zonally blocked. The now established fact that


the zonal wind stress is balanced by bottom formstress means that the Sverdrup balance is upset byinteractions with topography: meridional flowsdriven by the wind stress curl are returned in flowsbalanced by bottom pressure torques rather thanin viscous boundary layers. In addition, for a sig-nificant part of the northward Ekman flux toreturn in density layers that do not intersect topog-raphy at some latitude (as models and observa-tions indicate), a strong, deep-reaching ACC isneeded to maintain the interfacial form stressdivergence required for dynamical balance of thereturn flow. Without strong top-to-bottom flow,the ACC could not achieve a dynamical state thatis almost non-viscous, supercritical with respect toRossby wave propagation and meridionally con-strained in narrow fronts.

It has long been recognized that eddy fluxesplay an important part in the dynamics and ther-modynamics of the Southern Ocean. In the lastdecade the central role of eddies has become evenmore clear. Eddies carry heat poleward andmomentum downward across density surfaces; themomentum transfer helps establish the correlationbetween pressure and topography that providesthe bottom form stress to balance the wind; and asubstantial fraction of the meridional overturningcirculation is dynamically balanced by the diver-gence of interfacial form stress. In this sense, thedynamical analogy between the ACC system andthe mid-latitude troposphere is even more com-plete than previously appreciated.

The high-quality WOCE hydrographic andtracer sections have provided new insights into theformation and circulation of Southern Oceanwater masses. Buoyancy exchange with the atmos-phere drives substantial water mass transforma-tions, converting both light water to dense water(e.g. over the extensions of the subtropical westernboundary currents, and near the Antarctic margin)and dense to light (over much of the SouthernOcean, where northward Ekman transport com-bines with heat and freshwater input to convertdeep water to intermediate water). Tracers havebeen used to refine estimates of the rate of AntarcticBottom Water formation, to identify the main cir-culation pathways, and to quantify the ‘ventilationage’ of Southern Ocean water masses.

Many of these ideas have come together to pro-vide a new appreciation of the significance of theSouthern Ocean in the global climate system. The

ACC provides the interbasin connection requiredfor a global thermohaline circulation to exist. Per-haps a more important link between the SouthernOcean and the global overturning circulation is thetransformation of water masses driven by air–seaforcing and diapycnal mixing. Various lines of evi-dence suggest that diapycnal mixing rates in themain thermocline are too slow to support the tra-ditional view that sinking of deep water is bal-anced by upwelling uniformly distributed over theocean. Much of the conversion of cold to warmwater that is required to close the NADW over-turning circulation appears to take place in theSouthern Ocean, where deep water outcrops and isexposed to air–sea buoyancy forcing. The NADWcell, however, is overwhelmed in the zonal inte-gral by an even stronger deep overturning involv-ing conversion of upper deep water to denserlower deep and bottom water in the SouthernOcean, and conversion of lower to upper deepwater in the deep basins of the Indian and PacificOceans.

Analyis of WOCE data from the SouthernOcean is at an early stage. We anticipate furtherprogress will be made as the full suite of observa-tions collected during WOCE (e.g. hydrographyand tracers, floats, altimetry, moorings) are syn-thesized. Advances in theory and modelling of theACC also continue at a rapid rate. Among theimportant open questions to be addressed by theseanalyses and future observational programmes are:What is the absolute transport of the ACC, andhow and why does it vary in time? How sensitiveare the water mass conversions taking place in theSouthern Ocean, and the overturning circulationsof which they are part, to changes in atmosphericforcing? What are the relative contributions of‘deep’ and ‘bottom’ waters produced along theAntarctic margin to the ventilation of the deepsea? How representative are the WOCE-era mea-surements? Can we detect and interpret changesbetween measurements made during WOCE andhistorical or future observations?

Many, but not all, of these questions will beanswered as analysis of the WOCE data set contin-ues. However, while a major step forward, theWOCE observations are still sparse in space andtime. In a region as remote as the Southern Ocean,there will always be a strong reliance on remotesensing and autonomous instruments. For example,the development of profiling floats now provides



the opportunity to sample the variability of theSouthern Ocean on broad spatial scales for the firsttime. The floats complement, but do not replace, thehigh-density repeat sampling along fixed cruisetracks required for transport estimates. Direct,coherent and sustained in-situ measurements ofabsolute velocity on large spatial scales, as requiredto improve our understanding of the barotropicflow, remain beyond our present technologicalcapacity. However, the combination of a highlyaccurate and well-resolved geoid from planned satel-lite gravity missions and satellite altimetery will pro-vide an unprecedented opportunity to resolve theabsolute flow of individual jets in the ACC, and andmake possible quantitative studies of eddy–meanflow interactions in the Southern Ocean. The South-ern Ocean also remains a great challenge for oceanmodels, which suffer their greatest difficulties wherestrong currents and eddies interact with bottomtopography, a process central to the dynamics of theACC. While significant progress been made in the

theory, observation and modelling of the SouthernOcean, we still do not have a realistic theoreticalpicture of the dynamical processes that regu-late theresponse of the ACC to wind stress and thermoha-line forcing. Only with such a picture will it be pos-sible to place the WOCE observations into theirproper context and determine their limitations. Thechallenge is to use and extend the WOCE data andmodels to help build that picture.

Acknowledgements

We thank Trevor McDougall, Peter Baines, theeditors, and two anonymous reviewers for theircomments on an earlier draft. We also thankLouise Bell and Jennifer Moss for preparing sev-eral figures. This work is supported in part byEnvironment Australia through the CSIRO Cli-mate Change Research Program. Contribution No. 1628 from the Alfred-Wegener-Institute forPolar and Marine Research.


Documents

The Antarctic Circumpolar Current System - CCPOklinck/Reprints/PDF/rintoulOcnCirClim2001.pdfWe first describe recent observations of the structure and transport of the ACC (Section