North Brazil Current Ring Generation and Evolution Observed with Seawifs*

1058 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

North Brazil Current Ring Generation and Evolution Observed with SeaWiFS*

DAVID M. FRATANTONI AND DEBORAH A. GLICKSON Department of Physical Oceanography, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

(Manuscript received 25 April 2001, in ®nal form 22 August 2001)

ABSTRACT The earth's largest oceanic rings are formed by the retro¯ecting North Brazil Current (NBC) near 8ЊN in the western tropical Atlantic. The NBC ¯ows northward across the equator and past the mouth of the Amazon River entraining river-in¯uenced shelf water along its nearshore edge. Enhanced phytoplankton production associated with the nutrient-rich Amazon discharge results in near-surface chlorophyll gradients that delineate the trajectory of the retro¯ecting NBC. These large-scale gradients, visible from space using Sea-viewing Wide Field-of-view Sensor (SeaWiFS) ocean color imagery, enable visualization of NBC rings during the initial phases of their evolution and northwestward translation. Observations of 18 NBC rings identi®ed between September 1997 and September 2000 are summarized. Six rings formed each year. Although nearly circular at formation the rings quickly deformed as they translated at speeds near 15 cm sϪ1 toward the Caribbean Sea. Typical core radii of rings near 55ЊW were 100 km and 150 km in the across- and alongshore dimensions, respectively. The contribution of each ring to intergyre mass transport (1.0 Ϯ 0.4 Sv) was estimated using SeaWiFS derived surface areas and an estimate of vertical penetration (600 m) based on in situ tracer observations. Several rings were observed (using satellite-tracked surface drifters in combination with SeaWiFS imagery) to violently collide with the Lesser Antilles. At least one ring maintained an organized circulation while passing directly over the island of Barbados.

1. Introduction are destroyed through interaction with abrupt topography in the vicinity of the Lesser Antilles. Locally, NBC The North Brazil Current (NBC) is an intense western rings and their ®lamentary remains episodically disrupt boundary current and the dominant surface circulation surface circulation patterns in the eastern Caribbean, feature in the western tropical Atlantic Ocean. The NBC impact the distributions of salinity and icthyoplankton separates from the South American coastline at 6Њ±8ЊN and curves back on itself (retro¯ects) to feed the east- (e.g., Kelly et al. 2000; Cowen and Castro 1994; Borstad ward North Equatorial Countercurrent (NECC) and 1982), and pose a physical threat to expanding offshore close the anticyclonic (clockwise) wind-driven equa- oil and gas exploration on the South American conti- torial gyre (Fig. 1). The retro¯ection of the NBC is nental slope. Globally, the ®ve to six rings generated dynamically similar to the Agulhas Current south of annually by the equator-crossing NBC are responsible Africa (e.g., Lutjeharms 1996) and, like the Agulhas, for up to one-third of the equatorial-to-subtropical mass the NBC occasionally curves back upon itself so far as transport associated with the upper limb of the Atlantic to pinch off large warm-core vortices (Johns et al. 1990). meridional overturning circulation (MOC), a funda- These NBC rings, which can exceed 450 km in overall mental component of the earth climate system (Fratan- diameter and 2000 m in vertical extent, swirl anticy- toni et al. 1995, 1999, 2000; Goni and Johns 2001). clonically at speeds approaching 100 cm sϪ1 while trans- Due to their geographic location NBC rings are par- lating northwestward toward the Caribbean on a course ticularly dif®cult to investigate using satellite remote parallel to the South American coastline (Didden and sensing techniques. Rings shed from subtropical bound- Schott 1993; Richardson et al. 1994; Fratantoni et al. ary currents (e.g., the Gulf Stream, Kuroshio, Agulhas 1995, 1999). After translating for 3±4 months the rings Current, East Australian Current) are often discernable in satellite infrared imagery due to the contrast between their (warm or cold) core temperature and that of the * Contribution Number 10462 of the Woods Hole Oceanographic Institution. surrounding environment (e.g., Brown et al. 1983). Rel- atively weak surface temperature gradients in the western tropical Atlantic warm pool (e.g., Servain and Lukas Corresponding author address: Dr. David M. Fratantoni, Woods 1990) make sea surface temperature (SST) imagery in- Hole Oceanographic Institution, Dept. of Physical Oceanography, Woods Hole, MA 02543. effective for NBC ring identi®cation. Similarly, the rel- E-mail: [email protected] atively small sea surface height (SSH) signature asso-

᭧ 2002 American Meteorological Society

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1059

FIG. 1. Cartoon depicting the major upper-ocean circulation features in the western tropical Atlantic. The South Equatorial Current (SEC), North Brazil Current (NBC), and North Equatorial Countercurrent (NECC) close the wind-driven equatorial gyre. NBC rings provide a mechanism for mass and tracer transport from this gyre through the tropical gyre and into the southern extent of the North Atlantic subtropical gyre. ciated with the azimuthal velocities of the low-latitude trations and enhanced biological productivity. A plume NBC rings are dif®cult (but not impossible: see Didden of phytoplankton-rich, high-chlorophyll water is ad- and Schott 1993) to distinguish from an energetic back- vected northwestward along the nearshore edge of the ground eddy ®eld using satellite altimetry. Recently, NBC and into the interior as the NBC retro¯ects into Goni and Johns (2001) assembled a census of NBC rings the North Equatorial Counter Current (NECC).1 Re- using TOPEX/Poseidon SSH anomaly data merged with motely sensed surface chlorophyll measurements de- a mean ®eld derived from historical hydrography. While rived from ocean color observations reveal ®laments of this approach appears promising, the relatively low tem- highly productive Amazon-in¯uenced water adjacent to poral (10 days) and spatial (2Њ±3Њ longitude) measure- and surrounding relatively lifeless midocean water. This ment resolution inhibits detailed altimetric study of contrast permits visualization of the cyclic advance and NBC ring structure and evolution from altimetry alone. retreat of the NBC retro¯ection and the accompanying Pauluhn and Chao (1999) combined TOPEX altimetry formation of pinched-off NBC rings (Fig. 2). with an eddy-resolving numerical simulation and were In this article we present new observations detailing able to identify a number of translating NBC rings. the generation and evolution of NBC rings using three Because of the dif®culty of remote observation, NBC years of Sea-viewing Wide Field-of-view Sensor rings were not ``discovered'' until the late 1980s. Prior (SeaWiFS) ocean color measurements. We also present to this time it was generally understood that the western and make ancillary use of surface drifter trajectories and tropical Atlantic is an energetic and eddy-rich environ- hydrographic observations resulting from a recent Na- ment (e.g., Bruce et al. 1985). However, the essential tional Science Foundation±sponsored regional ®eld pro- evidence that discrete, westward-translating rings are gram. The acquisition, processing, and analysis of the shed by the retro¯ecting NBC resulted from the analysis SeaWiFS data is described in section 2. Our results, of Coastal Zone Color Scanner imagery (Johns et al. including a summary of ring characteristics and case 1990). As reported earlier by Muller-Karger et al. studies illustrating the genesis and demise of three NBC (1988), ocean color imagery of the western tropical At- rings, are presented in section 3. A brief discussion of lantic reveals a unique tracer of ocean circulation. The Amazon River discharges onto the continental shelf of 1 See Geyer and Beardsley (1995) and accompanying reports for equatorial Brazil resulting in elevated nutrient concen- a recent summary of Amazon River out¯ow plume observations.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1060 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

FIG. 2. A SeaWiFS chlorophyll a image depicting the NBC retro¯ection and a recently separated NBC ring. Relatively lifeless midocean water (dark blue; within the retro¯ection and the ring) contrasts sharply with the highly productive waters in¯uenced by the nutrient-rich Amazon out¯ow. Note also the plume of high-chlorophyll water spreading northwestward from the mouth of the Orinoco River. The ability to track NBC rings with SeaWiFS degrades signi®cantly in this region. these results is provided in section 4, and our conclu- 7-day composite images centered every fourth day. The sions are summarized in section 5. compositing algorithm retained the lowest value from each grid cell of the component images.2 The 7-day composite period is short compared to the 50±60-day 2. Data and methods period of NBC ring generation, but is approximately SeaWiFS measures chlorophyll a and water leaving one-half the rotation period of a translating NBC ring. radiances at six wavelengths (Hooker et al. 1992; Based on examination of several representative com- McClain et al. 1998). We obtained daily global ®elds posites and their component images we do not believe of SeaWiFS level-3 chlorophyll a data from the Goddard the smoothing and smearing of spatial features inherent Distributed Active Archive Center at the NASA/God- in the compositing process signi®cantly impacts the re- dard Space Flight Center for the period September sults presented herein. 1997±September 2000. The level-3 data product con- Each of the 265 composite images was manually an- sists of calibrated, atmospherically corrected data cor- alyzed to determine the position of the tracer fronts responding to a period of 1 day and stored in a global, delineating the onshore edge of the NBC, the southern equal-area grid with cells approximately 81 km2. From edge of the NECC, and the circumference of any trans- each daily global ®eld we extracted a regional sub®eld lating NBC rings. The position of the northwest corner encompassing the western tropical Atlantic and the east- of the NBC retro¯ection was recorded for each image, ern Caribbean Sea. The sole use of SeaWiFS imagery as were the center and major and minor diameters of in the present study is for identi®cation of near-surface any NBC rings present. Time series were constructed property gradients. Further details regarding the of the ring center position and along- and across-shore SeaWiFS instrument, calibration methodologies, and diameters for each individual ring. The time difference data processing can be found in Hooker et al. (1992) between the observed separation of a ring from the ret- and McClain et al. (1998). ro¯ection and the passage of the ring center through the Due to the intense convection associated with the 55ЊW meridian was used to compute northward and migrating intertropical convergence zone (ITCZ) the western tropical Atlantic is often obscured by clouds. Individual daily images in the study region ranged from 2 Unlike sea surface temperature imagery in which the coldest pix- cloud-free to 95% obscured with signi®cant daily and els generally signify clouds, the SeaWiFS level 3 imagery is distributed with cloud values set to an arbitrarily high ¯ag value. Hence a seasonal variation. To minimize the impact of occasional multiday composite that selects the lowest chlorophyll value elimi- cloudiness on our ability to discern large-scale spatial nates cloud-¯agged pixels in much the same way a warmest-pixel structures we generated a sequence of 265 overlapping SST compositing algorithm would function.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1061

TABLE 1. Rings identi®ed and tracked using SeaWiFS. Date tracked West velocity North velocity Speed Ring Date formed Date passed 55ЊW with SeaWiFS (cm sϪ1) (cm sϪ1) (cm sϪ1) A 10 Oct 1997 23 Nov 1997 96 8.8 0.9 8.8 B 2 Jan 1998 26 Jan 1998 44 13.4 3.8 13.9 C 3 Mar 1998 25 Mar 1998 44 14.6 0.6 14.6 D 16 Apr 1998 4 May 1998 50 21.4 6.4 22.4 E 1 Jul 1998 27 Jul 1998 64 14.8 2.5 15.0 F 1 Jan 1999 25 Jan 1999 56 11.5 5.1 12.5 H* 30 Mar 1999 25 Apr 1999 44 14.8 6.4 16.2 G 10 Feb 1999 10 Mar 1999 60 9.2 3.7 9.9 I 17 May 1999 29 May 1999 24 16.1 3.2 16.4 J 18 Sep 1999 7 Dec 1999 96 6.4 1.8 6.7 K 16 Jan 2000 11 Feb 2000 50 19.8 6.9 21.0 L 4 Mar 2000 10 Mar 2000 40 10.7 6.4 12.5 M 25 Apr 2000 31 May 2000 24 14.3 5.7 15.4 N 12 Jun 2000 12 Jul 2000 24 17.1 4.3 17.7 Ave (std dev) 13.8 (4.2) 4.1 (2.1) 14.5 (4.3)

* Ring H was not clearly observed at 55ЊW due to cloud cover. Estimated values of position and velocity are based on the trajectories of three satellite-tracked surface drifters launched in the ring core shortly after formation (see Glickson et al. 2000). westward translation speeds. East of 58ЊW the rings are graphic interaction. A centered, 90-day moving av- generally easy to visualize with well-de®ned cores of erage was subtracted from each observation to min- midocean (presumably South Atlantic/equatorial gyre) imize the contribution of seasonal concentration var- water surrounded by highly productive Amazon-in¯u- iations. The resulting anomaly time series describes enced water. West of this longitude ring tracking is more the mesoscale variability in ocean color at a site well dif®cult due to additional sources of elevated chloro- removed from the NBC retro¯ection and near the phyll (the Orinoco River discharge and enhanced pro- center of the NBC ring translation ``corridor.'' ductivity near the Lesser Antilles) and smearing of surface gradients by wind and lateral mixing. Surface drifter trajectories (discussed below) con®rm that the rings continue to translate at least as far west as Barbados 3. Results (61ЊW) while their surface color expression gradually erodes or is masked by competing processes. a. Ring generation The subjective manual interpretation of images is in- A total of 14 individual ring formation events were formative and relatively precise, but not particularly ef- visually identi®ed during the 3-yr study period (Table ®cient. As we hope to extend this analysis in the future, 1; Fig. 3). Displacement vectors depicting the overall we developed a semiobjective method for inferring NBC translation of each ring during the observed time period ring formation and translation statistics from ocean color are shown in Fig. 4. Ring formation dates were deter- imagery. Two indices were de®ned: mined subjectively by examination of individual 7-day 1) Retro¯ection position anomaly (RPA): For each 7- composite images. Formation was inferred when the day composite image we determined the distance low-chlorophyll ring core was visibly separated from from an arbitrary upstream reference point to the the low-chlorophyll interior of the NBC retro¯ection by northwest corner of the NBC retro¯ection. This lo- a band or ridge of relatively high-chlorophyll water. On cation was extracted manually from each image, al- several occasions a ring appeared to separate but was though in hindsight a relatively simple automated subsequently observed to reattach to the retro¯ection. method could be devised. The record-length (3-yr) These abortive separation events were not included in mean position was subtracted from each observation the statistics enumerated in Table 1. Rather, we cata- to form an anomaly describing the alongshore ex- loged only those formation events that could be linked tension/retraction of the NBC retro¯ection relative to a separated, freely translating ring observed at or to its mean position. downstream of 55ЊW. 2) Chlorophyll concentration anomaly (CCA): For each A majority of the 14 observed ring shedding events 7-day composite image we computed the area-av- are correlated with sudden, large (Ͼ250 km) south- eraged chlorophyll a concentration in a 3Њ latitude eastward retractions of the NBC retro¯ection and com- by 1Њ longitude subregion centered at 9.5ЊN, 55ЊW. mensurate reduction in RPA (Fig. 5). Similarly, nearly Rings that pass this meridian are completely sepa- all of the identi®ed rings are evident as 30±50-day pe- rated from the NBC retro¯ection but have not begun riods of negative CCA. These anomalous low periods to experience signi®cant deformation due to topo- correspond to passage of individual NBC ring cores

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1062 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

FIG. 3. (a) Repeating annual cycles of Amazon River discharge (solid) from Figueiredo et al. (1991) and NBC volume transport (dashed) as measured near 4ЊN by Johns et al. (1998). (b) Repeating annual cycles of wind direction (solid) and stress (dashed) at 49ЊW, 3 ЊN from the global climatology of Hellerman and Rosenstein (1983). Wind direction follows the oceanographic convention (i.e., the direction in which the wind is blowing). (c) A graphical depiction of the time period during which each identi®ed ring formed, passed the 55ЊW meridian, and was lost to remote observation. Several of these rings were observed in situ during a series of research cruises conducted between Nov 1998 and Feb 2000. The formation date of rings Q1±Q4 was approximated based on observed mean translation speeds.

(composed of relatively ``dead'' midocean water) dependence. This interesting and unexpected compli- through the 55ЊW monitoring domain. The retro¯ection cation is discussed further in section 4. retraction signaling ring generation precedes the arrival Although NBC transport varies signi®cantly during of a low-chlorophyll event at 55ЊW by approximately the year (Fig. 3a) we detect no particular seasonality in 1 month. We were unable to identify a consistent re- ring generation (Fig. 6). This ®nding is generally con- lationship between the amplitude of the retro¯ection re- sistent with a recent studies of NBC ring generation traction and the duration or the intensity of the subse- using TOPEX altimetry3 (Pauluhn and Chao 1999; Goni quent low-CCA event at 55ЊW. and Johns 2001) and with moored observations (Fra- Four additional low-CCA events, two each in the fall tantoni et al. 1995), but is markedly different from pre- of 1998 and 1999, could not be correlated with subjec- vious numerical results and inferences from sparse La- tive observations of NBC ring translation past 55ЊW, grangian observations (e.g., Richardson et al. 1994). nor could they be related to a retraction of the NBC Molinari and Johns (1994) demonstrated that the NBC retro¯ection. Yet, these low-chlorophyll events are al- retro¯ection is a consistent, year-round feature in the most identical in duration and amplitude to the 14 vi- western tropical Atlantic thermocline, thus dispelling sually veri®ed ring passage events. In Figs. 3 and 5 these previous notions that NBC ring generation must be a questionable CCA events have been labeled Q1±Q4. seasonal phenomenon. The few existing in situ obser- One of these features (Q2) was carefully surveyed dur- vations of NBC rings indicate signi®cant ring-to-ring ing a research cruise in December 1998 and found to structural variability. At present we are unable to de- have the expected hydrographic and velocity character- termine if this variability is purely random or a con- istics of an NBC ring. Based on this in situ validation and the general similarity of the four events, we surmise that Q1±Q4 are in fact NBC rings that were visually unrecognizable in the composite imagery. Given the 3 Goni and Johns (2001) segregate their observations into ``Rings'' seasonality of the Amazon discharge, the NBC trans- and ``Eddies'' depending on the altimetric observability of the NBC retro¯ection at the time an anticyclonic feature was identi®ed. For port, and the local wind ®eld it is not altogether sur- comparison with our SeaWiFS observations, a portion of which over- prising that our ability to visually detect NBC rings via lap the TOPEX measurements of Goni and Johns, we have regrouped their ocean color signature should exhibit a seasonal their census results neglecting this distinction.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1063

rings formed from one year to the next. While the present study found no year-to-year variation in ring formation, the 3-yr study period is hardly adequate to address interannual variability. It should also be noted that the Goni and Johns (2001) observations appear to sta- bilize in the late 1990s with approximately six rings formed per year from 1996 to the present (G. Goni 2000, personal communication).

b. Ring translation Once free of the retro¯ection the NBC rings were observed to move northwestward parallel to the continental slope at a mean speed of 14.5 Ϯ 4.3 cm sϪ1. This speed is consistent with previous estimates from surface drifters (Richardson et al. 1994), satellite altimetry (Did- den and Schott 1993; Pauluhn and Chao 1999; Goni and Johns 2001), and numerical simulations (Fratantoni et FIG. 4. Displacement of the 14 NBC rings enumerated in Table 1 al. 1995). The translation speed of individual rings during the period of observation with SeaWiFS imagery. The mean Ϫ1 position of the NBC retro¯ection is illustrated schematically for ref- varies from a minimum of 7 cm s to a maximum of Ϫ1 erence. 22 cm s in the vicinity of 55ЊW. As will be illustrated below in the form of case studies, the translation speed of a particular ring can vary considerably during its brief sequence of the underlying seasonality of the NBC and lifetime with speeds as high as 30 cm sϪ1 occasionally its environs. noted. A portion of the translation speed variability re- Questions remain as to the degree of interannual var- sults from simultaneous deformation of the overall ring iability in ring generation. Goni and Johns (2001) report geometry (see below). The mean trajectory traced by signi®cant (factor of 2) differences in the number of NBC ring centers is generally parallel to the continental

FIG. 5. Retro¯ection position anomaly (RPA) and chlorophyll concentration anomaly (CCA) time series (see text for explanation). Where possible, shaded lines indicate visual correlation between retraction of the NBC retro¯ection (indicated by sharp decline in RPA) and subsequent passage of a ring across the 55ЊW meridian (indicated by a local minimum in CCA). Individual measurements (circles) taken from SeaWiFS imagery are connected by a cubic spline. CCA events Q1± Q4 do not visually correspond to ring generation events. See text for details.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1064 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

c. Ring geometry A summary of geometric parameters corresponding to ring observations near 55ЊW is shown in Table 2. To simplify the characterization of ring geometry we approximated each ring as an ellipse. Mean major (minor) axis lengths of 304 (213) km yield a mean ring-core surface area of approximately 52 000 km2 at this longitude. Typical aspect ratios (ratio of minor to major axis length) near 0.7 suggest that the standard kinematic approximation of an oceanic ring as a circular, axisym- metric vortex may not apply in this region. There is a tendency for the major axis of the elliptical ring to align itself parallel to the sloping western boundary, particularly as the ring translates to the west of 55ЊW and encounters topography that is increasingly perpendic- ular to the ring's preferred westward translation direction. Several rings were observed to dramatically change their orientation as they evolved (see case studies below). When cast in terms of cross- and alongshore dimensions these observations suggest mean radii of 100 FIG. 6. Histograms depicting the quarterly frequency of observed and 150 km, respectively. The circulation associated NBC ring generation. Note that Goni and Johns (2001) do not spe- ci®cally report the date of ring generation, but rather the date on with the NBC retro¯ection and separated NBC rings which a ring is ®rst identi®ed using TOPEX altimetry (G. Goni 2000, extends beyond the low-chlorophyll core. In situ ve- personal communication). This difference in methodology is most locity measurements (e.g., Fratantoni et al. 1999; Wilson likely responsible for the differences shown. et al. 1999) indicate the overall scale of circulation associated with an NBC ring may exceed 450 km in diameter. shelfbreak in water 3000±4000 m deep (Fig. 4; Didden The observed ocean color gradients provide a means and Schott 1993; Fratantoni et al. 1995). to visualize the generation and evolution of NBC rings. The relative importance of NBC ring self-propagation But do the ocean color patterns we observe correspond versus advection by a background ¯ow was investigated in a meaningful way to the expected horizontal velocity brie¯y by Fratantoni et al. (1995) and in greater detail structure of an NBC ring? For example, may we assume by Jacob (1997). The latter study concluded that NBC that the color front that visibly de®nes the ring boundary rings self-propagate via the ␤ effect at a speed consistent is dynamically related to the radius of maximum ve- with analytical models (e.g., Cushman-Roisin et al. locity, typically regarded as the boundary between the 1990). There is therefore no requirement for a back- ring core and the external environment? As part of a ground northwestward ¯ow to explain their movement. recent in situ measurement program, synoptic surveys This provides support for the notion ®rst advanced by of hydrographic properties and absolute vector velocity Richardson et al. (1994) that the northwestward ``Guy- were obtained in and around the NBC retro¯ection and ana Current'' historically depicted in regional circula- several NBC rings (Fratantoni et al. 1999). On several tion schematics and pilot charts may be an artifact of occasions these in situ surveys coincided with relatively non-eddy-resolving measurements of NBC rings.4 The cloud-free atmospheric conditions permitting simulta- amplitude and seasonality of northwestward ¯ow over neous representation of the surface circulation with the wide and shallow (Ͻ100 m) continental shelf re- SeaWiFS. In Fig. 7 we show velocity vectors at a depth mains uncertain, although moored measurements up- of 15 m superimposed on a nearly simultaneous stream of the retro¯ection suggest an annual-mean ¯ow SeaWiFS image of the NBC retro¯ection. Note partic- of up to 5 Sv (Sv ϵ 106 m3 sϪ1) (Johns et al. 1998). ularly the line of stations near the northwestward extent Lagrangian observations of a coastal current (Limebur- of the retro¯ection (A) and across the southeastward ®rst ner et al. 1995; Glickson et al. 2000) are generally in- meander of the NECC (B). In both areas the maximum conclusive as most surface drifters launched on the shelf velocity of the NBC/NECC jet is aligned (within the upstream of the retro¯ection are quickly pulled offshore resolution of the survey) with the observed ocean color into the NBC. front. Velocities in the NBC upstream of the retro¯ection (C) are at a local maximum near the 100-m isobath, suggesting that the edge of the continental shelf is both 4 Recent observations in the western Indian Ocean have similarly demonstrated that the Mozambique Current is best described as a the onshore boundary for NBC transport and the off- succession of eddies rather than as a feature of the mean circulation shore boundary for high-chlorophyll Amazon-in¯u- (DeRuijter et al. 2002). enced water. These comparisons indicate that dynami-

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1065

TABLE 2. Geometric parameters of NBC rings observed near 55ЊW. a

Major axis Minor axis Surface area Volumeb Transportc Ring (km) (km) Aspect ratio (km2 ϫ 104) (m3 ϫ 1013) (Sv) A 345 222 0.64 6.01 3.61 1.14 B 400 289 0.72 9.08 5.45 1.73 C 311 233 0.75 5.70 3.42 1.08 D 222 167 0.75 2.91 1.75 0.55 E 256 189 0.74 3.79 2.28 0.72 F 211 156 0.74 2.58 1.55 0.49 G 367 278 0.76 8.00 4.80 1.52 I 311 167 0.54 4.07 2.44 0.77 J 278 233 0.84 5.09 3.06 0.97 K 311 245 0.79 5.97 3.58 1.14 L 333 167 0.50 4.37 2.62 0.83 Ave (std dev) 304 (58) 213 (47) 0.71 (0.10) 5.23 (2.01) 3.14 (1.21) 1.00 (0.38)

a Rings H, M, and N are not included in this table as they were obscured by clouds in the vicinity of 55ЊW. b Volume computed assuming ring is an elliptical cylinder with vertical height 600 m. This height is determined from consideration of oxygen anomaly pro®les obtained during in situ survey of three NBC rings (see Fig. 8). c Annualized per ring transport is computed by dividing the ring volume (m3) by the number of seconds in one year. cally relevant geometric parameters (e.g., Table 2) may a passive tracer that, at depth, can only be advected by be reasonably inferred from ocean color imagery. the ring's velocity ®eld or diffused. The three oxygen anomaly pro®les shown in Fig. 8 are similar in form, d. Ring volume transport each tending toward zero near a depth of 600 m. This suggests that the vertical segment of the swirling ring The effective transport of South Atlantic water into core that is hydrographically distinct from its surround- the subtropical North Atlantic by NBC rings has been ing environment may be approximated5 by the depth estimated at approximately 1 Sv per ring by several interval 0±600 m. This vertical interval was used in authors (Johns et al. 1990; Didden and Schott 1993; combination with the SeaWiFS surface area measure- Richardson et al. 1994; Fratantoni et al. 1995). Based ments to compute the ring volume and effective an- on such estimates, the annual mass ¯ux carried by 5± nualized per ring transports shown in Table 2. 6 rings amounts to roughly one-third of the inter- The resulting 1.0 Ϯ 0.4 Sv per ring transport value hemispheric transport required to close the upper limb deduced from our SeaWiFS observations is strikingly of the Atlantic MOC (e.g., Schmitz and McCartney similar to, yet completely independent of, earlier NBC 1993; Schmitz 1995). The relative consistency of the ring transport estimates. Necessary simpli®cations in published mass ¯ux estimates tends to obscure the fact the present calculation suggest that the ubiquitous 1 that they are based on very limited data and use dif- Sv per ring value we achieved is probably an upper ferent assumptions about horizontal scale, vertical pen- bound on the effective annual transport. In particular, etration, and number of rings that are formed in a typ- Richardson et al. (1994) and Fratantoni et al. (1995) ical year. noted that NBC rings observed with Lagrangian de- The present dataset provides a new opportunity to vices and a current meter mooring, respectively, in- estimate the transport potential of NBC rings using dicated reduced core diameter at depth relative to at methods independent from those employed by previous the surface. This suggests a ring core geometry more investigators. To compute ring volume we assume each similar to a truncated cone than to a cylinder, and there- ring may be approximated as an elliptical cylinder with fore smaller in volume. A signi®cant contribution of a known surface area (Table 2) and a vertical penetration the present study to the assessment of ring transport is depth estimated from recent in situ ring observations. our independent estimate of the number of rings Three different NBC rings were surveyed near 57ЊW during regional research cruises conducted between No- formed per year (6), a value that can only be obtained vember 1998 and February 2000 (Fratantoni et al. through remote observation and/or extended in situ 1999). In Fig. 8 we show vertical pro®les of tempera- monitoring. Additional detailed study of NBC ring wa- ture, salinity, and dissolved oxygen anomalies computed ter mass and velocity structure is required to re®ne this at the core of each ring relative to the ring's immediate coarse estimate of volume transport. background environment. As fundamental constituents of the density ®eld, temperature and salinity are inti- mately related to the velocity ®eld, and hence the pro- 5 For this calculation we ignore the apparent reversal in oxygen anomaly in the 100-m-thick surface mixed layer. This feature is most ®les of temperature and salinity anomaly re¯ect the dis- likely due to nonconservative upper-ocean sources and sinks of ox- parate azimuthal velocity structure of these three rings ygen rather than lateral mixing between the ring core and its envi- (Fratantoni et al. 1999). Dissolved oxygen, however, is ronment.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1066 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

FIG. 7. A composite SeaWiFS image (centered on 8 Dec 1998) of the NBC retro¯ection aug- mented by in situ absolute velocity measurements at depth 15 m. Note the correspondence between high gradients in ocean color and local maxima in the NBC jet. e. Ring evolution: Case studies 9. Just prior to the separation of Ring F in early January In this section we present case studies detailing the 1999, the NBC retro¯ection was in an extended state evolution of three NBC rings observed with SeaWiFS. reaching almost to 9ЊN (Fig. 10a). An indentation or The three rings (F, G, and K) are typical of the larger ``neck'' had formed where the ®rst meander of the population we observed. These particular rings were southeastward NECC approached the northwestward chosen for detailed study primarily because they were NBC. On approximately 1 January the retro¯ection re- extensively surveyed in situ as part of a regional ®eld tracted sharply (see Fig. 5) leaving behind a pinched- program (Fratantoni et al. 1999) and seeded with sat- off ring centered over the northeast corner of the De- ellite-tracked surface drifters (Glickson et al. 2000). A merara Rise. Ring F moved hesitantly toward the north- more detailed account of the demise of NBC rings upon west for about two weeks before its translation speed collision with the Lesser Antilles, utilizing both surface stabilized near 12 cm sϪ1 (Fig. 10d). This ring was one drifters and vertical arrays of acoustically tracked sub- of the smallest identi®ed during the study, with a surface surface RAFOS ¯oats, will appear in a forthcoming ar- area approximately one-half the average value observed ticle. Note that the Glickson et al. (2000) drifter data at 55ЊW. While approximately round shortly after for- report and a combined SeaWiFS-drifter animation loop mation (aspect ratio near 1.0 on 10 January) Ring F are publicly available online at gradually became more eccentric with time and rotated (http://science.whoi.edu/users/dfratantoni). such that its major axis was oriented parallel to the continental slope. This alongshore stretching is particularly evident in the trajectories of surface drifters launched in the ring in February 1999 (Fig. 9). The 1) RINGS F AND G drifters in the core of Ring F abruptly stopped their A time series of SeaWiFS images depicting the gen- looping behavior in early March as the center of cir- eration and evolution of Rings F and G is shown in Fig. culation neared Trinidad and the ring became extremely

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1067

FIG. 8. Vertical pro®les of temperature, salinity, and dissolved oxygen anomaly computed within the core relative to a pro®le on the ring's periphery. Three pro®les are shown, one each for rings observed near 57ЊW in 1998 (solid), 1999 (dotted), and 2000 (dashed). The oxygen anomaly pro®les suggest that the strongest water mass contrast between the core of an NBC ring and its surrounding environment occurs above a depth of 600 m. elongated. Just prior to the apparent collapse of its vor- could result in variability in the location at which ring tical circulation, Ring F extended as far north as Mar- core water is released to the surrounding environment. tinique (13ЊN) with an overall length of almost 600 km. As seen in Fig. 9, the trajectories of drifters intentionally Though we ®nd no evidence that Ring F penetrated the launched near the ring center indicate a closed circu- Lesser Antilles as a coherent vortex, the distribution of lation considerably smaller than the radius of maximum its fractured remnants (identi®ed by surface drifters) velocity. Over time the radius at which the drifters loop indicate that some fraction of the ring's core volume becomes larger but remains within the assumed radius entered the southeastern Caribbean through a series of of maximum velocity, which we associate with the low- shallow (50 m) passages in the Grenadines. chlorophyll ring core. By late February Rings F and G Ring G was generated immediately following ring F are separated by only a narrow (150 km) band of high- on approximately 10 February 1999. The retro¯ection chlorophyll water that, we assume, is experiencing con- did not extend as far north prior to the formation of this siderable horizontal shear resulting from the opposing ring (Fig. 5) nor was any signi®cant ``necking'' of the azimuthal velocities of the two rings. The demise of retro¯ection noted (Fig. 11). Nevertheless, in contrast Ring G is somewhat different from its predecessor with to Ring F, Ring G is one of the largest rings observed very few of the core drifters entering the Caribbean. during this study with a surface area almost 50% larger Instead, the ring maintains its vortical circulation while than average. This size discrepancy suggests the pos- moving northward parallel to and east of the Lesser sibility that the extreme elongation of Ring F near the Antilles. Most of the drifters in the ring core accelerate end of its life cycle may have been due in part to in- as they pass through a deep but narrow channel between teraction with its neighbor. One common feature of vor- the islands of Barbados and St. Lucia. tex±vortex interactions, studied extensively in the lab- While no drifter in Ring G demonstrated a complete oratory and in numerical simulations, is that of a large, loop around Barbados (but see Ring K, below) the strong vortex entraining ®laments of ¯uid from a neigh- movement of the low-chlorophyll ring core strongly boring, smaller, weaker vortex (e.g., Provenzale 1999). suggests that it passed directly over the island, com- The close proximity of Rings F and G (not atypical of pletely engul®ng it (see Fig. 9, 11±19 April). How is other observed NBC rings) suggests that such interac- this possible? At a depth of 1000 m (i.e., somewhat tion may be a common and important component of below the vertical center of the ring's azimuthal cir- downstream ring evolution. Such ring±ring interaction culation) the island of Barbados appears to the ring as

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1068 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

FIG. 9. Time series of composite SeaWiFS images depicting the evolution of Rings F and G. Color palette is identical to that used in Fig. 2. Seven-day segments of satellite-tracked surface drifter trajectories corresponding to the SeaWiFS image are also shown. The length of each line segment re¯ects the speed of the drifter. The most recent drifter position (i.e., the head of each 7-day ``worm'') is depicted by a small circle.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1069

FIG. 10. Summary diagram depicting the generation and evolution of Ring F. (a) Simpli®ed representation of the color front bounding the NBC retro¯ection and Ring F during various stages of its life cycle. Dashed contour corresponds to the last observed con®guration of the retro¯ection prior to identi®cation of a separated ring. (b) Movement of the ring center during the period of observation. Circles appear every 4 days. Approximated positions (required due to cloud cover) are denoted by open circles. (c±e) Time series of latitude (solid), longitude (dashed), translation speed, and aspect ratio (ratio of the major and minor axes of an elliptically shaped ring).

an obstacle approximately 40 km wide and 100 km long. 2) RING K With a core diameter at least ®ve times the width of this obstacle (and a total extent perhaps twice that size) The evolution of Ring K is depicted in a time series the ring is able to maintain a (possibly distorted) vortical of SeaWiFS images shown in Fig. 12. This ring was circulation. This does not appear to be the case when formed in January 2000 in a manner quite similar to an NBC ring meets the considerably larger ridge en- that of Ring F,accompanied by a clear extension, ``neck- compassing the Lesser Antilles, as in the case of Ring ing,'' and subsequent retraction of the retro¯ection F.A similar problem, that of Mediterranean Water eddies (Figs. 5 and 13a). Ring K accelerated rapidly toward (meddies) colliding with an isolated cylindrical sea- the west. The mean translation speed over its ®rst month mount, has been studied in the laboratory by Cenedese of existence was 21 cm sϪ1, making it one of the fastest (2002) for both advected and self-propagating vortices. observed during this study. This rapid translation was Cenedese found that for ratios of seamount diameter to accompanied by signi®cant deformation of its already vortex diameter of less than 0.2 (the regime correspond- elliptical shape, with an aspect ratio gradually decreas- ing to NBC Ring±Barbados interaction) the vortex ing from 0.8 to near 0.5. By early March the major axis moved past the topographic obstacle with minimal dis- of the ring had assumed a position parallel to the con- turbance. tinental slope east of Tobago.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1070 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

FIG. 11. As in Fig. 10 but for Ring G.

Near the middle of March the center of ring K passed contrast and the resulting pattern of successive ``missing directly over Barbados. Unlike Ring G, we have detailed rings'' (Q1±Q4) apparent in Fig. 3. In addition to hor- veri®cation of this event as a surface drifter in the ring izontal advection by wind and the NBC, surface chlo- core completes a closed loop around the island and then rophyll concentration on the equatorial continental shelf continues looping as the ring moves northward (see Fig. is assumed to be related to the Amazon discharge 10: 20 March). This anticyclonic circulation persisted through an unknown and probably nonlinear biological until the ring center reached about 17ЊN. From this point response to changing nutrient input as well as inherent the ring gradually decayed without further translation seasonality in phytoplankton productivity. It is reason- (not shown; see Glickson et al. 2000). None of the drift- able to hypothesize that the annually varying Amazon ers launched in Ring K entered the Caribbean, although discharge (Oltman 1968; Figueiredo et al. 1991), the three drifters grounded while heading in that direction, NBC transport (Johns et al. 1998), and the local wind one each on St. Lucia, Martinique, and Guadeloupe. stress (Lentz 1995a,b) conspire to reduce the amount of ``tagged'' chlorophyll-rich water on the outer continental shelf that is entrained into the NBC. This in turn 4. Discussion leads to a reduced ability to visually distinguish the NBC We have demonstrated that it is possible, albeit dif- retro¯ection and any pinched-off rings in ocean color ®cult, to extract useful information about NBC ring gen- imagery. eration and evolution from remotely sensed ocean color As shown in Fig. 3, events Q1±Q4 occur within a data. One particular source of dif®culty stems from a few months of the climatological annual minimum in recurrent July±October minimum in surface chlorophyll Amazon River discharge, shortly after the NBC trans-

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1071

FIG. 12. As in Fig. 9 but for Ring K. port begins to decline from its annual maximum, and charge during boreal summer and fall (Fig. 3) could during a period of relatively light southwestward winds. signi®cantly reduce the amount of Amazon-in¯uenced Lentz (1995b) examined seasonal variations in the con- water on the outer continental shelf that is available to ®guration of the Amazon plume using climatological become entrained into the NBC. Further observation, hydrographic data and found that, on average, the Am- analysis, and/or numerical simulation is required to fur- azon discharge was contained much closer to shore ther explore the relationship between the various forcing (200±300 km) during July±December than during Jan- mechanisms described above and their relative impact uary±June (400±500 km). This difference in offshore on surface chlorophyll distributions in the western equa- extent was attributed to a combination of high Amazon torial Atlantic. discharge and the more-frequent occurrence of southwestward winds during the ®rst half of the year. This 5. Summary and conclusions is consistent with synoptic observations of the Amazon plume by Lentz and Limeburner (1995) that clearly re- We have presented unique observations of North Bra- veal a minimum in the offshore extent of low-salinity zil Current ring generation and evolution using SeaWiFS water during November 1991 coincident with minimum ocean color imagery and satellite-tracked surface drift- Amazon discharge. We surmise that the large (50%) ers. These observations provide insight into the com- reduction in local wind stress and reduced Amazon dis- plete life cycle of an NBC ring from its genesis in the

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1072 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

FIG. 13. As in Fig. 10 but for Ring K. low-latitude Atlantic to its demise near the Lesser An- in chlorophyll concentration persists along the NBC tilles. In combination with recent in situ observations ring translation corridor during all seasons of the our relatively small collection of remotely observed year. rings enabled a new and independent estimate of the 3) Based on SeaWiFS observations and in situ tracer contribution of NBC rings to intergyre volume transport measurements, a reasonable upper bound on the ef- in the western tropical Atlantic. The main conclusions fective annualized transport of a single NBC ring is of this investigation can be summarized as follows: approximately 1.0 Ϯ 0.4 Sv. This value is both con- 1) Ocean color imagery of the western tropical Atlantic sistent with and independent from previous esti- reveals the dynamically relevant spatial and temporal mates. scales associated with NBC ring generation and evo- 4) Approximately Six NBC rings are formed each year. lution. Strong horizontal gradients in surface chlo- Successive rings evolve in close proximity to one rophyll can be used to infer the position of the NBC other. Our observations suggest that neighboring retro¯ection and the radius of maximum velocity of rings may exert substantial in¯uence on each other. NBC rings. Interaction between adjacent rings may control the 2) The effectiveness of SeaWiFS imagery for visual time and location at which trapped ring core water NBC ring identi®cation appears to have a seasonal is released to the tropical/subtropical gyre circula- limitation associated with the annual cycle of Am- tion. azon River discharge and/or wind-forced cross-shelf 5) Progressive deformation of the elliptical ring ge- transport. However, detectable mesoscale variability ometry occurs rapidly relative to the rate of north-

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC MARCH 2002 FRATANTONI AND GLICKSON 1073

westward ring translation. This implies that precise REFERENCES prediction of high-velocity events at a particular location (e.g., an island, oil rig, etc.) depends on Borstad, G. A., 1982: The in¯uence of the meandering Guiana Current knowledge of both the ring's translation and defor- and Amazon River discharge on surface salinity near Barbados. J. Mar. Res., 40, 421±434. mation histories. Brown, O. B., D. B. Olson, J. W. Brown, and R. H. Evans, 1983: Satellite infrared observation of the kinematics of a warm-core ring. Aust. J. Mar. Freshw. Res., 34, 535±545. The societal importance of a thorough understanding Bruce, J. G., J. L. Kerling, and W. H. Beatty, 1985: On the North of NBC ring generation and evolution extends beyond Brazilian eddy ®eld. Progress in Oceanography, Vol. 14, Per- their role as a link in the Atlantic MOC. For example, gamon, 57±63. Cenedese, C., 2002: Mesoscale vortices colliding with a seamount. expanding deep water oil and gas exploration efforts J. Geophys. Res., in press. offshore of Venezuela and Trinidad depend on accurate Cowen, R. K., and L. R. Castro, 1994: Relation of coral reef ®sh characterization and prediction of regional circulation larval distributions to island scale circulation around Barbados, variability in order to assure the safety and pro®tability West Indies. Bull. Mar. Sci., 54, 228±244. of these endeavors. In the absence of an extensive in Cushman-Roisin, N., E. P. Chassignet, and B. Tang, 1990: Westward motion of mesoscale eddies. J. Phys. Oceanogr., 20, 758±768. situ monitoring program, the ability to predict the move- DeRuijter, W. P. M., H. Ridderinkhof, J. R. E. Lutjeharms, M. W. ment and intensity of NBC rings on an operational basis Schouten, and C. Veth, 2002: Observations of ¯ow in the Moz- depends largely on our ability to collect and interpret ambique Channel. Geophys. Res. Lett., in press. remote observations. Individually, satellite altimetry Didden, N., and F. Schott, 1993: Eddies in the North Brazil Current retro¯ection region observed by GEOSAT altimetry. J. Geophys. and ocean color measurements provide important in- Res., 98, 20 121±20 131. formation regarding the generation and evolution of Figueiredo, A. G., M. Allison, and C. A. Nittrouer, 1991: Amazon NBC rings, and each approach has its bene®ts and lim- discharge: Internal report for AMASSEDS researchers. Univer- itations. Altimetric measurements are immune to cloud sity Fed. Fluminense Tech. Rep., Niteroi, Brazil. cover and allow direct estimation of circulation inten- Fratantoni, D. M., W. E. Johns, and T. L. Townsend, 1995: Rings of the North Brazil Current: Their structure and behavior inferred sity, but at marginally useful spatial and temporal res- from observations and a numerical simulation. J. Geophys. Res., olution. Ocean color observations provide high-reso- 100, 10 633±10 654. lution, near-synoptic views of surface circulation pat- ÐÐ, P. L. Richardson, W. E. Johns, C. I. Fleurant, R. H. Smith, S. terns, but provide no intensity information and are sus- L. Garzoli, W. D. Wilson, and G. J. Goni, 1999: The North Brazil Current Rings Experiment (abstract). Eos, Trans. Amer. Geo- ceptible to cloud interference and seasonal changes in phys. Union, 80, 43. discharge-related productivity. A logical next step in the ÐÐ, W. E. Johns, T. L. Townsend, and H. E. Hurlburt, 2000: Low- development of an operational ring tracking system latitude circulation and mass transport pathways in a model of might involve combination of the best features of these the tropical Atlantic Ocean. J. Phys. Oceanogr., 30, 1944±1966. two observational methods, perhaps using an assimi- Geyer, W. R., and R. C. Beardsley, 1995: Introduction to special section on physical oceanography of the Amazon shelf. J. Geo- lating numerical model to blend real-time surface ob- phys. Res., 100, 2281±2282. servations into a dynamically consistent three-dimen- Glickson, D. A., D. M. Fratantoni, C. M. Wooding, and P. L. Rich- sional picture of the tropical Atlantic circulation. ardson, 2000: North Brazil Current Rings Experiment: Surface drifter data report November 1998±June 2000. Woods Hole Oceanographic Institution Tech. Rep. WHOI-2000-10, 129 pp. Acknowledgments. This work was supported by the Goni, G. J., and W. E. Johns, 2001: A census of North Brazil Current National Science Foundation through Grant OCE 97- Rings observed from T/P altimetry: 1992±1998. Geophys. Res. 29765. The ocean color data were provided by the Lett., 28, 1±4. SeaWiFS Project at Goddard Space Flight Center. The Hellerman, S., and M. Rosenstein, 1983: Normal monthly wind stress over the World Ocean with error estimates. J. Phys. Oceanogr., data were obtained from the Goddard Distributed Active 13, 1093±1104. Archive Center under the auspices of the National Aero- Hooker, S. B., W. E. Esaias, G. C. Feldman, W. W. Gregg, and C. R. nautics and Space Administration. Use of this data is McClain, 1992: An Overview of SeaWiFS and Ocean Color. in accord with the SeaWiFS Research Data Use Terms NASA Tech. Memo. 104566, 24 pp. and Conditions Agreement. Supplemental support for Jacob, J. P., 1997: In¯uence of topography on the propagation of isolated eddies. Masters thesis, Rosenstiel School of Marine and the surface drifter measurements was provided by the Atmospheric Science, University of Miami, 116 pp. National Oceanic and Atmospheric Administration. The Johns, W. E., T. N. Lee, F. A. Schott, R. J. Zantopp, and R. H. Evans, in situ observations reported here result from the col- 1990: The North Brazil Current retro¯ection: Seasonal structure laborative data collection and analysis efforts of the and eddy variability. J. Geophys. Res., 95, 22 103±22 120. North Brazil Current Rings Experiment coinvestigators ÐÐ, ÐÐ, R. C. Beardsley, J. Candela, R. Limeburner, and B. Cas- tro, 1998: Annual cycle and variability of the North Brazil Cur- (P. L. Richardson, D. M. Fratantoni, W. E. Johns, S. rent. J. Phys. Oceangr., 28, 103±128. Garzoli, W. D. Wilson, and G. J. Goni). We acknowledge Kelly, P. S., K. M. M. Lwiza, R. K. Cowen, and G. J. Goni, 2000: the thoughtful comments made by P. L. Richardson and Low-salinity pools at Barbados, West Indies: Their origin, fre- W. E. Johns on an early version of this manuscript. We quency, and variability. J. Geophys. Res., 105, 19 699±19 708. Lentz, S. J., 1995a: The Amazon River Plume during AMASSEDS: thank Semyon Grodsky and an anonymous reviewer for Subtidal current variability and the importance of wind forcing. their insightful comments and suggestions for improve- J. Geophys. Res., 100, 2377±2390. ment. ÐÐ, 1995b: Seasonal variations in the horizontal structure of the

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC 1074 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 32

Amazon Plume inferred from historical hydrographic data. J. Oltman, R. E., 1968: Reconnaissance investigations of the discharge Geophys. Res., 100, 2391±2400. and water quality of the Amazon River. U.S. Geological Survey ÐÐ, and R. Limeburner, 1995: The Amazon River Plume during Circular 552, 16 pp. AMASSEDS: Spatial characteristics and salinity variability. J. Pauluhn, A., and Y. Chao, 1999: Tracking eddies in the subtropical Geophys. Res., 100, 2344±2375. north-western Atlantic Ocean. Phys. Chem. Earth, 24A, 415± Limeburner, R., R. C. Beardsley, I. D. Soares, S. J. Lentz, and J. 421. Candela, 1995: Lagrangian ¯ow observations of the Amazon Provenzale, A., 1999: Transport by coherent barotropic vortices. River discharge into the North Atlantic. J. Geophys. Res., 100, Annu. Rev. Fluid. Mech., 31, 55±93. 2401±2416. Richardson, P. L., G. E. Hufford, R. Limeburner, and W. S. Brown, Lutjeharms, J. R. E., 1996: The exchange of water between the South 1994: North Brazil Current retro¯ection eddies. J. Geophys. Res., Indian and South Atlantic Oceans. The South Atlantic: Present 99, 5081±5093. Schmitz, W. J., Jr., 1995: On the interbasin-scale thermohaline cir- and Past Circulation, G. Wefer, W. H. Berger, G. Siedler, and culation. Rev. Geophys., 33, 151±173. D. Webb, Eds., Springer-Verlag, 122±162. ÐÐ, and M. S. McCartney, 1993: On the North Atlantic circulation. McClain, C. R., M. L. Cleave, G. C. Feldman, W. W. Gregg, S. B. Rev. Geophys., 31, 29±49. Hooker, and N. Kuring, 1998: Science quality SeaWiFS data for Servain, J., and S. Lukas, 1990: Climatic Atlas of the Tropical Atlantic global biosphere research. Sea Technol., 39, 10±16. Wind Stress and Sea Surface Temperature 1985±1989, IFRE- Molinari, R. L., and E. Johns, 1994: Upper layer temperature structure MER, 133 pp. of the western tropical Atlantic. J. Geophys. Res., 99, 18 225± Wilson, W. D., S. L. Garzoli, G. J. Goni, W. E. Johns, R. H. Smith, 18 233. C. I. Fleurant, P. L. Richardson, and D. M. Fratantoni, 1999: Muller-Karger, F. E., C. R. McClain, and P. L. Richardson, 1988: The The North Brazil Current retro¯ection: Two recent surveys. Eos, dispersal of the Amazon's water. Nature, 333, 56±59. Trans. Amer. Geophys. Union, 80, p. 43.

Unauthenticated | Downloaded 10/02/21 06:18 AM UTC