Routes and rates of larval fish transport from the southeast to the northeast United States continental shelf

1774

Limnol. Oceanogr., 47(6), 2002, 1774–1789q 2002, by the American Society of Limnology and Oceanography, Inc.

Routes and rates of larval fish transport from the southeast to the northeastUnited States continental shelf

Jonathan A. HareNOAA, NOS, NCCOS, Center for Coastal Fisheries and Habitat Research, 101 Pivers Island Road, Beaufort,North Carolina 28516

James H. ChurchillWoods Hole Oceanographic Institution, Woods Hole, Massachusetts 02543

Robert K. CowenRosenstiel School of Marine and Atmospheric Sciences, University of Miami, 4600 Rickenbacker Causeway,Miami, Florida 33149

Thomas J. BergerScience Applications International Corporation, 615 Oberlin Road, Raleigh, North Carolina 27605

Peter C. CornillonGraduate School of Oceanography, University of Rhode Island, South Ferry Road, Narragansett, Rhode Island 02882

Paul DragosBattelle Ocean Sciences, 397 Washington Street, Duxbury, Massachusetts 02332

Scott M. GlennInstitute of Marine and Coastal Sciences, P.O. Box 231, Rutgers University, New Brunswick, New Jersey 08901

John J. GovoniNOAA, NOS, NCCOS, Center for Coastal Fisheries and Habitat Research, 101 Pivers Island Road, Beaufort,North Carolina 28516

Thomas N. LeeRosenstiel School of Marine and Atmospheric Sciences, University of Miami, 4600 Rickenbacker Causeway,Miami, Florida 33149

Abstract

Larval fish originating south of Cape Hatteras, North Carolina, are frequently found on the continental shelf northof Cape Hatteras, even as far north as the Scotian shelf. The Gulf Stream and associated warm-core rings arehypothesized as the physical mechanisms responsible for the northward transport of larvae. Specifically, larvaespawned along the southeast U.S. continental shelf are entrained into the Gulf Stream, transported to the northeast,regularly incorporated in warm-core ring streamers, transported across the slope region, and released along the shelfedge north of Cape Hatteras. This proposed transport route was evaluated using hydrographic data and drifter tracksthat were collected as part of other studies. Continental shelf water that originated south of Cape Hatteras was found

AcknowledgmentsPatricia Tester, Cecily Natunewicz, and David Chapman provided critical reviews of earlier drafts of this manuscript. The South Atlantic

Bight Recruitment Experiment was funded by NOAA Center for Sponsored Coastal Ocean Research, Coastal Ocean Program. The CapeHatteras Field Study was funded by the U.S. Department of the Interior, Minerals Management Service contract 14-35-0001-30599 toScience Applications International Corporation. The Frontal Eddy Dynamics Experiment was supported by the U.S. Department of theInterior, Minerals Management Service contract 14-12-0001-30349. Partial support for this work was provided by NOAA/CIMAS throughthe South Florida Ecosystem Restoration Prediction and Modeling Program, Contract NA67RJ0149. The larval fish work on the northeastUnited States continental shelf was supported by the NOAA Office of Sea Grant, U.S. Department of Commerce, under grants NA86AA-D-SG045 and NA90AA-D-SG078. Analyses at Woods Hole Oceanographic Institution were supported by the U.S. Department of Energythrough contract DE-FG02-92ER61418. This work was performed while the senior author held a National Research Council ResearchAssociateship. This is Woods Hole Oceanographic Institution Contribution 10372.

1775Lagrangian views of larval transport

along the edge of the Gulf Stream north of Cape Hatteras, and drifter tracks demonstrated each segment of theproposed transport route. A probability model was then developed to estimate transport times for larvae carriedover the proposed transport route using a combination of sea surface temperature images, statistics of drifter transportspeeds, and a larval mortality function. Modeled transport time distributions closely matched observed age distri-butions of larvae, further supporting the hypothesized transport route.

Two major oceanographic regimes occupy the eastern sea-board of the United States. The southeast U.S. continentalshelf (SEUSCS) extends from the Straits of Florida to CapeHatteras and contains relatively warm, salty water (Atkinsonet al. 1983). The northeast U.S. continental shelf (NEUSCS)reaches from Cape Hatteras to Georges Bank and containsrelatively cold, fresher water (Beardsley et al. 1985; Aikmanet al. 1988; Churchill et al. 1993). (The terminology south-east and northeast U.S. continental shelves is used in theLarge Marine Ecosystem literature [Sherman 1993] and isused here to avoid problems associated with the unofficialand inaccurate geographical names South and Middle Atlan-tic Bight [Richards 1999]. Both sets of place names refer tonearly identical geographic areas.) On the inner portion ofthe SEUSCS, circulation is predominantly wind and buoy-ancy driven (Blanton and Pietrafesa 1978; Lee et al. 1989;Werner et al. 1999). On the outer portion of the SEUSCS,circulation is dominated by the Gulf Stream, which flowsnorthward along the shelf edge; flow direction at a givenlocation, however, is variable owing to wavelike meanders(Lee et al. 1989). Large-scale circulation on the NEUSCS isnot directly influenced by the Gulf Stream, which separatesfrom the shelf edge in the vicinity of Cape Hatteras (Csanadyand Hamilton 1988). The mean current over the NEUSCSand upper slope is directed southwestward, along isobaths,and is associated with an along-shore pressure gradient(Beardsley et al. 1985; Aikman et al. 1988). The NEUSCSand SEUSCS flows converge near Cape Hatteras, and mostshelf and slope water is entrained into Gulf Stream flowsand carried northeastward (Churchill and Berger 1998).

The water mass boundary near Cape Hatteras coincideswith a boundary separating two distinct marine faunas: awarm-temperate fauna to the south and a cold-temperate fau-na to the north (Briggs 1974). Members of these faunas,however, move between oceanographic regimes. Adults ofsome cold-temperate species are found south of Cape Hat-teras (e.g., tautog, Tautoga onitis, Parker and Dixon 1998),and larvae of NEUSCS-spawning species can be transportedsouth of Cape Hatteras via wind and buoyancy-driven intru-sions into the SEUSCS (Stegmann and Yoder 1996; Gro-thues et al. 2002). Likewise, many warm-temperate speciesare found in the NEUSCS during late spring, summer, andearly fall (Able and Fahay 1998). Some of these warm-tem-perate species migrate northward into the NEUSCS as adultsin spring and return south with the onset of fall (e.g., Atlanticmenhaden, Brevoortia tyrannus, Ahrenholz 1991). However,many warm-temperate species are transported north of CapeHatteras as larvae or juveniles (Markle et al. 1980; Able andFahay 1998). Larvae of some SEUSCS-spawned species usenursery habitats in the NEUSCS and migrate southward infall to join adult populations (e.g., bluefish, Pomatomus sal-tatrix, Kendall and Walford 1979), whereas others are ex-

patriates and perish as water temperatures cool (e.g., spotfinbutterflyfish, Chaetodon ocellatus, McBride and Able 1998).

The physical mechanisms that transport larvae from theSEUSCS to NEUSCS are important to the ecology of manywarm-temperate species. For species whose larvae and ju-veniles successfully utilize NEUSCS habitats, the mecha-nisms responsible for northward transport could control lar-val supply to nursery habitats (Hare and Cowen 1996). Forspecies whose larvae are expatriated to the NEUSCS, north-ward larval transport represents a loss of propagules fromsouthern populations. Under a scenario of climatic warming,expatriation may provide a mechanism for warm-temperatespecies to colonize areas north of Cape Hatteras, similar tothe recent spread of many tropical species into the northernSEUSCS (Parker and Dixon 1998). Larval transport fromthe SEUSCS to the NEUSCS might also be important in thenorthward dispersion of introduced species (Whitfield et al.2002).

Several potential physical mechanisms could result in thetransport of SEUSCS and Gulf Stream water, and theirplankton, to the NEUSCS. Pietrafesa et al. (1994) found thatstrong northward winds can force SEUSCS water northwardinto the NEUSCS, in opposition of the NEUSCS’s meancurrent. However, such wind-driven excursions have notbeen observed more than 100 km north of Cape Hatterasand cannot account for observations of SEUSCS-spawnedfish larvae in the central or northern NEUSCS (Shima 1989;Hare and Cowen 1991; Cowen et al. 1993), or still furtherto the northeast over Georges Bank (Cowen et al. 1993) andthe Scotian Shelf (Markle et al. 1980). Water dischargedfrom Gulf Stream meanders often makes its way to theNEUSCS edge (Churchill and Cornillon 1991a, 1991b;Churchill et al. 1993). Using hydrographic data acquiredquarterly over 11 yr, Churchill and Cornillon (1991b) deter-mined that the probability of such transport decreases sharp-ly going northwards from Cape Hatteras, and found no in-trusions of discharged Gulf Stream water at the NEUSCSedge north of 398N (off Delaware Bay).

Many investigators have postulated that warm-core rings(WCRs), which are mesoscale eddies that form when GulfStream meanders ‘‘pinch-off’’ (Joyce et al. 1984), transportsubtropical and warm-temperate plankton, including larvae,to the northeast North American shelf edge (Cox and Wiebe1979; Markle et al. 1980). Specifically, Hare and Cowen(1991, 1996) and Cowen et al. (1993) proposed four seg-ments for northward larval transport associated with WCRs(Fig. 1). Larvae that are spawned on the continental shelfsouth of Cape Hatteras are (1) entrained into the GulfStream, (2) transported to the northeast along the edge ofthe Gulf Stream, (3) carried in WCR streamers from the GulfStream and across the Slope Sea (i.e., the region betweenthe Gulf Stream and the shelf edge north of Cape Hatteras),and (4) ejected from WCR streamers at the shelf edge where

1776 Hare et al.

Fig. 1. Transport route proposed by Hare and Cowen (1991,1996) and Cowen et al. (1993) for the movement of larvae spawnedalong the southeast U.S. continental shelf (SEUSCS) to the north-east U.S. continental shelf (NEUSCS) edge. Larvae that arespawned on the northern SEUSCS are (1) entrained into the GulfStream, (2) transported to the northeast along the edge of the GulfStream, (3) carried in WCR streamers from the Gulf Stream acrossthe slope region between the Gulf Stream and the shelf edge northof Cape Hatteras, and (4) ejected from WCR streamers at the shelfedge, where they enter the NEUSCS/slope frontal region. Coastlineand 200-m isobath are plotted, and place names used in the text areindicated. Fig. 2. (A) Tracks of 13 satellite-tracked drifters released in the

northern portion of the SEUSCS as part of three separate field pro-grams (see Data sources). These drifter tracks were used to examinesegments 1 and 2 of the proposed route (Fig. 1, Table 1). (B) Tracksof 11 drifters released in the slope region of the NEUSCS (see Datasources). These drifter tracks were used to examine segments 3 and4 of the proposed route (Fig. 1, Table 1). Coastline and 200-misobath are plotted with drifter tracks.

they enter the shelf/slope frontal region. WCR streamers arecomposed of Gulf Stream–derived water that moves aroundthe western periphery of the WCRs and result from a WCR–Gulf Stream interaction (Evans et al. 1985; Nof 1988). Agesof larval fish derived from otolith increments suggest thattransport over the proposed route can occur in as little as 8d (Hare and Cowen 1991). Although founded in a generalunderstanding of the physical oceanography of the region,there is no direct physical evidence for the proposed trans-port route (see discussion by Epifanio and Garvine 2001);nor is there independent verification of the speed over whichwater and larvae can travel over the proposed route.

The purpose of the present study was to evaluate the trans-port route proposed by Hare and Cowen (1991, 1996) andCowen et al. (1993) through an examination of water prop-erties and Lagrangian drifter tracks. These observations re-veal transport along each of the four segments of the pro-posed transport route, as well as transport connectionsjoining the segments. To further assess the likelihood thatSEUSCS-spawned larvae observed on the NEUSCS werecarried along this route, a probability model was developedthat estimated the age distribution of larvae arriving at theNEUSCS edge at the endpoint of the route. Model inputincluded all the available measurements of drifter speedalong the various segments of the proposed route, distancesfor each segment extracted from satellite-derived sea surfacetemperature (SST) images, and estimates of larval mortality

rate. Model results were then compared with measured agedistributions of SEUSCS-spawned larvae captured at theNEUSCS edge to assess the validity of the proposed trans-port route.

Materials and methods

Data sources—Hydrographic data: Hydrographic mea-surements taken along a cross-slope transect east of Virginiaduring May 1984 as part of the Mid-Atlantic Slope and Rise(MASAR) project (Csanady and Hamilton 1988) provided aview of SEUSCS water at the edge of the Gulf Stream, northof Cape Hatteras. CTD (conductivity, temperature, depth)casts and XBT (expendable bathythermograph) drops weremade, and observed temperature–salinity (TS) propertieswere compared to the TS properties of the water massestypically observed in the SEUSCS and southern NEUSCSduring springtime (Atkinson et al. 1983; Churchill et al.1993). To gain a sense of the speed of entrained SEUSCS


Table 1. Summary of drifters used in this study including original project, transmitter number,release date, water depth at release, drogue type and depth, general release location (see Fig. 1),and transport route segments for which the drifter was included in analysis. Drogue type refers toholey sock (HS), surface (S), and window shade (WS). Drogue depth refers to depth of the centerof the drogue.

ProjectTransmit-

ter no. Release date

Waterdepth at

release (m)

Drogue

depth(m) type Release location

Seg-ment

FREDFREDFREDFREDSABRESABRESABRESABREMMS-CHFSMMS-CHFSMMS-CHFSMMS-CHFSMMS-CHFS

87728773877587769806

175432600456335373540377435903593

23 May 8722 May 8712 May 8712 May 87

1 Feb 941 Feb 94

17 Dec 9417 Dec 94

5 May 9226 Aug 9230 Oct 93

8 Nov 9212 May 93

1302703035303030302525252525

10101010

00

1992.252.252.252.252.25

HSHSHSHSSSWSWSHSHSHSHSHS

Onslow BayOnslow BayRaleigh BayRaleigh BayOnslow BayOnslow BayOnslow BayOnslow BayRaleigh BayRaleigh BayRaleigh BayRaleigh BayRaleigh Bay

1,21,21,21,2*1,21,21†1,21,21,21,21,21,2

SEFCAR‡DWDS-EPADWDS-EPADWDS-EPADWDS-EPADWDS-EPADWDS-EPADWDS-EPADWDS-EPADWDS-EPADWDS-EPADWDS-EPA

231122273412738127421274512750127611276212763127681276912771

27 Feb 963 May 914 May 906 Jul 90

21 Jul 9013 Sep 90

3 Dec 908 Dec 904 Nov 908 Nov 90

20 Jan 913 Feb 91

.2000

.2000

.2000

.2000

.2000

.2000

.2000

.2000

.2000

.2000

.2000

2.251010101010101010101010

HSHSHSHSHSHSHSHSHSHSHSHS

Florida Bay106 DWDS106 DWDS106 DWDS106 DWDS106 DWDS106 DWDS106 DWDS106 DWDS106 DWDS106 DWDS106 DWDS

3,43§3§3,43,43,43§3,43,43,43,4

* Crossed from the Gulf Stream to the northeast U.S. continental shelf edge in a discharge of water emanatingfrom a Gulf Stream meander.

† After release, drifter moved southwest and was stranded on Frying Pan Shoals. This drifter was omitted fromthe residence time on the southeast U.S. continental shelf (segment 2).

‡ Released off south Florida, subsequently became entrained into Gulf Stream flows, and moved north of CapeHatteras, where it was incorporated into a WCR during ring formation.

§ One of three EPA-DWDS drifters that circuited around the periphery of a WCR and was not detrained in thevicinity of the shelf/slope front north of Cape Hatteras.

water, the geostrophic velocity component normal to thetransect line was calculated using the CTD station at whichSEUSCS water was encountered and the adjacent XBT sta-tions (the tight TS relation observed in the transect data al-lowed us to infer salinity from the XBT data).

Drifter data: Tracks of 13 drifters released in the northernSEUSCS during three separate experiments (Fig. 2A, spe-cifics of drifters and their release are provided in Table 1)were used to examine movements along segments 1 and 2of the proposed transport route (Fig. 1). Four of the trackswere from the Frontal Eddy Dynamics (FRED) experiment,which examined Gulf Stream frontal eddies between CapeFear and Cape Hatteras (Glenn and Ebbesmeyer 1994a,1994b). Four tracks were from the South Atlantic Bight Re-cruitment Experiment (SABRE), which examined biologicaland physical processes involved in the recruitment dynamicsof estuarine-dependent fishes that spawn on the SEUSCS(Werner et al. 1999). The final five drifter tracks were ac-

quired as part of the Mineral Management Service–CapeHatteras Field Study (MMS-CHFS), which evaluated the po-tential effect of oil and gas exploration over the shelf regionbetween Chesapeake Bay and Cape Lookout (Churchill andBerger 1998).

Drifter tracks from the Environmental Protection Agen-cy’s (EPA) Monitoring, Research, and Surveillance Study ofthe 106-mile Deep Water Dumpsite (EPA-DWDS) were usedto evaluate segments 3 and 4 of the proposed transport route(Fig. 1). The EPA-DWDS evaluated the potential effects ofsewage sludge dumping at a site on the continental slopeeast of New Jersey and south of New York (Berger et al.1996; Dragos et al. 1996). During the study, 58 drifters weredeployed in and around the 106-mile site. Tracks of 11 drift-ers, which were entrained in the Gulf Stream and then intoa WCR west of 658W, were used in this study (Table 1, Fig.2B).

Location data from all drifters were quality checked andlinearly interpolated to produce tracks at evenly spaced in-

1778 Hare et al.

Fig. 3. Five-day composite satellite-derived SSTs from June,July, and August 1988. Station locations are plotted where P. sal-tatrix pelagic juveniles were captured in June 1988 and X. novaculalarvae were captured in July and August 1988. Symbol size repre-sents concentration of larvae—zero larvae are denoted with an opencircle. The proposed transport route is indicated by a black line,and distances were extracted from these lines for use in a probabilitymodel. Red indicates warmer water and blue indicates cooler water.Land is shown as white, and the 200-m isobath is plotted.

Fig. 4. (A) Distribution of speeds as drifters crossed out of theSEUSCS. (B) Distribution of residence times of drifters released onthe SEUSCS. All drifters released on the SEUSCS were entrainedinto Gulf Stream flows and moved rapidly to the northeast. Onedrifter that was stranded on Frying Pan Shoals for about 1 week(Table 1) was not included in the determination of residence time.

tervals (6 h). Velocities were then estimated from the tracksby dividing the distance traveled in an interval by time.

Satellite-derived SSTs: Drifter tracks were superimposedon concomitant satellite-derived images of SSTs to relatesurface temperature features with movement of drifters overdifferent segments of the proposed transport route. Satellite-derived SST images were obtained from the SoutheastCoastWatch Program at the NOAA Center for Coastal Fish-eries and Habitat Research, from the Remote Sensing Lab-oratory at the University of Rhode Island, and from a dataset of 5-d composite imagery developed for NOAA byBrown and Podesta (1989).

Probability model—Conceptual framework: A probabilitymodel was developed to give estimates of the age distribu-tion of larvae that traveled along the proposed route to arrive


at a specified location on the NEUSCS edge. Specific sce-narios were developed for the model to correspond to spe-cific times when SEUSCS-spawned larvae were collected onthe outer NEUSCS. Modeled age distributions were thencompared to observed age distributions, which were deter-mined through otolith aging of captured larvae. If actual lar-val ages appeared at very low probabilities in the modeledage distribution, then transport along the proposed routecould be deemed unlikely. Conversely, coincidence of actualages with modeled ages of high probability supports the as-sertion that larvae had traversed the proposed route.

Conceptually, the model consists of two components: (1)probability distributions of transport times over the four seg-ments of the proposed route and (2) a larval mortality func-tion. To determine component 1, a transit time probabilityfunction was defined for each route segment (Fig. 1). Forroute segment i, this is denoted as pi(t), where pi(t)dt is theprobability that larvae are transported the length of the seg-ment in the forward direction (leading to the NEUSCS) overa period between t and t 1 dt. Consider transport across thefirst two segments. If p1(t) and p2(t) are independent and theprobability of going from one segment to the next is 1, thenthe likelihood per unit time that a larva may transit the twosegments in a period t is simply the sum of all possiblecombinations of p1(t1) p2(t2)dt, with t1 and t2 summing to t.This can be expressed as a convolution integral,

t

p (t) 5 p (t 2 «)p («) d« (1)1–2 E 1 2

0

where p1–2(t)dt is the probability of being carried in the for-ward direction over segments 1 and 2 within a period of tto t 1 dt, and « is a dummy variable for integration. Withthe inclusion of additional segments, the above is expandedto nested convolution integrals. The transit time distributionfor the four segments of the specified route is

p (t)1–4

t c «

5 p (t 2 c)p (c 2 «)p (« 2 k)p (k) dk d« dcE E E 1 2 3 4

0 0 0

(2)

with k and c also dummy variables of integration.A survival probability S(t) was defined to account for the

effect of mortality on the final transit time distribution (com-ponent 2). This gives the likelihood of survival for larvaethat cover the route in a time t. The transit time probabilityfunction with mortality included then becomes Eq. 3.

P(t) 5 p1–4(t)S(t) (3)

Residence time on the SEUSCS, p1(t): This segment wasrepresented as a normal distribution of the expected resi-dence time of water on the SEUSCS. This distribution wasfully specified by the mean and standard deviation of resi-dence time of drifters deployed on the SEUSCS. Residencetime was defined as the period between a drifter’s releaseand its crossing of 758W. FRED, SABRE, and MMS-CHFSdrifters were used, but the SABRE drifter that grounded onFrying Pan Shoals was excluded from the calculation (Table

1). Larvae were assumed to be spawned in the northern por-tion of the SEUSCS, and leaving the shelf is the first seg-ment of transport to the NEUSCS.

Transport time along the Gulf Stream’s edge, p2(t): Forthis, and the subsequent route segments, transit time proba-bility was determined from (1) the segment length and (2)the distribution of speeds along the segment. Segment lengthwas determined for three specific scenarios (June, July, andAugust 1988), times when SEUSCS-spawned fish were col-lected on the outer shelf of the NEUSCS. The procedureinvolved tracing the proposed route over SST images thatcoincided with larval collections (Fig. 3) and then determin-ing the distances along the edge of the Gulf Stream (segment2), from the Gulf Stream to the NEUSCS edge around aWCR (segment 3), and along the NEUSCS edge from theWCR to larval collection locations (segment 4).

The speed distribution for segment 2 was determined fromthe tracks of drifters that moved from 75 to 718W (rangechosen to include maximum number of drifters). The along–Gulf Stream speed was taken as the ratio of the length ofthe line connecting the crossings of these longitudes for eachdrifter and the time difference between the crossings. Thismethod was necessary to give a representative speed becausemovement of drifters along the edge of the Gulf Stream oftenwas not unidirectional. Speeds were computed from FRED,SABRE, and MMS-CHFS drifters.

The mean and standard deviation of the drifter speedswere used to specify a normal transport speed distribution.This was defined as q2(v), where q2(v)dv is the likelihoodthat larvae traverse segment 2 with a mean speed betweenv and v 1 dv. Because our interest was confined to transportin the forward direction along the route, this distribution wasconverted to one defined for speeds in the forward directiononly. To ensure that the ‘‘forward speed’’ distribution wouldintegrate to unity, it was determined as

q (v)2q9(v) 5 (4)2 `

q (v) dvE 2

0

Conversion of q (v) to a transit time probability was done92by noting that the equation

q (v)dv 5 p2(t)dt92 (5)

is satisfied for v 5 D2/t, where D2 is the segment 2 length.Evaluating dv/dt gives Eq. 6.

p2(t) 5 D2q (v)/t292 (6)

Warm-core ring transport time to the shelf edge, p3(t):None of the 13 drifters released in the SEUSCS were in-corporated into WCR streamers, and only two of these drift-ers passed a WCR. Thus, these drifters could not be used toexamine segments 3 and 4 of the proposed transport route(Table 1).

Examination of drifter tracks with concomitant SST im-ages indicated that at least 23 of the EPA-DWDS driftersmoved from the edge of the Gulf Stream into the circulationof a WCR before reaching 608W. Analyses herein were con-

1780 Hare et al.

Fig. 5. (A) Location of hydrographic stations sampled during MASAR cruise 2 (Csanady and Hamilton 1988). (B) Section of temper-ature and (C) salinity along the MASAR transect extending from the southern portion of the outer NEUSCS to the edge of the Gulf Stream.(D) Volumetric TS diagram of water found over the SEUSCS during May (from Atkinson et al. 1983). The number in each square givesthe volume (km3) of SEUSCS water with TS characteristics falling within the box’s range. Also shown is the TS profile of water encounteredat station M24. The near-surface portion of the M24 trace overlaps with data from the SEUSCS. (E) TS regimes of various water massesfound in the southern portion of the NEUSCS during late May and early June 1984 (from Churchill et al. 1993). TS profile from stationM24 is overlain. The near-surface portion of the M24 trace does not overlap with data from the NEUSCS.

ducted on only the 11 drifters that entered WCR circulationwest of 658W (Fig. 2B). WCR streamer speed was calculatedas the average of a drifter’s speed while in a WCR streamer.The resulting ensemble of drifter speeds was used to specifyp3(t) in the manner as described above for p2(t).

Transport time along the NEUSCS edge, p4(t): Transportspeed along the NEUSCS edge were calculated from EPA-DWDS drifters that exited WCR streamer circulation andmoved to the west-southwest parallel to the NEUSCS edge(n 5 8). For each drifter, along-shelf edge speed was cal-culated as its average speed while it traveled parallel to theshelf edge. The speed ensemble was used to specify p4(t) inthe manner as described above for p2(t).

Mortality: Following Peterson and Wroblewski (1984),larval fish survival probability was modeled as an exponen-tial function. The decline of a population with an initial N0

individuals can be expressed as N(t) 5 N0e2zt, where N(t) isthe number of individuals surviving to time t and z is theinstantaneous larval mortality rate. The survival probabilityfunction was expressed relative to the overall survival like-lihood (for all transit times) as

S(t) 5 e2zt/[# p1–N(t)e2ztdt]`0 (7)

Scenario-specific mortalities were calculated from theslope of the natural logarithm–transformed age distribution(Ricker 1975). For Xyrichtys novacula larvae (the July andAugust 1988 scenarios), mortality rate was estimated using


Fig. 6. (A) Tracks of two FRED drifters released in the northern SEUSCS. Track time is coincident with the underlying satellite SSTimage. Circles indicate locations at 24-h intervals. (B) Speed time series derived from the two drifter tracks shown in panel A. Both driftersmoved east-northeastward in association with the Gulf Stream front at speeds between 100 and 200 cm s21. (C, E) Tracks of two MMS-CHFS drifters released in Raleigh Bay. Both drifters exit the SEUSCS in association with the Gulf Stream; however, both are detrainedfrom the main Gulf Stream current downstream of Cape Hatteras in association with meander crests. Both were then re-entrained. Tracktime is coincident with the underlying satellite SST images. Circles indicate locations at 24-h intervals. (D, F) Speed time series derivedfrom the two drifter tracks shown in panels C and E, respectively. Time at which drifter leaves SEUSCS is indicated by vertical line labeledSEUSCS. Periods that correspond to positions and images shown in panels C and E are indicated by black bars.

data presented in Hare and Cowen (1991). Estimated instan-taneous mortality was 0.12 6 0.01 d21 (mean 6 SE, r2 50.89). For P. saltatrix, the age distributions of P. saltatrixlarvae collected in the SEUSCS during April 1989 andMarch 1990 were used (Hare and Cowen 1996, unpubl.data). Estimated P. saltatrix mortality was 0.11 6 0.02 d21

(r2 5 0.67).

Analysis of model results: The probability model esti-mated larval age distributions for three scenarios in whichSEUSCS-spawned larvae were collected over the outerNEUSCS. Shima (1989) reported the ages of SEUSCS-spawned P. saltatrix pelagic juveniles that were collected inthe vicinity of the Hudson Canyon during June 1988. Hareand Cowen (1991) reported the ages of expatriated X. no-vacula larvae that were captured in the Hudson Canyon re-gion during July and August 1988 (only larvae collected atthe outer three stations of the cross-shelf sampling grid wereincluded in the comparison with model results).

Sensitivity analyses were conducted to evaluate the effect

of various model parameters on predicted larval age distri-butions. Larval mortality rate (z) was varied for each of thethree time periods modeled: June, July, and August 1988.Values of z used in the sensitivity analyses were 0.05, 0.2,0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, and 1.0 d21, which covers therange of larval mortality rates commonly reported (Petersonand Wroblewski 1984; Morse 1989). Distance of the route,mean speed, and standard deviation of speed were also var-ied for each of the three time periods modeled. Values werevaried by 610, 25, 50, 75, and 100%. Values were variedas an ensemble; for example, all distances were increasedby 10% or all speeds were decreased by 25%. The numberof observed ages that were in the lower and upper 5% tailsof each modeled distribution was then calculated.

Results

Entrainment of water from the SEUSCS into the GulfStream—Upon release on the SEUSCS, most drifters movedto the northeast (n 5 10); however, some initially moved to

1782 Hare et al.

Fig. 7. Distributions of Gulf Stream–associated speeds for drift-ers released on the SEUSCS.

→

Fig. 8. (A–C) Track of an EPA-DWDS drifter released off the shelf of the NEUSCS. The drifter initially moved to the southwest beforebecoming incorporated into the flows associated with the Gulf Stream in the vicinity of Cape Hatteras (not shown). The drifter traveledeast-northeastward with the Gulf Stream but was then detrained in association with a Gulf Stream meander. Within 4 d, the drifter wasentrained into the cross-slope circulation of a WCR streamer. After crossing the Slope Sea, the drifter exited the WCR streamer and movedwestward parallel to the NEUSCS edge. In panels A, B, and C, red indicates warmer water and blue indicates cooler water. The black lineshows the entire track. Black circles show daily positions concomitant with the 5-d composite image. (D) Speed time series derived fromthe drifter track shown in panels A–C. The period corresponding to panels A–C are indicated by black bars. (E) Distribution of cross-slopespeed associated with the WCR streamer for the ensemble of 11 EPA-DWDS drifters. (F) The track of a drifter released off of SouthFlorida in January 1998. Once north of Cape Hatteras, the drifter moved northeastward with the Gulf Stream. The drifter was then

the southwest before turning to the northeast (n 5 3) (Fig.2A). All drifters eventually exited the SEUSCS into the vi-cinity of the western Gulf Stream front and exit speeds weretypical of reported Gulf Stream and Gulf Stream frontal ve-locities (;100 cm s21; Fig. 4A; Glenn and Ebbesmeyer1994a, 1994b). No particular SST feature could be linked tothe process of Gulf Stream entrainment owing to cloud coverduring entrainment events. Defining 758W as the eastern ex-tent of the SEUSCS, the mean drifter residence time on theouter SEUSCS was 8.5 d (Fig. 4B).

SEUSCS water at the Gulf Stream’s edge north of CapeHatteras—The temperature distribution along the MASARtransect (Fig. 5A) showed the Gulf Stream front approxi-mately 130 km east of the shelf edge (Fig. 5B). A surfacelayer of relatively fresh water extending to a depth of about10 m was situated along the western edge of the Gulf Streamfront (station M24; Fig. 5C). TS characteristics identified thislayer as entrained SEUSCS water and were within the rangeof TS properties of water found on the SEUSCS during May(Fig. 5D) but were far outside the range of TS properties ofNEUSCS and slope water during May–June (Fig. 5E).

Estimates of geostrophic velocity normal to the transectindicated that this entrained SEUSCS water was traveling at

a speed of roughly 200 cm s21. Between stations M23 andM24, estimated geostrophic velocity at 5 m was 195 cm s21

(750 m reference depth); between stations M24 and M25,estimated geostrophic velocity at 5 m was 211 cm s21 (330m reference depth).

Drifter tracks supported the geostrophic estimate ofSEUSCS water speed along the edge of the Gulf Stream butindicated that the path of this water can be complex. Mostdrifters entrained from the SEUSCS into the Gulf Streamremained in the Gulf Stream frontal region after passingCape Hatteras (Fig. 6A). The speed of all drifters exceeded100 cm s21 at some point downstream of 758W (Fig. 6B),and several traveled in excess of 200 cm s21. However, fewof the drifters maintained rapid downstream transport forextended periods of time; many drifters either slowed in theGulf Stream or left and then reentered the Gulf Stream cur-rent (Fig. 6C–F). Detrainment of drifters from the main GulfStream current occurred to both sides of the Gulf Stream(e.g., into the slope region between the Gulf Stream and theNEUSCS edge, and into the Sargasso Sea) and typically oc-curred near the crest of a meander. Similar detrainment andentrainment of deeper RAFOS drifters has been describedby Bower and Rossby (1989). As a result of the detrainmentfrom and subsequent re-entrainment into the main GulfStream current, the average speed between 75 and 718W was101.1 cm s21 (Fig. 7).

Cross-slope movement of water from the Gulf Stream edgeto the NEUSCS edge—Evidence of WCR-associated trans-port of water from the edge of the Gulf Stream to the shelfedge of the NEUSCS is frequently seen in satellite-derivedSST images (Joyce et al. 1984; Evans et al. 1985). Imagesfrom June to August 1988 provide relevant examples ofWCR streamers (Fig. 3), covering the period during whichP. saltatrix and X. novacula larvae were captured near Hud-son Canyon (Shima 1989; Hare and Cowen 1991, 1996).These SST images show a series of rings moving south-westward through the NEUSCS slope region, each with aband of entrained Gulf Stream water.

Drifter tracks from the EPA-DWDS study show evidenceof cross-slope transport in WCR streamers. Upon releaseover the NEUSCS upper slope, most of the 58 driftersmoved to the southwest. All were eventually entrained intothe Gulf Stream. Once in the Gulf Stream, these driftersmoved northeastward at the edge of the Gulf Stream, similarto the drifters released in the SEUSCS (Fig. 2B). Eleven ofthe EPA-DWDS drifters were incorporated into WCRstreamers west of 658W, some directly from the Gulf Stream


incorporated into the interior of a WCR upon formation, where it remained for 30 d. Note the difference in the fate of this drifter relativeto those that are transported across the slope in association with WCR streamers (Fig. 8A–C). Temporal series of track times are coincidentwith the underlying satellite SST images. Circles indicate locations at 24-h intervals. Red indicates warmer water and blue indicates coolerwater. (G) Track of a FRED drifter released in the northern SEUSCS in May 1987. The drifter moved across the slope region to theNEUSCS edge in association with a Gulf Stream discharge. The black line shows the entire track. Black circles show daily positionsconcomitant with the 5-d composite image. (H) Speed time series of FRED drifter shown in panel G. Period corresponding to panel G isindicated by a black bar.

1784 Hare et al.

Fig. 9. Distribution of along-NEUSCS speeds for EPA-DWDSdrifters that crossed the slope region in association with WCRstreamers and were ejected from WCR circulation in the vicinity ofthe NEUSCS edge.

Table 2. Summary of specific transport distances used for eachtransport segment for parameterization of the probability model.Transport distances were determined by tracing the proposed trans-port route (Fig. 1) on satellite-derived SST images and extractingdistances (Fig. 3). Mortality values used in the model are also pro-vided. Mortality rates were determined from abundance-at-age data(Ricker 1975).

Transport modelsegment

SegmentNo.

Transport distance (km)

Jun 88 Jul 88 Aug 88

Gulf StreamCross-slopeAlong-NEUSCS

234

668280172

553313

0

455285

0

Mortality rate (z) 0.11 0.12 0.12

front and others after detrainment from the Gulf Stream inassociation with a Gulf Stream meander (Fig. 8A,B). Fol-lowing entrainment into a WCR streamer, transport acrossthe Slope Sea was direct, with an average speed of 50.9 cms21 (Fig. 8D,E).

Cross-slope transport of water in WCR streamers was dis-tinct from the circulation of the ring itself. An example wasprovided by a drifter released off of the Florida Keys thattraveled along the northern edge of the Gulf Stream duringApril 1998 (Fig. 8F). This drifter moved into a meander crestas a WCR formed and remained in the anticyclonic circu-lation of the ring for at least 30 d. Such transport would traplarval fish in the ring until either the ring decayed or larvaedeveloped the ability to actively move out of the ring.

One drifter was ejected from the Gulf Stream near thecrest of a meander and moved across the Slope Sea to theNEUSCS edge within a filament of discharged Gulf Streamwater (Fig. 8G). While in the filament, the drifter’s cross-slope speed was slower (;25 cm s21) than typical cross-slope speeds within WCR streamers (;50 cm s21) (Fig. 8D,

H). This observation indicates that although the probabilityof cross-slope transport in a Gulf Stream discharge decreasesfrom Cape Hatteras northward (Churchill and Cornillon1991b), such transport can occur north of Delaware Bay.

Expulsion of water from a WCR at the NEUSCS edge—Drifter tracks indicate that the probability of water in a WCRstreamer being ejected from the ring into the shelf edge re-gion is high. Eight of the 11 EPA-DWDS drifters left theWCR circulation near the NEUSCS edge and moved west-southwestward (Fig. 8C). The remaining three drifters con-tinued around the WCR and were eventually re-entrainedinto the Gulf Stream current. Once along the NEUSCS edge,all eight drifters initially moved to the west-southwest withan average speed of 15.4 cm s21 (Fig. 9).

Probability model—The probability model used distancesdetermined from SST imagery (Fig. 3, Table 2), mortalityrates calculated from abundance at age (Table 2), and speedsdetermined from drifter statistics (Table 3). One potentialconcern with the model was the small number of observa-tions used to determine the speed components. However, thevalues reported here agree with observations from a numberof studies that used a variety of techniques (Table 4). Thus,we have confidence in the values used here.

Distributions of observed larval ages closely matched lar-val age distributions derived from the probability model(Fig. 10). Means of observed larval ages were within 4 d ofmean modeled ages, and standard deviations of observed andmodeled ages were within 3.5 d (Fig. 10). Patterns in ob-served larval age distributions between sampling times werealso replicated in the modeled transit time distributions.Maximum modeled and observed ages were of P. saltatrixcaught in June 1988 (Fig. 10). The older modeled ages ofthis scenario resulted from the length of the along-NEUSCScomponent (Fig. 3, Table 2). Minimum modeled and mea-sured ages were for X. novacula caught in August 1988 (Fig.10). The young modeled ages resulted from shorter GulfStream and cross-shelf distances and from the absence of analong-NEUSCS segment (Fig. 3, Table 2).

Larval mortality was an important component of the mod-el (Fig. 11). At the lowest mortality (z 5 0.05), the meansand standard deviations of the modeled ages exceeded themeans and standard deviations of observed larval ages (Fig.11, top row). As larval mortality increased, the mean andstandard deviation of model ages decreased. Model resultswere generally consistent with observations over a range oflarval mortality from 0.1 to 0.3 (Fig. 11), which agrees withthe mortality rates estimated in this study (Table 2) and withmortality rates reported in the literature (Peterson and Wrob-lewski 1984; Morse 1989).

Sensitivity analyses also demonstrated that the best fit be-tween modeled and observed larval ages occurred at modeldistance and speed values within 25% of calculated values(Tables 2, 3). Standard deviations in speed had little effecton modeled distributions (Fig. 11, second row). For transportdistances and mean speeds, changes of 650, 75, and 100%resulted in modeled larval ages that were different than ob-served larval ages (Fig. 11, third and fourth rows); a largenumber of observed ages fell within the lower and upper 5%


Table 3. Summary of general transport speed values used for each transport segment for param-eterization of the probability model. Transport speed values were derived from drifter tracks ana-lyzed in combination with SST images. Segment numbers refer to Fig. 1. SEUSCS residence timeis the residence time of a drifter on the SEUSCS. Along-NEUSCS speed is the speed along theNEUSCS shelf/slope frontal region.

Transport model segmentSegmentNumber Mean SD Sample size Source

SEUSCS residence timeGulf Stream speedCross-slope speedAlong-NEUSCS speed

1234

8.5 d101.1 cm s21

50.9 cm s21

15.4 cm s21

6.449.913.5

4.9

121311

8

Fig. 4Fig. 7Fig. 8Fig. 9

Table 4. Speed estimates for the various components of the proposed transport route as reported in other studies. The agreement betweenvelocities derived in this study (Table 3) and those of other studies provide confidence in the speed estimates used in the probability model.Additionally, Nof (1988) indicated that propogation rate around a WCR in a streamer equals one-half of the particle speed along theupstream front of the Gulf Stream. Based on the Gulf Stream speeds shown in this table, cross-slope velocities in WCR streamers wouldbe 20–75 cm s21, which agrees with the speed derived from drifter tracks (Fig. 8E). ADCP is Acoustic Doppler Current Profiler.

Transport modelsegment Segment No.

Estimated speed(cm s21)

Method (depth ofmeasurement, m) Source

Gulf Stream

Cross-slope

Along-NEUSCS

222233444

40–150100–130

40–13030–15051.1 (SD 5 26.7)40–8016.4 (SD 5 7.8)10.2 (SD 5 8.1)12.4 (SD 5 10.6)

PEGASUS (0)ADCP (50)ADCP (60)ADCP (100)SST (0)ADCP (50)DriftersCurrent meter (10)Current meter (10)

Halkin and Rossby 1985Hummon and Rossby 1998Pierce and Joyce 1988Joyce et al. 1990Zheng et al. 1984Hummon and Rossby 1998Lozier and Gawarkiewicz 2001Aikman et al. 1988Beardsley et al. 1985

tails of the modeled distributions. Greater transport speedsand lesser transport distances had less of an effect on themodeled age distribution of X. novacula larvae in July andAugust 1988, likely because larvae cannot be transported tothe NEUSCS edge much faster.

One potential concern with the model is that the transportdistances are for one point in time; yet, as larvae move overthe route, transport distances change. If we accept the dis-tances from June 1988 (July 1988) to represent the maxi-mum change that could have occurred in transport distancesfor larvae that were collected in July 1988 (August 1988),the sensitivity analysis indicates that modeled ages wouldnot be greatly affected by changing route length becausedistance varied by less than 25% between months (Table 2;Fig. 11, bottom row).

Discussion

Drifter tracks and hydrographic data provide direct phys-ical evidence of the proposed route by which larvae spawnedin the northern SEUSCS are entrained into the Gulf Stream,transported to the northeast, regularly incorporated in WCR-streamers, transported across the slope region and releasedalong the NEUSCS edge. All drifters released in the northernSEUSCS were entrained into Gulf Stream flows and werecarried rapidly to the northeast along the edge of the GulfStream. Hydrographic data confirmed the presence and speedof SEUSCS water along the edge of the Gulf Stream north

of Cape Hatteras. Similar observations of SEUSCS wateralong the western Gulf Stream front north of Cape Hatterashave been reported recently (Grothues and Cowen 1999;Hare et al. 2001). Drifters were incorporated in WCRstreamers from the edge of the Gulf Stream and then releasedfrom ring circulation in the vicinity of the NEUSCS edge.

All three observed larval age distributions fell withinmodeled larval age distributions, which were based on drift-er-derived transport speed distributions and SST image–de-rived transport distances. Sensitivity analyses found thatsmall changes in model parameters changed the modeleddistributions to more closely match observed distributions.The coincidence of actual and modeled ages supports theassertion that the larvae collected on the NEUSCS edge tra-versed the proposed route, and that the statistics of transportspeed over the route’s segments, derived from multiple yearsof drifter track data, are relatively stationary (not changingwidely over time).

The probability of SEUSCS-spawned larvae moving overthe proposed route is largely dependent on WCR–GulfStream interactions. All drifters released in the northernSEUSCS were entrained into Gulf Stream flows. Approxi-mately 75% of drifters moving in WCR streamers were re-leased from streamers along the NEUSCS edge and traveledsouthwestward along the shelf edge. Entrainment into aWCR streamer is the most variable step. Although none ofthe drifters released in the northern SEUSCS were incor-porated into WCR streamers, only two passed WCRs. One

1786 Hare et al.

Fig. 10. The results of a probability model (line) compared toobserved larval ages (histogram). Modeled and observed mean andstandard deviation are provided. In all cases, the model results arevery similar to observed ages, supporting the hypothesis that thelarvae were transported from the SEUSCS to the NEUSCS edgevia the transport route proposed by Hare and Cowen (1991, 1996)and Cowen et al. (1993) (Fig. 1).

interpretation is that the proposed transport route does notoccur. An alternative interpretation is that incorporation intoWCR streamers occurs at some probability ,1, and is fun-damentally dependent on the presence of a WCR. This latterinterpretation is supported by the tracks of the EPA-DWDSdrifters. These drifters moved northeastward in the GulfStream similar to the drifters released in the northernSEUSCS. During the period of the EPA-DWDS study,WCRs were prevalent (Berger et al. 1996) and 11 of 58drifters were incorporated into WCR streamers west of658W, indicating an approximate 20% probability of incor-poration into a WCR. These data indicate that entrainmentfrom the Gulf Stream into WCRs depends, first, on the pres-ence of a WCR and, second, on the location of a water parcel(or drifter) relative to the spatial structure of the WCR–GulfStream interaction (Hummon and Rossby 1998).

Because entrainment into a WCR streamer is the mostvariable component of the proposed transport route, the sup-ply of larvae from the SEUSCS to the NEUSCS edge islargely controlled by entrainment into a WCR streamer. Thisconclusion is consistent with the strong correlation betweenWCR streamer activity and recruitment of SEUSCS-spawned P. saltatrix to NEUSCS estuarine nursery habitats(Hare and Cowen 1996), and the correspondence betweenthe location of WCRs and the occurrence of SEUSCS-spawned larvae along the NEUSCS edge north of DelawareBay (Hare and Cowen 1991; Cowen et al. 1993). Larvalsupply of a wide variety of warm-temperate species to ju-venile habitats north of Cape Hatteras might be controlledby the cross-slope transport in WCR streamers (Atlanticmenhaden, Brevoortia tyrannus; American eel, Angullia ros-trata; conger eel, Conger oceanicus; crevalle jack, Caranxhippos; C. ocellatus; white mullet, Mugil cephalus; stripedmullet, Mugil curema; butterfish, Peprilus triacanthus). Be-cause WCRs are a universal element of western boundarycurrent systems (Olson 2001), the described transport routecould be an important source of larval supply to more pole-ward continental shelves throughout the world’s oceans.

Discharges of water from Gulf Stream meanders also ap-pear to be an important mechanism of larval movement fromthe Gulf Stream into the slope region of the NEUSCS (Hareet al. 2001), but the likelihood of transport to the NEUSCSedge via this mechanism decreases northward from CapeHatteras (Churchill and Cornillon 1991b). Discharges ofGulf Stream water might explain the recruitment ofSEUSCS-spawned bluefish to NEUSCS estuaries in years oflow WCR streamer activity (Hare and Cowen 1996). Onedrifter in this study provided an example of cross-slopetransport in discharged Gulf Stream water. If slower cross-slope transport were indicative of all discharges of GulfStream water, then fish arriving at the NEUSCS edge viaGulf Stream water discharges would be older (slower speed)and fewer in number (more cumulative mortality) than thosecarried across the slope in WCR streamers.

The short resident times of drifters released in the northernSEUSCS indicate that a majority of larvae spawned in thisregion might be lost to the Gulf Stream. Drift bottle studies(Barans and Roumillat 1976) and modeling studies (Check-ley et al. 1999; Hare et al. 1999) indicate that the residencetime of surface water in the northern SEUSCS is short and


Fig. 11. The sensitivity of model results to variations in larval mortality (z) (first row), standard deviation of transport speed (secondrow), mean transport speed (third row), and transport distance (fourth row). Columns are the results for the three different time periodsmodeled: June 1988, July 1988, and August 1988. The observed mean is indicated by a horizontal dotted line for comparison to the modelresults. Values used in the model (Fig. 10) are indicated by vertical dotted lines. The model mean (6SD) from the sensitivity analyses areshown by black circles. The number of observed ages falling within the 5% upper and lower tails of the modeled distributions are shownas the histograms.

decreases with distance from shore. Blanton and Pietrafesa(1978) estimated a complete flushing time of Onslow Bayof 60 d, with bottom waters moving in and surface watersleaving. Govoni and Spach (1999) and Grothues and Cowen(1999) estimated large fluxes of larvae off the SEUSCS. Fu-ture work should attempt to quantify loss relative to recruit-ment of species that spawn on the SEUSCS and depend onSEUSCS nursery habitats.

Drifter tracks also support the idea that the transport oflarvae onto the NEUSCS at the surface is rare. Gawarkiew-icz et al. (1996) noted high-salinity slope water above thepycnocline on the outer shelf (between 80- and 100-m iso-

baths) and identified the potential forcing mechanisms aswind and WCRs. No drifters examined here and only a fewof the total EPA-DWDS drifter set crossed the 100-m isobath(Berger et al. 1996; Dragos et al. 1996). Yet, many fish lar-vae cross from the slope onto the NEUSCS, which indicatesthat the physical mechanisms of cross-frontal exchange pre-dominantly operate below 10 to 15 m. Cowen et al. (1993)proposed that on-shelf movement of some species occurredvia an ontogenetic deepening of larval vertical distributionscoupled with slopewater intrusions along the summertimeNEUSCS thermocline (Flagg et al. 1994). For species thatcross onto the NEUSCS at the surface (e.g., bluefish, P. sal-

1788 Hare et al.

tatrix; white hake, Urophycis tenuis), Hare and Cowen(1996) and Hare et al. (2001) proposed active horizontalswimming as the mechanism. Drifter tracks indirectly sup-port both mechanisms of cross-shelf transport by demon-strating that passive, surface slopewater to NEUSCS trans-port is rare, but does occur (see Gawarkiewicz et al. 1996).

Larval mortality played an important role in the predictedtransport times, influencing both the mean and standard de-viation of predicted ages. Generally, larval mortality valuesfrom 0.1 to 0.3 resulted in predicted transport times (andstandard deviations) that most closely matched observedages, and these values are similar to observed larval mor-tality rates (Morse 1989). Inclusion of larval mortality isclearly important in estimating connectivity between localpopulations (sensu Cowen et al. 2000).

In a recent review, Epifanio and Garvine (2001) discussedthe importance of local wind and buoyancy forcing in thetransport of larval fish and invertebrates along the easterncoast of the United States. This present study and others(Hare and Cowen 1991, 1996; Cowen et al. 1993; Lee et al.1994; Stegmann and Yoder 1996; Govoni and Spach 1999)indicate that processes related to the region’s western bound-ary current are also involved in larval transport, particularlyon the outer shelf and where the shelf is narrow. A difficultyfacing the advancement of larval transport studies in thisregion, however, is a coupled understanding of the role oflocally and remotely forced processes in affecting larvaltransport.

The approach developed here allows hypotheses regardinglarval fish transport to be tested. If observed larval agesappeared at very low probabilities in the modeled age dis-tribution, then transport along the proposed route could bedeemed unlikely. Although there are valid criticisms of usingLagrangian measurements to follow larval fish (e.g., slippageof drifter owing to wind, drifter tracking fixed depth versuslarval fish moving vertically), the large-scale view of larvaltransport, both subjective and objective, provides valuableinsights into larval transport processes. A number of studieshave used Lagrangian measurements, both actual and mod-eled, to examine the mechanisms and outcomes of larvaltransport (e.g., Pepin and Helbig 1997; Hare et al. 1999;Cowen et al. 2000). Similarly, probabilistic approaches haveprovided insights into the population-level consequences oflarval transport (Levin 1983; Schultz and Cowen 1994). Asthese techniques continue to develop and are coupled withfield observations of larval fish distribution and abundance,the ability to test hypotheses regarding larval transportmechanisms will increase, thereby furthering our under-standing of the importance of larval transport in the ecologyof marine populations.

References

ABLE, K. W., AND M. P. FAHAY. 1998. The first year in the life ofestuarine fishes in the Middle Atlantic Bight. Rutgers Univ.Press.

AHRENHOLZ, D. W. 1991. Population biology and life history ofNorth American menhadens, Brevoortia spp. Mar. Fish. Rev.53: 3–19.

AIKMAN, F., H. W. OU, AND R. W. HOUGHTON. 1988. Current var-

iability across the New England continental shelf-break andslope. Cont. Shelf Res. 8: 625–651.

ATKINSON, L. P., T. N. LEE, J. O. BLANTON, AND W. S. CHANDLER.1983. Climatology of the southeastern U.S. continental shelfwaters. J. Geophys. Res. 88: 4705–4718.

BARANS, C. A., AND W. A. ROUMILLAT. 1976. Surface water driftsouth of Cape Lookout, North Carolina. South Carolina De-partment of Natural Resources Technical Report 12.

BEARDSLEY, R. C., D. C. CHAPMAN, K. H. BRINK, S. R. RAMP, AND

R. SCHLITZ. 1985. The Nantucket Shoals Flux Experiment(NSFE79). Part I: A basic description of the current and tem-pearture variability. J. Phys. Oceanogr. 15: 713–748.

BERGER, T. J., T. F. DONATO, P. DRAGOS, AND D. REDFORD. 1996.Features of the surface circulation in the Slope Sea as derivedfrom SST imagery and drifter trajectories. J. Mar. Environ.Eng. 2: 43–65.

BLANTON, J. O., AND L. J. PIETRAFESA. 1978. Flushing of the con-tinental shelf south of Cape Hatteras by the Gulf Stream. Geo-phys. Res. Lett. 5: 495–498.

BOWER, A. S., AND T. ROSSBY. 1989. Evidence of cross-frontal ex-change processes in the Gulf Stream based on isopycnal RA-FOS float data. J. Phys Oceanogr. 19: 1177–1190.

BRIGGS, J. C. 1974. Marine zoogeography. McGraw Hill.BROWN, O. B., AND G. P. PODESTA. 1989. Satellite sea surface tem-

perature (SST) climatology development for fisheries. NOAASoutheast Fisheries Science Center, Final Report Contract NA85-WC-77-06134.

CHECKLEY, D. M., P. B. ORTNER, F. E. WERNER, L. R. SETTLE, AND

S. R. CUMMINGS. 1999. Spawning habitat of the Atlantic men-haden in Onslow Bay, North Carolina. Fish. Oceanogr. (suppl.2) 8: 22–36.

CHURCHILL, J. H., AND T. J. BERGER. 1998. Transport of MiddleAtlantic Bight shelf water to the Gulf Stream near Cape Hat-teras. J. Geophys. Res. 103: 30,605–30,621.

, AND P. C. CORNILLON. 1991a. Water discharged from theGulf Stream north of Cape Hatteras. J. Geophys. Res. 96:22,227–22,243.

, AND . 1991b. Gulf Stream water on the shelf andupper slope north of Cape Hatteras. Cont. Shelf Res. 11: 409–431.

, E. R. LEVINE, D. N. CONNORS, AND P. C. CORNILLON. 1993.Mixing of shelf, slope and Gulf Stream water over the conti-nental slope of the Middle Atlantic Bight. Deep-Sea Res. 40:1063–1085.

COWEN, R. K., J. A. HARE, AND M. P. FAHAY. 1993. Beyond hy-drography: Can physical processes explain larval fish assem-blages within the Middle Atlantic Bight? Bull. Mar. Sci. 53:567–587.

, K. M. LWIZA, S. SPONAUGLE, C. B. PARIS, AND D. B.OLSON. 2000. Connectivity of marine populations: Open orclosed? Science 287: 857–859.

COX, J., AND P. H. WIEBE. 1979. Origins of oceanic plankton in theMiddle Atlantic Bight. Estuar. Coast. Mar. Sci. 9: 509–527.

CSANADY, G. T., AND P. HAMILTON. 1988. Circulation of slopewater.Cont. Shelf Res. 8: 565–624.

DRAGOS, P., F. AIKMAN, AND D. REDFORD. 1996. Lagrangian statis-tics and kinematics from drifter observations pertaining to dis-persion of sludge from the 106-mile site. J. Mar. Environ. Eng.2: 21–41.

EPIFANIO, C. E., AND R. W. GARVINE. 2001. Larval transport on theAtlantic continental shelf of North America: A review. Estuar.Coast. Shelf Sci. 52: 51–77.

EVANS, R. H., K. S. BAKER, O. B. BROWN, AND R. C. SMITH. 1985.Chronology of warm-core ring 82B. J. Geophys. Res. 90:8803–8811.

FLAGG, C. N., R. W. HOUGHTON, AND L. J. PIETRAFESA. 1994. Sum-


mertime thermocline salinity maximum intrusions in the Mid-Atlantic Bight. Deep-Sea Res. 41: 325–340.

GAWARKIEWICZ, G. G., C. A. LINDER, J. F. LYNCH, A. E. NEWHALL,AND J. J. BISAGNI. 1996. A surface-trapped intrusion of slopewater onto the continental shelf in the Mid-Atlantic Bight.Geophys. Res. Lett. 23: 3763–3766.

GLENN, S. M., AND C. C. EBBESMEYER. 1994a. The structure andpropogation of a Gulf Stream frontal eddy along the NorthCarolina shelf break. J. Geophys Res. 99: 5029–5046.

, AND . 1994b. Observations of Gulf Stream frontaleddies in the vicinity of Cape Hatteras. J. Geophys Res. 99:5047–5055.

GOVONI, J. J., AND H. L. SPACH. 1999. Exchange and flux of larvalfishes across the western Gulf Stream front south of Cape Hat-teras, USA in winter. Fish. Oceanogr. (suppl. 2) 8: 77–92.

GROTHUES, T. M., AND R. K. COWEN. 1999. Larval fish assemblagesand water mass history in a major faunal transition zone. Cont.Shelf Res. 19: 1171–1198.

, , L. J. PIETRAFESA, F. BIGNAMI, G. L. WEATHERLY,AND C. N. FLAGG. 2002. Flux of larval fish around Cape Hat-teras. Limnol. Oceanogr. 47: 165–175.

HALKIN, D., AND T. ROSSBY. 1985. The structure and transport ofthe Gulf Stream at 738W. J. Phys. Oceanogr. 16: 1814–1828.

HARE, J. A., AND R. K. COWEN. 1991. Expatriation of Xyrichtysnovacula (Pisces: Labridae) larvae: Evidence of rapid cross-slope exchange. J. Mar. Res. 49: 801–823.

, AND . 1996. Transport mechanisms of larval andpelagic juvenile bluefish (Pomatomus saltatrix) from South At-lantic Bight spawning grounds to Middle Atlantic Bight nurs-ery habitats. Limnol. Oceanogr. 41: 1264–1286.

, AND OTHERS. 1999. Larval transport during winter in theSABRE study area: Results of a coupled vertical larval behav-ior–three-dimensional circulation model. Fish. Oceanogr.(suppl. 2) 8: 57–76.

, M. P. FAHAY, AND R. K. COWEN. 2001. Springtime ich-thyoplankton of the slope region off the northeastern U.S.: Lar-val assemblages, relation to hydrography and implications forlarval transport. Fish. Oceanogr. 10: 164–192.

HUMMON, J., AND T. ROSSBY. 1998. Spatial and temporal evolutionof a Gulf Stream crest–warm core ring interaction. J Geophys.Res. 103: 2795–2809.

JOYCE, T. R., AND OTHERS. 1984. Rapid evolution of a warm-corering. Nature 308: 837–840.

, K. A. KELLEY, D. M. SCHUBERT, AND M. J. CARUSO. 1990.Shipboard and altimetric studies of rapid Gulf Stream vari-ability between Cape Cod and Bermuda. Deep-Sea Res. 37:897–910.

KENDALL, A. W., AND L. A. WALFORD. 1979. Sources and distri-butions of bluefish, Pomatomus saltatrix, larvae and juvenilesoff the east coast of the U.S. Fish. Bull. 77: 213–227.

LEE, T. N., E. WILLIAMS, J. WANG, R. EVANS, AND L. ATKINSON.1989. Response of South Carolina continental shelf waters towind and Gulf Stream forcing during winter of 1986. J. Geo-phys. Res. 94: 10,715–10,754.

, M. E. CLARKE, E. WILLIAMS, A. F. SZMANT, AND T. J.BERGER. 1994. Evolution of the Tortugas Gyre and its influenceon recruitment in the Florida Keys. Bull. Mar. Sci. 54: 621–646.

LEVIN, L. A. 1983. Drift tube studies of bay–ocean water exchangeand implications for larval dispersal. Estuaries 6: 364–371.

LOZIER, M. S., AND G. GAWARKIEWICZ. 2001. Cross-frontal ex-change in the Middle Atlantic Bight as evidenced by surfacedrifters. J. Phys. Oceanogr. 31: 2498–2510.

MARKLE, D. F., W. B. SCOTT, AND A. C. KOHLER. 1980. New andrare records of Canadian fishes and the influence of hydrog-

raphy on resident and nonresident Scotain Shelf ichthyofauna.Can. J. Fish. Aquat. Sci. 37: 49–65.

MCBRIDE, R. S., AND K. W. ABLE. 1998. Ecology and fate of but-terflyfishes, Chaetodon spp., in the temperate, western NorthAtlantic. Bull. Mar. Sci. 63: 401–416.

MORSE, W. 1989. Catchability, growth, and mortality of larval fish-es. Fish. Bull. 87: 417–446.

NOF, D. 1988. The propagation of ‘streamers’ along the peripheryof warm-core rings. Deep-Sea Res. 35: 1483–1498.

OLSON, D. B. 2001. Biophysical dynamics of western transitionzones: A preliminary synthesis. Fish. Oceanogr. 10: 133–150.

PARKER, R. O., AND R. L. DIXON. 1998. Changes in a North Car-olina reef fish community after 15 years of intense fishing—global warming implications. Trans. Am. Fish. Soc. 127: 908–920.

PEPIN, P., AND J. A. HELBIG. 1997. Distribution and drift of Atlanticcod (Gadus morhua) eggs and larvae on the northeast New-foundland shelf. Can. J. Fish. Aquat. Sci. 54: 670–685.

PETERSON, I., AND J. S. WROBLEWSKI. 1984. Mortality rate of fishesin the pelagic ecosystem. Can. J. Fish. Aquat. Sci. 41: 1117–1120.

PIERCE, S. D., AND T. M. JOYCE. 1988. Gulf Stream velocity struc-ture through inversion of hydrographic and acoustic dopplerdata. J. Geophys. Res. 93: 2227–2236.

PIETRAFESA, L. J., J. M. MORRISON, M. P. MCCANN, J. H. CHUR-CHILL, E. BOHM, AND R. W. HOUGHTON. 1994. Water masslinkages between Middle and South Atlantic Bights. Deep-SeaRes. 41: 365–390.

RICHARDS, W. J. 1999. Problems with unofficial and inaccurate geo-graphical names in the fisheries literature. Mar. Fish. Rev. 61:56–57.

RICKER, W. E. 1975. Computation and interpretation of biologicalstatistics of fish populations. Bull. Fish. Res. Board Can. 191:1–382.

SCHULTZ, E. T., AND R. K. COWEN. 1994. Recruitment of coral-reeffishes to Bermuda: Local retention or long distance transport?Mar. Ecol. Prog. Ser. 109: 15–28.

SHERMAN, K. 1993. Large marine ecosystems as global units formarine resources management—an ecological perspective, p.4–14. In: K. Sherman, L. M. Alexander, and B. D. Gold [eds.],Large marine ecosystems: Stress, mitagation, and sustainabil-ity. AAAS Press.

SHIMA, M. 1989. Oceanic transport of the early life history stagesof bluefish (Pomatomus saltatrix) from Cape Hatteras to theMid-Atlantic Bight. M.S. thesis. State Univ. of New York atStony Brook.

STEGMANN, P. M., AND J. A. YODER. 1996. Variability of sea-surfacetemperature in the South Atlantic Bight as observed from sat-ellite: Implications for offshore spawning fishes. Cont. ShelfRes. 16: 843–861.

WERNER, F. E., B. O. BLANTON, J. A. QUINLAN, AND R. A. LUET-TICH. 1999. Physical oceanography of the North Carolina con-tinental shelf during the fall and winter seasons: Implicationsfor the transport of larval menhaden. Fish. Oceanogr.(suppl 2)8: 7–21.

WHITFIELD, P., T. GARDNER, S. P. VIVES, M. R. GILLIGAN, W. R.COURTENAY, G. C. RAY, AND J. A. HARE. 2002. Biologicalinvasion of the Indo-Pacific lionfish (Pterois volitans) alongthe Atlantic coast of North America. Mar. Ecol. Prog. Ser. 235:289–297.

ZHENG, Q., V. KLEMAS, AND N. E. HUANG. 1984. Dynamics of theslope water off of New England and its influence on the GulfStream as inferred from satellite IR data. Remote Sens. Envi-ron. 15: 135–153.

Received: 15 June 2001Accepted: 1 May 2002Amended: 4 June 2002

Routes and rates of larval fish transport from the southeast to the northeast United States continental shelf

Documents