Diverse Early Life-History Strategies in Migratory Amazonian Catfish

19 Pages·2015·3.97 MB·English

RESEARCHARTICLE
Diverse Early Life-History Strategies in
Migratory Amazonian Catfish: Implications
for Conservation and Management
JensC.Hegg1*,TommasoGiarrizzo2,BrianP.Kennedy1,3
1 DepartmentofFishandWildlifeSciences,UniversityofIdaho,Moscow,Idaho,UnitedStatesofAmerica,
2 LaboratóriodeBiologiaPesqueira—ManejodosRecursosAquáticos,UniversidadeFederaldoPará
(UFPA),Belém,PA,Brazil,3 DepartmentsofGeologicalSciencesandBiologicalSciences,Universityof
Idaho,Moscow,Idaho,UnitedStatesofAmerica
* [email protected]
Abstract
Animalmigrationsprovideimportantecologicalfunctionsandcanallowforincreasedbiodi-
versitythroughhabitatandnichediversification.However,aquaticmigrationsingeneral,
OPENACCESS andthoseoftheworld’slargestfishinparticular,areimperiledworldwideandareoften
poorlyunderstood.SeveralspeciesoflargeAmazoniancatfishcarryoutsomeofthelon-
Citation:HeggJC,GiarrizzoT,KennedyBP(2015)
gestfreshwaterfishmigrationsintheworld,travellingfromtheAmazonRiverestuarytothe
DiverseEarlyLife-HistoryStrategiesinMigratory
AmazonianCatfish:ImplicationsforConservationand Andesfoothills.Thesespeciesareimportantapexpredatorsinthemainstemriversofthe
Management.PLoSONE10(7):e0129697. AmazonBasinandmakeuptheregion’slargestfishery.Theyarealsotheonlyspeciesto
doi:10.1371/journal.pone.0129697
utilizetheentireAmazonBasintocompletetheirlifecycle.Studiesindicateboththatthe
AcademicEditor:MarcusSheaves,JamesCook fisheriesmaybedecliningduetooverfishing,andthattheproposedandcompleteddamsin
UniversityHospital,AUSTRALIA
theirupstreamrangethreatenspawningmigrations.Despitethis,surprisinglylittleisknown
Received:June24,2014 aboutthedetailsofthesespecies’migrations,ortheirlifehistory.Otolithmicrochemistry
Accepted:May12,2015 hasbeenaneffectivemethodforquantifyingandreconstructingfishmigrationsworldwide
acrossmultiplespatialscalesandmayprovideapowerfultooltounderstandthemove-
Published:July8,2015
mentsofAmazonianmigratorycatfish.Ourobjectivewastodescribethemigratorybehav-
Copyright:©2015Heggetal.Thisisanopen
iorsofthethreemostpopulousandcommerciallyimportantmigratorycatfishspecies,
accessarticledistributedunderthetermsofthe
CreativeCommonsAttributionLicense,whichpermits Dourada(Brachyplatystomarousseauxii),Piramutaba(Brachyplatystomavaillantii),andPir-
unrestricteduse,distribution,andreproductioninany aíba(Brachyplatystomafilamentosum).WecollectedfishfromthemouthoftheAmazon
medium,providedtheoriginalauthorandsourceare
RiverandtheCentralAmazonandusedstrontiumisotopesignatures(87Sr/86Sr)recorded
credited.
intheirotolithstodeterminethelocationofearlyrearingandsubsequent.Fishlocationwas
DataAvailabilityStatement:Allrelevantdataare
determinedthroughdiscriminantfunctionclassification,usingwaterchemistrydatafromthe
withinthepaperanditsSupportingInformationfiles.
literatureasatrainingset.Wherewaterchemistrydatawasunavailable,wesuccessfullyin
Funding:Giarrizzoreceivedaproductivitygrantfrom
predicted87Sr/86Srisotopevaluesusingaregression-basedapproachthatrelatedthegeol-
CNPq(process:308278/2012-7)andwasfundedby
CAPES(PNPDandPró-Amazônia:biodiversidadee ogyoftheupstreamwatershedtotheSrisotoperatio.Ourresultsprovidethefirstreported
sustentabilidade),whichwereusedforsample otolithmicrochemicalreconstructionofBrachyplatystomamigratorymovementsintheAma-
permittingandcollection.Allotherresearchactivities
zonBasin.Ourresultsindicatethatjuvenilesexhibitdiverserearingstrategies,rearingin
werecompletedwithoutspecificfunding.Thefunders
bothupstreamandestuaryenvironments.Thiscontrastswiththeprevailingunderstanding
hadnoroleinstudydesign,datacollectionand
analysis,decisiontopublish,orpreparationofthe thatjuvenilesrearintheestuarybeforemigratingupstream;however,itissupportedby
manuscript.
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 1/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
CompetingInterests:Theauthorshavedeclared somefisheriesdatathathasindicatedthepresenceofalternatespawningandrearinglife-
thatnocompetinginterestsexist. histories.Thepresenceofalternatejuvenilerearingstrategiesmayhaveimportantimplica-
tionsforconservationandmanagementofthefisheriesintheregion.
Introduction
Animalmigrationprovidesmanyimportantecologicalfunctions:theycanbeastabilizing
strategyinseasonalenvironments;offertransitoryhabitatsforlargepopulations;oftentrans-
portmaterialsacrossecosystemboundaries;andmayincreasearegionsbiodiversity[1].Large-
scalemigrationsshedlightonecosystemconnectivityacrossscalesandcanbeusedasalensto
understandbroaderbehavioralresponsestotheenvironmentandlinkstophysicalprocesses
[2–4].However,migrationsworldwideareunderthreatfromthealterationofmigratorypath-
ways,habitatloss,climaticchangesandanthropogenicchangestothelandscape[5].Inaquatic
systems,changesinupstreamlanduseandtheplacementofdamshavehadsignificantimpacts
onecosystemsandmigrationsworldwide[6–11].Thisisparticularlytrueforlargemigratory
fishthatareunderthreatinmanyoftheworld'slargestriversystems[12–14].Despitethis,
manylargemigratoryfishspeciesarenotwellunderstood[14].
Globally,damsandwaterresourceschallengesinthetwolargestriversinChinaprovidean
exampleoftheongoingchangestolargeriversandtheireffectsonaquaticspecies,including
sturgeonandpaddlefish[15].InSouthAmerica,transnationalriversystemsandalackofcoor-
dinatedresearchofaquaticsystemsmayresultinlossestounspecifiedlevelsofbiodiversity
[16–20].Newdamspresentauniquechallengetomigratoryfishintheregion.Becausethe
youngofmanyAmazonianspeciesundergoadriftinglarvalstage,evenifadultscanpassabove
damsthelackofflowinreservoirscreatesabarrierthatdriftingjuvenilesareunabletosur-
mountontheirwaydownstream[21].
SeveralspeciesofAmazoniancatfishinthegenusBrachyplatystomacarryoutsomeofthe
longestfreshwaterfishmigrationsintheworld,travellingover4,500kmfromrearingareasin
theAmazonestuarytospawninggroundsinriversinthefoothillsoftheAndes[22–24].These
specieslargelyinhabitwhitewaterandclearwaterrivers(riverclassificationsfromSioli[25])
withintheAmazonBasin[26],withrarereportsintannicblackwaterrivers[23].Thesecatfish
speciesaretheonlyknownorganisms,terrestrialoraquatic,thatrequiretheentirelengthof
theAmazonbasintocompletetheirlifecycle[23].Theyarealsooneofthefewapexpredators
inthepelagicanddemersalzonesofthelargestAmazonianrivers,playinganimportantrolein
trophicdynamicsandecosystemfunctioningwithintheentirebasin[27].However,evidence
indicatesthatthefisheriesforthemostpopulousspeciesareindecline,potentiallyduetoover-
fishing[22,28].Therelianceofthesespeciesonheadwaterstreamsforspawningleavesadults
andlarvavulnerabletoblockingoftheirmigrationpathsbydamsandtheirreservoirs[21,29].
SurprisinglylittleisknownaboutthelifehistoryofmigratoryAmazoniancatfishgiventhat
thethreemostabundantBrachyplatystomataxasupportthelargestfisheriesintheAmazon
Basin[23,24].Dourada(Brachyplatystomarousseauxii)isapelagicpredatorfoundthroughout
thewhitewaterandclearwaterriversoftheAmazonandsupportsthelargestfisheryinthe
Amazon[27,29].Piramutaba(Brachyplatystomavaillantii)makeupasecondlargeexportfish-
eryandarefoundalmostexclusivelyintheAmazonRivermainstem,whitewatertributaries,
andtheestuary[30–32].Piraíba(Brachyplatystomafilamentosum)isthelargestofthemigra-
torycatfish,presentinwhitewaterriversthroughouttheAmazonbasin.Itisalsothemost
locallyexploitedandleastunderstoodofthesethreespecies[22].Theexpansivescaleofthe
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 2/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
Fig1.WatersamplingpointsandgeologyoftheAmazonRiverbasin.Mapsshow(A)thelocationof87Sr/86SrwatersampleswithintheAmazonRiver
basindigitizedbytheauthorsfromlocationdescriptionsintheliterature[51–53]andpointspredictedfromEq1.Thegeologicalageandcompositionofthe
basin(B)usedtopredictthe87Sr/86Srsignaturesofunsampledwatershedsisalsoshown.MapscreatedusingUSGSdatasets[59–61].
doi:10.1371/journal.pone.0129697.g001
AmazonBasin,andthelargesizeoftheriversthesefishinhabit,havemadetrackingandrecon-
structingtheirmovementsverydifficult[32].Ourcurrentunderstandingofthemigratory
behaviorofmigratoryAmazoniancatfishisbasedonfishingrecords(includingthecatchtim-
ingandsizeoffishacrosstheAmazonbasin)andagrowingnumberofscientificsampling
efforts[23,28,31,33–36].Afterhatchingintheupperreachesofthewhitewaterriversoriginat-
ingintheAndes,larvaeofthesespeciesdriftdownstreamfortwotofourweeksbeforereaching
theAmazonestuary.Juvenilesrearintheestuarybeforecommencinganupstreammigration
thatcoincideswiththeseasonalfloodpulse.Geneticdataindicatethatdouradamayhometo
nataltributariesinthebasintospawn[37,38].
Otolithmicrochemistryhasbeenaneffectivemethodforquantifyingandreconstructing
fishmigrationsworldwideacrossmultiplespatialscales[39–46].Strontiumratioinparticular
hasbecomeapowerfultoolfordeterminingmovementandlocationbecauseitisnotfraction-
atedbiologically.Thus,thesignaturesrecordedinotolithsmatchthewaterthroughwhichfish
pass[43,47–50].StudiesofgeologicalweatheringthroughouttheAmazonbasinhaveprovided
detailed,multi-yearrecordsofmicro-chemicalandisotopicchemistryinthelargestriversof
thebasin.Thesedataprovidetherequiredbackgroundsamplingnecessarytotieregional
otolithsignaturestogeographiclocation[51–53](Fig1A,Table1).Recentstudieshavealso
shownthefeasibilityofpredicting87Sr/86Srsignaturesofunknownwatershedsusingthegeo-
logicmakeupofthebasin,allowingresearcherstocharacterizestrontiumsignaturesof
unsampledareas[54,55].Theseadvancespointtootolithmicrochemistryasapotentiallypow-
erfultooltounderstandthemovementsofAmazonianmigratorycatfish.
Ourobjectivewastodescribethemigratorybehaviorsoflarge,migratorycatfishinthe
AmazonRiverbasinusingotolithmicrochemistry.Wefocusedourstudyonthethreemost
populousandcommerciallyimportantspeciesintheAmazonBasin.Wesoughttodetermine
thelocationofearlyrearingandsubsequentmovementindourada,pirimutaba,andpiraíba
usingsamplescollectedfromtwo,largefishmarketsatthemouthoftheAmazonRiverandin
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 3/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
Table1. Brachyplaystomaspp.sampleinformation.
Species SampleNumber Location TotalLength(cm) Weight(kg)
Brachyplatystomarousseauxii BR1 Belém 75 8.00
(dourada) BR3 Belém 75 8.00
BR6 Belém 75 8.00
BR7 Belém 75 8.00
BR8 Belém 75 8.00
BR10 Belém 75 8.00
BR12 Belém 75 8.00
BR14 Belém 75 8.00
BR16 Belém 75 8.00
BR18 Belém 100 22.00
BR19 Manaus 110 11.00
BR21 Manaus 100 6.00
BR23 Manaus 115 10.00
BR24 Manaus 75 3.70
BR25 Manaus 80 4.00
BR26* Manaus - -
Brachyplatystomavaillanti BV1 Manaus 75 4.19°
(piramutaba) BV2 Manaus 70 3.38°
BV3 Manaus 68 3.09°
BV4 Manaus 65 2.68°
BV5 Manaus 70 3.38°
Brachyplatystomafilamentosum BF1 Belém 220 110.00
(piraíba) BF3 Belém 250 130.00
BF5 Belém 90 20.00
*Asterisciotolith.AnalysiswasexcludedduetolowSrconcentrations
°Weightsareestimatedfromlength-to-weightratios
FishwerecollectedfromtwolocationsintheBrazilianAmazon;inthecitiesofBelémnearthemouthoftheAmazonRiver,andManausinthecentral
AmazonBasin.Piramutabawereguttedpriortootolithcollection.Theirweightsareestimatedfromalength-to-weightrelationshipfromPirker[56].
doi:10.1371/journal.pone.0129697.t001
thecentralAmazon.Wedeterminedthemovementpatternsoverthelifetimeofindividualfish
usinglaserablationisotopemassspectrometryoftheirotoliths.Areasofstablesignaturewere
identifiedstatisticallythroughoutthechemicalprofileoftheotolith,whichwerethenclassified
totheirlocationwithinthebasinusingdiscriminantfunctionanalysis.Thediscriminantfunc-
tionwascreatedusingatrainingsetof87Sr/86SrsamplesfromriversthroughouttheAmazon
basin.Thesesampleswereobtainedfromthegeologicalliterature.Whereriver87Sr/86Srvalues
wereunknown,weusedestablishedrelationshipsbetweensurfacewater87Sr/86Srvaluesand
theageandcompositionoftheunderlyingwatershedgeologytopredictthesesignatures.
Ethicalstatement
Ethicalapprovalwasnotrequiredforthisstudy,asallfishwerecollectedaspartofroutinefish-
ingprocedures.FishweresacrificedbytheartisanalfishermeninManausandBelémusing
standardfisheriespracticesanddonatedtotheauthors.
Nofieldpermitsweredemandedtocollectanysamplesfromanylocation,sinceallsamples
derivedfromcommercialcatch.Noneofthespeciesincludedinthisinvestigationarecurrently
protectedorendangered.Therefore,noadditionalspecialpermitswerenecessary.Permission
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 4/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
toexporttheotolithsampleswasgrantedbytheBrazilianGovernmentwithpermitnumber:
116217(MMA,IBAMA,CITES09/01/2013).
Methods
OtolithCollection
InMarch2012,atotalof24pairedlapillusotolithsamples(16pairsfordourada,5forpiramu-
tabaand3forpiraíba)werecollectedfromthetwomajorfishingportsofBrazilianAmazonia,
thecitiesofManausandBelém(Table1andFig1A).Thesecitiesarelocated1,606riverkilo-
metersapart.Manaus(03°05'39.60''S,60°01'33.63''W),isthelargestcityinthecentralAmazon,
locatedattheconfluenceofthewhitewaterSolimõesRiverwiththeblackwaterNegroRiver.
Belém(01°27'18.04''S48°30'08.90''W),issituatedonthebanksoftheAmazonestuaryandis
themainlandingportoflargemigratorycatfishesfisheriesinBrazil[30].Priortootolithcollec-
tionthetotallength(TL)andweight(W)ofeachfishwasrecorded:dourada:meanTL=83.6
cm,meanW=8.58kg;piramutaba:meanTL=67.6cm,meanW=3.34kg;piraíba:mean
TL=186.6cm,meanW=86.6kg.Piramutabawereguttedpriortocollectionsoweightwas
estimatedusingalength-to-weightratiofromPirker[56].
FishcollectedfromManauswerecapturedinthemainstemAmazonRiverbetweenthe
mouthoftheMadeiraRiverandManausasreportedbythefisherman.Thus,wewouldexpect
thechemicalsignaturesrepresentingtheendofthefish'slife(signaturesfromtheedgeofthe
otolith)offishcaughtinManaustorepresentsignaturesinthemainstemAmazonRiverorits
tributariesupstreamoftheMadeiraRiver.FishcollectedinBelémwerecapturedintheestuary
between60kmand150kmfromBelémaccordingtothefisherman.Theotolithedgechemistry
offishcaughtinBelémarethereforeassumedtomatchthesignatureoftheestuary,thelower
Amazontributaries,orthelowerAmazonRivermainstem.Becausethe87Sr/86Srsignaturecan
takedaystoweekstoequilibrateandaccumulateenoughmaterialtoreliablysample,itispossi-
blethatfishcouldexhibitsignaturesotherthanthelocationofcaptureiftheyhadrecently
movedfromahabitatwithadifferentsignature.
OtolithAnalysis
Theleftlapillusotolithfromeachsamplewaspreparedusingstandardmethodsofmounting,
transversesectioningwithahighprecisionsaw,andabrasivepolishingtorevealtherings
[34,57](Fig2).Iftheleftotolithwasmissingorunavailabletherightotolithwasusedforanaly-
sis.OtolithswerethenanalyzedattheGeoAnalyticalLaboratoryatWashingtonStateUniver-
sityusingaFinniganNeptune(ThermoScientific)multi-collectorinductivelycoupledplasma
massspectrometercoupledwithaNewWaveUP-213laserablationsamplingsystem(LA-M-
C-ICPMS).Weusedamarineshellstandardtoevaluatemeasurementerrorrelativetothe
globalmarinesignatureof0.70918[58].Repeatedanalysesofamarineshellsignatureprovided
anaverage87Sr/86Srvalueof0.70914duringthecourseofthestudy(N=22,St.Error=
0.00002).Thelaserwasusedtoablateasamplingtransectfromthecoreoftheotolithsection
totheedge(30μm/sscanspeed,40μmspotsize,0.262sintegrationspeed,~7J/cm).This
resultedinacontinuoustime-seriesof87Sr/86Srdatafromthebirthofthefish(core)toits
death(edge)whichwasusedforsubsequentanalysis.FormoredetailedmethodsseeHegget.
al[40].Theasteriscuswasusedforonesampleforwhichthelapilliwerenotavailable;however,
thestrontiumconcentrationwaslowandtheunreliableresultswerenotincluded.
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 5/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
Fig2.Otolithsectioningandanalysis.Representativetransversesectionfromadouradalapillusotolithshowingtheanalysisarea(inred)usedforall
otolithswiththelaser-ablationtracksindicated.Allanalyseswereperformedapproximatelyperpendiculartothegrowthrings.
doi:10.1371/journal.pone.0129697.g002
BaselineWaterSamplingandPrediction
Twenty-fourwatersamplingpointslocatedthroughouttheAmazonRiverbasin(Fig1A)from
threepublishedstudiesprovidedbaseline87Sr/86Srvaluesforourstudy(Table2andFig3).
Santosetal.[53]providedthirteensamplesfromtheOre-HYBAMproject(www.ore-hybam.
org),amulti-yearresearcheffortwithacomprehensivesamplingdesigncoveringthemainstem
AmazonRiverandallofthemajortributariesaboveObidos,Brazil.Nineadditionalsamples
fromGaillardetetal.[52]coveredthemainstemAmazonandthemouthsofthemajor
tributariesasfareastasSantarem,Brazil.Finally,Queirozetal.[51]providedtwosamples
fromtheLowerSolimõesandUpperPurusRivers.Ourintentionwastoincludesamplesthat
representedallmajorAmazontributariesataregionalscale,whileexcludingsmallertributaries
thatwereunlikelytoprovidelong-termhabitatforourstudyspecies.Smallertributariesinthe
AmazonBasinhavebeenshowntoexhibitmuchdifferentisotopicchemistryfromtheirmain-
stemrivers[51].Thescaleandgeologicheterogeneityofthesesmallertributariescouldjeopar-
dizeassignmentaccuracy[54].
Theisotopicchemistryofafewsignificantlocationswerenotavailableintheliterature.
NotablymissingweresamplesfromthemouthoftheAmazonRiver,itstributariesbelowObi-
dos,andtheTocantinsRiverwhichcontributestotheestuaryhabitatofourstudyspecies.To
accountforthe87Sr/86Sroftheselocationsweusedtherelationshipbetweenthegeologic
makeupofawatershedandits87Sr/86Srsignaturetopredictthesepoints,followingasimilar
regressionapproachtoHeggetal.[54].WatershedsweredelineatedinqGIS(http://www.qgis.
org),anopen-sourcegeographicinformationsystem,usingtheGRASSanalysisplugin,which
containsadvancedwatershedanalysistoolsfromtheopen-sourceGRASSGISplatform(http://
grass.osgeo.org).Allanalysislayerswereprocuredfromopen-accessdatasets.Watersampling
pointsweremanuallydigitizedbytheauthorsbasedonlocationdescriptionsfromSantosetal.
[53],Gaillardetetal.[52],andQueirozetal.[51].Topographylayersweretakenfromthe
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 6/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
Table2. IsotopicandgeologicmakeupofmajorwatershedsoftheAmazonRiverbasin.
SamplePoint Literature River RiverGroup 87Sr/86Sr St.Dev. N Carboniferous Tertiary Precambrian Mean
Name(From Source Classification Age
Literature) (Ma)
Amazon13 Gaillardet Amazon WesterTributaries 0.710728 - 1 1% 47% 21% 409
etal.1997 &Amazon
Mainstem
Amazon14 Gaillardet Amazon WesterTributaries 0.71112 - 1 1% 47% 21% 409
etal.1997 &Amazon
Mainstem
Amazon20 Gaillardet Amazon WesterTributaries 0.711478 - 1 2% 42% 25% 429
etal.1997 &Amazon
Mainstem
Amazon6 Gaillardet Amazon WesterTributaries 0.709172 - 1 0% 60% 17% 407
etal.1997 &Amazon
Mainstem
RioMadeira Gaillardet Lower Beni-Madeira& 0.720036 - 1 2% 20% 30% 412
etal.1997 Madeira LowerNegro
RioNegro Gaillardet Lower Beni-Madeira& 0.716223 - 1 0% 21% 68% 1243
etal.1997 Negro LowerNegro
RioTopajos Gaillardet Lower LowerAmazon 0.733172 - 1 10% 8% 48% 658
etal.1997 Topajos Tributaries
RioTrombetas Gaillardet Lower LowerAmazon 0.732295 - 1 3% 7% 78% 515
etal.1997 Trombetas Tributaries
Uracara Gaillardet Lower LowerAmazon 0.723584 - 1 2% 19% 62% 558
etal.1997 Uracara Tributaries
Purus Queirozetal. Lower WesterTributaries 0.711135 - 1 0% 92% 4% 564
2009 Purus &Amazon
Mainstem
Solimões* Queirozetal. Lower WesterTributaries 0.714461 - 1 0% 63% 1% 289
2009 Solimões &Amazon
Mainstem
Atalaya Santosetal. Upper WesterTributaries 0.70887 - 1 0% 20% 0% 331
2013 Solimões &Amazon
Mainstem
Borba Santosetal. Lower Beni-Madeira& 0.71762 - 1 2% 20% 30% 412
2013 Madeira LowerNegro
Borja Santosetal. Upper WesterTributaries 0.7085 - 1 0% 13% 0% 426
2013 Solimões &Amazon
Mainstem
Caracarai Santosetal. Upper Beni-Madeira& 0.72238 - 1 0% 0% 74% 1650
2013 Negro LowerNegro
Franciscode Santosetal. Upper WesterTributaries 0.70592 0.00037 26 0% 67% 10% 287
Orellana 2013 Solimões &Amazon
Mainstem
Itiatuba Santosetal. Lower LowerAmazon 0.72964 0.00587 27 10% 4% 51% 671
2013 Tapajos Tributaries
LaBrea Santosetal. Lower WesterTributaries 0.71012 - 1 0% 90% 6% 1126
2013 Purus &Amazon
Mainstem
Manacapuru Santosetal. Lower WesterTributaries 0.70907 0.00025 38 0% 72% 2% 312
2013 Solimões &Amazon
Mainstem
Obidos Santosetal. Amazon WesterTributaries 0.71154 0.00053 46 1% 46% 23% 408
2013 &Amazon
Mainstem
(Continued)
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 7/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
Table2. (Continued)
SamplePoint Literature River RiverGroup 87Sr/86Sr St.Dev. N Carboniferous Tertiary Precambrian Mean
Name(From Source Classification Age
Literature) (Ma)
PortoVelho Santosetal. Lower Beni-Madeira& 0.71677 0.00073 9 3% 17% 22% 379
2013 Madeira LowerNegro
Ruranbaque Santosetal. Upper Beni-Madeira& 0.71730 0.00126 38 17% 13% 1% 433
2013 Madeira LowerNegro
Serrinha Santosetal. Upper LowerAmazon 0.73183 0.00737 16 0% 19% 80% 614
2013 Negro Tributaries
Tabitinga Santosetal. Upper WesterTributaries 0.70881 0.00029 9 0% 75% 3% 200
2013 Solimões &Amazon
Mainstem
AmazonMouth Predicted Amazon WesterTributaries 0.71625 0.00787° - 2% 40% 29% 482
(Predicted) from &Amazon
regression Mainstem
Jari(Predicted) Predicted LowerJari LowerAmazon 0.72928 0.00873° - 0% 2% 89% 718
from Tributaries
regression
Paru(Predicted) Predicted LowerParu LowerAmazon 0.72703 0.00845° - 1% 4% 78% 515
from Tributaries
regression
Tocantins Predicted Lower LowerAmazon 0.72683 0.00884° - 13% 22% 36% 877
(Predicted) from Tocantins Tributaries
regression
Xingu Predicted Lower LowerAmazon 0.72633 0.00827° - 2% 13% 70% 1091
(Predicted) from Xingu Tributaries
regression
*Outlierdroppedfromregressionanalysis
°ValuesarepredictionintervalsoftheregressionfromEq1.
StrontiumratiosweretakenfromwatersamplesreportedintheliteratureforlocationsthroughouttheAmazonBasinandusedasbaselinestodetermine
thelikelylocationoffishmovement.Thesesampleswereclassifiedtothreestatisticallydistinguishablerivergroupclassificationsusingquadratic
discriminantfunctionanalysis.Unsampledlocationswerepredictedusinggeologicregression(Eq1).
doi:10.1371/journal.pone.0129697.t002
GTOPO30dataset[59]andstreamcoursesfromtheHydroSHEDdataset[60].Geologicinfor-
mationcamefromtheWorldEnergyAssessmentGeologicMapoftheAmazonRegion[61].
Weusedgeologicageastheprimarycandidateindependentvariablesinourregressionto
predictthe87Sr/86Srsignaturesforunsampledtributaries,alongwithverygeneralintrusiveand
extrusiverock-typecategories(Fig1B).OurmethodsdifferedfromHeggetal.[54],whoused
rocktypeastheprimaryexplanatoryvariableratherthanage.Wedidthisbecausethevery
genericdesignationsofintrusiveorextrusiverockavailableinourdatasetwereinsufficientto
provideexplanatorypower.Thevaluesforthesecandidatevariableswerecalculatedbycon-
vertingthegeologicagecodesfromthemapattributetabletocontinuousvariablesusingthe
meanage(Ma)ofthegeologicperiodsencompassedbyeachcodeusingtheInternational
ChronostratigraphicChart[62].Thepercentageareaofeachrockageandtypewasthencalcu-
latedwithinthewatershedupstreamofeach87Sr/86Srsamplepoint.Themeanageofeach
watershed,weightedbyarea,wasalsoincludedasapotentialexplanatoryvariableforthe
regression,leavingtwenty-fourpotentialexplanatoryvariablesfortheregression.
Modelselectionusedageneticalgorithmselectionprocedureinthe{glmulti}packageforR
[63].Welimitedmodelstofourtermstolimitthenumberofpotentialmodelsandincluded
interactionterms.ThegeneticalgorithmusesasearchalgorithmbasedonDarwiniannatural
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 8/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
Fig3.RiverisotopicsignaturesthroughouttheAmazonRiverbasin.Strontiumratiovalues(y-axis)for
eachsampledandpredictedwatershed(x-axis)inthecurrentstudy.Colorindicatestheclassificationtothree
rivergroupsusingquadraticdiscriminantfunctionanalysis.Soliderrorbarsindicatethestandarddeviation
wheresampleswererepeatedovertime(SeeTable1forsamplesizes).Dashederrorbarsindicatethe
predictionintervalsfromthegeologicregression(Eq1)usedtopredictthatpoint.Pointsborderedinblack
weremisclassifiedduringcrossvalidationofthequadraticdiscriminantfunction.
doi:10.1371/journal.pone.0129697.g003
selection,anefficientmethodformodeloptimizationwhenthenumberofpotentialmodelsis
large,aswasthecasewithourgeologicdata[64].Akaike’sInformationCriterionoptimizedfor
smalldatasets(AICc)wasusedastheoptimizationcriteriaforthegeneticalgorithm,acriterion
thatpenalizesover-parameterization[65].Onethirdofthesamplepointswererandomly
selectedasavalidationset,withheldfromthemodelselectionprocedure,andusedtoassess
predictionaccuracyofthebestmodel.Thebestmodelwasthenusedtocalculatethe87Sr/86Sr
valuesfortheunsampledpointsinthebasin,usingthegeologyupstreamofthesepoints.
GroupingofDistinguishableWatersheds
Thewatersamplepointsweregroupedintothreedistinguishablegeographicregionsusing
priorknowledgeofthegeographyandgeologyofthewatersheds(Table1,Fig1A,Fig3).River
basinsthatweregeographicallycontiguousandbroadlygeologicallyandchemicallysimilar
weregrouped.TheAmazonRivermainstemandwesterntributaries,allconsideredwhitewater
rivers[26],weregroupedtogetherduetotheoverwhelminginfluenceoftheAndesontheir
chemistry.TheBeni-MadeiraRiverandlowerNegroRiverweregroupedduetosimilarchemi-
calsignaturesfromamixofuplandmountainousgeologyandold,lowland,AmazonandGuy-
anashieldgeology.TheNegroRiver,beingblackwater,wouldnotbeexpectedtocontainlarge
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 9/19
DiverseEarlyLife-HistoryStrategiesinMigratoryAmazonianCatfish
numbersofourtargetspecies,whilethewhitewaterMadeiraisaknownfishery[23,26].The
lowerAmazontributaries(belowtheMadeiraRiver)weregroupedduetothetheirsimilarly
old,shieldgeologiesresultinginhigh87Sr/86Srvalues.Theseriversareallconsideredclearwater
tributaries[26].Thesegroupassignmentswerethenusedasthetrainingset,with87Sr/86Sras
thepredictorandgroupastheresponse,tocreateaquadraticdiscriminantfunction.Thisdis-
criminantfunctionwasthenusedinthefollowingsectiontoclassifythe87Sr/86Srsignatures
recoveredfromfishotolithstothesethreedistinguishablerivergroups.
DeterminingFishMovementandLocation
Thetransectof87Sr/86Srvaluesfromthecoretotherimofeachotolithwasanalyzedusinga
PELTalgorithmchangepointanalysis(usingthe{changepoint}packageinR[66])todetermine
whenthemean87Sr/86Srvalueschangedtoanewstablesignature.Thischangepointalgorithm
generatedmeanvaluesandstartingpointsforeachstableregionwithintheotolithtransect
usingapenaltyvalueof0.0001[67].Eachstablesignaturewasassumedtocorrespondtomove-
mentintoanewriversignature,withthefirststablesignaturecorrespondingtoearlyrearing.
Insomecasesthechangepointalgorithmreturnederroneousmeansforsmallportionsofthe
signature,inlocationswerethemeanswasobviouslyunstable.Suchfragmentsweremanually
removed.
Stableotolithsignatureswerethenclassifiedtotheirlikelylocationusingthediscriminant
functiondevelopedfromknownandpredictedwatersamplingpointsinthepriorsection.
Becauseaprioriprobabilityofgroupmembershipwasunknown,thepriorprobabilitiesforthe
discriminantfunctionweresettobeequalamonggroups.Afterclassification,theresultswere
plottedandthedatawereassessedtodeterminetrendsinearlyrearingandmovementboth
withinandamongthethreesampledspecies.
Results
BaselineWaterSamplingandPrediction
Themostparsimoniousmodelwithoutinteractiontermsexplained~80%ofthevariationin
thedatabutprovidedanunreasonablyhighpredictionforthemouthoftheAmazonRiver.We
hadnodirectevidenceofsignificantgeologicinteractions;however,weincludedinteractions
inasecondmodelselectionexerciseinhopesoffindingaparsimoniousmodelthatbetterfit
theavailabledata.Welimitedthemaximumnumberofmodeltermstofourtolimitthenum-
berofpotentialmodelsavailablefromthetwenty-fouravailablevariablesplusinteractions.
Thislimitisreasonablesincemoretermswouldriskoverparameterizationgiventhenumber
ofobservationsusedtobuildthemodel.UndertheseconditionstheAICcmodel-selection
algorithmselectedthreemodelsthatweregreaterthantwoAICcpointssuperiortothenext
mostparsimoniousmodel.Thetopmodel,
87Sr=86Sr¼ð0:0263ÞPrecambrian(cid:2)ð0:676ÞPrecambrian:Tertiary(cid:2)ð0:0011ÞMeanWatershedAgeðweightedbyareaÞþεð1Þ
explained89%(AdjustedR2)ofthevariationinthedata,providedthebestpredictionresiduals
forthevalidationset,andresultedinamorereasonablepredictionofthemouthofthe
Amazon.Thisequationwasusedtopredictthe87Sr/86Srsignaturesforthefiveunsampled
watersheds.
GroupingofDistinguishableWatersheds
Aquadraticdiscriminantfunctionprovidedthebestcross-validationerrorrate(3.6%)fordis-
criminatingallthewatershedsintothethreeregions.OnepredictedvaluefortheMadeira
PLOSONE|DOI:10.1371/journal.pone.0129697 July8,2015 10/19