GEOPHYSICAL RESEARCH LETTERS, VOL. 26, NO. 21, PAGES 3329-3332, NOVEMBER 1, 1999

The Intermediate Depth Circulation of the Western South Atlantic

O. BoebeP,R.E. Davis2, M. Ollitrault3, R.G. Peterson2, P.L. Richardson4, C. Schmid s and W. Zenk 6.

Abstract. The subsurfaceoceanic circulation is an important lantic, the pathwaysand advectivevelocity of intermediate part of the Earth climate system.Subsurface currents tradi- waterhave remained poorly known due to the complexitiesof tionally are inferredindirectly from distributionsof tempera- the South Atlantic circulation. ture and dissolvedsubstances, occasionally supplemented by In contrastto surfaceflow patterns[Peterson and Stramma, current meter measurements.Neutrally-buoyant floats how- 1991], which are available from satellitetracked drifters, ba- ever, now enable us to obtain for the first time directlymeas- sin-wide direct observationsof the intermediatelayer flow ured intermediatedepth velocityfields over large areassuch wereunobtainable until recentlyand alternativeindirect meth- as the western South Atlantic. Here, our combined data set odscarried significant uncertainties due to the lack of a refer- providesunprecedented observations and quantification of key encelayer of known velocity.Significant progress has now flow patterns,such as the SubtropicalGyre return flow (12 Sv; beenachieved through the trackingof large numbersof sub- 1 Sverdrup= 106m3s4), itsbifurcation near the Santos Plateau surfacefloats (Table 1), which drifted with the surrounding and the resultingcontinuous narrow and swift northwardin- watermass for periodsof oneto threeyears. Some floats were termediatewestern boundary current (4 Sv). This northward trackedacoustically [Rossby et al., 1986], andothers by satel- flowing water passesthrough complex equatorial flows and lites when the floatsperiodically ascended to the sea surface finally entersinto the North Atlantic. [Davis et al., 1996]. Our analysiscombines for the first time historicand recentfloat trajectoriesfrom variousindependent float programmes.The study concentrateson data located Introduction within the intermediatedepth layer between 650m and 1050 m [Reid, 1994]. Currentsin this layer exhibit little verti- Both deep wintertimeconvection and Nordic Seasoverflows cal shear [Reid, 1994] and, north of 50øS, representwell the in the North Atlantic form cold deep water which flows intermediatewater motions.The compositeview of all trajec- southwardacross the equatorand is balancedby northward tories(Plate 1) showsa complexbut structuredlarge-scale flow of warm upperlayer water [Broecker,1991]. The warm patternof flow. The resultingspace-time averaged flow pat- upperlayer water has its originsas thermocline(50 m - 500 m tern (Plate2) revealsthe probablemain routeof intermediate depth)and intermediate(500 m - 1200 m) water in the South waterthrough the SouthAtlantic, which is furtherclarified in Atlantic. Intermediatewater is thoughtto warm and upwellin a schematicinterpretative summary (Plate 3). Below we will the vicinity of the equatorinto thermoclinewater, which pro- follow this overallflow patternfrom southto north,compar- videsa sourcefor North Atlantic deepconvection and Nordic ing our findingwith indirectobservations and notions of the Seasinflow. In orderto understandand modelthe overturning SouthAtlantic intermediatedepth flow. cell and meridional heat flux variations we need to understand the origins,pathways and underwaymodifications of interme- The Confluence Zone and the Subtropical Gyre diate water. Although intermediatewater must flow from its formationsite in the south northwardthrough the South At- A regionthought to be importantto the exchangeof waters betweenthe AntarcticCircumpolar Current (ACC) and the SubtropicalGyre is the ConfluenceZone near the western boundaryat -38øS [Gordonand Greengrove,1986]. Mixing •Universityof CapeTown, Dept. of Oceanography,Ronde- bosch,South Africa; Now at: GraduateSchool of Oceanogra- phy,University of RhodeIsland, Narragansett, Rhode Island. Table 1. Float Data Obtainedat IntermediateDepth in the 2ScrippsInstitution of Oceanography,University of Cali- Western South Atlantic. forniaSan Diego, La Jolla,California. 3Laboratoirede Physiquedes Oc6ans, Ifremer Centre de Type of float Data coverage• Numberof Days of Brest, Plouzan6, France. floats data 4WoodsHole OceanographicInstitution, Woods Hole, Massachusetts. ALACE 2 1990 - 1996 70 21694 MARVOR 3 1994 - 1996 28 11762 •1OAA/AOML,Physical Oceanography Department, Mi- RAFOS 4 1992- 1996 75 22390 ami, Florida. SOFAR s 1989- 1992 13 4562 *Institutfiir Meereskundean der Universit•itKiel, Kiel, Germany. •Datacoverage indicates the period over which the array of instru- mentssampled data. It may be longerthan the samplingperiod of an Copyright1999 by theAmerican Geophysical Union. individualinstrument. 2Autonomous Lagrangian Circulation Explorer [Davis et al., 1996]. 3Bretonword for Seahorse[Ollitrault et al., Papernumber 1999GL002355. 1994].4Ranging and Fixing of Sound[Rossby et al., 1986].SSound 0094-8276/99/1999GL002355 $05.00 Fixing and Ranging[Richardson and Schmitz,1993].

3329 3330 BOEBEL ET AL.: INTERMEDIATE CIRCULATION OF THE WESTERN SOUTH ATLANTIC

The intermediatewater then flows eastwardaway from the Confluence Zone in the South Atlantic Current (SAC, red 0 • a hi,*' - - 2• •. , • • trajectoryalong 40øS). These flow patterns are confirmed by a statisticalanalysis of all the floatvelocity data. (Plate 2). The SAC (15_+6Sv at intermediatedepth, see caption of Plate2 for S.o'uthß a descriptionof thetransport estimates) and the adjacent ACC (19_+10Sv at intermediatedepth) both flow eastat speedsof 10- 20 cms '•, whichincrease with increasingsouthern lati- tude. The ACC derives most of its water from the Drake Pas- sage(red trajectory along 50øS), but part stems from the Con- fluence Zone where recirculated intermediate water is en- trained into the ACC. Duringtheir nearly parallel flow acrossthe Argentine Basin, the ACC and SAC repeatedlyexchange waters, in particular

•o

•W 55 • • • • • 25 • 15 lOW Plate 1. Summ• plotof trajectoriesof 186 subsu•acefloats in the South Atlantic Ocean based on four diCerent inst•ment t•es: ALAC• [Dd•is et alL, 1996],MARVOR [OHitrd,•tet alL, 1994], RAMOS [Boebe• et M., 1998] and SO•AR [Richdrdxo•• Schmit•, 1993] floats.•e launchpositions of MARVOR, RAMOS and SO•AR floats are indicatedby dots,with the subsequent,order of one to two yearstrajecto- ries based on daily subsu•ace positions. In addition MARVOR floats su•ace eve• two or threemonths to trans- mit data. •ach positionat time of subsequentdescent is markedby a dot. ALAC• floats provideunderwater dis- placementsover two to threeweeks inte•als, indicatedby a straightline with the descentpositions (sta• of driP) marked by smalldots. Selected floats in colourdepict prominent flow patterns(red if the total displacementis eastward,blue if westwardand thick blackif trappedin westernbound• cur- rents).Contour lines indicate 3000 m and 1000m depthand the coastline. •e continent and the area between the coast andthe 1000m isobathare •old and light •reen, respectively. Plate2. Space-timeaveraged mean velocities based on a total See Plate 3 for named features. of 170float years of velocitydata. For theinterior ocean, away from the 1500 m isobath,the box size is 2 ø of latitudex 4 ø of longitudeon a 1øx 2 ø grid.The arrows, based on at least60 of ventilated(younger and fresher) and recirculated (older and daysof data,are centred at therespective box centre and are saltier)intermediate water in theConfluence Zone is exempli- redif thezonal component is eastward(blue if westward) and fied by the thick black and red trajectories(Plate 1) which thespeed is greater 3 cm s 4. Transportestimates ofthe inter- convergeat the westernboundary near 38øS. Recirculated in- mediatedepth component of theACC, SACand the Subtropi- termediatewater (thick black trajectory)is derivedfrom the calGyre Return Current are latitudinal integrals of zonalaver- westwardReturn Currentof the SubtropicalGyre. The thick agesof thesegridded mean velocities. Integration limits are blacktrajectory approaches the ConfluenceZone aftertravel- derivedfrom minima in the zonal transportfunction which ling south-westwardunderneath the Brazil Current at a speed clearlyseparate three transport bands: 21øS-36øS (the Sub- of 10- 20 cms". Floats launched in theDrake Passage [Davis tropicalGyre ReturnCurrent), 37Sø-44øS (the SAC) and 45øS-53øS(the ACC). A layerthickness of 400 m is assumed. et al., 1996](red trajectory along the western boundary) depict Thetransport errors derive only from the statistical uncertainty a northwardflow of 30 - 40 cms", followingthe Falkland of themean gridded velocities and do not include uncertainties Currentup to 38øS[Peterson et al., 1996].Within the Conflu- in the layerthickness, current width or verticalvelocity shear. enceZone, the intermediatelayer is highlyturbulent with ve- Along the westernboundary at -200 km intervals,data lo- locitiesof 60 cms" andhigher. This vigorous eddy motion is catedbetween the coastand the 1500 m isobathwas averaged a mechanismby which freshintermediate water is injected within 200 km radius(black arrows,based on at least 15 days into the SubtropicalGyre [Boebel et al., 1999]. of data ). Contoursas in Plate 1. BOEBEL ET AL.: INTERMEDIATE CIRCULATION OF THE WESTERN SOUTH ATLANTIC 3331

•N turn Current splits into a narrow northwardIntermediate Western BoundaryCurrent (1WBC, northwardthick black trajectoriesin Plate1), anda widersouth-westward flow band (Plate 2, black and blue arrowssouth of 28øS).The splitting ratio is estimatedto be approximately1:2, with the larger amount (-9 Sv) being recirculatedsouth [Schmid, 1998]. Contraryto Wiist'sscheme [1935], no northwardflow is ob- servedalong the westernboundary between the Confluence Zone and 28øS. Wiist's flow schemewas first challengedafter 30 yearsby Taft [1963] and later by Reid [1994] who hy- pothesisedthe flow schemewhich is directlyobserved here for the first time, after another30 years.

The IWBC and the Equatorial Region The northwardflowing IWBC (Plate 2, black arrowsnorth of 28øS) is observedquasi-continuously from 28øS to 2øS by sevenfloats which were entrainedinto the boundarycurrent. Float trajectoriesand current meter records indicate a jet width on the order of 30+5 km (see Schmid [1998] for a de- tailed descriptionof the 1WBC width), a thicknessof 400 m [Boebelet al., 1997], and a mean speedof-30 cm S-1 (Plate 2). This suggestsa northwardtransport of 4+2 Sv for the in- termediatelayer, with the error includingboth the uncertainty in thecurrent's width and its speed (+10 cm s"). This estimate compareswell with estimates(7.7 Sv at 19øS,4.0 Sv at 21øS and 2.7 Sv at 24øS) of the 1WBC transportbased on recent Plate 3. Schematicdiagram of main flow patternsat interme- WOCE sections[Schmid, 1998], and fall in the lower rangeof diate depthin the westernSouth Atlantic. Contourlines and estimates(4-8 Sv) presentedby Fu [1981] and Roemmich hatching as in Plate 1. Abbreviationsas follows: Falkland [1983] for the 27.0-27.4 o0 densityrange at 24øS. Current, FC; Northern Intermediate Counter Current, NICC; Southof the Vit6ria-TrindadeRidge, small scalerecirculation EquatorialIntermediate Current, EIC; SouthernIntermediate cells are frequentlyfound offshorethe IWBC. Farther north, Counter Current, SICC; South Equatorial Under Current, between5øS and 17øS,the 1WBC occasionallydetrained floats SEUC; Intermediate Western Boundary Current, IWBC; into the interiorocean. This indicatesa mechanismby which North Brazil Under Current, NBUC. For comparison,an in- intermediatewater of subtropicalorigin is injected into the sertbased on a schematicdiagram by Petersonand Stramma tropical regime. In the interior tropical ocean,box averaged [ 1991], showsthe near-surfaceflow with the southwardBrazil -1 Current (BC), the north-westward North Brazil Current velocities are small (< 1 cm s , grey arrows in Plate 2) and (NBC) andthe Tropicaland Subtropicalgyres. randomlyoriented. Here intermediatewater can be gradually dispersedand loseits distinctivephysical properties. Before flowing northward into the subtropicalgyre of the North Atlantic, the intermediatewater must passthrough the near 45øS 37øW. There the flow circulatesaround the quasi- western ends of swirl zonal equatorialjets [Schott et al., barotropicZapiola Eddy [Davis et al., 1996;Flood and Shor, 1998]. For a long time it has been questionedwhether the 1988] (Plate3), a stationaryanticyclonic (anticlockwise) fea- northwardequatorial crossing of intermediatewater occurs di- ture centredat 45øS 42øW. While approachingthe Mid-At- rectly along the westernboundary or via zonal excursionsin lanticRidge, the ACC and SAC becomekinematically indis- the equatorial current system [Richardsonand Schmitz, tinguishableand encompass-10 ø of latitude.North of this 1993]. Floats turn westward at the north-eastern comer of broadband, along -35øS, the quiescentcentre of the anticy- South America and feed into the North Brazil Under Current clonicSubtropical Gyre separatesthe eastwardflow from the (NBUC). However, insteadof a direct cross-equatorialcon- slowerwestward Subtropical Gyre Return Current. tinuationof this flow, Plate 2 suggeststhat the NBUC retro- The-400 km wide Return Current is predominantlyzonal, flects at 2øS into the eastward Southern Intermediate Counter centeredat -30øS (Plate 2, blue arrows),and flows westward Current [Schott et al., 1998] (SICC, red arrows south of at 3-5 cms '• ratheruniformly, with transportestimates of equator).Subsequently, the SICC is seento feed northward 12_+3Sv at intermediatedepth. Towards the west, the Return (see Fig.3, float #115 in Boebel et al. [1998] for a higher Currentprogressively reaches farther north. However, in con- resolutiondepiction of this process)into the EquatorialInter- trastto the northernbranch of the SubtropicalGym at the sea- mediate Current (blue trajectoryalong equator in Plate 1) -! surface(Plate 3, insert), which flows north-westwarddiago- which flows west at 5-10 cm s mean velocities. However, in nally acrossthe SouthAtlantic [Reid, 1994], the intermediate agreementwith hydrographicstudies [Schott et al., 1998], the depthflow remainssouth of theVit6ria-Trindade Ridge. float data suggestan intermittent,seasonally reversing flow of Reachingthe continentalslope above the SantosPlateau, a theEquatorial Intermediate Current [Boebel et al., 1998]. previouslyhypothesised bifurcation [Reid, 1994] is now re- When the intermediatewater onceagain approachesthe shelf vealednear 28øS,which we suggestbe named,,Santos Bifur- near 44øW, it retroflects into the Southern or Northern Inter- cation"due to its vicinity to the SantosPlateau. Here, the Re- mediate Counter Currents or continues into the North Atlantic 3332 BOEBEL ET AL.: IN•RMED•• CIRCULATION OF THE WESTERN SOUTH ATLANTIC via the shelf-trappedNBUC or in Noah Brazil Currentrings Boebel,O., C. Schmid,G. Podestti,and W. Zenk, Intermediatewater [Richardsonand Schmitz, 1993]. Although no float has di- in the Brazil-Malvinas ConfluenceZone: a Lagrangian View, rectly crossedthe equatornear the westemboundary we can- Journal of GeophysicalResearch, 104 (C9), in press,1999. Broecker,W.S., The greatocean conveyor, Oceanography, 4, 79-88, not exclude that intermediatewater does chose this path 1991. sometimesand thereforeshow an interruptedNBUC at inter- Davis, R.E., P.D. Killworth, and J.R. Blundell, Comparisonof mediatedepth in Plate 3. AutonomousLagrangian Circulation Explorer and Fine Resolution Antarctic Model results in the South Atlantic, Journal of Geo- physicalResearch, 101,855 - 884, 1996. Summary Flood,R.D., and A.N. Shor, Mud wavesin the ArgentineBasin and theirrelationship to regionalbottom circulation patterns, Deep-Sea Fresh and ventilatedintermediate water is injectedinto the Research A, 35, 943-971, 1988. South Atlantic circulation at the Brazil-Falkland Confluence Fu, L.-L., The generalcirculation and meridionalheat transport of the Zone where this water is mixed with older, recirculatedinter- subtropicalSouth Atlantic determined by inversemethod, Journal mediate water. The freshened intermediate water then flows of PhysicalOceanography 11, 1171 - 1193, 1981. Gordon,A.L., and C.L. Greengrove,Geostrophic calculations of the east with the South Atlantic Current [Peterson and Strarnrna, Brazil-Falklandconfluence, Deep-Sea Research A, 33, 573-585, 1991] (15 Sv). Downstream,strong cross frontal exchanges 1986. betweenthe subtropicaland subpolarcirculation are observed, Ollitrault, M., G. Loa•c, and C. Dumortier,MARVOR, a multicycle suggestingadditional injection sites of newly formedinterme- RAFOSfloat, Sea Technology, 35, 39 - 44, 1994. Peterson,R.G., C.S. Johnson,W. Krauss,and R.E. Davis, Lagrangian diate waterinto the subtropicalgyre. East of the Mid-Atlantic Measurements in the Malvinas Current, in The South Atlantic: Ridge and beyondour area of studypart of this flow band is presentand past circulation,edited by G. Wefer,W.H. Berger,G. believedto turn noah and west, recirculatinga large amount Siedler,and D. Webb, pp. 122 - 162, Springer-Verlag,Berlin- of intermediatewater within the subtropicalgyre. At its east- Heidelberg,1996. ernmostposition, the Atlantic intermediatewater is subjectto Peterson,R.G., and L. Stramma,Upper-level circulation in the South AtlanticOcean, Progress in Oceanography,26, 1 - 73, 1991. influx from the Indian Ocean, which possiblymodifies the Reid, J.L., On the total geostrophiccirculation of the North Atlantic heatand saltbudget of the intermediatelayer water. Ocean:Flow patterns,tracers and transports,Progress in Ocean- The intermediateflow re-entersthe studyarea with the Sub- ography,33, 1 - 92, 1994. tropicalGyre ReturnCurrent near 30øS (12 Sv) which bifur- Richardson,P.L., and W.J. Schmitz,Deep cross-equatorialflow in the catesat the westernboundary with 9 Sv goingsouth and 4 Sv Atlantic measuredwith SOFAR floats,Journal of GeophysicalRe- search, 98, 8371 - 8388, 1993. going noah. The resultingnoahward flow betweenthe ,,San- Roemmich,D., The balanceof geostrophicand Ekman transportsin tos Bifurcation"and the equatoris within a narrowswift con- the tropicalAtlantic Ocean, Journal of PhysicalOceanography, tinuousintermediate western boundary current. This IWBC 13, 1534- 1539, 1983. flows counterto the overlyingsouthward Brazil Currentand is Rossby,T., D. Dorson,and J. Fontaine,The RAFOS System,Journal particularlyrelevant for the calculationand modellingof the of Atmosphericand OceanicTechnology, 3, 672-679, 1986. Schmid, C., Die Zirkulation des AntarktischenZwischenwassers im anomalous noahward heat flux found in the South Atlantic Stidatlantik,Ph.D. thesis,University of Kiel, Kiel, 1998. since it introduces a shallow level of no motion near 750 m Schott,F.A., J. Fischer,and L. Stramma,Transports and pathwaysof within the Brazil Current.Whether the sluggishoffshore flow the upper-layercirculation in the westerntropical Atlantic, Journal field of the Tropicsprovides an additionalnoahward flux of of PhysicalOceanography, 28, 1904-1928, 1998. AAIW, as suggestedby hydrographicstudies, has yet to be Taft, B.A., Distributionof salinityand dissolvedoxygen on surfaces of uniformpotential specific volume in the SouthAtlantic, South examined.Further studies are alsoneeded in the equatorialre- Pacific, and Indian oceans,Journal of Marine Research,21 (2), gion, where strongzonal excursionsand seasonalvariability 129-146, 1963. resultin a complexflow field, whichfinally guidesintermedi- Wrist, G., Schichtungund Zirkulationdes AtlantischenOzeans. Das ate water acrossthe equatorinto the Noah Atlantic. Bodenwasserund die Stratosphaere.,in WissenschaftlicheEr- gebnisseder DeutschenAtlantischen Expedition auf dem For- schungs-und VermessungsschiffMETEOR 1925-27, pp. 1-288, Acknowledgements.Our thanksare extended to all crewsand of- Walter de Gruyter und Co., Berlin, 1935. ricerswho significantlycontributed to the successof theseresearch ef- forts.The ALACE and SOFAR work was supportedby the National ScienceFoundation, USA. The MARVOR projectwas supportedby Ifremer, France. The RAFOS work was supportedby the Deutsche O. Boebel,Graduate School of Oceanography,University of Rhode Forschungsgemeinschaft,Germany and the Ministeriumfiir Bildung, Island,Narragansett, Rhode Island 02882 ([email protected]). Wissenschaft,Forschung und Technologie,Germany. Olaf Boebelis R.E. Davis and R.G. Peterson,Scripps Institution of Oceanogra- appreciativeof his currentsupport by the Alexandervon Humboldt- phy, La Jolla,California 92093. Foundation,Germany. M. Ollitrault,Laboratoire de Physiquedes Octans, Ifremer Centre de Brest, BP 70, 29280, Plouzant, France. P.L. Richardson,Woods Hole OceanographicInstitution, Woods References Hole, Massachusetts02543. C.Schmid, NOAA/AOML, Physical OceanographyDepartment, Boebel, O., C. Schmid, and W. Zenk, Flow and recirculation of Ant- 4301 RickenbackerCauseway, Miami, Florida33149. arctic IntermediateWater acrossthe Rio GrandeRise, Journal of W. Zenk, Institut ftir Meereskunde an der Universit•it Kiel, GeophysicalResearch, 102, 20967 - 20986, 1997. DtisternbrookerWeg 20, 24105 Kiel, Germany. Boebel, O., C. Schmid, and W. Zenk, Kinematic elementsof Antarctic IntermediateWater in the westernSouth Atlantic, Deep-Sea Re- search H, 46, 355-392, 1998. (ReceivedMay 25, 1999; acceptedAugust 16, 1999)