Caribbean Current and Eddies As Observed by Surface Drifters

Home , Antilles Current, Caribbean Current, Windward Islands, Wu (region)

ARTICLE IN PRESS

Deep-Sea Research II 52 (2005) 429–463 www.elsevier.com/locate/dsr2

Caribbean Current and eddies as observed bysurface drifters

P.L. RichardsonÃ

Department of Physical Oceanography, MS 29, Woods Hole Oceanographic Institution, 360 Woods Hole Road, Woods Hole, MA 02543, USA

Received 19 December 2003; accepted 29 November 2004

Abstract

Recent satellite-tracked surface drifter trajectories were analyzed to describe the mean currents and eddies in the 1 Caribbean Sea. The structure of the Caribbean Current and its variabilitywere determined from high-resolution 2- degree maps of the mean velocityand eddykinetic energy.Looping drifter trajectories were used to identifydiscrete cyclones and anticyclones, and their characteristics were described and related to the structure of the mean flow. The translation rate of eddies in different areas was found to be similar to the mean velocityof the local background flow fields, suggesting that the eddies were largelyadvected bythe background flow. Ten energetic anticyclonestranslated westward at 13 cm/s in the Venezuela and Colombia Basins. These anticyclones tended to lie in two bands, centered near 151N and 171N, coinciding with two jets of the Caribbean Current. The northern weaker jet contains water primarilyfrom the North Atlantic; the southern stronger jet contains water from the tropical and South Atlantic. The anticyclones are thought to have formed in the eastern Caribbean from the anticyclonic vorticity derived from North Brazil Current rings. The ring vorticityenters the eastern Caribbean through island passages and is probablyamplified by the anticyclonic shear on the northern side of the jets. Southwest of Cuba a cyclone–anticyclone pair was observed to translate slowly( 2 cm/s) westward into the Yucatan Current. The cyclone was tracked for 10.5 months with four drifters, making it the longest tracked of the Caribbean eddies. r 2005 Elsevier Ltd. All rights reserved.

1. Introduction (Wilson and Leaman, 2000)(Fig. 2). Other drifters were launched east of the Caribbean in other Recently, a large number of satellite-tracked, experiments including the North Brazil Current drogued, drifting buoys measured trajectories in (NBC) rings experiment (Fratantoni and Richard- the Caribbean Sea (Fig. 1). Manydrifters were son, 2004); these drifters passed through the launched in the Caribbean during 1998–2000 as Antilles Islands passages adding trajectories in part of the ‘‘Year of the Ocean-1999’’ study the Caribbean. Taken together, these trajectories provide a unique data set that reveals previously ÃTel.: +1 508 289 2546; fax: +1 508 457 2163. unknown details of the circulation and eddies in E-mail address: [email protected]. the different parts of the Caribbean. The intent of

430 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463

n l

s d

I s

r a

w rom the

s d

d n

n i l a W Is TI LLES ntours. The rd N a A 2000 m ° w

e R 63 e E L S S 2000 m m 200 E L ° 66 P.R. S

L ° Venezuela

I 69

T Basin

N Venezuela

A Republic Dominican ° 72 Haiti ° 75 Columbia

R Caribbean Sea E

T 2000 m Cuba A ° Jamaica

E 78 Basin

R Columbia G Panama °

81 2000 m 2000 °

Isle of Pines 200 m 200 Rica Basin Costa Yucatan Cayman Basin Nicaragua Honduras W84 ° 87 Yucatan Peninsula ° ° ° ° N ° 21 18 15 12 9 Fig. 1. Location diagram showing the major Caribbean basins, the surrounding countries and islands, and the 200 m (dashed) and 2000 m (dotted) depth co relativelyshallow Jamaica Ridge extendsVenezuela from and the Colombia Honduras Basins shelf of to the Hispaniola eastern and Caribbean. separates the Yucatan and Basins Cayman of the western Caribbean f ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 431 e western at entered the W ° 63 W ° 66 W ° 69 W ° 72 W ° 75 N. 1 Starting Points W ° 78 W ° 81 W ° 84 W and 13 that entered from the north across 22 1 W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 Fig. 2. Launch orCaribbean, 24 start in locations the of southernbox Colombia 212 from Basin, drifters the and east used 54 across in in 61 the this eastern study. Caribbean. Ninety-four One others hundred were and launched outside eighteen the drifters Caribbean were including 66 launched th in the Caribbean including 40 in th ARTICLE IN PRESS

432 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 this paper is to analyze these data with the goal of awayfrom the coast into deeper regions and vice learning more about the characteristics of Car- versa. This has important implications for the ibbean eddies in general and specificallyabout the dispersal of pollutants and fish larvae, e.g. Third, energetic anticyclones in the eastern Caribbean eddies appear to have some preferred pathways that might have been generated byNBC rings. through the Caribbean that cause bands of higher Satellite altimetrystudies suggest that some NBC and lower (rectified) mean westward velocity rings can pass coherentlythrough the Antilles coinciding with westward jets in the Caribbean passages into the Caribbean (Goni and Johns, Current and which suggests that eddy-mean flow 2003). The motivation for the present studyis to interactions could be important. Fourth, eddy see whether the drifter data could provide evidence forcing of the circulation of deeper layers in the that Caribbean anticyclones are generated by NBC Caribbean could be significant if the energetic rings. eddies extend deeplyinto the water column like Although the mean velocityfield and velocity some NBC rings. Fifth, Caribbean eddies can be variance have been mapped recentlyfor the advected into the Yucatan Current and influence Caribbean using surface drifters (Wilson and the path of the Loop Current in the Gulf of Leaman, 2000; Fratantoni, 2001; Centurioni and Mexico and the formation of Loop Current rings Niiler, 2003), and trajectories of drifters looping in there. eddies have been shown and brieflydescribed (Centurioni and Niiler, 2003), these data have not yet been used to thoroughly describe discrete 2. Background eddies in the Caribbean. The new drifter trajectories provide an important new Lagrangian The Caribbean Current is the major route by picture of Caribbean eddies, but the eddies were which South Atlantic water flows into the Florida discussed onlybrieflyin the short paper by Current and Gulf Stream and is therefore an Centurioni and Niiler. Theyused a high-resolution important conduit of the upper part of the 1 ( 2 degree) mapping grid to show detailed views northward-flowing meridional overturning circu- of velocityvectors in subregions of the Caribbean, lation (MOC) (Schmitz and Richardson, 1991; but no mean velocitymap of the whole Caribbean Schmitz and McCartney, 1993). The Caribbean was included. This present paper builds on these has two nearlyequal sources of inflow water (see earlier studies byfirst using drifter trajectories to Johns et al., 2002). North of (roughly) Martinique map the circulation and eddykinetic energy(EKE) near 151N Caribbean inflow is primarilyGulf 1 of the whole Caribbean at high resolution (2 Stream water returning southwestward in the degree) and then to systematically identify and North Equatorial Current. This water passes describe several different kinds of discrete eddies in through the Leeward Islands of the Lesser order to show where theyoccur in relationship to Antilles, through the Windward Passage between the mean current field and to estimate their Cuba and Hispaniola, and through the Mona translation rates, swirl velocities, and overall sizes. Passage between Hispaniola and Puerto Rico. Eddies in the Caribbean are important for South of 151N Caribbean inflow is primarilyof several reasons. First, the various source waters tropical and South Atlantic origin. South Atlantic of the Caribbean Current coming from the North water crosses the equator in the NBC and flows Atlantic, South Atlantic, and nearbyAmazon and northwestward along the continental margin of Orinoco Rivers are advected and stirred bythe South America in the form of a coastal current eddies’ swirl velocity, which can be much larger (Candela et al., 1992), in NBC rings (Johns et al., than the velocityof the mean field. Numerous 1990; Didden and Schott, 1993; Richardson et al., large and energetic eddies translating westward 1994; Fratantoni et al., 1995), plus some in Ekman through the Caribbean blend and mix the source transport in the ocean interior (Mayer and waters together. Second, eddies interact with the Weisberg, 1993). The NBC retroflects near 61N boundaries and often sweep near-shore water and feeds into the eastward-flowing North ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 433

Equatorial Countercurrent. Periodically, roughly Caribbean to the Gulf Stream are a consequence 8–9 times per year, large 400-km diameter NBC of the superposition of a realistic MOC and a rings pinch off from the retroflection and translate realistic wind-driven circulation. Without the northwestward toward the Caribbean at around MOC there were no NBC rings and much lower 15 cm/s (Johns et al., 2003; Garzoli et al., 2003; EKE. Thus, model simulations provide evidence of Goni and Johns, 2003; Fratantoni and Richard- a link between the MOC, the formation of NBC son, 2004). Most rings translate northward just rings, the variabilityof flow into the Caribbean, east of the Antilles to around 14–181N where the and the eddyfield there ( Fratantoni et al., 2000; rings stall and decay( Fratantoni and Richardson, Johns et al., 2002; Barnier et al., 2001). The eddy 2004; Goni and Johns, 2003). Some rings dis- field in the Caribbean thus could be caused by appear after theycollide with the continental rings (or their remnants) translating through the margin and islands south of 141N. Most South island passages into the Caribbean where the ring Atlantic water carried byrings enters the Car- vorticityacts as a perturbation for instabilities of ibbean, episodically, through island passages south the Caribbean Current (see Murphyet al., 1999 ). of 181N(Fratantoni and Richardson, 2004), NBC rings have been tracked with surface although some near-surface water in the rings drifters that were trapped and looped in the rings’ passes northward around the islands. swirl velocity( Richardson et al., 1994; Fratantoni Of the 28 Sv (1 Sv ¼ 106 m3/s) of total transport and Richardson, 2004), bysatellite altimetrythat through the Caribbean, 10 Sv enter the Caribbean measured the positive sea-level anomalies in the through the southern or Windward Islands pas- central regions of rings (Didden and Schott, 1993; sages, 8 Sv enter through the northern or Leeward Goni and Johns, 2001, 2003), and byocean-color Islands passages, and 10 Sv enter through the images that detected the chlorophyll-rich water Greater Antilles Islands passages (Johns et al., from the Amazon advected around the periphery 2002). Roughly, half of the Caribbean transport is of rings (Johns et al., 1990; Fratantoni and derived from the South Atlantic and around half Glickson, 2002). Some rings tracked byaltimetry (or more) of the South Atlantic water is trans- appeared to coherentlyenter the Caribbean, but ported byNBC rings ( Johns et al., 2002, 2003). their signal became weaker there, and theycould Maps of historical shipdrift velocityand EKE not be tracked for long (Goni and Johns, 2003). show a swift flow of water and a band of high The wide spacing between satellite altimeter lines EKE extending from the equator along the north- makes it difficult to see details of the ring eastern coast of South America through the anomalies as theyenter the Caribbean. Most Caribbean into the Gulf Stream (Wu¨ st, 1964; drifters that had been in rings near the Antilles Wyrtki et al., 1976). This pattern has recentlybeen passed through the island passages into the confirmed and refined using surface drifters and Caribbean (or grounded on the islands or were satellite altimetry(see Fratantoni, 2001). The retrieved at sea and taken ashore), implying that recent drifter, current meter, and altimetrydata most ring water enters the Caribbean. However, show that energetic NBC rings translating north- there has been no real evidence from drifters that ward along the northern coast of South America rings pass coherentlythrough island passages. up to the Caribbean are primarilyresponsible for Satellite altimetryhas been used to measure the the band of high EKE, and to a large extent anomalous sea-level signal of some eddies in the responsible for the significant (rectified) mean Caribbean (Nystuen and Andrade, 1993; Schott velocityfield of the Guiana Current ( Richardson and Molinari, 1996; Carton and Chao, 1999; et al., 1994). These rings appear to be responsible, Pauluhn and Chao, 1999; Andrade and Barton, either directlyor indirectly,for the high EKE in 2000; Astor et al., 2003). Several large eddies were the Caribbean (Fratantoni et al., 2000; Johns et al., identified that translated westward in the Carib- 2002). bean at speeds of 15 cm/s with some evidence of Model simulations suggest that NBC rings and the eddyanomalies increasing as theytranslated the high EKE band extending through the (Carton and Chao, 1999). Both anticyclones and ARTICLE IN PRESS

434 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 cyclones were observed, but anticyclones appear to (Fig. 3). The drifting buoydata were acquired be most prevalent. A difficultyin using altimetry from the Global Drifting BuoyData Assembly to observe eddies is the necessityof estimating the Center at the National Oceanic and Atmospheric seasonallyvarying sea level of the Caribbean, Administration Atlantic Oceanographic and Me- given its complicated structure, so that sea-level teorological Laboratoryin Miami Florida. The anomalies can be accuratelyestimated and majorityof the drifters were similar to the tracked. Despite these difficulties, some altimeter WOCE–TOGA Lagrangian drifter described by observations suggest a possible connection be- Sybrandy and Niiler (1991). Drogues were at- tween NBC rings and Caribbean anticyclones tached below the surface float centered at a depth (Carton and Chao, 1999). Satellite ocean-color of 15 m. Positions of a few drifters were based on images also have been used to identifya Caribbean satellite fixes obtained every3 days(to reduce eddy, which was confirmed by shipboard measure- costs), which provided some slightlyodd-looking ments (Corredor et al., 2004). interpolated trajectories compared to the usual Some of the first direct velocityobservations of ones based on several satellite fixes per day. Since eddies in the Caribbean were made bydrifters the looping period of the eddies was usuallymuch deployed in 1975–1976 when satellite tracking first greater than 3 days, these interpolated data were became available (Molinari et al., 1980, 1981; used where appropriate. Earlier trajectories from Heburn et al., 1982; Kinder, 1983; Kinder et al., 1975 to 1976 were not included because of 1985). The drifters meandered through the Car- different drogues, drogue depths, problems with ibbean and occasionallylooped in a few anti- drogue retention and drifter slip due to wind and cyclones illustrating time-dependent fluctuations. wave forces on the surface floats. Because the earlytrajectories were relatively Box averages of velocityand EKE were sparse, it was difficult to infer much about eddy generated bygrouping all available 6-hourly characteristics. The drifter observations were drifter velocities into boxes. Mean velocitywas grouped to map the mean velocity(and variance) calculated as the sum of all u velocitycomponents in 2-degree bins, creating one of the first directly in a box (in the x direction) divided bythe number measured velocityfields of the Caribbean ( Moli- of observations, similarlyfor v velocityin the y nari et al., 1980). Other earlymaps of directly direction. EKE was calculated byaveraging the u measured currents in the Caribbean were based on and v velocityvariances (variances about the mean historical ship drifts (see Wu¨ st, 1964). It is velocities) in each box. Standardp error of mean instructive to see the changes between the broad velocitywas estimated using 2st=N; where s is current in these earlymaps and the details of the variance of velocityabout the mean velocity, N smaller-scale features being resolved in the more is the number of 6-hourlyvelocityobservations recent higher-resolution ones as shown by Cen- divided by4, and t is the integral time scale of the turioni and Niiler (2003) and in this present paper. Lagrangian autocorrelation function, which was More complete descriptions of the Caribbean estimated to be 2 days. In practice, the number of circulation and aspects of its variabilityhave been degrees of freedom, N=t; was estimated by given by Wu¨ st (1964), Schott and Molinari (1996), summing the number of 2-dayintervals for which Mooers and Maul (1998), Centurioni and Niiler each drifter was within a box. (2003), and Gyory, Mariano and Ryan at the Rossbynumber ðRoÞ was estimated for some website: http://oceancurrents.rsmas.miami.edu. eddies by Ro ¼ z=f ; where z ¼ 2o is the relative vorticityof an eddy(assuming solid bodyrota- tion), o is the angular velocityof the eddy,and f is 3. Methods the Coriolis parameter, f ¼ 2O sin y: If a drifter was located outside the (approximately) solid body During the last 8 years (1996–2003) over 73,000 rotation core region of an eddy, then the 6-hourlypositions and velocities were measured calculated Rossbynumber could be an over with 212 drifting buoys in the Caribbean Sea estimate. The relative vorticityof an eddycan be ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 435 W ° ons come from W. Manydrifters looped 1 N81 W63 1 ° W66 ° W69 ° W72 ° Caribbean Drifter Trajectories W75 ° W78 ° W81 ° 84 W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 for long times in eddies in the eastern Caribbean and in the western Caribbean southwest of Cuba. Fig. 3. Summary of 212 surface drifter trajectories in and near the Caribbean and over 50 drifter years of data from the years 1986–2003. Most observati the years 1998–2000. Arrowheads are spaced at 10 days. Trajectories clearly show the cyclonic Panama–Colombia Gyre centered near 11 ARTICLE IN PRESS

436 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 expressed in terms of its period of rotation (T in appears to be only50-km wide. Numerous days) as f =ðT sin yÞ: T was estimated from the other loops in the gyre extended eastward to looping drifters. Sin y for the mid-part of the around 751W suggesting that the western part is eastern Caribbean (14–151N) is around 0.25. embedded in a much larger gyre. The eastern part contains significant time variabilitypartiallydue to two westward translating cyclonic eddies 4. Drifter data located there. Mooers and Maul (1998) have reported that the 4.1. Trajectories Panama–Colombia Gyre ‘‘consists of an intense cyclone that together with an adjoining anti- A summaryof drifter trajectories ( Fig. 3) cyclone and cyclone is embedded in a larger but illustrates some characteristics of currents and weaker cyclonic gyre’’. The cyclonic eddies ob- eddies. In general, the Caribbean is filled with a served bydrifters could be interpreted as the complex tangle of trajectories indicating time- ‘‘adjoining cyclone’’, but there are no obvious dependent fluctuations, which are primarilydue anticyclonic loops in the vicinity of the gyre and no to eddymotions. In particular, the region south- confirmation of the ‘‘adjoining anticyclone’’. west of Cuba appears to be denselypacked with Monthlymaps of trajectories in the gyre,not convoluted trajectories, which give little indication shown, do not reveal an obvious seasonal of anymean flow. In contrast to this are two variation, aside from the two cyclones, although regions where trajectories lie roughlyparallel to the data are rather sparse at this resolution. each other, indicating that mean flow dominates Thus, the cyclonic gyre was quasi-permanent over the eddies. One is in the southern Colombia during the 19-month period from June 1998 to Basin where numerous drifters circled the counter- December 1999 when the data are available. clockwise Panama–Colombia Gyre. The second is Maps of historical shipdrift velocities, summarized in the western Caribbean where drifters went by Rennell (1832) and Wu¨ st (1964), suggest westward just north of Honduras and then turned the gyre existed over much longer time scales and went northward in the Yucatan Current, (200 years). which exits the Caribbean between Yucatan and Ten drifters transited through the whole Car- Cuba. ibbean illustrating the main pathways of flow Expanded plots of the Panama–Colombia Gyre through the Caribbean Sea (Fig. 5). Four other are shown in Fig. 4. Almost all drifters that circled drifters almost completed the trip but grounded on the gyre were launched there. One circled the gyre the coast of the Yucatan Peninsula, due in part to 11 times during 13 months (Fig. 4B). Thirteen of the mean trade winds blowing onshore there. Most these drifters eventuallyleft the gyrebydrifting of the drifters looped in eddies including some northwestward. Onlyone drifter from the eastern large ones, 250 km in diameter, in the Venezuela Caribbean was entrained into the gyre circulation. Basin. On average the 10 drifters spent 6.2 months The gradual flushing out of the gyre of drifters in passing through the Caribbean, with a range of launched there and the lack of others entering the 3.4–8.0 months. gyre from the east is attributable to a mean The envelope of these 10 drifter trajectories northwestward near-surface flow driven bythe graduallynarrows to 100 km in the west mean easterlytrade winds. Manyother drifters where drifters were funneled through gaps in the grounded on the east-facing coast of Nicaragua Jamaica Ridge. West of the ridge the envelope due to onshore winds there. widens somewhat before narrowing again as Several drifters looped in the nearlycircular drifters were funneled into the 150-km wide western part of the gyre centered near 11.01N Yucatan Channel. Five of the drifters exiting in 81.51W and bounded by9–13 1N 80–841W. The the Yucatan Current returned southward again on southwestern part of this gyre is tightly confined to the east side of the channel indicating inflow to the the coast of Costa Rica, where the main current Caribbean there. ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 437 W ° meter cyclones ). Arrowheads are Fig. 11 W74 ° W that circled in the gyre 11 times during June 1998–July 1 W76 N 75.6 ° 1 W78 ° Panama - Colombia Gyre N. These cyclones, which are difﬁcult to see here, will be shown later without background drifters ( W80 1 ° W82 ° 84 N N N ° ° ° (A) 14 12 10 Fig. 4. (A). Drifter trajectories in the Panama–Colombia Gyre region. A major source of the variability in the eastern half of the gyre is two 200-km dia that drifted westward at 3–5spaced cm/s at near 1-dayintervals. 11 (B) Trajectoryof1999. a drifter This launched is in the the Panama–Colombia longest-lasting Gyrenear drifter 10.8 in the gyre. Arrowheads are spaced at 1-day intervals. ARTICLE IN PRESS

438 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 W ° 980701 980601 W74 ° 980815 990713 W76 ° 981128 ) Continued Fig. 4. ( 980917 W78 Panama - Colombia Gyre ° 26 9905 981017 990130 W80 ° W82 ° 84 N N N ° ° ° (B) 14 12 10 ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 439 W ° hrough island the eastern Caribbean. W63 ° W66 ° W69 ° W72 ° Long Trajectories W75 ° W78 ° W81 ° 84 W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 passages; seven others were launchedArrowheads in are the eastern spaced Caribbean. at Some 10 looping days. trajectories reveal the presence of eddies, in primarilyanticyclones, Fig. 5. Trajectories of 10 drifters that drifted through the Caribbean and exited through the Yucatan Straits. Three drifters entered the Caribbean t ARTICLE IN PRESS

440 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463

4.2. Speeds near Tobago and with the southern islands in that region. Two fast flows (425 cm/s) appear to enter A plot of drifter speeds along trajectories (Fig. the southeastern Caribbean passages near 121N 6) reveals regions where high and low speeds and 141N (611W) and merge into the main dominate. Although highest speeds (100 cm/s) Caribbean Current by66 1Wnear131N. are scattered throughout the Caribbean, they A second somewhat slower (25–30 cm/s) and appear to be concentrated in the Yucatan narrower band of westward flow, which lies in the Current and along the southern boundary northern Caribbean just south of Hispaniola especiallynoticeable in the eastward-flowing (16–171N), merges with the main southern part southern part of the Panama–Colombia of the Caribbean Current near 751W. This north- Gyre and in the Caribbean Current along the ern band of current appears to extend eastward to boundaryof Venezuela and Colombia. the islands near 171N where water enters from the Low speeds are seen throughout the Caribbean North Equatorial Current and from NBC rings but are dominant in the northeastern region inside that stall and decaynear 15–18 1N. The two bands and outside the Caribbean and also just south of of westward flow were identified by Centurioni Cuba. and Niiler (2003) and also appear in the lower- resolution velocitymap shown by Fratantoni 4.3. Mean velocity vectors (2001). Elsewhere the mean velocityis generallyslower The pattern of trajectories and their speeds in (o25 cm/s) to the west although the mean flow just different parts of the Caribbean look very south of Cuba is particularlyweak 2 cm/s and different, suggesting different current regimes difficult to see. A few red arrows are located near including different mean flows and different kinds the Windward Passage near 201N741W. The mean and numbers of eddies. To explore these geogra- velocityin this region (19.0–20.5 1N 73.5–75.01W) phical variations more quantitativelybin-averaged is 1375 cm/s southwestward, indicating significant mean velocityvectors were mapped in Fig. 7 inflow. Northeast of the Caribbean in the area of and velocityvariances about the mean, or EKE, the Antilles Current is 6 cm/s flow to the north- in Fig. 8. west. The swiftest part of the Caribbean Current, 425 cm/s, shown byred arrows in Fig. 7, flows 4.4. Eddy kinetic energy westward along the southern boundaryof the Venezuela Basin near 12–151N with mean speeds High values of EKE (4800 cm2/s2) shown by up to 80 cm/s. It continues through the mid-part of yellow–red colors are prevalent in the central the Colombia Basin near 14–161N, northwestward Venezuela and Colombia Basins and in a few through gaps in the Jamaica Ridge near 811W, areas of the Yucatan and Cayman Basins (Fig. 8). westward north of Honduras between 181Nand Coinciding with the high EKE values in the central 191N, and northward along Yucatan near Caribbean are numerous energetic anticyclones 86–871W where speeds reach 120 cm/s. In the discussed below. Low EKE values shown byblue Colombia Basin, the Caribbean Current vectors colors (200 cm2/s2) dominate the region north- merge with those in the westward-flowing northern east of the Caribbean, the region just south of limb of the Panama–Colombia Gyre, which Cuba and in the western Panama–Colombia Gyre. extends from 91N to roughly14 1N. In the southern The low EKE values in the gyre are located near part of the gyre mean speeds are around 75 cm/s. some of the largest mean vectors, indicating the The swift part of the Caribbean Current can be low temporal variabilityof the gyrestructure in traced back to the southeastern corner of the the west. Low values of EKE are also located in Caribbean where tropical and South Atlantic the swift part of the Caribbean Current north of water enters the Caribbean from the NBC and Honduras 18–191N, indicating low variabilityof from rings that collide with the continental margin this part of the current. ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 441 W ° peeds shown in red are W63 ° W66 ° W69 ° W72 ° Drifter Speeds W75 ° 100 cm/s are clustered along the southern boundaryof the Venezuela and Colombia Basins and along the 4 W78 ° W81 ° 84 0 to 20 20 to 40 40 to 60 60 to 80 80 to 100 100 plus cm/sec W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 Fig. 6. Drifter locations color-coded to indicate speed. Over 73,000 individual 6-hourlyvelocitymeasurements are shown as colored dots. Fastest s plotted over slowest speeds shown in blue. Fastest speeds western boundaryoff the Yucatan Peninsula. ARTICLE IN PRESS

442 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 W ° W63 ° W66 ° W69 ° -degree bins. Vectors are shown for all bins that contained more than three 1 2 25 cm/s) to ﬂow westward through the southern part of the Caribbean except north W72 ° 4 -degree by 1 2 Mean Velocity Vectors W75 ° W78 ° W81 ° 84 50 cm/s W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 Fig. 7. Mean velocityvectors calculated bygrouping 6-hourlyvelocityvalues into degrees of freedom (see text). Theof swift the Caribbean Panama–Colombia Current is Gyre shown located byred vectors in ( the southern Colombia Basin. ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 443 1900 1000 900 800 700 600 500 400 300 200 100 0 W ° W63 ° W66 ° ) 2 W69 ° /sec 2 -degree bins. EKE is large throughout the eastern Caribbean except in the 1 2 W72 ° -degree by 1 2 W75 ° Eddy Kinetic Energy (cm W78 ° W81 ° 84 ) estimated from drifter velocityvariances in 2 /s 2 W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 Fig. 8. EKE (cm Panama–Colombia Gyre and in the northeastern region north of the Caribbean where EKE is low. ARTICLE IN PRESS

444 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463

5. Caribbean eddies EKE of the looper with largest loops was around 1500 cm2/s2, suggesting that this anticyclone and In order to investigate Caribbean eddies each others are major contributors to the high EKE in drifter trajectorywas inspected for loops and the eastern Caribbean. cusps, which reveal the characteristic motion of a Another example of an anticyclone in the particle in an eddyconsisting of a swirl velocity, central Caribbean is shown in Fig. 10. Anticyclone around the eddycenter plus its translation. Time 5 was tracked with two drifters (one looper) for 4 series of velocitywere studied to estimate when a months as it translated at 14 cm/s from the eastern drifter entered and exited an eddy. Portions of Caribbean near 621W westward up to a location trajectories that contained two or more loops in near the Jamaica Ridge near 741W. Swirl speeds the same direction (loopers) were interpreted to were around 30 cm/s at a diameter of 100 km. The have been in the swirl velocityof a discrete eddy. combined trajectories suggest an overall diameter The term ‘‘eddy’’ is used to refer to both clockwise of around 200 km. The typical maximum swirl rotating anticyclones and counterclockwise rotat- speed of the eastern Caribbean anticyclones is ing cyclones. Loopers were used to obtain around 60 cm/s, somewhat faster than this second information about eddydiameter, rotation period, example. swirl velocity, translation velocity, and eddy trajectories. Since the period of rotation and swirl 5.2. Loopers velocityvarywith radius, these characteristics estimated from drifters are representative of the The drifter trajectories in these two anticyclones radius sampled bythe drifters. This should be kept are representative examples of the numerous in mind when comparing characteristics of differ- anticyclones that dominate the Venezuela and ent eddies. The primaryemphasis here is on the Colombia Basins. A summaryof all loopers in the numerous energetic anticyclones in the Venezuela Caribbean is shown in Fig. 11, and the inferred and Colombia Basins, but eddies in other areas of paths of the eddycenters in Fig. 12, subdivided by the Caribbean are also discussed. color into red anticyclones and blue cyclones. Overall 19% of the total drifter data in the large 5.1. Examples of anticyclones box (Fig. 3) are loopers, and these are almost evenlysplit into 28 cyclonic loopers and 29 A good example of a large and energetic anticyclonic loopers. These loopers were judged anticyclone in the central Caribbean is shown by to have been in 49 different eddies, including 25 trajectories of five different drifters during a 7- cyclones and 24 anticyclones. month period (Fig. 9). Trajectories of two loopers The mean westward velocityof all drifters in the are shown plus five shorter pieces of trajectories large box (Fig. 3) is 9.170.4 cm/s (Table 2). (not official loopers) considered to have been in Loopers translated at an average 5.270.9 cm/s, the swirl velocityof this anticyclonefor various anticyclonic loopers at 8.271.3 cm/s and cyclonic amounts of time. The anticyclone translated from loopers at 2.271.1 cm/s. A general trend of near the southeastern corner of the Caribbean, anticyclones translating at close to the mean near 641W, 1700 km westward up to the Jamaica velocityof all drifters and faster than cyclones Ridge near 801W. The longest looper consisting of also holds for the eastern Caribbean and appears 8.5 loops over 4 months at an average diameter of to be due to anticyclones being located within the 167 km (Table 1) went westward at 11 cm/s. The main Caribbean Current in the central basin and envelope of the looping trajectories suggests the the cyclones being located near the boundaries eddydiameter was nearly300 km or around half outside the regions of higher mean velocity. the north–south extent of the Caribbean there. The Anticyclones were observed to translate at very overall diameter of the anticyclone was probably different speeds in different areas but at velocities larger than this. Fastest swirl velocities were similar to the mean velocityin these areas. The around 70 cm/s at a diameter of 200 km. The implication is that eddies were advected bythe ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 445 W ° ranslated alf of the W60 ° 990701 990824 W63 ° ). This anticyclone was tracked with drifters for Table 1 990908 W66 ° 990929 W69 ° Drifters in Anticyclone 1 991029 W72 ° 991130 W75 ° 000110 000127 N) just north of anticyclone 1 and generated energetic near-inertial oscillations in it with an amplitude of 75 cm/s and period near 2 days. These W78 1 ° 81 N N N N ° ° ° ° 9 18 15 12 north–south extent of theeastward Venezuela (near Basin. 15 Arrowheads are spaced at 1-dayintervals. In mid-November, 1999, hurricane Lennywith winds of 135 knots t oscillations, which graduallydecayedover 2 weeks, are superimposed on the loops. anticyclone nearly7 months as it translated 1700 km across the Caribbean at an average velocityof 11 cm/s. The diameter of the largest loops was 300 km, around one-h Fig. 9. Trajectories of ﬁve different drifters looping in and around anticyclone 1 during July 1999–January 2000 ( ARTICLE IN PRESS

446 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463

Table 1 Eastern Caribbean anticyclones

Anticyclone Days Number Period Diameter Swirl velocity U V number tracked loops (days) (km) (cm/s) (cm/s) (cm/s)

1 119 8.5 14 167 43 À11.0 0.8 2 22 1.8 12 222 66 À11.0 À3.1 3 41 2.3 18 187 38 À21.4 0.7 4 70 4.2 17 109 24 À16.6 À0.7 5 101 7.3 14 108 28 À12.1 2.5 6 103 (10.8) 10 84 32 À10.0 3.2 7 51 2.7 19 235 45 À9.5 1.9 8 52 2.5 21 162 28 À13.6 5.2 9 56 3.0 19 265 52 À16.4 10.0 10 84 5.8 14 180 45 À12.8 5.5 Average 15.771.1 172719 4074 À13.471.2 2.671.2

Note: Average values (and standard errors) were calculated for the 10 major anticyclones in the eastern Caribbean (see Fig. 12). For anticyclone 1 the longest looper of the two identiﬁed was used in the estimates. Looping period of rotation was estimated by dividing the days of data by the number of loops. Swirl velocity was estimated from the square root of the sum of u and v velocityvariances about the mean velocityin the x and y directions, assuming that most velocityvariance was derived from the looping motion about the mean anticyclone advection velocity ðU; VÞ: Diameter was estimated from the swirl velocitytimes the period divided by p: The typical maximum swirl velocity in the anticyclones was around 60 cm/s, and the typical diameter of the largest loops was around 200 km. The Rossbynumber of the anticyclonesusing their average period of rotation is À0.25. local background current in which theywere translated at 1.373.0 cm/s; the low mean velocity embedded. is due to the eastward translation of one cyclone Details of the eddies are discussed below near the southern boundaryof Hispaniola and grouped into four regions—the eastern Caribbean another cyclone that was stationary there. Both of (Venezuela and Colombia Basins), the Panama– these cyclones appear to have formed when Colombia Gyre, the western Caribbean southwest anticyclones impinged on the southern coast of of Cuba, and the area of the Antilles Current in Hispaniola. The 10 main anticyclones in the the northeast. The data in these regions were eastern Caribbean are listed in Table 1. The grouped and some statistics summarized in average westward velocityof these is 13.4 7 Table 2. 1.2 cm/s, and northward velocityis 2.6 71.2 cm/s based on grouping the 10 average velocities of the individual anticyclones. 6. Eastern Caribbean anticyclones Half of the anticyclone trajectories started in the far eastern Caribbean near or east of 651W The summaryplot of loopers ( Fig. 11) shows including two just inside the Lesser Antilles, one that the Venezuela and Colombia Basins are near 13.01N 62.31W and the other near 15.61N dominated byanticyclones (red trajectories). Some 62.61W(Fig. 11). This adds support to the cyclones (blue) were observed there, but they are hypothesis that the anticyclones form from rem- fewer in number and tend to be located near the nants of NBC rings. Eight trajectories ended near boundaries. A box average of the central region the Jamaica Ridge, which appears to disrupt the between 651W and 751W reveals that 25% of all typical eddy circulation so that the drifters ceased the data are in loopers and that 71% of these are looping. Anticyclone 6 stalled near the ridge before anticyclonic. The mean westward velocity of the 10 the drifter stopped looping. The inferred anti- anticyclonic loopers in the box is 14.472.5 cm/s cyclone trajectories tend to lie along two bands compared to the mean velocityof all data in the near 151N and 171N, although three of the 151N box 15.470.9 cm/s. The three cyclonic loopers anticyclones drifted northwestward and merged ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 447 W ° -day ruary9, 1999 as 981017 W60 ° W63 ° 981126 W66 ° 990104 W69 ° Drifters in Anticyclone 5 990209 W72 ° W75 ° W78 ° 81 N N N N ° ° ° ° 9 18 15 12 it translated 1200 kmintervals. through the Caribbean at an average velocityof 14 cm/s. The diameter of largest loops was around 200 km. Arrowheads are spaced at 1 Fig. 10. Trajectories of two drifters looping in and around anticyclone 5. This anticyclone was tracked for nearly 4 months from October 17, 1998 to Feb ARTICLE IN PRESS

448 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 yclonic and W ° n Caribbean (Venezuela rs, and these were almost W63 ° W66 ° W69 ° W72 ° All Loopers W75 ° W78 ° W81 ° 84 W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 equally subdivided into cyclonic andand anticyclonic, Colombia although Basins) different appears distributions toanticyclonic were be loopers observed dominated in overlap by different southwest anticyclonic parts of of loopers, Cuba. the and Arrowheads Caribbean. cyclonic are The loopers easter spaced are at most 10 often days. found near the boundaries there. Numerous c Fig. 11. Drifter trajectories in 28 cyclonic loopers (blue lines) and 29 anticyclonic loopers (red lines). Overall, 19% of the drifter data were loope ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 449 . ° 63 Table 1 w cyclones there. r eddies that clearly N appear to translate 1 ° 66 ° 69 10 3 N, although three anticyclones near 15 ° 1 72 8 5 N and 17 1 7 ° 9 75 6 4 Eddy Trajectories ° 78 W near the Jamaica Ridge. Ten of the main anticyclones are numbered and details given in 2 1 1 ° 81 N ones near 75 1 ° 84 W appear to lie along two preferred paths centered near 15 1 W ° 87 ° ° ° ° N ° 21 18 15 12 9 translated. Note the preponderance of anticyclones translating westward through the Venezuela and Colombia Basins as compared to the relatively fe Fig. 12. Trajectories of 19 cyclones (blue) and 19 anticyclones (red) inferred from looping drifter trajectories. Trajectories were plotted only fo The anticyclones nearnorthwestward 70 and converge with the 17 ARTICLE IN PRESS

450 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463

Table 2 Looper statistics

Whole area 9–221N Eastern Caribbean Western Caribbean NE of Caribbean 61–881W 12–18.31N 65–751W 19–221N 79–871W 18.3–221N 61–741W

Total data 73,332 14,196 15,104 14,061 % Loopers 19 25 38 14 % Anticyclones 50 71 37 53 All U (cm/s) À9.170.4 À15.470.9 À2.070.7 À4.970.5 Loopers U À5.270.9 À10.672.0 À1.671.3 À4.771.4 Anticyclonic U À8.271.3 À14.472.5 À2.572.0 À5.972.0 Cyclonic U À2.271.1 À1.373.0 À1.171.6 À3.472.1

Note: Four regions are listed including the whole area (Fig. 3) and three smaller subregions consisting of the Eastern Caribbean south of Puerto Rica and Hispaniola which contains 10 energetic anticyclones, the Western Caribbean southwest of Cuba containing numerous cyclones and anticyclones, and the Antilles Current region northeast of the Caribbean which has a mixture of cyclones and anticyclones. Mean eastward velocity ðUÞ and standard error are listed for all the data in the region and for the loopers including anticyclonic and cyclonic. Northward velocity was omitted because it is usually smaller than the standard error and because the eddies translated mainlywestward. with the 171N ones near the Jamaica Ridge the northern, eastward-flowing, part of the anti- southwest of Haiti (near 171N751W). Curiously, cyclones that impinge on the southern boundaries the northern anticyclones translated at around the of Hispaniola and Puerto Rico. Data grouped same speed as the southern ones despite faster quarterlyinto 1-degree north–south bands be- westward velocities in the south. tween 651W and 751W suggest that the two jets 1 The two bands of anticyclones appear to shown in the 2-degree profile using non-looper coincide with the structure of the mean velocity data exist throughout the year. field. Specifically, the southern band of anti- The similarityof velocityprofiles raises the issue cyclones coincides with anticyclonic shear located of the relationship between the anticyclones and on the northern side of the main Caribbean the westward jets. One possibilityis that the Current as shown byred arrows in Fig. 7. The anticyclones cause the maxima and minima in northern band of anticyclone trajectories lies near the mean velocityprofile. The looper velocity the northern edge of the northern band of red profile in Fig. 13 shows that the swirl velocityof arrows near 171N(Fig. 7). The two bands of anticyclones, which is westward south of the eddy anticyclones merge where the northern current center and eastward north of the center, in increases in speed as shown byan increased combination with the westward translation of the number of red arrows in Fig. 7 near 751W. eddycenter results in a rectified mean velocity A meridional velocityprofile across the eastern profile that looks like the mean velocityprofile Caribbean (65–751W) using drifter data but created byexcluding the loopers. Presumably, excluding loopers (Fig. 13, right panel) clearly some other drifters that were not classified as shows two jets or maxima in the westward flow, loopers were advected bythe outer portions of the one of 3476 cm/s centered near 131N, the other of anticyclones and contributed to the mean (recti- 1972 cm/s centered near 171N, bounded on the fied) velocityprofile created byexcluding loopers. north byslow eastward flow 173 cm/s. The The idea is that the two bands of anticyclones add equivalent profile using onlylooper data ( Fig. 13, the double-jet structure to the Caribbean Current, left panel) has a similar structure, a maximum of which without the anticyclones might gradually 4977 cm/s near 131N, a second maximum of increase in westward velocitytoward the south. A 1875 cm/s near 171N, bounded on the north (near second possibilityis that the Caribbean Current, 181N) byswift 26712 cm/s eastward velocity. independent of anticyclones, consists of two The swift eastward current near 181N is caused by westward jets and that the anticyclones form and ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 451

Eastern Caribbean 65W to 75W Loopers Only Eastern Caribbean 65W to 75W Loopers Excluded 19 19

18 18

17 17

16 16

15 15 Latitude (degrees) Latitude (degrees)

14 14

13 13

12 12 -50 -40 -30 -20 -10 0 10 20 30 -40 -30 -20 -10 0 Eastward Velocity (cm/sec) Eastward Velocity (cm/sec)

1 Fig. 13. Meridional velocityprofile of eastward velocityaveraged bygrouping individual velocities in 2-degree north–south bins in the central Caribbean (65–751W). Note the similar double-jet structures of the profile created using onlylooper data and the profile using non-looper data. grow located in the anticyclonic shear on the Estimates were made of the energyexchange northern edge of the jets. The EKE of the between the eddies and the mean field using anticyclones increased around 37% as they trans- r/u0v0SqU/qy+r/v0v0SqV/qy (and omitting other lated westward, from around 676 cm2/s2 near terms), which expresses the production of EKE by 62.51W (averaged in 3-degree longitude bins) up interaction of the turbulent Reynolds stresses with to 923 cm2/s2 near 74.51W, which lends support to the shear of the mean flow and vice versa, where u0 the second possibilityand is consistent with the and v0 are fluctuations of velocityabout the mean results of an altimetric studywhich suggested that velocity U and V and /S indicates an average anticyclones intensified toward the west (Carton (see Hansen and Paul, 1984). Results using data in and Chao, 1999). 0.5-degree bins between 651W and 751W suggest a ARTICLE IN PRESS

452 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 conversion of EKE to mean kinetic energyand located in two areas of mean eastward currents, standard error 0.3070.26 10À3 erg/cm3 s south the southern Panama–Colombia Gyre and the of the southern jet maximum and near the south- Yucatan Current that flows northeastward. Three ern boundary12.0–13.0 1N. North of this between other clusters are associated with Caribbean 13.01N and 18.51N is a conversion of mean eddies. One is southwest of Cuba concentrated flow kinetic energyto eddyenergy 0.127 near the Isle of Pines. A second is centered between 0.11 10À3 erg/cm3 s. The ratio of the eddyenergy 17.01N and 18.31N just south of Hispaniola. A to the eddyproduction in the region 13.0–18.5 1N third is centered between 15.01N and 16.51N in the suggests a time scale of 49 days, which can be mid-Colombia Basin. A plot of eastward speeds interpreted as the e-folding time of exponentially over 30 cm/s in loopers (not shown) suggest that growing eddies. these three clusters are caused primarilybythe Previous studies of hydrographic sections across eastward component of the eddyswirl velocity.In the Caribbean also have observed two westward particular, the two clusters in the eastern Car- jets (Gordon, 1967; Morrison and Nowlin, 1982), ibbean appear to match the northern parts of and a recent velocitysection across the Caribbean anticyclones. Although there could be eastward near 661W in September 1997 also clearlyshowed flows not associated with eddies, the matching of two westward jets, one of 120 cm/s near 131Nand the swift eastward eddyswirl speeds with the the other of 20 cm/s near 171N separated bynearly clusters of eastward velocities suggests these zero flow near 15–161N(Hernandez–Guerra and counterflows are caused primarilybythe eddies. Joyce, 2000). The southern jet contained water Therefore, most occurrences of fast eastward flow from the Orinoco River, the tropics, and the South in the Caribbean outside of the Panama–Colombia Atlantic. The northern one contained Caribbean Gyre and Yucatan Current are likely due to the surface water and subtropical underwater with swirl velocityof Caribbean eddies, and in the sources in the North Atlantic. We are not aware of eastern Caribbean mainlyanticyclones. anyobservations or anymodel simulations that The looper data were used to estimate the reproduce the two bands of anticyclones, although population and formation rate of anticyclones in there is an indication of a double-jet structure of the eastern Caribbean. In 1998, four anticyclones the Caribbean Current in some (Murphyet al., were tracked with drifters, including two sets of 1999; Fratantoni et al., 2000). Fratantoni et al. two anticyclones tracked simultaneously (numbers (2000) and also Johns et al. (2002) discussed three 3 and 10, 4 and 5, Table 1). Since some anti- simulations that showed (1) the wind-driven cyclones could have been missed, this implies a transport (no MOC) from the North Atlantic to minimum formation rate of four anticyclones per be concentrated in the northern part of the year. In the box bounded by 651W and 751W Venezuela Basin, (2) the MOC-driven transport anticyclonic loopers comprised 17.5% of all the (no wind) from the South Atlantic to be concen- drifter data. This implies that around 17.5% of the trated in the southern part of the basin, and (3) the area, which is around 108,000 km2, comprised wind-plus-MOC-driven circulation to have trans- anticyclones. If the typical overall diameter is port concentrations in both the northern and 250 km, then there should be a population of southern parts, implying the existence of two around 2.2 anticyclones in the box at any one time, distinct jets. in agreement with the two pairs of tracked Eastward or counterflows in the Caribbean have anticyclones. At a typical speed of 13.4 cm/s been repeatedlyreported ( Gordon, 1967; Roem- anticyclones translate through the box in around mich, 1981; Morrison and Nowlin, 1982; Smith 3.0 months, which suggests a formation rate and Morrison, 1989; Morrison and Smith, 1990; around eight anticyclones per year. Choosing a Johns et al., 1999). In order to search for smaller diameter of 200 km results in around 12 substantial counterflows in the drifter data, anticyclones per year. The 8-per-year rate matches all eastward velocities over 30 cm/s were plotted that of NBC rings observed to translate toward the (Fig. 14). Major clusters of eastward velocityare Caribbean Islands (Johns et al., 2003), consistent ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 453 W ° velocities were also W63 ° W66 ° W69 ° W72 ° W75 ° Eastward Drift Velocity over 30 cm/sec W78 ° W81 ° 84 W ° 87 N N N N N ° ° ° ° ° 9 21 18 15 12 Fig. 14. Locations of eastward drifter velocitycomponents that are faster than 30 cm/s, indicating regions of counterﬂows. Plots of slower eastward generated, but those clusters of eastward velocitytended to be more diffuse. ARTICLE IN PRESS

454 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 with the hypothesis that the anticyclones form The Caribbean Current is maximum during July from remnants of the rings. and minimum during November (Fuglister, 1951; The number of looping days in anticyclones Johns et al., 2002); the 3-month lag between (and the percentage of total data in anticyclones) maximum currents and maximum number of varies seasonallyfrom low values, down to 17 days anticyclones could be the time required for anti- during February–April, to high values, reaching cyclones to become sufficiently energetic and trap 136 days during September–November (Fig. 15). particles in closed circulation. Although there is no The implication is that more anticyclones form obvious seasonalityin the number of NBC rings during summer and fall than during winter and formed, rings that form after the summer NBC spring, or possiblythe anticyclones are more maximum transport and while the retroflection is intense and better able to trap drifters in their still clearlyestablished (November–February)ap- swirl velocityduring the months of September– pear to be larger and deeper than those at other November. times (Johns et al., 2003). This seasonalityin ring The inferred seasonal variation of anticyclone structure together with variations of Caribbean population might be related to variations in the inflow might influence how long rings take to velocityof the Caribbean Current (and current penetrate into the Caribbean, resulting in season- shear), which could modulate anticyclone growth. alityof Caribbean anticyclones.

ANTICYCLONE LOOPER DAYS 140

120

100

60 Number of Looper Days

0 1 2 3 4 5 6 7 8 9 10 11 12 13 Month

Fig. 15. Number of looper days in the 10 main anticyclones in the eastern Caribbean (Table 1). The longest looper of the two available was used for anticyclone 1. ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 455

In summary, several characteristics of the anti- island passage at 30 cm/s (see Wilson and Johns, cyclones observed in the eastern Caribbean suggest 1997) in around 30 days, which seems consistent theycould have formed from the anticyclonic with what we know about rings. Some rings remnants of NBC rings. First, the anticyclones collided with the southern Caribbean islands and ﬁrst appeared in the eastern Caribbean just west of disappeared rapidlyin around a month; other the Antilles where rings collided with the islands rings passed northward to 15–181N where they and disappeared. Second, the estimated formation decayed or disappeared over a few months just rate and the measured translation rate of anti- east of the islands. cyclones are very similar to those of rings. Third, There is verylittle direct evidence from drifters the swirl velocityand rotation rate of anticyclones of NBC ring water forming an anticyclone inside are about half that of rings and the overall the Caribbean. The best evidence comes from a diameter of the anticyclones is somewhat smaller single drifter that had been looping in an NBC ring than that of rings. These observations are con- located east of the islands (Fig. 16). The drifter sistent with the concept that a ring’s vorticity looped and translated toward the island arc, decays somewhat in passing through the Antilles passed between Dominica and Guadeloupe Is- so that anticyclones formed on the inside of the lands, made two small (20 km) cyclonic loops islands from ring remnants would be weaker than immediatelywest of Dominica and then one-and- the original rings. Fourth, the anticyclones appear a-half larger (140 km) anticyclonic loops and a to lie in two bands that coincide with westward jets cusp in what could have been an anticyclone of the Caribbean Current, implying that anti- forming inside the Caribbean (red trajectory). The cyclones are energetic components of the circula- drifter then meandered westward through the tion. However, if anticyclones form on the north- Venezuela Basin leaving no further clues about ern anticyclonic shear side of the jets from the possible anticyclone. instabilities of the Caribbean Current, then we High-resolution numerical models of ring-like would also expect to see cyclones form along the eddies colliding with islands suggest that some cyclonic shear sides of the jets, which we do not eddies can pass coherentlythrough the island (except for one cyclone near the southern passages when the eddydiameter is much larger boundaryof the Venezuela Basin). It is possible than the islands and passages (Simmons and Nof, that instabilities of the anticyclonic shear of the 2002; Garraffo et al., 2003). Garraffo et al. report jets help organize and amplifythe injected anti- that inside the Caribbean the simulated sea-surface cyclonic vorticity of rings, and this could con- height anomalies are generallysmaller than those tribute to the energyand longevityof the anti- of rings, the anomalies propagate westward, and cyclones and their observed ampliﬁcation as they theyoften originate as part of a ring that remains translate westward. for some time east of the Antilles. How could NBC ring water pass through island It is also possible that local wind stress curl can passages and reform as an anticyclone? Results of form anticyclones. Oeyet al. (2003) , using a a laboratorystudyof an eddycolliding with numerical simulation of Caribbean circulation, several seamounts shows that water can be peeled found that a patch of anticyclonic wind stress curl off the outer part of the eddy, pass between located southwest of Hispaniola can force a seamounts as a streamer, and reform into an eddy depression of isopycnals through Ekman (or two eddies) on the other side of the seamounts pumping, resulting in an anticyclone. These Cenedese and Adduce (2002) see also Cenedese ‘‘Hispaniola anticyclones’’, as Oey et al. call them, (2002) and Wang and Dewar (2003). Streamers subsequentlydrifted westward through the Car- from some eddies passed through two passages ibbean at a rate of about 3–4 per year. Some of the and sometimes two eddies formed. The diameter of drifter-tracked anticyclones started farther east the central portion of a ring, inside the maximum than Hispaniola, but perhaps theycould have been swirl velocity, is around 200 km. This amount of augmented bythis process in the region southwest water could pass through a typical 40-km wide of Hispaniola. ARTICLE IN PRESS

456 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 ng OS ﬂoat, p implying that thwestward up to W ° 0712 0716 2-dayperiod). Arrowheads are spaced 0829 0909 W58 ° 0824 0919 0930 M Drifters in an NBC Ring D 0915 W60 G ° 62 1008 15-km anticyclonic cusps in the red trajectory in August are interpreted to be near-inertial oscillations ( 1023 W ° 64 N N N ° ° ° 18 14 16 Fig. 16. Two drifters looping in an NBC ring that translated southwestward toward the islands of Guadeloupe (G), Dominica (D), and Martinique (M) duri July–September 2000. One drifter (red) enteredsome the ring Caribbean between anticyclonic Guadeloupe vorticity andcontained Dominica gaps, entered and and the made the one-and-a-half Caribbean ﬂoat anticyclonicthe could and loops islands. not plus Some was be a small cus forming tracked with an adequate anticyclone resolution in there. the The Caribbean. data Its trajectoryadds from evidence the that other the ring drifter translated (blue), sou an undrogued RAF at 1-dayintervals. ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 457

7. Other eddies that theycould have formed as pieces of the eastern part of the gyre that temporarily separated 7.1. Eastern Caribbean cyclones from the more intense western part. The third is that theycould have been formed as inverse All of the cyclones in the eastern Caribbean (modeled) ‘‘Hispaniola anticyclones’’ by a local appear to begin or to lie near topographic features maximum in the wind stress curl that coincides of the basin. The cyclone near the southern side of with the cyclones (Chelton et al., 2004; Oeyet al., the Venezuela Basin, translating westward at 2003). Andrade and Barton (2000) suggested that 21 cm/s, could have been generated bythe cyclonic cyclones form seasonally during the part of the shear on the southern side of the main Caribbean year when strongest meridional gradients in Current where it flows along the boundary. Two salinityand wind occur. loopers in different cyclones south of Hispaniola began looping where the eastward swirl velocityof 7.3. Western Caribbean eddies two different anticyclones impinged on the southern coast of the island; one cyclone remained Two cyclones (diameter 100 km, swirl velocity stationary, the other was advected clockwise 30 cm/s) were observed in the southern part of around the anticyclone, eastward along the coast the western basin, and theypresumablyformed then southward. Both cyclonic loopers began with there on the southern, cyclonic shear side of the small 10–50-km diameters and fast 1.5–4.0-day Caribbean Current. The region north of this and loops ðRo ¼ 2:220:8Þ; which graduallyincreased southwest of Cuba is dominated by10 cyclonic in diameter and period. Two cyclonic loopers loopers and nine anticyclonic loopers. In the began near passages through the eastern islands, region 19–221N 79–871W, 38% of the data are in and two others were located just inside these loopers, and 63% of these are cyclonic. The mean islands, implying that cyclones are often formed by westward velocityof all loopers in the box is flows through the passages. Two other cyclones 1.671.3 cm/s, indistinguishable from the mean of with small loops were observed in the western all the drifters 2.070.7 cm/s. The implication is Colombia Basin one of which passed over the that slow background zonal flow caused the slow Jamaica Ridge. translation of eddies there. The loopers in the box 19–221N 79–871W were 7.2. Panama– Colombia Gyre cyclones located in seven different cyclones and six different anticyclones. Eight loopers were located in a Two cyclones were observed in the eastern part cyclone–anticyclone pair, which translated slowly of the cyclonic gyre near 11.51N 77.51W(Fig. 11). (2 cm/s) westward. The three longest looping Theytranslated westward at 3–5 cm/s toward the drifters in the Caribbean were in this pair, the central part of the gyre. Swirl speeds of 38 cm/s longest two at 8.6 and 7.9 months in the cyclone. were located at a diameter of around 200 km and Fig. 17 shows a summaryof the loopers in the the period of rotation was around 20 days ðRo ¼ cyclone and anticyclone and four monthly sum- 0:26Þ: The cyclones were observed in 1998 and maries of trajectories in these eddies during 1999 during the months of September–December, November 1999–February2000. The anticyclone suggesting seasonality. Another cyclone was ob- center was located around 200-km west of the served with altimetryin the gyrenear 14 1N during cyclone center. On several occasions drifters loop- August–November 1993 by Andrade and Barton ing around one eddyof the pair began to loop (2000). around the other eddy, suggesting that the near- Three possible formation mechanisms are sug- surface swirl velocityof the two eddies was gested. The first is that the cyclones could have connected. The drifters began to loop in the eddies formed on the southern cyclonic shear side of the in October 1999 when launched. The anticyclone main Caribbean Current jet as it flows westward translated westward and appeared to coalesce with awayfrom the coast of Colombia. The second is the Yucatan Current in February2000 when the ARTICLE IN PRESS

458 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 plots of N N N ° ° ° ° ° ° ° ° ° 22 20 18 22 20 18 22 20 18 ° ° 80 80 e colored gray. Two C 9910 - 0008 ° ° 82 82 C ° ° AC 84 84 AC ° ° W in November 1999 and another is located in the northeastern 1 N 81.0 1 W86 W86 ° ° Feb 00 All C Dec 99 88 88 ° ° 80 80 C 9910 - 0002 Cyclone-Anticyclone Pair ° ° C 82 82 AC ° ° 84 84 AC ° ° W86 Nov 99 Jan 00 All AC W86 ° ° 88 88 ° ° ° ° ° ° N N N ° ° ° 20 18 20 18 20 18 22 22 22 additional small cyclonic loopers are seen; one began near Grand Cayman Island near 19.3 loopers in the cyclone (blue)–anticyclone (red) pair as they translated westward toward the Yucatan Current. Background drifters (non-loopers) ar Fig. 17. Summaries of four loopers in the cyclone (C) and four in the anticyclone (AC) of a cyclone–anticyclone pair southwest of Cuba, and four monthly corner in December–February. Arrowheads are spaced at 5 days. ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 459 drifters accelerated to the north. The cyclone 17, November panel) and merged with the loops of translated westward until August 2000 when it the first cyclonic looper. Two additional cyclonic was located near 20.81N 84.51W and the drifters loopers were in this cyclone, which eventually stopped looping. It is possible that the cyclone also reached a diameter of 200 km. Four other drifters coalesced with the Yucatan Current. The cyclone’s near the southern coast of Cuba made small lifetime was at least from mid-October to the end cyclonic loops there, three of them when other of August 2000, a total of 10.5 months and a nearbydrifters indicated larger-scale anticyclonic record for Caribbean eddies. motion near the coast of Cuba. These observations The overall diameter of the anticyclone was are similar to those near Hispaniola and may 200 km, its fastest swirl velocitywas 40 cm/s, indicate an important formation mechanism of and its period 30 days (Ro À0:1). The loopers cyclones. The two successive cyclonic loopers that in the cyclone began in October 1999 with small merged in the cyclone suggest that a series of small 10-km loops, which increased to around 100 km intense cyclones began near the Isle of Pines, that during December 1999–April 2000 and to around the cyclones were advected eastward by the swirl 200 km afterward. Maximum swirl speeds were velocity of the anticyclone, and that the cyclones around 30 cm/s at 10-km diameter, increasing to coalesced growing into a larger long-lasting 60 cm/s at 60-km diameter in November 1999 and cyclone. Drifters in the anticyclone and that at 200 km during May–July 2000. The period of passed within around 10 km of the Isle of Pines rotation increased from around 1 day ðRo 3Þ for tended to loop in the cyclone; those farther away the earlysmall 10-km loops up to around 17 days tended to continue to loop in the anticyclone. Due ðRo 0:2Þ at 200-km diameter at the end. to its slow translation, the anticyclone interacted The western Caribbean anticyclones were ob- continuouslywith the Isle of Pines for 3 months, served northeast of the main Caribbean Current which appears to be sufficientlylong for a series of on its anticyclonic shear side, which suggests that cyclones to form, collect, and grow into the large theycould have formed from instabilities in this energetic cyclone observed. region. It is also possible that anticyclonic vorticity Rather similar small cyclones (40–50-km dia- from the eastern Caribbean anticyclones was meter) have been observed in the vicinityof warm advected over the Jamaica Ridge and reformed core Gulf Stream rings, possiblyformed bytheir as western anticyclones, perhaps amplified by the interactions with the continental margin there anticyclonic shear there. (Kennellyet al., 1985 ). Some other similar A possible formation mechanism of the cyclone cyclones, called frontal eddies and spin-off eddies, in the northern Yucatan Basin is suggested bythe also have been observed along the left edge of the drifter that began to make rapid (1-dayperiod), Loop Current, Florida Current, and Gulf Stream small (10-km diameter) cyclonic loops when the (looking downstream), although the eddies are northern part of the anticyclone (of the anti- usuallyadvected rapidlydownstream unlike the cyclone–cyclone pair) impinged on the southern cyclones observed southwest of Cuba (see e.g. Lee, boundaryof Cuba near the Isle of Pines (near 1975; Lee and Atkinson, 1983; Lee et al., 1991; 831W). Swift 70–80 cm/s eastward flow measured Fratantoni, 1998; Fratantoni et al., 1998). The bytwo drifters within 3 km of the Isle of Pines is large size, small translation velocity, and long life interpreted to have created intense cyclonic shear, of the Cuba cyclone appear to be different from which resulted in the formation of a small intense these other cyclones. cyclone. The cyclone was advected to the eastern side of the anticyclone where, by the end of 7.4. Eddies Northeast of the Caribbean November, the cyclonic loops increased in diameter to around 60 km and the period of rotation Northeast of the Caribbean in the upper right- increased to 3 days (Ro 0:9). Another drifter in hand corner of the large box (Fig. 11) is located six the anticyclone started to loop cyclonically near anticyclones and six cyclones including a cyclone the same place as the first but 15 days later (Fig. just north of Hispaniola translating toward ARTICLE IN PRESS

460 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463

Windward Passage. In this area (18.3–22.01N topography. The anticyclonic looper data suggest 61–741W), 14% of the data are loopers and 53% a formation rate of around 8–12 anticyclones per are anticyclones. The mean westward velocity of year with possibly more during September to all the data is 4.970.5, of looper data 4.770.5 cm/ November and fewer during Februaryto May. s, of anticyclones 5.972.0 cm/s, and of cyclones Probablythe most significant findings are that in 3.472.1 cm/s. Typical (median) values of swirl the eastern Caribbean there are two bands of speeds are 20 cm/s, diameters are 100 km, and anticyclones that match two bands of swift periods of rotation are 18 days ðRo ¼ 0:16Þ; westward flows or jets in the Caribbean Current; although a few cyclonic loopers were significantly that the northern one originates (primarily) in the smaller and rotated faster than this. wind-driven flow of the North Atlantic; and the southern one originates in the MOC-forced flow coming from the South Atlantic. The anticyclones 8. Summary appear to form in the far eastern Caribbean from remnants of NBC rings that collided with the Two hundred and twelve drifter trajectories Lesser Antilles; the anticyclones were perhaps were used to characterize the mean currents in the amplified byinstabilities of the anticyclonic shear Caribbean and its variability. Different geographi- of the Caribbean Current jets. cal patterns of trajectories and drifter speeds Cyclones tended to be located near the bound- suggested some characteristics of the flow field aries in the eastern Caribbean, and it is suggested that were shown more quantitativelyin maps of that theyformed on the cyclonic shear side of the the mean velocityand EKE. The Caribbean Caribbean Current or in the cyclonic shear zone Current was traced as a high-speed current created when the swirl velocityof anticyclones (425 cm/s) westward through the southern part impinged on Hispaniola. Some small cyclones of the Caribbean except where it merged with the were formed as water squirted through the counterclockwise flow of the Panama–Colombia passages of the Lesser Antilles. Gyre in the mid-Colombia Basin. A second band Eastward flows greater that 30 cm/s were of westward velocity(25–30 cm/s) observed in the observed in the eastern Caribbean. The southern northern Caribbean just south of Hispaniola Panama–Colombia Gyre with its swift mean merged with the main southern Caribbean Current eastward flow stands out among these. Two other near 751W. The northern band was traced east- clusters of eastward velocitycoincide with the ward to the Antilles where water enters from the northern eastward swirl velocityof anticyclones North Equatorial Current and from some NBC located in the two bands. Although there could be rings. The southern band was traced to the countercurrents not associated with anticyclones, southeastern Caribbean where South Atlantic the two clusters suggest that a major source of Water enters including some from NBC rings. eastward currents are anticyclones. Drifter trajectories with two or more loops were The western Caribbean area southwest of Cuba identified as eddies and subdivided into cyclonic was dominated byloopers. Thirty-eightpercent of and anticyclonic loopers. Overall, 19% of the data the data were in 19 loopers, including 10 cyclonic were in a total of 57 loopers, 29 anticyclonic and and nine anticyclonic. Eight loopers were located 28 cyclonic. The loopers were considered to have in a cyclone–anticyclone pair that translated been located in 49 different eddies, 24 anticyclones slowly 2 cm/s westward toward the Yucatan and 25 cyclones. The eastern Caribbean is Current, where both eddies disappeared and dominated by energetic anticyclones with typical probablycoalesced with it. The anticyclone swirl speeds of 60 cm/s, diameters of 200 km, and possiblyformed from anticyclonic vorticityad- periods of rotation of 16 days. The anticyclones vected over the Jamaica Ridge from eastern translated westward at around 13 cm/s up to the Caribbean anticyclones, enhanced by the anti- Jamaica Ridge where the drifters stopped looping cyclonic shear of the Caribbean Current in the implying that the eddies were disrupted by western Caribbean. The cyclone is thought to have ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 461 formed when the northern part of the anticyclone 29765 and OCE 01-36477. Chris Wooding pro- impinged on the Isle of Pines creating strong cessed the drifter data and created the figures. This cyclonic shear and a series of small intense research is an extension of work undertaken in cyclones, which were advected eastward and collaboration with other members of the NBC coalesced into a larger cyclone. Nine drifters Ring Experiment, AmyFfield, Dave Fratantoni, began looping cyclonically in small rapid loops Silvia Garzoli, Gustavo Goni, Bill Johns, and (down to around a 1-dayperiod) near topography Doug Wilson. Helpful comments on an earlier suggesting this could be a frequent source of version of this paper were provided byClaudia cyclones in the Caribbean. Cenedese, Jim McWilliams, and two anonymous The westward translation velocityof anti- reviewers. It is a pleasure to contribute to a cyclones varied from around 13 cm/s in the eastern collection of papers in honor of Walter Zenk and Caribbean to around 2 cm/s just south of Cuba. to remember with fondness our numerous colla- The velocitywas similar to the velocityof back- borations over the years. ground drifters, which suggests that the anticyclones were advected by the background flow in which theywere embedded. This implies that References there is not a typical speed of eddies translating through the Caribbean. Instead, the speed of an Andrade, C.A., Barton, E.D., 2000. Eddydevelopment and eddycan varysignificantlydepending on its route, motion in the Caribbean Sea. Journal of Geophysical faster in the main Caribbean Current jet, slower in Research 105 (C11), 26191–26201. Astor, Y., Muller-Karger, F., Scranton, M.I., 2003. Seasonal the regions outside of fast currents, especiallyslow and interannual variation in the hydrography of the Cariaco just southwest of Cuba. Basin: implications for basin ventilation. Continental Shelf Research 23, 125–144. Barnier, B., Reynaud, T., Beckman, A., Bo¨ ning, C., Molines, J.- 9. Conclusions M., Barnard, S., Jia, Y., 2001. On the seasonal variability and eddies in the North Brazil Current: insights from model intercomparison experiments. Progress in Oceanography48, The new drifter data provide an improved 195–230. picture of the distribution of Caribbean eddies Candela, J., Beardsley, R.C., Limeburner, R., 1992. Separation and their characteristics. Coupled with the detailed of tidal and subtidal currents in ship mounted acoustic mean current field, the observations of eddies Doppler current profiler (ADCP) observations. Journal of Geophysical Research 97 (C1), 769–788. suggest where and how theymight have been Carton, J.A., Chao, Y., 1999. Caribbean Sea eddies inferred formed. However, Caribbean eddies remain poorly from TOPEX/POSEIDON altimetryand a 1/6 1 Atlantic known because of so few in situ observations, Ocean model simulation. Journal of Geophysical Research especiallysubsurface ones. Further studyof the 104 (C4), 7743–7752. Caribbean anticyclones and cyclones is needed to Cenedese, C., 2002. Laboratoryexperiments on mesoscale vortices colliding with a seamount. Journal of Geophysical determine their physical structure and water mass Research 107 (C6). properties and to evaluate their importance to the Cenedese, C., Adduce, C., 2002. Influence of multiple islands general circulation. The anticyclones’ water prop- and their 3-D geometryon the bifurcation of eddies. Eos, erties need to be compared to those of NBC rings Transactions, American Geophysical Union 83 (47), Fall in order to help clarifythe origin of the anti- Meeting Supplement, Abstract OS52D-0250, p. F683. Centurioni, L.R., Niiler, P.P., 2003. On the surface currents of cyclones. the Caribbean Sea. Geophysical Research Letters 30 (6), 1279. Chelton, D.B., Schlax, M.G., Freilich, M.H., Milliff, R.F., Acknowledgements 2004. Satellite measurements reveal persistent small-scale features in ocean winds. Science 303, 978–983. Corredor, J.E., Morell, J.M., Lopez, J.M., Capella, J.E., Contribution 11102 from the Woods Hole Armstrong, R.A., 2004. Cyclonic eddy entrains Orinoco Oceanographic Institution. Funding was provided River plume in eastern Caribbean. Eos, Transactions, byNational Science Foundation grants OCE 97- American Geophysical Union 85 (20) 197, 201–202. ARTICLE IN PRESS

462 P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463

Didden, N., Schott, F., 1993. Eddies in the North Brazil Heburn, G.W., Kinder, T.H., Allender, J.H., Hurlburt, H.E., Current retroflection region observed byGeosat altimetry. 1982. A numerical model of eddygeneration in the Journal of Geophysical Research 98 (C11), 20121–20131. southeastern Caribbean Sea. In: Nihoul, J.C. (Ed.), The Fratantoni, D.M., 2001. North Atlantic surface circulation Hydrodynamics of Semi-Enclosed Seas. Elsevier, Amster- during the 1990s observed with satellite-tracked drifters. dam, pp. 299–328. Journal of Geophysical Research 106 (C10), 22067–22093. Hansen, D.V., Paul, C.A., 1984. Genesis and effects of long Fratantoni, D.M., Glickson, D.A., 2002. North Brazil waves in the equatorial Pacific. Journal of Geophysical Current ring generation and evolution observed with Research 89 (C6), 10431–10440. SeaWiFS. Journal of Physical Oceanography 32, Hernandez-Guerra, A., Joyce, T.M., 2000. Water masses and 1058–1074 /1058:NBCRGAS2.0.CO;2. circulation in the surface layers of the Caribbean at 661W. Fratantoni, D.M., Richardson, P.L., 2004. The evolution and Geophysical Research Letters 27 (21), 3497–3500. demise of North Brazil Current rings. Journal of Physical Johns, W.E., Lee, T.N., Schott, F.A., Zantopp, R.J., Evans, Oceanographysubmitted for publication. R.H., 1990. The North Brazil Current retroflection: Fratantoni, D.M., Johns, W.E., Townsend, T.L., 1995. seasonal structure and eddyvariability.Journal of Geophy- Rings of the North Brazil Current: their structure and sical Research 95 (C12), 22103–22120. behavior inferred from observations and a numerical Johns, E., Wilson, W.D., Molinari, R.L., 1999. Direct simulation. Journal of Geophysical Research 100 (C6), observations of velocityand transport in the passages 10633–10654. between the Intra-American Sea and the Atlantic Ocean, Fratantoni, P.S., 1998. The formation and evolution of 1984–1996. Journal of Geophysical Research 104 (C11), Tortugas eddies in the southern Straits of Florida and Gulf 25805–25820. of Mexico. Ph.D. Thesis, Universityof Miami, unpublished. Johns, W.E., Townsend, T.L., Fratantoni, D.M., Wilson, Fratantoni, P.S., Lee, T.N., Podesta, G.P., Muller-Karger, F., W.D., 2002. On the Atlantic inflow to the Caribbean Sea. 1998. The influence of Loop Current perturbations on the Deep-Sea Research I 49, 211–243. formation and evolution of Tortugas eddies in the southern Johns, W.E., Zantopp, R.J., Goni, G.J., 2003. Cross-gyre Straits of Florida. Journal of Geophysical Research 103 transport byNorth Brazil Current rings. In: Goni, G.J., (C11), 24759–24779. Malanotte-Rizzoli, P. (Eds.), Interhemispheric Water Ex- Fratantoni, D.M., Johns, W.E., Townsend, T.L., Hurlburt, change in the Atlantic Ocean. Elsevier Oceanographic Series H.E., 2000. Low-latitude circulation and mass transport 68, Elsevier, Amsterdam, pp. 411–441. pathways in a model of the tropical Atlantic Ocean. Kennelly, M.A., Evans, R.E., Joyce, T.M., 1985. Small-scale Journal of Physical Oceanography 30, 1944–1966 cyclones on the periphery of a Gulf Stream warm-core ring. /1044:LLCAMTS2.0.CO;2. Journal of Geophysical Research 90 (C5), 8845–8857. Fuglister, F.C., 1951. Annual variations in current speeds in the Kinder, T.H., 1983. Shallow currents in the Caribbean Sea and Gulf Stream System. Journal of Marine Research 10 (1), Gulf of Mexico as observed with satellite-tracked drifters. 119–127. Bulletin of Marine Science 33 (2), 239–246. Garraffo, Z.D., Johns, W.E., Chassignet, E.P., Goni, G.J., Kinder, T.H., Heburn, G.W., Green, A.W., 1985. Some aspects 2003. North Brazil Current rings and transport of southern of the Caribbean circulation. Marine Geology68, 25–52. waters in a high resolution numerical simulation of the Lee, T.N., 1975. Florida Current spin-off eddies. Deep-Sea North Atlantic. In: Goni, G.J., Malanotte-Rizzoli, P. (Eds.), Research 22, 753–765. Interhemispheric Water Exchange in the Atlantic Ocean. Lee, T.N., Atkinson, L.P., 1983. Low-frequencycurrent and Elsevier Oceanographic Series 68, Elsevier, Amsterdam, temperature variabilityfrom Gulf Stream frontal eddies and pp. 375–409. atmospheric forcing along the southeast US outer con- Garzoli, S.L., Ffield, A., Yao, Q., 2003. North Brazil Current tinental shelf. Journal of Geophysical Research 88 (C8), rings and the variabilityin the latitude of retroflection. In: 4541–4567. Goni, G.J., Malanotte-Rizzoli, P. (Eds.), Interhemispheric Lee, T.N., Yoder, J.A., Atkinson, L.P., 1991. Gulf Stream Water Exchange in the Atlantic Ocean. Elsevier Oceano- frontal eddyinfluence on productivityof the southeast US graphic Series 68, Elsevier, Amsterdam, pp. 357–375. continental shelf. Journal of Geophysical Research 96 Goni, G.J., Johns, W.E., 2001. A census of North Brazil (C12), 22191–22205. Current rings observed from TOPEX/POSEIDON altime- Mayer, D.A., Weisberg, R.H., 1993. A description of COADS try: 1992–1998. Geophysical Research Letters 28 (1), 1–4. surface meteorological fields and the implied Sverdrup Goni, G.J., Johns, W.E., 2003. Synoptic study of warm rings in transports for the Atlantic Ocean from 301Sto601N. the North Brazil Current retroflection region using satellite Journal of Physical Oceanography 23, 2201–2221 altimetry. In: Goni, G.J., Malanotte-Rizzoli, P. (Eds.), /2201:ADOCSMS2.0.CO;2. Interhemispheric Water Exchange in the Atlantic Ocean. Molinari, R.L., Atwood, D.K., Duckett, C., Spillane, M., Elsevier Oceanographic Series 68, Elsevier, Amsterdam, Brooks, I., 1980. Surface currents in the Caribbean pp. 335–356. Sea as deduced from satellite tracked drifting buoys. Gordon, A.L., 1967. Circulation of the Caribbean Sea. Journal Proceedings of the Gulf and Caribbean Fisheries Institute of Geophysical Research 72 (24), 6207–6223. 31, 106–115. ARTICLE IN PRESS

P.L. Richardson / Deep-Sea Research II 52 (2005) 429–463 463

Molinari, R.L., Spillane, M., Brooks, I., Atwood, D., Duckett, Roemmich, D., 1981. Circulation of the Caribbean Sea: a well- C., 1981. Surface currents in the Caribbean Sea as deduced resolved inverse problem. Journal of Geophysical Research from Lagrangian observations. Journal of Geophysical 86 (C9), 7993–8005. Research 86 (C7), 6537–6542. Schmitz Jr., W.J., McCartney, M.S., 1993. On the North Mooers, C.N.K., Maul, G.A., 1998. Intra-Americas Sea Atlantic circulation. Reviews of Geophysics 31, 29–49. circulation. In: Brink, K.R., Robinson, A.R. (Eds.), The Schmitz Jr., W.J., Richardson, P.L., 1991. On the sources of the Sea Volume II. Wiley, New York, pp. 183–208. Florida Current. Deep-Sea Research 38 (Suppl.), Morrison, J.M., Nowlin, W.D., 1982. General distribution of S379–S409. water masses within the eastern Caribbean Sea during the Schott, F., Molinari, R.L., 1996. The western boundary winter of 1972 and fall of 1973. Journal of Geophysical circulation of the subtropical Warmwatersphere. In: Krauss, Research 87 (C6), 4207–4229. W. (Ed.), The Warmwatersphere of the North Atlantic Morrison, J.M., Smith, O.P., 1990. Geostrophic transport Ocean. Gebru¨ der Borntraeger, Berlin, pp. 229–253. variabilityalong the Aves Ridge in the eastern Caribbean Simmons, H.L., Nof, D., 2002. The squeezing of eddies through Sea during 1985–1986. Journal of Geophysical Research 95 gaps. Journal of Physical Oceanography 32, 314–335 /0314: (C1), 699–710. TSOETGS2.0.CO;2. Murphy, S.J., Hurlburt, H.E., O’Brien, J.J., 1999. The Smith, O.P., Morrison, J.M., 1989. Shipboard acoustic Doppler connectivityof eddyvariabilityin the Caribbean Sea, the current profiling in the eastern Caribbean Sea, 1985–1986. Gulf of Mexico, and the Atlantic Ocean. Journal of Journal of Geophysical Research 94 (C7), 9713–9719. Geophysical Research 104 (C1), 1431–1453. Sybrandy, A.L., Niiler, P.P., 1991. WOCE/TOGA Lagrangian Nystuen, J.A., Andrade, C.A., 1993. Tracking mesoscale ocean Drifter Construction Manual, SIO Ref. 91/6, WOCE Rep. features in the Caribbean Sea using Geosat altimetry. 63. Scripps Institution of Oceanography, La Jolla, CA Journal of Geophysical Research 98 (C5), 8389–8394. (58pp). Oey, L.-Y., Lee, H.-C., Schmitz Jr., W.J., 2003. Effects of winds Wang, G., Dewar, W.K., 2003. Meddy–seamount interactions: and Caribbean eddies on the frequencyof Loop Current implications for the Mediterranean salt tongue. Journal of eddy shedding: a numerical study. Journal of Geophysical Physical Oceanography 33, 2446–2461 /2446:MIIFTMS Research 108 (C10). 2.0.CO;2. Pauluhn, A., Chao, Y., 1999. Tracking eddies in the subtropical Wilson, W.D., Johns, W.E., 1997. Velocitystructure and North-Western Atlantic Ocean. Physics and Chemistry of transport in the windward Island passages. Deep-Sea the Earth A 24 (4), 415–421. Research I 44 (3), 487–520. Rennell, J., 1832. An investigation of the Currents of the Wilson, D., Leaman, K., 2000. The tropical origins of the Gulf Atlantic Ocean, and of Those Which Prevail Between the Stream. Current 16 (1), 14–17. Indian Ocean and the Atlantic. J.G.&F. Rivington, London Wu¨ st, G., 1964. Stratification and Circulation in the Antil- (359pp). lean–Caribbean Basins. Colombia UniversityPress, New Richardson, P.L., Hufford, G.E., Limeburner, R., Brown, York (201pp). W.S., 1994. North Brazil Current retroflection eddies. Wyrtki, K., Magaard, L., Hager, J., 1976. Eddy energy in the Journal of Geophysical Research 99 (C3), 5081–5093. oceans. Journal of Geophysical Research 81 (15), 2641–2646.