Investigation of tropical eel spawning area in the South-Western Indian Ocean: Influence of the oceanic circulation

  • Published on
    11-Sep-2016

  • View
    212

  • Download
    0

Transcript

  • are d

    Centre de Recherches et dEnseignement sur les SystmecUniversit Pierre et Marie Curie-Paris 06, UMR 7208 (C43 rue Cuvier, CP26, 75231 Paris cedex 05, France

    a r t i c l e i n f o

    Article history:Received 6 November 2009Received in revised form 4 June 2010Accepted 9 June 2010Available online 6 July 2010

    coasts, undergoing during this phase a severe mortality. Lepto-cephali are thought to metamorphose into glass eels at their arrivalon the continental shelf. Later they colonize coastal and inlandwaters. After the settlement they become sedentary yellow eel.Their subsequent growth period lasts 38 years in males and 815 years in females of the temperate European eel (Feunteun,

    areas. Their sexual maturation occurs during this migration. Littleis known about this second oceanic migration except that it occursin the rst 1000 m in open ocean, which makes surveys difcultand expensive (Tesch, 1979). However, new tools as pop-up aredeveloped to follow individually eel adults, and this range ofmigration depths seems to be conrmed (Jellyman and Tsukamoto,2005; Aarestrup et al., 2009). Nevertheless, most of the keys tounderstand eel life history are related to reproduction and recruit-ment. Consequently, a particular attention has to be given to themarine phase. Indeed, larval dispersal by oceanic currents

    * Corresponding author.E-mail addresses: pous@mnhn.fr (S. Pous), feunteun@mnhn.fr (E. Feunteun),

    Progress in Oceanography 86 (2010) 396413

    Contents lists availab

    c

    elscellien@mnhn.fr (C. Ellien).spawning area. They all allow migration to each recruitment sites consistent with duration estimatedfrom otolith microstructure analyses. Nevertheless, there is substantial variability on intra-seasonal tointerannual timescale in simulated migration durations and arrival success, with specic amplitude toeach recruitment site and spawning location.

    2010 Elsevier Ltd. All rights reserved.

    1. Introduction

    Eels (Anguilla genus) are highly migratory diadromous species,i.e. a part of their life cycle takes place in fresh water and the otherpart in sea water. Spawning takes place at depths over 400 m inwarm oceans (Tsukamoto, 1992). After hatching, leptocephali aredriven by the oceanic currents towards the continental (or islands)

    2002), but this kind of information is not known for tropical eel.Moreover, duration of the growth period varies strongly, amongspecies and within species, according to environmental parameterssuch as food availability, water quality and temperature (Jellyman,1997; Tesch, 2003). After this growth period, a second metamor-phosis occurs: yellow eels change into silver eels and migratethousands of kilometres to go back to their oceanic spawning0079-6611/$ - see front matter 2010 Elsevier Ltd. Adoi:10.1016/j.pocean.2010.06.002s Ctiers, 38 rue du Port Blanc, 35800 Dinard cedex, FranceNRS-UPMC-MNHN-IRD), Dpartement Milieux et Peuplements Aquatiques, Musum national dHistoire naturelle,

    a b s t r a c t

    In the South-Western Indian Ocean (SWIO), four eel species of the genus Anguilla (i.e. Anguilla bicolorbicolor, Anguilla nebulosa labiata, Anguilla marmorata andAnguilla mossambica) were identied, while theirrespective oceanic spawning area remained unknown. Based on collected larvae, glass eel captures andhydrodynamical conditions, previous studies raised the hypothesis that the eel spawning area mightbe common to all of those freshwater eel species, and located East of Madagascar. An original modelingapproach, based on backward simulations, is developed to assess how the ocean circulation in the SWIOdetermines the location of the spawning areas and whether a common spawning area for each recruit-ment site where glass eels were found is possible. We use a hydrodynamical model, which reproducesrealistically the 3D open ocean circulation in the region, associated with a Lagrangian model that calcu-lates the possible migration pathways of larvae, represented by passive particles. Some biological param-eters, provided by previous otolith microstructures analysis, are taken into account to constrain oursimulations. Results suggest the existence of a common spawning area located between 13S and 19Sand westwards of 60.5E, although these boundaries vary on the interannual timescale. Salinity frontswere reported beside the boundaries, reinforcing this assumption. We explore the impact of hydrody-namic conditions on recruitment and migration durations from three specic regions within the commonExprimentation et Approches Numrique, (LOCEAN), CC100, 4 Place Jussieu, 75252 Paris cedex 05, FrancebMusum national dHistoire naturelle, UMR 7208 (CNRS-UPMC-MNHN-IRD), Dpartement Milieux et Peuplements Aquatiques,Investigation of tropical eel spawning areInuence of the oceanic circulation

    S. Pous a,*, E. Feunteun b, C. Ellien c

    aMusum national dHistoire naturelle, UMR 7159 (CNRS-IRD-UPMC-MNHN), Laboratoi

    Progress in O

    journal homepage: www.ll rights reserved.in the South-Western Indian Ocean:

    Ocanographie et de Climatologie:

    le at ScienceDirect

    eanography

    evier .com/locate /pocean

  • eanodetermines at least partly, the location of recruitment and theabundance of recruits. The long larval duration of leptocephalimakes the precise location of their spawning areas important sincespecic current or temperature conditions could affect their sur-vival and then, their recruitment success (Tsukamoto, 1992; Cas-tonguay et al., 1994; Dsaunay and Gurault, 1997). Then theleptocephalus stage needs to be better understood. Indeed, theduration of the oceanic larval migration is subject of controversy,particularly for Atlantic eels (Bonhommeau et al., 2008; McCleave,2008).

    The analysis of geographical trends of leptocephalus size distri-bution have been almost the only way to discover the eel spawn-ing locations (Schmidt, 1923; Jespersen, 1942; Tsukamoto, 1992;Kimura et al., 2006). In the Atlantic and Pacic oceans, the distri-bution of the youngest eel larvae suggests that spawning takesplace in specic areas that may facilitate mating and place larvaein the appropriate landwards-owing currents (Kimura et al.,1999, 2001; Tsukamoto et al., 2002; Tsukamoto, 2006; Friedlandet al., 2007). It has been suggested that maturing European (Angu-illa anguilla), American (Anguilla rostrata) and Japanese (Anguillajaponica) silver eels make long migrations to spawn in thesubtropical gyres of the North Atlantic or North Pacic oceans(McCleave et al., 1987; Tsukamoto, 1992). In the case of the twoAtlantic eel species, the surface expression of the 22.5 C isotherm,located near the frontal zone in the Sargasso Sea, forms the north-ern boundary of the spawning area (McCleave, 1993). Tempera-ture fronts in the Sargasso Sea may act as cues that help adulteels to locate the spawning area (Friedland et al., 2007). Jetsassociated with these fronts appear to transport a variety of lepto-cephali species eastwards (Miller and McCleave, 1994). Therefore,changes in the latitude or intensity of these fronts (and the asso-ciated jets) may affect both the spawning location and the subse-quent transport of the leptocephali to continental habitats, andmay partly explain the decline in eel recruitment recorded sincethe early 1980s (Dekker, 2003; Friedland et al., 2007). In the Paci-c Ocean, the young Japanese eel leptocephali are mostly foundwithin the margins of the North Equatorial Current, owing wes-terly towards the Japanese coasts, and at a depth ranging between50 and 150 m. Their spawning location corresponds to a salinityfront (Tsukamoto, 1992; Kimura et al., 2001). The reduced salinityin the frontal zone, or some associated features such as odour,may provide migrating eel adults with a cue which triggers cessa-tion of migration and initiation of spawning. As previously statedfor the Atlantic eels and the thermal front, changes in the locationof the salinity front may contribute to the decline in Japanese eelrecruitment observed in the recent decades (Kimura et al., 2001;Kimura and Tsukamoto, 2006; Friedland et al., 2007). The key sim-ilarity in Atlantic and Pacic oceans is the position of the spawn-ing areas connected to major currents, the Gulf Stream for theAtlantic Ocean and the NEC with its northern bifurcation (i.e.Kuroshio Current) for the Pacic Ocean, which can drive the off-springs to their growth habitats and thus complete the recruit-ment of eels (Tsukamoto, 1992).

    In the South-Western Indian Ocean (SWIO), four species of An-guilla have been identied: Anguilla bicolor bicolor,Anguilla nebulosalabiata, Anguilla marmorata and Anguilla mossambica (Ege, 1939;Robinet et al., 2007). The anthropogenic impact on those tropicaleel stocks is less documented than for the Atlantic and Pacic eelspecies (Moriarty and Dekker, 1997; Dekker, 2003). Nevertheless,recent observations explain the lack of large eels in most islandsof the Indian Ocean as a consequence of traditional sheries(Robinet et al., 2007). Indeed, those eel stocks are facing a growinginterest from the international markets in Madagascar and South

    S. Pous et al. / Progress in OcAfrica (Robinet et al., 2008). Of all shes that breed in the westernIndian Ocean, the genus Anguilla has probably the highest eco-nomic value per unit weight of sh, and for decades the world-wide demand exceeds the supply (Jackson, 1976).

    The eel spawning area in the SWIO has not been discovered yet,despite oceanographic cruises aiming at sampling eel larvae. How-ever, Jespersen (1942), following his observations on small lepto-cephali distribution and hydrographic conditions in this region,suggested that spawning area should be shared by all those fresh-water eel species and located East of Madagascar. Since then, noleptocephalus has been caught. Glass eels or riverine yellow eelsof those four species were sampled in estuaries and rivers of LaRunion Island, Mauritius Island, Mayotte Island, and on the east-ern coast of Madagascar between 2000 and 2006, highlighting thatthe composition of eel communities seems contrasted between thedifferent islands of this region (Robinet et al., 2008), as Ege (1939)previously suggested. This contrasted distribution may be due: (1)to different migratory routes followed by the eel larvae before theirestuarine recruitment (Robinet et al., 2003, 2008; Rveillac et al.,2008); (2) to different spawning areas or periods among species:A. marmorata and A. mossambica would share the same spawningarea located between Madagascar and the Mascarene ridge, withdifferent spawning periods, while A. bicolor bicolor might spawnfrom an area located westwards and closer to La Runion Island(Robinet et al., 2003); (3) to different behavioral traits, or intrinsicmetabolism (Robinet et al., 2003, 2008; Rveillac et al., 2008).These hypotheses are mainly based on otolith microstructure anal-yses, which allow to assess the leptocephalus stage duration, theperiod of recruitment as well as the spawning period (Robinet etal., 2003; Rveillac et al., 2008, 2009). However, eld data beingsparse, these hypotheses need to be confronted to modelingapproach.

    Hydrodynamic modeling provides quantitative methods ofapplying physical oceanographic information to the question oflarval dispersal (Tremblay et al., 1994; Young et al., 1998; Ellienet al., 2004). These models can provide a synoptic view of larvaldistribution for a large range of geographic, hydrodynamic and cli-matic conditions at different spatial and temporal scales. They canalso be used as a tool to determine the relative inuence of variouscomponents of the water circulation and biological factors on lar-val dispersal (Barnay et al., 2003; Ellien et al., 2004). Lagrangianmodels have already been used to explain the distribution of theeels and their larvae and to discuss the inuence of oceanographicpatterns on larval dispersal and recruitment in the Pacic (Kimuraet al., 1999; Kim et al., 2007) and Atlantic Ocean (Kettle and Haines,2006; Bonhommeau et al., 2009a,b).

    In this study, we use a state-of-the-art hydrodynamical model,which realistically reproduces the 3D circulation in the region,associated with a Lagrangian model that reproduces the possiblemigration pathways of larvae. The biological inputs, such as migra-tion durations, together with spawning and recruitment periods,have been deduced from the published results on otolith micro-structure analyses from all eel species. Hence simulations repro-duce migration of eel larvae of the SWIO with no distinction ofthe species.

    The aim of this study is to understand how the ocean circulationin the SWIO constraints the location of spawning areas for tropicaleels: (1) can there be a common spawning area to all freshwatereel species sampled in their different recruitment sites? (2) Whatare the leptocephali possible migration routes to the recruitmentsites? (3) What is the variability in the recruitment and migrationduration induced by the environmental conditions? As a rst stepto answer these questions, we consider a very simple parameteri-sation of the larvae behavior in order to focus on the impact ofenvironmental conditions alone. We nally discuss the effect of

    graphy 86 (2010) 396413 397more complex behavior (vertical diurnal migration and mortalityscenarios).

  • 2. Data, models and methods

    Our results are based on simulations of the oceanic circulationin the SWIO, from a state-of-the-art hydrodynamical model, usedto investigate migration pathways of eel larvae, taking into accountthe few observed biological parameters. In this section, we de-scribe the general features of the ocean circulation in the SWIO, re-view the available observations referring to eel early life traits,present the hydrodynamical and Lagrangian models and nallyintroduce the suite of experiments used in the present study.

    2.1. Data

    2.1.1. Circulation and hydrology in the SWIOThis study is located in the SWIO, between 5N and 27S, 34E

    and 75E (Fig. 1). We focus on the eastern coast of Madagascar,the island of Mayotte and the Mascarene Plateau including

    398 S. Pous et al. / Progress in OceanoFig. 1. Chart of the southwestern part of the Indian Ocean. Gray lines indicateisobaths 200 m (shaded), 1000 m and 2000 m from the bathymetry of DRAKKARMauritius and La Runion islands. The Mascarene Plateau is char-acterized by a series of shallow banks and shoals separated by dee-per ridges, encompassing over 2000 km from The Seychelles at thenorthern end, to Mauritius at the southern end.

    Circulation in the SWIO mostly consists of the South EquatorialCurrent (hereafter SEC), which is the northern boundary of the ba-sin-wide anticyclonic subtropical gyre (Stramma and Lutjeharms,1997). The SEC is a broad ow, heading westward, located between10S and 20S. New et al. (2007) showed that when the SEC passesacross theMascarenePlateaunear 60E, it splits into twocores form-ing a northern core between 10S and 14S (passing north of Saya deMalha Bank and between Saya de Malha and Nazareth Banks) and asouthern core between 17S and 20S (passing between CargadosCarajos Bank and Mauritius). When reaching Madagascar, thosecores form the North-East and South-East Madagascar Currents(hereafter NEMC and SEMC, Schott andMcCreary, 2001). The NEMCows northwards around Madagascar and reaches the ComorosArchipelagos (including Mayotte) whereas the SEMC ows south-wards aroundMadagascar. Near 25S, a part of SEMC seems to retro-ect and feed a shallow eastward jet (Palastanga et al., 2007), theSouth Indian Ocean Countercurrent (hereafter SICC).

    During the North-East (hereafter NE) monsoon, trade winds areweaker and located further south, which is expected to induce sea-sonal changes of the SEC. Nevertheless, the latter remains unclear(Schott and McCreary, 2001), as illustrated by the climatologicalsurface currents in February (NE monsoon, Fig. 2a and c) and Au-gust (Southwest (SW) monsoon, Fig. 2b and d). Indeed, year-longin situ observations of the NEMC and SEMC suggest that theirsimulations. Black arrows locate the main currents, namely the South EquatorialCurrent (SEC), the Northern East Madagascar Current (NEMC) and the Southern EastMadagascar Current (SEMC).seasonal cycles have small amplitude and cannot be signicantlydistinguished from the strong intra-seasonal variability (Schottet al., 1988).

    As found in the Atlantic and Pacic Oceans, the eel spawningarea is likely to be marked by thermal and/or salinity fronts. Inthe SWIO, there is no clear thermal front in the upper ocean asillustrated along WOCE section IR3 in April 1995 (Fig. 3a). Never-theless the SEC acts as a barrier, separating water masses of south-ern and northern origin with contrasted salinities (Wyrtki, 1971;Schott and McCreary, 2001; New et al., 2007). Tropical SurfaceWater (TSW), carried westwards by the SEC (between 5S and20S in the upper 50100 m) from the Indonesian Seas to the Afri-can coast, has relatively low salinity. It is anked on its northernside by Arabian Sea High Salinity Water (ASHSW), identied by asalinity maximum under the mixed layer (at 100 m deep approxi-mately) and on its southern side by Subtropical Surface Water(STSW), identied by a salinity maximum at 200250 m. The sub-sequent surface salinity fronts, located around 5S and 18S respec-tively (Fig. 3c) are also observed east of section IR3, on theMascarene ridge around 12S and 19S respectively in July 2002(New et al., 2007).

    2.1.2. Biological dataIn the SWIO, glass eels or riverine yellow eels of three species

    (i.e. Anguilla bicolor bicolor, A. marmorata and A. mossambica) havebeen captured (mostly by electroshing, but also by traditional netshing in Madagascar) in estuaries of four islands: La Runion is-land, Mauritius island, Mayotte and Madagascar (see Table 1 thatsummarizes published information regarding 237 glass eels or el-vers captured in nine recruitment sites of the SWIO from 2000 to2006). We review otolith microstructure analyses that give infor-mation on the recruitment date, the age at recruitment and theduration of marine larval life, based on the counting of dailyincrement rings, from which the spawning period can be deduced(Lecomte-Finiger, 1992; Arai et al., 2001; Kuroki et al., 2007; Rve-illac et al., 2008). In the SWIO, the maximum estimated migrationduration is 180 days, and recruitment mostly occurred from Janu-ary to April, corresponding to a hatching period from August toDecember.

    These biological data are based on the assumption that incre-ment rings are deposited daily in otoliths. This assumption is sub-ject of controversy, in particular for European eel (A. anguilla), asthe leptocephali migration duration assessed from modeling study(Kettle and Haines, 2006; Bonhommeau et al., 2009a), or eld stud-ies and cohort analyses (McCleave, 2008) is twice as long as thatrevealed by otolith microstructure analyses. This difference couldbe explained by the leptocephalus metabolic rate that may beslower in cold temperature, resulting in a deposit increment everyother day instead of daily as it is currently assumed (Bonhommeauet al., 2009a). However, in warmer waters such as the PacicOcean, the increment rings seem to be deposited on a daily basis.Indeed, Tsukamoto (2006) collected, on their spawning area, newlyhatched A. japonica pre-leptocephali, and the counting of theirgrowth rings showed that they had hatched between 2 and 5 daysbefore their catch. This direct observation validates the pertinenceof considering that increment rings are deposited daily, at least foreel leptocephali in warm waters. In the Indian Ocean, the watertemperature allows us to consider that the published data on oto-liths microstructure analyses constitute a pertinent basis for oursimulation biological inputs.

    2.2. Models

    graphy 86 (2010) 3964132.2.1. The hydrodynamic modelOceanic data used in this study (ocean currents, water temper-

    ature and salinity from surface to bottom) is provided by two mod-

  • E

    (b

    (d

    eano50 cm.s1

    Feb

    (a)

    30oE 35oE 40oE 45oE 50oE 55oE 60oE 65oE 70o

    5oS

    10oS

    15oS

    20oS

    25oS

    30oS

    50 cm.s1(c) 5oS

    S. Pous et al. / Progress in Oceling experiments of the DRAKKAR group (www.ifremer.fr/lpo/drakkar). The rst simulation, noted DRAK01 (the ORCA025-G70simulation referred by the DRAKKAR group, 2007) is used for years19882002. In order to have a longer dataset of hydrodynamicalconditions, we also use simulation DRAK02 that reproduces years20002007. The global ORCA025 coupled ocean/sea-ice model con-guration developed by the European DRAKKAR collaboration(DRAKKAR group, 2007) is used to perform DRAK01 and DRAK02simulations. An overall description of the model and its numericaldetails are given in Barnier et al. (2006). This model congurationuses the ORCA global tri-polar grid (Madec and Imbard, 1996) at 1/4 resolution (1442 1021 grid points and 46 vertical levels). Ver-tical grid spacing is ner near the surface (6 m) and increases withdepth to 250 m at the bottom. Horizontal resolution is 27.75 km atthe equator, 13.8 km at 60N, and gets to 10 km in the ArcticOcean. The ocean/sea-ice code is based on the NEMO frameworkversion 1.9. (Madec, 2008). It uses a partial step representation ofthe bottom topography and a momentum advection scheme whichboth yield signicant improvements (Penduff et al., 2007; Le Som-

    50 cm.s1

    Feb

    (e)

    30oE 35oE 40oE 45oE 50oE 55oE 60oE 65oE 70oE

    5oS

    10oS

    15oS

    20oS

    25oS

    30oS

    (f

    Feb

    30oE 35oE 40oE 45oE 50oE 55oE 60oE 65oE 70oE

    10oS

    15oS

    20oS

    25oS

    30oS

    Fig. 2. Climatological near surface circulation (at 15 m deep) in February (i.e. NE monsoodrifter-based climatology (a and b, Lumpkin and Garraffo, 2005), a 19932007 climatoloand d, OSCAR data obtained from JPL Physical Oceanography DAAC and developed by ESR2007).50 cm.s1

    Aug

    )

    30oE 35oE 40oE 45oE 50oE 55oE 60oE 65oE 70oE

    5oS

    10oS

    15oS

    20oS

    25oS

    30oS

    50 cm.s1) 5oS

    graphy 86 (2010) 396413 399mer et al., 2009). The bathymetry of the model is adapted to betterrepresent the different pathways between the banks of the Masca-rene Ridge (New et al., 2007). These global simulations are drivenwithout data assimilation by the hybrid interannual forcing DFS3described in detail in Brodeau et al. (2010) which blends dailyand monthly satellite-derived radiative uxes and precipitationswith 6-hourly 10-m atmospheric surface variables for turbulentuxes from the ERA40 reanalysis (DRAK01) or ECMWF analysis(DRAK02). Hence these two simulations only differ in the turbulentuxes (heat and momentum) at the surface. Circulation andhydrography in DRAK01 and DRAK02 are very similar, except thatthe circulation is slightly more intense in DRAK02 compared toDRAK01. Considering the three years in common of DRAK01 andDRAK02 (20002002), we are able to assess the inuence of thehydrodynamic forcings on the biological results (i.e. migrationduration).

    Barnier et al. (2006) highlighted the performance of the simula-tions in reproducing the strong ocean currents and the variabilitydue to eddies, even in comparison with higher resolution models.

    50 cm.s1

    Aug

    )

    30oE 35oE 40oE 45oE 50oE 55oE 60oE 65oE 70oE

    5oS

    10oS

    15oS

    20oS

    25oS

    30oS

    Aug

    30oE 35oE 40oE 45oE 50oE 55oE 60oE 65oE 70oE

    10oS

    15oS

    20oS

    25oS

    30oS

    n, left panels) and August (i.e. SWmonsoon, right panels) derived from a 19982003gy of ocean currents computed from various satellites and in situ measurements (c) and from DRAK01 simulation averaged from 1993 to 2002 (e and f, Penduff et al.,

  • eano10N

    20N

    30N

    400 S. Pous et al. / Progress in OcThe main features of the global circulation are well represented aswell as the interannual variability (Drakkar group, 2007; Penduffet al., 2007). In the SWIO, the mean simulated circulation comparesreasonably well with that observed, considering that the two cli-matologies derived from observations are not fully consistent witheach other (Fig. 2). The regions of maximum current are very sim-ilar to those found in observations, although the southern core of

    30E 40E 50E 60E30S

    20S

    10S

    EQ

    Fig. 3. Hydrography along section WOCE IR3 (location indicated in top panel) in April, 2www.ewoce.org/) and DRAKKAR-G70 simulation (b and d).graphy 86 (2010) 396413the SEC owing westward south of 17S is weaker than observed.This southern core feeds the southward current along Madagascar(SEMC), which is also underestimated in the model. Note that thehorizontal resolution of the model remains too coarse to reproducethe circulation in the vicinity of small islands (20 km offshore)such as La Runion, Mauritius and Mayotte, but this is likely toaffect only the very last days of migration. Consequently, for the

    70E 80E

    Oce

    an D

    ata

    View

    18, 1995: temperature and salinity from, respectively, WOCE data (a and c, http://

  • e 3 Aveill

    Stsa

    G

    GGGE

    G

    GGG

    G

    G

    G

    G

    GGGGG

    eanopurpose of this study that is to understand better the impact of theopen ocean circulation on the migration of eel larvae in the SWIO,we are condent that the model is reliable. Simulated temperature

    Table 1Oceanic larval migration duration and hatching periods, reported in literature, for thmodeling study. References: (1) Robinet et al. (2003), (2) Robinet et al. (2008), (3) R

    Species Sampling locations Months of capture(N individuals)

    A. bicolor bicolor La Runion From November 2000 toApril 2001 (11)April 2005 (1)

    Mauritius April 2005 (9)Mayotte April 2005 (12)

    April 2005 (3)

    A. marmorata La Runion From November 2000 toApril 2001 (9)April 2005 (15)

    Mauritius April 2005 (30)Mayotte April 2005 (29)Madagascar (from northto south)(a) Mananjary From November 2005 to

    February 2006 (15)(b) Farafangana January 2006 (15)

    A. mossambica La Runion From November 2000 toMarch 2001 (12)April 2005 (1)

    Madagascar (from northto south)(a) Andevoranto December 2005 (15)(b) Mahanoro December 2005 (15)(c) Mananjary December 2005 (15)(d) Manakara December 2005 (15)(e) Vangaindrano December 2005 (15)

    S. Pous et al. / Progress in Ocand salinity of the upper ocean also compares well with observa-tions (Fig. 3).

    The large-scale climate variability in the Indian Ocean is wellrepresented in the model, according to the time uctuations ofthe commonly used climatic indices. The tropical Indian Ocean Di-pole index (hereafter IOD), dened as the difference in sea surfacetemperature (hereafter SST) anomaly in June-November betweenthe western (5070E, 10N10S) and the eastern (90110E, 010S) tropical Indian Ocean (Webster et al., 1999; Saji et al., 1999)is in the model positive in 1991, 1994, 1997, 2006 and 2007, andnegative in 1992, 1996 and 1998, which corresponds to observa-tions (http://www.jamstec.go.jp/frcgc/research/d1/iod/). The sec-ond major climatic index for the Indian Ocean, which is thesouthern Indian Ocean Dipole index (hereafter SIOD), dened asthe difference in SST anomaly in DecemberMarch between thewestern (5565E, 2737S) and the eastern (90100E, 010S) sub-tropical Indian Ocean (Behera and Yamagata, 2001), is in the modelpositive in 1993, 1997, 1999, 2001, 2006 and 2007, and negative in1995, 1998, 2000, 2002 and 2003, which corresponds to observa-tions (http://www.jamstec.go.jp/res/ress/behera/siod.html). Thesetwo indices relate to independent phenomenon, hence are not cor-related to each other (Behera and Yamagata, 2001). They bothinuence the circulation in the SWIO but with different time lags:the intensity of SEC, SEMC and NEMC tends to increase (respec-tively decrease) in phase (i.e. the same year) with a positive (neg-ative) SIOD event (Hermes and Reason, 2005) and 12 years after anegative (positive) IOD event (Palastanga et al., 2006).

    2.2.2. The Lagrangian modelWe used the Lagrangian tool ARIANE to simulate the trajectory

    of passive particles based on 3D streamlines, in a given velocityeld and subsequent water mass characteristics (http://www.u-niv-brest.fr/lpo/ariane, Blanke and Raynaud, 1997). In our study,the passive particles are assumed to represent eel larvae. Trajecto-ries are calculated off-line from the 3D 5-days-averaged oceanvariables of simulations DRAK01 and DRAK02. This method allows

    nguilla species in the SWIO. These data are used as biological inputs in the presentac et al. (2008), (4) Rveillac et al. (2009).

    age atmpling

    Oceanic larvalmigration durations(mean SD, minmax)(in days)

    Hatching periods Ref.

    lass eels 46.2 5.8 (3957) September 2000January 2001 1

    lass eels 151 September 2004 2lass eels 110.6 16.7 (87139) OctoberDecember 2004 2lass eels 101.8 9.2 (87117) OctoberDecember 2004 2lvers 97.0 4.4 (92100) MarchAugust 2004 2

    lass eels 96.9 26.4 (60135) SeptemberDecember 2000 1

    lass eels 111.1 15.9 (94142) AugustDecember 2004 3lass eels 139.2 24.0 (91180) AugustDecember 2004 3lass eels 120 13.1 (104151) SeptemberDecember 2004 3

    lass eels 122.5 15.2 (92145) SeptemberNovember 2005 2

    lass eels 108.8 8.4 (94120) AugustSeptember 2005 2

    lass eels 102.1 17.2 (72130) JulyOctober 2000 1

    lass eels 113 October 2004 2

    lass eels 82.5 8.2 (7096) August 2005 4lass eels 84.8 10.8 (64102) AugustSeptember 2005 4lass eels 92.5 18.8 (73131) JulyAugust 2005 4lass eels 103.0 20.3 (79152) JulyAugust 2005 4lass eels 96.6 14.5 (78127) JulyAugust 2005 4

    graphy 86 (2010) 396413 401to perform a large number of experiments, which would not bepossible using on-line calculation. Backward trajectories are easilycomputed by simply changing the sign of the velocity eld. Basedon the location of the recruitment site and the larval life duration,the backward strategy allows us to simulate the migration path-ways of larvae back in time, from each recruitment site to the posi-tion of possible spawning areas.

    ARIANE tool offers two types of statistical analysis. The rst oneis a qualitative approach (hereafter Qualitative), where one particlestands for one larvae. The initial position of each particle is set (lat-itude, longitude, depth and time), then ARIANE calculates their tra-jectory. As the position in space and time of each particle isrecorded, only a few particles (of the order of 10,000 for our choiceof release) can be followed in a single experiment. Hence they maynot illustrate all the possible pathways of larvae migration.

    The Lagrangian tool is also used in a quantitative mode (hereaf-ter Quantitative). In that case, particles are associated with smallwater parcels (500m3) instead of representing larvae directly.The whole water mass of a given geographical section is discretizedin those water parcels, one parcel being released for each elemen-tary transport across the section (Ds, 1995; Blanke and Raynaud,1997). Trajectories of the water parcels are calculated until the par-cel reaches a given geographical section. Hence quantitative exper-iments give access to uxes between different sections rather thanto individual trajectories. The amount of particles released dependson the intensity of the water ow near each initial geographicalsection and on the duration of the experiment. On average,35,000600,000 particles are released in the Quantitative experi-ments, which allow to assume that all the possible pathways areinvestigated.

    There are two limitations to the Lagrangian model. First, it uses5-day averaged velocity elds from the hydrodynamical model,hence current anomalies at higher frequencies are not taken into

  • eanoaccount when calculating off-line the Lagrangian trajectories. Thismay introduce biases in the displacement of particles. However,such high frequency processes imply oceanic adjustment at spatialscales of the order of 10 km (sub-mesoscale processes), which can-not be resolved presently in multidecadal simulations. Besides,Lagrangian calculations are done with a purely advective schemewhich implies that horizontal and vertical diffusion of the particlesare not represented. This may affect the displacement of larvae.Nevertheless, using a purely advective scheme is the only way todo backward calculation.

    2.2.3. Biological parameterisationsThere are indications of vertical migration on a diurnal time

    scale during the ontogenetic development of the Atlantic and Jap-anese eel larvae (Castonguay and McCleave, 1987; Otake et al.,1998; Tsukamoto, 2009). Nevertheless there is no observation ofthe behavior of eel larvae in the Indian Ocean. As a result, we rstassume that eel larvae may be represented by strictly passive par-ticles and focus on the inuence of environmental conditionsalone. Considering the larvae as passive particles implies that theyeventually migrate on the vertical following the ocean currents. Inmost experiments, we release the larvae in the upper layer of theocean (0100 m depth). However, in Section 4, we discuss theeffect of implementing a particle behavior that mimics the verticaldiurnal migration of larvae, making particles ip every 12 h from50 m to 200 m. This range of depths was chosen based on lepto-cephalus migration in the Atlantic and Pacic Oceans as there isa lack of knowledge about the leptocephalus migration for the dif-ferent eel species in the SWIO. In the Atlantic and Pacic Oceans,leptocephali migrate up to 50 m, which value we choose as upperlimit for vertical migration of leptocephali in the SWIO. The lowerlimit of vertical migration differs from the Atlantic Ocean,where leptocephali reach 300 m deep (Castonguay and McCleave,1987), to the Pacic Ocean where particles only reach 150 mdeep (Kimura et al., 2001). We choose an arbitrary intermediatevalue of 200 m as lower limit of larvae vertical migration in theSWIO.

    No realistic mortality is taken into account for the simulationsof larval dispersal, as no data is available on this demographicparameter. However, when the migration duration of a particle ex-ceeds 180 days, the particle vanishes, which can be viewed as acoarse parameterisation of mortality (0% mortality before 180 daysand 100% onwards). In Section 4, we discuss the effect of imple-menting a more realistic parameterisation of mortality: the cumu-lative mortality from 0 to time t, being 1 eMt, where M is theannual mortality rate (Peterson and Wroblewski, 1984).

    Otolith analyses of newly hatched Japanese eel leptocephali sug-gests that the spawning is discrete and occurs around 4 days beforenew moon, every month within the spawning period (Tsukamotoet al., 2003; Tsukamoto, 2006). In the forward experiments, spawn-ing is represented as a continuous phenomenon occurring from Au-gust to December in order to investigate all the possiblehydrodynamical conditions and larvae migration pathways,

    Lagrangian simulations reproduce a number of particles thatlargely exceeds the number of glass eels captured in the SWIO.The histogram of simulated migration durations exhibits one orseveral modes, suggesting that the mode(s) may be statisticallymore signicant than the mean to characterize migration duration.Nevertheless, as observations are insufcient to calculate the modeof observed migration durations, we choose to compare the meansimulated migration duration (and eventually the minimum) tothe observed migration durations. Besides, we are aware that lim-iting migration durations to 180 days may introduce a bias in the

    402 S. Pous et al. / Progress in Ocmean migration duration. Nevertheless, this coarse parameterisa-tion of mortality is unavoidable to ensure that the water particlessimulated by the Lagrangian model mimic eel larvae.2.3. Simulation design

    We rst test whether a common spawning area for each recruit-ment site (i.e. Mayotte, Mauritius, La Runion and the southeastcoast of Madagascar (Mananjary and Vangaindrano), Fig. 1) canbe identied, considering the ocean currents in the SWIO. For thatpurpose, we run qualitative backward experiments BQL1 (coveringyears 19892002 using DRAK01, see Table 2 for a description of allexperiments) and BQL2 (covering years 20012007 using DRAK02).The initial positions of the particles are in the vicinity of therecruitment sites. Particles are released once a month between Jan-uary and April, at 4 depth levels from the surface to 100 m deep. Atotal of 2000 particles are released at each recruitment site everyyear, which corresponds to about 10 particles per grid cell. Thelocation of each particle is registered every day and for 180 days.The trajectories of the particles back in time allow us to describethe possible migration routes of eel larvae to the different recruit-ment sites. We illustrate these migration routes in 2006 (date atrecruitment) because it reproduces all larvae trajectories observedover the 19 simulated years, and then we evaluate the possibilityof a common spawning area.

    The eastward extension of the possible common spawning areais investigated by running backward quantitative experiments(Experiments BQN1-8, Table 2). The initial geographical sectionsconsist in four recruitment areas along the Madagascar coast(North-West, North-East, East and South-East) and the three is-lands previously identied (Mayotte, Mauritius and La Runion).Particles are released at recruitment sites continuously from thesurface to 100 m deep and from January to April, correspondingto a continuous spawning period from August to December. We de-ne four nal sections at xed longitude ranging westward from70E to 60.5E (the location of the central Mascarene ridge). Exper-iments are run for years 19892002 using DRAK01 (BQN1-4) andfor years 20012007 using DRAK02 (BQN5-8).

    We pursue our investigations of the inuence of environmentalconditions on the larvae migration, by evaluating the success forparticles released in selected regions of the spawning area to reacheach recruitment site. For this purpose, we run forward quantita-tive experiments (Experiments FQN1-5, Table 2) with initial geo-graphic sections located at three chosen spawning sites and nalgeographic sections at the seven recruitment sites. Particles are re-leased continuously from the surface to 100 m deep and from Au-gust to December. Experiments FQN1-3 are run for each year from1988 to 2001 using DRAK01, while Experiments FQN4-5 are run foryears 20002006 using DRAK02 to test the sensitivity to differenthydrodynamic forcings. For each year, we examine the proportionsin which the successful particles (i.e. particles that succeed inreaching one of the recruitment sites) share out between therecruitment sites. We also calculate the migration durations fromthe chosen spawning sites to the recruitment sites and comparewith observations for years 2004 and 2005 (Experiments FQN4-5).

    Additional sensitivity experiments are discussed in Section 4.We run experiments FQL6-7 to test the effect of vertical diurnalmigration of the particles (particles ip every 12 h between 50and 200 m). In these qualitative experiments, particles are releasedat the same initial location as in reference quantitative experi-ments with no larvae behavior FQN4 and FQN5 (similar to FQN1and FQN3 respectively, but covering years 20002006), with thesame frequency of release. Because this implies tracking the posi-tions of 450,000 particles, we only run FQL6-7 for years 2005 and2006. In all experiments above, the mortality does not depend onthe duration of migration as long as they do not exceed 180 days.We also test the inuence of a more realistic parameterisation of

    graphy 86 (2010) 396413mortality by running quantitative experiments FQN8 (whereM = 3.8 yr1) and FQN9 (where M = 8 yr1), which are otherwisesimilar to FQN4.

  • relea

    Fina

    E1(resE1(resMadMayMadMayMadMayMadMayMadMayMadMayMadMayMadMayMadMay

    eanoTable 2List of experiments.

    Experiment Type Direction Initial conditions of particle

    Locationb

    Initial

    BQL1 Qualitative Backward Madagascar (2) MauritiusMayotte Runion

    BQL2 Qualitative Backward Madagascar (2) MauritiusMayotte Runion

    BQN1-4 Quantitative Backward Madagascar (4) MauritiusMayotte Runion

    BQN5-8 Quantitative Backward Madagascar (4) MauritiusMayotte Runion

    FQN1 Quantitative Forward B1

    FQN2 Quantitative Forward B2

    FQN3 Quantitative Forward B3

    FQN4 Quantitative Forward B1

    FQN5 Quantitative Forward B3

    FQL6 Qualitative Forward B1

    FQL7 Qualitative Forward B3

    FQN8 Quantitative withmortality M = 3.8 yr1

    Forward B1

    FQN9 Quantitative withmortality M = 7.6 yr1

    Forward B1

    S. Pous et al. / Progress in Oc3. Results

    3.1. Backward qualitative migration routes: observation ofconvergence areas

    We rst illustrate the trajectories that reach the recruitmentsites of Mayotte, Mauritius, La Runion and the southeast coastof Madagascar (Mananjary and Vangaindrano, Fig. 1) in 2006(Fig. 4). To simplify Fig. 4, we only show trajectories passingthrough the Mascarene Ridge, which constitute a large majorityof the trajectories, the others spreading away in various directions.

    Most of the particles reaching Mayotte are transported by thenorthern core of the SEC, between Saya De Malha and NazarethBank (10S15S) directly North of Madagascar, and then by theNEMC to Mayotte (Fig. 5d). Nevertheless, some particles are trans-ported by the southern core of the SEC (17S20S), passing southof Carcados Carajos Shoals, before reaching the northern part ofMadagascar (Fig. 4e and f). As the northern route is shorter andpresumably faster because the currents are more intense, migra-tion duration via the northern route is shorter than via the south-ern route (the minimum migration duration from the MascareneRidge is 60 days via the northern route vs. 110 days via the south-ern route). Besides, the Lagrangian trajectories of particles recruit-ing in Mayotte suggest that the NEMC is partly fed by the southerncore of SEC.

    Both recruitment sites in Madagascar are fed by the same twopathways. Particles transported by the northern core reach thecoast of Madagascar near 16S while other particles transportedby the southern core reach the coast near 17S (Fig. 4g). Then bothstreams merge and ow southwards banked along the coast. Here

    a Lagrangian simulations are performed for each year of the period 19882002 (respHereinafter, the year of each experiments is the calendar year at the time of spawning f

    b Two (resp. four) locations on Madagascar are considered for the qualitative (respbackward (forward) experiments are shown in Fig. 6a (resp. Fig. 7).

    c Particles are released at xed dates for the qualitative experiments and continuouslythe quantitative experiments.

    d Particles are released at xed depth for the qualitative experiments, continuously in tquantitative experiments BQN1-4, FQN1-5 and FQN8-9, and migrate every 12 h betweese at Spawning/Recruitment sites Periodsa

    Datec Depthd

    l

    January 5, February 5,March 5, April 5

    12, 25, 45,100 m

    19892002

    January 5, February 5,March 5, April 5

    12, 25, 45,100 m

    20012007

    E2E3E4pectively)

    DecemberApril 0100 m 19892002

    E2E3E4pectively)

    DecemberApril 0100 m 20012007

    agascar (4) Mauritiusotte Runion

    AugustDecember 0100 m 19882001

    agascar (4) Mauritiusotte Runion

    AugustDecember 0100 m 19882001

    agascar (4) Mauritiusotte Runion

    AugustDecember 0100 m 19882001

    agascar (4) Mauritiusotte Runion

    AugustDecember 0100 m 20002006

    agascar (4) Mauritiusotte Runion

    AugustDecember 0100 m 20002006

    agascar (4) Mauritiusotte Runion

    AugustDecember 50/200 m 20052006

    agascar (4) Mauritiusotte Runion

    AugustDecember 50/200 m 20052006

    agascar (4) Mauritiusotte Runion

    AugustDecember 0100 m 20002006

    agascar (4) Mauritiusotte Runion

    AugustDecember 0100 m 20002006

    graphy 86 (2010) 396413 403again, the northern pathway is the fastest (the minimum is125 days vs. 160 days for the southern route). Interestingly, thissuggests that the SEMC is partly fed by the northern core of SEC.

    Most particles recruiting in Mauritius and La Runion are alsocoming from the southern core of the SEC (Fig. 4d and f). As Mau-ritius is located close to the Mascarene Ridge, migration durationmay be as short as 50 days for the year considered, which is shorterthan the migration duration estimated from observations (Table 1).The shortest migration duration to La Runion is slightly longerthan to Mauritius (90 days). Another migration route appears forLa Runion site, at the beginning similar to the northern one de-scribed for the Madagascar recruitment sites and then, near 21S,turning east toward La Runion island (Fig. 4fh). Migration dura-tion along this second pathway is longer (150 days for the fastestpathway).

    If there exists a common spawning area for all recruitmentsites, it has to be located simultaneously within the envelope ofall the possible locations of particles up to 180 days beforerecruitment, relative to each recruitment site (Fig. 5a). As theintersection of all envelopes is not empty in 2006 (shaded areain Fig. 5a), a common spawning area (or several ones) may existin the SWIO. The intersection of all envelopes, referred to as theconvergence zone, is roughly centered near the central Mascareneridge but extends southwestwards to the northern region of Mau-ritius and La Runion island. Note that the convergence zone is re-stricted to the North-West of La Runion Island for migrationdurations of 90 days and progressively extends to the North-East,near the Mascarene ridge, for longer migration durations (Fig. 5band c). Considering the 19 years of experiments BQL1 and BQL2individually, there is substantial interannual variability in the

    . 20002007) with velocity from the model conguration DRAK01 (resp. DRAK02).or forward experiments and at the time of recruitment for backward experiments.. quantitative) experiments. Location of nal (initial) conditions for quantitative

    in the given time range (depending on the transport across each initial section) for

    he given depth range (depending of the transport across each initial section) for then 50 and 200 m deep in experiments FQL6-7.

  • eano404 S. Pous et al. / Progress in Ocpathways of particles: the shape of trajectories may shift to thenorth or to the south and the eastward extension also exhibitsgreat variability (not shown). These induce substantial changesin the shape of the convergence zone, but the latter always exists,suggesting that there may always be a common spawning area foreel larvae in the SWIO.

    Fig. 4. Simulated trajectories from the recruitment sites (a) backward in time after (b) 10for recruitment in 2006.graphy 86 (2010) 3964133.2. Backward quantitative experiment: extension of the convergencearea east of the Mascarene ridge

    East of the Mascarene Ridge, at sections E1, E2 and E3 (Fig. 6a),particles coming from the recruitment sites are conned between9S and 17S for recruitment years 20012007 (Fig. 6b based on

    days, (c) 20 days, (d) 60 days, (e) 90 days, (f) 120 days, (g) 160 days and (h) 180 days

  • eanoS. Pous et al. / Progress in Ocexperiments BQN5-8, Table 2), as well as for recruitment years19892002 (not shown), which corresponds to the location ofthe SEC (Fig. 2f). The bulk of particles slightly migrates southwardfrom E1 to E3, due to the southward component of the currentsvelocities. As previously described from the qualitative experi-ments, particles reaching Mauritius are rather located in thesouthern SEC. Closer to the Mascarene ridge, at section E4, the lat-itudinal distribution of the particles is wider, and several peaksappear in the distribution: one at 14S and several south of16S. They correspond to the northern and southern routes tothe recruitment sites.

    Considering the 19 years of experiments BQN1-8 individually,there are years when particles released from a particular recruit-ment site do not reach (backward in time) the nal sections E1,E2 or E3, while particles always reach section E4. If the spawningarea was located at section E3 (respectively E2 and E1), there are2 (respectively 4 and 10) years out of 19 for which no recruitmentwould occur in La Runion. Similarly, there are few years for whichthere would be no recruitment at Madagascar South-East, Mada-gascar East and Mauritius for a spawning area located at sectionsE2 and E1. Considering these probability results, section E4 is themost favorable eastern boundary for the spawning area in the cen-tral Mascarene ridge.

    As expected from the location of the sections, the mean migra-tion duration to the recruitment sites, averaged over years 2004and 2005 to compare with observedmigration durations in Table 1,

    Fig. 5. Envelope of the location of particles up to 180 days before recruitment in 2006 foat 90 (b) and 130 (c) days before recruitment. Cross hatched area locates the convergengraphy 86 (2010) 396413 405increases from section E4 to section E1 (Fig. 6c). At sections E3 andE4, the distribution of migration durations as a function of the lat-itude indicates that migration is the shortest for MadagascarNorth-East, Mayotte and Madagascar North-West, via the directnorthern route, and for Mauritius via the southern route, especiallyfor a spawning area at section E4. The structure is more complexfor the other recruitment sites, and depends on the section thatis considered. Among all particles released in 2005 at La Runion,none reaches (backward in time) section E1, which is incompatiblewith eld observations. Particles released in 2005 at La Runionthat reach section E2 have an average migration duration of175 days and a minimum migration duration of 171 days (notshown), which largely exceeds observations. These two points givemore credit to a common spawning area located in the vicinity ofsections E3 and E4 (assuming that the common spawning area islocated east of the Mascarene Ridge). Indeed, for a spawning arealocated near section E4 (respectively E3), there are particles thatrecruit in 4 (respectively 3) sites with an average migration dura-tion of 110 days that corresponds to the ensemble average ofobservations over all species and years. Concerning the recruit-ment sites of Madagascar East and South-East, the average migra-tion duration exceeds 140 days even for a spawning area locatednear section E4, but the minimum migration duration for particlesthat recruit in 2005 (respectively 88 and 101 days, not shown) iscompatible with observations. However, the backward simulationto sites located at the west of E4 was technically impossible.

    r Mauritius, Mayotte, La Runion and two sites of Madagascar (a). Same for particlesce zone.

  • eano 0o

    5oS

    10oS

    15oS

    20oS

    25oS

    406 S. Pous et al. / Progress in Oc3.3. Forward quantitative migration: arrival success for a spawningarea on the Mascarene ridge

    From a spawning area B1 that encompasses the bulk of back-ward trajectories through section E4 (Fig. 7), arrival success tothe different recruitment sites exhibits interannual variability(Fig. 8a and b, left). A large majority of particles that recruit alwaysreaches Mayotte or the North-East site of Madagascar (respectively45% and 33% of particles that recruit on average over years 19882006). Arrival success to Mauritius (respectively Madagascar East)ranges from 4% to 12% (respectively from 1% to 11%). La Runion,North-West and South-East sites of Madagascar receive always lessthan 6% of successful particles and there are years when the arrivalsuccess drops to less than 1%. If the southern extension of thespawning area B1 is reduced (16S instead of 17S), the arrival suc-cess in Mauritius decreases from 8% to 4% on average over years

    0 20 40 0 20 40

    30S

    25S

    20S

    15S

    10S

    5S

    0

    E4 (60.5oE) E3 (63oE)

    Probability densit

    1 60100140180 6010014018030S

    25S

    20S

    15S

    10S

    5S

    0

    Migration d

    E4 E

    35oE 40oE 45oE 50oE 55oE 60oE

    Fig. 6. Location of the recruitment sites (Mayotte in dark blue, Mauritius in light blue, Lasections (black dashed lines) used for the quantitative backward experiments BQN1-8 (a)recruitment site in quantitative backward experiments BQN5-8: probability density fuduration (c, days) on average from 2004 to 2005. (For interpretation of the references to(a)

    graphy 86 (2010) 39641319882006, and there are years when it drops to less than 1% aswell as for La Runion (not shown). Hence recruitment in Mauri-tius and La Runion depends crucially on the southernmost exten-sion of the spawning area. For each site, arrival success alsodepends on the spawning date. However, the best spawning date,based on the arrival success, changes from one year to the other,as illustrated in Fig. 9 (left).

    On average over years 20002006 and over all particles that re-cruit (experiment FQN4), simulated migration durations to therecruitment sites of Mayotte and Mauritius are of the same orderof magnitude as observed, while those to La Runion and Madagas-car are overestimated (Fig. 8b, right and Tables 1 and 3). For aspawning event in 2004, simulated migration duration is the short-est for Mayotte and the North-East site of Madagascar (91 and93 days resp.) and the longest for the East and South-East sites ofMadagascar (163 and 165 days), while in 2005, the shortest migra-

    0 20 40 0 20 40

    E2 (66oE) E1 (70oE)

    y function (no unit)

    20002006 (b)

    Mayotte Mauritius La Runion Madagascar NWMadagascar NEMadagascar EMadagascar SE

    60100140180 6010014080

    20042005 (c)

    uration (day)

    3 E2 E1

    65oE 70oE

    Runion in green and Madagascar in red, gray, purple and light pink) and the nal. Statistical analysis of particles passing through sections E1, E2, E3 and E4 from eachnction over years 20002006 as a function of latitude (b, no unit) and migrationcolour in this gure legend, the reader is referred to the web version of this article.)

  • tion duration is recorded for the recruitment site of Mauritius(97 days in average), the longest being still in Madagascar Eastand South-East with 159 and 166 days respectively. However, inthe observed data set, La Runion seems to be the recruitment sitewith the shortest migration durations, the longest durations beingobserved either in Madagascar, Mayotte or Mauritius depending on

    82 for La Runion for a spawning in 2004, 88 days for MadagascarEast and 102 days for Madagascar South-East for a spawning in2005 (which roughly corresponds to observations on glass eels col-lected in Andevoranto and Mananjary).

    Migration duration depends on the spawning date on the intra-seasonal time scale (Fig. 8a and b, right), but the function linkingmigration duration to the spawning date is very noisy and doesnot exhibit a clear intra-seasonal signal. In addition, the migrationduration for a specic spawning date varies from one year to theother (Fig. 9, right), showing important interannual variability.

    There is no simple link between interannual uctuations in themean migration duration to each recruitment site and the large-scale climatic indices for the Indian Ocean IOD and SIOD (Fig. 10,correlations are not signicant at 95% level), presumably becausethese two climatic indices are not related to each other and inu-ence the circulation in the SWIO with different time lags. Never-theless in 1997, IOD and SIOD both contribute to a strengtheningof the SEC, NEMC and SEMC (IOD is negative in 1996 while SIODis positive in 1997, Fig. 10a), which may induce the decrease inthe mean simulated migration duration to all recruitment sites re-corded in 1997 (Fig. 10b). This illustrates that the IOD and SIODhave an impact on the eel larvae migration in the SWIO throughthe ocean circulation, although this impact is complex.

    3.4. Forward quantitative migration: arrival success for otherspawning areas

    B1B2B3

    35oE 40oE 45oE 50oE 55oE 60oE 65oE 70oE

    0o

    5oS

    10oS

    15oS

    20oS

    25oS

    Fig. 7. Location of the recruitment sites (Mayotte in dark blue, Mauritius in lightblue, La Runion in green and Madagascar in red, gray, purple and light pink) andthe spawning areas B1 B2 and B3 (black boxes) used for the quantitative forwardexperiments FQN1-5, FQL6-7 and FQN8-9. (For interpretation of the references tocolour in this gure legend, the reader is referred to the web version of this article.)

    S. Pous et al. / Progress in Oceanography 86 (2010) 396413 407the years of recruitment and the eel species (Table 1). In experi-ment FQN1, average migration durations are 419 days longercompared to FQN4, except for La Runion (Fig. 8a, right and Table3).

    It is important to note that, in spite of the differences betweenthe mean simulated and observed migration durations, for eachrecruitment site, there are always particles that recruit withinthe observed migration duration. Indeed, the simulated minimummigration duration is 47 days for Mayotte, 51 days for Mauritius,

    100

    80

    s (%

    )1988 1990 1992 1994 1996 1998 2000

    60

    40

    20

    0

    Arriv

    al s

    ucce

    s Mayotte MauritiusLa RuniMadagasMadagasMadagasMadagas

    2000 2002 2004 2006

    100

    80

    60

    40

    20

    0

    Arriv

    al s

    ucce

    ss (%

    )

    Mayotte MauritiusLa RuniMadagasMadagasMadagasMadagas

    Fig. 8. (a) Arrival success for each site in 19882001 (left) and migration duration (days)that the spawning area is at B1 (see Fig. 7). (b) same for 20002006.We test the arrival success and migration duration from twoother spawning areas within the convergence zone, to the recruit-ment sites: spawning area B3 is located in the westernmost part ofthe convergence zone while spawning area B2 is located in be-tween B1 and B3 (Fig. 7). In both cases (spawning area in B2 orB3), there are 70% of particles released that recruit to all sites onaverage over years 19882001, which is higher than for a spawn-ing area in B1 (where only 35% of released particles recruit on aver-age over the same time period).

    on car NWcar NEcar Ecar SE

    08/11 09/05 09/30 10/25 11/19 12/14

    60

    80

    100

    120

    140

    160

    180

    Mig

    ratio

    n du

    ratio

    n (d

    ay)

    Hatching date

    (a)

    on car NWcar NEcar Ecar SE

    08/11 09/05 09/30 10/25 11/19 12/14

    60

    80

    100

    120

    140

    160

    180

    Mig

    ratio

    n du

    ratio

    n (d

    ay)

    Hatching date

    (b)on average from 1988 to 2001 as a function of the spawning date (right), assuming

  • 08/10/04 09/04/04 09/29/04 10/24/04 11/18/04 12/13/046080

    100120140160180

    Mig

    ratio

    n du

    ratio

    n (d

    ay)

    Hatching date

    08/11/05 09/05/05 09/30/05 10/25/05 11/19/05 12/14/056080

    100120140160180

    Mig

    ratio

    n du

    ratio

    n (d

    ay)

    Hatching date

    igration duration (days) as a function of the spawning date (right), for two consecutive

    2001 for experiments FQN1, FQN2 and FQN3 and over 20002006 for experiments FQN4tment site for experiments FQN4 for years 2004 and 2005 when observations are available.

    uration at recruitment sites (mean/minima)

    Madagascar Mauritius La Runion

    NW NE E SE

    eanography 86 (2010) 39641308/10/04 09/04/04 09/29/04 10/24/04 11/18/04 12/13/040

    5

    10

    15

    20

    25Ar

    rival

    suc

    cess

    (%)

    Hatching date

    08/11/05 09/05/05 09/30/05 10/25/05 11/19/05 12/14/050

    5

    10

    15

    20

    25

    Arriv

    al s

    ucce

    ss (%

    )

    Hatching date

    Mayotte Mauritius La Runion Madagascar NWMadagascar NEMadagascar EMadagascar SE

    Fig. 9. Arrival success for each site as a function of the spawning date (left), and myears: 2004 (top) and 2005 (bottom) when observations are available.

    Table 3Mean and minimum migration duration at each recruitment site averaged over 1988and FQN5. We also indicate the mean and minimummigration durations at each recrui

    Experiment Spawning site Periods Depth Migration d

    Mayotte

    408 S. Pous et al. / Progress in OcFor a spawning area located in B2, arrival success to the recruit-ment sites of Madagascar and La Runion is generally higher thanfor a spawning area in B1 (Fig. 11a, left). This is consistent withqualitative backward experiments, as B2 is located closer to LaRunion and Madagascar. Actually, the North-East site of Madagas-car recruits most of the particles released in B2 (42% of all particlesthat recruit on average over years 19882001). Arrival success inMayotte is lower for a spawning area located in B2 than for aspawning area located in B1, which directly feeds the recruitmentsite via the northern route. Arrival success is the poorest at Mauri-tius, where it averages to 3% and is larger than 1% only in 1988,1992 and 1997. This is consistent with the circulation in the regionas currents passing through B2 are mostly westward, hence theydo not directly reach Mauritius (Fig. 2). Migration durations for aspawning area in B2 towards all recruitment sites are, on average,1030 days shorter than for a spawning area located in B1, exceptfor Mayotte where it is 13 days longer (see Table 3). As for aspawning area in B1, migration durations depend on the spawningdate and do not show any intra-seasonal modulation whatever therecruitment site considered (Fig. 11a, right). Again, there is largeintra-seasonal and interannual variability in both arrival successand migration durations (not shown).

    For a spawning area located in B3, arrival success to Mayotte islower than for a spawning area located in B2, while the East andSouth-East recruitment sites of Madagascar receive most of the re-leased particles (65% of all particles that recruit on average overyears 19882001, Fig. 11b, left). Arrival success at La Runionand Mauritius are higher for a spawning area in B3 compared toB2. Migration durations from a spawning area located in B3 to-wards all recruitment sites of Madagascar and Mayotte are 1030 days shorter than for a spawning area located in B2 (Fig. 11b,

    FQN1 B1 19882001 0100 m 118/61 143/86 122/56 150/96 156/110 112/49 140/80FQN2 B2 19882001 0100 m 131/70 134/81 113/58 120/68 137/86 102/46 121/63FQN3 B3 19882001 0100 m 121/76 110/70 89/49 91/42 112/60 100/45 84/24FQN4 B1 20002006 0100 m 103/50 136/70 103/40 146/88 151/103 107/43 143/76FQN4 B1 2004 0100 m 91/47 128/70 93/42 163/106 165/133 108/51 137/82FQN4 B1 2005 0100 m 109/58 144/77 112/53 159/88 166/102 97/53 140/80

    1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 20083

    2

    1

    0

    1

    2

    3

    (a)

    Nor

    mal

    ized

    inde

    x (n

    o un

    it)

    IODSIOD

    1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 20083

    2

    1

    0

    1

    2

    3

    Nor

    mal

    ized

    mea

    n m

    igra

    tion

    dura

    tion

    (no

    unit)

    (b)

    Mayotte Mauritius La Runion Madagascar NWMadagascar NEMadagascar EMadagascar SE

    Fig. 10. Normalized Indian Ocean Dipole (IOD) and Southern Indian Ocean (SIOD)indices (a, no unit) and normalized mean migration duration to each recruitmentsite assuming that the spawning area is at B1 (b, no unit) from 1988 to 2006.

  • yotte uritius Runidagasdagasdagasdagas

    yotte uritius Runidagasdagasdagasdagas

    ys)

    eanoright). Migration duration from spawning area in B3 to Mauritius issimilar to that from spawning area in B2, and averages to 100 days.Most interestingly, migration duration from spawning area in B3 toLa Runion is 84 days on average, hence shorter than migration

    1988 1990 1992 1994 1996 1998 2000

    100

    80

    60

    40

    20

    0

    Arriv

    al s

    ucce

    ss (%

    )MaMaLaMaMaMaMa

    1988 1990 1992 1994 1996 1998 2000

    100

    80

    60

    40

    20

    0

    Arriv

    al s

    ucce

    ss (%

    )

    MaMaLaMaMaMaMa

    Fig. 11. Arrival success for each site in 19882001 (left), and migration duration (dathat the spawning area is: (a) at B2 and (b) at B3 (see Fig. 7).

    S. Pous et al. / Progress in Ocduration to Mauritius, consistent with observed data, which isnot the case for spawning areas in B1 nor B2 (see Table 3). Besides,migration durations increase for recruitment sites from the North-East to the South-East Madagascar, as for a spawning area in B1and B2. As for the other spawning areas, there is substantial in-tra-seasonal to interannual variability in both arrival success andmigration durations (not shown).

    4. Discussion

    4.1. Sensitivity to hydrodynamic model and biologicalparameterisations

    Two simulations of the hydrodynamical model (DRAK01 andDRAK02) have been used to reproduce the trajectories of passiveparticles in the SWIO from the possible spawning areas to recruit-ment sites, knowing that these two simulations differ only in theturbulent uxes at the surface. Forward quantitative simulationFQN4, based on DRAK02, gives migration durations similar to thosein FQN1 (based on DRAK01), the difference being of the order of 020 days. This may be interpreted as the standard deviation of themigration durations as a function of the hydrodynamic forcings.As noted above, migration duration from spawning area at B1 torecruitment sites, averaged over years 20002006, is 419 daysshorter than those averaged over 19882001 (excepted for La Ru-nion). This should not be interpreted as decadal variability of thelarval dispersion, as the difference is of the order of the standarddeviation of the migration durations related to the hydrodynamicmodel.

    In all experiments discussed above, larvae are considered aspassive particles, transported only by the advective currents onthe horizontal as well as on the vertical, although eel larvae areknown to be capable of vertical migration. Field studies have re-ported evidences of diel vertical migration, for Japanese and Atlan-tic eel leptocephali, from the upper ocean at night to depths from150 to 300 m (respectively for Pacic and Atlantic eel leptocephali)

    on car NWcar NEcar Ecar SE

    08/11 09/05 09/30 10/25 11/19 12/14

    60

    80

    100

    120

    140

    160

    180

    Mig

    ratio

    n du

    ratio

    n (d

    ay)

    Hatching date

    (a)

    on car NWcar NEcar Ecar SE

    08/11 09/05 09/30 10/25 11/19 12/14

    60

    80

    100

    120

    140

    160

    180

    Mig

    ratio

    n du

    ratio

    n (d

    ay)

    Hatching date

    (b)

    on average from 1988 to 2001 as a function of the spawning date (right), assuming

    graphy 86 (2010) 396413 409during the day (Castonguay and McCleave, 1987 for Atlantic eel,Otake et al., 1998; Kimura et al., 2001 for Japanese eel). The variousprocesses that contribute to the control of the vertical distributionof eel leptocephali, depend on the larval stage (Castonguay andMcCleave, 1987 for Atlantic eels, Otake et al., 1998; Tsukamoto,2009; Yamada et al., 2009 for Japanese eels); they may facilitatethe larvae survival and growth and nally help to ensure a success-ful migration. We test this hypothesis by running additional exper-iments with a parameterisation of the larvae vertical behavior(FQL6-7 compared to FQN4-5 for years 20052006, see Table 2).For a spawning area located in B3, there are indeed more particlesreleased that recruit to all sites. Arrival success to all recruitmentsites is very similar to the reference experiments with no verticalbehavior, except for Mauritius where arrival success decreasesfrom 2% to 0% for the 2 years of the sensitivity experiment (spawn-ing in 2005 and 2006). The difference in migration duration (0 to20 days) is within the range of variability related to the hydrody-namic conditions. The effect of including the larvae vertical behav-ior is different for a spawning area located in B1: there are lessparticles released that recruit to all sites. The arrival success de-creases (eventually to 0) for all recruitment sites but Mayotte, be-cause particles following the northern route cannot pursue theirmigration southward at depth along the coast of Madagascar, pre-sumably due to a vertical shear in ocean coastal currents. Thisactually questions the realism of imposing vertical migrationsevery 12 h even in the presence of a vertical shear in the ocean cur-rents considering the present resolution of the model. The meanmigration durations increase by 1545 days, presumably becausemore particles follow the southern route that is longer, exceptfor Mayotte where the mean migration duration remains the same.Finally, it is important to keep in mind that there are no observa-tions of the behavior of eel larvae in the Indian Ocean, hence little

  • eanois known about how this behavior depends on the hydrodynamicconditions, which is required to assert the realism of the parame-terisation that we tested here. Also the absence of continentalslope in the vicinity of volcanic islands in the Indian Ocean mayinuence the larval behavior at the time of recruitment, but noinformation is available on this phenomenon.

    The results that we present in this study do not include a real-istic larval mortality. We ran additional experiment FQN8 that in-clude a more realistic parameterisation of mortality, with constantmortality rateM = 3.8 yr1, which is the most realistic value for eelleptocephalus in Atlantic Ocean, assessed from a combination ofLagrangian simulations and eel ecology parameters (Bonhommeauet al., 2009b). When compared to the similar experiment with norealistic mortality (FQN4), there are less particles released that re-cruit to all sites (15% in FQN8 and 43% in FQN4), but the propor-tions in which successful particles share out between thedifferent recruitment sites is very similar. The mean migrationduration to all sites over years 20002006 in FQN8 is of the orderof 10 days smaller than in FQN4, while the minimum migrationdurations remain the same. This value of 3.8 yr1 belongs to thevery low mortality rates among sh species but in regards to thevery long leptocephalus migration duration in Atlantic Ocean (i.e.21 months), it is the only way to ensure arrival of enough juvenilesin their recruitment sites (Bonhommeau et al., 2009b). In the In-dian Ocean, eel larval duration is shorter (i.e. maximum of6 months), so that the mortality rate could be stronger, withouthaving such drastic consequences on larvae arrival success. Inaddition, the water temperature in the Indian Ocean is warmerthan in Atlantic Ocean, presumably leading to a faster metabolismfor leptocephali of the different eel species, hence a stronger mor-tality. Indeed, it has been shown that the mortality rate is stronglyrelated to temperature, among other environmental parameters(Pepin, 1991; Wegner et al., 2003), and increases as water temper-ature increases (Houde and Zastrow, 1993). Besides, mortality rateis likely to be different between the different eel species of the In-dian and Atlantic Oceans. For example, among the tuna of thegenus Thunnus, instantaneous larval mortality rate of Atlantic Blue-n tuna (Thunnus thynnus) has been assessed to 0.2 d1 (Scott et al.,1993), while it ranges between 0.16 and 0.41 d1 for Yellowntuna (Thunnus albacares) (Lang et al., 1994) and between 0.5 and0.66 d1 for Southern Bluen tuna (Thunnus maccoyi) (Daviset al., 1991). As a result, we also tested a mortality rates of7.6 yr1 (experiment FQN9). There are only 6% of the released par-ticles that recruit to all sites. The mean migration durations to allsites decreases further compared to the experiment with no realis-tic mortality rate, but the trends among recruitment sites still re-mains the same, hence our results are still valid. For highermortality rates, there are years when arrival success decreases to0 for some recruitment sites, which affects our results, but maxi-mum migration durations are not consistent with observationsany more (less than 143 days for MP 19 yr1).

    4.2. Interpretation of our results in the context of biological knowledge

    Following our simulations results, a common spawning area canbe dened, boundaries of which are variable from one year to an-other, but which remains centered near the central region of theMascarene Ridge (i.e. located westwards of 60.5E, while northernand southern boundaries oscillated respectively around 13S and19S) (Fig. 5). This nding is consistent with Jespersens hypothesis(1942). Recent studies have shown that for Atlantic (resp. Pacic)eel species, thermal (resp. salinity) fronts are supposed to be iden-tied as landmarks by the eels to locate the spawning areas (McC-

    410 S. Pous et al. / Progress in Ocleave, 1993; Friedland et al., 2007 for the Atlantic and Tsukamoto,1992; Kimura et al., 2001 for the Pacic). In the SWIO, there is nothermal front associated to the SEC (Fig. 3a). On the other hand,and as discussed above, two salinity fronts are associated withTSW that is advected westwards by the SEC, both in observations(Antonov et al., 2006) and in the hydrodynamic model. Duringthe spawning period (AugustDecember), the TSW vein is narrow,and the associated salinity fronts are located near 12S at 60E, andnear 19S at 56E (Fig. 12). These two fronts are located North andSouth of the common spawning area dened here. It is importantto note, though, that there is substantial interannual variabilityin the location of these salinity fronts. Besides, the presence ofthe Mascarene Ridge may act as a geographic or topographic land-mark for the eels in the SWIO.

    The hypothesis of a single spawning area in the SWIO for A.mossambica is highly probable, as this species is endemic of the re-gion (observed only on the eastern coasts of Africa: from Kenya toSouth Africa, in Madagascar, Maurice and La Runion, Fishbasedata: www.shbase.org, version (09/2009), Froese and Pauly,2009). The hypothesis must be discussed for A. marmorata as thisspecies is widely distributed through the Indo-Pacic region, fromthe western Indian Ocean to the North Pacic and South Pacicoceans (Ege, 1939). Such a pandemic species is likely to have sep-arate spawning areas in the different regions of the Indian Ocean,since tropical eels are characterized by shorter larval migrationduration than temperate species (Kuroki et al., 2006). For the In-dian Ocean, Jespersen (1942) suggested the existence of at leasttwo spawning areas located respectively Southwest of Sumatrafor the leptocephali captured in the eastern Indian Ocean, and Eastof Madagascar for the one caught in the western Indian Ocean. Re-cent genetic studies on this species (Gagnaire et al., 2009) conrmthe existence of two populations, leading to the hypothesis of twogeographically and/or temporally separated spawning areas be-tween SWIO and Sumatra populations. Hence it is conceivable thatthe common spawning area described in our modeling results in-clude the SWIO spawning area for A. marmorata. Beside, Minegishiet al. (2008) and Gagnaire et al. (2009) observed genetic similari-ties between SWIO and Sumatran populations. In particular,Gagnaire et al. (2009) observed cytonuclear desequilibrium onindividuals fromMauritius and La Runion suggesting the possibil-ity of larval exchanges from Sumatra to these easternmost islandsof SWIO. Backward simulations from the SWIO have been run totest this hypothesis (not shown). It appears that even with a larvalmigration duration as long as 365 days (which exceeds observa-tions), passive particles do not reach Indonesia. Such transoceanicmigrations, if they exist, are necessarily rare and depend on verystrong meteorological events, which are not taken into accountin our model, due to the 5-days averaged hydrodynamical data.

    We explore the impact of hydrodynamic conditions on recruit-ment and migration duration from the common spawning area tothe recruitment sites. Because the size and boundaries of the com-mon spawning area change with time, we test three potentialspawning areas (B1, B2 and B3), xed in time, located within theaverage common spawning area. There are remarkable differencesin larval supply and migration durations to the different recruit-ment sites depending on the spawning area. Indeed, our simula-tions highlight that some years, the potential spawning areas B1and B2 do not provide larval supply with realistic migration dura-tions to all recruitment sites. Nevertheless, the observed distribu-tion of eel communities in estuaries and rivers of the SWIO iscontrasted between the different recruitment sites, with A. bicolorbicolor and A. nebulosa labiata being dominant grossly North ofMadagascar whereas A. marmorata and A. mossambica dominatein the South (Watanabe, 2003; Robinet et al., 2007 and 2008). Thissuggests that different spawning areas may be linked to differenteel species. For example, as B1 seems to favor larval supply on

    graphy 86 (2010) 396413Mayotte and North-East of Madagascar (Fig. 8) and A. bicolor bicoloris dominant in the North of Madagascar (Watanabe, 2003; Robinetet al., 2007, 2008), it is conceivable that its spawning area is

  • 34

    t) d

    eanolocated near B1. At the opposite, the absence of A. marmorata inSeychelles Islands strongly suggests that the spawning area for thisspecies is not located in the vicinity of B1, and may be in moresouthern areas than A. bicolor bicolor.

    On the other hand, A. marmorata and A. mossambica recruit in thesame rivers and estuaries in the southern region of SWIO. Indeed,during the Dana expedition in the SWIO (December 1929January1930), Jespersen (1942) captured leptocephali of A. marmorata andA. mossambica in the same region (i.e. North and West-Southwestof Madagascar), but with a difference in their development stages:leptocephali of A. mossambica were considerably more advancedthan those ofA.marmorata.More recently, in LaRunion Island, Rob-inet et al. (2003) observed an earlier recruitment for A. mossambicathan for A. marmorata, with the same time lag in their hatching per-iod (deduced from otoliths analyses), suggesting a 2 months differ-ence in their respective spawning dates. The possibility of acommon spawning area with a temporal difference in the spawningperiods for A. marmorata and A. mossambica is conceivable, as oursimulations highlight the great variability in larvalmigration trajec-tories according to the release period. This temporal variability inthe migration routes may indeed explain the observed differencesin the proportion of the two species, in La Runion Island in 20002001, as it was suggested by Robinet et al. (2003). However our re-sults do not exclude the possibility of a separate spawning area inmore southwestern zones. Indeed, this ts with the shorter migra-tion durations observed in this species (Robinet et al., 2008; Rveil-lac et al., 2009) than in A. marmorata.

    35oE 40oE 45oE 50oE 55oE 60oE 65oE

    5oS

    10oS

    15oS

    20oS

    25oS

    Fig. 12. Mean salinity at the surface (left) and along a meridional section at 58E (righDRAK01 simulation. Gray lines indicate isobaths 1000 m and 2000 m.

    S. Pous et al. / Progress in OcIn Madagascar, Rveillac et al. (2009) observed that larval dis-persal durations for A. mossambica increase from north to south onthe eastern coast. The same gradient in larval dispersal durationsis observed in our simulations covering years 19882001, whateverthe spawning area (B1, B2, and B3). Indeed, for a spawning area in B1(resp B2 and B3) simulatedmigration durations increase on averagefrom 150 to 156 days (resp, from 120 to 137, and from 91 to112 days) between the recruitment sites located East and South-East of the Malagasy eastern coast. This modeling result suggeststhat this gradient in larval dispersal duration is only related to thehydrodynamical conditions. The migration duration is related tothe length of themigratory pathway along the eastern coast ofMad-agascar, as the current ows southwards along this coast.

    From ageing studies, it has been observed that larval migrationduration is shorter to reach La Runion than Mauritius (Table 1).Among the three potential spawning areas tested within this study,only B3 allows such pattern, with a mean migration duration of84 days towards La Runion, vs. 100 days to reach Mauritius. How-ever, to conrm this observation, more biological data are needed.Indeed, this assumption lays on few individuals observed on differ-ent years c.a. for A. bicolor bicolor, 11 individuals were observed in2000 in La Runion, leading to a mean migration duration of 46.2(5.8) days, and 1 individual in 2005 with a migration duration of151 days. These data were compared to nine individuals capturedin Mauritius in 2005, with a mean migration duration of 110.6(16.7) days. This assumption ismore convincingwhen comparisonis made on A. marmorata on 2005: on 15 individuals captured in LaRunion, the mean migration duration reached 111.0 (15.8) days,vs. 139.0 (24.0) days for 30 individuals captured in Mauritius.

    The present study shows that larval migration duration de-pends on the spawning date and the inherent hydrodynamic con-ditions. Such a larval migration duration plasticity has alreadybeen reported by Rveillac et al. (2008) on A. marmorata in differ-ent recruitment sites of the SWIO, and also between eel species fora same recruitment site in La Runion (Robinet et al., 2003). Thisplasticity has been explained by Rveillac et al. (2008) by boththe individuals intrinsic metabolism and the environmental condi-tions such as sea temperature and food availability. In our model,as larvae are considered as passive particles, individual variabilityis not taken into account. Still, variability in larval migration dura-tion has been highlighted, which is only linked to the hydrodynam-ical variability and the different migratory pathways followed bylarvae. Besides, our results clearly show that migration durationas a function of the spawning date is very noisy, and does not ex-hibit a clear intra-seasonal signal (Fig. 8a and b, right), suggestingthat there is no best spawning date for each spawning area to eachrecruitment site. This is consistent with the non signicant sea-sonal variation of larval duration observed in A. mossambica that.5 34.6 34.7 34.8 34.9 35.0 35.1 35.2 35.3 35.4 35.5 35.6 35.7

    25oS 20oS 15oS 10oS 5oS

    Dep

    th (m

    )

    0

    100

    200

    300

    400

    uring the spawning period (August to December) from 1988 to 2002, calculated from

    graphy 86 (2010) 396413 411recruited in Madagascars East coast in 2005 (Rveillac et al., 2009).

    5. Conclusion

    As the eel recruitment in the Atlantic and Pacic has beendeclining in the recent decades, eels stocks in the SWIO are facinga growing interest from international sheries. Although severalstudies described the eel community distribution, and despiteoceanographic cruises aiming at sampling eel larvae, little isknown about the eel spawning area in the SWIO. The objective ofthis study is to investigate the eel spawning area in the SWIO usinga modeling approach that simulates migration trajectories consis-tent with available observations.

    Preliminary experiments, which calculate trajectories from therecruitment sites of the SWIO backward in time, suggest that theremay be a common spawning area for all recruitment sites, roughlybounded between 13S and 19S and located westwards of 60.5E,which yields migration durations to the recruitment sites roughlyconsistent with observations. This area is bounded by salinityfronts that may serve as cue for adult eels to locate the spawning

  • glass eels that recruit, but observations are insufcient to test this

    tion durations are different when taking into account diel vertical

    eanomigration of the eel larvae. Still, observations are required to assertthe realism of the vertical behavior that was tested here. Also,implementing a more realistic larvae behavior would require ahydrodynamical model that reproduces more realistically the cir-culation in the coastal regions.

    The important highlight of this study is the substantial variabil-ity on intra-seasonal to interannual timescale in simulated migra-tion durations and arrival success, with amplitude specic to eachrecruitment site. There is no best spawning date for each spawningarea to each recruitment site that would be common to all years.Migration pathways to the recruitment sites are inuenced by boththe IOD and the SIOD that are climatic phenomena independent ofeach other, which seem to induce a complex interannual variabilityin arrival success and migration durations. Such variability cannotbe validated yet because observations are too sparse. To concludeon the most probable location of the spawning area, more oceano-graphic surveys and estuarine samplings are needed on all therecruitment sites, to assess (1) the spatial distribution of eel larvaein the SWIO and (2) the temporal variability of larval supply to thedifferent recruitment sites. A spatio-temporal monitoring of therecruitment is highly recommended.

    Acknowledgments

    We thank the Drakkar group for providing the hydrodynamicaldata. The Drakkar model has been run at the IDRIS computing cen-tre of CNRS, Orsay, France. We also thank B. Blanke and N. Grimafor helping us with Lagrangian tools. Fig. 3 was created usingODV software (Schlitzer, R., Ocean Data View, http://odv.awi.de,2010). We thank E. Rveillac and T. Robinet for their work on oto-lith microstructure analyses. We appreciate the detailed commentsof Sylvie Dufour in helping us to improve the manuscript. Wethank the two anonymous reviewers for their very constructivecomments.

    References

    Aarestrup, K., Okland, F., Hansen, M., Righton, D., Gargan, P., Castonguay, M.,Bernatchez, L., Howey, P., Sparholt, H., Pedersen, M., McKinley, R., 2009. Oceanicspawning migration of the European eel (Anguilla anguilla). Science 325, 1660.

    Arai, T., Aoyama, J., Ishikawa, S., Miller, M.J., Otake, T., Inagaki, T., Tsukamoto, K.,2001. Early life history of tropical Anguilla leptocephali in the western PacicOcean. Marine Biology 138, 887895.hypothesis.We explore the impact of hydrodynamic conditions on recruit-

    ment and migration duration from three specic regions of thecommon spawning area, located on the Mascarene ridge and Northof Mauritius and La Runion. The migration duration and arrivalsuccess to each recruitment site depends on the region that weconsider. The spawning area located North of La Runion (B3)seems to yield migration durations to all recruitment sites thatt the best with observed migration durations. Nevertheless,observations suggest that there may be different spawning areasdepending on the eel species considered, which is not excludedby the simulations, as long as they are within the large commonspawning area that we dene. Besides, arrival success and migra-area. Interannual simulations show that the common spawningarea has varying size and boundaries, but always exists from1988 to 2006. We describe several migration pathways from thespawning area to each recruitment site of the SWIO. Note that thiscould imply multi-modal distribution of some characteristics (e.g.migration duration, age at recruitment, spawning period) of the

    412 S. Pous et al. / Progress in OcAntonov, J.I., Locarnini, R.A., Boyer, T.P., Mishonov, A.V., Garcia, H.E., 2006. WorldOcean Atlas 2005, Salinity, vol. 2. In: Levitus, S. (Ed.), NOAA Atlas NESDIS 62, USGovernment Printing Ofce, Washington, DC (182 pp).Barnay, A.-S., Ellien, C., Gentil, F., Thibaut, E., 2003. A model study on variations inlarval supply: are populations of the polychaete Owenia fusiformis in the Englishchannel open or closed? Helgoland Marine Research 56, 229237.

    Barnier, B., Madec, G., Penduff, T., Molines, J.M., Trguier, A.-M., Le Sommer, J.,Beckmann, A., Biastoch, A., Boening, C., Dengg, J., Derval, C., Durand, E., Gulev, S.,Remy, E., Talandier, C., Theetten, S., Maltrud, M., McClean, J., DeCuevas, B., 2006.Impact of partial steps and momentum advection schemes in a the global oceancirculation model at eddy permitting resolution. Ocean Dynamics 56 (56),543567.

    Behera, S.K., Yamagata, T., 2001. Subtropical SST dipole events in the southernIndian Ocean. Geophysical Research Letters 28, 327330.

    Blanke, B., Raynaud, S., 1997. Kinematics of the Pacic equatorial undercurrent: aEulerian and Lagrangian approach from GCM results. Journal of PhysicalOceanography 27, 10381053.

    Bonhommeau, S., Chassot, E., Rivot, E., 2008. Fluctuations in European eel (Anguillaanguilla) recruitment resulting from environmental changes in the Sargasso Sea.Fisheries Oceanography 17 (1), 3244.

    Bonhommeau, S., Blanke, B., Trguier, A.-M., Grima, N., Rivot, E., Vermard, Y.,Greiner, E., Le Pape, O., 2009a. How fast can the European eel (Anguilla anguilla)larvae cross the Atlantic Ocean? Fisheries Oceanography 18 (6), 371385.

    Bonhommeau, S., Le Pape, O., Trguier, A.-M., Blanke, B., Grima, N., Gascuel, D.,Vermard, Y., Rivot, E., 2009b. Estimates of the mortality and the duration of thetrans-Atlantic migration of European eel leptocephali using a particle trackingmodel. Journal of Fish Biology 74, 18911914.

    Brodeau, L., Barnier, B., Penduff, T., Trguier, A.-M., Gulev, S., 2010. An ERA40 basedatmospheric forcing for global ocean circulation models. Ocean Modelling 31,88104.

    Castonguay, M., McCleave, J.D., 1987. Vertical distributions, diel and ontogeneticvertical migrations and net avoidance of leptocephali of Anguilla and othercommon species in the Sargasso Sea. Journal of Plankton Research 9, 195214.

    Castonguay, M., Hodson, P., Couillard, C., Eckersley, M., Dutil, J., Verreault, G., 1994.Why is recruitment of American eel, Anguilla rostrata, declining in the St-Lawrence-River and Gulf. Canadian Journal of Fisheries and Aquatic Sciences 51,479488.

    Dekker, W., 2003. Did lack of spawners cause the collapse of the European eel,Anguilla anguilla? Fisheries Management and Ecology 10, 365376.

    Dsaunay, Y., Gurault, D., 1997. Seasonal and long-term changes in biometrics ofeel larvae: a possible relationship between recruitment variation and NorthAtlantic ecosystem productivity. Journal of Fish Biology 51, 317339.

    Ds, K., 1995. Inter-ocean exchange of water masses. Journal of GeophysicalResearch, 100(C7), 13, 499514.

    Drakkar group, 2007. Eddy-permitting Ocean circulation hindcasts of past decades.Clivar Exchanges No 42, 12(3), 810.

    Ege, V., 1939. A revision of the genus Anguilla Shaw: a systematic, phylogenetic andgeographical study. Dana Report. Carlsberg Foundation 16, 8256.

    Ellien, C., Thibaut, E., Dumas, F., Salomon, J.-C., Nival, P., 2004. A modelling study ofthe respective role of hydrodynamic processes and larval mortality on larvaldispersal and recruitment of benthic invertebrates: example of Pectinaria koreni(Annelida: Polychaeta) in the Bay of Seine (English Channel). Journal ofPlankton Research 26 (2), 117132.

    Feunteun, E., 2002. Management and restoration of European eel population(Anguilla anguilla): an impossible bargain. Ecological Engineering 18, 575591.

    Friedland, K.D., Miller, M.J., Knights, B., 2007. Oceanic changes in the Sargasso Seaand declines in recruitment of the European eel. ICES Journal of Marine Science64, 519530.

    Froese, R., Pauly, D. (Eds.), 2009. FishBase version 09/2009. Online Dataset, WorldWide Web Electronic Publication .

    Gagnaire, P.A., Minegishi, Y., Aoyama, J., Rveillac, E., Robinet, T., Bosc, P.,Tsukamoto, K., Feunteun, E., Berrebi, P., 2009. Ocean currents drive secondarycontact between Anguilla marmorata populations in the Indian Ocean. MarineEcology Progress Series 379, 267278.

    Hermes, J.C., Reason, C.J.C., 2005. Ocean model diagnosis of interannual coevolvingSST variability in the South Indian and South Atlantic Oceans. Journal of Climate18, 28642882.

    Houde, E.D., Zastrow, C.E., 1993. Ecosystem and taxon specic dynamic andenergetics properties of larval sh assemblages. Bulletin of Marine Sciences 53,290335.

    Jackson, P.B.N., 1976. The distribution and growth of eels of the genus AnguillaShaw, 1830, in the Western Indian Ocean. In: Guz, P. (Ed.) Biologie Marine etExploitation Des Ressources de lOcan Indien Occidental, vol. 47. Paris:ORSTOM, p. 189192 (371 p).

    Jellyman, D., 1997. Variability in growth rates of freshwater eels (Anguilla, spp.) inNew Zealand. Ecology of Freshwater Fish 6 (2), 108115.

    Jellyman, D., Tsukamoto, K., 2005. Swimming depths of offshore migrating longneels Anguilla dieffenbachia at sea. Marine Ecology Progress Series 286, 261267.

    Jespersen, P., 1942. Indo-Pacic Leptocephalids of the Genus Anguilla: Systematicand Biological Studies. The Carlsberg Foundations Oceanographical ExpeditionRound the World 19281930 and Previous Dana Expedition; Dana Report N22,88105.

    Kettle, A.J., Haines, K., 2006. How does the European eel (Anguilla anguilla) retain itspopulation structure during its larval migration across the North AtlanticOcean? Canadian Journal of Fisheries and Aquatic Sciences 63, 90106.doi:10.1139/F05-198.

    graphy 86 (2010) 396413Kim, H., Kimura, S., Shinoda, A., Kitagawa, T., Sasai, Y., Sasaki, H., 2007. Effect of ElNio on migration and larval transport of the Japanese eel (Anguilla japonica).ICES Journal of Marine Science 64, 13871395. doi:10.1093/icesjms/fsm091.

  • Kimura, S.K., Ds, K., Coward, A.C., 1999. Numerical simulation to resolve the issueof downstream migration of the Japanese eel. Marine Ecology Progress Series186, 303306.

    Kimura, S., Inoue, T., Sugimoto, T., 2001. Fluctuation in the distribution of low-salinity water in the north equatorial current and its effect on the larvaltransport of the Japanese eel. Fisheries Oceanography 10, 5160.

    Kimura, S., Tsukamoto, K., 2006. The salinity front in the north equatorial current: alandmark for the spawning migration of the Japanese eel (Anguilla japonica)

    Peterson, I., Wroblewski, J.S., 1984. Mortality rate of shes in the pelagic ecosystem.Canadian Journal of Fisheries and Aquatic Sciences 41, 11171120.

    Rveillac, E., Feunteun, E., Berrebi, P., Gagnaire, P.A., Lecomte-Finiger, R., Bosc, P.,Robinet, T., 2008. Anguilla marmorata larval migration plasticity as revealed byotolith microstructural analysis. Canadian Journal of Fisheries and AquaticSciences 65, 21272137. doi:10.1139/F08-122.

    Rveillac, E., Robinet, T., Rabenevanana, M.-W., Valade, P., Feunteun, E., 2009. Clues

    S. Pous et al. / Progress in Oceanography 86 (2010) 396413 413related to the stock recruitment. Deep Sea Research II 53, 315325.Kimura, Y., Miller, M.J., Minagawa, G., Watanabe, S., Shinoda, A., Aoyama, J., Inagaki,

    T., Tsukamoto, K., 2006. Evidence of a local spawning site of marine eels alongnortheastern Japan, based on the distribution of small leptocephali. FisheriesOceanography 15 (2), 183190.

    Kuroki, M., Aoyama, J., Miller, M.J., Wouthuyzen, S., Arai, T., Tsukamoto, K., 2006.Contrasting patterns of growth and migration of tropical anguillid leptocephaliin the western Pacic and Indonesian Seas. Marine Ecology Progress Series 309,233246.

    Kuroki, M., Aoyama, J., Wouthuyzen, S., Sumardhiharga, K., Miller, M.J., Tsukamoto,K., 2007. Age and growth of Anguilla bicolor bicolor leptocephali in the easternIndian Ocean. Journal of Fish Biology 70, 538550.

    Lang, K.L., Grimes, C.B., Shaw, R.F., 1994. Variations In the age and growth ofyellown tuna larvae Thunnus albacares, collected about Mississippi Riverplume. Environmental Biology of Fishes 39, 259270.

    Lecomte-Finiger, R., 1992. Growth history and age at recruitment of European glasseels Anguilla anguilla as revealed by otolith microstructure. Marine Biology 114,205210.

    Le Sommer, J., Penduff, T., Theetten, S., Madec, G., Barnier, B., 2009. Howmomentumadvection schemes inuence current-topography interactions at eddy-permitting resolution. Ocean Modelling 29, 114. doi:10.1016/j.ocemod.2008.11.007.

    Lumpkin, R., Garraffo, Z., 2005. Evaluating the decomposition of tropical Atlanticdrifter observations. Journal of Atmospheric and Oceanic Technology I 22,14031415.

    Madec, G., Imbard, M., 1996. A global ocean mesh to overcome the north polesingularity. Climate Dynamic 12, 381388.

    Madec, G., 2008. NEMO reference manual, ocean dynamic component: NEMO-OPA.Preliminary version, Tech. Rep. 27, Note du ple de modlisation, Institut PierreSimmon Laplace (IPSL), France, ISSN No 1288-1619.

    McCleave, J.D., Kleckner, R.C., Castonguay, M., 1987. Reproductive sympatry ofAmerican and European eels and implications for migration and taxonomy.American Fisheries Society Symposium 1, 286297.

    McCleave, J.D., 1993. Physical and behavioral controls on the oceanicdistribution and migration of leptocephali. Journal of Fish Biology 43 (A),243273.

    McCleave, J.D., 2008. Contrasts between spawning times of Anguilla speciesestimated from larval sampling at sea and from otolith analysis of recruitingglass eels. Marine Biology 155, 249262.

    Miller, M.J., McCleave, J.D., 1994. Species assemblages of leptocephali in thesubtropical convergence zone of the Sargasso Sea. Journal of Marine Research52, 743772.

    Minegishi, Y., Aoyama, J., Tsukamoto, K., 2008. Multiple population structure of thegiant mottled eel Anguilla marmorata. Molecular Ecology 17, 31093122.

    Moriarty, C., Dekker, W., 1997. Management of the European eel. Fishery Bulletin15, 1110.

    New, A.L., Alderson, S.G., Smeed, D.A., Stanseld, K.L., 2007. On the circulation ofwater masses across the Mascarene Plateau in the South Indian Ocean. Deep-Sea Research Part I 54, 4274.

    Otake, T., Inagaki, T., Hasumoto, H., Mochioka, N., Tsukamoto, K., 1998. Diel verticaldistribution of Anguilla japonica leptocephali. Ichthyological Research 45, 208211. doi:10.1007/BF02678565.

    Palastanga, V., van Leeuwen, P.J., de Ruijter, W.P.M., 2006. A link between low-frequency mesoscale eddy variability around Madagascar and the large-scaleIndian Ocean variability. Journal of Geophysical Research 111, C09029.doi:10.1029/2005JC003081.

    Palastanga, V., van Leeuwen, P.J., Schouten, M.W., de Ruijter, W.P.M., 2007. Flowstructure and variability in the subtropical Indian Ocean: instability of theSouth Indian Ocean countercurrent. Journal of Geophysical Research 112,C01001. doi:10.1029/2005JC003395.

    Penduff, T., Le Sommer, J., Barnier, B., Treguier, A.-M., Molines, J.M., Madec, G., 2007.Inuence of numerical schemes on current-topography interactions in 1=4 globalocean simulations. Ocean Science 3, 509524.

    Pepin, P., 1991. Effect of temperature and size on development, mortality andsurvival rates of the pelagic early life history stages of marine sh. CanadianJournal of Fisheries and Aquatic Sciences 48, 503518.to the location of the spawning area and larval migration characteristics ofAnguilla mossambica as inferred from otolith microstructural analyses. Journalof Fish Biology, 74(9), 18661877. doi:10.1111/j.1095-8649.2009.02285.x.

    Robinet, T., Lecomte-Finiger, R., Escoubeyrou, K., Feunteun, E., 2003. Tropical eelsAnguilla spp. Recruiting to Runion Island in the Indian Ocean: taxonomy,patterns of recruitment and early life histories. Marine Ecology Progress Series259, 263272.

    Robinet, T., Feunteun, E., Keith, P., Marquet, G., Olivier, J.-M., Rveillac, E., Valade, P.,2007. Eel community structure, uvial recruitment of Anguilla marmorata andindication for a weak local production of spawners from rivers of Runion andMauritius islands. Environmental Biology of Fish 78, 93105.

    Robinet, T., Rveillac, E., Kuroki, M., Aoyama, J., Tsukamoto, K., Rabenevanana, M.W.,Valade, P., Gagnaire, P.A., Feunteun, E., 2008. New clues for freshwater eels(Anguilla spp.) migration routes to eastern Madagascar and surrounding islands.Marine Biology, 154, 453463. doi:10.1007/s00227-008-0938-7.

    Saji, N.H., Goswami, B.M., Vinayachandran, P.N., 1999. A dipole mode in the tropicalIndian Ocean. Nature 401, 360363.

    Schmidt, J., 1923. The breeding places of the eel. Philosophical Transactions of theRoyal Society of London, Series B 211, 179208.

    Schott, F.A., Fieux, M., Kindle, J., Swallow, J., Zantopp, R., 1988. The boundarycurrents east and north of Madagascar 2. Direct measurements and modelcomparisons. Journal of Geophysical Research 93, 49634974.

    Schott, F.A., McCreary Jr., J.P., 2001. The monsoon circulation of the Indian Ocean.Progress in Oceanography 51 (1), 1123. doi:10.1016/S0079-6611(01)00083-0.

    Scott, G.P., Turner, S.C., Grimes, C.B., Richards, W.J., Brothers, E.B., 1993. Indices oflarval bluen tuna, Thunnus thynnus, abundance in the Gulf of Mexico:modelling variability in growth, mortality, and gear selectivity. Bulletin ofMarine Science 53, 912929.

    Stramma, L., Lutjeharms, J.R.E., 1997. The ow eld of the subtropical gyre of theSouth Indian Ocean. Journal of Geophysical Research 102, 55135530.

    Tesch, F.W., 1979. Tracking of silver eels (Anguilla anguilla) in different shelf areas ofthe Northeast Atlantic. Rapports et Procs-Verbaux des runions, ConseilPermanent pour lExploration de la Mer 174, 104114.

    Tesch, F.W., 2003. The Eel, fth ed. Blackwell Science, Oxford. 434 p.Tremblay, M.J., Loder, J.W., Werner, F.E., Naimie, C.E., Page, F.H., Sinclair, M.M., 1994.

    Drift of sea scallop larvae Placopecten magellanicus on Georges Bank: a modelstudy of the roles of mean advection, larval behavior and larval origin. Deep SeaResearch II 41, 749.

    Tsukamoto, K., 1992. Discovery of the spawning area for Japanese eel. Nature 356,789791.

    Tsukamoto, K., Aoyama, J., Miller, M.J., 2002. Migration, speciation and the evolutionof diadromy in anguillid eels. Canadian Journal of Fisheries and Aquatic Sciences59, 19891998.

    Tsukamoto, K., Otake, T., Mochioka, N., Lee, T.W., Fricke, H., Inagaki, T., Aoyama, J.,Ishikawa, S., Kimura, S., Miller, M.J., Hasumoto, H., Oya, M., Suzuki, Y., 2003.Seamounts, new moon and eel spawning: the search for the spawning site ofthe Japanese eel. Environmental Biology of Fishes 66, 221229.

    Tsukamoto, K., 2006. Spawning of eels near a seamount. Nature 439, 929.Tsukamoto, K., 2009. Oceanic migration and spawning of anguillid eels. Journal of

    Fish Biology 74, 18331852.Watanabe, S., 2003. Taxonomy of the freshwater eels, Genus Anguilla Schrank, 1798.

    In: Aida, K., Tsukamoto, K., Yamauchi, K. (Eds.), Eel Biology. Springer-Verlag,Tokyo, pp. 318.

    Webster, P.W., Moore, A.M., Loschnigg, J.P., Lebben, R.R., 1999. Coupled oceanatmosphere dynamics in the Indian Ocean during 19971998. Nature 401, 356360.

    Wegner, G., Damm, U., Purps, M., 2003. Physical inuences on the stock dynamics ofplaice and sole in the North Sea. Scientia Marina 67 (Suppl. 1), 219234.

    Wyrtki, K., 1971. Oceanographic Atlas of the International Indian Ocean Expedition.National Science Foundation, Washington, DC. p. 531.

    Yamada, Y., Okamura, A., Mikawa, N., Utoh, T., Horie, N., Tanaka, S., Miller, M.J.,Tsukamoto, K., 2009. Ontogenetic changes in phototactic behavior duringmetamorphosis of articially reared Japanese eel Anguilla japonica larvae.Marine Ecology Progress Series 379, 241251.

    Young, E.F., Bigg, G.R., Grant, A., Walker, P., Brown, J., 1998. A modeling study ofenvironmental inuences on bivalve settlement in the Wash, England. MarineEcology Progress Series 172, 197214.

    Investigation of tropical eel spawning area in the South-Western Indian Ocean: Influence of the oceanic circulationIntroductionData, models and methodsDataCirculation and hydrology in the SWIOBiological data

    ModelsThe hydrodynamic modelThe Lagrangian modelBiological parameterisations

    Simulation design

    ResultsBackward qualitative migration routes: observation of convergence areasBackward quantitative experiment: extension of the convergence area east of the Mascarene ridgeForward quantitative migration: arrival success for a spawning area on the Mascarene ridgeForward quantitative migration: arrival success for other spawning areas

    DiscussionSensitivity to hydrodynamic model and biological parameterisationsInterpretation of our results in the context of biological knowledge

    ConclusionAcknowledgmentsReferences

Recommended

View more >