DemographyandPopulationStatusofPolarBearsinWesternHudsonBay,

Canada

NicholasJ.Lunn,SabrinaServanty,EricV.Regehr,SarahJ.Converse,Evan
Richardson,IanStirling

EnvironmentCanadaResearchReport

AuthorAffiliation:
NicholasJ.Lunn1,SabrinaServanty2,3,EricV.Regehr4,SarahJ.Converse3,EvanRichardson1,
IanStirling1,5

1
WildlifeResearchDivision,Science&TechnologyBranch,EnvironmentCanada,CW405
BiologicalSciencesCentre,UniversityofAlberta,Edmonton,ABT6G2E9,Canada

2
ColoradoStateUniversity,ColoradoCooperativeFishandWildlifeResearchUnit,FortCollins,
CO80523,USA

3
U.S.GeologicalSurvey,PatuxentWildlifeResearchCenter,12100BeechForestRoad,Laurel,
MD20708,USA

4
U.S.FishandWildlifeService,MarineMammalsManagement,1011E.TudorRd.,Anchorage,
AK99503,USA

5
DepartmentofBiologicalSciences,UniversityofAlberta,Edmonton,ABT6G2E9,Canada

SUMMARY
WeevaluatedthedemographyandpopulationstatusoftheWesternHudsonBay(WH)
polarbearsubpopulationfortheperiod19842011,usingliverecapturedatafrom
researchstudiesandmanagementactions,anddeadrecoverydatafrompolarbears
harvestedforsubsistencepurposesorremovedduringhumanbearconflicts.

WeusedaBayesianimplementationofmultistatecapturerecapturemodels,coupled
withamatrixbaseddemographicprojectionmodel,tointegrateseveraltypesofdataand
toincorporatesamplinguncertainty,anddemographicandenvironmentalstochasticity
acrossthepolarbearlifecycle.Thisapproachallowedforestimationofasuiteofvital
rates,includingsurvivalandreproduction.Thesevitalrateswereusedtoparameterizea
Bayesianpopulationmodeltoevaluatepopulationtrendsandprojectpotential
populationoutcomesunderdifferentenvironmentalscenarios.

Survivaloffemalepolarbearsofallageclasseswassignificantlycorrelatedwithseaice
conditions;particularlywiththetimingofseaicebreakupinthespringandformationin
thefallandtheinteractionofthetwo.Thisisconsistentwithpreviousfindingslinking
bodyconditionandsurvivalofWHpolarbearstoenvironmentalchangesassociatedwith
climaticwarmingandsupportstheecologicaldependenceofpolarbearsonthe
availabilityofseaice.

Survivalofmalepolarbearswasnotcorrelatedwithseaiceconditions.Thiswaslikely
becauseahigherproportionofmortalityformaleswascausedbyhumansratherthanby
naturalfactors.Forexample,approximately73%ofmortalityforyoungmalebears(i.e.,
59yearsold)wasduetodirecthumancausedremovals,largelybecauseofsexselectivity
inthesubsistenceharvest.

ThedecliningtrendinsizeoftheWHsubpopulationovertheperiod19872004was
similartoapreviousanalysis(Regehretal.2007),suggestingconsistencybetweenthe
twodemographicevaluations.Pointestimatesofabundanceweresomewhatlowerusing
theupdatedstatisticalapproach.Itisimportanttorecognizethattheanalyzeddatawere
notcollectedinamannerthatisoptimalforestimatingabundanceandthatthegoalof
thecurrentanalysiswastoestimatevitalratesanddemographictrends.

EstimatesofpopulationgrowthratewerealsoderivedusingaBayesianpopulationmodel
basedonestimatedsurvivalandreproductiveratesfromthemultistatecapture
recapturemodel.Fortherecentdecade20012011,thegrowthrateofthefemale
segmentofthepopulationwas1.02(95%CI=0.981.06).Apparentlystabletopositive
populationgrowthforfemalesmaybedueinlargeparttononlinearity(i.e.,shortterm
stability)inthelongtermobservedandforecastedtrendtowardearlierseaicebreakup
inwesternHudsonBay.
The2011abundanceestimatefromthisanalysiswas806bearswitha95%Bayesian
3

credibleintervalof653984.Thisislowerthan,butbroadlyconsistentwith,the
abundanceestimateof1,030(95%confidenceinterval=7451406)froma2011aerial
survey(Stapletonetal.2014).Thecapturerecaptureandaerialsurveyapproacheshave
differentspatialandtemporalcoverageoftheWHsubpopulationand,consequently,the
effectivestudypopulationconsideredbyeachapproachisdifferent.

INTRODUCTION

Polarbears(Ursusmaritimus)aredistributedthroughouttheicecoveredwatersofthe

circumpolarArcticin19relativelydiscretesubpopulations(Obbardetal.2010).Asthey
dependonseaiceasaplatformfromwhichtohuntseals(StirlingandDerocher2012),changes
inthedistributionandextentofseaiceandthepatternsoffreezeupandbreakuphavethe
potentialtoinfluencethepopulationecologyofpolarbears(StirlingandDerocher1993,2012;
Derocheretal.2004;StirlingandParkinson2006;Laidreetal.2008;Rodeetal.2010,2012,
2013).Theimpactsofalongtermwarmingtrendinspringtemperatures(Skinneretal.1998;
GagnonandGough2005)onthedatesofseaicebreakupandfreezeupinHudsonBayhave
beenwelldocumented(Hochheimetal.2010,2011).Theseinturnimpactbodycondition,
reproduction,anddemographyofpolarbearsintheWesternHudsonBay(WH)subpopulation
(Stirlingetal.1999,2004;Regehretal.2007).WHpolarbearsarespendingprogressively
longerperiodsoftimeonland,duetoearlierarrivalonshoreandlaterdeparturefromland
relatedtodeclinesintheavailabilityofseaice(Cherryetal.2013).

Duringthe1960sand1970s,InuitreportedthattheabundanceoftheWH

subpopulationhadincreased(Tyrrell2006;NirlungayukandLee2009),likelyduetoreduced
huntinganddisturbancefollowingclosureoftheYorkFactoryfurtradingpost,withdrawalof
militarypersonnelfromChurchill,andtheclosureofhuntinginManitoba(Stirlingetal.1977;
DerocherandStirling1995).DerocherandStirling(1995)estimatedthemeanpopulationsize
oftheWHsubpopulationfor19781992tobe100051.Asthisestimatewasconsideredtobe
conservativebecausethestudyhadnotincludedthesoutheasternportionoftherangeeastof
theNelsonRiver(Figure1),thepopulationsizewaslateradjustedto1200formanagement
purposes(Calvertetal.1995;Wiigetal.1995).Regehretal.(2007)estimatedthatabundance
haddeclinedfromabout1194(95%CI=1020,1368)in1987toabout935(95%CI=794,1076)
in2004.Theyfoundthatsurvivalratesofcubs,subadults,andoldbears(>20years)ofboth
4

sexeswerecorrelatedwiththedateofseaicebreakup,anddecreased25%foreachweek
thatseaicebrokeupearlierthanaverage(Regehretal.2007).

AspartofacollaborativeeffortbetweenEnvironmentCanada,theUSFishandWildlife

Service,andtheUSGeologicalSurvey,weanalyzeddatafromthecaptureoffreerangingpolar
bearscollectedbyEnvironmentCanadaanditspartnersforthreedecades,aspartofalong
termresearchprogramontheecologyandstatusofpolarbearsinwesternHudsonBay.
Analysisofthesedataallowsforquantificationofrelationshipsbetweenenvironmentalchange
anddemography,anassessmentofthehealthofthesubpopulationbyconsideringchangesin
keyvitalratessuchassurvivalandreproductiverates,andestimationofthesizeandtrendof
theeffectivestudypopulation,asdefinedbytheextentandtimingofcapturerecapture
sampling,toinformmanagementdecisions.Ouranalysisincludedsevenadditionalyearsof
data(20052011)thatwerenotavailableinpreviousanalyses(Regehretal.2007).Weuseda
hierarchicalmultistatecapturerecapturemodel(e.g.,Arnason1972;Hestbecketal.1991;
Kendalletal.2006;Converseetal.2009;KryandSchaub2012),ratherthanthesinglestate
modelusedpreviously.Themultistatemodelingframeworkinvolvesspecificationofmultiple
statesinthiscase,lifestageswhichcanmoreeffectivelyaccountforheterogeneityby
allowingdemographicparameterstovarybetweenstates.Thismodelalsoallowsfor
estimationofdemographicparametersofprimaryinterest,suchassurvivalandreproductive
rates,whichdescribetransitionsbetweenstates.Ourapproachincorporatedabroaderrange
ofdatathanpreviousanalyses:dataforbearsthatwerecapturedandreleasedaliveunder
multiplesamplingprotocols,forbothresearchandmanagementpurposes,anddataforbears
thatwereharvestedforsubsistencepurposesorkilledforreasonsofhumansafety.The
multistatemodelallowedustoexplicitlyconsiderthesex,age,andreproductivestatusofbears
andthehierarchicalstructureallowedustomoreefficientlyaccountforannualvariationin
parameterssuchasdetectionprobability.Weusedtheresultsofthisestimationprocedureto
constructapopulationmodel,usingaBayesianPopulationViabilityAnalysis(BPVA)approach
(i.e.,demographicinputstotheviabilityanalysiscomefromtheBayesianestimationprocedure;
Wade2002;KryandSchaub2012;Servantyetal.2014),toprojectpotentialpopulation
outcomesunderdifferentseaicescenarios.Theprimaryobjectiveswereto:
5

1.estimatesurvivalrates,ratesofmortalitydirectlyattributabletohumans,reproductive
rates,populationsize,andpopulationtrend;
2.evaluaterelationshipsbetweendemographicparametersandseaiceavailability;and,
3.assessthepotentialimpactsofchangesinseaiceavailabilityonthestatusoftheWH
subpopulation.

METHODS
Studyarea
ThecurrentboundariesoftheWHsubpopulation(Figure1)arebasedonextensive
recordsofcapture,recapture,andharvestoftaggedbears(Stirlingetal.1977;Derocherand
Stirling1990,1995;TaylorandLee1995;Lunnetal.1997).Thissubpopulationappearstobe
largelysegregatedgeographicallyfromtheSouthernHudsonBay(SH)subpopulationtothe
southeastandtheFoxeBasin(FB)subpopulationtothenorthduringtheopenwaterseason,
althoughallthreesubpopulationsmixtosomedegreeontheHudsonBayseaiceduringwinter
andspring(Stirlingetal.1977;DerocherandStirling1990;StirlingandDerocher1993;Taylor
andLee1995;Peacocketal.2010).
Fieldmethodsanddatasources
WeanalyzedliveencounteranddeadrecoverydataforWHpolarbearsthatwere
collectedfromdifferentsourcesfrom19842011.EachyearbetweenlateAugustandearly
October,EnvironmentCanada(EC)andtheUniversityofAlbertacapturedandreleasedpolar
bearsina12,000km2areabetweenChurchill,ManitobaandtheNelsonRiver(Figure1;AreaC)
aspartofalongtermresearchprogram(e.g.,Stirlingetal.1977;RamsayandStirling1986,
1988;DerocherandStirling1990,1995;Stirlingetal.1999).Althoughtherewaslimited
geographiccoverageinthesamplingeffortwithinthisareain19841986,theannualsampling
effortwasmoreevenlydistributedsince1987(Regehretal.2007).AdditionalsamplingbyEC
occurredintheareabetweenChurchillandtheManitobaNunavutborder(Figure1;AreaB)in
1999,and20052011;and,betweentheNelsonRiverandtheManitobaOntarioborder(Figure
1;AreaD)in19841986,19941995,and20032005.
Intheautumnwhenresearchoccurred,polarbearsweregreatlyconcentratedrelative
6

totheirdispersedwinterdistributionovertheseaiceandwereeasilysightedagainstasnow
freebackground.Freerangingbearswerelocatedbyhelicopterandcapturedusingstandard
chemicalimmobilizationtechniques(Stirlingetal.1989).Allpolarbearsobservedduring
fieldworkwerecapturedwhenitwassafetodoso,exceptforsomepregnantfemalesthattook
refugeinmaternaldensexcavatedintheground.Capturedpolarbearswereindividually
markedusingplasticeartagsandpermanenttattoosontheinnersurfacesoftheupperlip.
Eachbearwastemporarilymarkedwithpainttoavoidrecapturingthatindividualthesame
season.Avestigialpremolarwasextractedfromuntaggedanimalsolderthanoneyearforage
determination(CalvertandRamsay1988)whereastheagingofcubsoftheyear(COY;
approximately9monthsoldinautumn)andyearlingcubs(approximately21monthsoldin
autumn)wasbasedonbodysizeanddentition.
Veryhighfrequency(VHF)collars(AdvancedTelemetrySystems,Isanti,MNorTelonics,
Inc.,Mesa,AZ;useofproductnamesdoesnotimplyendorsementbytheUSorCanadian
governments)orsatellitecollars(Telonics,Inc.)weredeployedonsomeadultfemales.Collars
werenotappliedtosubadultbearstoavoidinjuryduringgrowth,ortoadultmalesastheir
necksarelargerincircumferencethantheirheads.MostVHFcollarsremainedactiveforfive
yearsandsatellitecollarsfortwoyears.However,activecollarsdidnotguaranteerecapture,
socaptureprobabilityforthisgroupwasexpectedtobe<1.Wederivedthetimevarying
individualcovariatetelemetrytoindicatewhenadultfemaleswerelikelytobeavailablefor
targetedrecapturebytelemetry,basedoncollardeploymentinformationandexpectedbattery
life.Thisapproachwasexpectedtoaccountformostindividualheterogeneityinrecapture
probabilityresultingfromliveencountersofbearslocatedbytelemetry.Itdidnotaccountfor
someyearlingslocatedbytelemetryinassociationwiththeirmotherscollarbecausewedid
notknowaprioriwhichcollaredfemaleshaddependentyoung,nordiditaccountforvariation
inbatterylifeorcollarretention.
AllcaptureandhandlingmethodswerereviewedandapprovedannuallybytheEC
PrairieandNorthernRegionAnimalCareCommitteeandtheUniversityofAlbertaBioSciences
AnimalPolicyandWelfareCommittee.ManitobaConservationandParksCanadaAgency
issuedannualwildliferesearchpermitsunderwhichthesestudieswereconducted.
7

Wealsoincludeddatathatwerecollectedbyotheragenciesusingsimilarmethods:in
AreaCbyUniversityofSaskatchewanin19891996(e.g.,AtkinsonandRamsay1995;Cattetet
al.1997;Polischuketal.2002);inAreaDbyOntarioMinistryofNaturalResourcesin19841986
and20032005;and,alongtheNunavutcoastofwesternHudsonBay(Figure1;AreaA)bythe
GovernmentofNunavutin2007.Toensurethatvitalrateswereestimatedusingaconsistent
studypopulation,weonlyincludedliveencountersofpolarbearsoutsidethestudyarea(Area
C)iftheyhadbeenpreviouslycapturedwithinit.
WealsoincludeddataonbearscapturedbyManitobaConservation(MB)staffnear
ChurchillthroughthePolarBearAlertProgram(Kearney1989).Polarbearswereimmobilized
fromthegroundandeitherdetainedinaholdingfacilityortransportedoutofChurchill,
generallytocoastalareasupto50kmnorth.Someproblembearsweresubjecttolethal
removal.MarkinganddatacollectionproceduresweresimilartothoseusedbyEC.Previous
analysessuggestedthathandlinginChurchillwascorrelatedwithlowersurvival(Regehretal.
2007),perhapsbecausebearsseekingsupplementalfoodintownweremorelikelytobein
poornutritionalcondition(LunnandStirling1985)orbecausebearsfrequentingChurchillhada
greaterchanceofbeingkilledbyhumansnearcommunities.HandlinginChurchillwasalso
previouslyassociatedwithhigherrecaptureprobability(Regehretal.2007),likelybecause
polarbearshandledthereweremorelikelytoreturnforpotentialfoodrewardsandtherefore
besusceptibletorecapturebyMB(Regehretal.2007).Toaccountforthispotentialvariation,
wederivedatimevaryingindividualcovariate,Churchill,whichwassetto0priortoand
includingthesamplingoccasiononwhichabearwasfirstcapturedbyMBandsetto1onall
subsequentsamplingoccasions.
Finally,additionaldatacamefrompolarbearsharvestedeachyearaspartofalegal,
regulatedsubsistencehuntbyInuitlivingalongtheNUcoastofwesternHudsonBay(Figure1;
AreaA;Derocheretal.1997;Tayloretal.2008).Biologicalsamplesandotherinformation
wereobtainedfromharvestedbears,andtheidentitiesofresearchmarkedbearsthatwere
harvestedwereprovidedtoEC.
Wefittedmodelstoasubsetofthetotaldatabasedonthefollowingcriteria.We
considereddatafrom19842011,forconsistencywithpreviousanalyses(e.g.,Regehretal.
8

2007),andbecauseECandMBsamplingprotocolswererelativelyconsistentovertheseyears.
Weexcludedlivecapturesfromoutsidetheannualsamplingperiodof1Augustto15
November,thusincludingapproximately95%ofautumnliveencounters,tobettermeetthe
assumptionofinstantaneoussamplingcommontolivecapturerecapturemodelsforopen
populations(Williamsetal.2002).Ifanindividualhadmorethanoneliveencounterinayear
withadifferentstateineachencounter(e.g.,hadadependentcubinoneencounterbutnotin
another),weusedtheearlierencountertodefinethebearsstate(whenafemalehas
dependentcubs,observationofthefemaleisexpectedtoguaranteeobservationofthecubsas
well).Ifaknownfemalewasencounteredbutidentitiesofherdependentyoungwere
unknown(e.g.,duringavisualencounterofafamilygrouplocatedbytelemetry),weusedthe
fieldestimatedageclassofthedependentyoungtoinformthestateoftheadultfemale(e.g.,a
femalewithcubsoftheyearversusafemalewithout),anddidnototherwiseincludethe
dependentyounginthedataset.
Weassignednumericagesto16yearlingsandtwoyearoldsthatlackedtoothderived
ageinformation,basedontheirfieldestimatedageclass.Weassignednumericagesto66
olderbearslackingageinformationbasedonthemediantoothderivedageofotherbears
capturedonthesamesamplingoccasionwiththesamefieldestimatedageclass(subadultor
adult)anddegreeoftoothwear(subjectiveindex13).
Wehandleddeadrecoveriesintwoways.First,individualcapturehistorieswereright
censored(i.e.,ignoredintheanalysis)followinginadvertentdeathsduringcapture,sothese
removalsdidnotaffectparameterestimates.Second,weincludedahumancausedmortality
stateforpurposefulhumancausedremovals(i.e.,bearsintheNUharvestandproblembears
killedbyMB).Deadrecoveriesthatoccurredafterthestartofthesamplingperiodincalendar
yearjwereassignedtothehumancausedmortalitystate(forwhichweestimatedatransition
probability)incalendaryearj+1,whichensuredthatestimatesofhumancausedmortality
includedbearsfirstmarkedinyearjandsubsequentlyremovedinthesameyear.Since
capturerecapturemodelsestimatedemographicparametersbasedoncapturehistoriesof
individuallyidentifiedanimals,weonlyincludeddeadrecoverydataforpolarbearsthatwere
encounteredandmarkedinthestudyareaafter1984.
9

Seaice
WeevaluatedtrendsinseaiceavailabilityintheWesternHudsonBaysubpopulation
managementzonefrom19792012usingpassivemicrowavesatelliteimagery(resolution25x
25km)fromtheNationalSnowandIcedataCenter(NASATeamalgorithm),whichprovides
dailyseaicecoverconcentrationstothenearestpercentage(Cavalierietal.2012).Seaice
imagerywassampledeachdayusing381samplingpointsthatprovidedcompletecoverageof
theWesternHudsonBaymanagementzone(Figure1).Themeaniceconcentrationoverall
381pixelswascalculatedforeachdayoftheyeartodeterminetheaverageiceconcentration
withinthemanagementzone.Fromthesedatawederivedthreeseaicemetrics:i)breakup
date:definedastheordinaldateatwhichspringseaicereached50%concentrationandstayed
belowthatconcentrationforatleastthreeconsecutivedays(Etkin1991;Stirlingetal.1999;
GagnonandGough2005;StirlingandParkinson2006),ii)freezeupdate:theordinaldateat
whichautumnseaicereached50%concentrationandremainedabovethatconcentrationfor
atleastthreeconsecutivedays,andiii)icedecay:theabsolutevalueoftheslopeofanordinary
leastsquaresregressionoftherateofseaicelossfrom1MayofeachyearuntiltheWestern
HudsonBaymanagementzonewascompletelyicefreeinthesameyear.Correlationsbetween
these3variableswereweakwiththeexceptionofthecorrelationbetweenbreakupdateand
icedecay,whichwas0.93.However,weincludedbothbreakupdateandicedecayinthe
modelstodeterminewhichmightbeabetterpredictorofdemographicparameters,viamodel
selectionasdescribedbelow.Wealsoincludedinteractionsbetweenbreakupandfreezeup
dateinsomemodelstoaccountforthepotentialinteractiveeffectofhavingbothanearly
breakupandalatefreezeup(forexample)inagivenyear,andbreakupandicedecayto
accountforthepotentialinteractiveeffectofboththetimingandrateofseaicebreakup.
Multistatemodelstructure
Wedevelopedseparatemultistatestructuresformalesandfemales(Figures24).
Transitionsamongstatesarerepresentedbyarrowsanddependonfivetypesofparameters
(Table1).Femalebears(Figure2)canfirstenterthedatasetwhentheyareasyoungascubsof
theyear(~9mooldcubsdependentontheirmother;abbreviatedasFC).Oneyearlater,
conditionalontheirsurvivalS,thesecubswillhaveeitherbecomeindependent(i.e.,weaned;
10

F1I)withprobabilityWorremaineddependentontheirmother(F1D)withprobability1W.
Threesubsequentclassesreflectannualageincreases(F2,F3,F4)andtransitionsbetween
thesestatesaredeterministic(i.e.,thebearmustgetolder)conditionalonsurvival.Femalesin
theF4state(i.e.,at4+yearsofage)transitionthefollowingyearintooneofthreeadultstates,
conditionalonsurvival.Conditionalonnotbreeding,withprobability1BwhereBisdefinedas
breedingprobability,femalestransitionintothefemalewithnocubs(FnY)state.Conditional
onbreeding,withprobabilityB,femalescanbeobservedinautumnwith2cubs(i.e.,twins,
withprobabilityT;F2Y)orwithasinglecub(withprobability1T;F1Y).Weemphasizethat
breedingisdefinedhereastheprobabilityofhavingatleastonecubthatsurvivestoautumn,
andsoistheproductoftheprobabilitiesofgivingbirthtoatleastonecubandofhavingat
leastoneofthecubsproducedsurvivetoautumn.AdultbearsinthestateFnYincludeamixof
femaleswithoutdependentyoung,andfemaleswithoneormoreyearlings,whichmaybe
eitherdependent(andsowouldbeobservedifthemotherisobserved)orindependent(and
mayormaynotbeobserved,independentofthemother).Includingafemalewithyearling
statewouldrequireaccountingforstateuncertainty(Pradel2005)toappropriatelyrepresent
stochasticityintheweaningprocessandresultinguncertaintyinsurvivalofyearlings.As
includingthisstatewouldhaveresultedinamodelstructurethatwasprohibitivelycomplex,it
waspreferableforourpurposestonotincludeitandacceptsomeheterogeneityintheFnY
state.AconsequenceofthisapproachisthatBrepresentsanaveragebreedingprobabilityfor
alladultfemalesinstatesF4andFnY,someofwhich(e.g.,thefemaleswithyearlingsinFnY)
werenotlikelyphysiologicallycapableofbeingpregnantinthatyear.

Thestatestructureformalesissimpler(Figure3),andincludesstatesanalogoustoFC,

F1I,andF1D(MC,M1I,andM1D,respectively).Beyondtheageof21months(i.e.,beyond
statesM1DandM1I),allmales,conditionalonsurvival,enteranadultmalestate(MA).

Withtheinclusionofmortalityinformation,bothfemaleandmalemultistatestructures

include2deathstates(Figure4),eitherofwhichcanbeenteredfromanylivestate,conditional
onmortality(1S).Theseincludeanobservabledeathstate(FoDandMoD,forfemalesand
malesrespectively),whichbearsenterwithprobabilityH,conditionalonmortality,indicating
theprobabilitythatmortalitywashumancaused(eitherthroughhuntingormanagement
11

removals).Weassumethatallbearsenteringthisstatearedetected,giventhehighadherence
toreportingrequirementsforbearstakenundertheregulatedsubsistenceharvest(Stirlingand
Prestrud1994;Peacocketal.2011;VongravenandPeacock2011)orformanagement
purposes.Bearscanalsoenteranunobservabledeadstate(FuDandMuD),iftheirmortalityis
notduetoahumancause,withprobability1H.
Survival,S,reflectsannualapparentsurvival,theprobabilitythatanindividualsurvives
andremainsinthestudyareafromyearttoyeart+1.Weaning,W,istheprobabilitythat,
conditionalonsurvival,anindividualinacubstate(FCorMC)inyearthasweanedandentered
stateF1IorM1Ibyyeart+1.Breedingprobability,B,istheprobabilitythatafemalebear
withoutacubattheprevioustimestepproducesacubthatsurvivesto9monthsofage.
Twinningprobability,T,istheprobability,conditionalonbreeding,thatafemaleproduces2
(or,inrarecases,3)youngthatsurviveto9monthsofage.
Modelfitting
Wetestedavarietyofcovariatesandeffectswhichdescribedpolarbearbiology,
environmentalconditions,oraspectsofstudydesign.Covariateswereusedtogetherwiththe
statestructureinthelifecyclegraphstoexplaintemporal,group,andindividualvariationinthe
parameters(Table2).Ageeffectswerepartiallyrepresentedbystatesinthelifecyclegraph
(statesFCthroughF4reflectannualagesforfemales,andstatesMC,M1I,andM1Dreflect
annualagesformales).Forstatesthatincludedmultipleages(MA,F1Y,F2Y,andFnY)more
detailedageeffectswereincludedwithinthestates.Thestructureoftheseageeffectsis
describedinTable2.Overallagestructurewassimilartoormoredetailedthanprevious
analyses(Regehretal.2007)andvariedamongtheparametersS,B,H,TandW.Furthermore,
weincludedinteractionsbetweenagestructureandenvironmentalcovariates,toallowfor
differingeffectsoficeconditionsonbearsofdifferentages.Seaicecovariateswere
transformedtoastandardnormaldistributionbeforeinclusioninthemodel.
Thelargesizeofthedatasetmadeitimpracticaltoanalyzeasasingleunit,soall
analyseswereconductedseparatelyforfemalesandmales.WefittedthemodelsusingMarkov
chainMonteCarlo(MCMC)simulationsinaBayesiananalyticalcontext(e.g.,McCarthy2007;
RoyleandDorazio2008).Wechosestandardvaguepriordistributionsfortheparameters.
12

ThreeMCMCchainswithrandominitialvaluesweregenerated,withconvergenceassessed
basedonaGelmanandRubinstatisticbetween1and1.1(Gelman1996;GelmanandHill2007).
ThesimulationswereperformedusingJAGS3.3.0(Plummer2003).TheRpackagerjags
(Plummer2013)wasusedtocallJAGSandexportresultstoR2.14.2(RDevelopmentCore
Team2012).Weusedempiricalmeansand95%Bayesiancredibleintervals(95%BCI)to
summarizeposteriordistributions.
WeusedprogramRELEASE(Burnhametal.1987)toevaluatehowwelltheCormack
JollySeber(CJS)capturerecapturemodelforopenpopulationsfitasubsetofthedatathat
excludedhumancausedremovalsandanimalscapturedusingtelemetry.Whenappliedtofour
strataconsistingoffemales4years,females5years,males4years,andmales5years,
thesummedchisquarestatisticsdividedbythetotaldegreesoffreedomestimatedavariance
inflationfactor( )of0.94(2=446.8,df=474).Similartopreviousanalyses(Regehretal.
2007),thissuggeststhatastandardCJSmodelprovidesanadequatefittothedatawhen
partitionedtoallowsexandagebasedvariationinsurvivalandrecaptureprobabilities.
BecausethemultistatemodelsusedhereweremoregeneralthanthestandardCJSmodel(e.g.,
theyallowedadditionalvariationasafunctionofreproductivestate),weconcludedthatthe
mostgeneralmultistatemodelwouldalsofitthedatawellandthereforedidnotincludea
varianceinflationfactorinanalyses.
Modelselectionandparameterestimation
Weusedthedevianceinformationcriterion(DIC)calculatedasGelmansapproximation
(Gelmanetal.2004)formodelselection,andcompletedmodelselectionforeachparameterin
asteppedfashion(Tables3,4).Inallcases,weusedageneralmodelfordetectionprobability,
p,whichincludedstateandageeffects,randomtimeeffects,andthetelemetry(femalesonly)
andChurchillcovariates.Wethenusedastepbystepmodelselectionapproachtoselectthe
mostsupportedmodelstructureforoneparameterwhileusingageneralstructurewithstate
andageeffectsfortheparametersthathadnotyetbeenevaluated.Thisapproachfocusedon
evaluatingpotentialrelationshipsbetweenenvironmentalconditionsandpolarbearvitalrates,
aprimaryobjectiveoftheanalysis. Byfittingarelativelysmallnumberofgeneralmodels,we
allowedformajorsourcesofvariationinthedatabasedonapriorihypothesesaboutbiology
13

andstudydesign,whilereducingthenumberof(computationallyintensive)modelsweran.
Formales,wefirstselectedthebestmodelstructureforsurvival,thenhumancausedmortality,
andthenweaning.Forfemaleswefirstselectedthebestmodelstructureforsurvival,then
humancausedmortality,breeding,twinning,andfinallyweaning.
WederivedabundanceestimatesfromthetoprankedmodelsusingtheHorvitz
Thompson(HT)estimator,similartoRegehretal.(2007)butextendedtotheBayesian
framework.Thisapproachestimatesabundanceattimetbydividinganindicatorvariablefor
eachindividual(equalto1iftheindividualwascapturedattimet,andequalto0otherwise)by
theestimatedrecaptureprobabilityforthatindividual,andsummingtheresultsoverall
individuals.ThisBayesianbasedapproachallowedforanestimationofvariancethat
accountedforthevariationinbothannualsamplesizeandrecaptureprobabilityandallowed
ustoestimatetheuncertaintyaroundpopulationsize.Becausemultistatemodelsconditionon
firstcaptureandthereforedonotproduceestimatesofpforCOYs,whicharebydefinitionfirst
timecaptures,weestimatedthenumberofCOYsateachsamplingoccasionbasedonthe
numberofadultfemaleswithoneortwoCOYs(stateF1YandF2Y)andthemeanlittersizefor
femalesinstateF2Y(2.04COYs,becausethisstateincludedafewfemaleswithtripletlitters).
Wecalculatedpopulationtrend,intermsofthemeangeometricobservedpopulationgrowth
rate,basedonabundanceestimates.Thiswasachievedbyexponentiatingthemeanofthe
ratiosofthenaturallogofabundanceforsequentialsamplingoccasions(Humbertetal.2009).
Populationprojectionmodeling
Wedevelopedapostbreedingmatrixbasedpopulationprojectionmodel(Caswell
2000)basedonthefemaleandmalelifecyclegraphs(Figures2,3)tobothestimateandproject
populationgrowthrate()usingthesurvivalandreproductiveparametersfromthemultistate
capturerecapturemodeling.Thisapproachusesthesuiteofvitalratesestimatedfromthe
multistatemodelframework,includingthepotentialeffectsofenvironmentalcovariateson
parameters,andallowsustoprojectpopulationgrowthundervaryingenvironmental
conditions.Weprojectedthepopulationsoastoaccountforsamplinguncertainty,
demographicstochasticity,andenvironmentalstochasticity.Samplinguncertaintywas
accountedforthroughinclusionofthefullsamplingdistributionasrepresentedbythesamples
14

intheMCMCchains.Demographicstochasticitywasaccountedforthroughinclusionof
binomialtrialsforeachofthedemographicprocesses(survival,weaning,breeding,etc.)and
environmentalstochasticitywasaccountedforbyselecting,ateachtimestepinthepopulation
projection,theestimatedsetofdemographicparameterscorrespondingtotheseaicevariables
atthattimestep.Webuilttwoseparateprojectionmodels,oneusingparameterestimates
fromthetoprankedmodelidentifiedbythemodelselectionproceduredescribedabove,and
oneusingaglobalmodelthatincludedallhypothesizedeffects.Thisprovidedageneral
assessmentofthesensitivityofpopulationprojectionstothestructureoftheestimation
model.
Weinitializedpopulationprojectionsusingthemeanstateandagecompositionofthe
populationforthethreeyearperiod19851987,ascalculatedfromthecapturesamplesand
estimatesofpfromthemostsupportedmodel,usingtheHorvitzThompsonestimator.This
proportionalcompositionwasappliedtothestartingyearforagivenprojectionbymultiplying
bytheestimatednumberoffemalesandmalesinthatyear.
Toevaluatethepopulationleveleffectsofpotentialchangesinenvironmental
conditions,weranthepopulationmodelunderthreedifferentseaicescenarios.First,we
sampledfromtheentiretimeseries(19842010)ofseaicevariables,withreplacement,for
each50yrrunofthepopulationmodel.Wesampledinamannertoretainthenatural
correlationsintheseaicevariables(i.e.,ifabreakupdatewasselectedforyeart,the
correspondingfreezeupdatewasselectedaswell).Second,weranahighseaicescenario,
whichincludedtheseaicevariablessampledfromthe19842010timeseriesusingthevalues
fromthoseyearsthatmeteachof3conditions:theywereintheupper50%quantileofbreak
up(i.e.,laterseaicebreakup),thelower50%quantileoffreezeup(i.e.,earlierfreezeup),and
thelower50%quantileofice.decay(i.e.,slowicedecay).Theseconditionsrepresentyears
withrelativelyhighavailabilityofseaiceandthusrepresentfavorableenvironmental
conditionsforpolarbearswithrespecttotheirabilitytoaccessprey.Last,weranalowsea
icescenario,whichincludedseaicevariablessampledfromoppositequantilestothoseusedin
thehighscenario,torepresenticeconditionsthatpreviousstudieshavesuggestedarelikely
tohavenegativeeffectsonpolarbears.
15

RESULTS
ThemodeledsubsetofWHdataconsistedofindividualcapturehistoriesfor3,034polar
bears,including6,224liveencounters(62ofwhichwereincidentalremovals)and519
purposefulhumancausedremovals.Oftheliveencounters,493werebearsagedoneyearor
olderthatweretargetedforcapturebytelemetry.Thecovariatetelemetryprovidedcoverage
forapproximately75%oftelemetryencounters,suggestingthatitexplainedmostofthe
individualvariationinpassociatedwithradioandsatellitetelemetry.Approximately34%of
individualbearswereencounteredbyMBatsomepointand,therefore,hadnonzeroentriesin
theChurchillcovariatesusedtomodelvariationinpandH.
Seaice
ChangesinseaicechronologyonwesternHudsonBayovertheperiod19792012
showedasignificanttrendtowardsearlierbreakupinthespringandlaterfreezeupinthe
autumn(Figure5).Thedateofseaicebreakuphasbeenoccurring5.5daysearlierperdecade
(t=3.359,p=0.002)andvariedfrom2Juneto11July,withameanof22June(SE=1.8days).
Thedateofseaicefreezeuphasbeenoccurring4.1dayslaterperdecade(t=2.655,p=0.013)
andvariedfrom10Novemberto16December,withameanof28November(SE=1.6days).
Overthelastdecade(20012010),thedateofseaicebreakupvariedbetween4June
and2Julybutdidnotexhibitasignificanttrend(linearregression,p=0.584).Similarly,thedate
ofseaicefreezeupvariedbetween23Novemberand9Decemberwithoutasignificanttrend
(linearregression,p=0.132).
Modelselection
Forfemalepolarbears,modelselectionledtothemostsupportedmodelM3(Table5)
withsurvival(S)asafunctionofstateandageeffects(COY,yearlingdependent,yearling
independentto4yr,519yrwithoutCOY,519yrwithCOY,20yrwithoutCOY,and20yr
withCOY)andthetimevaryingcovariatesbreakup,freezeup,andtheinteractionbetween
breakupandfreezeup;timeconstanthumancausedmortality(H)asafunctionofstateand
ageeffects(COYanddependentyearling,independentyearlingto4yr,5yrwithoutCOY,and
5yrwithCOY);timeconstantbreeding(B)asafunctionofstateandageeffects(4yr,59yr,
1019yr,and20yr);andtimeconstanttwinning(T)andweaning(W)probabilitieswithno
16

stateoragestructure.

Formalepolarbears,modelselectionledtothemostsupportedmodelM1(Table6)

withtimeconstantsurvival(S)asafunctionofstateandageeffects(COY,yearlingdependent,
yearlingindependentto4yr,59yr,1019yr,and20yr);timeconstanthumancaused
mortality(H)asafunctionofstateandageeffects(COYanddependentyearling,independent
yearlingto4yr,59yr,and10yr);andtimeconstantweaning(W)probabilitieswithnostate
oragestructure.
Parameterestimates
Forfemalepolarbears,totalapparentsurvival(S)variedwithtimeasafunctionofsea
iceconditions(Figure6,arepresentativegraphforadultfemales519yearsoldwithoutCOY).
Thestrongestseaicerelationshipwasbetweenearlierbreakupandlowersurvivalforallage
classes(Figure7,arepresentativegraphforadultfemales519yearsoldwithcubsoftheyear
andindependentfemales14yearsold).Recaptureprobabilitiesforadultfemalesvariedby
state:femalesaged519yearsoldwithoutCOY(andwithoutaradiocollarorpreviouscapture
inChurchill)hadp=0.11,comparedtop=0.37forfemaleswithCOY.Thissupportsthe
hypothesisoflowerpforpregnantadultfemalesthatmayseekrefugeinmaternaldensand
thereforebelesssusceptibletocapture.
Formalebears,themodelselectionprocesssupportedtimeconstantsurvival.
TimeinvariantestimatesofSwerederivedforfemalebearsusingtheMCMCchains
frommodelM3withoutthecoefficientsassociatedwithseaicecovariates(i.e.,usingthe
interceptandstateandagecoefficientsonly).Thisallowedforcomparisonofsurvivalamong
statesandageclasses,andwithtimeinvariantestimatesofSfromthemostsupportedmodel
M1formales(Table7).Whilewefoundevidenceforincreasingsurvivalwithage,followedby
senescentdeclinesforbears20yearsold,therewasnoindicationofdifferencesinadult
femalesurvivalasafunctionofreproductivestatus,withtheexceptionofhigherSforolder
adultfemaleswithcubscomparedtoolderadultfemaleswithoutcubs.Estimatesoftotal
apparentsurvivalwerelowerforallageclassesofindependentmalesthanforindependent
females,likelyduetotheeffectsofthesexselectiveharvest.

17

Estimatesofhumancausedmortality
Theprobabilityofhavingbeenpurposelykilledbyahuman,conditionalondying(H),
variedbysexandage(Table8).Forfemales,independentyearlingsthrough4yroldsexhibited
thehighestH.Formales,youngadultsaged59yearsexhibitedthehighestH,followedby
independentyearlingsthrough4yrolds.Thesepatternsareconsistentwithpreviouswork
suggestingthatyoungerbears,particularlymales,aremostlikelytoencroachuponhuman
settlementsandthusriskbeingkilled(LunnandStirling1985),andaredisproportionately
representedinthesubsistenceharvestduetoasexselectivemanagementapproach(Derocher
etal.1997;Tayloretal.2008;Peacocketal.2010).TherelativelyhighestimatesofHforsome
sexandageclassessuggestthatalargecomponentofoverallmortalityforsomesegmentsof
theWHsubpopulationisduetohumancausedremovals.Forexample,H=0.73foryoungadult
males(59years),indicatingthatforevery100youngadultmalesthatdieeachyear,
approximately73ofthemwerepurposelykilledbyhumans.
Reproductiveparameters
Estimatesofbreedingprobabilityforadultfemalesweretimeconstantanddidnotvary
asafunctionofseaiceconditions.Breedingprobabilityincreasedwithadultfemaleage,
followedbyasenescentdeclineforfemales20yearsold(Table9).Theprobabilityof
producingtwins(T)wasalsotimeconstantoverthecourseofthestudy,andconstantacross
statesandages,at0.47(95%BCI=0.430.52).Theprobabilityofweaning(W)forCOYswas
timeconstantat0.22(95%BCI=0.140.40)forfemalesand0.28(95%BCI=0.180.41)for
males.
Abundanceoftheeffectivestudypopulation
Abundanceestimatesfromthecapturerecapturestudyapplytoaneffectivestudy
populationdefinedasanimalswithanonnegligibleprobabilityofoccurringinthestudyarea
duringautumnsampling(i.e.,thesuperpopulationperKendalletal.1997).Thisdefinition
includesanimalswithlongtermfidelitytothestudyareathatwererandomlyoutsideofthis
areainanygivenyear,butdoesnotincludeanimalsthathadhighfidelitytootherareasand
wereconsequentlyunlikelytoeverenterthestudyareaorourcapturesample.Wederived
abundanceestimates(Figure8)usingtheHTestimatorappliedusingestimatesofpforthe
18

mostsupportedmodels;M3forfemalesandM1formales.Theestimateofoverallabundance
declinedfrom1185polarbears(95%CI:9931411)in1987to806bears(95%CI=653984)in
2011.Althoughtherewasvariationintheannualabundanceestimatesfrom20042011,the
periodoftimesincethepreviousdemographicassessmentoftheWHpopulation,therewasno
cleartrend(Figure8).
Althoughtheabundanceestimatesfromtheupdatedmodelingapproachfortheyears
19872004arelowerthanpreviousestimatesinRegehretal.(2007),properinferencecanonly
bemadewithineachofthemodels.Themeanobservedpopulationgrowthratefortheperiod
19872004forthecurrentstudyis0.973,comparedto0.981basedonpointestimatesof
populationsizefromRegehretal.(2007).Thisindicatesaconsistentevaluationofpopulation
trendbetweenthetwoanalyses.Amoredetailedcomparisonofpointestimatesofpopulation
size,onanannualbasis,iscomplicatedbydifferencesinthedatasetsandthemodeling
approaches.Inparticular,smallchangesinhowpismodeledcanresultinlargechangesinthe
abundanceestimates,becauseabundanceestimatesproducedinthiswayarehighlysensitive
toheterogeneityincapture(Williamsetal.2002).Althoughthecurrentestimateofabundance
in2011islowerthantheestimateof935(95%CI=7941076)fortheyear2004fromRegehret
al.(2007),comparisonofthesenumbersistenuous.Amorereliableindicatorofrecenttrendis
themeanobservedpopulationgrowthrate,whichis1.01overtheyears20042011.
Populationprojectionmodel
Asanadditionalandcomplementaryapproachforunderstandingpopulationtrends
(SandercockandBeissinger2002)andalsoforprojectingtrendsunderdifferentpossibleseaice
scenarios,weusedmatrixbasedpopulationprojectionmodelstoestimatepopulationgrowth
ratebasedonestimatesofsurvivalandothervitalratesfromthemultistatemodels.
Weusedpostbreedingmatrixbasedprojectionmodelstomodelpopulationgrowth
rate()overdifferenttimeframes(i.e.,pastandfuture)andoveravarietyofpossible
environmentalconditions.Theseestimatesare,therefore,derivedfromestimatesofSand
othervitalratesfromthemultistatemodels.
Forthecurrentanalysis,wefocusedonestimatingforthefemalesegmentofthe
population,duetothecriticalimportanceoffemalebearstoreproductionandpopulation
19

growth.Forfemales,weestimated=1.02(95%CI=0.981.06)fortheperiod20012010.This
isbasedonthemostsupportedfemalemodelM3andlikelyrepresentsthemostreliable
assessmentofrecentpopulationtrend,suggestingthatthefemalepopulationremainedstable
orincreasedslightlyduringthelastdecade.Althoughthisestimatereflectsbothnaturaland
purposefulhumancausedmortality,itappliestofemalesonlyanddoesnotreflectpotentially
lowerformalebearsduetotheeffectsofmalebiasedharvestmortality,assuggestedby
higherestimatesofHformalescomparedtofemales(themeanobservedpopulationgrowth
ratefortheentirepopulationoverthe20012010period,calculatedfromtheabundance
estimatesasdescribedabove,was0.99).Forcomparison,asimilarestimateof=1.02(95%CI
=0.981.06)wasderivedfromthemoregeneralfemalemodelthatincludediceeffectsonthe
reproductiveparametersB,W,andT.Thissuggeststhatestimatesofforfemalesarerobust
tothemodelselectionprocessandthatthemostsupportedmodeldidnotexclude,on
statisticalgrounds,potentialbiologicallymeaningfuleffectsonreproductiveparametersofthe
limitednumberofseaicemetricsthatweevaluated.
Wealsoprojectedpopulationsizeforwardintimeforthefemalesegmentofthe
population,andforfemalesandmalestogether(i.e.,thetotalpopulation),undertwo
assumptionsforfutureseaiceconditions(Table10).Thisrepresentsasensitivityanalysiswith
regardtotheprojectedtrendoftheWHsubpopulation,givenarangeofrelevantbut
hypotheticalfutureseaiceconditions.Thisapproachassumedthattherelationshipbetween
icecovariatesanddemographicparametersasestimatedfromthe19842011datawillremain
stable,aswouldotherfactorssuchasthelevelofhumancausedremovals.Forthecombined
femaleandmaleprojections,weusedparameterestimatesfromamalemodelthatincluded
seaiceeffectsonsurvival,asinfemales,toreflecttheirpotentialbiologicalimportance,despite
seaicecovariatesnotbeingsupportedduringthemodelselectionprocedure.Iffutureseaice
conditionsarehigh,asrepresentedbyasubsetofseaiceconditionsobservedfrom1984
2010,thelongtermpopulationgrowthrateisestimatedtobeapproximately=1.02,or2%
populationgrowthperyear.Similarly,ifseaiceconditionsarelow,longtermpopulation
growthrateisestimatedtobeapproximately=0.97,representinga3%populationdecline
peryear.
20

DISCUSSION
Populationtrend

OurupdatedanalysissuggestedthatthegrowthrateoftheWHsubpopulationhas,

overall,beennegativesincecapturerecapturestudiesbeganinthemid1980s,andhasbeen
relativelystableoverthelastdecade(Figure8).Themajordriverofpopulationchangefor
femalepolarbearswasthetimingofseaicebreakupandformation,whichwascorrelated
fromyeartoyearwithsurvivalofallageandreproductivecategoriesoffemales,butnotwith
annualreproductiverates(cubproductionandsurvivaltofirstautumn).Incontrast,the
survivalofmalepolarbearswasnotcloselylinkedtoseaice,butwasinsteadprimarilya
functionoftheirageclass.Thismaybearesultofhigherratesofdirecthumancausedmortality
inmalesthanfemales,whichfromyeartoyeararebasedonmanagementdecisionsrather
thanenvironmentalvariationandmayleadtocompensatoryeffectsduetodensitydependence
ortheselectiveremovalofanimalsinpoorernutritionalconditions.Theseresultsunderscore
theimportanceofconsideringenvironmentalfactorswithindemographicpopulationmodelsin
ordertoaccountforimportantdriversofdemographicrates.

Thelongtermresponseofpolarbearstoclimatechangeisexpectedtovaryintimeand

space,bothamongandwithinsubpopulations(Amstrupetal.2010;IUCN/SSCPolarBear
SpecialistGroup2013)andcasestudiescontinuetoimproveourunderstandingofthe
complexityofthespeciesresponsetoecologicalchange(Peacocketal.2012;Rodeetal.2012,
2013).Thisstudyrepresentsoneofthelongesttermandhighestresolutionanalysesofpolar
beardemographytodate,providinginsightintohowdemographicchangesmayoccurand
providingadditionalevidenceforlinkagesbetweenseaiceconditionsandpolarbearsurvival.
Thematrixprojectionmodelresultssuggestthatthelackofanegativetrendinspringseaice
breakupdatefrom20012010resultedinastablefemalepopulation,andthiswasgenerally
supportedbytherelativelystableabundanceestimates,suggestingconsistencybetweenthese
twodifferent(butnotindependent)methodsofinference.Itislikelythatthisperiodofrelative
stabilityinseaiceconditionsallowedforsufficientsurvivalandreproductionthathuman
causedremovaloffemaleswassustainableoverthisdecade.Evidenceforthedependenceof
theWHsubpopulationonseaiceconditions,combinedwithforecastsofdecreasingduration
21

andextentoficecoverinsouthernandwesternHudsonBayfromregionalclimatemodels(Joly
etal.2011),suggeststhatthelongtermpopulationtrendislikelytobenegative.However,our
resultssuggestthattheWHpopulationisabletorespondpositivelywhenclimaticandseaice
conditionsimprove;assumingthattheabundanceandavailabilityofpreypopulationsalso
respondsimilarly.Thishighlightsthesensitivityofpopulationprojectionstoassumptionsabout
futureenvironmentalconditionsandtherelationshipslinkingtheenvironmentalconditionsto
vitalrates.

Theestimatesoftotalapparentsurvivalusedinmatrixmodelsrepresenttheprobability

ofremainingaliveandinthestudyarea.Permanentemigrationoutofthestudyareawillresult
inestimatesofapparentsurvivalthatarelowerthannaturalsurvival,whichmayintroduce
negativebiasintopopulationprojectionsifnotcompensatedbyincludingestimatesof
immigrationintothestudyarea.Temporarymovementoutofthestudyareacanalsoresultin
biasedsurvivalestimatesinsomecases(Kendalletal.1997).Thedegreetowhichmovements
haveintroducedbiasintosurvivalestimatesinourcaseisnotknown,butthestrongsitefidelity
ofthesebearsduringautumn(DerocherandStirling1990;Stirlingetal.2004;Cherryetal.
2013),combinedwithperiodiccapturesofbearsoutsidethestudyarea(e.g.,inAreaD)likely
minimizestheseeffects.
Abundanceoftheeffectivestudypopulation
ItisimportanttoestimateboththeactualnumberofpolarbearswithintheWH
subpopulation(e.g.,pointestimatesofpopulationsize)andthetrendinnumbersovertime
(i.e.,populationgrowthrate)toinformharvestmanagementdecisions.Abundancecanbe
estimatedusingdifferentapproaches,suchascapturerecaptureanalysisandaerialsurveys,
whichprovidedifferenttemporalandspatialperspectivesandinvolvedifferentassumptions
andcaveats.Theestimatesofabundancedevelopedusingthecurrentcapturerecapture
frameworkaredefinedwithrespecttotheeffectivestudypopulationwhichmaybedifferent
thanthepopulationasdefinedfrombiologicalormanagementperspectivesanddependon
thespatialcoverageofthecapturerecapturesamplingarea,thetimingofcapturerecapture
sampling,andthemovementofanimalsinandoutofthisareaovermultipleyears.Estimates
ofabundancefromcapturerecapturestudiesmayalsobeinfluencedbyotherfactorsresulting
22

inunmodeledcaptureheterogeneity,whichcanintroducebiasthatisgenerallynegative
(Schaubetal.2004).Althoughwedonotknowthedegreetowhichourabundanceestimates
maybebiased,wedonotthinkitwasasignificantfactorinthisanalysisbecausetheestimated
pvalueswerereasonablyhighandbecausewetriedtominimizethepotentialeffectsof
captureheterogeneitybyincludingseveralcovariates(telemetry,Churchill)andbothstateand
ageeffectsinthemodels.
PreviouscapturerecaptureanalysesfortheWHsubpopulation(Lunnetal.1997;
Regehretal.2007)consideredestimatesofabundancetoreflectthepopulationasdefined
fromabiologicalormanagementperspective,whichwasbasedonearlierobservationsthat
relativelyfewpolarbearsexhibitedconsistentlyhighfidelitytoareasofthecoastthatwere
outsideoftheprimarystudyareabutstillwithintheWHmanagementboundary(Derocherand
Stirling1990;Stirlingetal.2004).However,arecentaerialsurvey(Stapletonetal.2014)found
evidenceforsignificantnumbersofbearslocatedoutsidetheECsamplingareatotheeastof
theNelsonRiverin2011,andestimatedalargersizeforthepopulationofpolarbears
inhabitingtheWHmanagementarea.Comparisonofthe2011pointestimateof806(95%CI=
653984)fromthisstudywiththeestimateof1030(95%CI=7541406)fromthe2011aerial
surveyrequirescarefulinterpretation.Theaerialsurveylikelyprovidesanaccuratesnapshot
estimateofthetotalnumberanddistributionofpolarbearsintheWHmanagementareaatthe
timeofthesurvey.Thisdiffersfromthepointestimateofpopulationsizefromcapture
recapturemodels,whichrepresentsthegroupofbearswithanonzeroprobabilityofmoving
throughthecapturerecapturesamplingarea(amuchsmallerareathanthatcoveredbythe
aerialsurvey;definedasthesuperpopulation;e.g.,Williamsetal.2002).Furthermore,the
2011populationsizeestimatesfromthetwoapproachescannotdefinitivelybesaidtobe
different,asevidencedbyoverlapintheirconfidenceintervals.Wesuggestthatconsidering
resultsfromthecapturerecaptureandaerialsurveystogether,keepingthestrengthsand
limitationsofeachmethodinmind,providesamorecompletepictureofthedistribution,
status,andtrendoftheWHpopulation.
Influenceofseaiceconditionsonsurvival
Wefoundthatsurvivaloffemalepolarbearsinallageclasseswascorrelatedwiththe
23

timingofseaicebreakup,freezeup,andtheinteractionofbreakupandfreezeup.Thisis
consistentwithpreviousstudiesthatlinkedbodyconditionandstatusofWHpolarbearsto
changesindurationofseaicecoverassociatedwithclimaticwarming(e.g.,Stirlingetal.1999;
Regehretal.2007).Continuingreductionofseaiceextentanddurationwouldrepresenta
significantthreattotheWHsubpopulation(StirlingandDerocher1993,2012;Derocheretal.
2004;Molnretal.2010;Peacocketal.2010;Molnretal.2011;CastrodelaGuardiaetal.
2013)becausepopulationgrowthrateforpolarbearsandsimilarlonglivedanimalsdepends
primarilyonsurvivalandproductivityofadultfemales(e.g.,Hunteretal.2010).
Trendsofdecliningseaicedurationanddeclinesinpolarbearsurvivalandreproduction
havealsobeendocumentedintheSouthernBeaufortSeasubpopulation(Regehretal.2010;
Rodeetal.2010).CorrelationsbetweenseaiceandbodyconditionintheBaffinBayandDavis
Straitsubpopulationssuggestenvironmentalconditionscanlimitthenutritionalstatusand
productivityofpolarbearpopulations(Rodeetal.2012),althoughtheroleofdensityeffects
arepoorlyunderstood,especiallyinDavisStraitwherepolarbearabundancehasincreased
(Peacocketal.2013).Environmentallimitationitselfisanaturalandubiquitousphenomenon
requiredtomaintainpopulationswithintherangeofenvironmentalcarryingcapacity;it
becomesaconservationconcernwhencarryingcapacityexhibitslongtermdeclinessuchas
appearstobethecasefortheWHandSouthernBeaufortSeasubpopulations.IntheChukchi
Seasubpopulation,polarbearsappeartohavemaintainedbodyconditionandreproduction
overa20yearperioddespitelargedeclinesinseaice(Rodeetal.2013),possiblyduetohigh
biologicalproductivity,therelativelyrecentadventoficefreedaysinpreferredpolarbear
habitatsoverthecontinentalshelf,ordensityeffectsassociatedwithpreviouslyhighlevelsof
humancausedremovals.IntheNorthernBeaufortSeaandSHsubpopulations,demographic
analysessuggestedequivocalorabsentsupportforrelationshipsbetweensurvivalandice
conditions,respectively,althoughconcernsremainregardingthelikelihoodofnegativeeffects
ofcontinuedseaicelossonpolarbearsinbothsubpopulations(Stirlingetal.2011;Obbardet
al.2007).InareassuchasHudsonBay,whereseasonalseaicedynamicsresultinextendedice
freeperiods,polarbearstransitionbetweenpositive(onice,feeding)andnegative(onshore,
fasting)energystates.Longerperiodsonshorewillresultingreaternegativeimpactson
24

energybudgetsandconsequentlyonsurvivalandproductivity(Molnretal.2010,2011;Castro
delaGuardiaetal.2013).
Despitethegrowingbodyofliteratureonmarinemammalresponsestoclimatechange,
seaicechange,andsubsequentshiftswithinArcticmarineecosystems(e.g.,Fergusonetal.
2005,LaidreandHeideJrgensen2005;KovacsandLydersen2008;Laidreetal.2008;Wiiget
al.2008;Molnretal.2010,2011;StirlingandDerocher2012),predictinglongterm
demographicchangesisdifficult.Therelationshipswedetectedbetweenseaiceandpolar
bearsinwesternHudsonBaymaychangeovertime.Forexample,theremaybecritical
thresholdsforthelengthofthefastingperiod,whichforWHpolarbearsisdeterminedbysea
iceconditions,beyondwhichbearscannotaccumulatesufficientbodyfattosurviveand
reproduce(Robbinsetal.2012).Furthermore,whiletheavailabilityofringedseals,themain
preyofpolarbearsinHudsonBay,islikelytobenegativelyaffectedbyclimatechange
(Fergusonetal.2005),thereareinsufficientdataavailabletoassesshoworatwhatratethese
andotherchangesatlowertrophiclevelswillfurtherimpactpolarbears(StirlingandDerocher
2012).Wesuggestthatvariabilityinrelationshipsbetweenseaiceandpolarbearpopulation
statusthathavebeenestimatedtodate,aswellasuncertaintyaboutthefutureformofsuch
relationships,doesnotdiminishlongtermconcerns(e.g.,Amstrupetal.2008;Stirlingand
Derocher2012)basedonforecastsofcontinuedseaiceloss(Hollandetal.2006;Zangand
Walsh2006;Stroeveetal.2007,2012)andthefundamentalrelianceofpolarbearsonseaice
asaplatformfromwhichtoaccessenergyrichmarineprey(Amstrup2003).Casestudiessuch
asthecurrentanalysesarekeytounderstandingthecomplexitiesofsuchrelationshipsand
providingmanagerswiththeinformationnecessarytodevelopconservationplansthatreflect
bothnearertermvariationandlongertermconcerns.
Influenceofseaiceandharvestonmalepolarbearsurvival

AlthoughRegehretal.(2007)foundeffectsofchangesindurationofseaiceonsurvival

ofdependent,juvenile,andsenescentmalepolarbears,wedidnotdetectarelationship
betweensurvivalofmalebearsandseaiceconditions.Thereareanumberofpossible
explanations.First,inrecentyearstheserelationshipsmayhaverelaxedgivenrelativelystable
seaiceconditions.Second,thecurrentapproachofmodelingfemaleandmaledataseparately
25

mayhaveimprovedourabilitytodetectsexspecificeffects.Third,thedirectinclusionof
detaileddataonhumancausedmortalitymayhavehadanimpactontheestimationofthese
effects.Inparticular,thenumberofmalebearskilledbyhumans(theprobabilityofadead
malebearhavingdiedduetodirecthumancausesrangedfrom0.24to0.73,dependingonage;
Table8)maybesufficientlylargetoincurcompensatoryeffectsthatdampenfluctuationsin
naturalsurvivalduetoenvironmentalvariation,makingrelationshipsbetweensurvivalandsea
icedifficulttodetect.Harvestoccursprimarilyintheautumnasbearsmovenorthwardalong
thecoastofwesternHudsonBayinanticipationoffreezeup.Theselectionformalesinthe
harvest,combinedwiththetendencyfornutritionallystressedbearstoapproachhuman
settlementswheretheyareexposedtoharvest,couldeffectivelyremovealargeproportionof
youngermalesthatotherwisewouldhavebeensusceptibletonaturalmortalityinthecoming
winter,thusmakingharvestatleastpartiallynonadditivetonaturalmortality.Inaddition,
largeradultmalepolarbearsmaybemorebufferedfromenvironmentalfluctuationscompared
tofemalesbecauseoftheextraenergeticdemandsonfemalesraisingcubs.Basedonbody
massandfatcontentofWHpolarbearscollectedinthe1980sand1990s,Robbinsetal.(2012)
estimatedthatadultmalescanspendtwiceaslongfasting(240days)aslactatingfemales(120
days)beforestarvationoccurs.Thissuggeststhatfullygrownadultmaleshavelowersize
specificenergeticdemandswhileonland,andmaybethegroupleastimpactedbychanging
seaiceconditionsexperiencedtodate.
Influenceofseaiceconditionsonreproduction

Wedidnotfindthatreproductionwascorrelatedwithvaryingseaiceconditionsover

theperiod19842011.Thismayhavebeeninfluencedbyheterogeneityinthebreeding
probabilityofFnYindividuals,astheyincludedbothsingleadultfemalesthatwerepotentially
pregnantandabletoproduceCOYthefollowingyearandadultfemaleswithdependent
yearlingsthatwereunlikelytobepregnant.Thisstatesimplificationwasnecessarytoavoid
buildingamodelincludingstateuncertainty,whichwouldhavesignificantlyincreasedthe
numberofunobservablestagestoconsider.Itwouldhavelikelydecreasedstatisticalpowerto
detectpatternsinsurvival,theprimarydriverofpopulationdynamicsforpolarbears,and
raisedidentifiabilityproblems(Gimenezetal.2009).Consequencesofthisapproachmayhave
26

beenreducedabilitytodetecttemporalvariationinbreedingprobability(B),aswellasa
nuancedinterpretationofB.AsBwasestimatedforamixedgroupoffemalesthatincluded
someindividualsthatwerephysiologicallyincapableofbeingpregnant,duetotheircontinued
careforyearlings,theparametercannotbeinterpreteddirectlyastheprobabilityofbreeding
forsingleadultfemales.TheseissuesshouldnotintroducebiasintoestimatesofBoroverall
estimatesofpopulationgrowthratefrommatrixprojectionmodels,althoughuseoftime
constantBmightslightlyobscuretemporalpatternsinpopulationgrowthrateoverthecourse
ofthestudy.Wealsonotethateffectsonbreedingprobabilitymaybeinherentlymoredifficult
todetect,becausebreedingprobabilityisestimatedfromasmallergroupofbears(onlythose
bearsintheFnYstate)andsomayhavelessstatisticalprecision,whereassurvivalisestimated
forallbears.ComparisonsofobservationsofmeanlittersizeinFB,SH,andWHintheearly
2000s(Peacocketal.2010)andmorerecently(Stapletonetal.2014)indicatethattheWH
subpopulationiscurrentlylessproductive.Therelativelylownumberofcubsrecordedduring
the2011WHaerialandcoastalsurveyswasnotedbyStapletonetal.(2014).

Thelatespringearlysummerperiodiscriticalforpolarbears,asitisduringthistime

thattheyaccumulateatleasttwothirdsoftheenergythattheyrequirefortheentireyear
(Stirlingandritsland1995).Itissuggestedthatchangesinbodyconditionofadultmales,
adultfemaleswithcubs,andsolitaryadultfemales(Stirlingetal.1999),anddeclinesinmassof
solitaryadultfemales(StirlingandParkinson2006),reflecttheamountoftimethatbearshave
spentontheseaicepriortocomingashore.Ouranalysisdidnotincludesuchmorphometric
covariates.Rodeetal.(2013)examinedbodysize,condition,andrecruitmentofpolarbearsin
twoadjacentsubpopulationsChukchiSeaandSouthernBeaufortSeaduringaperiodof
decliningseaicehabitat.Theyfounddifferingresponsesandconcludedthatdeclinesinseaice
extentdidnotcompletelyexplainobservedpopulationproductivity,andsuggestedthatpolar
bearsmayshowcomplexandnonlinearresponsestoclimatechangethatareinfluencedby
factorssuchasbiologicalproductivity.Werecommendthatfutureresearcheffortsconsider
incorporatingamechanisticenergeticsmodelwithinthedemographicframeworkusedto
projectfuturepopulationstatus(e.g.,Molnretal.2010).

Integratingpopulationinformationfrommultiplesourcesatdifferentspatialand
27

temporalscalesisnecessarytoeffectivelyunderstandthestatusandtrendinCanadaspolar
bearsubpopulations.Thestrengthofthemultistatemodelingapproachusedhereistheability
toassesslinkagesbetweenchangesinvitalrates,environmentalcorrelates,andpopulation
trend.Thisallowstheprocessesunderlyingpopulationchangetobeidentified,allowingforthe
developmentofappropriatemanagementactions.Thestrongdemographiclinkagebetween
seaiceconditionsandfemalesurvivalinWHpolarbearsenableseffectiveforecastingofthe
outcomeofdifferentmanagementscenariosandtheirimplicationsforsubsequentchangesin
populationsize,andprovidesadditionalevidenceofthesensitivityofthepopulationtrendsin
WesternHudsonBaytochangesinseaiceconditions.

ACKNOWLEDGEMENTS

FinancialandlogisticalsupportwasprovidedbyBuschEntertainmentCorporation/Sea

World,CarefortheWildInternational,theChurchillNorthernStudiesCentre,EarthRangers
Foundation,EnvironmentCanada,theIsdellFamilyFoundation,ManitobaConservation,
NationalFishandWildlifeFoundation,NaturalSciencesandEngineeringResearchCouncil,
NunavutWildlifeResearchTrustFund,ParksCanadaAgency,QuarkExpeditions,Schad
Foundation,U.S.FishandWildlifeService,WildlifeMedia,WorldWildlifeFundArctic
Programme,andWorldWildlifeFundCanada.Thisreportbenefittedfromusefulcommentson
earlierdraftsbyR.Elliot,E.Krebs,J.Nichols,andK.Rode.Manythankstothenumerousco
workers,fieldassistants,andhelicoptercrewswhomadeitpossibletocollectthreedecadesof
dataunderchallengingfieldconditions.

FinancialandlogisticalsupportfordataanalysiswasprovidedbytheUSFishand

WildlifeService,USGSPatuxentWildlifeResearchCenter,andtheColoradoCooperativeFish
andWildlifeResearchUnit.OurappreciationgoestoL.L.Bailey,W.A.Link,J.D.Nichols,J.A.
Royle,andM.C.Rungeforhelpfuldiscussionsandsupport.

Useoftrade,product,orfirmnamesisfordescriptivepurposesonlyanddoesnotimply

endorsementbytheCanadianorU.S.Governments.Thefindingsandconclusionsinthisreport
arethoseoftheauthorsanddonotnecessarilyrepresenttheviewsofEnvironmentCanadaor
theU.S.FishandWildlifeService.
28


LITERATURECITED
Amstrup,S.C.2003.Polarbear,Ursusmaritimus.Pp.587610inG.A.Feldhamer,B.C.
Thompson,andJ.A.Chapman,eds.MammalsofNorthAmerica:Biology,Management,
andConservation.JohnHopkinsUniversityPress,Baltimore,USA.
Amstrup,S.C.,E.T.DeWeaver,D.C.Douglas,B.G.Marcot,G.M.Durner,C.M.Bitz,andD.A.
Bailey.2010.Greenhousegasmitigationcanreduceseaicelossandincreasepolarbear
persistence.Nature468:955U351.
Amstrup,S.A.,B.G.Marcot,andD.C.Douglas.2008.ABayesiannetworkmodelingapproach
toforecastingthe21stcenturyworldwidestatusofpolarbears.Pp.213268inE.T.
DeWeaver,C.M.Bitz,andL.B.Tremblay,eds.ArcticSeaIceDecline:Observations,
Projections,Mechanisms,andImplications.GeophysicalMonographSeries180.
AmericanGeophysicalUnion,Washington,USA.
Arnason,A.N.1972.Parameterestimatesfrommarkrecaptureexperimentsontwo
populationssubjecttomigrationanddeath.ResearchesonPopulationEcology13:97
113.
Atkinson,S.N.,andM.A.Ramsay.1995.Theeffectsofprolongedfastingonthebody
compositionandreproductiveperformanceofpregnantfemalepolarbears(Ursus
maritimus).FunctionalEcology9:559567.
Burnham,K.P.,D.R.Anderson,G.C.White,C.Brownie,andK.H.Pollock.1987.Designand
analysismethodsforfishsurvivalexperimentsbasedonreleaserecapture.American
FisheriesSocietyMonograph5,Bethesda,Maryland,USA.
Calvert,W.,andM.A.Ramsay.1988.Evaluationofagedeterminationofpolarbearsbycounts
ofcementumgrowthlayergroups.Ursus10:449453.
Calvert,W.,M.Taylor,I.Stirling,G.B.Kolenosky,S.Kearney,M.Crte,andS.Luttich.1995.
PolarbearmanagementinCanada198892.Pp.6179in.Wiig,E.W.Born,andG.W.
Garner,eds.PolarBears:ProceedingsoftheEleventhWorkingMeetingoftheIUCN/SSC
PolarBearSpecialistGroup.IUCN,Gland,SwitzerlandandCambridge,UK.
CastrodelaGuardia,L.,A.E.Derocher,P.G.Myers,A.D.TerwisschavanScheltinga,andN.J.
29

Lunn.2013.FutureseaiceconditionsinwesternHudsonBayandconsequencesfor
polarbearsinthe21stcentury.GlobalChangeBiology19:26752687.
Caswell,H.2000.Prospectiveandretrospectiveperturbationanalyses:theirrolesin
conservationbiology.Ecology81:619627.
Cattet,M.R.L.,S.N.Atkinson,S.C.Polischuk,andM.A.Ramsay.1997.Predictingbodymassin
polarbears:ismorphometryuseful?JournalofWildlifeManagement61:10831090.
Cavalieri,D.J.,C.L.Parkinson,N.DiGirolamo,andA.Ivanoff.2012.Intersensorcalibration
betweenF13SSMIandF17SSMISforglobalseaicedatarecords.IEEEGeoscienceand
RemoteSensingLetters9:233236.
Cherry,S.G.,A.E.Derocher,G.W.Thiemann,andN.J.Lunn.2013.Migrationphenologyand
seasonalfidelityofanArcticmarinepredatorinrelationtoseaicedynamics.Journalof
AnimalEcology82:912921.
Converse,S.J.,W.L.Kendall,P.F.Doherty,Jr,andP.G.Ryan.2009.Multistatemodelsfor
estimationofsurvivalandreproductionintheGreyHeadedAlbatross(Thalassarche
chrysostoma).TheAuk126:7788.
Derocher,A.E.,N.J.Lunn,andI.Stirling.2004.Polarbearsinawarmingclimate.Integrative
andComparativeBiology44:163176.
Derocher,A.E.,andI.Stirling.1990.Distributionofpolarbears(Ursusmaritimus)duringthe
icefreeperiodinwesternHudsonBay.CanadianJournalofZoology68:13951403.
Derocher,A.E.,andI.Stirling.1995.Estimationofpolarbearpopulationsizeandsurvivalin
westernHudsonBay.JournalofWildlifeManagement59:215221.
Derocher,A.E.,I.Stirling,andW.Calvert.1997.Malebiasedharvestingofpolarbearsin
westernHudsonBay.JournalofWildlifeManagement61:10751082
Etkin,D.A.1991.BreakupinHudsonBay:itssensitivitytoairtemperaturesandimplications
forclimatewarming.ClimatologicalBulletin25:2134.
Ferguson,S.H.,I.Stirling,andP.McLoughlin.2005.Climatechangeandringedseal(Phoca
hispida)recruitmentinwesternHudsonBay.MarineMammalScience21:121135.
Gagnon,A.S.,andW.A.Gough.2005.Trendsindatesoficefreezeupandbreakupover
HudsonBay,Canada.Arctic58:370382.
30

Gelman,A.1996.Inferenceandmonitoringconvergence.Pp.131143inW.R.Gilks,S.
Richardson,andD.J.Spiegelhalter,eds.MarkovChainMonteCarloinPractice.
ChapmanandHall,London,UK.
Gelman,A.,J.B.Carlin,H.S.Stern,andD.B.Rubin.2004.BayesianDataAnalysis.Second
Edition.Chapman&Hall/CRC,BocaRaton,USA.
Gelman,A.,andJ.Hill.2007.DataAnalysisUsingRegressionandMultilevel/Hierarchical
Models.CambridgeUniversityPress,NewYork,USA.
Gimenez,O.,B.J.T.Morgan,andS.P.Brooks.2009.Weakidentifiabilityinmodelsformark
recapturerecoverydata.Pp.10551067inD.L.Thomson,E.G.Cooch,andM.J.Conroy,
eds.ModelingDemographicProcessesinMarkedPopulations.Springer,NewYork,USA.
Hestbeck,J.B.,J.D.Nichols,andR.Malecki.1991.Estimatesofmovementandsitefidelity
usingmarkresightdataofwinteringCanadageese.Ecology72:523533
Hochheim,K.,D.G.Barber,andJ.V.Lukovich.2010.ChangingseaiceconditionsinHudsonBay,
19802005.Pp.3951inS.H.Ferguson,L.L.Loseto,andM.L.Mallory,eds.ALittleLess
Arctic:TopPredatorsintheWorldsLargestNorthernInlandSea.Springer,NewYork,
USA.
Hochheim,K.P.,J.V.Lukovich,andD.G.Barber.2011.Atmosphericforcingofseaiceinhudson
bayduringthespringperiod,19802005.JournalofMarineSystems88:476487.
Holland,M.M.,C.M.Bitz,andB.Tremblay.2006.Futureabruptreductionsinthesummer
arcticseaice.GeophysicalResearchLetters33,L23503,doi:10.1029/2006GL028024.
Humbert,J.Y.,L.S.Mills,J.S.Horne,andB.Dennis.2009.Abetterwaytoestimatepopulation
trends.Oikos118:19401946.
Hunter,C.M.,H.Caswell,M.C.Runge,E.V.Regehr,S.C.Amstrup,andI.Stirling.2010.Climate
changethreatenspolarbearpopulations:astochasticdemographicanalysis.Ecology
91:28832897.
IUCN/SSCPolarBearSpecialistGroup.2013.Summaryofpolarbearpopulationstatusasper
2013.IUCN/SSCPolarBearSpecialistGroup,http://pbsg.npolar.no/en/status/status
table.html(accessed23rdFebruary2014).
Joly,S.,S.Senneville,D.Caya,andF.J.Saucier.2011.SensitivityofHudsonBaySeaiceand
31

oceanclimatetoatmospherictemperatureforcing.ClimateDynamics36:18351849.
Kearney,S.R.1989.ThePolarBearAlertProgramatChurchill,Manitoba.Pp.8392inM.
Bromley,ed.BearPeopleConflicts:ProceedingsofaSymposiumonManagement
Strategies.NorthwestTerritoriesDepartmentofRenewableResources,Yellowknife.
Kendall,W.L.,P.B.Conn,andJ.E.Hines.2006.Combiningmultistatecapturerecapturedata
withtagrecoveriestoestimatedemographicparameters.Ecology87:169177.
Kendall,W.L.,J.D.Nichols,andJ.E.Hines.1997.Estimatingtemporaryemigrationusing
capturerecapturedatawithPollocksrobustdesign.Ecology78:563578.
Kry,M.,andM.Schaub.2012.BayesianPopulationAnalysisUsingWinBUGS:AHierarchical
Perspective.AcademicPress,SanDiego,USA.
Kovacs,K.M.,andC.Lydersen.2008.ClimatechangeimpactsonsealsandwhalesintheNorth
AtlanticArcticandadjacentshelfseas.ScienceProgress91:117150.
Laidre,K.L.,andM.P.HeideJrgensen.2005.Arcticseaicetrendsandnarwhalvulnerability.
BiologicalConservation121:509517.
Laidre,K.L.,I.Stirling,L.F.Lowry,.Wiig,M.P.HeideJrgensen,andS.H.Ferguson.2008.
QuantifyingthesensitivityofArcticmarinemammalstoclimateinducedhabitatchange.
EcologicalApplications18(Supplement):S97S125.
Lunn,N.J.,andI.Stirling.1985.Thesignificanceofsupplementalfoodtopolarbearsduringthe
icefreeperiodofHudsonBay.CanadianJournalofZoology63:22912297.
Lunn,N.J.,I.Stirling,D.Andriashek,andG.B.Kolenosky.1997.Reestimatingthesizeofthe
polarbearpopulationinWesternHudsonBay.Arctic50:234240.
McCarthy,M.A.2007.Bayesianmethodsforecology.CambridgeUniversityPress,Cambridge,
UK.
Molnr,P.K.,A.E.Derocher,G.W.Thiemann,andM.A.Lewis.2010.Predictingsurvival,
reproduction,andabundanceofpolarbearsunderclimatechange.Biological
Conservation143:16121622.
Molnr,P.K.,A.E.Derocher,T.Klanjscek,andM.A.Lewis.2011.Predictingclimatechange
impactsonpolarbearlittersize.NatureCommunications2:186.
Nirlungayuk,G.,andD.S.Lee.2009.ANunavutInuitperspectiveonWesternHudsonBaypolar
32

bearmanagementandtheconsequencesforconservationhunting.Pp.135142inM.
M.R.Freeman,andL.Foote,eds.Inuit,PolarBears,andSustainableUse:Local,National
andInternationalPerspectives.CCIPress,UniversityofAlberta,Edmonton,Canada.
Obbard,M.E.,T.L.McDonald,E.J.Howe,E.V.Regehr,andE.S.Richardson.2007Polarbear
populationstatusinSouthernHudsonBay,Canada.U.S.GeologicalSurvey
AdministrativeReport.U.S.GeologicalSurvey,Reston,VA.
Obbard,M.E.,G.W.Thiemann,E.Peacock,andT.D.DeBruyn.(Eds.).2010.PolarBears:
ProceedingsoftheFifteenthWorkingMeetingoftheIUCN/SSCPolarBearSpecialist
Group.IUCN,Gland,SwitzerlandandCambridge,UK.
Peacock,E.,A.E.Derocher,N.J.Lunn,andM.E.Obbard.2010.Polarbearecologyand
managementinHudsonBayinthefaceofclimatechange.Pp.93115inS.H.Ferguson,
L.L.Loseto,andM.L.Mallory,eds.ALittleLessArctic:TopPredatorsintheWorlds
LargestNorthernInlandSea.Springer,NewYork,USA.
Peacock,E.,A.E.Derocher,G.W.Thiemann,andI.Stirling.2011.Conservationand
managementofCanadaspolarbears(Ursusmaritimus)inachangingArctic.Canadian
JournalofZoology89:371385.
Peacock,E.,J.Laake,K.L.Laidre,E.W.Born,andS.N.Atkinson.2012.Theutilityofharvest
recoveriesofmarkedindividualstoassesspolarbear(Ursusmaritimus)survival.Arctic
65:391400.
Peacock,E.,andM.Taylor.2007.PolarbearsofwesternHudsonBay:surveyextension
investigation.GovernmentofNunavutunpublishedreport.
Peacock,E.,M.K.Taylor,J.Laake,andI.Stirling.2013.Populationecologyofpolarbearsin
DavisStrait,CanadaandGreenland.TheJournalofWildlifeManagement77:463476.
Plummer,M.2003.JAGS:aprogramforanalysisofBayesiangraphicalmodelsusingGibbs
sampling.Proceedingsofthe3rdInternationalWorkshoponDistributedStatistical
Computing(DSC2003),March2022,Vienna,Austria.ISSN1609395X.
Plummer,M.2013.Packagerjags:BayesiangraphicalmodelsusingMCMC.Version3.10.
Availableathttp://sourceforge.net/projects/mcmcjags/files/.
Polischuk,S.C.,R.J.Norstrom,andM.A.Ramsay.2002.Bodyburdensandtissue
33

concentrationsoforganochlorinesinpolarbears(Ursusmaritimus)varyduringseasonal
fasts.EnvironmentalPollution118:2939.
Pradel,R.2005.Multievent:Anextensionofmultistatecapturerecapturemodelstouncertain
states.Biometrics61:442447.
Prestrud,P.,andI.Stirling.1994.TheInternationalPolarBearAgreementandthecurrentstatus
ofpolarbearconservation.AquaticMammals20:113124.
RDevelopmentCoreTeam.2012.R:Alanguageandenvironmentforstatisticalcomputing.R
FoundationforStatisticalComputing,Vienna,Austria.ISBN3900051070,URL:
http://www.Rproject.org.
Ramsay,M.A.,andI.Stirling.1986.Onthematingsystemofpolarbears.CanadianJournalof
Zoology64:21422151.
Ramsay,M.A.,andI.Stirling.1988.Reproductivebiologyoffemalepolarbears(Ursus
maritimus).JournalofZoology,London214:601634.
Regehr,E.V.,C.M.Hunter,H.Caswell,S.C.Amstrup,andI.Stirling.2010.Survivalandbreeding
ofpolarbearsinthesouthernBeaufortSeainrelationtoseaice.JournalofAnimal
Ecology79:117127.
Regehr,E.V.,N.J.Lunn,S.C.Amstrup,andI.Stirling.2007.Effectsofearlierseaicebreakupon
survivalandpopulaitonsizeofpolarbearsinwesternHudsonBay.JournalofWildlife
Management71:26732683.
Robbins,C.T.,C.LopezAlfaro,K.D.Rode,.Tien,andO.L.Nelson.2012.Hibernationand
seasonalfastinginbears:theenergeticcostsandconsequencesforpolarbears.Journal
ofMammalogy93:14931503.
Rode,K.D.,S.C.Amstrup,andE.V.Regehr.2010.Reducedbodysizeandcubrecruitmentin
polarbearsassociatedwithseaicedecline.EcologicalApplications20:768782.
Rode,K.D.,E.Peacock,M.Taylor,I.Stirling,E.W.Born,K.L.Laidre,and.Wiig.2012.Ataleof
twopolarbearpopulations:icehabitat,harvest,andbodycondition.PopulationEcology
54:318.
Rode,K.D.,E.V.Regehr,D.C.Douglas,G.Durner,A.E.Derocher,G.W.Thiemann,andS.M.
Budge.2013.VariationintheresponseofanArctictoppredatorexperiencinghabitat
34

loss:feedingandreproductiveecologyoftwopolarbearpopulations.GlobalChange
Biologydoi:10.1111/gcb.12339.
Royle,J.A.,andR.M.Dorazio.2008.HierarchicalModelingandInferenceinEcology:The
AnalysisofDatafromPopulations,MetapopulationsandCommunities.AcademicPress,
SanDiego,USA.
Sandercock,B.K.,andS.R.Beissinger.2002.Estimatingratesofpopulationchangefora
Neotropicalparrotwithratio,markrecaptureandmatrixmethods.JournalofApplied
Statistics29:589607.
Schaub,M.,O.Gimenez,B.R.Schmidt,andR.Pradel.2004.Estimatingsurvivalandtemporary
emigrationinthemultistatecapturerecaptureframework.Ecology85:21072113.
Servanty,S.,S.J.Converse,andL.L.Bailey.2014.Demographyofareintroducedpopulation:
movingtowardmanagementmodelsforanendangeredspecies,thewhoopingcrane.
EcologicalApplications24:927937.
Skinner,W.R.,R.L.Jefferies,T.J.Carleton,R.F.Rockwell,andK.F.Abraham.1998.Prediction
ofreproductivesuccessandfailureinlessersnowgeesebasedonearlyseasonclimatic
variables.GlobalChangeBiology4:316.
Stapleton,S.,S.Atkinson,D.Hedman,andD.Garshelis.2014.RevisitingWesternHudsonBay:
Usingaerialsurveystoupdatepolarbearabundanceinasentinelpopulation.Biological
Conservation170:3847.
Stirling,I.andA.E.Derocher.1993.Possibleimpactsofclimaticwarmingonpolarbears.Arctic
46:240245.
Stirling,I.andA.E.Derocher.2012.Effectsofclimatewarmingonpolarbears:areviewofthe
evidence.GlobalChangeBiology18:26942706.
Stirling,I.,C.Jonkel,P.Smith,R.Robertson,andD.Cross.1977.Theecologyofthepolarbear
(Ursusmaritimus)alongthewesterncoastofHudsonBay.CanadianWildlifeService
OccasionalPaper33,64pages.
Stirling,I.,N.J.Lunn,andJ.Iacozza.1999.Longtermtrendsinthepopulationecologyofpolar
bearsinwesternHudsonBayinrelationtoclimaticchange.Arctic52:294306.
Stirling,I.,N.J.Lunn,J.Iacozza,C.Elliott,andM.Obbard.2004.Polarbeardistributionand
35

abundanceontheSouthwesternHudsonBayCoastduringopenwaterseason,in
relationtopopulationtrendsandannualicepatterns.Arctic57:1526.
Stirling,I.,T.L.McDonald,E.S.Richardson,E.V.Regehr,andS.C.Amstrup.2011.Polarbear
populationstatusinthenorthernBeaufortSea,Canada,19712006.Ecological
Applications21:859876.
Stirling,I.,andN.A.ritsland.1995.Relationshipsbetweenestimatesofringedseal(Phoca
hispida)andpolarbear(Ursusmaritimus)populationsintheCanadianArctic.Canadian
JournalofFisheriesandAquaticSciences52:25942612.
Stirling,I.andC.L.Parkinson.2006.Possibleeffectsofclimatewarmingonselectedpopulations
ofpolarbears(Ursusmaritimus)intheCanadianArctic.Arctic59:261275.
Stirling,I.,C.Spencer,andD.Andriashek.1989.Immobilizationofpolarbears(Ursusmaritimus)
withTelazolintheCanadianArctic.JournalofWildlifeDiseases25:159168.
Stroeve,J.,M.M.Holland,W.Meier,T.Scambos,andM.Serreze.2007.Arcticseaicedecline:
fasterthanforecast.GeophysicalResearchLetters34,L09501,
doi:10.1029/2007GL029703.
Stroeve,J.C.,V.Kattsov,A.Barrett,M.Serreze,T.Pavlova,M.Holland,andW.N.Meier.2012.
TrendsinArcticseaiceextentfromCMIP5,CMIP3andobservations.Geophysical
ResearchLetters39,L16502.doi:10.1029/2012GL052676,2012.
Taylor,M.,andJ.Lee.1995.DistributionandabundanceofCanadianpolarbearpopulationsa
managementperspective.Arctic48:147154.
Taylor,M.K.,P.D.McLoughlin,andF.Messier.2008.Sexselectiveharvestingofpolarbears
Ursusmaritimus.WildlifeBiology14:5260.
Tyrrell,M.2006.Morebears,lessbears:Inuitandscientificperceptionsofpolarbear
populationsonthewestcoastofHudsonBay.JournalofInuitStudies30:191208.
Vongraven,D.,andE.Peacock.2011.DevelopmentofapanArcticmonitoringplanforpolar
bears:backgroundpaper.CircumpolarBiodiversityMonitoringProgramme,CAFF
MonitoringSeriesReportNo.1,CAFFInternationalSecretariat,Akureyri,Iceland.
Wade,P.R.2002.Bayesianpopulationviabilityanalysis.Pp.213238inS.R.Beissinger,andD.
R.McCullough,eds.PopulationViabilityAnalysis.UniversityofChicagoPress,Chicago,
36

USA.
Wiig,.,J.Aars,andE.W.Born.2008.Effectsofclimatechangeonpolarbears.Science
Progress91:151173.
Wiig,.,E.W.Born,andG.W.Garner.(Eds.).1995.PolarBears:ProceedingsoftheEleventh
WorkingMeetingoftheIUCN/SSCPolarBearSpecialistGroup.IUCN,Gland,Switzerland
andCambridge,UK.
Williams,B.K.,J.D.Nichols,andM.J.Conroy.2002.AnalysisandManagementofAnimal
Populations.AcademicPress,SanDiego,USA.
Zhang,X.,andJ.E.Walsh.2006.TowardaseasonallyicecoveredArcticocean:scenariosfrom
theIPCCAR4modelsimulations.JournalofClimate19:17301747.

37


Figure1.MapofHudsonBayshowingthemanagementboundaryoftheWesternHudsonBay
polarbearsubpopulation(dashedline)andareasofresearchandsurveyeffort.Mostofthe
capturerecapturedatacomefromanimalshandledinAreaC(studyarea),withlimited
capturesinAreaB.Geographiccoverageforcapturerecapturedatawasextendedtoinclude
AreaDin198486,199495,and200305foranimalsthathadfirstbeencapturedinAreasB
andC.The2011aerialsurveycoveredtheAreasA,B,C,andD(Stapletonetal.2014).
38

F1Y
S*B*(1T)
F1I
S*W

S*B*(1T)

S
F2

FC

F3

F4

S*(1B)

S*(1W)

S*(1B)
FnY

F1D

S*B*T

S*B*T

F2Y

Figure2.Multistatemodelstructureforthefemalecompartment.Parametersaredefinedin
Table1.

39

M1I

S*W
S

S
MC
MA

S*(1W)
M1D

Figure3.Multistatemodelstructureforthemalecompartment.Parametersaredefinedin
Table1.

40

All

(1S)*H
States

FoDor

MoD

All
States
1

FuDor
(1S)*(1H)

MuD

Figure4.Multistatemodelstructureformortality.ParametersaredefinedinTable1.

41

Day of year

200

(a)Breakup:y=0.551x+1273.4

190
180
170
160
150
1980

1985

1990

1995

2000

2005

2010

2000

2005

2010

Year

Day of year

340

(b)Freezeup:y=0.412x490.0

330

320

310

300
1980

1985

1990

1995

Year
Figure5.Dateof(a)seaicebreakup(50%seaiceconcentration)inspringand(b)seaice
freezeup(50%seaiceconcentration)inautumninwesternHudsonBayfrom19792012,
estimatedfrompassivemicrowavesatelliteimagery(datasource:NationalSnowandIceData
Center,Boulder,Colorado;http://nsidc.org).

42

Figure6.Totalapparentsurvivalforadultfemalesaged519yearsoldandwithoutCOYs,
estimatedfromliverecaptureanddeadrecoverydatafortheWesternHudsonBaypolarbear
subpopulationfrom19842011usingmultistatecapturerecapturemodels.Thisstudyshowed
thatinterannualvariationinsurvivalisafunctionofseaiceconditions.

43

1.0

(a)Independentfemalebears,14yearsold

Survival

0.9

0.8

0.7

0.6
150

155

160

165

170

175

180

185

190

195

190

195

Ordinal date of sea ice break-up

1.0

Survival

0.9

0.8

0.7

(b)Adultfemalebears,519yearsold,withcubsoftheyear
0.6
150

155

160

165

170

175

180

185

Ordinal date of sea ice break-up

Figure7.Survivalratesof(a)independentfemalepolarbears14yearsoldand(b)adult
femalebears519yearsoldwithcubsoftheyearinrelationtodateofseaicebreakup,
westernHudsonBay,19842011.

44

1500

Number of polar bears

1200

900

600

300

0
1985

1990

1995

2000

2005

2010

Year
Figure8.Estimatedpopulationsize,derivedbyapplyingaHorvitzThompsonestimatorusing
recaptureprobabilitiesestimatedfromliverecaptureanddeadrecoverydatafortheWestern
HudsonBaypolarbearsubpopulationfrom19842011,usingmultistatecapturerecapture
models.Pointestimatesofabundanceand95%confidenceintervalsareshownfor19872011
only,becausethe19851986pointestimateswerebiasedbyincompletesamplingofthecore
studyareaandarenotcomparable(Regehretal.2007).

45

Table1.ParametersestimatedfromliverecaptureanddeadrecoverydatafortheWestern
HudsonBaypolarbearsubpopulationfor19842011,usingmultistatecapturerecapture
modelsbasedonthelifecyclegraphsinFigures24.

Parameter
Description
S
Totalapparentsurvival:theprobabilitythatanindividualaliveintheautumn
ofyeartsurvivestotheautumnofyeart+1anddoesnotpermanently
emigratefromthestudypopulation
W
Weaning:theprobabilitythatadependent9mo.cubintheautumnofyeart
(stateFCorMC)becomesanindependentyearlingintheautumnofyeart+1
(stateF1IorM1I),conditionalonsurvival
B
Breeding:theprobabilitythatanadultfemalegivesbirthinthespringof
yeartandthatatleastonememberofalittersurvivesuntiltheautumnof
yeart
T
Twinning:theprobabilitythattwoormoremembersofalittersurviveuntil
autumn,conditionalontheadultfemalegivingbirthinthespring,andat
leastonememberofthelittersurvivesuntilautumn
H
Humancausedmortality:theprobabilitythatanindividualthatdiesinthe
intervalttot+1waspurposelykilledbyahuman(e.g.,subsistenceharvest
ordefensekill)
p
Recapture:theprobabilitythatanindividualwasrecapturedandreleased
aliveonsamplingoccasiont,conditionalonbeingaliveandnothaving
permanentlyemigratedfromthestudypopulation

46

Table2.Covariatesandeffectsusedtoexplainvariationinparametersestimatedfromlive
recaptureanddeadrecoverydatafortheWesternHudsonBaypolarbearsubpopulationfor
19842011.

Covariateoreffect
Description
age.maleeffects
AdditionalageeffectswithintheadultmalestateMAconsistingof
subadults(24yr),youngadults(59yr),primeadults(1019yr),and
senescentadults(20yr).
age.femaleeffects
Additionalageeffectswithintheadultfemalestates(FnY,F1Y,F2Y)
consistingofyoungadults(59yr),primeadults(1019yr),and
senescentadults(20yr).
churchill
Individualandtimevaryingcovariate.Thevaluewas0ifan
individualhadneverbeencapturedaroundthecommunityof
Churchill,and1forallsamplingoccasionsfollowingthefirstcapture
aroundChurchill.
telemetry
Individualandtimevaryingcovariate,appliedonlytoadultfemales
5years.Thevaluewas1ifafemalewasequippedwitha
functionalradiocollarandavailableforrecaptureusingVHFor
satellitetelemetry,and0otherwise.
breakup
Juliandateforcalendaryeartonwhichseaiceextentinthe
WesternHudsonBaymanagementareadeclinedbelow50%
coverage.
freezeup
Juliandateforcalendaryeartonwhichseaiceextentinthe
WesternHudsonBaymanagementareaincreasedtoabove50%
coverage.
icedecay
Absolutevalueincalendaryeartfortheslopeoftheicedecay
functionbetweenMay1andthedateonwhichtheWesternHudson
Baymanagementareaiscompletelyicefree.
randomtime
Arandomeffectofyear,includedonlyinthemodelofdetection
probability.

47

Table3.Steppedmodelselectionforfemalepolarbears.

Firststep:Settingupthefirstmodelwithwhichweregoingtodothecomparison
Modelnotation
Effectsincludedintheprobabilityofrecapture
M1
p=f(State/ageeffect+Churchillrecap+telemetry+randomtime)
S=f(State/ageeffect)
H=f(State/ageeffect)
B=f(State/ageeffect)
T,W
Secondstep:Selectionofthebestmodelforsurvival
Modelnotation
Effectsincludedinsurvival
M2
S=f(State/ageeffect+Breakup)
M3
S=f(State/ageeffect+Breakup+Freezeup+Breakup*Freezeup)
M4
S=f(State/ageeffect+Breakup+Icedecay+Breakup*Icedecay)
Thirdstep:Selectionofthebestmodelfortheprobabilityofdyingduetohumancauses
Modelnotation
Effectsincludedintheprobabilityofdyingduetohunting
M5
H=f(State/ageeffect+Churchillhunting)
Fourthstep:Selectionofthebestmodelfortheprobabilitytoreproduceandhavealive9mo
oldcub
Modelnotation
Effectincludedintheprobabilitytoreproduceandhavealive9moold
cub
M6
B=f(State/ageeffect+Breakup)
M7
B=f(State/ageeffect+Breakup+Icedecay+Breakup*Icedecay)
Fifthstep:Selectionofthebestmodelfortheprobabilityoftwinning
Modelnotation
Effectincludedintheprobabilityoftwinning
M8
T=f(State/ageeffect+Breakup)
Sixthstep:Selectionofthebestmodelfortheprobabilityofweaning
Modelnotation
Effectincludedintheprobabilityofweaning
M9
W=f(Breakup)

48

Table4.Steppedmodelselectionformalepolarbears.

Firststep:Settingupthefirstmodelwithwhichweregoingtodothecomparison
Modelnotation Effectsincludedinthedifferentparameters
M1
p=f(State/ageeffect+Churchillrecap+randomtime)
S=f(State/ageeffect)
H=f(State/ageeffect)
W
Secondstep:Selectionofthebestmodelforsurvival
Modelnotation Effectsincludedinsurvival
M2
S=f(State/ageeffect+Breakup)
M3
S=f(State/ageeffect+Breakup+Freezeup+Breakup*Freezeup)
M4
S=f(State/ageeffect+Breakup+Icedecay+Breakup*Icedecay)
Thirdstep:Selectionofthebestmodelfortheprobabilityofdyingduetohumancauses
Modelnotation Effectsincludedintheprobabilityofdyingduetohunting
M5
H=f(State/ageeffect+Churchillhunting)
Fourthstep:Selectionofthebestmodelfortheprobabilityofweaning
Modelnotation Effectincludedintheprobabilityofweaning
M6
W=f(Breakup)

49

Table5.Selectionofmostsupportedmodelforfemalepolarbearsusingdevianceinformationcriterion(DIC).Ageneralmodelfor
recaptureprobabilitywasusedthatincludedstateandageeffects;randomtimeeffects;effectstoaccountforsightingsinChurchill;
andeffectsduetoradiotelemetry.Additionalparameterswereaddedusingastepbystepmodelselectionapproach(seealsoTable
3).
Firststep:Selectionofthebestmodelforsurvival
Model
Effectsincludedinsurvival

Deviance

M3
State+ageeffect+Breakup+Freezeup+Breakup*Freezeup
M4
State+ageeffect+Breakup+Icedecay+Breakup*Icedecay
M2
State+ageeffect+breakup
M1
Stage+ageeffect
Secondstep:Selectionofthebestmodelforhumanrelatedmortality
Model
Effectsincludedinhumanrelatedmortalitywhileusingthebestmodel
forsurvival(modelM3)
M3
State+ageeffect
M5
State+ageeffect+Churchillcovariate
Thirdstep:Selectionofthebestmodelfortheprobabilityofbreeding
Model
Effectsincludedinprobabilityofbreedingwhileusingthebestmodelfor
survivalandhumanrelatedmortality(modelM3)
M3
State+ageeffect
M7
State+ageeffect+breakup+icedecay+breakup*icedecay
M6
State+ageeffect+breakup
Fourthstep:Selectionofthebestmodelfortheprobabilityoftwinning
Model
Effectsincludedinprobabilityoftwinningwhileusingthebestmodelfor
probabilityofsurvival,humanrelatedmortalityandprobabilityof
breeding(modelM3)
M3
State+ageeffect
M8
State+ageeffect+breakup
Fifthstep:Selectionofthebestmodelfortheprobabilityofweaning
Model
Effectsincludedinprobabilityofweaningwhileusingthebestmodelfor
survival,humanrelatedmortality,probabilitiesofbreedingandtwinning
(modelM3)
M3
State+ageeffect
M9
State+ageeffect+breakup

50

18326.73
18318.63
18319.73
18337.45

(Variance
deviance)/2
2145.889
2276.129
2344.892
2399.979

DIC

DeltaDIC

20472.62
20594.76
20664.62
20737.43

122.14
192
264.81

Deviance

(Variance
DIC
deviance)/2
18326.73 2145.889
20472.62
18319.38 2341.344
20660.73

188.11

Deviance

DIC

DeltaDIC

20472.62
20517.36
20782.61

44.74
309.99

(Variance
deviance)/2
18326.73 2145.889
18317.48 2199.88
18322.36 2460.246
Deviance

(Variance
DIC
deviance)/2

18326.73 2145.889
18323.47 2493.004
Deviance

20472.62
20816.48

(Variance
DIC
deviance)/2

18326.73 2145.889
18324.16 2301.032

20472.62
20625.19

DeltaDIC

DeltaDIC

343.86
DeltaDIC

152.5699

Table6.Selectionofmostsupportedmodelformalepolarbearsusingdevianceinformationcriterion(DIC).Ageneralmodelfor
recaptureprobabilitywasusedthatincludedstateandageeffects;randomtimeeffects;andeffectstoaccountforsightingsin
Churchill.Additionalparameterswereaddedusingastepbystepmodelselectionapproach(seealsoTable4).
Firststep:Selectionofthebestmodelforsurvival
Model
Effectsincludedinsurvival

Deviance

M1
State/ageeffect
M3
State+ageeffect+Breakup+Freezeup+Breakup*Freezeup
M2
State+ageeffect+breakup
M4
State+ageeffect+Breakup+Icedecay+Breakup*Icedecay
Secondstep:Selectionofthebestmodelforhumanrelatedmortality
Model
Effectsincludedinhumanrelatedmortalitywhileusingthebestmodel
forsurvival(modelM1)
M1
State+ageeffect
M5
State+ageeffect+Churchillcovariate
Thirdstep:Selectionofthebestmodelfortheprobabilityofweaning
Model
Effectsincludedinprobabilityofweaningwhileusingthebestmodelfor
survivalandhumanrelatedmortality(modelM1)
M1
State+ageeffect
M6
State+ageeffect+breakup

51

13143.31
13143.66
13146.44
13155.33

(Variance
deviance)/2
1533.99
1552.334
1632.879
1735.959

DIC

DeltaDIC

14677.3
14695.99
14779.32
14891.29

18.69
102.02
213.99

Deviance

(Variance
DIC
deviance)/2
13143.31 1533.99
14677.3
13154.41 1565.002
14719.41

42.11

Deviance

DeltaDIC

(Variance
DIC
deviance)/2
13143.31 1533.99
14677.3
13143.91 1623.598
14767.5

DeltaDIC

90.2

Table7.Estimatesoftimeinvarianttotalapparentsurvival(S),estimatedfromliverecapture
anddeadrecoverydatafortheWesternHudsonBaypolarbearsubpopulation19842011using
multistatecapturerecapturemodels.Ageclassesrepresentacombinationofstatestructure
andageeffectcovariates.

Female
S
95%CI 95%CI
Ageclass
(mode) lower upper
COY(9mo.)
0.56
0.48
0.66
Yearling(dependent)
0.71
0.61
0.81
Yearling(independent)to4yr
0.82
0.79
0.85
519yrwithoutCOY
0.94
0.92
0.96
519yrwithCOY
0.94
0.89
0.99
20yrwithoutCOY
0.77
0.71
0.82
20yrwithCOY
0.89
0.73
0.99

Male
S
95%CI 95%CI
Ageclass
(mode) lower upper
COY(9mo.)
0.52
0.46
0.58
Yearling(dependent)
0.79
0.71
0.87
Yearling(independent)to4yr
0.75
0.72
0.77
59yr
0.93
0.91
0.95
1019yr
0.90
0.88
0.91
20yr
0.72
0.67
0.76

52

Table8.Estimatesoftheprobabilityofhumanrelatedmortalityconditionalondeath(H),
estimatedfromliverecaptureanddeadrecoverydatafortheWesternHudsonBaypolarbear
subpopulation19842011usingmultistatecapturerecapturemodels.Ageclassesrepresenta
combinationofstatestructureandageeffectcovariates.

Female
95%CI
95%CI
Ageclass
H(mode)
lower
upper
COY(9mo.)andyearling(dependent)
0.05
0.03
0.07
Yearling(independent)to4yr
0.28
0.22
0.35
5yrwithoutCOY
0.08
0.05
0.11
5yrwithCOY
0.15
0.08
0.99

Male
95%CI
95%CI
Ageclass
H(mode)
lower
upper
COY(9mo.)andyearling(dependent)
0.05
0.03
0.08
Yearling(independent)to4yr
0.44
0.38
0.49
59yr
0.73
0.58
0.91
10yr
0.24
0.20
0.29

53

Table9.Estimatesoftheprobabilityofgivingbirthandhavingatleastonememberofalitter
surviveuntilautumn(B),estimatedfromliverecaptureanddeadrecoverydataforthe
WesternHudsonBaypolarbearsubpopulation19842011usingmultistatecapturerecapture
models.Ageclassesrepresentacombinationofstatestructureandageeffectcovariates.

Female
95%CI
95%CI
Ageclass
B(mode)
lower
upper
4yr
0.04
0.02
0.10
59yr
0.24
0.18
0.31
1019yr
0.31
0.25
0.39
20yr
0.27
0.20
0.36

54

Table10.Futurepopulationgrowthrate,frommatrixbasedpopulationprojectionmodelsusing
parametersestimatedfromliverecaptureanddeadrecoverydatafortheWesternHudsonBay
polarbearsubpopulation19842011,usingmultistatecapturerecapturemodels.

Population
Futuresea Futuretime
Population Lower95%CI Upper95%CI
segment
ice
frame(years) growthrate
conditions
()
female
high
10
1.02
0.98
1.05
female
high
20
1.02
0.99
1.04
female
high
50
1.02
1.00
1.05
female
low
10
0.97
0.85
1.02
female
low
20
0.96
0.85
1.01
female
low
50
0.97
0.92
1.01
femaleandmale
high
10
1.02
0.99
1.05
femaleandmale
high
20
1.02
0.99
1.05
femaleandmale
high
50
1.02
1.00
1.05
femaleandmale
low
10
0.96
0.85
1.01
femaleandmale
low
20
0.96
0.88
1.01
femaleandmale
low
50
0.97
0.92
1.01

55