REVIEWOpenAccessObservingtheunwatchablethroughaccelerationloggingofani - PDF document

REVIEWOpenAccess Observingtheunwatchablethroughacceleration loggingofanimalbehavior DanielleDBrown 1* ,RolandKays 2,3,4 ,MartinWikelski 4,5,6 ,RoryWilson 7 andAPeterKlimley 8 Abstract Behaviorisanimportantmechanismofevolutionanditispaidforthroughenergyexpenditure.Nevertheless,field biologistscanrarelyobserveanimalsformorethanafractionoftheirdailyactivitiesandattemptstoquantify behaviorformodelingecologicalprocessesoftenexcludecrypticyetimportantbehavioralevents.Overthepast fewyears,anexplosionofresearchonremotemonitoringofanimalbehaviorusingaccelerationsensorshas smashedthedecades-oldlimitsofobservationalstudies.Ani mal-attachedaccelerometersmeasurethechangeinvelocity ofthebodyovertimeandcanquantifyfine-scalemovementsan dbodyposturesunlimitedbyvisibility,observerbias,or thescaleofspaceuse.Pioneeredmorethanadecadeago,a pplicationofaccelerometersasaremotemonitoringtool hasrecentlysurgedthankstothedevelopmentofmoreaccessiblehardwareandsoftware.Ithasbeenappliedtomore than120speciesofanimalstodate.Accelerometermeasur ementsaretypicallycollectedinthreedimensionsof movementatveryhighresolution�(10Hz),andhavesofarbeenappliedtowardstwomainobjectives.First,the especificbehaviorsthroughanimalmovementandbody posture.Second,thevariationinaccelerometerwavefo rmmeasurementshasbeenshowntocorrelatewithenergy expenditure,openingupasuiteofscientificquestionsinspeciesnotoriouslydifficulttoobserveinthewild.Todate, studiesofwildaquaticspeciesoutnumberwildterrestrialspe ciesandanalysesofsocialbehaviorsareparticularlyfewin number.Researchersofdomesticandcaptivespeciesals otendtoreportmethodologymorethoroughlythanthose studyingspeciesinthewild.Therearesubstantialchalle ngestogettingthemostoutofaccelerometers,including validation,calibration,andthemanagementandanalysisofl argequantitiesofdata.Inthisreview,weillustratehow accelerometerswork,provideanoverviewoftheecologicalquestionsthathaveemployedaccelerometry,andhighlight theemergingbestpracticesfordataacquisitionandanalysi s.Thistooloffersalevelofdetailinbehavioralstudiesof free-rangingwildanimalsthathaspreviouslybeenimpossible toachieveand,acrossscientificdisciplines,itimproves understandingoftheroleofbehavioralmechanismsinecologicalandevolutionaryprocesses. Keywords: Accelerometer,Activity,Animalbehavior,Bio-loggi ng,Deadreckoning,Energyexpenditure,Ethogram, Remoteobservation,Telemetry danielle.brown@wku.edu 1 DepartmentofBiology,WesternKentuckyUniversity,1906CollegeHeights Blvd.#11080,BowlingGreen,KY42101-1080,USA Fulllistofauthorinformationisavailableattheendofthearticle ©2013Brownetal.;licenseeBioMedCentralLtd.ThisisanopenaccessarticledistributedunderthetermsoftheCreative CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginalworkisproperlycited. Brown etal.AnimalBiotelemetry 2013, 1 :20 http://www.animalbiotelemetry.com/content/1/1/20 Abstract Resumen: Elcomportamientoesunmecanismoimportantedelaevoluciónyquesepagaatravésdelgastode energía.Sinembargo,losbiólogosdecamporaramenteobservanlosanimalesdurantemásdeunafraccióndesus actividadesylosintentosdecuantificarelcomportamientoparaelmodeladodelosprocesosecológicosa menudoexcluyeneventoscrípticosperoimportantes.Enlosúltimosañosseprodujeronavancesimportantesenel monitoreoremotodelcomportamientodelosanimales,utilizandosensoresdetelemétrodeaceleración (acelerómetros)queempujanloslímitestradicionalesdelosestudiosobservacionales.Acelerómetrosunidosalos animalesmidenelcambiodelavelocidaddelcuerpoeneltiempoypuedencuantificarlosmovimientosaescala finayposturascorporalesilimitadasporlavisibilidad,elsesgodelobservador,olaescaladelautilizacióndel espacio.Comopionerohacemásdeunadécada,laaplicacióndelosacelerómetroscomounaherramientade monitoreoremotohaaumentadorecientementedebidoaldesarrollodehardwareysoftwaremásaccesibles.Seha aplicadoamásde120especiesdeanimaleshastahoy.Medidasdelosacelerómetrosserecogentípicamenteen tresdimensionesdemovimientoamuyaltaresolución�(10Hz),yhastaahorasehanaplicadohaciadosobjetivos principales.Primero,lospatronesdelasformasdelosacelerómetrosdeondasepuedenutilizarparadeducir comportamientosespecíficosatravésdemovimientodelosanimalesylaposturacorporal.Segundo,seha demonstradoquelavariaciónenlasmedidasdeformadelosacelerómetrosdeondasehademostradoquese correlacionaconelgastodeenergía,abriendounaseriedepreguntasdecaráctercientíficosobreespeciesmuy difícilesdeobservarenlanaturaleza.Hastalafecha,losestudiosdelasespeciesacuáticassilvestressuperanalas especiesterrestressilvestres,ylosanálisisdeloscomportamientossocialessonmuypocosennúmero.Los investigadoresdelasespeciesdomésticasyencautiveriotiendenareportarmetodologíamáscompletaquelos queestudianlasespeciessilvestres.Hayretosimportantesparaconseguirelmáximorendimientodelos acelerómetros,incluyendolavalidación,calibraciónygestiónyanálisisdegrandescantidadesdedatos.Enesta revisiónseilustracómofuncionaelacelerómetro,seproporcionaunavisióngeneraldelasinvestigaciones ecológicasquehanempleadolosacelerómetrosysedestacanlasmejoresprácticasemergentesparalaadquisición yanálisisdedatos.Estaherramientaofreceunniveldedetalleenlosestudiosdecomportamientodelosanimales salvajesquehansidohastaahoraimposiblesdealcanzary,entodaslasdisciplinascientíficas,quemejorala comprensióndelpapeldelosmecanismosdecomportamientodelosprocesosecológicosyevolutivos. Palabrasclaves: Acelerómetro,actividad,bio-registro,comportamientoanimal,gastoenergético,etograma, navegaciónaestima,observaciónadistancia,telemetría. Review “ Mangoestonaturetolearnwhatnatureis,but,inso doing,heintroducespossibilitiesofdistortionthroughhis ownpresence. ”– T.C.Schneirla(p.1022,[1]). Naturalistshavelongbeenawarethattheirpresencecan affectanimalbehavior[1,2].Directobservationpresents obviousdifficultieswhenanimalsperceivehumansaspred- ators[3]orwhentheyarenaturallysecretiveandelusive [4,5].Habituatingindividualstoanobserverissometimes possiblebutitislabor-intensive,andcanrequirelong-term study[6,7].Furthermore,thoughthesubjectsunderstudy maybehabituated,humanpresencecanstillaffecttheir behavioralinteractionswithothernon-habituatedpredator, preyorcompetitorspecies[8].Theobserverisrarelyun- detectableandevenanimalsthatdonotappeartoreactto humanpresencemaystillchangetheirbehaviorinsubtle ways[9,10].Directobservationsarealsobiasedbyourown physicallimitations[11-13]andtendenciestoattendto someeventsandsubjectsmorethanothers[14]. Thefieldofbiotelemetrygrewoutoftheneedtolo- cateanimalsatwillandobserveandrecordtheirhabitsdes- pitetheirabilitiestotravelrapidlyandwidelyininclement weather,underwater,oratnight[12,13,15].Locatingani- malsinspacehasprogressedfrommanualtrackingof animal-borneradio-oracousti csignalstoautomateddepth andgeomagneticloggersandsatellite-basedpositioningsys- temsthatpracticallyeliminatetheobservereffectandcan nowprovidepreciseworldwidelocationswithfewtemporal orspatialconstraints[11,16].Nevertheless,arecordofani- mallocationsoradepthprofiletellswheretheanimalwas andhowlongitstayedthere,butthebehavioralcontextis absentandmusteitherbeinferredordemandsareturnto directobservationmethods[17].Theseissuesunderscore theneedforremotemeasurementofanimalbehaviortore- duceoreliminatethepotentialeffectsofobserverpresence whilemaintainingahighlevelofdetailindatarecording thatiscomparabletodirectobservation[18].Overthepast fewyears,therehasbeenanexplosionofresearchonre- motemonitoringofanimalbehaviorusingmeasurements ofacceleration(Figure1)[19, 20].Thistool,theaccelerom- eter,hasrepeatedlycircumventedmanyoftheage-old limitsofdirectobservationofanimalsinthefield. Figure2providesabasicexplanationofhowanaccel- erometerworks[21].Anaccelerometerisaspring-like Brown etal.AnimalBiotelemetry 2013, 1 :20 Page2of16 http://www.animalbiotelemetry.com/content/1/1/20 piezoelectricsensor.Whendeformed,thesensorgener- atesawave-likevoltagesignalthatisproportionaltothe acceleration(changeinvelocity)itexperiences[22].The sensorisdeformedbothbygravitationalaccelerationas wellasinertialaccelerationduetomovement.Fromone tothreeofthesesensorsarealignedorthogonallytoone anotherandaffixedtoananimalsothateachsensor measuresaccelerationinasingleplane,ordimension,of movement(surge,heave,andsway(Figure2)).Allthree sensorscollectingsimultaneousmeasurementscanrep- resentthree-dimensionalmovementrealistically[20,23]. Thesensorscanbeuser-programmedtosampleacceler- ationatfrequenciesrangingfrom0.5to10,000Hz,and canbesettorecordcontinuouslyorinrepeatedbursts (e.g.,every2min).Thevoltagesignals,alsoknownas rawaccelerometeroutput,maybeusedintheirraw state,orconvertedtoactualaccelerationiftheunitis carefullycalibrated(measuredin g ;1 g =9.8m/s 2 ).Under staticcircumstances,suchasduringrestorafterdeath,the accelerometersignalonlyrepresentsthegravitationalforce actingonthesensors.Whenananimalismoving,sensor outputrepresentsaccelerationduetogravitycombined withtheinertialaccelerationgeneratedbymovement[23]. Accelerometerstypicallyincorporateamicroprocessorand digitalmemorytostoreloggedmeasurementsuntilthein- strumentisretrieved[24]. Measurementofaccelerationisawell-establishedre- searchtoolinbiomechanics[25,26]andexercisescience [27,28].Thefirst(wired)accelerometerswereusedto examinethebiomechanicsofmovementinhumans[29] andfish[30]andthentoascertainthecorrelationbe- tweenbodilyaccelerationandoxygenconsumptionin humansubjects[28].Theintroductionofair-bagtech- nologyinpassengervehiclesleadtothedevelopmentof relativelyinexpensiveaccelerometersthatuseverylittle power[31].Thesewerequicklyadoptedforstudiesout- sidethelaboratoryenvironment,becausetheyare “ small, low-costinstrumentsthatprovidequantitative[andob- jective]measurements[ofactivity] ” p.679,[32].Animal studiesusingthesemodern,trulyportableacceleration sensorsdidnotappearintheliteratureuntilthelate 1990s[33,34].Initially,animalstudieswereconfinedto captiveanddomesticatedspecies,aswellasaquatictaxa, forwhomfewotherbehavioralobservationmethods werepossible[12,35,36].Sincethen,theongoingreduction inthesizeofcomputermicroprocessorsandimprove- mentsinbatterysize,weight,andlongevitycombinedwith thesesmallsolid-stateaccelerationsensorshaveresulted inamodernaccelerometerthatcanweigh0.7g(withouta battery)andmeasure9.5×15×4mm(availablefrom: http://www.technosmart.eu/axy.php).Modernaccelerome- tersalsoconsumeverylittledigitalmemorywitheach measurement[37],sodatacollectionanddatastorageon- boardtheinstrumentitselfispossibleforuptoseveral monthsoryears,dependingonthesamplingschedule.Ac- celerometersthatsimplylogtheirdatamustberetrieved afterthesamplingperiod,aswithothertypesoftelemetric dataloggers.However,someaccelerometersincorporate ultra-highfrequencydatadownloadtechnologyinsimilar frequencybandwidthsasthoseusedincellularphones [38].Thisfeaturemakesitpossibletodownloadthedata fromtheaccelerometerfromareasonabledistance(upto 500m,personalobservation)evenifthedeviceandits bearerarenotvisibleortheinstrumentcannotbere- trievedbecauseithasbeendiscardedinatreecavity,for example[21].Radiooracousticbeaconsarecommonly usedonloggersthatmustberetrieved[21,24]. Movementisthefundamentalbehavioralresponseto bothinternalmotivationsandtheexternalenvironment [13,17].Usingaccelerometers,biologistscanmeasurethe movementbehaviorofwildanimalsoverbiologicallyand Figure1 Primarypapersusingaccelerometersinanimal behaviorresearch1998 – 2012. Figure2 Howanaccelerometerworksandtypicalorientation ofinstrumentaxesonaterrestrialmammal( Tamandua mexicana )[21]. Brown etal.AnimalBiotelemetry 2013, 1 :20 Page3of16 http://www.animalbiotelemetry.com/content/1/1/20 ecologicallysignificanteventsandperiods,practicallyun- limitedbyvisibility,observerbias,orgeographicscale.Ac- celerometerscanbedeployedwithothersensors,suchas thoserecordinglocation(GPS,acoustictelemetry,water depth),physiologicalmeasurements(heartrate,bodytem- perature),andenvironmentalvariableslikeairtemperature, lightlevelsandmagneticheading[24,39,40].Particularly whencombinedwithotherinstruments,measurementsof accelerationcanprovideawiderangeofdetailedinforma- tionontheenvironmentalcontextofanimalbehaviorand physiologythatcanexceedthedescriptiveabilitiesofthe humanobserveranddeepenourknowledgeevenforwell- knownspeciessuchasdomesticanimals.Here,wereview howaccelerometershavebeenusedtodateinthestudyof animalbehavior,includingthetaxonomicandresearch trendsintheliteratureandweillustratethetypeofdata producedbythistechnologyfrominstrumentsdeployed onavarietyofspecies.Further,weprovideasummary ofthecurrentlyavailabletechniquesfordatacalibration, managementandanalysis,andsuggestkeydirectionsfor futureresearch. Methods WeaccessedBIOSIS®PreviewsandISIWebofKnow- ledge®onlineandransearchesforanypublicationcontain- ingreferencestoaccelerometryinthetitle,abstractor keywords.Welimitedouranalysistoprimaryresearch publishedinpeer-reviewedjournalsandbookchapters throughDecember2012.Fromthose,weselectedstudies utilizinganimal-bornesensorsappliedtonon-humanspe- cies.Weassessedtheresultingworksforthefollowing:i) studypurpose;ii)speciesandwhethersubjectswerecap- tive/domesticorfree-ranging,andaquaticorterrestrial; iii)numberofaccelerationaxes;iv)samplingfrequency utilized;v)thebehavioralresolutionoftheresultingmea- surements;vi)theparametersoftheaccelerometerdata usedforanalysis;vii)whetherornotbehavioralclassifica- tionaccuracywasreported(ifpertinent);andviii)whether accelerometrywascombinedwithothertelemetrysensors. Resultsarepresentedaspercentages;notallpercentages willsumto100becausenotallcategoriesweremutually exclusive. Results Wediscovered176accelerometrystudiesandcounted125 animalspeciesthathaveborneaccelerometers(Additional file1).Studieswererelativelyevenlysplitbetweenaquatic (48.3%)andterrestrial(52.8%)habitatsandbetweenfree- rangingwildanimals(50%)anddomesticated/captivewild animals(33/27.3%),buttherewerebiasesamongtaxafor thesecategories(Figures3and4).Mammalsrepresented 45.6%ofallstudyspecieswithdomesticcattleandPinni- pedsbeingthemost-studiedamongthemammals(14%of studiesand18%ofspecies,respectively).Birdscomprised 33.6%ofallstudyspeciesand38%ofavianspecieswereei- therSphenisciformesorSuliformes.Fishesincluded11.2% ofspeciesandhalfofallfishspecieswereElasmobranch sharks.Eightreptilespecies,fiveofthemChelonians,com- prised6.4%ofstudysubjects.Giantcuttlefish( Sepia apama ),Humboldtsquid( Dosidicusgigas ),Kingscallop ( Pectenmaximus )andCanetoad( Bufomarinus )werethe fourstudyspeciesremainingoutsideofthesefourtaxon categories. Morethanhalfofallstudies(62.3%)utilized3-axisac- celerometers;90.3%ofstudiesutilizedeither2-or3-axis accelerometers.Samplingfrequenciesrangedfrom0.5Hz to10,000Hz,with60%ofstudiesusingoneofthefollow- ingmostcommonsamplingfrequenciesof8,10,16,32, 64or100Hz.Forty-eightpercentofstudiescollected Figure3 Accelerometrystudiesperformedonwildfree-ranging animalscomparedtodomestic/captiveanimalsbytaxon. Figure4 Accelerometrystudiesperformedonaquaticanimals comparedtoterrestrialanimalsbytaxon. Brown etal.AnimalBiotelemetry 2013, 1 :20 Page4of16 http://www.animalbiotelemetry.com/content/1/1/20 accelerationdatacontinuouslyand13.3%collecteddatain discreteburstsorintervals;38.7%ofstudiesdidnotclearly reporttheircollectionmethod.Sixty-threepercentofstud- iescombinedanaccelerometerwithothertelemetricin- struments;however,free-rangingwildspecieswere3times morelikelythancaptivewildspeciesand6timesmore likelythandomesticatedspeciestobeoutfittedwithtelem- etrydevicesthatcontainedmultiplesensors.Themost commonremotesensorsusedintandemwithaccelerome- tersmeasureddepth(35.6%ofstudies),travelspeed(16% ofstudies)andtemperature(14.7%ofstudies). Surveyofquestionscurrentlyservedbyaccelerometry: bodypostureandbodymovement Theaccelerationwaveformsovershort(millisecond)to long(minutes)periodscanbeusedtodeducebehavior- specificbodyposturesandbodymovements(Figures5 and6)[41,42].Acrosstaxa,36.4%ofstudiesreportedac- celerationethogramsoracceleration-baseddescriptions ofbehavior.Justunderhalf(46.6%)ofallstudiesutilized theaccelerometerwaveformstodetermineactivitybudgets. AsshowngraphicallyinFigu re5,accelerometervoltage outputofinactiveorrestbehav iorismoreorlessconstant, whilewhole-bodymovementofanykindproducesfluctuat- ingaccelerationwaveformswithhighlevelsofvariance amongmeasurements.Ofthestudiesexaminingactivity budgets,35%ofauthorsusedth isvariancecharacteristicof accelerometerwaveformstosimplyidentifythetimingof activityvs.rest[43-46].Sixty- fivepercentofauthorsidenti- fieddistinctwaveformsforspecificbehaviorsandthenesti- matedtheamountoftimeanimalsspentengagedinthese behaviorssuchaschasingpreyorfeeding,flight,swimming, walking,running,climbing,standing,lyingdown,thermo- regulationandsleeping.Quan tifyingforagingeffortisan applicationofaccelerometrythatfewothertelemetrytech- nologiescanaccomplishandisparticularlyusefulforani- malsthatforageorhuntoutofsight.Researchershave documentedforagingstrategiesthatdifferbyspecies,age orsex[47-52].Otherstudiesplacedaccelerometersensors onthehead/mandibletodirectlymeasureattemptsatfood capture[53],althoughforagingeffortdidnotnecessarily correlatewithforagingsuccess[54].Thereareseveral methodsforidentifyingandcategorizingwaveformsthat representspecificbehaviors(see ‘ BestPractices ’ ,below). Onaverage,thesestudieswereabletoidentifyfourdistinct accelerationwaveformprofiles(range2to7),typicallyfall- ingunderthebroadbehavioralcategoriesoflocomotion, resting,andfeeding/foraging[55-57].Ingeneral,whilethe accelerometerpatternsofactivelocomotorybehaviors (walking,running,climbing,swimmingandflying)are clearlydistinguishablefrominactivebehaviorssuchas sleep,thermoregulationanddigestion,thewaveformsof thesetypesofrelativelyimmobilebehaviorsarenotpar- ticularlydistinctfromoneanother[19]. Whenanaccelerometeriscombinedwithothersen- sorsonataggedanimal,researcherscandescribethe broaderecologicalcontextofaccelerometer-determined behaviors.Lightlevelandambienttemperaturesensors intandemwithaccelerometrypermitexaminationofactiv- itytiminginrelationtoenvironmentalconditions[58-61]. Accelerometersandremotely-sensedlocationviaGPS, compass,depthoracousticsensorsprovidethespatialdis- tributionsofaccelerometer-determinedbehaviors[62-66] andcanleadtonovelinsightsaboutspecies ’ behavioral ecology.Forexample,traditionalobservation-onlyresearch ofthelittle-knownoilbird( Steatorniscaripensis )ledtothe hypothesisthatthesenocturnalfrugivorousbirdswerenot seeddispersersbecausetheseedsfromtheirdietwerere- gurgitatedinthedarkcaveswherethebirdsroost.Holland Figure5 Accelerometer-baseddeterminationofbodyposture andthetimingofrestvs.activity. Dataarefromastudyofthe northerntamanduaanteater Tamanduamexicana [21].Acceleration wassampledat19Hzfor~3secondsevery2min.Forsimplicity,the y-axisisnotshownandwaveformsrepresenttheaveragevoltage measuredevery2minforeachaxis. Figure6 Heave-axisaccelerationwaveformsofthreebehaviors oftheSwallow-tailedgull (Creagrusfurcatus) . They-axisshowsthe unit-freevoltageoutputoftheaccelerometersensor.Photograph anddatapreparedbySebastiánCruz(unpublished). Brown etal.AnimalBiotelemetry 2013, 1 :20 Page5of16 http://www.animalbiotelemetry.com/content/1/1/20 etal.[38]determinedthatoilbirdsoutfittedwithGPS/ac-celerationloggersspentonlyeverythirddayincaves,otherwiseremainingintherainforestwheretheyregurgi-tatedseedsontotheforestflooratconsiderabledistancesfrombothfeedingsitesandcaveroosts.Theauthorsmain-tainedthatoilbirdsshouldbereconsideredasanimportantlong-distanceseeddisperserinNeotropicalforests,anovelhypothesisfortheecologyofthatecosystem.Behavioralanalysisappliedtomonitoranimalwelfarewasasignificantcomponentofaccelerometryresearchonterrestrialanimals;80%oftheterrestrialstudies(and25%ofallstudies)examinedthewelfareofdomesticatedspecies.Typically,authorsusedaccelerometrytomonitorbehavioralchangesassociatedwithreproduction[67,68]orbehavioralresponsestoveterinaryorhusbandryprac-tices[69-74].Instudyingwelfareoffree-rangingwildspe-cies,mortalitysensorsareacommonfeatureoftelemetersandtypicallyprovideaspecialsignaltoalertresearcherstotheanimalsdemise[75].Theadvantageofusingaccel-erometerstodetectmortalityisthatitincludesarecordofbehaviorleadinguptothetimeofdeath,providingarichercontextthatasimplelocationandtimeofdeathoftendonot.Forexample,Kroneetal.[76]wereabletoidentifyachangeinactivityand,ultimately,themomentofdeath,duetotoxinexposureinawhite-tailedseaeagleHaliaeetusalbicillaWhatislargelyabsentfromthisbodyof82articlesaboutactivitybudgetsisthemeasurementofsocialbehav-iors.Whilenumerousstudiescomparedbehaviorbudgetsduringparticularreproductivestatesandreportedetho-gramsforbroodingornestpreparation[52,57,66,77-81],onlytwostudiesexaminedwhethermatingbehaviorhadacharacteristicaccelerationprofile[82,83].Thescarcityofpublishedaccelerometryethogramsforaggressiveinterac-tions,territorialorcourtshipdisplays,andplayandparent-offspringbehavior[84,85]couldbebecausethesesocialbehaviorsweregenerallyrareinthemajorityofthespeciesthathavebeenstudied,theaccelerationwaveformsofsocialbehaviorswereindistinguishablefromthoseofnon-socialbehaviors,orbecauseitwasnotfeasibletotagmultipleanimalsinthesamegroup.Inter-individualtel-emetry,withanimalsbearingtagsthatareabletorecordthedateandtimeofproximitytoothertaggedanimalshasrecentlybeenreportedforacoustictransmitters[86].Theapplicationofaccelerometerstostudiesofsocialbe-haviorwouldbenefitmightilyfromaccelerometertagsthathavetheabilitytorecordproximity,identityorevenbehavioroftaggedindividualsincontactwiththeanimalthatbearstheprimarytag.Surveyofquestionscurrentlyservedbyaccelerometry:biomechanicsandtheenergeticsofmovementEnergeticshavelongbeenofinteresttobehavioralecolo-gists[87,88]becauseallmovementsrequireenergy,andprudentallocationofenergytospecificactivitiessuchasforaginghasdirectconsequencesforfitnessandnaturalselection[77,89,90].Priortotherecentdevelopmentsinaccelerometry,measuringenergyexpenditureinwildani-malsinthefieldinvolveddoubly-labeledwaterorheartratetelemetry,bothofwhichhavelogisticallimitationsthathaverestrictedtheiruse[91,92].Accelerometertechnologyhasdramaticallyadvancedourunderstandingoftheroleofenergyinbehavioralstrategiesbymakingitpossibletostudyfine-scale,behavior-specificenergyexpenditureout-sidethelaboratoryindiversetaxa[93].WilsonandHalseyetal.havetestedforcorrelationsbetweenbodilyacceler-ationandoxygenconsumed(assumingatleastpredomin-antlyaerobicmetabolicpathways)acrossawiderangeofspeciesfromaquaticmammals[94,95],birds[64,77,96-99],fishes[100,101],reptiles[52,102-104]andabivalve[105],toterrestrialmammals,birds[56,106-108]andoneam-phibian[109].Althoughthestrengthoftherelationshipbe-tweenbodilyaccelerationandoxygenconsumption(asaproxyformetabolicrate)variesanddependsonanumberoffactors,therelationshipisvalidacrossallspeciesexam-inedtodate[110].WilsonsmetricOverallDynamicBodyAcceleration(ODBA)[77]hasbecomethemostcom-monlyusedacceleration-basedproxyofmetabolicrate(en-ergyexpenditure)andseveralarticleshavebeendevotedtostandardizingthisproxyorvariantsofit[19,106,110-112].Thecurrentavailableresearchindicatesthatbodilyacceler-ationcanqualitativelyassesshowtheamountofmechan-icalworkperformedbythebodydiffersamongactivelocomotivebehaviors,adistinctimprovementonoldertechniquesthatwerenotbehavior-specific(Figure7)[110].Seventy-threearticlesappliedaccelerometrytobio-mechanicalresearch(42.7%ofallarticlesexamined).Asmallminorityofthesestudies(7),eschewedapplicationstometabolismandinsteadremainedwithinthetraditionalrealmofevaluatingperformance:runninginracehorses[113-117],swimminginseasnakes[118],andflightinProcellariformseabirds[119].Theremaining90%ofarti-clesfocusedonenergyefficiencyduringlocomotionfortravelorforaging[31,35,94,120-124].Inordertobetterunderstandtheselectionpressuresoncurrentpatternsoflocomotorbehavior,researcherscomparedmovementen-ergeticsacrossspecies,movementstrategies,demographicclassesandbehaviors[42,48,102,107,122-127].Therewasastronghabitatbiasinexistingaccelerometry-basedresearchonbiomechanicalenergetics,withaheavyemphasisonmarinedivinganimalssuchasPinnipedsandCetaceans[36,95,128,129],penguins[97,130,131],Pelicaniformbirds[77,99,126,131-133],andmarinetur-tles[79,134].Terrestrialtaxa,mainlybirds,wererepre-sentedinonlyfive(of66)studiesofmovementenergetics[56,106-108].Oftheterrestrialspecies,wewereabletoidentifyonlyasinglepublishedfieldstudyofenergeticsfornon-volantterrestrialanimals:canetoads(Bufomarinusetal.AnimalBiotelemetryPage6of16http://www.animalbiotelemetry.com/content/1/1/20 [109].Batterysizeandweightstillmostlyprecludeaccel- erometryenergeticsstudiesofthesmallestwildmammals (particularlybats)andbirds.Afurtherlimitationisthatac- celerometersdonotappeartobeaparticularlygoodproxy ofenergyexpendedduringimmobilebutstillenergetically costlybehaviorssuchasthermoregulationorgestation [107,135]. Potentialapplicationofaccelerometry:position andlocation Accelerationmeasurementscanbeusedtoderiveanimal speed,which,togetherwithcompassanddepth/altitudinal information,couldbeusedto ‘ dead-reckon ’ ananimal ’ s position.Thereareseveralexistingmethodsforlocating animalsinspaceandtimeincludingradiotelemetry[136], satelliteorgeographicpositioningsystems[11,137,138] andacousticarrays[139].Noneofthesemethods worksforallspeciesandhabitatsand,consequently, travelpathsarefrequentlyreconstructedbybridging sporadicpointsandhavelowspatio-temporalresolution [140,141]. Dead-reckoning(alsoknownaspathintegration)uses vectorcalculationsfromvelocityandthechangeinheight ordepthtogetherwithaknownstartposition(usuallythe animalreleasepoint)toderivenewpositionswithrespect tothosepreviouslyknown[24,142].Locationsobtainedby dead-reckoning,therefore,arenotsubjecttothesamecon- straintsofreceiverlocationorsatelliteaccessandrepresent analternativemethodforstudyingmovementpathswhen radio-orsatellite-basedtelemetrymethodsareunsuitable. Dead-reckoningusessensorson-boardthetelemetrytag thatrecordheading/direction(usuallymeasuredwithmag- netometers),altitudeordepth(usuallymeasuredwithpres- suresensors),andspeed.Intheory,speedcanbecalculated bytakingthederivativeofaccelerationoveraknowntime interval[130]orbyusingaknownstridelengthandthe accelerometer-measuredstridefrequency[142].However, speeddeterminedinthiswaycanbesubjecttolargeerrors duetovariationinslopeandsubstrateduringtravel [67,139].Theseerrorsareparticularlyunpredictablein aquaticorvolantspecies,duetodriftcausedbywaterand windcurrentsratherthananimallocomotion[142].In terrestrialsystems,terraininclineandsubstrateimpact stridelength,affectspeedcalculationsandconsequently thedeterminationofdistancemoved.Furthermore,these errorsaccumulateovertime,makinglocationestimates increasinglyworsefurtherfromthelastknownlocation. Becauseoftheseproblems,dead-reckoningfromaccele- rometrydatahasbeenusedinfrequentlyandmost researchersinterestedinmovementspeedhaveadded separatespeedsensors(smallexternalpropellers)tothe telemetrytags[24,137,141-144].AsGPStechnology becomesmorewidelyintegratedintoaccelerometertags, thegreatestpotentialfordead-reckonedanimallocation comesinrecreatingtheexacttravelpathbetween subsequentGPSlocationscollectedatshortintervalse.g., 5min[24,139]. Bestpracticesindataacquisitionanddataanalysis:tag attachmentandtandemsensors Attachingtelemetrytagstoanimalsisacomplicatedand delicateprocessthatrequirescaretoreducetheinflu- enceoftheequipmenttotheanimal.Consultationwith experiencedfieldbiologistsandtagcompanies,notto mentionproperliteraturereview,iscriticalduringthe planningstageofanytaggingstudy.Inadditiontothe standardconcernsofattachment longevity,deviceretrieval andwhethertagattachmentaffectsanimalbehavior[138], tagattachmentforaccelerometersensorsisespeciallysensi- tivebecauseshiftsofthetagrelativetothepositionofthe animalcouldimpacttheinterpretationofthethree-axis data.Extensivepreliminaryresearchonreadilyobservable animalsisoftenneededtofine-tuneanewattachment methodforagivenspecies[145].Commonmethodsof accelerometertagattachmenti ncludeneckcollars[20,55], legbracelets[146],harnesses[109,145],andtape-[118], clamp-[147]orglue-ontags(Figure2)[21,102].Arigidat- tachmentensuresthatoncethetagisdeployed,theaxes,or dimensions,ofmovementbeingmeasureddonotchange overthedeploymentperiodandthataccelerationofthetag Figure7 Overalldynamicbodyaccelerationshownforahoppingandnon-hoppingcanetoad Bufomarinus . Thisstudywasthefirstto useaccelerometrytoestablishabehavioraltimebudgetandassignenergycoststothosebehaviorsforanon-volantterrestrialanimal.Graphic reprintedwithpermissionfrom[109]. Brown etal.AnimalBiotelemetry 2013, 1 :20 Page7of16 http://www.animalbiotelemetry.com/content/1/1/20 independentoftheanimal(byacollarbouncingupanddownontheneck,forexample)iskepttoaminimum[111].Forspeciesthatmustwearcollarsorbracelets,acompletelyrigidattachmentisnotpossibleunlessthecol-larcanbepreventedfromturningaroundtheneck/leg[55],whichmaypresentawelfareconcernforfree-ranginganimals.Forsomequestions,forexamplethetimingofac-tivity/rest,therequirementofrigidattachmentmaybere-laxed.Finally,accelerometertagscanalsobedeployedinsidethebodycavityofsomespecies[61,148],whichmayreduceconcernsabouttagmovementsthatareirrelevanttotheresearchquestion.Internaldeploymentsmaypro-videtheadvantageofrecordingaccelerationsduetophysiologicalprocessessuchasheartbeatandmovementsofsmoothmuscleduringdigestion[148],butcanalsohavethedisadvantageofnecessitatingsurgicalproceduresfortagdeploymentandretrieval/removal,whichcanaffectanimalbehaviorandwell-being.Theorientationofaxesistypicallyplacedsothatthesurgeaxisisalignedwiththelongitudinalbodyaxisandswaywiththehorizontalbodyaxis(Figure2)[20].Ensuringthattagpositionisassimilaraspossiblebetweenindivid-uals,especiallythoseofverydifferentbodysizes,improvesthesignal-to-noiseratiooftheaccelerometeroutputandminimizeserrorsininterpretation[20,63,101,131].Beyonditsorientationonthebody,thespecificanatomicallocationoftheattachedaccelerometertaglargelydetermines,whatbehaviorscanbedistinguishedbytheiraccelerometrypat-terns.Bothspeciesmorphologyandtagplacementwillde-terminethenumberandtypeofbehaviorswithdistinctaccelerationprofiles[44].Forexample,tagsattachedtoansback,asinFigure2[38,109],willnotprovideac-celerationpatternsoffine-scalefeedingbehaviorsthatonlyinvolvemovementofthemouth.Ontheotherhand,accel-erationsofchewingmovementsmaybedetectablewithneckcollars[33,55].Inhumans,ithasbeenwellestablishedthatpreciseaccelerometer-baseddescriptionsoffull-bodymovementrequireatleastfiveaccelerationsensors,onemountedonthetrunkofthebodyandoneoneachex-tremity[26].Studiesoffree-rangingwildanimalaretypic-allylimitedtoonetelemetrytagperindividual;however,multipleaccelerometerinstrumentshavebeenusedondo-mesticatedanimals[149-152]andinahandfulofwildmar-inespecies[53,54,153-157],improvingtheprecisionofbehaviormeasurements.Evenwhencontainedinasingletag,mostmodernac-celerometersarecombinedwithothertypesofsensorstoenhancetheamountofinformationcollectedsimul-taneouslyfromtheenvironment,suchaslight,air/waterpressure,externalair/watertemperature,relativehumidityandmagneticfield[24],aswellasfromtheanimal,suchasbodytemperature,heartrateandmouth/jawmove-ments[79,99,101,133,134,157-160].Inmoderntelemetrytags,eachofthesedatasensors,includingeachaxisoftheaccelerometer,havetheirownseparatechannelsfordatare-cording,sothataccelerometerdataarecollectedindepend-entlyofotherinformationlikeGPS[24,160].Asaresult,evenifonesensormalfunctionsorcannotacquireinforma-tionmomentarily(forexample,theGPSunitspendsseveralminutesattemptingtoaccesssatellitesandobtainaloca-tion),theothersensorscontinuetorecorddataonsched-ule.Insometags,theactivitylevelsoftheanimalasdeterminedbytheaccelerometercanbeusedtosetthere-cordingschedulesofothersensorsdynamically.Forex-ample,theGPSscheduleissettoacquirelocationsmorefrequentlyduringactivebehaviorssuchasforagingandtravelandlessfrequentlyduringrest,improvingtheoverallperformanceandbatterylongevityofthetelemeter[21].Bestpracticesindataacquisitionanddataanalysis:samplingaxes,samplingintervalandsamplingfrequencySamplingofallthreeaxesofacceleration(tri-axial)isthemostaccurateandprecisewayofmeasuringbehav-iorthatoccursinthreedimensionsaswellasestimatingenergyexpenditure[20,161].Forsomeresearchques-tionsorforrelativelyimmobilespecies,oneortwoaxesmaybesufficienttocharacterizethebehavior(s)ofinter-est[23,80].However,theefficiencyofmodernacceler-ometersensorsmeanthatlittleisgained,intermsofbatterylife,byusingfeweraxes.Themajorityofstudiesintheliteraturesampledaccel-erationcontinuously,atfrequenciesabove1Hz[97].Thistypeofsamplingproducesanextremelyhighvol-umeofdata;becauseeachaccelerometeraxisisseparate,threeaxesrecordingat1Hzproducethreemeasure-mentspersecond,whichrapidlyaccumulateintomil-lionsofloggedmeasurementsforatagdeployedoverseveraldays.Inpractice,continuousdataaretypicallysub-sampledoraveragedoverseveralsecondsworthofmea-surementstocreatearunningmean[111]soanalternativetocontinuoussamplingistorecordforafewsecondsatin-tervalsofoneormoreminutes.Byrecordingathighreso-lution(e.g.,60Hz)butshortduration(e.g.,1to3sec)thisstrategyaimstosamplejustonebehaviortypeandavoidbehavioraltransitions(e.g.,restingtowalking)thatcouldcomplicateautomatedclassificationstatistics.Eachdiscretesamplingperiodisthencalledaoranepoch[22,158].Becauseburststudiesrecordfewerdataovertheentirestudyperioditispossibletodownloadthedatare-motelythroughwirelessconnections[21,40],whereascon-tinuousaccelerometerdatatypicallyareloggedoverdaysorweeksandmanuallydownloadedupontagretrieval[143].Ifanimalsareexpectedtoremainwithinthevicinityofafixedreceiver,thencontinuousdatamaybetransmittedwirelesslyatintervals[161].Ifproximitytoareceiverisproblematic,aswithmarineanimalsthatcanrangeoververylongdistances,datacanstillbecollectedathighreso-lution(highsamplingfrequencyandcontinuoussamplingetal.AnimalBiotelemetryPage8of16http://www.animalbiotelemetry.com/content/1/1/20 interval)aslongastheentiresamplingperiodmatchesde-vicestoragecapacity,orthereisperiodicoffloadingofdataviamobilereceiverssuchassatellites.Generally,thesmallerthesubject,thefasterthemove-mentandthehigherthesamplingfrequencynecessarytoaccuratelycharacterizethepatternofacceleration[98,123,137].Fromsignalprocessingtheorywehavetherule-of-thumbthatforadequatereconstructionofacon-tinuouswaveformsuchasacceleration,thesamplingfre-quencyoughttobeatleasttwicethatofthehighestfrequencymovementbeingclassified[162].Satoetal[123]measuredthedominantstrokefrequenciesforsev-eralspeciesofaquaticbirdsandmarinemammalsandtheyrangedfrom0.2Hzforspermwhalesto9.3Hzinguillemots.Meanwhile,thethreemostcommonsamplingfrequenciesintheliteraturewere10,16,and32Hzbuttherewaslittleapriorijustificationforthechoiceofsam-plingfrequency.Halseyetal.foundthataccelerometer-samplingfrequenciesof2to10Hzwereadequateforcharacterizingenergyexpenditureinchickens[19].Thesestudiessuggestthatsamplingfrequencieshigherthan50to60Hzareprobablyunnecessaryformostresearchquestionsandthatinsuchcasestheadditionaldatagener-atediswastefulofdigitalstoragespace.However,authorsrecommendedthattheresearchquestionanddesiredtempo-spatialresolutionofthedatashouldultimatelydic-tatethesamplingfrequency(andsamplinginterval)[19].Bestpracticesinaccelerometerdataanalysis:describingthewaveformsForbothcontinuousandburstsamplingschemes,irre-spectiveofsamplingfrequency,accelerationsensorspro-ducerawdatainawave-likesignalwithunitsinvoltage.Priortoanalysis,researchersmayusecalibrationequationstoconvertthissignalintoactualaccelerationmeasuredinunitswhere1=9.80665m/s[57,127,154,159].Thiscalibrationandconversionmayberequiredformea-suringtheactualbiomechanicalforcesexperiencedbyani-malsduringdifferentmovements,forexample,theairtowatertransitionforadivingseabird,orthestrikeforceonhorseshooveswhenrunning.Thedynamicbodyacceler-ationmetricsalsousethesignalconvertedtoaccelerationastheproxyformetabolicexertion[96].Alternatively,forsimpleaccelerationethogramsordeterminingactivitybudgets,thesignalmaybeusedintherawvoltagestateasdepictedinFigure6.Todescribetheaccelerationwaveformpatterns,re-searchersthencalculateawidevarietyofsummarystatisticsusingeachburstspopulationofvaluesorsubsamplesofthecontinuousmeasurements(Table1).ThestatisticslistedinTable1couldbecalculatedforeachaxisindividuallyorcombinedtorepresentmultipleaxessimultaneously[149].Thereisadichotomyintheliteratureonhowre-searchersprocessanddescribethewave-likepropertiesofaccelerometeroutput.Someresearchershaveuti-lizedsimplewaveformstatistics,suchasthenumberofpeaks(frequencyofthemovement),themeanvalueofthewaveform(bodyangle),andtheirvariances[23,35,121,131,146,160].Othershaveusedspecializedpro-gramstoperformcomplexanalysesonthewaveforms,resultinginalargenumberofadditionaldescriptivesta-tistics[34,46,55,127,130,144,156,161,165].Therearenu-merouscomplextechniquesforanalyzingdatathat,likeacceleration,existinatimeseries[167,168].ThemostcommonlyusedmethodwithaccelerometerdataisthefastFouriertransformation.Fouriertransformationsiden-tifytheindividualfrequenciesthatarepresentintherawaccelerationwaveformanddeterminethepowerspectraldensitiesofthosefrequencies,i.e.,howmuchofthetotalsignalispresentineachfrequency[162].Anothercom-plexapproachiscontinuouswavelettransformation,whichidentifiesnotonlywhichfrequenciesarepresent,butalso Table1StatisticsusedtodescribeaccelerationSummarystatisticRepresentativeMean[Runningmeanforcontinuousdata[Minimum,Maximum,Range[Variance[StandarddeviationInversecoefficientofvariation[Overalldynamicbodyacceleration[Vectordynamicbodyacceleration[Subsequent-measurementautocorrelation[Trend(linearregressioncoefficientthroughaxisdata)Pair-wisecorrelationsbetweenaxesdataInclination,azimuthofresultantandtheircircularvariancesFrequencypowerFastFouriertransformation[129Continuouswavelettransformationon164]
acceleration[frequencyWaveformfrequency[Waveformperiodandamplitude[Areaunderthewaveformcurve[Skewnessandkurtosisofthewaveform[Signalmagnitudearea[WaveformlengthWaveforminheritanceetal.AnimalBiotelemetryPage9of16http://www.animalbiotelemetry.com/content/1/1/20 whenduringthesignaltheyarepresent[164].Shepardetal.[20]andLaichetal.[23]suggestthatthesecomplexanalysesarenotessentialandthatsimplerstatisticsarebothintuitivelyandpracticallymoreaccessibleforthebroadestrangeofpotentialusers.However,theyacknowl-edgedthatwhenbehaviorsaretransitoryand/orhighlyvariablep.36[23],oraremeasuredusingonlyoneaxisofacceleration,themorecomplicatedtechniquesandtheadditionalstatisticstheyprovidemayprovehelpfulforidentifyingordistinguishingdifferentbehaviors.DatafiltersforseparatinggravitationalaccelerationfrominertialaccelerationRecallthattheaccelerometerwaveformoutputduringmovementisacombinationofaccelerationduetograv-ityandinertialaccelerationduetoanimalmovement(dynamicacceleration).Whenisolated,gravitationalac-celerationcanbeusedtodeterminetheorientationofthebodyinspace(postureorbodyangle)[23,55,57].Thegravitationalcomponentcanbeisolatedby:i)applyingalow-passfiltersuchas0.1Hzthatremoveshighfrequencyacceleration[41,78,79,121,164];orii)bysmoothing(i.e.,calculatingarunningmean)overalargesetofmeasure-ments[57].Foraccelerationsampledoverafewsecondsinaburst,takingthemeanvalueofasingleburstsmeasure-mentscansufficeforisolatingmomentarygravitationalac-celeration,orbodyangle[21].OnecanseehowthisworksinFigure5;notetherelativelyflatslopeofvoltageoutputforallthreeaxeswhentheanimalismoreorlessmotion-less(leftsidedesignatedresting).Betweenminutes2and4,themeanvalueoftheheaveaxisshiftsdramaticallyastheanimalchangedpositionduringrestfromafeet-upposturetoafeet-downposture.Thechangeinthemeanvalueoftheheaveaxisrepresentsachangeinvoltageout-putstimulatedbygravitationalaccelerationafterthetag(theanimal)changedorientation.Determiningexactbodyanglerequirescalibrationofaccelerometervoltageoutputasthetagispassedthrough360degreesalongeachaxis.Usingthismethod,researcherscalculatedbodypitchanglefromtheheaveorsurgeaxesandbodyrollanglefromtheswayaxis,alsocorrectingforthepositionofthetagontheanimal[41,163].Conversely,researchersusedhigh-passfilterstoexamineaccelerationsduetomovementinisolationfromthegravi-tationalcomponentofacceleration[63,159].Thisdynamiccomponentwasusedtocalculatethemeasuresofdynamicbodyaccelerationinthemajorityofthestudiesonenerget-ics[19,20,24,77,97,107,109,111,142,159].Frequencyfilterswerealsousedtoreducethenoiseintheaccelerationsignalcreatedbynon-rigidattachmentsofaccelerometercollars[31]andtoisolatethepatternofoneparticulartypeofdy-namicbehaviorthatoccurredsimultaneouslywithothermovements,namelypreycaptureeventsduringswimming[155,156].Spectralandotherwaveformanalysesdiscussedintheprevioussectionareoftenconductedondynamicac-celerationafteritsisolationbyhigh-passfrequencyfiltering.Bestpracticesinaccelerometerdataanalysis:validationandassigningcharacteristicwaveformstobehaviorTheadvantageofaccelerometersisthattheyprovidearemotelycollectedrecordofbehavior:largesetsofaccel-erationwaveformsthatweremostlynotobservedbythehumaneye.Tounderstandhowtheaccelerationrecordandthestatisticalpropertiesofthewaveformsrelatetoobservablebehavior,researchersusingthistoolmusthaveawayofassigningthewaveformstospecificbehaviorsorbehavioralcategorieswithahighdegreeofaccuracy(valid-ation).Thistaskrequiressomepriorknowledgeofthebe-haviorsanimalsperformandstudies,todate,havegenerallyobtainedthisinformationfromdeploymentsonsimilardo-mesticatedanimals,captiveindividualsandbriefperiodsofobservationonfree-rangingwildanimals,someviavideo[36].Carefullysynchronizedobservationsandaccelerom-eterrecordingsvalidatewhatbehaviorscorrespondtowhataccelerometermeasurements,forexample,therelativelyflatwaveformsthatoccurduringrestcomparedtothevari-ablewaveformsthatoccurduringactivity(Figure5).Thisprocessalsomustquantifytowhatextentaccelerometerwaveformsforthesamebehaviorvarywithinanindividual,orbetweenindividualsorspecies[19,109].Thisvalidationprocessisacriticallyimportantpartofusingaccelerome-ters.Theaccuracyoftheconclusionsdrawnfromassigningbehaviorsorenergyexpenditurestoaccelerometerwave-formsdependenormouslyontheaccuracyoftheassign-ments(seediscussionbelowonmethodsreporting).Thisfactnotwithstanding,wildanimals,particularlyaquaticspecies,maynotbeobservedatallbetweenre-leaseandtagrecovery[42].Evenwhenanimalsarebeingobserveddirectly,itishardtobecertainthatallpossiblerelevantbehaviorshavebeenwitnessed[169],especiallywhenextrapolatingbehaviorincaptivitytobehaviorinthewild.Asaccelerometryhasmatured,researchershavedevelopedspecialsoftwaretoolstoaddressthisobstacleandreducethetimeandlabornecessaryfordirectob-servation[164,170].Withknowledgeofi)generalbodyshape,ii)formoflocomotion(bipedal,quadrupedal,etc.),andiii)howthetagisattachedtothebody,thesesoftwareprogramscanhelpresearchersvisualizethemovementoftheirstudyanimalsaccordingtotheaccelerometersignalsrecordedduringtagdeployment[170].Manualexaminationofaccelerometerdataisessentialinthepilotphasesofastudy,butanautomaticsystemtocategorizewaveformpatternsandassignthemtodif-ferentbehaviorsquicklybecomesnecessaryduetothelargesizeoftheaccelerationdatasets.AftercalculatingwaveformcharacteristicssuchasthoselistedinTable1,therearetwomajorapproachestoautomaticwaveformclassificationintheliterature.Thefirstistousestatisticaletal.AnimalBiotelemetryPage10of16http://www.animalbiotelemetry.com/content/1/1/20 algorithmstoclusteraccelerometerwaveformswithsimi-larcharacteristicsandthenassigneachclustertoageneralbehavioralgroup[163].Forexample,Sakamotoetal[164]usedanunsupervisedk-meansclusteringalgorithmtoassignaccelerometerwaveformsfromadivingseabirdto20differentgroups,whichtheymatchedtosimultan-eouslyrecordeddepthprofilesandthenlabeledwithdif-ferentbehaviorgroupsincludinginflightunderwaterdivingandonland.Thesecond,morecommonap-proachistousetheaccelerometerwaveformsgeneratedfromknown(observed)behaviorsofsimilardomesticatedorcaptiveindividualstotrainanalgorithmthatwillassigntheremainingwaveformsinthedatasettothosespecificbehavioralcategories.Forexample,Nathanetal.[56]ob-servedwildandcaptivevulturesexhibitavarietyofbe-haviorswhilewearingaccelerometertagsandthenusedvarioussupervisedstatisticalalgorithmstocategorizetheaccelerometerwaveformsaseitheractiveflight,pas-siveflight(soaring-gliding),eating,lyingdown,preening,standingorrunning.Bothmethodslessentheburdensofextendeddirectobservationsandmanualanalysisofaccel-erometerdata,however,theformerhasthepotentialtodetectpreviouslyunknownorunobservedbehaviorsandbehavioralsequenceswhilethelatterhastheadvantagethatbehavioralcategoriescorresponddirectlytoobserva-tions.Bothrelyonthevalidationprocessforaccuratecon-clusions.Table2summarizestheassignmentmethodsandalgorithmsrepresentedintheliterature;Nathanetal[56]reviewsandcomparesseveralofthesupervisedalgo-rithmsindetail.Regardlessofthemethodusedtoassignaccelerometerwaveformstobehaviors,eachgroupofresearchersde-velopsitsownsetofwaveformstatisticstofeedtowhatarelargelycustom-designedautomaticclassificationsys-tems.Algorithmsthatdealwithoneparticulardomainofactivities(e.g.,divinginaquaticanimals)maynotbeeasilyadaptedforadifferentenvironmentordifferentsetofmovements[32].Furthermore,58%ofscientistsworkingwithdomesticspeciesand47.6%ofthosework-ingwithcaptivewildspeciesreportedtheperformanceandreliabilityoftheirchosenautomatedclassificationsystems[55,72],whileonly9.1%ofthosestudyingwildspeciesdidso[23,44,57,83].Lackofmethodsreporting,fromtheaccelerometer-recordingscheduletowhetherandhowaccelerometerdataisvalidated,stymiesdirectcomparisonsbetweenstudiesandbetweenanalyticalap-proaches.Aspreviouslydiscussed,validationisanessen-tialpartofusingaccelerometerasastand-infordirectobservation.Ifaccelerometersaretoachievewidespreaduseinstudiesoffree-rangingwildspecies,therewillhavetobemorecompletereportingofmethods,particu-larlyfortheclassificationphaseofanalysis.Bestpracticesinaccelerometerdataanalysis:datavisualizationandstorageWenotedearlierthatthehighresolutionofaccelerometryresultsinalargevolumeofdataaccumulatingoverashortperiod.Forexample,asingle-axisaccelerometertagre-cordingcontinuouslyat8Hzfor8hoursand40minutesresultedin249,988measurements[170].Whencombinedwiththedatafromsensorsdeployedintandemwithtri-axialaccelerometry,suchasGPS,depth,ortemperaturetelemeters,thedatasetinitsentiretycaneasilyover-whelmbasicspreadsheetandstatisticalprogramsanditbecomesdifficulttovisualizemorethanonedatastreamatatime.IgorPro(WaveMetrics,LakeOswego,OR,USA),(RFoundationforStatisticalComputing,Vienna,Austria)andMatlab(MathWorks,Natick,MA,USA)arecommonlyusedforaccelerationdataanalysisandcanhan-dlelargedatasets[83,164,165],althoughallthreeprogramshaveaconsiderablelearningcurve.Weareawareoftwoweb-basedoptionsgearedtowardsanimal-bornetelemeterdatavisualization,storageandanalysis.MOVEBANK(availableathttps://www.movebank.org/)isafree,onlinedatabaseofanimaltrackingdatathathelpsresearcherstomanage,selectivelyshare,protect,analyzeandarchivetheirdata.WithMOVEBANK,researcherscanlinkanimalbehaviorfromaccelerometerdatawithanimallocationdatafromGPSandinformationfromglobalenvironmen-taldatasets,suchasweathermodelsandsatelliteimagery,makingiteasiertoexplorehowanimalsmovementsrelatetotheirenvironment.Gaoetal.presentedanotheronlineaccelerometerdatastorageandanalysissystem,theSemanticAnnotationandActivityRecognitionsystem[166].Theirinteractivewebinterfaceenablesecologiststovisualizeandcorrelatetri-axialaccelerometerdatastreamswhilealsofacilitatingaccelerometerdataanalysiswithasupportvectormachineclassificationalgorithm.Amajorbenefittousingtheseweb-basedrepositoriesisthatthe Table2MethodsforassigningaccelerometerwaveformstobehavioralcategoriesbasedonwaveformstatisticsMethodRepresentativesource(s)Referencepatterning[Fixed-threshold[UnsupervisedmachinelearningalgorithmsClusteranalysis[SupervisedmachinelearningalgorithmsClassificationandregressiontrees[Randomforests[Linearorquadraticdiscriminantanalysisanalysis55,132,149]Logisticregression[Supportvectormachines[Artificialneuralnetworks[etal.AnimalBiotelemetryPage11of16http://www.animalbiotelemetry.com/content/1/1/20 averagebiologisttrackingahandfulofanimalsgetsaccesstocollaborationswithotherbiologists,statisticians,engi-neersandcomputerprogrammerswhocancollectivelycontinuetodevelopthistoolandthehardwareandsoft-warethatmakethemostofaccelerometryspotential.ConclusionsandfuturedirectionsAccelerometryisatoolforfine-scaleobservationsofbe-havior,unlimitedbyanimalvisibility,terrain,climate,observerbiasorthescaleofspaceuse.Todate,acceler-ometertagshavebeenappliedtomorethan120speciesindiversetaxainordertodeducebodypostures,behav-iorsandenergeticsinthefield.Accelerometryalsoshowspotentialasamethodtodeadreckonananimalsexacttravelpathwhenappliedintandemwithsatellite-basedlo-cationsystems.Inalloftheresearchdescribed,accelero-metryprovidedfine-scalebehavioralmeasurementsthat,priortoitsdevelopment,wererarelyattainableoutsideofthelaboratorysettingandwithouttheinfluenceofthere-searcherspresence.Theliteratureshowsseveraltaxonomicbiasesinwhatresearchquestionshavebeenexaminedandhowthere-sultshavebeenreported.Studiesofwildaquaticspeciesoutnumberstudiesofwildterrestrialspecies.Researchonaquaticanimals(whethercaptiveorwild)hasfocusedondescribingthebiomechanicsandenergeticconsequencesofbehavior,whileinterrestrialsystemsthefocuswasondeterminingactivitybudgets.Bothatseaandonland,feeding,locomotion,andactivity/restwerethebehaviorcategoriesmostfrequentlyanalyzed;socialbehaviors(par-entalcare,territorial,matingandcourtshipbehaviors,andantagonisticexchanges)arenearlyabsentfrombothetho-gramsandenergybudgets.Researchersofdomesticandcaptivespeciestendedtoreportanalysismethodsmorethoroughlythanthosestudyingspeciesinthewild.Therearesubstantialchallengestogettingthemostoutofaccelerometerdata,includingdeviceretrievalanddatacalibration,validation,managementandanalysis.Nu-meroustechniquesforaddressingthesechallengeshavealreadybeenpublishedinbothhumanandanimalstudiesandnewmethodscontinuetodevelopandareawaitingbroadapplicationtothefield.Withmorethoroughre-portingofmethodologyandhabitualuseofweb-baseddatarepositories,universalpracticesareboundtoemergeashardwareandsoftwarecontinuestomatureandbe-comemorebroadlyavailableacrossresearchgroups.FuturedirectionsTherapiddevelopmentofthistoolinthefield,thusfar,leadsustoanticipatetwopromisingbreakthroughapplica-tionsthatwillopenevenmoredoorsinbehavioralresearch.Thefirstistheincorporationofinter-individualtelemetersinthestudyofsocialbehavior.Device-to-devicedatashar-ingandproximitysensorsexistintheconsumerelectronicsindustryandhavealreadybeenincorporatedintoacoustictelemeters[86].Mostmoderntelemetrytagshavemultipledatachannelsandcouldbemodifiedtoincludethesefea-tures.Accelerometertagsthathavetheabilitytorecordproximity,identityandevenbehavioroftaggedindividualsincontactwiththeanimalthatbearstheprimarytag,wouldreducetheburdenofextensivedirectobservation,yetpermitscientiststodirectlyquestionhowindividualsinteractandhowthoseinteractionsshapebehavioracrossalargenumberofsocialandterritorialspecies.Secondly,werecommendthatresearchersenteringthefieldofwildlifetelemetrylooktoexplicitlylinkenergeticexpenditureinwildanimalswithbehavioralresponsestohuman-alteredhabitats[163].Whetherconsideringclimatechange,re-sourcecompetitionoranti-predatordefenses,thepotentialtollonfitnessshouldmanifestinenergeticexpenditureandallowawindowontothelonger-termconsequencesofourimpactsonotherlong-livedspecies.Accelerometryof-fersalevelofdetailinbehavioralstudiesoffree-rangingwildanimalsthathaspreviouslybeenimpossibletoachieveandithasprovenitselfinfurtheringourunderstandingoftheroleofbehavioralmechanismsinecologicalandevolu-tionaryprocesses.AdditionalfileAdditionalfile1:SpreadsheetofallstudiesanalyzedforthisColumnheadingsindicatetheinformationtakenfromthestudy,studiesarelistedinrowsbyfirstauthor.CompetinginterestsAPK,MW,andRWareontheeditorialboardofAnimalBiotelemetry.Theauthorsdeclarethattheyhavenoothercompetinginterests.scontributionsDBcollectedthearticlesusedinthisreview,carriedoutthedataanalysis,anddraftedthemanuscriptandFigures1to5.RKinitiallyconceivedoftheideatowriteareviewofthesubjectandheavilyinfluencedtheorganizationoftheinformation.MW,RW,andAPKmadecriticalrevisionstodraftsforcomprehensivenessandpresentationoftheinformation.Allauthorsreadandapprovedthefinalmanuscript.AcknowledgementsThisreviewwassupportedbytheNewYorkStateMuseum;theMaxPlanckInstituteforOrnithology;SwanseaUniversity;anNSFpredoctoralfellowship(DDBrown)andtheBiotelemetryLaboratoryattheUniversityofCalifornia,Davis.WethankSebastiánCruzforcontributingFigure6andLewisHalseyforpermissiontoreusehisfigure(Figure7)aswellashelpfulcommentsonanearlierdraft.AuthordetailsDepartmentofBiology,WesternKentuckyUniversity,1906CollegeHeightsBlvd.#11080,BowlingGreen,KY42101-1080,USA.NorthCarolinaMuseumofNaturalSciences,11W.JonesStreet,Raleigh,NC27601,USA.Wildlife&Conservation,NorthCarolinaStateUniversity,Raleigh,NC,USA.SmithsonianTropicalResearchInstitute,Apartado0843-03092,Balboa,Ancón,Panamá,RepúblicadePanamá.MaxPlanckInstituteforOrnithology,VogelwarteRadolfzell,Schlossallee2,Radolfzell78315,Germany.ChairofOrnithology,KonstanzUniversity,ConstanceD-78457,Germany.Biosciences,CollegeofScience,SwanseaUniversity,Swansea,SA28PP,UK.ofWildlife,Fish,&ConservationBiology,1334AcademicSurge,UniversityofCalifornia,Davis,CA95616,USA.etal.AnimalBiotelemetryPage12of16http://www.animalbiotelemetry.com/content/1/1/20 Received:23July2013Accepted:7October2013Published:10December20131.SchneirlaTC:Therelationshipbetweenobservationandexperimentationinthefieldstudyofbehavior.AnnalsNYAcadSci2.CarpenterCR:Afieldstudyofthebehavioralandsocialrelationsofhowlingmonkeys(AlouattapalliataCompPsycholMonographs1934,10:168.3.CaroTM:DemographyandbehaviourofAfricanmammalssubjecttoBiolConserv4.MaffeiL,NossAJ,CuéllarE,RumizDI:Ocelot(Felispardalis)populationdensities,activity,andrangingbehaviourinthedryforestsofeasternBolivia:datafromcameratrapping.JTropEcol5.ChapmanFM:Whotreadsourtrails.NatGeogrMag6.JackKM,LenzBB,HealanE,RudmanS,SchoofVAM,FediganL:TheeffectsofobserverpresenceonthebehaviorofCebuscapucinusinCostaRica.AmJPrimatol7.CrofootMC,LambertTD,KaysR,WikelskiMC:Doeswatchingamonkeychangeitsbehaviour?Quantifyingobservereffectsinhabitutatedwildprimatesusingautomatedradiotelemetry.AnimBehav8.IsbellLA,YoungTP:Humanpresencereducespredationinafree-rangingvervetmonkeypopulationinKenya.AnimBehav9.CaineNG:Unrecognizedanti-predatorbehaviorcanbiasobservationalAnimBehav10.MartinP,BatesonPPG:MeasuringBehaviour.3rdedition.Cambridge:CambridgeUniversityPress;2007.11.CagnacciF,BoitaniL,PowellRA,BoyceMS:AnimalecologymeetsGPS-basedradiotelemetry:aperfectstormofopportunitiesandchallenges.PhilTransacRoyalSocB-BiolSci12.KooymanGL:Genesisandevolutionofbio-loggingdevices:1963MemoirsNatInstPolarResSpecialIssue13.CookeSJ,HinchSG,WikelskiM,AndrewsRD,KuchelLJ,WolcottTG,ButlerPJ:Biotelemetry:amechanisticapproachtoecology.TrendsEcolEvol2004,19:334343.14.AltmannJ:Observationalstudyofbehavior:samplingmethods.15.HartKM,HyrenbachKD:Satellitetelemetryofmarinemegavertebrates:thecomingofageofanexperimentalscience.EndangeredSpecRes16.TomkiewiczSM,FullerMR,KieJG,BatesKK:Globalpositioningsystemandassociatedtechnologiesinanimalbehaviourandecologicalresearch.PhilosophicalTransacRoyalSocB-BiolSci17.FryxellJM,HazellM,BorgerL,DalzielBD,HaydonDT,MoralesJM,McIntoshT,RosatteRC:Multiplemovementmodesbylargeherbivoresatmultiplespatiotemporalscales.ProcNatlAcadSciUSA2008,105:1911419119.18.AguiarLM,Moro-RiosRF:ThedirectobservationalmethodandpossibilitiesforNeotropicalCarnivores:aninvitationfortherescueofaclassicalmethodspreadoverthePrimatology.Zoologia2009,26:587593.19.HalseyLG,GreenJA,WilsonRP,FrappellPB:Accelerometrytoestimateenergyexpenditureduringactivity:bestpracticewithdataloggers.PhysiolBiochemZool20.ShepardELC,WilsonRP,QuintanaF,GomezLaichA,LiebschN,AlbaredaDA,HalseyLG,GleissA,MorganDT,MyersAE,NewmanC,MacdonaldDW:Identificationofanimalmovementpatternsusingtri-axialaccelerometry.EndangeredSpecRes21.BrownDD,LaPointS,KaysR,HeidrichW,KümmethF,WikelskiM:Accelerometer-informedGPStelemetry:reducingthetrade-Offbetweenresolutionandlongevity.WildlifeSocBull22.DowC,MichelKE,LoveM,BrownDC:Evaluationofoptimalsamplingintervalforactivitymonitoringincompaniondogs.AmJVetRes23.LaichAG,WilsonRP,QuintanaF,ShepardELC:IdentificationofimperialcormorantPhalacrocoraxatricepsbehaviourusingaccelerometers.EndangeredSpecRes24.WilsonRP,ShepardELC,LiebschN:Pryingintotheintimatedetailsofanimallives:useofadailydiaryonanimals.EndangeredSpecRes25.MorrisJRW:Atechniqueforthemeasurementofhumanbodymovements.JBiomech26.ChenKY,BassettDR:Thetechnologyofaccelerometry-basedactivitymonitors:currentandfuture.MedSciSportsExer27.ServaisSB,WebsterJG,MontoyeHJ:Estimatinghumanenergyexpenditureusinganaccelerometerdevice.JClinEng28.MeijerGA,WesterterpKR,KoperH,HoorFT:Assessmentofenergyexpenditurebyrecordingheartrateandbodyacceleration.MedSciSportsExerc29.CavagnaG,SaibeneF,MargariaR:Athree-directionalaccelerometerforanalyzingbodymovements.JApplPhysiol30.DuboisAB,CavagnaGA,FoxRS:Locomotionofbluefish.JExpZool31.SellersWI,CromptonRH:Automaticmonitoringofprimatelocomotorbehaviourusingaccelerometers.FoliaPrimatologia32.MathieMJ,CellerBG,LovellNH,CosterACF:Classificationofbasicdailymovementsusingatriaxialaccelerometer.MedBiolEngineerComp33.ScheibeKM,SchleusnerT,BergerA,EichhornK,LangbeinJ,ZottoLD,StreichWJ:ETHOSYS(R)newsystemforrecordingandanalysisofbehaviouroffree-rangingdomesticanimalsandwildlife.ApplAnimBehavSci34.SellersWI,VarleyJS,WatersSS:Remotemonitoringoflocomotionusingaccelerometers:apilotstudy.FoliaPrimatol35.YodaK,SatoK,NiizumaY,KuritaM,BostCA,MahoYL,NaitoY:monitoringofporpoisingbehaviourofAdéliepenguinsdeterminedusingaccelerationdataloggers.JExpBiol36.DavisRW,FuimanLA,WilliamsTM,CollierSO,HageyWP,KanatousSB,KohinS,HorningM:HuntingbehaviorofamarinemammalbeneaththeAntarcticfastice.Science37.KempB,JanssenAJMW,VanDerKampB:Bodypositioncanbemonitoredin3Dusingminiatureaccelerometersandearth-magneticfieldsensors.ElectroencephalogrClinNeurophysiol38.HollandRA,WikelskiM,KümmethF,BosqueC:Thesecretlifeofoilbirds:Newinsightsintothemovementecologyofauniqueavianfrugivore.PLoSONE39.JohnsonMP,TyackPL:Adigitalacousticrecordingtagformeasuringtheresponseofwildmarinemammalstosound.IEEEJOceanicEngineer40.MurchieKJ,CookeSJ,DanylchukAJ,SuskiCD:Estimatesoffieldactivityandmetabolicratesofbonefish(Albulavulpes)incoastalmarinehabitatsusingacoustictri-axialaccelerometertransmittersandintermittent-flowrespirometry.JExperimentMarineBiolEcol41.SatoK,MitaniY,CameronMF,SiniffDB,NaitoY:FactorsaffectingstrokingpatternsandbodyangleindivingWeddellsealsundernaturalJExpBiol42.MitaniY,AndrewsRD,SatoK,KatoA,NaitoY,CostaDP:restingbehaviourofnorthernelephantseals:driftinglikeafallingleaf.BiolLett43.vanOortBEH,TylerNJC,StoreheierPV,StokkanK-A:Theperformanceandvalidationofadataloggerforlong-termdeterminationofactivityinfree-rangingreindeer,RangifertarandusL.AppAnimBehavSci2004,89:299308.44.GervasiV,BrunbergS,SwensonJE:Anindividual-basedmethodtomeasureanimalactivitylevels:atestonBrownBears.WildlifeSocBull45.SakamotoY,KunisakiT,SawaguchiI,AoiT,HarashinaK,DeguchiY:AnoteondailymovementpatternsofafemaleAsiaticblackbear(ursusthibetanus)inasuburbanareaofIwateprefecture,northeasternJapan.MammalStudy46.WhitneyNM,PapastamatiouYP,HollandKN,LoweCG:Useofanaccelerationdataloggertomeasuredielactivitypatternsincaptivewhitetipreefsharks,Triaenodonobesus.AquaticLivingRes2007,20:299305.47.WeimerskirchH,ShafferSA,TremblayY,CostaDP,GadenneH,KatoA,Ropert-CoudertY,SatoK,AuriolesD:Species-andsex-specificdifferencesinforagingbehaviourandforagingzonesinblue-footedandbrownboobiesintheGulfofCalifornia.MarineEcolProgSer48.LaichAG,QuintanaF,ShepardELC,WilsonRP:Intersexualdifferencesinthedivingbehaviourofimperialcormorants.JOrnithol49.ZimmerI,Ropert-CoudertY,KatoA,AncelA,ChiaradiaA:DoesforagingperformancechangewithAgeinfemalelittlepenguins(eudyptulaPLoSONE50.ByrnesG,LimNTL,YeongC,SpenceAJ:Sexdifferencesinthelocomotorecologyofaglidingmammal,theMalayancolugo(GaleopterusJMammaletal.AnimalBiotelemetryPage13of16http://www.animalbiotelemetry.com/content/1/1/20 51.LeVaillantM,WilsonRP,KatoA,SarauxC,HanuiseN,PrudHommeO,LeMahoY,LeBohecC,Ropert-CoudertY:Kingpenguinsadjusttheirdivingbehaviourwithage.JExperimentBiol52.FossetteS,SchofieldG,LilleyMKS,GleissAC,HaysGC:AccelerationdatarevealtheenergymanagementstrategyofamarineectothermduringFunctionalEcol53.KokubunN,KimJH,ShinHC,NaitoY,TakahashiA:Penguinheadmovementdetectedusingsmallaccelerometers:aproxyofpreyencounterrate.JExperimentBiol54.ViviantM,TritesAW,RosenDAS,MonestiezP,GuinetC:PreycaptureattemptscanbedetectedinStellersealionsandothermarinepredatorsusingaccelerometers.PolarBiol55.WatanabeS,IzawaM,KatoA,Ropert-CoudertY,NaitoY:Anewtechniqueformonitoringthedetailedbehaviourofterrestrialanimals:acasestudywiththedomesticcat.AppAnimBehavSci56.NathanR,SpiegelO,Fortmann-RoeS,HarelR,WikelskiM,GetzWM:tri-axialaccelerationdatatoidentifybehavioralmodesoffree-ranginganimals:generalconceptsandtoolsillustratedforgriffonvultures.JExperimentBiol57.YodaK,NaitoY,SatoK,TakahashiA,NishikawaJ,Ropert-CoudertY,KuritaM,MahoYL:Anewtechniqueformonitoringthebehavioroffree-rangingAdéliepenguins.JExperimentBiol2001,204:685690.58.KappelerPM,ErkertHG:Onthemovearoundtheclock:correlatesanddeterminantsofcathemeralactivityinwildredfrontedlemurs(fulvusrufusBehavioralEcolSociobiol59.ErkertHG,KappelerPM:Arrivedinthelight:dielandseasonalactivitypatternsinwildVerreauxssifakas(Propithecusv.verreaux;Primates:BehavioralEcolSociobiol60.GillyWF,ZeidbergLD,BoothJAT,StewartJS,MarshallG,AbernathyK,BellLE:LocomotionandbehaviorofHumboldtsquid,dosidicusgigas,inrelationtonaturalhypoxiainthegulfofCalifornia,Mexico.JExperimentBiol2012,215:31753190.61.BarasE,TogolaB,SicardB,BenechV:BehaviouroftigerfishHydrocynusbrevisintheRiverNiger,Mali,asrevealedbysimultaneoustelemetryofactivityandswimmingdepth.62.MoreauM,SiebertS,BuerkertA,SchlechtE:Useofatri-axialaccelerometerforautomatedrecordingandclassificationofgoatsgrazingbehaviour.AppAnimBehavSci2009,119:158170.63.OTooleAC,MurchieKJ,PullenC,HansonKC,SuskiCD,DanylchukAJ,CookeSJ:Locomotoryactivityanddepthdistributionofadultgreatbarracuda(Sphyraenabarracuda)inBahamiancoastalhabitatsdeterminedusingaccelerationandpressurebiotelemetrytransmitters.MarineFreshwaterRes64.WilsonRP,QuintanaF,HobsonVJ:Constructionofenergylandscapescanclarifythemovementanddistributionofforaginganimals.ProcRoySocB-BiolSci65.KaysR,JansenPA,KnechtEMH,VohwinkelR,WikelskiM:TheeffectoffeedingtimeondispersalofVirolaseedsbytoucansdeterminedfromGPStrackingandaccelerometers.ActaOecologica-IntJEcol66.Ropert-CoudertY,GremilletD,KatoA,RyanPG,NaitoY,MahoYL:Afine-scaletimebudgetofCapegannetsprovidesinsightsintotheforagingstrategiesofcoastalseabirds.AnimBehav67.RothwellES,BercovitchFB,AndrewsJRM,AndersonMJ:EstimatingdailywalkingdistanceofcaptiveAfricanelephantsusinganaccelerometer.ZooBiol68.TakahashiM,TobeyJR,PisacaneCB,AndrusCH:EvaluatingtheutilityofanaccelerometerandurinaryhormoneanalysisasindicatorsofestrusinaZoo-housedkoala(phascolarctoscinereusZooBiol69.ThiermanJL,CroweTG,StookeyJM,ValentineB:Quantificationoftheresponseofelkduringvelvetantlerremoval.CanAgriEngineer70.BergerA,ScheibeK-M,MichaelisS,StreichWJ:Evaluationoflivingcondi-tionsoffree-ranginganimalsbyautomatedchronobiologicalanalysisofBehavReseMethodsInstrComp71.SchaerBLD,RyanCT,BostonRC,NunamakerDM:Thehorse-racetrackinterface:apreliminarystudyontheeffectofshoeingonimpacttraumausinganovelwirelessdataacquisitionsystem.EquineVetJ72.CornouC,Lundbye-ChristensenS:Classifyingsowsactivitytypesfromaccelerationpatternsanapplicationofthemulti-processkalmanfilter.AppAnimBehavSci73.WhiteBJ,CoetzeeJF,RenterDG,BabcockAH,ThomsonDU,AndresenD:Evaluationoftwo-dimensionalaccelerometerstomonitorbehaviorofbeefcalvesaftercastration.AmJVetRes74.SullivanEL,CameronJL:Arapidlyoccurringcompensatorydecreaseinphysicalactivitycounteractsdiet-inducedweightlossinfemaleAmJPhysiolRegIntegComparPhysiol75.CookeSJ:Biotelemetryandbiologginginendangeredspeciesresearchandanimalconservation:relevancetoregional,national,andIUCNRedListthreatassessments.EndangeredSpecRes76.KroneO,BergerA,SchulteR:Recordingmovementandactivitypatternofawhite-tailedSeaeagle(haliaeetusalbicilla)byaGPSdatalogger.JOrnithol77.WilsonRP,WhiteCR,QuintanaF,HalseyLG,LiebschN,MartinGR,ButlerPJ:Movingtowardsaccelerationforestimatesofactivity-specificmetabolicrateinfree-livinganimals:thecaseofthecormorant.JAnimEcol78.TsudaY,KawabeR,TanakaH,MitsunagaY,HiraishiT,YamamotoK,NashimotoK:Monitoringthespawningbehaviourofchumsalmonwithanaccelerationdatalogger.EcolFreshwaterFish79.YasudaT,AraiN:Changesinflipperbeatfrequency,bodyangleandswimmingspeedoffemalegreenturtlesCheloniamydasMarineEcolProgressSer80.ZimmerI,Ropert-CoudertY,PoulinN,KatoA,ChiaradiaA:Evaluatingtherelativeimportanceofintrinsicandextrinsicfactorsontheforagingactivityoftoppredators:acasestudyonfemalelittlepenguins.MarineBiol81.JensenMB:Behaviouraroundthetimeofcalvingindairycows.AppAnimBehavSci82.WhitneyNM,HaroldLPJr,PrattTC,CarrierJC:Identifyingsharkmatingbehaviourusingthree-dimensionalaccelerationloggers.EndangeredSpecRes83.LagardeF,GuillonM,DubrocaL,BonnetX,BenKaddourK,SlimaniT,ElMoudenEH:Slownessandacceleration:anewmethodtoquantifytheactivitybudgetofchelonians.AnimBehav84.IsmailA,RahmanF,MiyazakiN,NaitoY:Initialapplicationofbio-loggingtechniquesoncaptiveMilkyStork(Mycteriacinerea)inMalaysia.TropEcol85.RushenJ,dePassilleAM:Automatedmeasurementofaccelerationcandetecteffectsofage,dehorningandweaningonlocomotorplayofAppAnimBehavSci86.HollandKN,MeyerCG,DagornLC:Inter-animaltelemetry:resultsfromfirstdeploymentofacousticbusinesscardEndangeredSpecRes87.DeRouffignacC,MorelF:Acomparativestudyofwaterturnoverin4rodentspeciesofwhich2arefromthedesertMeriones-Shawiratmousegerbil.JdePhysiologie(Paris)88.ArnouldJPY,BoydIL,SpeakmanJR:TherelationshipbetweenforagingbehaviourandenergyexpenditureinAntarcticfurseals.JZool(London)89.BrownJH,GilloolyJF,AllenAP,SavageVM,WestGB:Towardametabolictheoryofecology.90.AltmannS,AltmannJ:Thetransformationofbehaviourfieldstudies.AnimBehav91.SpeakmanJR:DoublyLabelledWater:TheoryandPractice.CambridgeUniversityPress;1997.92.ButlerPJ,GreenJA,BoydIL,SpeakmanJR:Measuringmetabolicrateinthefield:theprosandconsofthedoublylabelledwaterandheartrateFunctionalEcol93.Ropert-CoudertY,WilsonRP:Trendsandperspectivesinanimal-attachedremotesensing.FrontEcolEnviron94.HindleAG,RosenDAS,TritesAW:SwimmingdepthandoceancurrentsaffecttransitcostsinStellersealionsEumetopiasjubatus.AquaBiol95.FahlmanA,WilsonR,SvardC,RosenDAS,TritesAW:ActivityanddivingmetabolismcorrelateinStellersealionEumetopiasjubatus.AquaBiol96.HalseyLG,WhiteCR,EnstippMR,WilsonRP,ButlerPJ,MartinGR,GremilletD,JonesDR:Assessingthevalidityoftheaccelerometrytechniqueforestimatingtheenergyexpenditureofdivingdouble-crestedcormorantsphalacrocoraxauritus.PhysiolBiochemZool2011,84:230237.etal.AnimalBiotelemetryPage14of16http://www.animalbiotelemetry.com/content/1/1/20 97.WilsonRP,ShepardELC,LaichAG,FrereE,QuintanaF:Pedallingdownhillandfreewheelingup;apenguinperspectiveonforaging.AquaBiol98.ShepardELC,WilsonRP,QuintanaF,LaichAGM,FormanDW:fortimeorsavingonfuel:fine-scaleenergybudgetsshedlightoncurrenciesinadivingbird.ProcRSocSerBBiolSci99.ShepardELC,WilsonRP,LaichAG,QuintanaF:Buoyedupandsloweddown:speedlimitsfordivingbirdsinshallowwater.AquaBiol100.GleissAC,NormanB,WilsonRP:Movedbythatsinkingfeeling:variabledivinggeometryunderliesmovementstrategiesinwhalesharks.FunctionalEcol101.GleissAC,DaleJJ,HollandKN,WilsonRP:Acceleratingestimatesofactivity-specificmetabolicrateinfishes:testingtheapplicabilityofaccelerationdata-loggers.JExperimentMarineBiolEcol102.FossetteS,GleissAC,MyersAE,GarnerS,LiebschN,WhitneyNM,HaysGC,WilsonRP,LutcavageME:Behaviourandbuoyancyregulationinthedeepest-divingreptile:theleatherbackturtle.JExperimentBiol103.EnstippMR,CiccioneS,GinesteB,MilbergueM,BallorainK,Ropert-CoudertY,KatoA,PlotV,GeorgesJY:Energyexpenditureoffreelyswimmingadultgreenturtles(Cheloniamydas)anditslinkwithbodyacceleration.JExperimentBiol2011,214:40104020.104.HalseyLG,JonesTT,JonesDR,LiebschN,BoothDT:MeasuringenergyexpenditureinSub-adultandhatchlingSeaturtlesviaaccelerometry.PLoSONE105.RobsonAA,ChauvaudL,WilsonRP,HalseyLG:Smallactions,bigcosts:thebehaviouralenergeticsofacommerciallyimportantinvertebrate.JRSocInterface106.HalseyLG,ShepardELC,QuintanaF,GomezLaichA,GreenJA,WilsonRP:Therelationshipbetweenoxygenconsumptionandbodyaccelerationinarangeofspecies.ComparBiochemiPhysiolPartAMoleIntegrPhysiol107.GreenJA,HalseyLG,WilsonRP,FrappellPB:Estimatingenergyexpenditureofanimalsusingtheaccelerometrytechnique:activity,inactivityandcomparisonwiththeheart-ratetechnique.JExpBiol108.HalseyLG,PortugalSJ,SmithJA,MurnCP,WilsonRP:Recordingraptorbehavioronthewingviaaccelerometry.JFieldOrnithol109.HalseyLG,WhiteCR:Measuringenergeticsandbehaviourusingaccelerometryincanetoadsbufomarinus.PLoSONE110.HalseyLG,ShepardELC,WilsonRP:AssessingthedevelopmentandapplicationoftheaccelerometrytechniqueforestimatingenergyComparBiochemPhysiolMoleIntegrPhysiol111.ShepardELC,WilsonRP,HalseyLG,QuintanaF,GomezLaichA,GleissAC,LiebschN,MyersAE,NormanB:Derivationofbodymotionviaappropriatesmoothingofaccelerationdata.AquaBiol112.QasemL,CardewA,WilsonA,GriffithsI,HalseyLG,ShepardELC,GleissAC,WilsonR:Tri-axialdynamicaccelerationasaproxyforanimalenergyexpenditure;shouldwebesummingvaluesorcalculatingthevector?PLoSONE113.LeleuC,BarillerF,CotrelC,BarreyE:Reproducibilityofalocomotortestfortrotterhorses.VetJ114.LeleuC,CotrelC,BarreyE:RelationshipsbetweenbiomechanicalvariablesandraceperformanceinFrenchStandardbredtrotters.LivestockProdSci115.WitteTH,KnillK,WilsonAM:Determinationofpeakverticalgroundreactionforcefromdutyfactorinthehorse(Equuscaballus).JExperimentBiol116.RatzlaffMH,WilsonPD,HuttonDV,SlinkerBK:Relationshipsbetweenhoof-accelerationpatternsofgallopinghorsesanddynamicpropertiesoftheAmJVetRes117.CottinF,MetayerN,GoachetAG,JulliandV,SlawinskiJ,BillatV,BarreyE:OxygenconsumptionandgaitvariablesofArabianendurancehorsesmeasuredduringafieldexercisetest.EquineVetJ118.BrischouxF,KatoA,Ropert-CoudertY,ShineR:Swimmingspeedvariationinamphibiousseasnakes(laticaudinae):asearchforunderlyingmecha-JExperimentMarineBiolEcol119.SatoK,SakamotoKQ,WatanukiY,TakahashiA,KatsumataN,BostC-A,WeimerskirchH:Scalingofsoaringseabirdsandimplicationsforflightabilitiesofgiantpterosaurs.PLoSONE120.HoytDF,WicklerSJ,CoggerEA,GoehringME:Areexaminationofthetrot-galloptransition:insightsfromthestudyoflocomotiononanincline.AmZool121.TanakaH,TakagiY,NaitoY:SwimmingspeedsandbuoyancycompensationofmigratingadultchumsalmonOncorhynchusketarevealedbyspeed/depth/accelerationdatalogger.JExperimentBiol122.Ropert-CoudertY,KatoA,WilsonRP,CannellB:Foragingstrategiesandpreyencounterrateoffree-rangingLittlePenguins.MarineBiol(Berlin)123.SatoK,WatanukiY,TakahashiA,MillerPJO,TanakaH,KawabeR,PonganisPJ,HandrichY,AkamatsuT,WatanabeY,MitaniY,CostaDP,BostCA,AokiK,AmanoM,TrathanP,ShapiroA,NaitoY:Strokefrequency,butnotswimmingspeed,isrelatedtobodysizeinfree-rangingseabirds,pinnipedsandcetaceans.ProcRSocSeriesBBiolSci2007,274:471477.124.HindleAG,YoungBL,RosenDAS,HaulenaM,TritesAW:Diveresponsediffersbetweenshallow-anddeep-divingStellersealions(EumetopiasJExperimentMarineBiolEcol125.KatoA,Ropert-CoudertY,GremilletD,CannellB:Locomotionandforagingstrategyinfoot-propelledandwing-propelledshallow-divingseabirds.MarineEcolProgSeries126.WatanukiY,WanlessS,HarrisM,LovvornJR,MiyazakiM,TanakaH,SatoK:Swimspeedsandstrokepatternsinwing-propelleddivers:acomparisonamongalcidsandapenguin.JExperimentBiol127.KawabeR,NashimotoK,HiraishiT,NaitoY,SatoK:Anewdeviceformonitoringtheactivityoffreelyswimmingflatfish,JapaneseflounderParalichthysolivaceusFisheriesSci128.WilliamsTM,DavisRW,FuimanLA,FrancisJ,LeBoeufBL,HorningM,CalambokidisJ,CrollDA:Sinkorswim:strategiesforcost-efficientdivingbymarinemammals.Science129.GoldbogenJA,CalambokidisJ,ShadwickRE,OlesonEM,McDonaldMA,HildebrandJA:Kinematicsofforagingdivesandlunge-feedinginfinJExperimentBiol130.AraiN,KurokiM,SakamotoW,NaitoY:AnalysisofdivingbehaviorofAdéliepenguinsusingaccelerationdatalogger.PolarBiosci2000,13:100.131.WatanukiY,TakahashiA,DauntF,WanlessS,HarrisM,SatoK,NaitoY:Regulationofstrokeandglideinafoot-propelledaviandiver.JExperimentBiol132.SoltisJ,WilsonRP,Douglas-HamiltonI,VollrathF,KingLE,SavageA:AccelerometersincollarsidentifybehavioralstatesincaptiveAfricanLoxodontaafricanaEndangeredSpeciesRes133.GreenJA,WhiteCR,BunceA,FrappellPB,ButlerPJ:consequencesofplungedivingingannets.EndangeredSpeciesRes134.FossetteS,GasparP,HandrichY,LeMahoY,GeorgesJ-Y:DiveandbeakmovementpatternsinleatherbackturtlesDermochelyscoriaceaduringinternestingintervalsinFrenchGuiana.JAnimEcol2008,236246.135.GleissAC,WilsonRP,ShepardELC:Makingoveralldynamicbodyaccelerationwork:onthetheoryofaccelerationasaproxyforenergyMethodsEcolEvol136.KaysR,TilakS,CrofootMC,FountainT,ObandoD,OrtegaA,KuemmethF,MandelJ,SwensonG,LambertT,HirschB,WikelskiM:Trackinganimallocationandactivitywithanautomatedradiotelemetrysysteminatropicalrainforest.CompJ137.ShiomiK,SatoK,PonganisPJ:Pointofnoreturnindivingemperorpenguins:isthetimingofthedecisiontoreturnlimitedbythenumberofstrokes?JExperimentBiol138.Ropert-CoudertY,WilsonRP:Subjectivityinbio-loggingscience:dologgeddatamislead?MemoirsNatInstPolarRes139.BidderOR,SoresinaM,ShepardEL,HalseyLG,QuintanaF,GomezLaichA,WilsonRP:Theneedforspeed:testingaccelerationforestimatinganimaltravelratesinterrestrialdead-reckoningsystems.140.ToblerMW:NewGPStechnologyimprovesfixsuccessforlargemammalcollarsindensetropicalforests.JTropEcol141.NarazakiT,SatoK,AbernathyKJ,MarshallGJ,MiyazakiN:SeaturtlescompensatedeflectionofheadingattheseasurfaceduringdirectionalJExperimentBioletal.AnimalBiotelemetryPage15of16http://www.animalbiotelemetry.com/content/1/1/20 142.WilsonRP,LiebschN,DaviesIM,QuintanaF,WeimerskirchH,StorchS, LuckeK,SiebertU,ZanklS,MüllerG,ZimmerI,ScolaroA,CampagnaC, PlötzJ,BornemannH,TeilmannJ,McMahonCR: Allatseawithanimal tracks;methodologicalandanalyticalsolutionsfortheresolutionof movement. DeepSeaResPartIITopicStudOceanogr 2007, 54: 193 – 210. 143.ShiomiK,SatoK,MitamuraH,AraiN,NaitoY,PonganisPJ: Effectofocean currentonthedead-reckoningestimationof3-Ddivepathsofemperor penguins. AquaBiol 2008, 3: 265 – 270. 144.ShiomiK,NarazakiT,SatoK,ShimataniK,AraiN,PonganisPJ,MiyazakiN: Data-processingartefactsinthree-dimensionaldivepathreconstruction fromgeomagneticandaccelerationdata. AquaBiol 2010, 8: 299 – 304. 145.HoughtonJDR,LiebschN,DoyleTK,GleissAC,LilleyMKS,WilsonRP,Hays GC: Harnessingthesun:testinganovelattachmentmethodtorecord finescalemovementsinoceansunfish(molamola). In Taggingand TrackingofMarineAnimalswithElectronicDevices. NewYork:Springer: NielsenJL,ArrizabalagaH,FragosoN,HobdayA,LutcavageM,SibertJ (Eds.);2009:229 – 242. 146.dePassilléAM,JensenMB,ChapinalN,RushenJ: Useofaccelerometersto describegaitpatternsindairycalves. JDairySci 2010, 93: 3287 – 3293. 147.GleissAC,NormanB,LiebschN,FrancisC,WilsonRP: Anewprospectfor tagginglargefree-swimmingsharkswithmotion-sensitivedata-loggers. FisheriesRes(Amsterdam) 2009, 97: 11 – 16. 148.SignerC,RufT,SchoberF,FluchG,PaumannT,ArnoldW: Aversatile telemetrysystemforcontinuousmeasurementofheartrate,body temperatureandlocomotoractivityinfree-rangingruminants. MethodsEcolEvol 2010, 1: 75 – 85. 149.WatanabeN,SakanoueS,KawamuraK,KozakaiT: Developmentofan automaticclassificationsystemforeating,ruminatingandresting behaviorofcattleusinganaccelerometer. GrasslandSci 2008, 54: 231 – 237. 150.LedgerwoodDN,WincklerC,TuckerCB: Evaluationofdataloggers, samplingintervals,andeditingtechniquesformeasuringthelying behaviorofdairycattle. JDairySci 2010, 93: 5129 – 5139. 151.KeeganKG,KramerJ,YonezawaY,MakiH,PaiPF,DentEV,KellermanTE, WilsonDA,ReedSK: Assessmentofrepeatabilityofawireless,inertial sensor-basedlamenessevaluationsystemforhorses. AmJVetRes 2011, 72: 1156 – 1163. 152.PrestonT,BaltzerW,TrostS: Accelerometervalidityandplacementfor detectionofchangesinphysicalactivityindogsundercontrolled conditionsonatreadmill. ResVetSci 2012, 93: 412 – 416. 153.OkuyamaJ,KawabataY,NaitoY,AraiN,KobayashiM: Monitoringbeak movementswithanaccelerationdatalogger:ausefultechniquefor assessingthefeedingandbreathingbehaviorsofseaturtles. EndangeredSpeciesRes 2009, 10: 39 – 45. 154.SuzukiI,NaitoY,FolkowLP,MiyazakiN,BlixAS: Validationofadevicefor accuratetimingoffeedingeventsinmarineanimals. PolarBiol 2009, 32: 667 – 671. 155.SkinnerJP,NorbergSE,AndrewsRD: Headstrikingduringfishcapture attemptsbyStellersealionsandthepotentialforusingheadsurge accelerationtopredictfeedingbehavior. EndangeredSpeciesRes 2010, 10: 61 – 69. 156.IwataT,SakamotoKQ,TakahashiA,EdwardsEWJ,StanilandIJ,TrathanPN, NaitoY: Usingamandibleaccelerometertostudyfine-scaleforaging behavioroffree-rangingAntarcticfurseals. MarineMammalSci 2012, 28: 345 – 357. 157.SatoK,DauntF,WatanukiY,TakahashiA,WanlessS: Anewmethodto quantifypreyacquisitionindivingseabirdsusingwingstrokefrequency. JExperimentBiol 2008, 211: 58 – 65. 158.RobertB,WhiteBJ,RenterDG,LarsonRL: Evaluationofthree-dimensional accelerometerstomonitorandclassifybehaviorpatternsincattle. CompElectronAgri 2009, 67: 80 – 84. 159.HalseyLG,ShepardELC,HulstonCJ,VenablesMC,WhiteCR,JeukendrupAE, WilsonRP: Accelerationversusheartrateforestimatingenergy expenditureandspeedduringlocomotioninanimals:testswithaneasy modelspecies,Homosapiens. Zoology 2008, 111: 231 – 241. 160.GleissAC,GruberSH,WilsonRP: Multi-channeldata-logging:towards determinationofbehaviourandmetabolicrateinfree-swimmingsharks. In TaggingandTrackingofMarineAnimalswithElectronicDevices. NewYork: Springer:NielsenJL,ArrizabalagaH,FragosoN,HobdayA,LutcavageM, SibertJ;2009:211 – 228. 161.ScheibeKM,GromannC: Applicationtestingofanewthree-dimensional accelerationmeasuringsystemwithwirelessdatatransfer(WAS)for behavioranalysis. BehavResMethods 2006, 38: 427 – 433. 162.YostM,CooperRA,BremnerFJ: Fourieranalyses:amathematicaland geometricexplanation. BehavResMethodsInst 1983, 15: 258 – 261. 163.WatanabeS,SatoK,PonganisPJ: Activitytimebudgetduringforaging tripsofemperorpenguins. PLoSONE 2012, 7 (11):e50357. 164.SakamotoKQ,SatoK,IshizukaM,WatanukiY,TakahashiA,DauntF,Wanless S: Canethogramsbeautomaticallygeneratedusingbodyacceleration datafromfree-rangingbirds? PLoSONE 2009, 4 (4):e5379. 165.MartiskainenP,JärvinenM,SkönJ-P,TiirikainenJ,KolehmainenM,MononenJ: Cowbehaviourpatternrecognitionusingathree-dimensionalaccelerom- eterandsupportvectormachines. AppAnimBehavSci 2009, 119: 32 – 38. 166.GaoL,CAmpbellHA,BidderOR,HunterJ: AWeb-basedsemantictagging andactivityrecognitionsystemforspecies ’ accelerometrydata. EcologicalInfo 2012, 13: 47 – 56. 167.KaplanHL: Correlations,contrasts,andcomponents:Fourieranalysisina morefamiliarterminology. BehavResMethodsInst 1983, 15: 228 – 241. 168.HartT,MannR,CoulsonT,PettorelliN,TrathanP: Behaviouralswitchingin acentralplaceforager:patternsofdivingbehaviourinthemacaroni penguin(Eudypteschrysolophus). MarineBiol 2010, 157: 1543 – 1553. 169.JuleKR,LeaSEG,LeaverL: Usingabehaviourdiscoverycurvetopredict optimalobservationtime. Behaviour 2009, 146: 1531 – 1542. 170.GrundyE,JonesMW,LarameeRS,WilsonRP,ShepardELC: Visualisationof sensordatafromanimalmovement. IEEE-VGTCSymposiumVisual 2009, 28: 815 – 822. 171.Ropert-CoudertY,GremilletD,RyanP,KatoA,NaitoY,LeMahoY: Between airandwater:theplungediveofthecapegannetmoruscapensis. Ibis 2004, 146: 281 – 290. 172.NadimiES,SogaardHT,BakT: ZigBee-basedwirelesssensornetworksfor classifyingthebehaviourofaherdofanimalsusingclassificationtrees. BiosystemsEngineer 2008, 100: 167 – 176. doi:10.1186/2050-3385-1-20 Citethisarticleas: Brown etal. : Observingtheunwatchablethrough accelerationloggingofanimalbehavior. AnimalBiotelemetry 2013 1 :20. Submit your next manuscript to BioMed Central and take full advantage of: € Convenient online submission € Thorough peer review € No space constraints or color “gure charges € Immediate publication on acceptance € Inclusion in PubMed, CAS, Scopus and Google Scholar € Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit Brown etal.AnimalBiotelemetry 2013, 1 :20 Page16of16 http://www.animalbiotelemetry.com/content/1/1/20