ebook img

Seasonal and circadian biases in bird tracking with solar GPS-tags PDF

19 Pages·2017·1.97 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Seasonal and circadian biases in bird tracking with solar GPS-tags

RESEARCHARTICLE Seasonal and circadian biases in bird tracking with solar GPS-tags RafaSilva1*,IsabelAfa´n2,JuanA.Gil3,JavierBustamante1,2 1 DepartmentofWetlandEcology,Estacio´nBiolo´gicadeDoñana(EBD-CSIC),Seville,Spain,2 GISand RemoteSensingLab(LAST-EBD),Estacio´nBiolo´gicadeDoñana(EBD-CSIC),Seville,Spain,3 Fundacio´n paralaConservacio´ndelQuebrantahuesos,Zaragoza,Spain *[email protected] a1111111111 a1111111111 Abstract a1111111111 a1111111111 GlobalPositioningSystem(GPS)tagsarenowadayswidelyusedinwildlifetracking.This a1111111111 geolocationtechniquecansufferfromfixlossbiasesduetopoorsatelliteGPSgeometry, thatresultintrackingdatagapsleadingtowrongresearchconclusions.Inaddition,new solar-poweredGPStagsdeployedonbirdscansufferfromanew“batterydrainbias”cur- rentlyignoredinmovementecologyanalyses.WeuseaGPStrackingdatasetofbearded OPENACCESS vultures(Gypaetusbarbatus),trackedforseveralyearswithsolarGPStags,toevaluatethe Citation:SilvaR,Afa´nI,GilJA,BustamanteJ causesandtriggersoffixanddataretrievallossbiases.Wecomparetwomodelsofsolar (2017)Seasonalandcircadianbiasesinbird GPStagsusingdifferentdataretrievalsystems(ArgosvsGSM-GPRS),andprogrammed trackingwithsolarGPS-tags.PLoSONE12(10): e0185344.https://doi.org/10.1371/journal. withdifferentdutycycles.Neitherofthemodelswasabletoaccomplishthedutycyclepro- pone.0185344 gramedinitially.Fixanddataretrievallossrateswerealwaysgreaterthanexpected,and Editor:AntoniMargalida,UniversityofLleida, showednon-randomgapsinGPSlocations.Numberoffixespermonthoftrackingwasa SPAIN badcriteriontoidentifytagswithsmallerbiases.Fix-lossrateswerefourtimeshigherdueto Received:July18,2016 batterydrainthanduetopoorGPSsatellitegeometry.Bothtagmodelswerebiaseddueto theunevensolarenergyavailablefortherechargeofthetagthroughouttheannualcycle, Accepted:September11,2017 resultingingreaterfix-lossratesinwintercomparedtosummer.Inaddition,wesuggestthat Published:October11,2017 thebiasfoundalongthediurnalcycleislinkedtoacomplexthree-factorinteractionofbird Copyright:©2017Silvaetal.Thisisanopen flightbehavior,topographyandfixinterval.Morefixeswerelostwhenvultureswereperch- accessarticledistributedunderthetermsofthe ingcomparedtoflying,inruggedversusflattopography.Butlongfix-intervalscaused CreativeCommonsAttributionLicense,which permitsunrestricteduse,distribution,and greaterlossoffixesindynamic(flying)versusstaticsituations(perching).Toconclude,we reproductioninanymedium,providedtheoriginal emphasizetheimportanceofevaluatingfix-lossbiasincurrenttrackingprojects,and authorandsourcearecredited. deployingGPStagsthatallowremotedutycycleupdatessothatthemostappropriatefix DataAvailabilityStatement:Allrelevantdataare anddataretrievalintervalscanbeselected. publiclyavailablefromtheSpanishNational ResearchCouncil(CSIC)digitalrepositoryatthe followingURL:http://digital.csic.es/handle/10261/ 155119. Funding:Fundingfortrackingdevicesandfield Introduction workwasprovidedbyFundacio´nparala Conservacio´ndelQuebrantahuesos(FCQ), ManywildlifemovementstudiesemployGlobalPositioningSystem(GPS)locationdata[1]. GobiernodeArago´n,MinisteriodeMedio Bothbasicecologyandconservationstudiesuselocationfixesobtainedfromanimalwithtags Ambiente,Fundacio´nBiodiversidad,ObraSocial carryingGPSreceivers.Amongothers,therearestudiesonhabitat-selection[2,3],animal CajaMadrid,Telefo´nicaandLIFE12NAT/ES/ 000322.RafaSilvawassupportedbyaPhD behavior[4,5]orhuman-wildlifeconflict[6,7].However,obtainingaGPSfix,andhowreliable PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 1/19 BiasesinsolarGPS-tags contractfromtheMinisteriodeEducacio´n,Cultura thislocationis,dependsonmanyfactorsthatarenotalwaystakenintoaccountinthesestud- yDeporte(FPU13-03484).Thefundershadnorole ies.NewGPS-trackingtagstendtogivethousandsoflocationswithveryhighspatialaccuracy. instudydesign,datacollectionandanalysis, Theycangivethefalseimpressionofanunbiasedaccountofthebehaviorandwhereaboutsof decisiontopublish,orpreparationofthe theanimalswhichcarrythem.But,iffixesarenotobtainedatrandom,morefixesarenotnec- manuscript. essarilybetter.Thiscanbespeciallytrickyifanimalbehaviororhabitatusehasaninfluenceon Competinginterests:Theauthorshavedeclared theprobabilityoflosingafix. thatnocompetinginterestsexist. Onepotentialtrackingbiasisduetothelossoffixesifthereceiveronthetagfailstoreceive signalsfromaminimumofthreesatellitesduringthelimitedtimetheGPSreceiverison,and thus,cannotcalculateafix.Thisfact,thatnotallscheduledlocationsareobtained,hassome- timesbeencalledintheliteratureas“fix-ratebias”[8].Inthisworkwewillusetheterm“fix- lossrate”(FLR)todescribetherateatwhichscheduledlocationsarenotobtained,calculated astheinverseoffixsuccessrate,andusethe“fixratebias”torefertoanybiasinthefixingrate causedbyenvironmental,technical,orbehavioralfactors.Theexistenceofbiasinthefixes thatareroutinelylossisanimportanteffectthathastobeconsideredintheanalysisofanimal movement,andsolutionshavetobedesignedtomitigatethem[9]. ManystudiesthathavegainedaninsightintoFLRbiashavebeenbasedonbatteryoperated GPS-tagsdeployedonterrestrialmammals[8,10–12].TheyhaveconcludedthatFLRbiasis influencedmainlybyGPSsatellitegeometry,fixinterval(intervalbetweensuccessivefixes), habitatuse,animalbehavior,orevenGPS-tagpositionandorientation.Itiswellknownthatas theamountofvisibleskyfortheGPSreceiverdecreases(creatingwhatisknownasapoor GeometricDilutionofPrecision,orpoor-GDOP)theriskoflosingthefixincreases[13]. Thus,environmentalfactorsthatpreventaclearviewofthesky,suchasruggedtopographyor canopyclosure,havebeendocumentedashavinganinfluenceonFLR[14,15].Somestudies haveshownthatasmallerfixintervalsyieldsasmallerFLR[14,16,17].Ithasalsobeendemon- stratedthatinactiveanimalshavehigherFLRthanactiveones[18]. Althoughthesestudieshaveresolvedmostoftheissuesinfixratebiasinbattery-operated trackingdevicesusedmainlyinterrestrialmammals,biasinsolar-poweredtagsfrequently usedforbirdtrackingremainsvirtuallyunexplored.Solardeviceswithrechargeablebatteries areessentialtotrackbirds,whichrequirelightlong-lastingtags.Birdsalsomoveinathree dimensionalspaceandtagretrievalisdifficult.Therefore,remotedataretrievalsystemsare alsoneeded,requiringextrabatteryfordatadownloadcomparedtodeviceswhichlogthedata andarelatterrecovered.Asaresult,onlydevicesusingsolarrechargeablebatteriesseemable toofferanadequatesolutiontothisproblem.Torechargethebatterysolartagsrequiresolar radiation,whichvariesextremelyovertheyear,isaffectedbyweather,andcanbeinfluenced bydeviceattachment(e.g.becoveredbyfeathers),andbirdbehavior.Researchersareusually keentogetasmuchdataaspossiblebutintensiveschedulescancompromisetheenergyavail- ablefortagoperationandcreatedatagaps.Todate,wehaveonlyfoundtwostudiesmention- ingthenegativeinfluenceofbatterydrainonFLR[17,19].Insummary,althoughthetrendin tagdevelopmenthasleadtowardsanewgenerationoflightertrackingdeviceswithmoreeffi- cientpowersolutions,technologyhasnotadvancedenoughyetsothattheriskoflosingfixes orlosingdataretrievalattemptsinanon-randomwayduetoaninsufficientbatteryperfor- mancecanbesafelyignored. Lowbatterycapacityandperformance,andpoorGDOParenotgenerallydirectcausesof poorGPS-tagperformancebuttheresultingeffectsofacombinationofotherfactorsrelatedto theelectronics,firmware,andqualityofthecomponents.Forexample,thesuboptimalperfor- manceoftheunitscanbeaconsequenceofnonoptimalarrangementoftheelectronics. Antennadesignisoftenconstrainedbythesizeoftheunitsandmightnotbeoptimalcausing lowantennasensitivityandthisincreasesthetimeneededforsearchingsatellitesandacquiring ephemerides(increasingbatterydrain).Alowantennasensitivityalsoreducesthenumberof PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 2/19 BiasesinsolarGPS-tags satellitesreceived(thusaffectingDOP).Batterychargeisalsodeterminedbytheefficiencyof thecircuitrycontrollingthesolarrecharging,anddevicesfromdifferentcompaniescanuse differentcomponentsfortheircircuitry.Alltheseaspectscreatedifferencesbetweentagmod- elsthatcanbeevaluatedinthelab,buttheeffectofbirdbehaviororhabitatselectionontag efficiencycanonlybeevaluatedwithrealdeploymentsonbirds. Besidesfixinterval,thedataretrievalinterval(intervalbetweensuccessivedataretrievals)is anotherparameterwhichisusuallyscheduledinthetagdutycycle.Mosttagscanstoreon boardtheGPSlocationsandonlydeletethemoncesuccessfullytransmitted.Dependingon thestudyobjectives,dataretrievalintervalcanbecriticaltoachievesuccessinthetrackingpro- gram.Speciesconservationandmanagementprogramsusuallyrequestaquasi-real-time trackingtopreventorquicklyidentifypotentialmortalityeventsduetopoisoning,accidents withpowerlinesorillegalhunting[20–22],mainlyatcriticallife-stagessuchasjuveniledis- persalorduringbreeding[23].Thismeansmanagerswouldaskforfrequentdataretrievals, every1–2daysintheworstscenario.Argos(www.argos-system.com)PlatformsTransmitters Terminals(PTTs)hasbeentheonlysystemavailableforanimaltrackingsincethe1970sfor studiesrequiringworldwidecoverageandinthemid1990sthesystemintegratedGPSreceiv- ersonthePTTs[24]toobtainhigh-spatial-resolutiontrackingdata.Inthelastdecade,some GPStagsdeployedonanimalshavestartedtousemobilenetworkdataservicesfordata retrieval[25].GSM/SMS(GlobalSystemforMobilecommunications/ShortMessageService) andthemorerecentGSM/GPRS(GeneralPacketRadioServiceforlargerdatasets)have emergedascheaperalternativestotheArgossystem.Unfortunately,therearestillextensive areasthroughouttheworld(e.g.areasinAmerica,NorthAsia,Australia,Africa,andpracti- callyalloceans)withoutGSMservices[26].Inthesameway,dataretrievalsystemsusingradio modemtechnologyasVHF/UHFornewprotocolslikeZigbeeorBluetoothareasolidalterna- tive.Theyaresecure,robust,andcheapersystems,andtheyrequirelesspower[27–29]allow- inghigherdatatransferratesthanArgosorGSM.Howevertheyareonlyappropriatefor short-rangedownloads,andhence,inadequateforspecieswithunpredictablemovements. InthispaperouraimistoevaluatetheexistenceofFLRbiasinsolar-poweredGPS-tagsand determineitscausesusingdatafromtwotagmodelsdeployedonbeardedvultures(Gypaetus barbatus).Beardedvulturesareparticularlyadequateforthistest.Theyinhabitmountain rangeswitharuggedtopographythatcreatesdifficultiesforGPSlocation.Thestronglyvari- ableweatherconditions,andtheseasonalvariationinsolarradiationmakesdifficulttopredict theperformanceofsolar-poweredGPStagsthatarefrequentlyusedinconservationprojects. Beardedvulturesrequireslopewindsandthermalupdraftstofly,creatingacircadianpattern inflightbehaviorthatcouldinfluencethevisibilityofGPSsatellitesforthetags.Weattemptto findoutwhyGPStagsdeployedonbeardedvultureslosefixesandifthistakesplacewitha non-randompattern.WeevaluateFLRandtrytoattributeitto"batterydrain"(thetaghasnot enoughenergytoattemptafix)ortopoorGPSsatellitegeometry(thetagisnotabletocontact withenoughsatellites,poor-GDOPhereafter).Weevaluatetheinfluenceofotherinteracting factorssuchasseasonalvariationinsolarradiation,fixinterval,topography,andbirdflight behavior.OurfirsthypothesisisthatFLRinsolartagsdeployedonbeardedvulturesisnon- random,neitheralongtheyear,norduringthedailycycle,andisstronglydependentonthe energyavailabilityforthetag.Accordingly,weexpectthathighcloudcoverageandshorter daysinwinterwouldproducehigherFLR,becausesolartagsdonotfullyrecharge,andcreate aseasonalbiasinnumberoffixesacquired.OursecondhypothesisisthatFLRisalsostrongly influencedbyhabitatuseandbirdbehavior.Differencesintheintensityofflightbehavior alongthediurnalcycleexposethetagstodifferentrelativesatellite-receivergeometries (GDOP),andpotentiallyalsotodifferentbatteryrechargeopportunities.Inthisworkwealso assesstheperformanceoftwodifferentdataretrievalsystemsthathavebeenused(Argosand PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 3/19 BiasesinsolarGPS-tags GSM/GPRS).Relatedtoourfirsthypothesis,thelossofdataretrievals,dataretrievallossrate (hereafterDRLR),shouldfollowthesamepatternasFLRthroughouttheyear,astheenergy consumptionfordatatransmissionisevengreaterthanforfixcollection. Materialsandmethods Ethicsstatements Thisstudywasconductedasapartofalong-termconservationandresearchprogramleaded bytheFundacio´nparalaConservacio´ndelQuebrantahuesos(www.quebrantahuesos.org)in accordancewiththecompetentauthoritiesinthemanagementofthisspecieslistedasendan- gered(EN)intheNationalCatalogueofThreatenedSpeciesofSpain(R.D.139/2011,BOE n.46,23thFebruary2011),aswellasinAnnexeIoftheEuropeanBirdsDirective(Directive 2009/147/EC)andSPEC-3category(Europeanthreatenedspecies).Alltheprocedureshave beenspecificallycarriedoutwithauthorizationoftheNatureConservationAuthorityofthe GovernmentofAragon,incompliancewiththeregionalDecree45/2003(BOAn.29,25thFeb- ruary2003)andfollowingtheprotocolsestablishedbytheStrategyforConservationofthe BeardedVultureinSpain(NationalCommissionforProtectionofNature,4thJune2000). Speciesandstudyarea Thebeardedvultureisaspecializedandterritorialscavengerfeedingmainlyonbonesand inhabitingruggedmountainousareas,inwhichtakesadvantageofslopesoaringinorderto exploitwithlowenergycoststhelargeareasthatmakeuptheirterritories.Nestsareusually locatedonremoteoverhungcliffledgesorincaves.Itisanendangeredspeciesremainingina fewmountainrangesinEurope,Asia,andAfrica[30].InEurope,thelargestbreedingnatural populationisinthePyrenees(170breedingpairs),withothersmallerpopulationsinCreteand Corsica.TherehavebeensuccessfulreintroductionprojectsintheAlpsandAndalusianmoun- tains,wherethespecieswaseradicated.Thespeciesmovementshavebeenstudiedsincethe 1980’swiththehelpofconventionalradiotracking[31],batterypoweredArgosPTTs[32],and morerecentlysolar-poweredGPStags[23,33–35].Trackingstudieshaveaimedtostudythe basicmovementecologyofthespeciesaswellastosolveconservationproblemsofwildand reintroducedpopulations(poisoning,leadintoxication,collisionswithpowerlines,andfood shortage). ThestudyareaincludestwodifferentmountainrangesinthenorthoftheIberianPenin- sula,thenaturalpopulationinthePyreneesandreintroducedindividualsintheCantabrian Mountains(S1Fig).Bothregionshavearuggedtopography,aseasonalclimate,andvarying weatherconditions,withelevationsranginguptoapproximately3,300minPyreneesand 2,500minCantabrianMountains. Fieldprocedures Between2006and2012,13beardedvultures(threeadults,oneimmature,andninenestlings) weretaggedwithtwodifferentmodelsofsolarpoweredGPS-tags(S1Table).Allnestlings weretaggedattheirnests(reintroducedindividualsatthehackingcage)whentheywere between85and105daysold.Adultsandimmaturebirdsweretrappedwithacannon-netin vulturerestaurants.Tagsweremountedonbackpack-styleharnessesusing5mmsiliconcord coveredbyTeflonribbon(BallyRibbonMills,Bally,Pennsylvania,U.S.A.),followingthe methodologydescribedbyBo¨gel[36],butnotusingtheweaklink.InthePyrenees(0.85˚W 42.51˚N),eightindividuals(threeadults,oneimmature,andfournestlings)wereequipped withGPS-ArgosPTT-10070gtags,hereafterPTT,(MicrowaveTelemetryInc.,Columbia, PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 4/19 BiasesinsolarGPS-tags Maryland,U.S.A.).From2010to2012fiveotherbeardedvulturenestlings(threeinthePyre- neesandtwointheCantabrianMountains,6.00˚W42.94˚N)wereequippedwithGPS-GSM/ GPRSCTT-11001stGen100gtags,hereafterCTT(CellularTrackingTechnology,LLC.,Somer- set,Pennsylvania,U.S.A.).Exceptforthedataretrievalsystem,both,CTTsandPTTshavesim- ilartechnicalcharacteristics(GPSsensor,solarpanels,rechargeablebattery,theystore-on- boardGPSlocationsuntilsuccessfullytransmitted,etc.).GPStagswereprogramedwithdiffer- entdutycyclesfollowingtherecommendationofthemanufacturertoobtainthebestperfor- mance.Oncedeployed,tagscouldnotbereprogrammedremotely.AllPTTswereprogramed withthesamedutycycle(PTT#1),providingafixevery2hona12hON/12hOFFcycle from7:00to19:00(CoordinatedUniversalTime,UTC,coincidentinthestudyareawithsolar time),andadataretrievaleverytwodays.InrelationtoCTTtags,twowereprogramedinthe firstyear(2010)toprovideafixevery30sfromsunrisetosunset,anddataretrievaleveryday (CTT#1).Alightsensorswitchedthetagoffduringthenight.Inthesecondyear(2011)asin- gleCTTwasfitted.Duetopoorperformanceofpreviousschedule,itwasprogramedtopro- videafixevery15minfromsunrisetosunsetanddataretrievaleverydayifaminimum numberoflocations,setbythemanufacturer,hadbeenrecorded,withamaximumdata retrievalfrequencyofonceperday(CTT#2).InpracticeCCT#2scheduleattemptedtransmis- sioneveryotherday.Inthethirdyear(2012),twoCTTswerefittedandtheywereprogramed toprovideafixevery15minanddataretrievaleveryday(CTT#3). Allindividualswerealsoequippedwithconventional20gTW51VHFradio-trackingtrans- mitters,manufacturedbyBiotrack(Wareham,Dorset,U.K.;4-yrbatterylifeexpectancy),to allowthebirdstobelocatedincaseofinjuryortechnicalfailureoftheGPS-tag.Birdswere alsomarkedwithmetalandcolordarvicringsandpatagial/humeraltags.Thetotalweightof allthemarkswasaround3%ofthebird’sbodymass,belowthegenerallyaccepted5%limit [37]. Trackingdata Toavoidproblemswithdaysormonthswithunequalsampling,wediscardedmonthsthat wereincompleteatthebeginningorendofthetrackingperiodforeachindividual(S2Table). Forindividualstaggedasnestlings,westartedthetrackingperiodonSeptember1stofthetag- gingyear,inordertonotincludedataneitherduringthenestingnorthehackingperiods.In thecaseofindividualstaggedasadultsorimmatures,whichwereusuallytrappedinvulture restaurantsatanytimeoftheyear,weusedonlydatafromthemonthfollowingthedeploy- mentdate.TheendofthetrackingperiodforouranalyseswasJune30th,2014forcurrently activetagsand“lastdataretrieval”dateforinactivetags.OnlydatarecordedduringthePTT ONcyclewereused(07:00–19:00UTC).Fixesweregroupedin2hintervalsinordertoallow comparisonsbetweenthetwodifferenttagmodels.Wecalculatedfixanddataretrievalloss rates(FLRandDRLR,respectively)perindividualasthefractionofscheduledfixesordata retrievalsthatwerelostforeachdutycycle.Wedifferentiateinbetweenfixeslostduetobattery drain(BatteryFLR),andfixeslostduetothetagGPSreceiverbeingunabletofindenoughsat- ellitestocalculatethefix(GeometryFLR),usingtheinformationprovidedbydownloaded data.Weconsideredafixwaslostduetobatterydrainwhenascheduledfixwasnotrecorded inthetagdataloggerorwhenatimedregisterwithoutGPS-fixindicatedaninsufficientbat- teryvoltage“lowbatt”.Regardingfixeslostduetoapoorgeometry,theywerecodedas“9999” byCTTsandas“nofix”byPTTs.Additionally,acquiredfixesweretaggedbybothtagmodels withfixquality:2Dwhenonlythreesatelliteshadbeenusedtocalculatethefixandnoaltitude estimatewasprovided,and3Dwhenfourormoresatelliteswereusedtocalculatethefix. OnlyCTTsprovidedtheHorizontalDilutionofPrecision(HDOP)foreachfix.Wecalculated PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 5/19 BiasesinsolarGPS-tags theeffectivetimelagbetweensuccessivefixesanddataretrievalstoevaluatetheperformance ofthedifferentdutycycles(S3Table)andcomparedittoprogramedschedules.Birdbehavior ateachfix(classifiedasperchingorflying)wasestimatedforbothtagmodelsonthebasisof theinstantaneousGPS-speedprovidedbythetag.Speedthresholdbetweenbothbehaviors wasestablishedfollowingavisualinspectionofinstantaneousGPS-speedhistogramsof3D- fixes,taking1.39m/sasthelimitbetweenperchingandflyingfixes.Tagbatteryvoltageand birdflyingaltitudewerecalculatedfromthevaluesprovidedbybothtagmodelsforeachfix.It isnotpossibletoknowwherethevulturewasorwhathappenedtotheGPS-tagwhenafixwas notrecorded,butweassumewecandeducethecausesoflosingfixesfromthequalityinforma- tionprovidedbyrecordedfixes. DatarecordedbythetagswereautomaticallyuploadedintoMovebank(www.movebank. org)withinthestudynamed“BeardedVulture(Gypaetusbarbatus),PyreneesandCantabrian Mountains”throughCTTGSMandArgoslivefeedsinordertofollowestablishedrecommen- dationsforanimalmovementdata[38]. Environmentaldata Wedownloadeddataonmeanmonthlypotentialsolarradiationfromthedigitalclimateatlas oftheIberianPeninsula[39].Wefirstlycalculatedthe90%-MCP(MinimumConvexPolygon) foreachindividualusing3D-fixes.ThenweextractedtheradiationvalueswithineachMCP foreachtag.Elevation(inmetersaboveWGS84ellipsoid)wasdownloadedfromthe30-m spatialresolutionASTERGlobalDigitalElevationModel(GDEM)v.2[40]toestimatethevul- tures’flyingaltitudeabovegroundlevel.Elevationdatawasalsousedtoestimatetheterrain roughness,calculatedasthestandarddeviationoftheelevationwithina500-mbufferaround each3D-fix. GPStests GPStheoryexplainsthatGPSreceiversrequiremoreTimeToobtainaFix(TTF)indynamic thaninstaticconditions[41].TTFisalsoshorterwhenthescheduledfixintervalisshorter, becausetheGPSreceiverneedstoreadephemerisandalmanacdatafromeachGPSsatellite beingtrackedatleastonceperhour.DependingontheGPSreceivertype,collectionofephem- erisandalmanacdata,cantakefrom30sto3min.BecauseGPSreceiversprogramedwith shortfixintervals(below1h)canusepreviouslytransmitteddata,theyareabletominimize TTFmorethanthoseprogramedwithlongfixintervals(above1h).Afirstexploratoryanalysis ofthedatasetsuggestedthattherecouldbeaninteractionbetweenfixintervalandbirdbehav- ior(perchingorflying)inFLRbias,becausePTTtagsthathavelongerfixintervals(2h) recordedahigherfractionofperchingfixesthanCTTtags.Whenthetagismovingandfix intervalislong,TTFincreasesdramaticallyandthereisahigherprobabilityoflosingafix.We designedafieldexperimenttotestforaninteractionbetweenfixintervalandinstantaneous GPS-speed(classifiedasstaticordynamic)inTTF.WeusedaGarminGPSMAP162shand- heldGPSdevicewiththesamethreefixintervalsasprogramedinthebeardedvultures’tags. WemeasuredtimesincetheGPSreceiverwasswitchedonuntila3D-fixwithanerrorbelow 5mwasobtained.30TTFmeasurementsweretimedwithastopwatchforeachstate(static anddynamic)andfixinterval(30s,15minand2h)(n=180)byperformingalternativesta- tionaryanddynamictests.Stationarytestswerealwaysconductedatthesamepoint(6.56˚W, 37.89˚N)inordertoavoidthebiasproducedbytopography.Dynamictestswerecarriedout drivingtheGPSonacaronafixed10-kmrouteataspeedof50(km/h).MaximumGPStime- outwasestablishedas180s.Afterthistime,ifnofixwasachieved,theGPSwasturnedoffand thefixwastakenaslost. PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 6/19 BiasesinsolarGPS-tags Statisticalanalysis Toestimatethepotentialseasonalandcircadianbiasesinfixesanddataretrievalsattemptswe fittedGeneralizedLinearMixedModels(GLMMs).WeusedBatteryFLR(fixeslostduetobat- tery/fixesscheduled),GeometryFLR(fixeslostduetopoorGDOP/fixesscheduled)and DRLR(dataretrievallost/dataretrievalsscheduled)asresponsevariablesinthemodelswith “binomial”errorsand“logit”links.TotestfortheexistenceofaseasonalbiasinFLRand DRLRduetobatterydrain,wefittedaGLMMusingasresponsevariablesBatteryFLRand DRLR,andaspredictorvariablesthetagmodelandmonth,includingtheindividualasaran- domfactor.TotestfortheexistenceofacircadianbiasinFLRduetopoor-GDOP,wefitteda GLMMusingasresponsevariableGeometryFLR,andaspredictorvariablestagmodeland solartime(HourUTC),includingtheindividualasarandomfactor.Statisticalsignificanceof predictorswasmadewithanANOVAtest,usingtheChi-squaredtesttoselectamongalterna- tivemodels. Todeterminetherelationshipbetweensolarradiationandbatteryvoltage,usingmean monthlyvaluesforeachindividualtag,weusedlinearregressionanalysis.Theinfluenceof birdflightbehavior(basedontheinstantaneousGPS-speedclassificationas“perching”or“fly- ing”)onGeometryFLRwasexaminedwithlinearregressionanalysis.WerelatedHDOPwith terrainroughnessandvultures’flyingaltitudeforfixesclassifiedas"perching"and"flying" respectively.Previously,wetestedwithaMann-Whitney’sUtest,iftherewereanydifferences inmeaninstantaneousGPS-speedvaluesbetweenbothtagmodels.ToanalyzeourfieldGPS experiment,weusedatwo-wayANOVAtotestfordifferencesinTTFinrelationtofixinterval andstate(staticanddynamic)andalsoitsinteraction.Toavoidhavinganunbalanceddataset, weconservativelyusedTTF=180sforthefixesthatwerenotacquiredusingthemaximum GPStimeout. AlldataanalyseswereperformedwithRv.3.2.4[42]accessingtomovebankdatastorebyR package“move”[43]. Results The13beardedvulturestrackedina7yearperiodfrom2007to2014providedatotalof 83,231GPSfixes(48.1%wererecordedbyCTTsand51.9%byPTTstags).CTTtagsprovided anaverageof256.3fixespermonthoftracking,whilePTTtagsprovidedanaverage126.6fixes permonthoftracking(S1andS2Tables). Trackingefficiency CCTswereprogramedtoprovide,dependingondutycycle,from2,880(CTT#1)to48 (CTT#3)fixesperday,whilePTTswereallprogramedtoprovide7(PTT#1)fixesperday.Nei- thertagmodelaccomplishedthistask.Regardingdays-with-fixes,CTTsonlycollectedfixesin 31.88±33.30%(mean±SD)ofthedaysoftracking,providingaFLRof0.93±0.07,while PTTscollectedfixes85.54±10.67%ofthedays,providingaFLRof0.40±0.12(Fig1and Table1).GapsbetweenfixeswereusuallylongerandmorefrequentinCTTsthaninPTTs. Themaximumperiodwithoutfixesrangedfrom12to307daysforCTTs,andfrom7to11 daysinPTTs.Thestandarddeviationofthetimelagbetweensuccessivefixeswas83.65hfor CTTsand7.68hforPTTs.WithregardtothelocationaccuracyofGPSfixes(estimatedas thepercentageof3Dfixesfromtotalfixes),CTTsprovidedmoreaccuratelocations(%3D fixes=88%±9%)thanPTTs(%3Dfixes=77%±6%). AbetterperformanceondataretrievalwasfoundinPTTs(DRLR=0.20±0.11)than inCTTs(DRLR=0.89±0.06).Thetimelagbetweensuccessivedataretrievalswascloser inPTTstotheprogramedschedule(scheduledeverytwodays,meanretrievaltime- PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 7/19 BiasesinsolarGPS-tags Fig1.Trackingefficiency,trackingqualityandcausesoffixlossofGPSsolartags.Tagmodels: CelltracktechGSM-GPRS(CTT)andMicrowaveArgos(PTT).TrackingEfficiency:meanrateofscheduled fixesthatwereacquired(blue)andlost(red).TrackingQuality:Basedonacquiredfixes(blue),thepercentage ofrealfixesthatare3Dfixes(solid)versus2Dfixes(hatched).Basedonlostfixes(red),thepercentagethat werelostduetobatterydrain(norecord,ortaggedas"lowbatt")(solid),andduetopoor-GDOP(recordwith nolocation,ortaggedas"timeout")(hatched). https://doi.org/10.1371/journal.pone.0185344.g001 Table1. Fixlossrate(FLR),dataretrievallossrate(DRLR)andperformanceofindividualGPSsolartags. Individual Dutycycle Dayswithfix(%) FLR %3D Time-lag DRLR Time-lag(days) max(days) SD(hours) max SD Carrodilla PTT#1 61 0.66 78 9 13.86 0.46 22 2 Asterix PTT#1 88 0.38 75 8 5.35 0.12 14 1 Goriz PTT#1 91 0.39 75 9 6.95 0.18 14 1 Rover PTT#1 83 0.33 90 8 7.59 0.20 10 1 Eva PTT#1 87 0.45 73 11 8.31 0.21 12 1 Ixeia PTT#1 87 0.44 69 11 8.36 0.19 14 1 Sevil PTT#1 96 0.3 76 7 5.59 0.11 10 1 Maria PTT#1 91 0.26 78 7 5.41 0.11 10 1 PTTmean 86 0.4 77 9 7.68 0.20 13 1 PTTSD 11 0.12 6 2 2.79 0.11 4 0 Deva CTT#1 16 0.99 89 243 65.79 0.85 246 25 Coto CTT#1 7 0.99 96 230 85.19 0.93 230 40 Luisa CTT#2 90 0.88 93 12 8.49 0.93 59 8 Cotiella CTT#3 24 0.82 90 182 62.15 0.80 183 18 Atilano CTT#3 22 0.96 73 307 196.62 0.93 310 49 CTTmean 32 0.93 88 195 83.65 0.89 206 28 CTTSD 33 0.07 9 111 69.27 0.06 94 16 Dutycycle:PTT#1=onefixevery2hona12hON/12hOFFcyclefrom7:00to19:00UTC,Dataretrievalevery2days,CTT#1=onefixevery30sfrom sunrisetosunset.Dataretrievaleveryday,CTT#2=onefixevery15minfromsunrisetosunset.Dataretrievalonceacertainnumberoffixesarestored, CTT#3=onefixevery15minfromsunrisetosunset.Dataretrievaleveryday.Dayswithafix(%ofdaysobtainingatleastoneGPSfix),FLR=FixLoss Rate(fractionofscheduledfixesthatarelost),%3D(%offixesthatareacquiredwiththreeormoresatellitesandhavealtitudeinformation).DRLR=Data RetrievalLossRate(fractionofscheduleddataretrievalsthatareunsuccessful).FixTime-lag=timelagbetweenconsecutivefixes.RetrievalTime- lag=timelagbetweenconsecutivedataretrievals. https://doi.org/10.1371/journal.pone.0185344.t001 PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 8/19 BiasesinsolarGPS-tags lag=2.44±1.32days)thaninCTTs(scheduledeveryday,meanretrievaltime-lag=10.54± 27.93days).Retrievaltime-lagsrangedbetween10–22daysforPTTsand59–310daysfor CTTs(Table1). Seasonalandcircadianbiasesinfixanddataretrievallossrates Fix-lossrate(FLR)wasingeneralmorenoticeableduetobatterydrain(90%ofthefixeslost byCTTsand68%ofthoselostbyPTTs)thanduetopoor-GDOP(10%forCTTsand32%for PTTs)(Fig1). Seasonalbias. SeasonalanalyseswithGLMMsshowedasignificanteffectofthemonth andthetagmodelinBatteryFLRandinDRLR(Table2),withmorefixesanddataretrievals lossesinautumnandwinterthaninspringandsummer(Fig2andS2Fig).BestGLMMsesti- matedthatBatteryFLRwas4timeshigher(4.29±0.62,mean±SE),andthatDRLRwas almost4timeshigher(3.73±0.34)inCTTsthanPTTs,duetotheunevenrechargeofthebat- teryalongtheyear. Batteryvoltage. Althoughtheincidentmeansolarradiationwasequalforalltags(S3A Fig),batteryvoltagewasonaveragelowerinCTTsthaninPTTs(S3BFig).Theresultsof regressionanalysesshowedapositivesignificantcorrelationbetweenmeanmonthlysolar radiationandmeanmonthlybatteryvoltage(r=0.41,p<0.05,n=156),bothinPTTs (r=0.68,p<0.05,n=96)andinCTTs(r=0.45,p<0.05,n=60,Fig3). Circadianbias. CircadiananalyseswithGLMMsshowedasignificanteffectofsolartime inGeometryFLR.GLMMshighlightedabiasthroughoutthediurnalcycle.Morefixeswere lostduetopoor-GDOPintheafternoonthaninthemorningandthegreatestnumberwere lostclosetodusk.ThepatternwassimilarinbothCTTsandPTTs(Fig4Aand4B).Onlythe timeofdayshowedastatisticalsignificanteffect,withnosignificantdifferencebetweentag models(Table3). Terrainroughness. ASpearman’scorrelationtestshowedapositivesignificantcorrela- tionbetweenterrainroughnessandHDOP(r=0.12,p<0.001,n=7,398)forthefractionof 3Dfixesclassifiedas"perching",andasignificantnegativecorrelationbetweenflyingaltitude andHDOP(r=−0.15,p<0.001,n=28,451)forthefractionof3Dfixesclassifiedas“flying”. Flightbehavior. InstantaneousGPS-speedof3Dfixesallowedustodifferentiatebetween “perching”and“flying”behavior.Thisallowedustoestimatetheeffectofbirdbehaviorin FLR.Themeanspeedoffixesclassifiedas“perching”was0.01m/s,(n=29,391)whilefor thoseclassifiedas“flying”21.1m/s,(n=39,711),withnosignificantdifferencesbetweentag models(U=8976871,p<0.4096).Thecomparisonoffixesclassifiedas“flying”versus“perch- ing”indicatedadifferentflightactivitytimebudgetinbeardedvulturesaccordingtotag model.Usingalltrackingdata,CTTscollecteda80%of3D-fixesinflight,whilePTTscol- lectedonly20%.Thecircadianpatterninflightbehavior,withhigherflightactivityaround Table2. GLMMsfittedtofixlossrateduetobatterydrain(BatteryFLR)andtodataretrievallossrate(DRLR)toevaluateseasonalbias. Model Deviance χ2 Df p-value BatteryFLR~+(1|Individual)Nullmodel 92929.92 BatteryFLR~Month+(1|Individual) 72096.11 20833.81 11 <0.001 Model Deviance χ2 Df p-value DRLR~+(1|Individual)Nullmodel 3366.57 DRLR~Month+(1|Individual) 3153.18 213.39 11 <0.001 DRLR~Month+Tagmodel+(1|Individual) 3123.4 29.77 1 <0.001 Bestmodels(inbold)wereselectedbyChi-squaredtest. https://doi.org/10.1371/journal.pone.0185344.t002 PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 9/19 BiasesinsolarGPS-tags Fig2.SeasonaltrackingbiasduetobatterydrainpredictedbytheGLMM.Meanmonthlyregression coefficients(solidline),withstandarderrorbars.(A)RelativeBatteryFLR=fix-lossrateduetobatterydrain, and(B)RelativeDRLR=dataretrieval-lossrate.Januarywasfixedasintercept. https://doi.org/10.1371/journal.pone.0185344.g002 noon,wasapparentforbothtagmodels(Fig4C),butsuggestedadifferentproportionofflying vs.perchingbehavior.Throughouttheday,bothCTTsandPTTsrecordedahigherpercentage of3Dfixesatnoonthaneitherduringthemorningorevening(Fig4D).Whenwecompared %3Dfixesforeach2-hinterval,PTTswereclosertoahypotheticalunbiasedand100%-effi- cienttagmodel(FLR=0),i.e.,whichmustacquirethesameamountof3Dfixesateach2-h interval,conformingwiththeregularfixintervalprogramed. InteractioneffectofGPSspeedandfixinterval. Thetwo-wayANOVAperformedon thesampleofourfieldexperimentshowedasignificantinteractionbetweeninstantaneous GPS-speed(classifiedasstaticordynamicstate)andfixinterval(30s,15minand2h) (Table4).FieldGPStestsdemonstratedthattheeffectofthelongestfixintervalonthefix-rate biasishigherinadynamictest(whenGPSismoving)thaninstationarytest(Fig5).Our PLOSONE|https://doi.org/10.1371/journal.pone.0185344 October11,2017 10/19

Description:
greater loss of fixes in dynamic (flying) versus static situations (perching). To conclude . There have been successful reintroduction projects in the Alps and Andalusian moun- tains, where the formance of Televilt GPS-Simplex™ collars on grizzly bears in western and northern Canada. Wildl Soc.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.