Annual Cycle and Variability of the North Brazil Current

JANUARY 1998 JOHNS ET AL. 103

W. E. J OHNS AND T. N . L EE Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, Florida

R. C. BEARDSLEY,J.CANDELA, AND R. LIMEBURNER Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

B. CASTRO Instituto Oceanogra®co Universidade SaÄo Paulo, SaÄo Paulo, Brazil (Manuscript received 18 October 1996, in ®nal form 5 June 1997)

ABSTRACT Current meter observations from an array of three subsurface moorings located on the Brazil continental slope near 4ЊN are used to describe the annual cycle and low-frequency variability of the North Brazil Current (NBC). The moored array was deployed from September 1989 to January 1991, with further extension of the shallowest mooring, located over the 500-m isobath near the axis of the NBC, through September 1991. Moored current measurements were also obtained over the adjacent shelf for a limited time between February and May 1990. The NBC has a large annual cycle at this latitude, ranging from a maximum transport of 36 Sv (Sv ϵ 106 m3 sϪ1) in July±August to a minimum of 13 Sv in April±May, with an annual mean transport of approximately 26 Sv. The mean transport is dominated by ¯ow in the upper 150 m, and the seasonal cycle is contained almost entirely in the top 300 m. Transport over the continental shelf is 3±5 Sv and appears to be fairly constant throughout the year, based on the available current meter records and shipboard ADCP surveys. The NBC transport cycle is in good agreement with linear wind-driven models and appears to be in near-equilibrium with remote wind stress curl forcing across the tropical Atlantic for much of the year. However, the mean transport of the NBC is 15 Sv larger than can be explained by wind forcing alone, indicating a strong thermohaline component. Mesoscale variability in the region is dominated by ¯uctuations with periods near 25±40 days and 60±90 days. The 25±40-day ¯uctuations are strongly surface trapped and are most energetic in early summer during the acceleration phase of the NBC. The lower-frequency ¯uctuations have a deeper reaching baroclinic structure, are present year-round, and are associated with the propagation of large anticyclonic eddies northwestward along the coast. It is hypothesized that these features may serve as a catalyst for the eddy shedding process seen in the NBC retro¯ection in earlier observations.

1. Introduction conduit for cross-equatorial transport of South Atlantic The North Brazil Current is a major low latitude west- upper-ocean waters as part of the Atlantic meridional ern boundary current in the Atlantic that transports up- overturning cell. This northward thermohaline return per-ocean waters northward across the equator. Its coun- ¯ow originates from a combination of intermediate wa- terparts in the other ocean basins are the Somali Current ters ¯owing though Drake Passage from the Paci®c in the Indian Ocean (DuÈing and Schott 1978; Swallow Ocean (Rintoul 1991) and warmer thermocline waters et al. 1991) and the New Guinea Coastal Current in the ¯owing westward around South Africa from the Indian Paci®c (Lukas et al. 1991). However, due to differences Ocean (Gordon 1986; Gordon et al. 1992). On the mean in basin geometry and forcing, each of these current these two sources, plus northward ¯ow of Antarctic Bot- systems is rather distinct in character. In the Atlantic tom Water, must combine to balance an export of ap- 6 3 Ϫ1 the NBC plays a dual role, ®rst, in closing the wind- proximately 15 Sv (Sv ϵ 10 m s ) of North Atlantic driven equatorial gyre circulation and feeding a system Deep Water from the Atlantic (Roemmich and Wunsch of zonal countercurrents and, second, in providing a 1983; Schmitz and McCartney 1993). The formation region of the NBC is generally agreed to be near 10ЊS where waters ¯owing westward in the South Equatorial Current (SEC) ®rst begin to concen- Corresponding author address: Dr. William E. Johns, RSMAS, University of Miami, 4600 Rickenbacker Causeway, Miami, FL trate into a northward boundary current (Fig. 1a). Near 33149. 5ЊS the northward transport of the NBC determined by E-mail: [email protected] geostrophic calculations and current pro®ling is ap-

᭧ 1998 American Meteorological Society

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 104

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 105 proximately 15±20 Sv (Stramma 1991; Stramma et al. variation is determined by remote wind stress (i.e., 1995; Schott et al. 1995). Most of this transport is con- Sverdrup forcing) in the interior, versus local wind tained in the thermocline, with only 3±5 Sv occurring forcing in the coastal region. For the tropical Atlantic in the surface layer; hence Stramma et al. (1995) refer between 0Њ±5ЊN, linear Sverdrup dynamics predicts a to the northward boundary current here as the North maximum northward boundary current transport dur- Brazil Undercurrent (NBUC). Between 5ЊS and the ing boreal summer and fall, when the ITCZ shifts to equator the NBC increases its transport due to predom- its northernmost position near 10ЊN, and a reversal to inantly surface in¯ow from the SEC, leading to an an- a weak southward boundary current during boreal nual mean transport at the equator of 32 Sv in the upper spring when the ITCZ retreats to the equator (Mayer 600 m (Schott et al. 1993, hereafter S93). Seasonal vari- and Weisberg 1993). No evidence for such a reversal ability of the NBC between 10ЊS and the equator appears has been found for the NBC, suggesting that either to be quite small; for example, S93 ®nd an annual cycle the remote wind forcing signal is overpowered by at the equator of only Ϯ3 Sv, with maximum transport northward thermohaline ¯ow carried by the NBC, or during June±August and minimum transport during De- that it is obscured by propagation delays from the cember±February. interior or local forcing effects. North of the equator water begins to leave the NBC In an effort to determine the pathways of the upper- and ¯ow into the interior, feeding a system of zonal ocean thermohaline ¯ow through the tropical Atlantic, countercurrents: the North Equatorial Countercurrent and the manner in which this ¯ow interacts with the (NECC), the Equatorial Undercurrent (EUC), and North seasonal wind-driven circulation, an international ®eld Equatorial Undercurrent (NEUC)(Metcalf and Stalcup program (WESTRAX, Brown et al. 1992) was con- 1967; Cochrane et al. 1979). At the surface the NBC ducted during 1989±91 concentrating on the western continues to 6Њ±7ЊN where it retro¯ects seasonally into boundary region between 0Њ and 10ЊN. Moored current the NECC from approximately June±January. Shipdrifts meter observations were collected in the North Brazil and drifter trajectories show that eastward ¯ow in the Current near the equator (S93), near 4ЊN (this study), NECC ceases in the western Atlantic from February± and near 6ЊN (Colin et al. 1994), to investigate the struc- May and is replaced by weak westward ¯ow (Richard- ture and variability of the NBC at several locations along son and Walsh 1986); therefore, any surface water trans- the coast. Earlier current meter measurements in the ported northward in the NBC at this time is presumed NBC retro¯ection region near 8ЊN are also available to continue northward along the coast and join the west- and those results are reported in Johns et al. (1990, 1993) ward ¯ow of the North Equatorial Current entering the and Fratantoni et al. (1995). In addition, measurements Caribbean. In the thermocline layer a portion of the were collected from moored current meters deployed on NBC ¯ow separates at or just north of the equator to the North Brazil shelf near 4ЊN during February±May feed the EUC, while another portion continues to 3Њ± 1990, and from shipboard ADCP surveys of the shelf 4ЊN where it appears to feed seasonally into the NEUC. region during the WESTRAX period, as part of a sep- Some of this thermocline ¯ow may also recirculate back arate international ®eld program (AMASSEDS, Nit- into the NBC around a semipermanent anticyclonic eddy trouer et al. 1991) designed to investigate the transport or gyre (the ``Amazon'' eddy: Bruce et al. 1985) cen- of freshwater and suspended sediment from the Amazon tered near 2ЊN. Mesoscale energy is known to be large River into the Atlantic. in the region (Wyrtki 1976; Johns et al. 1990; Didden The emphasis in this paper is on the currents and and Schott 1993) and has made the identi®cation of transports in the upper water column, extending from seasonal mean ¯ow patterns from synoptic ship surveys the surface to approximately 800 m, where the transition dif®cult. from the northward ¯owing NBC regime to the south- Models (e.g., Philander and Pacanowski 1986) in- ward ¯owing deep western boundary current (DWBC) dicate that the NBC has a large seasonal cycle north regime occurs. Mean ¯ows and statistics from the deeper of the equator, related to the seasonal migration of current measurements are also presented, but a detailed the intertropical convergence zone (ITCZ) and asso- discussion of these records is deferred until further syn- ciated changes in the wind stress curl across the in- thesis with regional measurements collected in the terior of the basin. A question left to be answered for WESTRAX time frame. the NBC, and for low-latitude western boundary cur- The organization of the paper is as follows. Data rents in general, is the extent to which their transport sources used in the analysis are described in section

←

FIG. 1. (a) Schematic of near-surface currents in the equatorial Atlantic Ocean (after Richardson and Walsh 1986). (b) The study region. Current meter mooring locations are indicated (sites A1, A2, and A3 are AMASSEDS moorings deployed over the Amazon shelf; sites S1, S2, and S3 are STACS-11 moorings deployed in the offshore NBC region). Shipboard ADCP sections in the offshore NBC region available during the period of the study (in Sep 1989, Sep 1990, and Jan 1991) are shown in bold lines. The stippled region on the shelf shows the area covered by AMASSEDS shipboard ADCP surveys in Aug 1989, Mar 1990, May 1990, and Nov 1991. (c) Mooring con®guration over topography, showing the instrument types and depths. The AMASSEDS moorings are shown on expanded vertical scale in the inset.

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 106 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

TABLE 1a. STACS-11 moored current meter statistics. Seasonal Instrument Record length Annual Mar±May Aug±Nov depth Length (m) (days) Var Mean Std dev Mean Std err Mean Std err Mean Std err S1 Lat 3Њ52.4ЈN, Long 48Њ43.5ЈW: Water depth 480 m 60 691.0 u Ϫ54.6 24.0 Ϫ55.3 3.6 Ϫ36.8 6.0 Ϫ69.3 8.2 v 47.9 26.0 48.7 4.3 28.7 5.2 59.0 9.5 110 735.5 u Ϫ40.5 21.8 Ϫ40.5 2.8 Ϫ24.7 7.1 Ϫ56.2 6.4 v 37.5 22.9 37.4 3.4 20.3 6.3 51.1 9.7 310 496.0 u Ϫ2.0 23.2 Ϫ2.5 4.0 9.9 8.0 Ϫ8.2 6.9 v 6.5 18.6 6.7 3.0 Ϫ3.0 3.4 11.2 6.9 S2 Lat 4Њ28.7ЈN, Long 48Њ18.5ЈW: Water depth 2037 m 50 499.5 u Ϫ26.0 27.2 Ϫ25.1 4.9 Ϫ5.5 11.4 Ϫ38.1 12.3 v 27.0 34.5 26.6 6.9 7.7 12.8 34.8 15.4 100 499.5 u Ϫ25.3 25.5 Ϫ23.7 4.6 Ϫ6.2 10.8 Ϫ39.2 11.0 v 17.7 30.3 17.6 6.1 Ϫ1.5 9.2 25.2 13.6 150 499.5 u Ϫ13.0 26.8 Ϫ11.5 4.2 0.6 15.2 Ϫ25.0 10.1 v 26.5 28.7 28.2 4.3 4.3 7.7 35.7 13.1 300 499.5 u Ϫ3.7 19.8 Ϫ3.6 3.3 0.1 9.7 Ϫ8.9 5.0 v 17.6 23.3 19.4 3.8 14.8 6.8 22.0 7.4 400 499.5 u Ϫ5.1 19.4 Ϫ5.1 3.2 Ϫ3.8 8.9 Ϫ10.0 6.0 v 16.6 20.7 18.6 3.5 17.1 7.4 20.3 7.8 800 499.5 u Ϫ0.7 9.9 Ϫ0.4 1.6 Ϫ1.0 4.3 Ϫ2.7 2.2 v 3.1 8.4 2.9 1.5 0.5 2.7 2.8 2.2 1400 499.5 u 17.8 14.1 19.9 2.1 16.2 7.4 18.9 3.4 v Ϫ7.9 7.9 Ϫ8.5 1.4 Ϫ7.4 4.6 Ϫ10.8 1.4 1800 445.0 u 17.8 16.1 18.7 2.7 18.0 10.5 24.6 3.9 v Ϫ7.2 7.9 Ϫ7.6 1.4 Ϫ7.4 5.0 Ϫ11.1 1.9 S3 Lat 5Њ17.6ЈN, Long 46Њ48.6ЈW: Water depth 3421 m 160 501.5 u 11.3 40.8 8.5 6.5 Ϫ8.1 8.4 1.6 20.5 v Ϫ9.8 35.1 Ϫ9.2 6.1 Ϫ0.4 13.9 Ϫ20.2 12.8 310 501.5 u 10.7 16.0 10.1 2.6 1.4 7.9 9.1 3.9 v Ϫ5.5 16.7 Ϫ6.7 2.6 8.2 5.0 Ϫ9.9 4.2 860 444.5 u Ϫ2.9 6.5 Ϫ2.7 1.3 Ϫ1.6 2.6 Ϫ5.6 2.4 v Ϫ1.6 7.0 Ϫ2.0 1.4 Ϫ0.1 3.5 Ϫ0.5 2.8 1460 501.5 u Ϫ3.4 11.3 Ϫ4.2 2.2 Ϫ6.7 6.8 Ϫ10.5 1.9 v 2.1 12.9 4.1 1.9 9.3 7.5 2.5 3.7 2060 501.5 u Ϫ2.8 9.3 Ϫ3.6 1.8 Ϫ4.2 3.8 Ϫ9.2 2.1 v 1.5 12.1 3.2 1.7 7.7 4.7 4.3 3.7 2860 138.0 u 1.3 6.7 Ð Ð Ð Ð 3.4 7.1 v Ϫ1.0 7.2 Ð Ð Ð Ð 2.2 5.2 3360 501.5 u 1.8 3.8 2.1 0.7 3.6 2.1 0.8 1.2 v Ϫ1.7 4.8 Ϫ1.6 0.9 Ϫ1.2 2.8 Ϫ0.9 1.2

2. A discussion of the seasonal current variation in 2. Observations the upper ocean follows in section 3. Methods used to calculate the transport of the NBC and principal a. Moored current meters results on the annual cycle of the NBC, its vertical The principal data used for this study are current me- structure, and the partitioning of the transport on and ter records recovered from an array of moorings de- off the shelf are given in section 4. Section 5 includes ployed across the continental shelf and slope of Brazil a brief discussion of the intraseasonal (mesoscale) near 4ЊN (Fig. 1). The continental margin in the region variability of the NBC at this location and describes of study consists of a broad inner shelf of less than 30 the dominant timescales and characteristics of the m depth that extends 100 km seaward of the coast, nar- eddy ®eld. Section 6 concludes with a summary and rowing gradually to the north. Near the 30-m isobath discussion of the results, with emphasis on the annual the shelf descends to an outer shelf platform (depths transport cycle of the NBC and it relationship to bas- 60±100 m) of nearly uniform width along the coast, inwide wind and thermohaline forcing, including followed by a shelf break at the 100-m isobath. The some relevant model results. total width of the shelf at this location is approximately

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 107

TABLE 1b. AMASSEDS moored current meter statistics. due to a tape recorder failure. Current pro®les from the Record length Common length ADCP were recorded at a nominal vertical resolution (10 Feb±16 Apr) of 8 m and subsequently interpolated to 10-m depth Instrument Record depth length Std Std intervals between 50 and 400 m. Data closer to the (m) (days) Var Mean dev Mean dev surface could not be used due to biases introduced by sidelobe interference from surface re¯ections (Johns A1 Lat 3Њ4.8ЈN, Long 50Њ18.6ЈW: Water depth 17 m 1988). The processing and editing of the data is sum- 3 101.5 u Ϫ25.7 19.9 Ϫ24.7 22.1 v 32.8 20.8 29.1 23.6 marized in Zantopp et al. (1993). 16 128.0 u Ϫ4.9 6.9 Ϫ5.9 8.7 Current meter moorings were also deployed over the v 4.2 6.4 5.8 7.6 north Brazil shelf during a portion of the STACS-11 A2 Lat 3Њ23.4ЈN, Long 49Њ56.4ЈW: Water depth 65 m period, from mid-February to mid-June 1990, as part of 3 65.0 u Ϫ46.8 27.3 Ϫ46.8 27.3 AMASSEDS (A Multidisciplinary Amazon Shelf Sedi- v 57.2 31.1 57.2 31.1 ment Study). Results from these moorings have been 32 65.0 u Ϫ36.6 9.3 Ϫ36.6 9.3 described in several papers dealing with the structure v 28.1 8.2 28.1 8.2 62 182.5 u Ϫ11.5 6.4 Ϫ12.4 5.3 and variability of the Amazon River plume (Lentz v 4.6 5.7 5.8 3.2 1995a,b) and tidal variability (Beardsley et al. 1995) A3 Lat 4Њ4.2ЈN, Long 49Њ37.2ЈW: Water depth 100 m over the shelf, and in a review of factors affecting the 32 127.5 u Ϫ59.9 18.5 Ϫ65.2 14.5 evolution and fate of the Amazon plume (Geyer et al. v 63.7 22.0 69.7 15.0 1996). Moorings A1 and A2 at the inner and midshelf 93 127.5 u Ϫ14.9 6.9 Ϫ16.9 6.6 were deployed by the Woods Hole Oceanographic In- v 10.1 8.3 11.1 8.1 stitution (WHOI), and mooring A3 at the outer edge of the shelf was deployed as a joint project between WHOI and the Instituto Oceanogra®co Universidade SaÄo Paulo 200 km. Seaward of the shelf break and extending to (IOUSP), Brazil. A variety of instrumentation was used depths in excess of 3000 m is the Amazon Cone, a on the moorings (Fig. 1c), including the use of paired depositional feature formed during Quarternary low sea surface and subsurface moorings at sites A1 and A2. level periods. Data recovery and processing of the AMASSEDS cur- Between September 1989 and January 1991, an array rent meter records are summarized in Alessi et al. of three moorings (labeled S1, S2, and S3 in Fig. 1b) (1992). The A2 surface mooring failed in mid-April was deployed in waters offshore of the shelf break by 1990, and as a result only the ®rst part of the deployment the University of Miami. The shallowest of these moor- is suitable for constructing transport estimates over the ings (S1) was placed over the 500-m isobath, near the shelf; this period (10 February±16 April 1990) is re- expected core of the NBC based on previous results ferred to hereafter as the AMASSEDS common time from shipboard pro®ling in the region (Flagg et al. period. 1986). This mooring (S1) was maintained for a full two- The orientation of the isobaths is roughly north- year period through September 1991. Mooring S2 was westward over the continental shelf and upper slope, located 60 km farther offshore over the 2000-m isobath, turning more westward offshore due to the Amazon and mooring S3 was located approximately 280 km sea- Cone. The direction of the mean ¯ow over the shelf and ward of the shelf break in 3400-m water depth (Fig. upper slope, as well as in the offshore region dominated 1c). We will henceforth refer to this array as the STACS- by the NBC, is predominantly northwestward. There- 11 Array [denoting its order in a sequence of mooring fore, all time series plots of vector currents are rotated deployments begun in the early 1980s by the NOAA by 45Њ into alongshore (positive northwest; 315ЊT) and Subtropical Atlantic Climate Studies (STACS) pro- cross-shore (positive northeast; 45ЊT) components for gram], and refer to the September 1989±January 1991 display purposes. All current meter records have been period as the STACS-11 common time period. low-pass ®ltered with a 40-h Lanczos ®lter to eliminate Instrumentation on these moorings consisted of a tidal and shorter period variations, and then subsampled mixture of vector averaging current meters (VACMs), at 12-h intervals. The mooring locations, instrument Aanderaa current meters, and one R. D. Instruments depths and durations, and ®rst-order statistics of all the self-contained 150-kHz acoustic Doppler current pro- low-passed current records are given in Tables 1a and ®ler (ADCP) looking upward from a depth of 400 m 1b. on S2. Sampling was at half-hour intervals for the VACMs, and at hourly intervals for the Aanderaa cur- b. STACS-11 array data rent meters and ADCP. All of the instruments returned high-quality records except for the top instrument on Vector time series of the upper-ocean (surface to 310 S3 (160 m), which had a noisy direction record requiring m) currents from the STACS-11 array are displayed in substantial editing. Four of the records were short by Fig. 2a. The current pro®les from the ADCP at site S2 up to several months due to battery failures, and one are subsampled to 50-m intervals and only those at 50, (the 2860-m record on S3) was only four months long 100, 150, and 300 m are shown. The presence of the

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 108 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

FIG. 2. Vector time series of (a) upper-ocean currents on moorings S1, S2, and S3, and (b) subthermocline currents on moorings S2 and S3 (below 800 m). All time series were ®ltered with a 40-h low-pass ®lter. Northwest is up.

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 109

FIG. 3. Alongshore component (heavy lines; positive northwest) and cross-shore component (lighter lines; positive northeast) of current from the 2-yr record at mooring S1.

strength of the NBC ¯ow over the upper continental slope, and this variation as well as shorter-term ¯uctuations in alongshore velocity are more clearly il- lustrated in Fig. 3. The two-year record at S1 shows similar behavior in both years: a minimum in ¯ow strength in late spring (May), followed by an acceleration of the upper-layer ¯ow through the summer to a maximum strength in late summer and fall. Fluc- tuations in the alongshore component of ¯ow occur throughout the record but are particularly pronounced during the acceleration phase of the NBC during summer. Seasonal variation is not as obvious in the near- surface current records at S2 although there is a minimum in ¯ow strength in April±May 1990 followed by relatively strong northwestward ¯ow during July± September (Fig 2a). The remainder of the records tend to be dominated by eddy variability and do not show an obvious annual cycle. An abrupt transition in the character of the current records is seen between the surface layer and the subthermocline layer near 300-m depth. For example, at site S1, where the surface signature of the NBC is strongest, there is no indication of a persistent mean ¯ow at 310-m depth, and the ¯uctuations there are large- FIG.2.(Continued) ly uncorrelated with those in the surface layer. Exam- ination of the ADCP current pro®les from S2 reveals that this transition occurs near 150 m, roughly at the NBC is clearly indicated in the strong, directionally sta- depth of the main thermocline, and that currents in the ble northwestward currents recorded at 60-m and 110-m upper 100 m and in the layer between 200±400 m are depth at site S1. Near-surface currents from the ADCP highly coherent among themselves. The layers above at site S2 also show a predominance of northwestward and below the main thermocline therefore appear to be ¯ow but exhibit more directional variability and occa- dynamically uncoupled, and as will be shown later in sional reversals. The current record from 160 m at site section 5, the dominant space and time scales of the S3 farther offshore shows prominent reversals on rel- variability in these layers are also different. atively long timescales and appears to be seaward of Deep currents in the region are expected to be in¯u- the in¯uence of the mean NBC boundary ¯ow regime. enced by the deep western boundary current (DWBC) Substantial seasonal variation is evident in the ¯owing equatorward and carrying tracer signatures of

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 110 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28 recently ventilated water from the North Atlantic. Pre- vious measurements along the tropical Atlantic western boundary have shown that high chloro¯uorocarbon (CFC) concentrations are found in two cores at depths of approximately 1400±1800 m and 3500±4000 m, associated with Upper and Lower North Atlantic Deep Water (UNADW and LNADW), respectively (Fine and Molinari 1988; Molinari et al. 1992). Current meters were placed on moorings S2 and S3 at levels spanning the upper CFC core, and at S3 near the top of the lower CFC core (just above the bottom), to observe the strength and variability of the DWBC at this location. Measurements were also made near 800-m depth on both moorings, and at 2860 m on S3, though the available record at the latter site is only 4 months long. Strong deep ¯ows were recorded at the 1400-m and FIG. 4. Alongshore components of ¯ow from moorings A1, A2, 1800-m levels on S2, indicating an intense mean jet and A3 over the Brazilian shelf, at the indicated depths. The time ¯owing southeastward along the topography at the UN- periods of AMASSEDS shipboard ADCP surveys of the shelf in Feb± ADW core depth (Fig 2b). Maximum currents exceeding Mar 1990 and May 1990 are shown by stippling. 40 cm sϪ1 occurred during a prolonged period of southeastward ¯ow from January to April 1990, and at various other times during the latter half of the record. These strong ¯ow periods were broken by two periods small vertical separations of 15±30 m on A1 and A2. of weak or reversed ¯ow during September±December Apparently, the strong halocline at the base of the plume 1989 and April±May 1990. Similarly strong mean decouples the plume ¯ow from the subplume ¯ow southeastward ¯ow with intermittent behavior was (Lentz and Limeburner 1995). Lentz (1995a) found that found by Colin et al. (1994) in the depth range 1400± the alongshore current variations in the plume layer 2000 m from moored current meter records near 6ЊN, were highly correlated with local alongshore winds and 51ЊW during March 1990±September 1991. The cause that a simple slab model for the plume layer with a of this intermittency is unknown, but as is discussed small linear ``interfacial'' friction coef®cient of 0.002 Ϫ1 further in section 5, it is believed to be related to off- cm s (consistent with estimates of the entrainment shore meandering of the UNADW rather than a ceasing velocity at the base of the plume; Lentz and Limeburner or discontinuation of the DWBC ¯ow along the bound- 1995) could explain most of the observed variability in ary. the near-surface plume currents. After removing this wind-driven component of variability the mean 3-m alongshore currents at A1 and A2 for the AMASSEDS c. AMASSEDS array data common period were found to be 46 and 83 cm sϪ1, Current records over the shelf from the AMASSEDS respectively, which are slightly larger than the recorded moorings are displayed in Fig. 4. As shown by Lentz mean currents (Table 1b) due to the fact that the mean (1995a), subtidal velocity ¯uctuations over the shelf are wind was southeastward during the period and counter strongly polarized in the alongshelf direction and there- to the mean ¯ow direction. These large alongshore mean fore only the alongshore velocity components are shown currents over the inner and midshelf, as well as those here. The near-surface (3 m) currents at moorings A1 below the plume layer, are thought to be driven by a and A2 were strongly in¯uenced by the Amazon plume. large-scale alongshore pressure gradient related to the Salinity measurements on these moorings as well as NBC (Geyer et al. 1996). The vertical pro®le of this regional surveys of the shelf during AMASSEDS in- mean background ¯ow is strongly sheared, decreasing dicated that these instruments were almost always in the nearly linearly with depth over the midshelf. Mean cur- Amazon plume, and that the near-bottom instrument on rents at A3 near the shelf break were also strongly A1 (16 m) was near the bottom front separating the low- sheared and northwestward during this period, but the salinity Amazon wedge over the inner shelf from higher ¯uctuations there show no obvious correlation with the salinity (ϳ36 psu) oceanic water over the outer shelf winds despite the absence of a capping effect by the (Lentz 1995a). The deeper instruments on A2 (32 and Amazon plume (Lentz 1995a). This suggests that other 62 m), as well as those on A3 near the shelf break, were processes besides wind-forcing, including variations in in oceanic water and not directly in¯uenced by the the NBC farther offshore, are responsible for the vari- plume. ability near the shelf break and in the subplume layer The currents in the near-surface plume are more vari- over the midshelf. Further discussion of the relationship able than those in the subplume layer, and there is no between the shelf currents and offshore variations in the obvious correlation between them even at the relatively NBC is given in section 3.

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 111 d. Shipboard ADCP data As part of AMASSEDS, four shipboard ADCP surveys were conducted over the North Brazil shelf between the equator and 5ЊN (Fig. 1b, see also Fig. 8). The results are used here to augment the moored current meter measurements and to investigate the seasonality of the NBC transport over the shelf. Results from two of these surveys (February±March 1990 and May 1990) were reported in Candela et al. (1992), where a method to remove the tidal signal from the data and isolate the structure of the lower-frequency (presumably geostrophic) ¯ow ®eld over the shelf was developed. The same method is used here to remove tidal aliasing from the other two surveys, conducted in August 1989 and No- FIG. 5. Annual mean currents for the STACS-11 array at the in- vember 1991, respectively. dicated depths. Only those mean currents with magnitudes greater Several large-scale shipboard ADCP surveys were than 3 cm sϪ1 are plotted. also conducted in the offshore region during WES- TRAX (Wilson et al. 1994). During the STACS-11 common time period, three sections extending across the the months of September±January, and one year of data moored array to near the shelf break were obtained (Fig values (from 1990 only) for February±August. 1b). The ®rst and third sections occurred during mooring For most sites the record-mean and annual-mean cur- deployment and recovery operations (in August/Sep- rents are very similar with typical differences of 10% tember 1989 and January 1991, respectively), while the or less. The annual mean currents at S1 (Fig. 5) show second occurred in September 1990 along a slightly the northwestward ¯ow of the NBC decreasing with different track during replacement of the S1 mooring. depth from 74 cm sϪ1 at 60 m to 55 cm sϪ1 at 110 m, The results from these ADCP sections are used primarily then dropping to a weak (insigni®cant) mean ¯ow at as quantitative checks on the moored transport calcu- 310 m. At S2, the mean currents in the top 100 m are lations. also northwestward at about half the strength of those seen at S1. Below this the currents veer clockwise with 3. Annual and seasonal mean ¯ows depth to nearly a northward orientation at 400 m. Unlike the ¯ow at S1, the mean currents at S2 decay more Mean currents recorded for the total duration of each gradually with depth below the surface layer and sig- of the STACS-11 current meter deployments are given ni®cant mean ¯ows extend to at least 400 m. At S3, the in Table 1a. Standard deviations about these mean val- mean currents at 160 and 310 m show a similar (weak) ues as well as standard errors based on an integral time- southeastward ¯ow of 12 cm sϪ1 at each level, opposite scale of 25 days are also shown. Integral timescales in direction to the mean ¯ow farther onshore at the same derived from the autocorrelation functions of individual depths. While no data are available in the upper 100 m time series ranged from approximately 15 to 40 days, at this site, it is assumed that this offshore reversal re- with only the two records at the UNADW level on S2 ¯ects the return ¯ow of the NBC retro¯ection and that (1400 and 1800 m) showing values in excess of 30 days. the surface currents at this site would have a similar As there was little signi®cant variation in these esti- mean direction with somewhat larger magnitude. mates, a median value of 25 days was used in the cal- Deep mean ¯ows at the offshore sites show the afore- culation of all standard errors. mentioned southeastward jet at the 1400-m and 1800- While a few instruments on S1 returned data for a m levels on S2, with mean speeds of 19 cm sϪ1 parallel full two years, most of the records are 17 months long, to the local isobaths. Mean currents in the same depth and the mean currents derived from these records may range on S3 show weak northwestward ¯ow, indicating contain seasonal biases due to uneven sampling of the possible recirculation of UNADW offshore of the west- annual cycle. Therefore, we also calculate and show in ern boundary. The deeper currents on S3, at 2860 m Table 1a an ``annual'' mean current for each site, de®ned (short record) and 3360 m, showed weak southeastward as the average of the monthly mean currents at each mean ¯ows of approximately 2 cm sϪ1, suggesting very site. This procedure helps to remove seasonal biases in little mean in¯uence from the deep (LNADW) core of the mean currents by giving equal weight to each month the DWBC. The weakness of these mean ¯ows probably of the calendar year. Monthly mean currents are cal- indicates that the measurements were above the velocity culated simply as the average of all data values falling core associated with the LNADW tracer maximum in within a given month over the total record length. Thus, this area (found between 3500 and 4000 m: Molinari et the monthly mean currents derived for the STACS-11 al. 1992), or that the LNADW velocity core was located common period consist of two years of data values for farther offshore over deeper isobaths. At both of the

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 112 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

FIG. 7. Vector time series of currents at 60-m depth on moorings A2 (midshelf), A3 (outer shelf), and S1 (500-m isobath). The full record lengths at moorings A2 and A3 are shown (northwest is up). The 60-m current shown for A3 was estimated by linear interpolation between the current records at 32 and 93 m on the mooring. Note the visual correlation between the strong ¯ow events in NBC on the upper slope (S1) and those on the shelf as the NBC begins its seasonal onset in early summer.

over the upper continental slope, beneath the NBC surface layer, during the NBC ¯ow minimum in boreal spring. At S2, the near-surface currents are notably weaker in spring with little signature of the NBC in the upper 150 m, while the currents at 300 and 400 m are northward or northwestward in both seasons at 10±20 cm sϪ1 and similar to the annual mean currents at these FIG. 6. Seasonally averaged upper-ocean currents for the STACS- depths. Farther offshore at S3, the upper-ocean ¯ow 11 array, for the months of (a) Mar±May and (b) Aug±Nov. Mean reverses from weakly northward or northwestward in current vectors over the shelf shown in (a) are from the AMASSEDS common time period 10 Feb±16 Apr 1990. spring to southeastward (160 m) or southwestward (310 m) in fall, which is consistent with the expected development of the NBC retro¯ection during this period. offshore sites the mean ¯ow near 800-m depth was weak The overall picture given by these data is that of a (3 cm sϪ1) and varied in direction from northward at considerably weaker and more coastally trapped NBC S2 (roughly the same direction as the 400-m mean ¯ow) during spring, with much of its transport con®ned over to southwestward at S3. the shelf and upper continental slope. In fall, the current Seasonal changes in the upper-ocean currents are il- is broader, with the near-surface ¯ow spanning most of lustrated in Fig. 6, and listed in Table 1a. The averaging the width of the continental slope, and extending in periods of March±May and August±November are depth through the full water column over the upper somewhat subjective but were chosen to illustrate the slope. extremes of the annual cycle in boreal spring and fall. The relationship of the shelf currents to offshore vari- Mean currents over the North Brazil shelf are also ability in the NBC at S1 was investigated by both Lentz shown in Fig. 6a (Table 1b) for the AMASSEDS com- (1995a) and Geyer et al. (1996). No clear statistical mon period, which does not coincide exactly with the correlation between the shelf and slope currents was March±May averaging period of the offshore currents found in these studies for the limited period of overlap but is of similar duration and in the same season. between the AMASSEDS and STACS-11 measure- The surface layer ¯ow at S1 is about twice as strong ments. However, for some of the longer shelf records, in fall as in spring, and the ¯ow at 310 m reverses from there is a strong visual correlation between offshore ¯ow northwestward in the fall (the same direction as the events in the NBC and those on the shelf (Fig. 7). The surface ¯ow) to southeastward in spring. The standard ¯ow bursts at S1, which characterize the acceleration errors for the seasonal 310-m currents are comparable phase of the NBC in early summer, are clearly felt over to the mean values (Table 1a), so these changes are only the outer and midshelf for the event in late May to early marginally signi®cant. However, this seasonal reversal June and at the midshelf (and therefore also presumably suggests that a mean subsurface countercurrent develops at the outer shelf) for the following event in early July.

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 113

This suggests that the stronger ¯ow events in the NBC typical shelf conditions for this time of year and prob- penetrate over the shelf and that at least a portion of ably dominated by a transient feature. the shelf current and transport variability is driven by Estimates of the total transport over the shelf in the offshore variations in the NBC. The ¯ow maximum over vicinity of the AMASSEDS and STACS-11 arrays were the shelf during these events appears to lag the offshore made for each survey by spatially averaging the along- ¯ow maximum by 5±7 days, which could be due to shore currents along isobaths between 2Њ and 5ЊN and onshore propagation but is more likely due to down- integrating the results from the coast to the 100-m iso- stream propagation of the events within the NBC, as the bath. The resulting transports are shown in Table 2, shelf moorings line is displaced somewhat downstream along with a similar estimate derived from Flagg et al.'s of the STACS-11 mooring line (Fig. 1). (1986) data across the shelf near 4ЊN (their section C). The current records over the shelf are not long enough Apart from the November 1991 survey, the transport to determine whether a seasonal change in the ¯ow over estimates are all similar, ranging from 2.7 to 4.0 Sv, the shelf occurs similar to that observed over the upper with a mean value (excluding November 1991) of 3.4 slope. The visual correlation between the shelf and slope Ϯ 0.6 Sv. While this is obviously a very limited number currents seen during the NBC acceleration phase sug- of realizations of the shelf ¯ow, there does not appear gests that strengthening of the ¯ow over the shelf during to be any convincing evidence for a signi®cant change summer and fall is possible but far from certain since in the seasonal transport of the NBC over the north the dynamics of the cross-shelf coupling on the seasonal Brazil shelf near 4ЊN. timescale may be different from that on shorter (mesoscale) timescales. The poor statistical correlation 4. NBC structure and transportÐThe seasonal found between the shelf and slope currents during the cycle AMASSEDS common period may be largely due to the fact that this period corresponded with the minimum The moored time series were combined to produce ¯ow phase of the offshore NBC regime. Notably, the vertical cross sections of the NBC ¯ow structure as well mean northwestward ¯ow over the shelf during this pe- as time series of volume transport in various depth rang- riod is actually stronger than in the offshore NBC re- es. The methodology for the calculations is as follows. gion. First, a smooth vertical pro®le was ®t to the northwest Some insight into the seasonal variation on the shelf (315ЊT) component of the 12-h low-passed current me- is available from the four AMASSEDS shipboard ADCP ter data on each mooring, using a shape-preserving cu- surveys (Fig. 8) and from additional velocity data col- bic spline function (Akima 1970). To extend the pro®les lected by Flagg et al. (1986) in the same region in De- to the surface, the vertical shear at the top instrument cember 1980. During August 1989, March 1990, and derived from the spline ®t was used for extrapolation, May 1990 the ¯ow structure over the shelf is quite sim- and this shear was assumed to remain constant to the ilar and is characterized by northwestward ¯ow over the surface. Velocity was set to zero at the bottom. For the entire shelf region mostly parallel to the isobaths. In all shelf moorings with less than three measurement levels, three surveys the northward transport over the shelf be- linear interpolation and upward extrapolation was per- tween 2Њ and 5ЊN is considerably stronger than that formed along with linear decay of the pro®le to zero at observed in the region between 2ЊN and the equator. the bottom. This observation, coupled with the relative uniformity The resulting velocity pro®les were then gridded onto of the transport ®eld in the 2Њ±5ЊN region, implies that a 10 m vertical by 10 km horizontal grid for each 12- much of the transport over the shelf near 4ЊN is derived h time step using triangular (linear) interpolation, with from in¯ow across the shelf break between the equator the velocity set to zero at each grid point on the bottom. and 2ЊN. These gridded cross sections provide ``snapshot'' views During November 1991, the shelf ¯ow was more dis- of the current structure and its variability; time aver- organized between 3Њ and 5ЊN, with smaller-scale vari- aging of the data for speci®ed periods was then used to ations than are apparent in the prior surveys. In partic- determine the mean ¯ow structure. Volume transports ular, there is an eddy or some other localized circulation were estimated by summing the transport contributions feature centered over the shelf edge near 3Њ±4ЊN that causes southwestward ¯ow onto the shelf at that location and seems to divert the northward ¯ow on the shelf TABLE 2. NBC shelf transport estimates near 4ЊN. south of that location toward the interior. It seems un- Date Source Transport (Sv) likely that this ¯ow pattern is characteristic of the sea- Aug 1989 AMASSEDS 1 3.2 sonally averaged ¯ow pattern over the shelf during this Mar 1990 AMASSEDS 2 4.0 time period. For example, Flagg et al. (1986) found a May 1990 AMASSEDS 3 2.7 consistent pattern of northwestward ¯ow over the shelf Nov 1991 AMASSEDS 4 0.7 in this same region during December 1980 at almost the Dec 1980 Flagg et al. (1986) 3.6 same time of year. This suggests that the ¯ow observed Average 2.8 Ϯ 1.3 in the November 1991 survey was unrepresentative of Average (excl. Nov 1991) 3.4 Ϯ 0.6

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 114 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

FIG. 8. Maps of vertically integrated volume transport (per unit width, m 2 sϪ1) over the North Brazil shelf derived from detided AMAS- SEDS ADCP surveys: (a) Aug 1989, (b) Feb±Mar 1990, (c) May 1990, and (d) Nov 1991. over the entire grid or speci®c parts of the grid (e.g., approximately 2 to 6 Sv during the AMASSEDS period, for the determination of transport in different depth with a mean value of 4.8 Ϯ 0.8 Sv (Fig. 10). Most of ranges or different distances from the coast). this transport occurs in the upper 50 m, with only about The lack of moored current meter data over the shelf 1 Sv of ¯ow, on average, between depths of 50±100 m during most of the STACS-11 period requires some as- over the outer shelf. The March 1990 AMASSEDS sumption about the transport over the shelf in order to ADCP survey occurred during this period (Fig. 10), and estimate the total NBC transport. For the AMASSEDS the average moored transport over the period of that common period, the shelf moorings and two nearest survey (4.7 Sv) is in reasonable agreement with the slope moorings (S1 and S2) provide a fairly well re- spatially averaged shelf transport for that survey (4.0 solved picture of the NBC ¯ow within 300 km of the Sv). Since the available shipboard surveys in this region coast (Fig. 9a). During this period, the core of the NBC do not show an obvious seasonal variation in shelf trans- was located over the shelf break with an average surface port, we will hereafter assume a constant transport on speed of approximately 90 cm sϪ1. The calculated trans- the shelf, which will be added to the offshore transports port over the shelf (coast to 100-m isobath) varied from calculated from the STACS-11 array to estimate the total

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 115

FIG. 9. Cross sections of mean alongshore (positive northwest) velocity for (a) the AMASSEDS common period (Feb±Apr 1990) and (b) the STACS-11 common period (Sep 1989±Jan 1991). The vertical dashed line at 420 km shows the offshore limit of integration for the calculated NBC transport, where the vertical mean (0±800 m) ¯ow reverses in (b).

NBC transport. From the available estimates of shelf offshore. Farther offshore at S3, the current reverses to transport, any value in the range 3±5 Sv could be used, southeastward, indicative of the mean in¯uence of the and for consistency with the directly calculated trans- NBC retro¯ection. Including this southeastward mean ports during the AMASSEDS common period we ¯ow in the NBC transport calculation would bias the choose to adopt the mean value of 4.8 Sv derived from transports toward lower values and not provide an ac- the moored data. [Note that the smaller shelf transport curate measure of the strength of the boundary current of 2.7 Sv found during the May 1990 AMASSEDS sur- itself. Therefore, the transport integration was truncated vey occurred during a period of anomalously weak ¯ow at a point lying near the transition zone between the over the outer shelf (Fig. 4), and as such this transport northward boundary current and the southward offshore probably represents a lower than average value for this time of year.] For the entire STACS-11 common period, the ¯ow in the offshore region is enhanced compared to the spring season (Fig. 9a) and the total width of the northwestward boundary ¯ow extends approximately 400 km

FIG. 10. Transport over the North Brazil shelf (coast to 100-m FIG. 11. Moored transport time series for the NBC, for the STACS- isobath) for the AMASSEDS common period, including the transport 11 common period. Upper panel: total transport (0±800 m); lower contributions in the top 50 m, and between 50 and 100 m over the panel: 0±100-m and 0±300-m transport. Transport estimates for the outer shelf. Stippling indicates the period of the Feb±Mar 1990 upper part of the NBC derived from available shipboard ADCP sec- AMASSEDS shipboard ADCP survey. tions across the STACS-11 array are indicated by dots (lower panel).

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 116 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

TABLE 3a. Average transport (Sv) of the NBC in various depth record. The 0±300 m transport follows the same pattern ranges, for total record length and for annualized (seasonally adjusted) but shows a larger overall seasonal range, as the ther- average. Standard deviations and standard errors are also given. mocline layer adds little transport in April±May while Depth Record-length average Annual average adding typically 10 Sv to the surface layer transport range during other periods. Average transports in these layers (m) Mean Std dev Std err Mean Std dev Std err over the duration of the record (Table 3a) yield a total 0±100 13.8 5.1 1.1 13.2 3.6 1.0 mean transport of 26.4 Sv for the NBC, which is divided 0±300 22.0 9.0 2.0 21.2 7.8 2.0 0±800 26.4 13.8 3.1 26.1 11.0 2.9 approximately 50%, 30%, and 20% among the three 100±300 8.2 5.4 1.2 8.0 5.0 1.3 layers, respectively. The layer 300±800 m, while ac- 300±800 4.4 7.7 1.7 4.8 5.8 1.5 counting for only a ®fth of the total transport, exhibits about twice the transport variance of either of the upper two layers. Annual mean transports, de®ned in the same ¯ow regime. After experimentation with different ap- way as before for the mean currents (Table 3a), yield a proaches, a ®xed point (located 420 km from the coast), similar value of 26.1 Sv for the total 0±800-m transport where the time-mean, vertically integrated (0±800 m) with essentially the same partitioning among layers. alongshore ¯ow reversed sign, was chosen to terminate Comparison with available shipboard ADCP transects the transport integration (Fig. 9b). This choice mini- across the moored array shows good agreement in cal- mized eddy contamination and interpolation errors and culated transports for the top 100 m and top 300 m, the provided robust estimates of transport on seasonal maximum depth of good ADCP data (Fig. 11). The timescales. mean difference between the shipboard ADCP and To summarize, moored transport estimates for the moored estimates for the layer 0±300 m was 2.1 Sv, NBC were made by 1) integration of the gridded with only one cruise (January 1991) showing a transport STACS-11 velocity data from the shelf break (100-m difference greater than 2 Sv. It should be noted that isobath) to a distance of 420 km from the coast (holding these comparisons actually serve only as checks on the the velocities observed at S1 constant to the shelf break) offshore (STACS-11) part of the NBC transport cal- and 2) assuming a constant transport on the North Brazil culation, since none of the WESTRAX ADCP transects shelf of 4.8 Sv as derived from the AMASSEDS moored extended over the shelf. In calculating the shipboard current data. The resulting transport time series are dis- ADCP transports, the same constant shelf contribution played in Fig. 11 for three depth ranges (0±100 m, 0± assumed for the moored transport calculation was ap- 300 m, and 0±800 m); the ®rst corresponds to the sur- plied (4.8 Sv), and the ADCP measurements were ex- face layer, the second to the surface plus thermocline trapolated to the shelf break by assuming that the ob- layer, and the third to the depth to which signi®cant served velocity pro®le at the shoreward end of the NBC transport extends. The choice of 300 m for the ADCP section extended without change to the shelf base of the thermocline layer is somewhat arbitrary, and break. The agreement between these estimates never- it is used here mainly for comparison with other pub- theless suggests that the moored transport estimates are lished results in the region. fairly accurate despite the sparse resolution in the off- Total (0±800 m) transport through the moored section shore region between S2 and S3. varies from approximately Ϫ10 Sv to 60 Sv with large Seasonal cycles of the transports were computed by ¯uctuations on 1±3-month timescales. Seasonal trans- taking monthly means of the respective time series after port variation is most apparent in the upper layers and ®ltering them with a 90-day low-pass ®lter to reduce becomes less evident with depth due to increased con- month-to-month variability (Fig. 12). The annual range tamination by eddy ¯uctuations. The surface layer trans- of the total NBC transport is approximately 23 Sv, with port decreases gradually from 20 Sv in September 1989 an asymmetric seasonal cycle ranging from a low of 13 to about 10 Sv in April±May 1990, followed by an Sv in April±May to a high of 36 Sv in July±August. increase up to 20 Sv again by August 1990 with oscil- Nearly all of this seasonal signal is contained in the top lations in the range 10±25 Sv for the remainder of the 300 m. Uncertainties in the monthly mean transports,

TABLE 3b. Annual and semiannual harmonics of NBC tranpsort, plus characteristics of seasonal cycle. Amplitudes are in Sverdrup, phases (␾) are Julian day of maximum or minimum, respectively. Depth Annual Semiannual Seasonal cycle range (m) Ampl ␾ Ampl ␾ Max ␾ Min ␾ Range % Var 0±100 3.5 285 1.6 21 16.3 234 8.2 110 8.1 35.1 0±300 7.2 272 3.5 38 30.1 239 11.6 116 18.5 46.3 0±800 8.5 263 2.9 63 36.7 254 17.8 121 19.0 21.5 100±300 3.9 260 2.3 49 13.8 241 3.3 123 10.6 37.5 300±800 1.7 225 2.7 102 8.1 278 0.1 16 8.0 7.7

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 117

ventional procedure of Fourier analyzing the time series and retaining only the annual and semiannual harmonics (Table 3b). The annual range resulting from this calculation is somewhat smaller (19 Sv), and the minimum transport in April±May is larger (18 Sv). Addition of the third annual harmonic, which helps to resolve the sharp transition in late spring to early summer, leads to results very similar to those shown in Fig. 12. Vertical pro®les of the NBC transport in 50-m depth increments, for both the annual mean and the two extremes of the seasonal cycle, are shown in Fig. 13. Dur- ing the spring minimum of the NBC (Fig. 13b), most of the transport (8.3 Sv, or 70%) is contained in the top 100 m, and of that more than half occurs over the shelf (Fig. 13a). Below 100 m, the transport pro®le in the spring is nearly uniform with depth, with some indication of an intermediate maximum near 400±500 m. FIG. 12. Seasonal cycles of NBC transport for the top 300 m (dashed The transport increase between spring and fall is con- line), for the 300±800-m depth range (short dashed line), and for the tained entirely in the top 400 m and is largest through total upper ocean (0±800 m). These seasonal cycles were created by computing monthly means of the respective NBC transport series the thermocline (Fig. 13a). Hence, in addition to having after smoothing with a 90-day low-pass ®lter. Standard errors for the a large seasonal cycle of total transport near 4ЊN, the monthly mean total (0±800 m) transport are indicated at right. vertical structure of the NBC changes seasonally from a surface-trapped ¯ow in spring, to a deeper reaching ¯ow with substantial transport in the thermocline during estimated from running 3-month standard errors of the the fall. Similar seasonal changes in the vertical struc- difference between the 40-h and 90-d low-passed time ture of the NBC near 6ЊN were found by Colin et al. series (i.e., containing ¯uctuations in the mesoscale (1994). band), are approximately Ϯ6 Sv for the total transport and Ϯ4 Sv for the layer 0±300 m. The most prominent characteristic of the seasonal cycle is the sharp increase 5. Intraseasonal variability that occurs between May and July, seen in both the a. Upper ocean upper 300-m transport and total transport. For comparison with other published results from the region (e.g., Superimposed on the seasonal cycle are energetic S93) seasonal cycles were also calculated using the con- ¯uctuations on timescales of 1 to 3 months that often

FIG. 13. Pro®les of NBC transport in 50-m vertical increments: (a) annual mean transport pro®le, with the portion occurring over the shelf in black; (b) seasonal transport maximum in Jul±Aug (hatched) and minimum in Apr±May (black). The transport on the north Brazil shelf shown in (a) is assumed to be constant throughout the year (4.8 Sv; see text).

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 118 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

appear to have a wavelike character. Varying degrees of vertical coherence are seen among the surface, thermocline, and subthermocline layers for these ¯uctuations, which suggests that no single vertical mode structure is dominant in the region. In addition, there appears to be a distinct seasonality of certain wave or eddy features related to the spinup phase of the NBC in early summer. Here we discuss brie¯y the energy-containing timescales and give a ®rst-order description of their vertical structure and kinematic properties. Kinetic energy spectra for the 50-m and 300-m levels on moorings S1 and S2 illustrate the main differences between the surface and subsurface layers (Fig. 14). In the surface layer, the energy is concentrated in a spectral peak centered near 30 to 40 days, while at 300 m the energy is concentrated at longer periods ranging from 50 to 100 days. The same low-frequency peak with re- duced amplitude is also seen at the 300-m level on S3. While there is no record in the surface layer on S3, it is assumed that the higher-frequency ¯uctuations seen at S1 and S2 are also present there. A secondary peak at a period of about 30 days is also observed at the 300- m level on mooring S1 (Fig. 14b); however, the ¯uctuations in this period band are incoherent with those in the surface layer. The spectra at 800 m and deeper (not shown) show dominant timescales in the 50±100- day range similar to those at the 300-m level. Volume transport spectra for the three layers 0±100 m, 0±300 m, and 0±800 m (Fig. 14c) show increasing dominance by the lower-frequency ¯uctuations with increasing depth, as could be anticipated from the above results. The total 0±800-m transport has most of its variance between periods of 50 and 100 days, with a peak at 80 days. Thus, while the higher-frequency variations dominate in the surface layer, they do not con- tribute much to the overall transport variation of the NBC due to their limited vertical scale. Pro®les of eddy kinetic energy in two period bands, 25±40 days and 60±90 days, are shown in Fig. 15 to better illustrate the vertical structure in the dominant spectral bands. The surface trapping in the higher-frequency band is clearly evident at both moorings S1 and S2. The lower-frequency band is characterized by nearly uniform amplitude through the upper 400 m at S2, with some scatter, while at S1 the energy at 300 m is considerably larger than that in the surface layer. There is a distinct seasonal variation in the intensity of the 25±40-day ¯uctuations (Fig. 15c), with maximum energy occurring during late spring and summer during the acceleration phase of the NBC. Maxima in the 25± 40-day bandpassed energy at the 60-m level on S1 occur

←

FIG. 14. Kinetic energy spectra at moorings S1 and S2: (a) the surface layer (60 m at S1; 50 m at S2); (b) the subsurface layer (310 m at S1; 300 m at S2), and (c) transport spectra of the NBC in depth ranges 0±100 m, 0±300 m, and 0±800 m.

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 119

FIG. 15. Pro®les of eddy kinetic energy at moorings S1 and S2 in two period bands: (a) 25± 40 days and (b) 60±90 days; (c) time series of 25±40-day period bandpassed alongshore current variations at 60 m on mooring S1. in May±August of both 1990 and 1991, and a maximum pography, while those at S2 and S3 have a rotary char- is also found at the upper levels on S2 in May±August acter that is counterclockwise at S2 and clockwise at 1990 (not shown). This seasonal variation is essentially S3. These characteristics suggest a kinematic interpre- the same as that found by S93 near the equator; that is, tation of these events as anticyclonic eddies propagating near-surface currents there were dominated by vari- northwestward along the Brazil margin that are embed- ability with 25±35-day periodicity and with maximum ded in, or adjacent to, the NBC. energy occurring in late spring and summer. No evi- A presentation of the time series from Fig. 2 that dence was found for signi®cant seasonal modulation of supports this description is shown in Fig. 16. Here the longer period (60±90 day) ¯uctuations. the vector currents at 300 m are displayed in a spatial The kinematic properties of the 60±90-day ¯uctua- perspective diagram where the time axis is mapped tions are most clearly revealed by the measurements at into alongstream spatial distance using an assumed the 300-m level, since they are relatively uncontami- alongstream phase velocity. The technique is an ad- nated by the higher frequency surface motions. Cross- aptation of the ``frozen ®eld'' approximation ®rst in- spectra at this level show that these ¯uctuations are troduced by Taylor (1921). It can be seen in Fig. 16 laterally coherent between moorings S1, S2, and S3 (this that for a choice of alongstream propagation speed of can be seen visually in Fig. 2, particularly during the 8cmsϪ1 , the ¯uctuations at moorings S1, S2, and S3 latter half of the record). The sense of this variation is are consistent with the passage of nearly symmetric that simultaneous pulses of alongshore (northwestward) anticyclonic eddies with diameters of roughly 300± ¯ow at S1, and north or northwestward ¯ow at S2, occur 400 km. The choice of 8 cm sϪ1 for the alongstream with pulses of southeastward ¯ow at S3. The ¯uctua- propagation speed is arbitrary, since the moored array tions at S1 are nearly rectilinear and parallel to the to- provides no information on downstream phase delays,

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 120 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

FIG. 16. Spatial perspective diagram of currents at 50-m and 300-m nominal depth from the STACS-11 array, for the period May±Oct 1990. The time axis is mapped into an equivalent alongstream coordinate, as shown at upper right, using an assumed constant alongshore propa- Ϫ1 Ϫ1 gation speed of c0 ϭ 8kmd . The velocity scale is 100 cm s for the 50-m level and 50 cm sϪ1 for the 300-m level. The behavior of the currents is suggestive of anticyclonic eddies propagating northwestward along the continental margin, as outlined by the bold dashed lines. and was chosen simply because it makes the eddies occur in the subsurface layer. Our interpretation of appear roughly circular in shape. Signi®cantly in- this complex variability is that there are two separate creasing or decreasing the assumed phase speed scales of eddy motion, in terms of both vertical and would make the eddies appear more asymmetric, with temporal scales (and possibly also horizontal scale) their long axes pointing in the alongshore or cross- that are superimposed upon the NBC and propagate shore directions, respectively. northward along its offshore edge as essentially A similar display is shown in Fig. 16 for the 50-m closed eddylike features. The monthly timescale fea- level currents at moorings S1 and S2, and while there tures that dominate in the surface layer are clearly is no offshore record at mooring S3 at this level, it seasonal in nature, while those in the underlying ther- can be seen that the higher-frequency ¯uctuations that mocline layer are not seasonal but have a more ob- occur in early summer have much the same character. vious eddylike character during summer and fall when In particular, the picture at both levels seems to be the NBC is at its seasonal maximum. inconsistent with what one would expect for simple Another observation from Fig. 16 is that the bursts meandering of the NBC. This display also serves to of southeastward ¯ow at 300 m on S1 occur regularly emphasize the decoupling between the surface and at times just preceding, or between, the arrival of the subsurface layers; during mid-May to July, there are subsurface anticyclonic eddy features. This suggests that two eddy events passing in the surface layer while the prevailing mean current on the upper slope is south- only one event occurs in the subsurface layer. Simi- eastward at the 300-m level, counter to the surface NBC larly, the surface currents between August and Oc- ¯ow, and that it is periodically arrested or diverted off- tober do not show any evidence of the two events that shore by the passage of these eddies.

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 121 b. Deep ocean this context, the present study at 4ЊN is located downstream of the equator, where much of the thermocline The most prominent feature of the deep current vari- ¯ow is thought to be drawn off to feed the EUC, and ability is the intermittency of the DWBC ¯ow at depths upstream of the NBC retro¯ection where surface waters of 1400±1800 m on mooring S2, and the associated are drawn seasonally into the NECC. variability at the same depths on mooring S3. It can be The combined WESTRAX and AMASSEDS moored seen in Fig. 2 (and Table 1) that the eddy energy is current measurements show that the NBC has a large elevated at these depths relative to that at 800 m, and seasonal cycle at 4ЊN, with maximum transport in boreal also relative to the deeper levels on S3. The energy fall and minimum transport in boreal spring. The annual spectrum at S2 does not show a de®nable peak, while mean transport of the NBC at 4ЊN is approximately 26 that at S3 shows a peak near 60±90 days similar to that Sv, with an annual range of 23 Sv between the seasonal found in the thermocline layer on all moorings. Despite minimum in April±May (13 Sv) and seasonal maximum this similar periodicity, we ®nd no statistical correlation in July±August (36 Sv). The NBC varies from a shallow, between the deep ¯uctuations and those in the ther- coastally trapped ¯ow in the spring, with nearly half of mocline layer, rather, they appear to be related to me- its transport con®ned over the Brazilian shelf, to a andering or eddy variability spawned by the deep UN- broader, deeper-reaching ¯ow in the fall. The transport ADW jet along the boundary. in the surface layer (0±100 m) increases from 8 to 16 For example, during the September±December 1989 Sv between the spring minimum and fall maximum, period when the DWBC ¯ow near the boundary is weak, while that in the thermocline layer (100±300 m) in- the offshore ¯ow is predominantly southeastward and creases from less than 2 Sv to almost 14 Sv. Most of relatively strong, suggesting that the core of the UN- the seasonal transport variation is con®ned to the upper ADW jet had meandered offshore to near the location 300 m. of mooring S3. During the other period of weak bound- The cause of this large seasonal variation is assumed ary ¯ow in April±May 1990, a southeastward ¯ow of to be a dynamic adjustment to changes in wind forcing comparable strength is not observed offshore, but there over the tropical Atlantic, and both local and remote is a weak reversal at that time, which suggests that the wind stress forcing may play a role in the total response. DWBC may have meandered offshore but not as far as Recent studies of the response of western boundary cur- S3. In general, periods of strongest southeastward ¯ow rents to varying winds have generally focused on remote along the boundary are associated with periods of forcing aspects, including changes in the structure and strongest north or northwestward ¯ow offshore (statis- intensity of the interior wind stress curl distribution and tically the alongshore current variations at S2 and S3 remote coastal winds (Anderson and Corry 1985; Great- are out of phase). This suggests that the mean north- batch and Goulding 1989; BoÈning et al. 1991). Forcing westward recirculation found offshore at the UNADW of low-latitude western boundary currents by local and depth (Fig. 5) is enhanced when the DWBC is ¯owing remote winds has also been speci®cally studied with strongly southeastward along the boundary. This could most extensive application to the Somali Current regime involve changes in the large-scale recirculation pattern in the Indian Ocean (Lighthill 1969; Hurlburt and or could simply re¯ect the presence of smaller-scale Thompson 1976; Anderson and Rowlands 1976). cyclonic circulation features adjacent to the western It is now well established that, at midlatitudes, the boundary, perhaps generated by meandering and pinch- baroclinic response of western boundary currents to an- ing off of eddies from the DWBC itself. nual interior wind stress curl changes is small, and that Owing to the sparsity of direct observations at this the dominant response is contained in the barotropic level, we can only speculate on the details of these pro- mode (cf. Lee et al. 1996 for observational evidence in cesses and their causes. However, it appears that much the North Atlantic subtropical gyre). In particular, the of the variability may be plausibly explained by periodic application of a ``stationary'' Sverdrup model, in which offshore meandering of the DWBC and local eddy ef- the annual cycle of upper-ocean boundary currents fects rather than necessarily involving a large-scale vari- would simply follow the interior Sverdrup transport cy- ation of the through¯ow of the DWBC on seasonal or cle, has been shown to be inappropriate at midlatitudes other timescales. due to the slow westward propagation of baroclinic planetary waves there. However, at low latitudes, baro- 6. Summary and discussion clinic adjustment can occur rapidly due to the much faster westward propagation of baroclinic planetary a. Seasonal cycle of the NBC waves, including equatorially trapped modes. The grav- A central goal of the current meter observations car- est meridional mode baroclinic Rossby wave can cross ried out in WESTRAX was to document the structure the width of the equatorial Atlantic in a little over 100 and seasonality of the NBC at several key locations days (Philander and Pacanowski 1980), suggesting that along the Brazil margin and to determine its role in a lagged response to interior wind stress curl should be feeding the equatorial countercurrents and interhemi- a strong factor in the NBC's seasonal response. spheric through¯ow along the western boundary. Within A comparison of the observed NBC transport cycle

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 122 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

based on Hellerman and Rosenstein (1983) winds. COADS winds over the tropical Atlantic during the period of this study (1990±91) did not show any signi®cant differences in amplitude or phase of the Sverdrup transport or interior wind stress curl distribution relative to climatological winds. According to the stationary Sver- drup theory, the NBC would have a maximum northward transport of about 15 Sv in August, when the ITCZ is in its northernmost position and the anticyclonic equatorial gyre and NECC are well established north of the equator (Mayer and Weisberg 1993). A reversal to approximately 10 Sv of southward transport would follow in winter when the ITCZ retreats to the equator and the region near 4ЊN is under the in¯uence of the cyclonic tropical gyre. The AC model, which includes the basic linear dynamics of the annual adjustment, closely follows the onset and peak phase of the NBC predicted by stationary Sverdrup theory but differs in the absence of a broad winter minimum. It is obvious from Fig. 17a that neither of these wind-driven predictions reproduces the observed mean transport of the NBC, which is larger by 15±20 Sv. We attribute this ``excess'' transport primarily to a quasi-steady thermohaline component of the NBC, as discussed later. Two curves are shown in Fig. 17 for the AC model, including just the transport in the top layer, and the transport in both layers (i.e., over the total water column). A strict comparison with the observations is not possible because the upper layer in the AC model is only 100 m thick, and thus all of the baroclinic seasonal variation is con®ned to this layer, whereas the observed seasonal NBC transport variation extends to at least 300- m depth. The minor differences between the two curves illustrate that the seasonal transport variation is predominantly baroclinic and con®ned to the top layer. Comparison of either of these curves with the observed transport cycle shows remarkable agreement in the amplitude and phase of the annual cycle. In contrast to the Sverdrup cycle, a spring minimum in transport in April or May now occurs, consistent with the observations, and the maximum transport occurs in August or Sep- tember, perhaps slightly lagged from the observations. It therefore appears that the essential features of the FIG. 17. (a) Seasonal transport cycle of the NBC (reproduced from NBC's seasonal cycle can be captured by a linear wind- Fig. 12), compared with the inverse of the climatological Sverdup transport across 4ЊN derived from Hellerman and Rosenstein (1983) driven model. The main difference between the station- winds (HR), and with the NBC transport cycle at 4ЊN from the An- ary Sverdrup cycle and the linear model result, that is, derson and Corry (1985) model (AC). Results from the AC model the more gradual decay of the NBC in late fall and are shown for the top 100 m (dotted line) and for the total water winter, is probably attributable to a propagation delay column (dashed line). (b) Seasonal cycles of wind stress curl in the western (45ЊW), central (30ЊW), and eastern Atlantic (15ЊW) along of the remote wind stress curl response from the eastern 3ЊN, and local alongshore wind stress (positive northwest) in the NBC part of the basin (Fig. 17b). The small secondary peak region (3ЊN, 49ЊW), taken from Hellerman and Rosenstein (1983). in the AC model transport in December (Fig. 17a) could also be due to this effect. Nonlinear models produce a very similar annual response of the NBC at this latitude with the inverse of the climatological Sverdrup transport (Philander and Pacanowski 1986; H. Hurlburt 1995, per- at 4ЊN, and with a prediction of the NBC transport from sonal communication). the linear, two-layer model of Anderson and Corry Local wind forcing in the western boundary region (1985; hereafter AC model), is shown in Fig. 17a. Both may also be important in the NBC response and these the Sverdrup transport relation and model results are effects are implicitly contained in both linear and non-

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 123 linear models. Studies in the western equatorial Indian els indicates that a signi®cant component of the NBC Ocean have shown that the initial onset of the Somali is due to cross-equatorial ¯ow of warm waters from the Current during the southwest monsoon is caused by lo- South Atlantic associated with the Atlantic meridional cal alongshore winds, followed by a buildup of the in- overturning cell (MOC). The difference between the terior wind stress curl response over a period of several observed transport and the linear model is approxi- months (Schott et al. 1990; Anderson and Rowlands mately 15 Sv in the annual mean and fairly constant 1976). In the Atlantic, the annual migration of the ITCZ through the year, suggesting a nearly steady thermo- leads to the development of positive (northwestward) haline contribution that is in line with available esti- alongshore wind stress in the western equatorial region mates of the net export of NADW from the North At- that peaks in summer (Fig. 17b), similar to that of the lantic (Schmitz and McCartney 1993). The vertical interior wind stress curl. However, the onset of these structure of this warm return ¯ow is not readily obtain- coastal winds begins in May, about two months before able from the observations since it is combined with a the wind stress curl reverses in the western and central seasonally varying wind-driven component. However, Atlantic. This could possibly explain the rapid onset of if it is assumed that the vertical distribution of the MOC- the NBC in early summer and that there is no appre- related ¯ow is also nearly constant through the year, ciable lag between the interior wind stress curl and the then some indication of its vertical structure can be taken seasonal onset of the NBC. The magnitude of the re- from the NBC transport pro®le in spring (Fig. 13a). sponse to local winds is uncertain but some guidance During this season, the wind-driven transport of the can be drawn from model applications in the Somali NBC should be nearly zero, so that the observed trans- Current region. Hurlburt and Thompson (1976), for export pro®le should be primarily re¯ective of the MOC ¯ow structure. The transport pro®le in spring is domi- ample, show that the linear response of the equatorial nated by near-surface ¯ow with a secondary maximum ocean to northward cross-equatorial winds is symmetric below the thermocline near 500-m depth. This distri- about the equator and leads to a northward boundary bution is roughly consistent with the levels at which current in the near-surface layer, whose strength decays Schmitz and Richardson (1991) found modi®ed South away from the equator on the scale of the equatorial Atlantic water ¯owing into the southern Caribbean and deformation radius. An equilibrium response is reached Florida Current. Their results indicate a total interhemi- after about a week, which, for reasonable model param- spheric warm water transport of 13 Sv, comprising 7 eters, yields a transport response at a distance of 4Њ from Sv in the surface layer (Ͼ24ЊC), 5 Sv in the lower Ϫ2 the equator of approximately 1 Sv per 0.1 dyn cm of thermocline (7Њ±12ЊC), and a small amount (1 Sv) in alongshore wind stress. If similar dynamics hold for the the main thermocline (12Њ±24Њ C). The corresponding NBC, a transport contribution of perhaps 5 Sv may oc- transports from the April±May NBC transport pro®le cur due to local forcing in summer with initial onset in are approximately 8 Sv for the Ͼ24ЊC layer (0±100 m), May. This is relatively small compared to the O(25 Sv) 1 Sv for the 12Њ±24ЊC layer (100±200 m), and 3 Sv for annual variation in interior Sverdrup transport, but may the 7Њ±12ЊC layer (200±500 m), using climatological impact the phase and amplitude of the seasonal cycle. temperature pro®les for reference. An additional 1.5 Sv A more thorough understanding of the response of the occurs below 500 m at temperatures between 5Њ±7ЊC. NBC to the combined effects of local and remote forcing The pathways by which this thermohaline ¯ow tran- will probably require a dedicated model study. If local sits the equatorial Atlantic to enter the North Atlantic forcing is important, however, it is surprising that there subtropical gyre system remain a subject of considerable is no clear evidence for a seasonal cycle in the NBC debate. Schott et al.'s estimate of the mean transport of transport over the shelf, since one would expect most the NBC at the equator is 32 Sv, some 20 Sv larger than of the locally forced transport signal to be contained the annual mean Sverdrup transport there. This suggests there. It is possible that such a signal exists, but that it that the thermohaline ¯ow is concentrated in the NBC is simply not revealed by the sparse data available on as it crosses the equator and that it continues northward the shelf during boreal summer and fall. along the western boundary to at least 4ЊN. Below we In summary, we ®nd that the annual variation of the describe the seasonal pathways of this ¯ow and con- NBC transport is well described by a simple, linear, nections to zonal interior currents for three layers; the baroclinic model (Anderson and Corry 1985). The main surface layer (0±100), the thermocline layer (100±300 features of the NBC annual cycle can be explained in m), and the subthermocline layer (300±800 m). A sche- terms of a lagged response to Sverdup (wind stress curl) matic for the surface and thermocline layers is given in forcing in the interior, with perhaps a secondary con- Fig. 18, which summarizes recent results from long-term tribution by local forcing in the coastal region. moored current measurements (S93 and this study) and seasonal shipboard surveys of the region (Schott et al. 1995; Wilson et al. 1994). b. Thermohaline ¯ow contribution to the NBC and seasonal pathways c. The surface layer The large mean offset between the observed NBC In the surface layer (0±100 m), the NBC transport is transport at 4ЊN and that predicted by wind-driven mod- comparable between the equator (11.4 Sv: S93) and 4ЊN

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 124 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

Didden and Schott 1993; Richardson et al. 1994; Fra- tantoni et al. 1995), and offshore retro¯ection into the NECC followed by northward Ekman transport in the ocean interior (Mayer and Weisberg 1993). During the NBC maximum in summer and fall, most of the NBC surface ¯ow is thought to separate from the western boundary near 6ЊN and feed the eastward ¯owing NECC between 3Њ and 9ЊN (Richardson and Walsh 1986; Limeburner et al. 1995). The surface velocity maximum in the NECC occurs in August along both 28Њ and 38ЊW (Richardson et al. 1992; Richardson and Walsh 1986), and eastward ¯ow is present there from June though January. Reversal to a weak westward ¯ow is observed at both longitudes from February through May. Katz (1993) found a similar annual cycle for the transport of the NECC along 38ЊW, with a maximum transport of approximately 22 Sv from August to No- vember, and weak westward transport during April and May. The corresponding maximum NBC surface layer transport at 4ЊN is 16 Sv (July±November). Whether all of the NBC surface transport is required to feed into NECC during summer and fall is not clear, since some portion of the NECC appears to be supplied by westward ¯ow from the NEC that turns southward and eastward in the western Atlantic near 10ЊN. For example, Wilson et al. (1994, hereafter W94) found from a shipboard survey in August 1989 that the eastward transport across 44ЊW in the NECC was 24 Sv, composed of 16 Sv of water that had retro¯ected from the NBC and 8 Sv that had joined from the NEC. However, no evidence for through¯ow of the NBC along the western boundary was evident in the W94 study. This, together with the overall consistency of the above transport estimates, suggests that little direct leakage of NBC water occurs along the western boundary during the NBC±NECC maximum ¯ow season in summer and fall. Whatever leakage does take place at this time appears to occur in the form of rings that pinch off from the NBC retro- ¯ection beginning in late fall, which may account for FIG. 18. Transport schematics for the surface (0±100 m) and ther- 3±4 Sv of northward transport on an annualized basis mocline (100±300 m) layers during the minimum (April) and max- (Fratantoni et al. 1995). imum (August) phases of the NBC at 4ЊN. Seasonal transport values In spring, the surface layer ¯ow of the NBC has no for the NBC are taken from moored current measurements at the equator (Schott et al. 1993) and at 4ЊN (this study). Transport esti- apparent pathway into the interior, and it must be con- mates for the NECC and NEUC in August are from Wilson et al. cluded that this ¯ow continues northward along the (1994). Small differences in the transports shown along pathways western boundary. Speci®cally, during April±May, (e.g., for the along boundary surface through¯ow in April; shown as when all available measurements indicate westward sur- 10 Sv at the equator versus 8 Sv at 4ЊN) are within measurement uncertainty. Areas of signi®cant uncertainty in transports or pathways face layer ¯ow in the interior, our results suggest that are labeled by question marks. approximately 8 Sv of surface ¯ow continues northward along the boundary to join the westward NEC ¯ow entering the southern Caribbean Sea. Schott et al. (1995, (13.8 Sv), indicating only a small mean in¯ow from the hereafter S95) arrived at similar conclusions from their SEC in this latitude range. The seasonal cycle in the March 1994 survey of the western equatorial Atlantic; surface layer is somewhat larger at 4ЊN, varying over they found 14 Sv of surface ¯ow crossing the equator an annual range of approximately 8 Sv at 4ЊN versus at 44ЊW and only 2±4 Sv returning in an eastward coun- only 4 Sv at the equator. Possible pathways for north- tercurrent to feed the upper portion of the EUC. Thus, ward continuation of this surface ¯ow include a coastal they concluded that 10±12 Sv of South Atlantic water boundary current (Csanady 1985; Candela et al. 1992), must continue northward in the NBC during this season. rings shed from the NBC retro¯ection (Johns et al. 1990; This agrees well with S93's moored estimate of ap-

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 125 proximately 10 Sv of net surface transport across 44ЊW timately joins the EUC farther east or remains in the in winter and spring. The mechanism or combination of NEUC is not known and could not be determined by mechanisms that carry this western boundary through- the W94 study. Climatological maps of the thermocline ¯ow beyond 4ЊN have yet to be fully documented. Per- topography in the western tropical Atlantic presented haps half of the through¯ow could be carried by a con- by Molinari and Johns (1994) also suggest that sepa- tinuous coastal boundary current con®ned over the shelf ration of ¯uid from the thermocline of the NBC occurs (3±5 Sv based on this study) or on the upper continental farther south along the western boundary in spring than slope shoreward of the 1000-m isobath (Johns et al. in summer and fall. Their map for March (their Fig. 6) 1990). The remainder may be carried in NBC rings that shows water on the 10ЊC surface leaving the NBC be- continue to form as the retro¯ection subsides in late tween the equator and about 4ЊN, whereas during Sep- winter, with ring formation events having been recorded tember this occurs between 4Њ and 8ЊN. Thus, the pref- as late as April (Didden and Schott 1993; Fratantoni et erential loss of ¯uid from the thermocline layer of the al. 1995). Further study, particularly in the coastal re- NBC between the equator and 4ЊN during spring appears gion just north of the NBC retro¯ection, is needed to to be related to the more southerly location of the NBC resolve this issue. retro¯ection into the EUC/NEUC current systems during that season. Despite this apparent seasonal variation in the separation latitude, net northward leakage of ther- d. The thermocline layer mocline water from the NBC to the subtropics along Between depths of 100 to 300 m, an annual volume the western boundary appears to be small throughout ¯ux of approximately 5 Sv is lost from the NBC to the the year (Cochrane et al. 1979; W94; S95). interior between the equator and 4ЊN. S93 ®nd a mean Feeding of the NEUC or possibly the EUC by an transport for the thermocline layer (100±300 m) of 13 undercurrent ¯owing southeastward along the upper Sv with small (Ϯ1 Sv) seasonal variation, while at 4ЊN slope beneath the surface NBC and carrying waters of the mean transport in this layer is 8 Sv and ranges from Northern Hemisphere origin has also been suggested by 2 Sv in April±May to 14 Sv in July±August. Thus, the observations (J90; W94; Colin et al. 1994) and model loss of thermocline water from the NBC between the results (Schott and BoÈning 1991). The prevailing south- equator and 4ЊN appears to be strongly seasonal and eastward current observed at 300 m on mooring S1 (Fig. mainly occurs during winter and spring. 7b) is consistent with penetration of this undercurrent The thermocline ¯ow that is diverted to the interior to at least 4ЊN in spring. A sporadic signature of this could either feed the EUC on the equator or the NEUC undercurrent also seems to be present during summer between 2Њ and 5ЊN. Both of these currents are per- and fall between the passage of thermocline eddies manent features of the equatorial Atlantic circulation through the region (Fig. 16), even though the mean ¯ow (Cochrane et al. 1979), although the NEUC appears to at this time is northwestward. S93 ®nd no evidence for be weakest in boreal spring (op. cit.; Molinari and Johns this coastal undercurrent at the equator (see also S95), 1994). S93 concluded that most of the thermocline ¯ow indicating that it probably turns offshore into the NEUC passing through their NBC array along 44ЊW must even- rather than continuing to the equator to join the EUC. tually retro¯ect and merge with the EUC. S95's March As indicated in Fig. 18a, the magnitude of the NEUC 1994 survey clearly supports this conclusion, as all 14 and the relative amount of supply from the NBC and Sv of thermocline water found in the NBC at that time NEC during spring still needs to be established. retro¯ected within 2Њ of the equator to merge into the EUC. Only a weak signature of the NEUC (approxi- e. The subthermocline layer mately 2 Sv) was found between 3Њ and 4ЊN in that survey, which could not be clearly traced back to the Transport in the subthermocline (300±800 m) layer NBC along the western boundary. However, during Au- of the NBC at 4ЊN is estimated to be 4.4 Sv, although gust 1989, W94 found the thermocline ¯ow in the NBC it is not well resolved in this study and is mainly de- to continue northward to approximately 6ЊN before re- termined by vertical interpolation between instruments tro¯ecting into the interior, where it fed into the NEUC at depth 300 or 400 m and 800 m. The transport in this as a subsurface branch of the NBC retro¯ection. That layer has no obvious seasonal cycle but rather is dom- the transport in the thermocline layer of the NBC at the inated by large intraseasonal ¯uctuations. Estimates of equator (S93) and 4ЊN (this study) is nearly equal during the transport crossing the equator in the NBC at similar summer is consistent with this description and suggests depths range from 3.7 to 11.0 Sv for the layer 300±600 that whatever ¯uid has been drawn off into the EUC in m, with a mean of 6.7 Sv from three cruises (S93), and the vicinity of the equator during summer occurs up- 12.4 Sv for the ␴␪ ϭ 26.8 (ϳ300 m)±1000 m layer from stream of S93's array location along 44ЊW. In contrast one additional cruise (S95). No moored transport esti- to the S95 spring survey, the NEUC during August 1989 mate was made for this layer by S93 at the equator. was well developed, with a total eastward transport of South of the equator, at 5Њ±10ЊS, S96 estimated a trans- approximately 23 Sv across 44ЊW (W94). Whether the port of approximately 6 Sv in the subthermocline (␴␪ thermocline water retro¯ecting near 6ЊN in summer ul- ϭ 26.8±1000 m) layer of the NBC. While the above

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 126 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28 estimates are mostly larger than the mean transport ob- the same dominant (monthly) timescale in the surface tained at 4ЊN, they are well within the range of observed layer, with the same seasonal modulation, they did not variability there, and it is not possible to determine if note a signi®cant peak at 60±90 days in the thermocline, a signi®cant change occurs between the equator and and their transport time series for the NBC was entirely 4ЊN. Interaction of the subthermocline ¯ow in the NBC dominated by the monthly timescale ¯uctuations in the with the equator, including the Antarctic Intermediate surface layer. Fluctuations near 8ЊN at the northern limit Water (AAIW), appears to be complex, with evidence of the NBC's seasonal retro¯ection show, on the other for eastward ¯ow in off-equatorial countercurrents bor- hand, a single dominant timescale of 40±60 days, which dering the equator and westward return ¯ow in an equa- is located in between the two period bands of energetic torial intermediate current below the EUC (S95). Sub- variability found here. Moreover, the ¯uctuations near stantial upwelling and conversion of intermediate water 8ЊN are vertically coherent from the surface to 1000 m is also indicated in the equatorial Atlantic; for example, and have a fairly simple vertical mode structure similar Roemmich (1983) found a conversion of 3 Sv of inter- to that of the ®rst baroclinic dynamical mode for the mediate water into lower thermocline water between 8ЊS region (Johns et al. 1990). These regional differences and 8ЊN, while Schmitz and McCartney (1993) hypoth- in the fundamental properties of the ¯uctuations are puz- esized a conversion of 5 Sv of intermediate water to zling and do not easily lend themselves to a uni®ed lower thermocline (7Њ±12ЊC) water near the equator, description of the evolving eddy ®eld along the NBC's which is transported northward from the equator in the path. What seems to be clear from the data is that the NBC. The weak transport signal at AAIW depths in the monthly timescale ¯uctuations that dominate near the NBC at 4ЊN may be re¯ective of this conversion process. equator lose their importance as one progresses north- As noted at the outset of this section, the MOC-related westward in the NBC and gradually give way to lower- through¯ow of the NBC inferred from this study is frequency ¯uctuations with a deeper-reaching modal largely consistent with Schmitz and Richardson's (1991) structure. The energy source for the monthly ¯uctua- ®nding that the dominant contributions to interhemi- tions is thought to be primarily a shear instability of the spheric exchange occur in the surface (Ͼ24ЊC) and low- SEC/NECC zonal current system in the interior, which er thermocline (7Њ±12ЊC) layers, rather than directly in produces large ¯uctuations in meridional velocity at this the intermediate water layer. periodicity beginning in May or June (Weisberg and Weingartner 1988) and has characteristics similar to mixed Rossby±gravity (Yanai) waves with westward f. Intraseasonal variability phase propagation and eastward energy propagation. Variability of the NBC near 4ЊN on timescales shorter The appearance of these ¯uctuations in the western than seasonal is found to be concentrated in two spectral equatorial boundary layer shortly following their gen- bands that are different in the surface and subsurface eration in the interior is suggestive of westward advec- layers. In the surface layer, the variability is dominated tion of this wave energy in the SEC and then northward by ¯uctuations of roughly monthly timescale (25±40 along the boundary in the NBC. The source of energy days), while below the surface layer the characteristic for the lower-frequency ¯uctuations and eddies north of timescales are longer, between 60 and 90 days. The the equator in the NBC is not yet clear and could be transport variability of the NBC tends to be dominated related either to a local instability of the NBC, westward by these lower-frequency ¯uctuations due to the strongly propagation of Rossby waves from the interior along surface-trapped nature of the higher-frequency ¯uctu- the latitude of the NECC (McClean and Klinck 1995), ations, even though the mean NBC transport itself is or propagation of eddylike features northward along the concentrated in the surface layer. coast from the equatorial region (Carton 1992). One new The monthly timescale ¯uctuations exhibit a distinct insight that can be drawn from this study is that the seasonal modulation in their amplitude, being most en- shedding of anticyclonic rings from the NBC retro¯ec- ergetic in early summer during the onset phase of the tion may not be the result of a completely localized NBC. The lower-frequency ¯uctuations in the thermo- process, as previously thought (Johns et al. 1990), but cline layer do not show a signi®cant seasonal variation that precursors of these eddies are present in the NBC and are present throughout the year. The kinematic prop- upstream of the retro¯ection, which may serve as a cat- erties of the ¯uctuations in both layers, but particularly alyst for the eddy-shedding process at the NBC retro- those of the subsurface ¯uctuations during summer and ¯ection. Analysis of satellite observations, particularly fall, are suggestive of the propagation of large (diameter altimetry data, may be able to shed further light on the approximately 300±400 km) anticyclonic eddies north- dynamics of the eddy ®eld in this region. westward along the continental margin that are attached For the deep ¯ow, we can only make limited conclu- to the offshore edge of the NBC. sions at this time due to the inadequacy of the present The intraseasonal variability near 4ЊN is notably dif- array measurements in resolving the DWBC structure. ferent in several ways from that seen near the equator The dominant timescale of variability in the deep layers (S93) and also farther north in the NBC retro¯ection is 60±90 days, similar to that found in the thermocline, region (Johns et al. 1990). For example, while S93 found but it is not coherently related to the upper-ocean ¯uc-

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC JANUARY 1998 JOHNS ET AL. 127 tuations. This timescale appears to be ubiquitous in the Beardsley, R. C., J. Candela, R. Limeburner, W. R. Geyer, S. J. Lentz, B. M. Castro, D. Cacchione, and N. Carneiro, 1995: The M tide deep layers of the tropical Atlantic, and its possible 2 on the Amazon Shelf. J. Geophys. Res., 100, 2283±2319. causes are discussed elsewhere (S93; Whitehead and BoÈning, C. W., R. DoÈscher, and H. J. Isemer, 1991: Monthly mean Worthington 1982; Johns et al. 1993). In the present wind stress and Sverdrup transports in the North Atlantic: A case these ¯uctuations appear to be related to mean- comparison of the Hellerman±Rosenstein and Isemer±Hasse cli- dering or eddy variability associated with the upper core matologies. J. Phys. Oceanogr., 21, 221±235. of the DWBC that ¯ows equatorward along the western Brown, W. S., W. E. Johns, K. D. Leaman, J. P. McCreary, R. L. Molinari, P. L. Richardson, and C. Rooth, 1992: A Western Trop- boundary at depths of 1400±2000 m. The mean ¯ows ical Atlantic Experiment (WESTRAX). Oceanography, 5(1), 73± observed at this level near the western boundary, and 77. in the immediate offshore region, are consistent with Bruce, J. G., J. L. Kerling, and W. H. Beatty III, 1985: On the North the presence of a mean cyclonic recirculation in the Brazil eddy ®eld. Progress in Oceanography, Vol. 14, Pergamon, 57±63. western basin at this latitude just seaward of the DWBC Candela, J., R. C. Beardsley, and R. Limeburner, 1992: Separation (Richardson and Schmitz 1993; McCartney 1993). Vari- of tidal and subtidal currents in ship-mounted acoustic Doppler ations in the strength of the upper DWBC along the current pro®ler (ADCP) observations. J. Geophys. Res., 97(C1), boundary appear to be related in part to the presence of 769±788. transient cyclonic eddies offshore of the DWBC, which Carton, J. A., 1992: Tropical Atlantic Ocean eddies collide with the coast of South America (abstract). Eos Trans. Amer. Geophys. are embedded in the western reaches of this mean re- Union, 73(51), (Ocean Sciences Meeting Suppl.), 22. circulation. Whether these eddies are generated by a Cochrane, J. D., F. J. Kelly, and C. R. Olling, 1979: Subthermocline local instability of the DWBC or are imported into the countercurrents in the western equatorial Atlantic Ocean. J. region along the DWBC from farther north is unknown Phys. Oceanogr., 9, 724±738. and is a topic for future study. Colin, C., B. Bourles, R. Chuchla, and F. Dangu, 1994: Western boundary current variability off French Guiana as observed from moored current measurements. Oceanol. Acta, 17, 345±354. Acknowledgments. Support for the STACS-11 mea- Csanady, G. T., 1985: A zero potential vorticity model of the North surements was provided by the National Oceanic and Brazilian Coastal Current. J. Mar. Res., 43, 553±579. Atmospheric Administration through Cooperative Didden, N., and F. Schott, 1993: Eddies in the North Brazil Current retro¯ection region observed by Geosat altimetry. J. Geophys. Agreement NA85-WC-H-06134 with the Cooperative Res., 98, 20 121±20 131. Institute for Marine and Atmospheric Studies (CIMAS). DuÈing, W., and F. Schott, 1978: Measurements in the source region The AMASSEDS physical oceanographic program was of the Somali Current during the monsoon reversal. J. Phys. supported by the National Science Foundation through Oceanogr., 8, 278±289. Grants OCE88-12917 and OCE91-15712, and by the Fine, R. A., and R. L. Molinari, 1988: A continuous deep western boundary current between Abaco (26.5ЊN) and Barbados (13ЊN). Funda cËaÄo de Amparo aÁ Pesquisa do Estado de SaÄo Deep-Sea Res., 35(9), 141±1450. Paulo through Grant FAPESP 89/3508-4. Additional Flagg, C. N., R. L. Gordon, and S. McDowell, 1986: Hydrographic support for preparation of this paper was provided by and current observations on the continental slope and shelf of Grant OCE-93-13671 from the National Science Foun- the western equatorial Atlantic. J. Phys. Oceanogr., 16, 1412± dation (R.B., J.C., R.L., and B. Castro), and from Con- 1429. Fratantoni, D. M., W. E. Johns, and T. L. Townsend, 1995: Rings of tract N00014-89-J1139 with the Of®ce of Naval Re- the North Brazil Current: Their structure and behavior inferred search (W.J.). Rainer Zantopp and David Fratantoni of from observations and a numerical simulation. J. Geophys. Res., the University of Miami assisted in the data analysis 100(C6), 10 633±10 654. and preparation of ®gures. Doug Wilson of NOAA/ Geyer, W. R., R. C. Beardsley, S. J. Lentz, J. Candela, R. Limeburner, AOML kindly provided us with shipboard ADCP results W. E. Johns, B. M. Castro, and I. D. Soares, 1996: Physical oceanography of the Amazon Shelf. Contin. Shelf Res., 16(5/6), across the STACS-11 moored section, and we thank 575±616. him, Fritz Schott, Bob Molinari, and Bernard Bourles Gordon, A. L., 1986: Interocean exchange of thermocline water. J. for helpful comments and discussion on the paper. Geophys. Res., 91, 5037±5046. , R. F. Weiss, W. M. Smethie Jr., and M. J. Warner, 1992: Ther- mocline and intermediate water communication between the South REFERENCES Atlantic and Indian Oceans. J. Geophys. Res., 97, 7223±7240. Greatbatch, R. J., and A. Goulding, 1989: Seasonal variations in a Akima, H., 1970: A new method of interpolation and smooth curve linear barotropic model of the North Atlantic driven by the Hell- ®tting based on local procedures. J. Assoc. Comput. Mach., 17, erman and Rosenstein wind stress ®eld. J. Phys Oceanogr., 19, 589±602. 572±595. Alessi, C. A., S. L. Lentz, R. C. Beardsley, and B. M. Castro, 1992: Hellerman, S., and M. Rosenstein, 1983: Normal monthly wind stress A Multidisciplinary Amazon Shelf SEDiment Study (AMAS- over the world ocean with error estimates. J. Phys. Oceanogr., SEDS): Physical Oceanography Moored Array Component. Tech 13, 1093±1104. Rep. 92-36, 87 pp., [Available from Woods Hole Oceanographic Hurlburt, H. E., and J. D. Thompson, 1976: A numerical model of Institute, Woods Hole, MA 02543.] the Somali Current. J. Phys. Oceanogr., 6, 646±664. Anderson, D. L. T., and P. B. Rowlands, 1976: The Somali Current Johns, W. E., 1988: Near-surface current measurements in the Gulf response to the southwest monsoon: The relative importance of Stream using an upward-looking acoustic Doppler current pro- local and remote forcing. J. Mar. Res., 34, 395±417. ®ler. J. Atmos. Oceanic Technol., 5, 602±613. , and R. A. Corry, 1985: Seasonal transport variations in the , T. N. Lee, F. A. Schott, R. J. Zantopp, and R. H. Evans, 1990: Florida Straits: A model study. J. Phys. Oceanogr., 15, 773± The North Brazil Current retro¯ection: Seasonal structure and 786. eddy variability. J. Geophys. Res., 95(C12), 22 103±22 120.

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC 128 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 28

, D. M. Fratantoni, and R. J. Zantopp, 1993: Deep western bound- , and W. J. Schmitz Jr., 1993: Deep cross-equatorial ¯ow in the ary current variability off northeastern Brazil. Deep-Sea Res., Atlantic measured with SOFAR ¯oats. J. Geophys. Res., 98(C5), 40, 293±310. 8371±8387. Katz, E., 1993: An interannual study of the North Equatorial Coun- , S. Arnault, S. Garzoli, and J. G. Bruce, 1992: Annual cycle of tercurrent. J. Phys. Oceanogr., 23, 116±123. the Atlantic North Equatorial Countercurrent. Deep-Sea Res., Lee, T. N., W. E. Johns, R. J. Zantopp, and E. Fillenbaum, 1996: 39(6), 997±1014. Moored observations of volume transport and variability in west- , G. E. Hufford, R. Limeburner, and W. S. Brown, 1994: North ern boundary currents of the subtropical North Atlantic at Brazil Current retro¯ection eddies. J. Geophys. Res., 99(C3), 26.5ЊN. J. Phys. Oceanogr., 26, 446±466. 5081±5093. Lentz, S. J., 1995a: The Amazon River plume during AmasSeds: Rintoul, S., 1991: South Atlantic interbasin exchange. J. Geophys. Subtidal current variability and the importance of wind forcing. Res., 96, 2675±2692. J. Geophys. Res., 100, 2377±2390. Roemmich, D., 1983: The balance of geostrophic and Ekman transports , 1995b: Seasonal variations in the Amazon plume structure in- in the tropical Atlantic Ocean. J. Phys. Oceanogr., 13, 1534±1539. ferred from historical hydrographic data. J. Geophys. Res., 100, , and C. Wunsch, 1985: Two transatlantic sections: Meridional 2391±2400. circulation and heat ¯ux in the subtropical North Atlantic Ocean. , and R. Limeburner, 1995: The Amazon River plume during Deep-Sea Res., 32, 609±664. AmasSeds: Spatial characteristics and salinity variability. J. Geo- Schmitz, W., Jr., and P. Richardson, 1991: On the sources of the phys. Res., 100, 2355±2375. Florida Current. Deep-Sea Res., 38 (Suppl.), S379±S409. Lighthill, M. J., 1969: Dynamic response of the Indian Ocean to the , and M. S. McCartney, 1993: On the North Atlantic circulation. onset of the southwest monsoon. Philos. Trans. Roy. Soc. Lon- Rev. Geophys., 31(1), 29±49. don, A265, 45±92. Schott, F. A., and C. W. BoÈning, 1991: The WOCE Model in the Limeburner, R., R. C. Beardsley, I. Soares, S. Lentz, and J. Candela, western equatorial Atlantic: Upper layer circulation. J. Geophys. 1995: Lagrangian ¯ow observations of the Amazon River discharge Res., 96(C4), 6993±7004. into the North Atlantic. J. Geophys. Res., 100(C2), 2401±2415. , J. C. Swallow, and M. Fieux, 1990: The Somali Current at the Lukas, R., E. Firing, P. Hacker, P. L. Richardson, C. A. Collins, R. equator: Annual cycle of currents and transports in the upper Fine, and R. Gammon, 1991: Observations of the Mindanao 1000 m and connection to neighbouring latitudes. Deep-Sea Res., Current during the western equatorial Paci®c Ocean circulation 37, 1825±1848. study. J. Geophys Res., 96(C4), 7089±7104. , J. Fischer, J. Reppin, and U. Send, 1993: On mean and seasonal Mayer, D. A., and R. H. Weisberg, 1993: A description of COADS currents and transports at the western boundary of the equatorial surface meteorological ®elds and the implied Sverdrup transports Atlantic. J. Geophys. Res., 98(C8), 14 353±14 368. , L. Stramma, and J. Fischer, 1995: The warm water in¯ow into for the Atlantic Ocean from 30ЊSto60ЊN. J. Phys. Oceanogr., the western tropical Atlantic boundary regime, spring 1994. J. 23, 2201±2221. Geophys. Res., 100(C12), 24 745±24 760. McCartney, M. S., 1993: Crossing of the equator by the deep western Stramma, L., 1991: Geostrophic transport of the south equatorial boundary current in the western Atlantic Ocean. J. Phys. Ocean- current in the Atlantic. J. Mar. Res., 49, 281±294. og., 23, 1953±1974. , J. Fischer, and J. Reppin, 1995: The North Brazil Undercurrent. McClean, J. L., and J. M. Klinck, 1995: Description and vorticity Deep-Sea Res. Part I, 42, 773±795. analysis of 50-day oscillations in the western tropical region of Swallow, J. C., F. A. Schott, and M. Fieux, 1991: Structure and the CME Model. J. Phys. Oceanogr., 25, 2498±2517. transport of the East African coastal current. J. Geophys. Res., Metcalf, W., and M. Stalcup, 1967: Origin of the Atlantic Equatorial 96(C12), 22 245±22 257. Undercurrent. J. Geophys. Res., 72, 4959±4975. Taylor, G. I., 1921: Diffusion by continuous movements. Proc. Lon- Molinari, R. L., and E. Johns, 1994: Upper layer temperature structure don Math. Soc., 2, 196±211. of the western tropical Atlantic. J. Geophys. Res., 99(C9), Weisberg, R. H., and T. J. Weingartner, 1988: Instability waves in the 18 225±18 233. equatorial Atlantic Ocean. J. Phys. Oceanogr., 18, 1641±1657. , R. A. Fine, and E. Johns, 1992: The deep western boundary Whitehead, R. H., and V. Worthington, 1982: The ¯ux and mixing current in the tropical North Atlantic Ocean. Deep-Sea Res., rates of Antarctic Bottom Water within the North Atlantic. J. 39(11/12), 1967±1984. Geophys. Res., 87, 7903±7924. Nittrouer, C. A., D. J. DeMaster, A. G. Figueiredo, and J. M. Rome, Wilson, W. D., E. Johns, and R. L. Molinari, 1994: Upper layer 1991: AmasSeds: An interdisciplinary investigation of a com- circulation in the western tropical North Atlantic Ocean during plex coastal environment. Oceanography, 4, 3±7. August 1989. J. Geophys. Res., 99(C11), 22 513±22 523. Philander, S. G. H., and R. C. Pacanowski, 1980: The generation of Wyrtki, K., L. Magaard, and J. Hager, 1976: Eddy energy in the equatorial currents. J. Geophys. Res., 85(C2), 1123±1136. oceans. J. Geophys. Res., 81, 2641±2646. , and , 1986: A model of the season cycle in the tropical Zantopp, R. J., W. E. Johns, and T. N. Lee, 1993: Moored current Atlantic Ocean. J. Geophys. Res., 91, 14 192±14 206. meter observations off Northeast Brazil (STACS-11 Array). Richardson, P. L., and D. Walsh, 1986: Mapping climatological sea- RSMAS, University of Miami, Tech. Rep., 93-005, 56 pp. sonal variations of surface currents in the tropical Atlantic using [Available from Dr. W. E. Johns, RSMAS, University of Miami, ship drifts. J. Geophys. Res., 91, 10 537±10 550. 4600 Rickenbacker Cswy., Miami, FL 33149.]

Unauthenticated | Downloaded 09/25/21 07:05 PM UTC