The warm water inflow into the western tropical Atlantic boundary regime , spring 1994

During March 1994 a survey of the western boundary of the tropical Atlantic, between JO"N and JO'S, was carried out by conductivity-temperature-depth and current profiling using shipboard and lowered acoustic Doppler current profilers. In the near­ surface layer, above ac-l = 24.5, the inflo\v into the boundary regime came dominantly from low latitudes; out of the 14 Sv that crossed the equator in the upper part of the North Brazil Current (NBC), only 2 Sv originated from south of 5°S, while 12 Sv came in from the east at [-55 with the South Equatorial Current (SEC). After crossing the equator near 44W, only a minor fraction of the near-surface NBC retroflected eastward, \\·hile a net through fl.o\v of about 12 Sv above a<-l = 24.5 continued northwest\vard along the boundary. By contrast, in the isopycnal range u, _ � 24.5-26.8 encompassing the Equatorial Undercurrent (EUC), the source waters of the equatorial circulation were dominantly of higher-latitude South Atlantic origin. While only 3 Sv of eastern equatorial w·ater entered the region through the SEC at 3°-5 5, there was an inflow of 10 Sv of South Atlantic water�in the No�th Brazil Undercurrent (NBUC) along the South American coast that originated south of l0S. The transport of 14 Sv arriving at the equator along the boundary in the undercurrent layer \vas almost entirely retroflected into the EUC with only marginal northern water additions along its path to 35 W. The off­ equatorial undercurrents in the upper thermocline, the South and North Equatorial Cndercurrents carried only small transports across 35W, of 5 Sv and 3 Sv, respectively, dominantly supplied out of SEC recirculation rather than out of the boundary current. Still deeper, three zonal undercurrents ,vcre observed: the ,vest\vard-fl.owing Equatorial Intermediate Current (EiC) in the depth range 200-900 m below the EUC, and two off ­ equatorial cast\vard undercurrents, the Northern and Southern Intermediate Countercurrents (NICC, SICC) at 400-1000 m and ] -3 latitude. In the lower part of the '-:BUC there was an Antarctic Intermediate Water (AAIW) inflow along the coast of 6 Sv, and there was a clear connection at the AAIW level to the SICC by low salinities and high o:xygens and a ,vcaker suggestion also that some supply of the NJCC might be through AAIW out of the deep NBUC.


I, Introduction
The near-surface circulation of the \vestern tropical Atlantic, -;outh of the equator, is characterized by broad \\'eSh\•ard inflow through the South Equatorial Current (SEC) that carries \\•ater masses \\•ith subsurface salinity and oxygen values resembling those of the eastern South Atlantic [ Reid, 1989: Tsuch ( ra,  19S6J.Near the boundary.this flo\v splits up into a northerly component that supplies the cross-equatorial v, :estcrn bound ary current.the North Brazil Current (NBC) and into a south erly n:circulation as part of the antic y clonic subtropical South ,\tlantic gyrc.South of the equator, l\vo east\\•ard flowing zonal currents ha\•e been ohscn•ed.the South Equatorial Undercur rent (SEUC) in the latitude range 3 °-5 v S. for \\-•hich Cochrane  et al. [1979J reported a transport of 15 Sv eash\•ard in the longitude range 25"-33 °\V, and a South Equatorial Counter current (SECC) in the latitude range 7 '°-9 °S. for \,-hich.ho\\' cver.only sparse evidence has hcen produced so far for the \\CStern tropical South Atlantic [e.g., .\lo!i11ari. 1982: Stranuna, 1991].
Cop�right 1995 by the American Gcophy�ical Union.
Paper numhcr 95JC02XOJ.(l\--tS-1)227 •95 95JC-(l2:,.\IJ3SO.:'i.1Kl As regards the seasonal signal south of the equator, ship drift currents [Richardson and Walsh, 1986;Anuudt. 1987].evaluation of the historical expendable bathythermograph (XBT) data file [.\fo!inari and Johns, 1994].and satellite altim ctry [ e.g .. I<o111a11ees. ,en. 1993] suggest that it is small compared \vith the region of the North Equatorial Countercurrent (NECC) regime.After an evaluation of the high-resolution \Vorld Ocean Circulation Experiment Community I\fodeling Eff ort (\VOCE Ct\1E) model of the North Atlantic revealing that the north\vard flo\ving boundary current off North Brazil, bet\\Ten 5 °S and lO c S. \\•as in fact an equatonvard undercur rent [ Schott and Bbning, 1991 ]. recent obsen'ational programs confirmed that undercurrent structure from direct velocity pro files I Stranuna et al., 1995]; it also can be seen in geostrophic results of older Nanscn hottlc sections [ Sill•eira et al., 1994).The North Brazil Undercurrent (NBUC) \Vas found to be quasi-continuous along the coast.bct¥;een lO"S and s�s, v,: ith vel)' little inflo\v out of the SEC in that latitude range.I lo\\-' ever.after passing c•ape San Roque at 5 °S. the SEC overrides the NBUC, causing its structure to change from an undercur rent to a surface-intensified current, the North Brazil Current.
The NBC, after crossing the equator, partially continues northv.•est\vard, partially rctroflccts into the Equatorial Under-current (EUC) and the NECC; part of it may also end up in the North Equatorial Undercurrent (NEUC).How this partition works and what kind of seasonality it has are still to be deter mined.The NECC is heavily supplied via cross-equatorial in flow during July to December or January, expressed by a strong surface current retroflection near 3 °-5 °N [Johns et al., 1990;Didden and Schott, 1992], while in spring to early summer the NECC is weak or even flowing westward [ Richardson and Reverdin, 1987].The NECC can also be supplied from the north, as was determined by Wilson et al. [1994] for August 1989, when significant transports out of the NEC turned south ward near the western boundary and joined the NECC.Part of this NEC recirculation even transited through the eastern Car ibbean.This southwestward NEC recirculation was also repro duced in the CME model [Schott and Boning, 1991].
Water mass transfer out of the retroflection zone partially occurs through eddy shedding, and detached anticyclonic rings have been observed to propagate along the South American coast toward the entrances of the Caribbean by surface color measurements [Muller-Karger et al., 1988;Johns et al., 1990], by trapped drifters [Richardson et al., 1994], and in Geosat altim etry [Didden and Schott, 1993].The weakness and even reversal of the NECC in spring, combined with only weak seasonality of the SEC inflow, was presumed to result in increased north westward transport out of the NBC toward the Caribbean in spring [Richardson and Philander, 1987], but so far, evidence for this transport cycle could not be produced [e.g., Didden and Schott, 1992].
At intermediate levels, Antarctic Intermediate Water (AAIW), identifiable by its oxygen maximum and salinity min imum, spreads northward along the western boundary [Tsuch iya, 1989], involving complex pathways in the western tropical Atlantic, as evidenced by the 800-m-level float tracks of Rich ardson and Schmitz [1993].Below, there arc the northern ex tensions of upper Circumpolar Deep Water (uCDW), charac terized by nutrient maxima [Reid, 1989].The deep reaching NBUC is one of the current branches affecting water mass redistribution down to these levels.
In the Pacifi c, three zonal intermediate-depth currents have been documented in the equatorial regime [Firing, 1987;Go11-ri011 and Toole, 1993]: the Equatorial Intermediate Current (EiC), flowing westward underneath the EUC, at 300-to 700-m depth, and two off-equatorial eastward intermediate currents, the Southern and Northern Intermediate Counter currents (SICC, NlCC), at 1.5 °-2 ° latitude and 400-to 1500-m depth.In the Atlantic, evidence in the past has been produced for the existence of the EiC [Stalcup and Metcalf, 1966;Hisard and Rua[, 1970]; as for the off-equatorial intermediate coun tcrcurrents, it is still an open question whether they also exist in the Atlantic and, if so, what role they play in the AAIW circulation.
Quantitatively, not very much is known about these diff erent branches of the near-boundary circulation because geostrophy degenerates in the near-equatorial zone and direct current observations were not available until recently.The transport of the NBC at the equator was determined by Schott et al. [1993] with a moored acoustic Doppler current profiler (ADCP) and current meter array near 44 °W.It amounted to an annual mean of 24 Sv for the upper 300 m with an annual cycle of only 3 Sv amplitude.Ship sections of the region show that a large fraction of this transport retroflects into the EUC; hence only part of it is water mass gain for the northern hemisphere.Ship profiling measurements from Meteor cruise M22/2 from No-vember 1992 across the NBUC at 5 °S and l0 °S were evaluated by Stramma et al. [1995].The transport of the upper 1000 m was 22 Sv with no net inflow joining the NBUC between lO"S and 5 °S, indicating a source supply even farther south along the coast.Silveira et al. (1994] computed an average geostrophic transport of the equatorward boundary current of 21 Sv for May-July 1986. A survey of the NB UC-NBC system and the near-equatorial zonal currents during a different season, in March 1994, that is.when the NECC is at its minimum or flowing westward, was carried out on Meteor cruise M27/3 (Figure 1).Currents and transports were directly measured by ADCP profiling along several sections.This paper presents currents, transports, and water mass distributions from these measurements, and it at tempts to determine connections among the different current branches and an assessment of pathways of the major water masses involved for this particular season.Some comparisons will be made with recent findings in the western Pacific because significant similarities of current patterns among the two oceans are found.

Observations
The tropical survey began on February 26 at 12 °N, continued along the 44 °W section (Figure 1 ), and terminated that section on March 5 at the equator.The 40 °W section was done during March 7-11 and the 35 °W section during March 13-17; finally.the measurements around the southern box, between 5 °S and l0 °S, were carried out during March 18-25.The conductivity temperature-depth (CTD) system used was Neil Brown Mark III with bottle calibrations.The pressure and temperature sen sors were calibrated in the laboratory before the expedition.and these calibrations were verified by reversing thermometer temperature measurements and thermometric pressure mea surements.The conductivity cell was calibrated from salinity bottle samples, as the cell showed some pressure dependence and a slight time dependence.The resulting salinity accuracy is :'::0.003.The oxygen sensor was calibrated from bottle oxygen samples analyzed by a modified Winkler titration.The accuracy of the calibrated sensor data is estimated at :'::0.07 mL L-1 • Current profiling was carried out by undcrway shipboard ADCP and by lowered ADCP (LADCP) on stations.mounted to the CTD/rosette (Figure 1 ).The depth range of the ship board 153.6-kHzADCP of the Meteor was much improwd compared with prior cruises because a protective cover over the transducers had been removed in the shipyard just prior to the cruise.Hence in fair weather, ranges of 350-400 m were reached routinely where the uppermost reliable measurement was from 25-m depth.The vertical resolution was 8 m.Ap proximately 250 individual profiles were averaged into 5-min ensembles for storage.
In order to remove the contributions of the ship's motion.the ADCP measurements have to be corrected for any trans ducer-gyrocompass misalignment before subtracting speed over ground (from Global Positioning System (GPS) position ing).A heading calibration similar to that of Pollard and Read [1990] was applied.For this procedure it is assumed that the vertically average current at a CTD station (at low ship speeds) is the same as that just before arriving at or after leaving the station.Any difference in the mean currents should then come from a deviation of the transducer axis relative to the gyro compass.By rotating the currents so that the differences between on-station and off-station currents are minimized, a mean mis alignment angle (or heading bias) is determined.Usually, all stations of a cruise are used to obtain a best estimate of the heading bias unless the transducer orientation changes during the cruise.During M27, however, a much larger error occurred due to a failure in the ship's gyrosystem.The automatic speed error compensation unit was not functioning, resulting in course and speed dependent heading deviations of up to 5 ° between reversing (north to south) section orientations.Along the individual sections of the cruise the deviation was fairly constant and was not expected to have an effect on the quality of the raw data.Therefore we decided to calibrate each section separately for the combined effect of transducer misalignment and gyro error.
The accuracy of the heading calibration depends crucially on the number of stations of the sections; it was smallest (-0.2 °) along the major sections and about twice as large along the connecting lines, that is, along 4 °30'N and 30 °\V (Figure I).On the transit leg along the shelf between 44 °\V and 40 °\V, no calibration could be performed, and the data were omitted.After the heading calibration, speed over ground from the ship's GPS system was subtracted to obtain absolute currents, which then were smoothed by a two-dimensional Gaussian filter (cutoff scales were 0.2 ° horizontally and 20 m vertically) and regridded to 0.1 ° (-10 km) resolution.The estimated error of the end product has two constituents, a bias arising from any residual heading misalignment of -2-4 cm s-1 , and a stochastic error (order of 5 cm s-1 ) resulting from GPS inaccuracies and from noise in the ADCP data.While the latter will be reduced by integration along sections, the bias will introduce systematic errurs in the transport calculations.
For the LADCP, two 153.6-kHzsystems were available, one narrowband (NB) and one broadband (BB).The BB system showed a temperature dependent failure that had already oc curred on a previous cruise but had been presumably corrected by the manufacturer.Unfortunately, along the 40 °\V section, the NB also failed for some stations but it was back in opera tion for the rest of the cruise.The processing of the LADCP data and derivation of absolute profiles of north and cast components were carried out as described by Fischer and Vis beck [1993].LADCP errors of the order of 5 cm s I are a combination of noise in the ADCP measurements, of OPS position inaccura cies, and of internal wave noise [Send, 1994].Because these errors are random, their effect on transport calculations will be reduced according to the number of stations used for the integration.

Water Mass Distributions
The differences between water masses of southern origin that are introduced into the region via the NBUC from the south and those from the northern hemisphere are very obvi ous in their salinities and oxygens.A potential temperature salinity (0-S) diagram (Figure 2a) shows the typical linear relation in the thermocline range representing the Central Water, which is of slightly higher salinity for the northern hemisphere [e.g., Emery and Meineke, 1986].The salinity max imum water is formed as high-salinity surface water in the subtropics and progresses equatorward as a subsurface salinity maximum, while the water above is salinity-poor owing to the high precipitation in the tropics.In the study region the salinity maximum is located at the top of the thermocline, at densities slightly below u < •> = 25.0, and it reaches salinities above 37.0 at 5 °-10 °S while staying below 36.7 north of the equator (Figure 2a).At the lower end of the warm water sphere, near u� = 27.3, the Antarctic Intermediate Water shows up as a salinity minimum (Figure 2a).
While salinity differences between both regions are small in the linear range of the 0-S diagram, the oxygen differences in the salinity range 35-36 are striking (Figure 2b): at 5 °-10 °S near the coast the oxygen values exceed 4 mL L -1 , while they stay below 3 mL L -1 in the thermocline layers of the north western boundary regime.The equatorial water flowing in from the east with the SEC is also characterized by lower oxygen than southern water in the Central Water range.The maximum at 6 mL Li is associated with the North Atlantic Deep Water (NADW) (Figure 2b).
The water mass distributions will be described here by prop erty distributions on three core isopycnal surfaces.They are marked in the 0-S diagram and are also inserted into a com posite plot of zonal currents with salinity along 35 °W (Figure 3a) and oxygen (Figure 3b).The surface u 0 = 25.0, at a mean depth of 100 ::!:: 19 m, lies just underneath the surface mixed layer in the northern part of the survey area and shows the distributions for the near-surface circulation and salinity max• imum water; it also passes through the velocity core of the EUC (Figure 3a).The surface <re" = 26.7, at a mean depth of 216 ::!:: 51 m, runs through the cores of both off-equatorial undercurrents and passes underneath the EUC (Figure 3).Finally, the <re-> = 27.28 surface, at a mean depth of 747 ::!:: 42 m, runs through the low-salinity core of the AAIW (Figure 3).At that level the NBUC is still quite vigorous in the equator ward direction, and there are three zonal undercurrents: the Equatorial Intermediate Current (EiC) and two off-equatorial currents, the Northern and Southern Intermediate Counter currents (NICC, SICC) at 2 °-3 °N and I .5 °-2.5 °S, respectively (Figure 3).
On each of the three selected surfaces.we present the dis tributions of salinity and oxygen as well as a classification by water mass groups.The latter are superimposed on current vector maps derived from the shipboard ADCP for the upper two isopycnals and from LADCP for the intermediate layer.
For <re" = 25.0,these distributions are shown in Figure 4. Beginning with the distributions south of the equator, the salinity maximum water spreads northward along the South American coast within the NBUC between 10 °S and 5 °S, and it reaches 44 °W near the southern boundary via the NBC with a core maximum still at 36.87 (Figure 4a).The highest oxygen values do not enter the region along the coast at J0 °S but come from offshore and also extend along the coast to 44 °W.It is interesting to note that the highest oxygen values in the near shelf core are located offshore of the salinity maximum, sug gesting two veins of boundary supply.The secondary O>-')'gen maximum of >4.5 mL L I extends southeastward from 2 ° at 44 °W (Figure 4b).From the northwest, along 3 °-4 ° , a wedge of salinities >36.5 protrudes into the region across 44 °W but this wedge is located north of the offshore oxygen maximum (Figure 4b).
In the near-surface circulation determined by shipboard ADCP (Figure 4c), the NBUC, which is weak at this near surface level, flows northward along the coast between 10 ° and 5 °S, while the SEC enters the region across the 35 °W section between 5 °S and the equator.The joint current flows westward along the southern boundary, then crosses the 44 °W section and suggests a relroflection at 2 °-4 °N along that section.Al though we cannot definitely conclude on the existence of that retrofleclion from the data at that level, we know from other work in progress that il exist west of 44 °W.That retroflection connects with eastward flows across 40 °W just north of the equator and between l.5 °S and 2 °N across the 35 °W section.This is the upper part of the EUC (Figure 3).North of the EUC, the northern SEC branch supplies westward surface flow, including the regime normally occupied by eastward flow ing ECC waters.
As regards the classification into water mass classes by their oxygens and salinities, one has to keep in mind that in the northern part of the region the a <-> = 25.0 surface is only 10 -20 m below the surface mixed layer; hence properties al this level might nol be conserved everywhere.Two dominant water mass characteristics can be distinguished at this level: those from the equatorial belt and northwestern regime, and those from the southern hemisphere.If the water masses are marked as south ern (Figures 4a and 4b) by their high oxygen (>4.5 mL L-1 ) and high salinities (>36.8), and as equatorial and northern by relatively low oxygen ( <4.3 mL L -1 ) and salinities ( <36.7), the class distribution as in Figure 4c results.From this classifica tion the southern inflow at a 0 = 25.0 originates not only from along the coast but also from the east between 5 °S and l0 °S (Figure 4c).Further, the boundary flow seems to continue to the west of 44 °W because water masses of the retroflection differ from those along the boundary.Retroflection water at this level has mixed-water characteristics.
The surface a 0 = 26.7 passes underneath the NBUC core at 250-m depth off Brazil, underneath the EUC at 200-m depth in the 35 °W section, and through the core of the off-equatorial undercurrents (Figure 3).Its mean depth is 216 ± 51 m.At this level, southern water is marked by low salinity and high oxygen (Figure 2).This low-salinity (Figure 5a) and high-oxygen (Fig ure 5b) water spreads northward along the western boundary, and via the retroflection into ea tward flow at 1.5 °-3.5 ° , which at this level and time of observation connects with the NEUC, while at the axis of the EUC at 40 ° and 35 °W we observed weak westward currents (Figure 5c).The boundary Figure Sc shows the SEUC current core in the 35 °W section at 3 °-4 °S, but an association of the SEUC at 35 °W with bound ary water mass properties is not possible; the SEUC core oxygen at 35 °W is only at 3.0 mL L 1 , compared with >4.0 mL L I in the NBUC (Figure Sb).This suggests that the SEUC is not supplied via a retroflection out of the BUC, but rather is supplied by recirculation of interior water masses coming in with the lower part of the SEC.
At the a f-1 = 27.28 level, salinities <34.45 (Figure 6a), ac companied by oxygen values >4.2 mL L 1 (Figure 6b) mark the propagation of AAIW along the coast around Cape San Roque.Along 2 °S, an eastward extension of salinities <34.45 across the 35 °W meridian is observed.This alinity minimum along 2 °S is accompanied by an oxygen maximum (Figure 6b).This property tongue is associated with the southern hemi spheric eastward current core (Figures 3 and 6c), suggesting eastward AAIW leakage out of the boundary current south of •.:

North Brazil Undercurrent and North Brazil Current
The currents of the upper 1200 m, as measured by shipboard ADCP in the top 350 m and augmented by LADCP for the deeper levels, are shown in Figure 7, where LADCP data are only available for four of the five sections presented.Transport numbers for individual clearly identifiable current branches are inserted and listed in Table L For upper layer transports, the near-surface currents above 25 m were estimated by upward shear extrapolation.The isopycnals <T e = 24.5 and <T e = 26.8,which serve as layer boundaries for budget calculations later, and <T tt = 27.28 which is the boundary for as the AAJW core, are also shown.In the fol lowing the individual current branches are discussed, and composite pictures for the three layers will be attempted.
At 10 °S the NBUC current maximum with core speeds >0.7 m s-1 (north component >0.5 m s -1 ) lies at 250 -350 m (Fig ure 7a), and at the AAIW level, speeds are still 0.2 m s I (Figure 7c).Hence a large transport fraction occurs below the shipboard ADCP range, and Figure 7a is augmented by LADCP measurements below 350 m.In the near-surface layer, above <T e = 24.5,only 0.6 Sv flow northward, compensated by an offshore southward recirculation, and in the <T e = 24.5-26.8layer, which encompasses the current core, the transport amounts to 10.5 Sv.Below u 8 = 26.8there is still northward transport of 6.6 Sv near the boundary, yielding a total of 17.1 Sv (Table l).I lowevcr, cast of 35 °W there is a second branch of northward flow of 5.5 Sv below u <-> = 26.8, and between . ,, .' .. At 5 °S the BUC transport above cr tt = 24.5 has increased against the l0 °S value, to 2.2 Sv (Figure 7b), but in the cr tt = 24.5-26.8range there is nearly the same transport as at lO"S: and between cr 6 = 26.8 and 1000 m the transport is also  [1991].The winds off Brazil were from the southeast during the survey, and the corresponding Ekman transports were 0.8 Sv and 0.7 Sv south ward across the 5 °S and 10 °S sections, respectively.North of Cape San Roque, the westward transport increases considerably through the addition of the SEC flow.As can be seen in Figure 7c, a clear distinction between the NBUC flow arriving from the south and the SEC inflow is not possible.Taking the limit at 3.5 °S yields a total of 18.2 Sv above cr 8 = 26.8,adding 5.9 Sv to what arrived above ere = 26.8along the coast at 5 °S.Despite the SEC superposition, the boundary current at 35 °W and 40 °W (Figure 7d) maintained a subsurface maximum, while at 44 °W the clear undercurrent character was lost (Figure 7e), in agreement with the moored ADCP obser vations of upper layer flow at that site by Schott et al. [ 1993].
Across 40 °W the westward transport south of l 0 N above CT 9 = 26.8amounts to 27.1 Sv, of which the flow above cr 8 = 24.5 was 10.5 Sv, and in the ere = 24.5-26.8layer the flow associated 8.4 8.4 with the boundary current core was 16.6 Sv westward, but an additional 3.2 Sv flowed westward underneath the southern part of the EUC (Figure 7d).The total flow could not be evaluated for this section because the LADCP was not func tioning along 40 °W, thus limiting the measurement range to that of the shipboard ADCP.Across 44 °W, transport is 13.6 Sv along the boundary and south of 2 °N in the layer above cr 0 = 24.5, and 14.4 Sv in the layer cr 0 = 24.5-26.8(Figure 7e), indicating a near-continuity of the boundary flow above <T e = 26.8 between 40 °W and 44 °W but with nearly compensating increase above and decrease below ere = 24.5.There is still a large transport across 44 °W between ere = 26.8 and the 1000-m level, amounting to 14.2 Sv (Figure 7e), yielding a total NBC boundary flow above 1000 m of 42.2 Sv.
North of 1.5 °N, the flow is eastward up to 4 °N, with a sub surface maximum at 200 m (Figure 7e).This subsurface max imum also occurred in the Wilson et al. [1994] observations.While they attributed it to the NEUC for the boreal summer situation, it is dominantly a source of the EUC in our case.The eastward flow across 44 °W in the cr 0 < 24.5 layer amounts to only 3.9 Sv, leaving a net northwestward through flow of 9.7 Sv along the boundary in that layer.In terms of water mass trans fer, this exchange might actually be larger, if the eastward return flow were not all retroflected NBC water, but partially northern water.By contrast, there is an exact match in the 24.5-26.8layer between the 14.4 Sv of NBC fl ow and the eastward return flow.Below, between a-0 = 26.8 and 1000 m, there is large eastward fl ow between 1.5 ° and 3.5 °N, totalling 27.8 Sv, followed by compensating westward flow of 27.7 Sv to the north, in the 3.5 °-5.5 °N latitude range.
The flow budgets of the NBC and the offshore reversals have to be seen within the context of the large temporal variations observed within the NBC at 44 °W, where Schott et al. [1993] determined variations of upper layer transports of 50% against the mean at periods of a few weeks.

The Zonal Near-Surface Currents
As was shown for the NBUC, there is very little addition to the equatorward boundary current in the 5 °-10 °S latitude band, amounting to less than 2 Sv.North of 5 °S, westward surface flow was observed on both sides of the Equatorial Undercurrent (Figures 4c and 7c).Between 5 °S and l 0 S the westward flow above a-0 = 24.5 was 13.1 Sv and 20.6 Sv for the a-0 = 24.5-26.8range.Without the NBUC transport of 10.5 Sv above a-0 = 26.8 that arrives from the l0 °S section, the net SEC transport across 35 °W south of the equator is 23.2 Sv; that is, 32% of the westward flow is of NBUC origin and 68% from the tropical interior South Atlantic.Above a-0 = 24.5, however the coastal fraction is near zero, because the 2.2 Sv crossing 5 °S were already SEC inflow.
At 40 °W the southern SEC is even less distinguishable from the coastal NBUC inflow than at 35 °W, except for a separate core of 3.2 Sv underneath the southern part of the retroflec tion.In terms of composition of the NBC at 44 °W, 70% of it is of southern coastal origin in the isopycnal range 24.5-26.8,and of the 13.6 Sv above a-0 = 24.5, 84% is of low-latitude ( <5 °S) origin.
A southern hemispheric zonal current reported earlier by Molinari [1982] was the existence of a South Equatorial Coun tercurrent at 25 ° to 30 °W, a band of eastward flow between 7 ° and 9 °S.As can be seen from the current vector map in Figure 4c as well as from the transport budgets, such a current was not observed during the March 1994 survey west of 30 °W.
Across the 35 °W section, between l.5 °N and 4.5 °N, west ward SEC transports of 4.0 Sv were determined above <Te = 24 .5 and 3.9 Sv at a-0 = 24.5-26.8,within which the NEUC is embedded (Figure 7c).In northern summer the northern part of that section would show vigorous eastward surface currents as part of the North Equatorial Countercurrent, but in March the NECC is weak or even reversed, as was found in our observations.In the 40 °W section the northern SEC is found north of 2.5 °N, transporting about half of the 35 °W volume above <Te = 26.8.There is a net outflow to the north across the 4 °30'N section, amounting to 5.7 Sv above a-e = 26.8, that is, larger than the difference between both zonal transports.At 44 °W, the westward SEC-NECC transport amounts to 12.2 Sv above a-0 = 26.8south of 7 °30'N (Figure 7e).
At 44 °W the zonal SEC and NECC westward inflow plus the excess of the NBC over the retroflection return flow lead to a net westward flow between the coast and 7 °N of 16.7 Sv above a-0 = 24.5 and 5.2 Sv in the second layer.

The Equatorial Undercurrents
The survey fell into the time period of near-equatorial loca tion of the intertropical convergence zone.During the passage along the 35 °W section, winds on the equator were unsteady, though generally eastward.As a consequence, the EUC in spring is nearer to the surface than in fall in the western Atlantic, and eastward winds at times cause an eastward sur face jet [e.g., Richardson and Reverdin, 1987;Schott and Bon ing, 1991] similar to the semiannual jet in the Indian Ocean.
During M27, equatorial currents along 35 °W were eastward between the surface and 200 m with a total transport of 21.4 Sv.At this longitude, the core of the EUC was located north of the equator owing to its northern hemispheric retroflection source (Figures 4c and 7c).At 40 °W, the equatorial eastward flow had a complex shape with two eastward cores, centered at l.2 °N and 2.5 °N, respectively.The undercurrent transport below a-0 = 24.5 of about 14.5 Sv is the same at 44 °W and 40 °W and only slightly lower, at 13.7 Sv, 550 km farther east, when that flow crosses 35 °W.Overall, based only on the transports, the EUC supply in its core layer is totally out of the southern hemisphere via the NBC retroflection at the western equato rial end, as was already suggested by Metcalf and Stalcup [1967], on the basis of oxygen distributions alone.
Underneath the EUC the presence of the westward flowing Equatorial Intermediate Current was established in repeat sec tions across the equatorial Pacific.A band of westward flow is also obvious at 35 °W in our observations (Figure 7c).Evidence of the EiC in the Atlantic exists at least since 1963, when Stalcup and Metcalf [1966] observed steady westward currents below the EUC.The measurements were obtained from short duration (6-60 days) current meter moorings near the equator at 35 °W and farther east.While this flow was not called EiC, the authors noted "that, at 405 m depth, the westward current appears to be concentrated at the Equator."The name EiC was used by Hisard and Rua/ [1970] for the description of the observations in the Atlantic by Stalcup and Metcalf [1966].In our observation, the EiC transport between 1 °s and 2 °N and for the depth range 200-1000 m, is quite large, at 20.4 Sv.In the Pacific, at 165 °W, mean EiC transports were only 4.5 Sv [ Gouriou and Toole, 1993].Hence our individual estimate may just be one view of a time-varying deep current system on the equator that shows strong variability at periods of weeks to months as found in the moored deep records of Schott et al. [1993] and as also seen in recently retrieved current meter records from a station deployed on the equator at 37"W.

The Off-Equatorial Undercurrents
The core of the SEUC, between 3 ° and 4 °S and at depths of 150-600 m, and that of the NEUC, between 3 ° and 4 °N and at 90-350 m, are recognizable in the 35 °W section (Figure 7c).Transport of the SEUC was 4.9 Sv, with only 1.7 Sv above <Te = 26.8, and for the NEUC the total transport was 2.7 Sv, of which 1.8 Sv fell into the <Te < 26.8 range.However, it bas to be noted that LADCP station spacing below 300 m was not adequate for these narrow currents.While isopycnal maps of Ts uchiya [1986] for oxygen and salinity show maxima extending eastward at the latitudes of the SEUC and NEUC, connecting their water mass properties to the NBC, such connection is not obvious from our measurements.In fact, the oxygen and sa linity at the SEUC core resemble much more the interior tropical environment than the high values of >4 mL L -• and <35.0 typical for the NBUC (Figures 3 and 5).This suggests supply through recirculation rather than via NBUC retroflec tion.In the southern part of the 40 °W section a small core near a-e = 26.8 can be identified in Figure 7d, embedded in the westward SEC flow.This suggests the root of the SEUC to be as far west as 40 °W.For the NEUC, partial supply could split off from the north ern part of the NBC retroflection, as water masses show south-em properties at some observation points in the western NEUC region (Figure Sc).However, the NEUC at 35 °W ap pears more dominated by low-oxygen equatorial waters (Fig ure 3b) than southern waters.
At intermediate depths, the Northern and Southern Inter mediate Countercurrents are seen at 35 °W in the depth range 400-900 m (Figure 7c).The SICC between 1.2 and 2.8 °S car ries 11.2 Sv, while the NICC, between 1.5 ° and 3.0 °N, trans ports 8.4 Sv.For comparison, in the Pacific at 159 °W the SICC mean was determined to 8.7 Sv, and the NICC mean was determined to be 3.0 Sv [Firing, 1987].
As can be seen in the water mass distribution along 35 °W (Figure 3) and the distributions on cr 0 = 27.28, the SICC is characterized by high oxygen values and low salinities, relating it to AAIW waters in the deeper part of the NBUC.At the location of the NICC axis, by contrast, salinities are higher and oxygen values are lower than in the NBUC, indicating that its supply is dominantly of northern origin and out of recirculation of the EIC, but admixtures of NBUC water are suggested by the meridional property gradients across the NICC (Figures 3  and 6).
The reconciliation of the deep zonal transports with those crossing 44 °W is not straightforward.At 44 °W the deep east ward flow out of the retroflection at l.5 °-3.5 °N carries 27.8 Sv below <T a = 26.8.If that supplies the NICC in about the same latitude range, there is 7 Sv of net eastward inflow into the 35 °-44 °W longitude range and south of 4 °N left from the west.An attempt to compose a consistent flow pattern on the basis of transports and water masses will be presented at the end of the paper.

Antarctic Intermediate Water Transport
The AAIW is clearly identifiable in the water mass distribu tions by its salinity minimum (Figures 3 and 6).In the currents the AAIW depth range is part of the downward decaying NBUC profiles, and there is no secondary velocity maximum associated with it in the sections at l0 °S (Figure 7a) or 5 °S (Figure 7b).Limiting the AAIW by the <Te = 26.8 and <.Te = 27.4 levels [e.g., Roemmich, 1983], which would put the lower limit near 1000-m depth, would yield an incoming transport into the equatorial regime of 6.0 Sv.Putting the upper level at <.T tt = 27.1, the boundary between the South Atlantic Central Water and AAIW, and the lower level at cr 1 = 32.15,following Stramma et al. [1995], results in a slightly larger number, 6.8 Sv, at lQ°S.As to the further fate of the AAIW, the LADCP data are too sparse to determine it, but there is a continuity of the deep NBUC and NBC from l0 °S to the equator along the boundary (Figure 7c), and salinity minimum water is observed in the Caribbean and Florida Current [Tsuchiya, 1986;Schmitz and Richardson, 1991].Part of the AAIW enters the SICC as indicated by the low-salinity and high-oxygen core at <.Te = 27.28 along 2 °S (Figure 6).The meridional gradients of both, salinity and oxygen across the latitude band occupied by the NICC (Figures 3 and 6) suggest admixtures of AAIW contri butions to waters of northern origin or out of the recirculation from the EiC.Both the SICC and NICC transports as well as the deep NBC along 44 °W are larger than the AAIW inflow at I Q°S; hence the AA1W can make up only a small fraction of the deep off-equatorial countercurrents.

Summary and Conclusions
In a survey with CTD and current-profiling methods during spring 1994, the boundary current between l0 °S and 10 °N of the western tropical Atlantic was investigated, and an attempt was made to budget the warm water routes that carry southern hemispheric waters toward the equator and of their coupling to the zonal equatorial circulation.
The vertical structure of the northward boundary current along the Brazilian coast between 5 ° and 10 °S is that of an undercurrent with the current core of 80 cm s-1 at about 200-m depth and small near-surface currents.This structure was also found in earlier surveys [Silveira et al., 1994;Stramma et al., 1995] and is also reproduced in a high-resolution numer ical model [Schou and Boning, 1991 ]. Part of the reason for the existence of this structure, which has its equivalent in the New Guinea Undercurrent in the western tropical Pacific [Lind strom et al., 1987;Ts uchiya et al., 1989], could be the fact that the Ekman transport in the western tropical region of both oceans is directed against the geostrophic flow, leading to the typical dynamics of an upwelling regime with a poleward sur face pressure gradient which is compensated by an equator ward subsurface pressure gradient that then causes an equa torward subsurface current.
The transport budgeting for the different current branches was carried out between isopycnal layers; for the near-surface layer above <.Te = 24.S, for the undercurrent layer in the isopy cnal range <.Te = 24.5-26.8,and for the Antarctic Intermediate Water layer between cr 0 = 26.8 and 1000 m, which corresponds roughly to cr 6 = 27.4 in the region.Despite careful calibration of the decomposition of the shipboard ADCP currents, signif icant errors can arise from small compass misalignment resid uals.This effect is of course most severe for the weaker off shore currents that require integration over long stretches of ship track.
In the equatorial regime there is a further difficulty for mean current and transport determination, the presence of an ener getic field of equatorial waves of periods ranging from weeks to months in the region [e.g., We isberg and We ingartner, 1988;Schott et al., 1993].It prohibits the application of equatorial geostrophy unless averaged over a sufficient number of sec tions and may also be responsible for a fair part of the trans port discrepancies found in our study.
In the near-surface layer the inflow stems dominantly from low latitudes (Figure 8a).At 10 °S, there is no net transport along the boundary, and between 10 ° and 5 °S only 2 Sv flow in from the east.However between 5 °S and 1 °S the SEC trans ports 11 Sv westward into the region.Hence 13 Sv leave the 35 °W section westward, and 14 Sv arrive at 44 °W.However, only 11 Sv cross the 40 °W meridian.While the decrease from 35 °W to 40 °W could be explained by recirculation into the shallow EUC, the increase from 40 °W to 44 °W is hard to explain; it cannot come from the eastward retroflection flow because there is an eastward increase of 2 Sv, not decrease in that branch between 44 °W and 40 °W.The total gain of the upper layer between 40 °W and 44 °W was 5 Sv, almost com pensated by a deficit of the same magnitude for the layer below; we come back to this point below.North of the EUC, transports are westward, with 4 Sv arriving across 35 °W, south of 4.5 °N.Farther north, another 3 Sv join to make the total transport 7 Sv that crosses 44 °W above cr 0 = 24.5 (Figure 8a).Hence in agreement with earlier studies of drifters [Richardson and Wa lsh, 1986], there is no eastward flow at this time of year in the latitudinal band usually occupied by the NECC.
The transport difference between the westward NBC flow across 44 °W and the eastward flow offshore is 10 Sv above <Te = 24.5.This is through flow of southern hemispheric water  1).
toward the Caribbean.Since there is no identifiable significant eastward return flow across our section even farther north than 7 °N, this flow did apparently continue northwestward along the boundary in spring 1994.The 10 Sv of through flow is a lower limit, obtained when assuming that all of the 4 Sv of the eastward counterflow offshore (Figure 8a) is retroflected wa ter.However, part of it has to be northern water as shown by the water masses on er 8 = 25.0 (Figure 4).In the schematic graph we acknowledge this partition by counting only half of the 4 Sv of eastward flow across 44 °W as retroflection water, bringing the total through flow above era = 24.5 to 12 Sv.For the undercurrent layer (Figure 8b ), the low-salinity, high-oxygen source waters originate dominantly from latitudes exceeding 10 °S: 10 Sv arrived as boundary flow in the NBUC and were only marginally augmented from the east in the 10 °-5 °S latitude range.Within the SEC, 3 Sv join the NBC at 35 °W just north of 5 °S, while 9 Sv of high-salinity, low-oxygen waters joined from the east in this layer, including the 2 Sv below the EUC (Figure 7c).There are also inconsistencie in the transport budgets for this layer between 4D°W and 44 °W.It can be stated that 14 Sv arrive at 44 °W and that there is matching eastward transport in the offshore flow that continues at about the same amount over the 1000-krn distance to the 35 °W section.The tran port agreement, combined with the southern water mass properties of that flow indicate that the 14 Sv of EUC flow in the er a = 24.5-26.88).

24,759
The transport inconsistency in the NBC between 40 °W and 44 °W also applies to the undercurrent layer; here there is an unexplained decrease of 5 Sv (Figure Sb; Table 1) between both longitudes, while in the upper layer there was an increase of 5 Sv.Although measurement errors may account for 1-2 Sv in the main current branches, it can almost be excluded that they were responsible for these inconsistencies.The main source of errors were GPS navigation and compass inaccura cies, both of which are not depth dependent and therefore could not explain the sign change of the transport inconsisten cies between the layers.Upper layer transports also depend on how representative the shear extrapolation over the top 25 m was, but again this could not explain the sign change of the transport imbalance.This leaves three possible reasons for these (and other) discrepancies in both layer transports: First, the surface mixed layer in the northwestern part of the survey region reached down very closely to the level a 0 = 24.5, mean ing that mass was probably not fully conserved in each of the layers.This, together with the downstream rise of the NBC and EUC cores suggests that upwelling might play a role, leading to some exchange between the two layers.Second, nonsynopticity of the survey was certainly a factor, and third, flow could escape onto the wide shelf region at 40 °W and 44 °W, but not at 35 °W.In October 1992 and in June 1991 the transport over the shelf at 44 °W was only about 0.5 Sv [Schott et al., 1993], while Candela et al. [1992]  As regards the off-equatorial North and South Equatorial Undercurrents, the water masses show clearly for the SEUC and not quite so clearly for the NEUC that they do not orig inate out of the boundary current but instead originate from recirculation of low-oxygen equatorial waters [Tsuchiya, 1986], and we mark them correspondingly in Figure Sb.The situation for the NEUC was found different from our observations for the summer circulation [Wilson et al., 1994], when the NECC is fully developed and not flowing westward as during our survey.The transport of the SEUC of 4.9 Sv at 35 °W is small com pared with the mean geostrophic transport of 15 Sv by Co chrane et al. [1979], which in part may result from their mea surement locations further east.Molinari [ 1982] described a range between 5 and 23 Sv f o r the SEUC also further east, at 25 ° to 28 °W.
The Antarctic Intermediate Water inflow below a 0 = 26.8 that arrived in the region as the deep part of the NBUC had a transport of 6 Sv at 10 °S and 5 °S.That number is lower than the 10 Sv obtained in inverse calculations for the Tropical Atlantic at 8 °S by Roemmich [1983] and the 8-14 Sv obtained by Schlitzer [1988] while in agreement with inverse results ob tained by Fu [198 1] using International Geophysical Year data from the subtropical South Atlantic.This flow turns around Cape San Roque (Figure Sc), and then there appears to be interaction with the two deep off-equatorial undercurrents, the Southern and Northern Intermediate Countercurrents.The SICC in particular is clearly related, by its high oxygen values and low salinities (Figure 6), to the AAIW in the boundary current.For the NICC the sources appear dominantly north ern and out of the recirculation of equatorial waters, but ad mixtures of AAIW are suggested by a slight oxygen maximum offshore in the 44 °W section, (Figure 6b).Other presentations of AAIW level distributions (Le Floc ' h and Merle, 1975] indi cate a branching of the AAJW eastward south of the equator by an eastward extending oxygen maximum and salinity mini-mum while showing no such eastward extension at the NICC latitude [Tsuchiya, 1989].
The transports for the SICC and NICC are quite large,at 11 and 8 Sv,respectively.Wedged in between the NICC and SICC is the westward flowing Equatorial Intermediate Current.While the NICC and SICC were not previously reported to exist in the Atlantic, the EiC was already documented in the observations of Stalcup and Metcalf(l966] and Hisard and Rua/ [1970], although transports could not be determined in these earlier measurements.In our section that EiC transports 19 Sv westward between a 0 = 26.8 and 1000 m, part of which recir culates into the SICC and NICC and part of which continues to merge with the NBC yielding a cross-equational flow of 14 Sv at 44 °W.Although the transports in that layer nearly balance in the 35 °W section, one has again to keep aware of the equato rial wave system that can overlay small mean currents in the directly measured current profiles.Hence we present these numbers for the equatorial intermediate currents as first indi cations, not more. The equatorial divergence in the Atlantic amounts to an upwelling of about 25 Sv that needs to be compensated by subsurface (above a 0 = 26.8)geostrophic flow toward the equator [e.g., Roemmich, 1983].Between our 44 °W and 35 °W sections, we could not detect the upwelling, for example, by a rising of the subsurface salinity maximum across isopycnal sur faces, although the water mass transfer of 5 Sv indicates some upwelling.In terms of the overall warm water transport the three layers show (Figures 7a and 7b) that 16 Sv arrive from the south along the coast, about the right number for the Atlantic overturning cell [e.g., Roemmich and Wu nsch, 1985;Rintoul, 1991 ].The low-latitude inflow into the western bound ary regime that is added to the boundary flow north of 5 °S {by the SEC) must then dominantly be recirculation of equatorial water, stemming from the equatorial divergence, out of the EUC rather than a "detour branch" of the "conveyor belt" circulation.This is clearly supported by the difference in the water mass properties of the SEC and NBUC branches.

Figure 3 .
Figure 3. Distributions of water mass properties in the upper 1200 m along 35 °W: (a) salinity and (b) oxygen in milliliters per liter.Current cores of different current branches are identified within the 20 cm s-1 contour in the upper 300 m, except for SEUC and NEUC ( 10 cm s -1 contour) and within the 10 cm-1 s -1 contour in the range 300-1200 m.Also included (dashed lines) are the isopycnals of a 0 = 25.0, a 0 = 26.7,and a 0 = 27.28.BC, North Brazil Current; SEC, South Equatorial Current; EUC, Equatorial Undercurrent; SEUC and NEUC, South and North Equatorial Undercurrent; EiC, Equatorial Intermediate Current; SICC and NICC, Southern and Northern Intermediate Countercurrents.

Figure 4 .
Figure 4. Distribution on isopycnal a<•> = 25.0 of (a) salinity, (b) oxygen, and (c) current vectors from shipboard ADCP and water mass class (sec text for details).Open circles in Figure 4c denote equatorial and northern water mass properties; solid circles denote southern water mass properties.

Figure 7 .
Figure 7. Current section plots in meters per second from combined shipboard and lowered ADCP of (a) meridional component at J0 °S, (b) meridional component at 5 °S, (c) zonal component at 35 °W, (d) zonal component at 40"W, and (e) zonal component at 44 °W.Transports in sverdrups of current branches are indicated in plots (large bold numbers; l Sv = 10 1 ' m 3 s -1 ); boundaries are marked by dotted lines.Meridional transports are positive northward; zonal transports are positive eastward.

Table 1 .
Current Branch Transports in Western Tropical Atlantic, Spring 1994 [Stramma et al., 1995]o cr H = 27.4 isopycnal.conservedagainst10°S.Overall, the total of 18.5 Sv for the NBUC is only 1.7 Sv larger than l0 °S boundary current trans port, mostly owing to an addition in the very near-surface layer.The total transport is below the 22 Sv found for November I 992 by Stramma et al.[1995], who found even less zonal inflow between 10 °S and 5 °S.The smaller transport in austral fall is in agreement with the observations from historical hydrographic data[Stramma et al., 1995]and with the evaluation of the WOCE CME model by Schott and Boning layer are fed out of the boundary flow and that 10 Sv of it originate from farther south than lO °S.The total EUC transport of 21.4 Sv at 35°W in March 1994 is in good agreement with the model results and direct observation at this season[Schott and Boning, 1991, Figure obtained 3.0 Sv in May and March from ADCP sections farther northwest at 48 °-52 °W.