Calibration ofvisuallyestimateddistancestomigrating seabirds with radarmeasurements

Mar´ıaMateos,1,2,4GonzaloM.Arroyo,1,2AlejandroRodr´ıguez,3DavidCuenca,1

andAndre´sdelaCruz2

1BiologyDepartment, Facultyof MarineandEnvironmentalSciences,University ofCa´diz,Av.RepublicaSaharaui,s/n,

11510PuertoReal,Ca´diz,Spain

2Fundacio´nMigres,ComplejoHuertaGrande,Ctra.N-340Km96.7,11390Pelayo,Algeciras,Ca´diz,Spain

3Department ofConservationBiology,Estacio´nBiolo´gicadeDon˜ana,CSIC,Am´ericoVespucios/n,IsladelaCartuja,

41092 Sevilla,Spain

ABSTRACT. Censusingseabirdsfromcoastalareasrequiresreliableestimatesofbirdnumbersandthedistances ofthebirdsfromthecoastline.Logisticalconstraintsmakevisualestimationofdistancestheonlyfeasiblemethod inmanystudies.Wetestedtheaccuracyofvisuallyestimatedoffshoredistancesofsix migratory seabirdspeciesin theStraitofGibraltarusingsimultaneousmeasurementsobtainedbyradar.Mostbirds(91%)weredetectedwithin

3kmofthecoastandwetruncatedourcalibrationatthisdistance.Wefoundastrongcorrelation betweenradar

andvisualestimates(R2=0.83,P0.0001).Themagnitudeoferrorsinvisualestimateswasmoderateandranged

from0.08to0.20fordifferentdistancesandobservers.Amongthefactorspotentiallyaffectingtheaccuracyof

visualestimatesofdistancetoseabirdinourstudywereobserveridentity,birdspecies,birdbehavior,andweather;

themostparsimoniousmodelinourstudyincludedobserver identityastheonlypredictor,andnomodelwithmore

thanonepredictorhadasmallerAkaike’sinformationcriterionvalue.Radarcanbeusedtohelptrainobserversand

toreducebiasesinvisualestimatesofdistancesbymeansofcalibration.Whennoothermethodsareavailableto

accuratelymeasuredistancestoseabirds,visualestimatesofdistances,recordedbyexperiencedobserversand once

calibratedwithradar(orotherground-truthingmethods),maybeacceptablefordifferentspeciesunderawide

rangeofenvironmentalconditions.

RESUMEN. Utilizacio´ndemedidasderadarparalacalibracio´ndedistanciasaavesmarinas migratorias estimadasvisualmente

Censaraves marinasdesdea´reascosterasrequiereestimacionesfiablesdelnu´merodeavesydeladistanciade estasacosta.Restriccioneslog´ısticashacenqueaveceslasestimaciones visualesdelasdistanciasseanelu´nico me´todofactiblepara llevaracaboestosestudios.Pusimosaprueba, enelEstrechodeGibraltar,laexactitudde lasestimaciones visualesdesdecosta,utilizandoseisespeciesdeavesmarinas,concurrentemente,conmedidas obtenidasporradar.Lamayor´ıadelasaves(91%)fuerondetectadasdentrodeuna´reade3kmdelacostay ajustamosnuestracalibracio´naestadistancia.Encontramosunafuertecorrelacio´nentrelasestimaciones visualesy lasdelradar(R2 =0.83,P0.0001).Lamagnituddeloserroresvisualesalestimarlasdistanciasfuemoderada yvariaronde 0.08a 0.20paradiferentesdistanciasyobservadores.Entrelosfactoresqueafectarona laexactitud delasestimacionesvisuales,seencuentranelobservador,la especiedeave,la conductaexhibidaporelaveylas condicionesmeteorolo´gicas.Elmodeloconmayorparsimoniaincluyo´laidentidaddelobservadorcomoelu´nico vaticinadoryningu´notromodelo,conma´sdeunvaticinador,tuvounvalordeAICmenor.Elradarpuedeser utilizadoparaentrenaralpersonaly reduceelsesgoenlasestimacionesvisualesmediantelacalibracio´n.Cuando nohaydisponibleotrosme´todosparamedirconexactitudladistanciaaquesemueven lasavesmarinas, elusode observadoresexperimentadosyunacalibracio´nconradar(uotrome´todo),puedenseraceptablesparatrabajarcon diferentesespeciesbajounaampliagamadecondicionesambientales.

Keywords: coastalmigratoryroutes,distancecalibration,distanceunderestimation,seabirdabundance,sea- clutter,StraitofGibraltar

Distancesamplingisawidespread method for estimatinganimalabundance(Buckland et al.2001, 2004), and maybe usefulfor estimatingdensitiesof marinebirds(Komdeur

4Corresponding author. Email: maria.mateos@

uca.es

etal.1992,Camphuysen etal.2004).Along seabirdmigratorypathwaysorfrequentlyused flightpaths,suchasbetweenforagingareas andbreedingcolonies,surveysfromstationary observationplatformsarerequiredto census birds(Camphuysenetal.2004).Thesecounts mayyieldabundanceestimateswithhighac- curacyprovidedthatdistancestobirdscanbe

measuredwithoutbiases(Bucklandetal.2004). Differentprocedureshavebeenusedtoobtain thosedistances,includingvisualestimates,opti- calrangefinders(Heinemann1981), binocular reticules(Yenetal.2004),radars(Petersenetal.

2006),andlaserrangefinders(Ransomand Pinchak2003).Characteristicsofthesurveyed populations,environmentalconditions,andthe accuracyrequiredwilldictatewhichmethods shouldbeused(e.g.,Bibbyetal.1992).Estimat- ingdistancebyeyehasbeencriticized,butcould betheonlyoptionundersome circumstances (Bibbyetal.1992),andnearlyallseabirdcensus work,whetherusingshore-basedorshipboard observers,reliesonvisuallyestimateddistances. However,noassessmentoftheaccuracywith whichobserversestimatethesedistancesexists. Previousattemptstomeasureerrorindistance estimatestoanimalsatsea,andthepossible effectsoffactorssuchasweatherontheaccuracy ofvisualestimatesofdistance,havefocusedon cetaceans(Schweder1997,BairdandBurkhart

2000,Williamsetal.2007).Becausedetection conditionsandmethodsforestimatingdistances tocetaceansdifferfromthoseusedforseabirds, calibratingdistanceestimatesto birdsatsea mayrequire specificstudiesaswell.However, toourknowledge,nostudyhasassessederrors indistanceestimatestoseabirdsorhasexamined thefactorsthatmayinfluencetheseerrors.

Differenttypesoferrormayarisefromvisual estimationofdistances,withbiaseddistance estimatesposingthemainproblembecausethey aredifficulttodetectandwilltransfererrorsto theestimatesofbirdabundance.Sucherrors canbecorrectedeitherinthefieldthrough testingprocedures and calibration(Marques

2004,Williamsetal.2007)orthroughstatis- ticaltechniquesemployedduringdataanalysis (Chen1998,Marques2004).Posthocanalytical approachesarelessrobustthan field-testing programs(Marques2004,Marquesetal.2006, Williamsetal.2007).Calibrationofdistance estimatestoseabirdsistherefore neededtoim- proveestimatesofbirdabundance.Calibration willbefurtherenhancediferrorsindistance estimatescouldbepredictedfrom factorssuch asbirdsize,birdbehavior,weatherconditions, orobserverability.

Werecordedandcomparedvisualandradar estimatesofdistancestoflyingseabirdsalonga migratorypathway.Ourspecificaimswereto: (1)testtheaccuracyofvisuallyestimatedoff-

shoredistancestoseabirdsbymeans ofradar measurements,and(2)determinetheeffectof birdsize,birdbehavior,weather,andobserver skillon theaccuracyofvisualestimates.

METHODS

Ourstudywasconductedfrom5Marchto

20April and16Octoberto18November2006

attheStraitofGibraltar(Fig.1).Thischannelis

14kmwideatitsnarrowestpointandistheonly

connectionbetweentheAtlantic Oceanandthe

MediterraneanSea.

FieldworkwascarriedoutatTarifaIsland

(southwestSpain),thesouthernmostpointof

thenorthcoastoftheStraitofGibraltar(Fig.1).

Fromthispoint,simultaneousobservationsof

migratingseabirdswerecollectedbyanobserver

andaradaroperator.Theradarantennaand

observerswereatthesamevantagepoint.Ob-

serversscannedtheseawithandwithouttheaid

ofbinoculars(VictoryFL10×42T;Zeiss,

Oberkochen,Germany)andtelescopes (Zeiss

20–60×85).Eachseabirdflockdetectedwas

identifiedtospecieslevelanditsdistancefrom

thecoastwasestimated bytheobserverwho

thencommunicatedthepositionoftheflock

anditsflightdirectiontotheradaroperator.

Ifthetargetwasfoundonthescreen,distance

wasalsomeasuredsimultaneously byan S-

bandsurveillanceradar(FR-2137SBB,Furuno,

Nishinomiya,Japan;peakpoweroutput30kW,

variablepulse0.07- 1.0 jis, transmittingat

3050MHz)locatedona platform10mabove

sealevel.Distanceestimatesweremadewhen

targetscrossedanimaginarylineperpendicular

tothecoastattheobservationpoint.

Atthebeginningofeachobservationperiod,

theradaroperator adjustedtheradarparameters

(range,gain,andseaandrainfilter)to optimize

the radarpicture.Wecombinedboththe range

functionandtheoff-centerfunction,placing

theorigin(antennaposition)inthelowerpart

ofthescreen, toreachascanningradiusof

eitherupto4.5km(shortrange) or9.1km

(longrange),withscanningradiusalternated

with eachobservationperiod.Gain,thereceptor

sensitivitytotheenergyreflected,wasadjusted

from45to100,where100correspondedtothe

maximumsensitivity.Sensitivitywasbetween

51and7075%ofthetime,between45and

5010%ofthetime,andbetween71and100

15%ofthetime.Theseaclutterfilterreduced

Fig.1. DetailsoftheAfricanandSpanishcoastlineintheareaofTarifa.Notethecharacteristicnarrow funnel-shapeoftheStraitofGibraltar.TarifaIslandisthemostefficientlocationforsurveyingseabirdsthe StraitofGibraltarfromthecoast (ProgramaMigres2009).

theclutterproducedbywaves(0=off;100= maximum).However,usingthisfiltermayresult inthe lossofweakechoes.Thebestsettingwas achievedwhenthesizeofthespotsproducedby seaclutterontheradarscreenwasminimized whileweakechoesunequivocallyproducedby flyingbirdsremainedvisible.Duringourstudy, theseaclutterfilter wassetatvaluesranging from0to40,withthefiltersetbetween11and

3075%ofthetime.Therainfilter(range=0–

100)reducedtheclutterproducedbytherain,

andwasalwayssetatvalues 30(andwassetat

zero90% oftime).

Datawerecollectedunderawiderangeof

meteorologicalconditions,andvisualobserva-

tionsweremadebythreeexperiencedobservers.

Allobservers hadcensusedseabirdsformore

than 4yearsatthestudysite,werefamiliarwith

thebirdspeciesandenvironmentalconditions

occurringinthestudyarea,andhadworked

withthe radarfor4monthsbeforethe onsetof

thisstudy.

Werestricted ouranalyses tothesixmost

abundantseabirds,includingNorthernGannets

(Morusbassanus),Cory’sShearwaters(Calonec-

trisdiomedea), BalearicShearwaters (Puffinus mauretanicus),AtlanticPuffins(Fraterculaarc- tica),Razorbills (Alcatorda),andGreatSkuas (Catharactaskua).Due to thedifficultyofiden- tificationand the frequentoccurrenceofmixed flocks,AtlanticPuffinsandRazorbillswerecon- sideredtogetherandreferredtoasauks.Foreach sighting,observersrecordedspecies,flocksize, flightdirection,andestimateddistance.

Eachhour,werecordedthespeedanddi- rectionofwindattheobservationsite.After examiningthefrequencydistributionofwind directionsduring2006attheTarifameteo- rologicalstation,winddirectionswerecatego- rizedaseithereastorwestbecausewindscame fromthesedirections94% oftime.Easterly winds(51.2%ofrecords)hadameandirection (±1SD)of79.8±22.2◦,andwesterlywinds (42.8%ofrecords)ameandirectionof271.0±

39.8◦.Allflyingbirdswere observedheading eithereastorwest.Flightdirectionswerecat- egorizedaseitherwithaheadwindortailwind (Spear andAinley1997).Visibilitywasassigned tooneofthreecategoriesbasedonhowclouds, haze, andlightlevelsaffected visibility ofthe

Africancoastlocated14kmaway:(1)good,with theAfricancoastclearlyvisible,(2)moderate, withonlythesilhouetteofthecoastvisible,and (3)bad,withtheAfricancoastnotvisible,but theseasurfacevisibleatadistanceofabout7km. Noobservationsweremadewhenboats,which typicallyfollowaroutealongthecenterof the strait,couldnotbeseen.

Althoughradarrecordswereusedtocalibrate visual estimates, radarmeasurementsarenot freefromerror.Thetechnicalspecificationsof ourradarestablishaninherentmeasurement errorof1%,withamaximumerrorof30m. Furthermore,birds appearonaradarscreen asatwo-dimensionalechoinsteadofaone- dimensionalpoint,especiallywheninflocks.We measuredthedistancetothecenteroftheecho. Thus,theremightbeasystematicerrorinthese radardistancesbecausethesizeofanechoonthe screencouldvarywiththedistancetothetarget duetothelevelofenergyreflected.Totestthe magnitudeofthiserror,werandomlyselecteda sampleof100targets.Wemeasuredthe echosize onthenorth–southaxisontheradarscreenwith theaidofageographical informationsystem (ArcView3.2.;EnvironmentalSystemsResearch Institute1999).Wefinallyexaminedwhether echosizevariedwithdistancetothetargetas measuredbyradar.

Statistical analyses. Theerrorofthe visualestimateofdistancewascalculated as

Error=|radarestimate–visualestimate|/radar estimate.Onlybirdsorflocksthatweredetected visuallyandhaddistancesestimatedwithradar wereincludedintheanalysis.Wedidnotassume thattherelationshipbetweenvisualestimatesof distance,ortheirerror,andthecorresponding radarestimateswerelinear.Therefore,weana- lyzedthisrelationshipwithgeneralizedadditive modelsthatmayfitnonlinear,smoothfunctions (HastieandTibshirani1990).

Second,weexaminedwhetherobserveriden-

tity,birdspecies,birdbehavior (flocksizeand

flightdirectionrelativetowinddirection),and

meteorologicalconditions(windspeedandvisi-

bility)influencedtheaccuracyofvisualestimates

ofdistance.Weusedvisuallyestimateddistances

asthe dependentvariableand includedthe

distanceestimatedbyradarasacovariatesothat

theeffectofpredictorswasassumedtoexplain

measurementerrors.Webuiltonemodelfor

eachpredictorandcomparedtheirfittothedata

withtheAkaike’sinformationcriterion(AIC).

Wethenexaminedwhetherthefitimproved whentwoormorepredictorswerecombined inthesamemodel. Weselectedthemostparsi- moniousmodelasthathavingthelowestAIC value(BurnhamandAnderson2002).

Analyses wereperformedwiththepackage

“gam”inR2.6.2(RDevelopmentCoreTeam

2008).Webuiltmodels assuming agamma dis-

tribution,usingthelog-linkfunction andthin

plateregression splinesassmoothsfunctions.

Weallowedforamaximumofthreedegreesof

freedomforthesmoothfunctions.Valuesare

reportedasmeans±1SD.

RESULTS

Observersmade1173 sightingsthat were subsequentlylocatedbyradar,including402

NorthernGannets,375Cory’sShearwaters,265

BalearicShearwaters, 117auks,and14Great

Skuas.Overall,42.4%ofsightingsweresingle

birdsandtherestwereflocks.

Wemade3.4%ofobservationsduringcalm

conditions(windspeed0.5ms−1),44.9%

duringperiodswitheasterlywinds(range=0.8–

15.8ms−1,mean=3.9±2.8ms−1),and51.7%

duringperiodswithwesterly winds(range=

0.6–9.4m s−1, mean=3.6 ±2.4 m s−1).

Visibilitywasgoodfor28.1%oftheobser-

vations,moderatefor 45.5%, and bad for

26.5%.

Radarestimatesofdistanceto birdsight-

ingsrangedfrom183to6913m(mean=

1559± 1009m),andvisuallyestimateddis-

tancesrangedfrom100 to 7000 m (mean

1438±978m).Mostcontacts(91%)were

within3000mofthecoast,as estimated by

radar,and,forfurtheranalysis,wetruncated our

dataatthatdistance.

Theestimatedsizeofechoesvariedfrom9to

33m(mean=17.9±7.8m,N=100flocks)

fordistancestothetargetsfrom105to2950m.

Echosizeontheradarscreendidnotvarywith

distancetothetarget(R2 =0.004,P=0.65).

Theerror indistanceestimatesattributedtothe

sizeof theechowassmall,always0.04,and,

becausemosttargetspassedwithin500m,this

errorwasoften0.02.

Visualestimatesofdistancewerecorrelated

withthecorrespondingradarestimates(R2 =

0.83,F =1135,P0.0001,N =1068;

Figs.2AandB).Theabsolutemeandifference

betweenvisualandradarestimatesofdistance

Fig.2. Relationshipbetweenradarestimatesofdistancetoflyingseabirds andcorresponding visualestimates intherange0–3km.Thebrokenlineindicatesanunbiasedrelationship.(A)Datarecordedbyobserver1, themostaccurateobserver.(B)Datarecordedbyobserver3,theleastaccurateobserver.

increased withincreasingdistance,withdif- ferencesof141m, 200m, and327m for distances0–1,1–2,and2–3km,respectively. The error,however,decreasedwithdistance, witherrorvaluesof0.20,0.15,and0.13for thethreedistanceintervals,respectively(overall meanerror=0.17±0.15).

Amongthefactorspotentiallyaffectingtheac-

curacyofvisualestimatesofdistanceto seabird,

themostparsimoniousmodelincludedobserver

identityas theonlypredictor(Table1),and

no modelwithtwoormorepredictorshad

asmaller AICvalue.Allobservers tendedto

underestimatedistances,andunderestimation

washigheratgreaterdistances(2km;Figs. 2A andB). Errorsinestimatingdistancewereless forobserver1(meanerror=0.13)thanfor observers2and3(meanerror =0.18and0.20, respectively)foralldistancesupto3km.Error differencesbetweenobserverswereexacerbated atgreaterdistances.Forthe2–3kmdistance interval,meanerrorswere0.08,0.19,and 0.20 forobservers1,2,and3, respectively.

DISCUSSION

Wefoundthatobserverscouldestimatedis- tanceswith reasonableaccuracy,asreflectedby

Table1. Generalizedadditivemodelsexamining theeffectoffourtypesofpredictors,namelyvisual observer,birdspecies,birdbehavior (flocksizeand flightdirection),andweatherconditions(windspeed andvisibility)onvisualestimatesofdistancesto seabirds.Distancesestimatedwithradarwerefittedas acovariateinallmodels,takingtheformofaspline with3df.

andBurkhart2000,Bucklandetal.2001),the threeobserversinourstudytendedtounderes- timatedistances.Underestimationincreasedat greaterdistances,whichwas alsoobservedin flaggingexperiments(Camp2007).Moreover, wedetectedanobserver-specificeffectofunder- estimationthatproduced asystematicerrorthat

shouldbecorrectedbeforeusingdatatoesti-

M odel Predictors AICa AAICb wic

aAkaike’sinformationcriterion.

bAkaikedifferences.

cAkaikeweights.

thesignificant correlation betweenradarand visualestimates. Furthermore,theerrorofvi- sualestimateswasrelativelylow(between0.08 and0.20,dependingontherangeof distances andobserver)andsimilar to errorsreported forvisualestimatesofdistancestocetaceans (Williamsetal.2007).Therequiredlevelof accuracyindistanceestimatesvariesamongspe- cificecologicalapplications.However,Marques (2007)suggestedthaterrorslessthan10%are neededbeforeapplyingestimatestodistance samplingmethodology.Errorsindistancees- timates produceerrorsofasimilarmagnitudein densityestimates,thatis,distanceconsistently overestimatedby10% underestimatesdensi- tiesby9%,anddistancesunderestimatedby

10%producedensitiesoverestimatedby11% (Bucklandetal.2001,2004).Errorsofvisual estimatesofdistancein our studywereon averagehigherthanrecommendedforareliable useofdistancesampling methods,andmay translateintounacceptablybiasedestimatesof populationsize.However,anumberofauthors havesuggestedthatsucherrorscanbecorrected throughcalibration,specificallybyusingregres- sionequationsobtainedfromthecomparisonof visualestimatesofdistancestoseabirdswitha ground-truthingmethod(Chen1998,Williams etal.2007).

Observeridentityaffectedthefitbetweenvi-

suallyestimateddistancesanddistancemeasured

byradar.Asalso reported inotherstudies(Baird

mateabundance(Bucklandetal.2001). Inour study,themostparsimoniousmodel(model2, Table1)maybeusedas acalibrationmodel tocorrecttheonlysourceofbiaswefound (Mateos2009).Differencesinaccuracybetween observersmaynotbeattributedtotheirdegreeof trainingorexperience,whichwassimilaramong thethreeobservers,buttovariationinvisual acuityorsharpness,perceptionofdistance,or otheraspectsof theirsensoryabilities.Lessex- periencedobserverswouldlikelyhaveproduced largererrors.Previousstudiesoftheeffectof trainingontheaccuracyofdistanceestimation onland(GibsonandBergman1954,Gibson etal.1955)andatsea(OienandSchweder1992, Schweder1996,1997)haveclearlyshownthat trainingimprovesestimationofdistance.Baird andBurkhart(2000)foundasimilareffectof experiencewhencensusingwhales.

Calibrationexperimentsinvolvingtheuseof

radarorotherground-truthingmethodscanbe

alsousefulfortrainingobserversandreducing

differencesindistanceestimates.Theaccuracy

ofvisualestimatesmightalsobeimprovedby

specifictrainingprograms, whereobserversare

toldthedistancesmeasuredwiththeradarafter

theyestimatethedistancetothetarget.Such

trainingwouldhelpobserverstobetterestimate

distances(Camp2007).

For investigatorswithout accessto radar-

estimateddistances,othermethodsareavailable

fortrainingobserverswhenmeasuringdistances

(e.g.,usingbuoysorfloatsplacedatknown

distancesfromtheobservationpoint)andwe

encouragetheimplementationoftrainingpro-

gramspriortofieldwork(eitherwithradar

orothermethods) forimprovingtheaccuracy

ofdistanceestimates.Furthermore,ourresults

suggestthat,whenpossible,selectingobservers

whoseinherentability toestimatedistanceis

bestamongallpotentialcandidatesisadvisable.

Ourresultsmay bealsousefulfordetermining

stripwidthwhendataaretakeninthefieldonly

bydistanceintervals(Bucklandetal.2001).For

example,inourcase,becausethehighesterror

valuewas0.2within3kmofthecoast,trying toestablishstripwidthsnarrowerthan600m wouldnotmakesense.

Distanceestimatesin our studywerenot

species-dependent,evenforspeciesthatvaried

considerablyinsize(e.g.,398gand549mm

wingspanforAtlanticPuffinsand3010gand

1850mmwingspanforNorthernGannets)and

flightbehavior(e.g.,fast-flappingauks,interme-

diateflap-glidingNorthernGannetsandglide-

flappingBalearic Shearwaters,and dynamic-

soaringCory’sShearwaters). Similarly,theac-

curacyofdistancesestimatedvisuallywasnot

affectedbyweatherconditions,flightdirection,

orwinddirection.Consequently,wesuggest

thatwhennoothermethodispractical,visual

estimatescouldbereliablyusedforawiderange

ofseabirdspeciesandfieldconditions,provided

thatobserversaresufficientlytrained.

Measuringdistancewith devicessuch as

telemeters andradarinthefieldisnotfreefrom

errorandthepossibleconsequencesofthiserror

ontheestimationof birdabundancearerarely

considered.Particularlywhenusingradar,the

signalontheradarscreenappearsas aspotthat

occupiesaspacegenerallylargerthanabirdor

flockinthefield.Nevertheless,themagnitudeof

thiserrorinourstudywaslow(generally2%),

anddidnotsubstantiallyaffectthevalidityof

distancemeasurements.

Tostatisticallyaddressmeasurementerrors,

manyobservationsare needed(Williamsetal.

2007).Ontheotherhand,collectingsufficient

ground-truthingdatatocalibratevisually es-

timateddistancescouldbecostly(Chenand

Cowling2001).However,usingradar,wefound

itpossibletoobtainaconsiderableamountof

independentdata,suitableforcalibration,with

relativelylittleeffort,andlargesamplespermit

morereliablecalibrationthansmallersamples.

Weconcludethat,whennoothermethods

areavailabletoaccuratelymeasuredistancesin

seabirdstudies,surveysbyexperiencedobservers

canprovidereasonably accuratedistanceesti-

matesunderawiderangeofenvironmental

conditionsandfordifferentspecies.Ourcon-

clusionscouldbeapplicabletoat-seasurveysfor

seabirdswhenobserversareonamovingship,as

longaslandmarkslocatedatknowndistances,

measuredwithradar,areavailable.Thismight

bethecaseforstudiesfromboatsclosetothe

coastorwhereshiptrafficisrelativelycommon.

Finally,theabilityofradartodetectlow-flying

seabirdsis limitedwhenwavescausesea-clutter, butradarcanbeusedtocalibratevisualob- servationsthat,inturn,canbeobtainedunder conditionswhenradarcannotbeoperated.

ACKNOWLEDGMENTS

Thisstudywasconductedwithinacollaborationagree- mentbetweentheMigresFoundationandtheUniversity ofCa´diz.TheradarfacilitywassuppliedbyCeowind CapitalEnergyOffShoreCompany.M.Mateoswassup- portedbytheJuntadeAndaluc´ıawithaFPUfellowship. WethanktheMigresFoundationtechnicalstafffortheir helpwiththefieldworkandB.Brudererforextensive commentsonearlydrafts.Themanuscriptwasgreatly improvedthankstothesuggestionsofA.E. Burger,one anonymousreviewer,andtheeditor.

LITERATURECITED

BAIRD,R.W.,AND S.M.BURKHART.2000.Biasand variabilityindistanceestimationonthewater:impli- cationsforthemanagementofwhalewatching.Paper SC/52/WW1,IWCScientificCommittee,Adelaide, Australia.

BIBBY,C.J.,N.D.BURGESS,ANDD.A.HILL.1992.Bird censustechniques.AcademicPress,London,UK.

BUCKLAND,S.T., D. R.ANDERSON,K.P.BURNHAM, J.L.LAAKE,D. L.BORCHERS,AND L.THOMAS.2001. Introductiontodistancesampling,estimatingabun- danceofbiologicalpopulations.OxfordUniversity Press,Oxford,UK.

———.2004.Advanceddistancesampling.OxfordUni- versityPress,Oxford,UK.

BURNHAM, K. P., AND D. R. ANDERSON. 2002.

Modelselectionandmultimodelinference:apracti-

cal information-theoreticapproach.Springer-Verlag,

NewYork,NY.

CAMP,R.J.2007.MeasurementerrorsinHawaiianforest

birdsurveysandtheireffectondensityestimation.

Hawai‘iCooperativeStudiesUnitTechnicalReport

HCSU-005,UniversityofHawai‘iatHilo,Hilo,HI.

CAMPHUYSEN,C.J.,A.D.FOX,M.F.LEOPOLD,AND

I.K.PETERSEN[online].2004.Towardsstandardised

seabirdsat seacensustechniques in connection

with environmental impact assessments for

offshorewindfarms intheUK.RoyalNetherlands

InstituteforSeaResearchandtheDanishNational

Environmental ResearchInstitute, Crown Estate

Commissioners,London,UK.Availableat:

phase1_final_04_05_06.pdf(10March2010).

CHEN, S.X.1998.Measurementerrorsinlinetransect

surveys.Biometrics54:899–908.

———,ANDA.COWLING.2001.Measurementerrorsin

linetransectsurveyswhere detectabilityvarieswith

distanceandsize.Biometrics57:732–742.

ENVIRONMENTAL SYSTEMSRESEARCHINSTITUTE. 1999.

ArcViewGIS3.2.EnvironmentalSystemsResearch

Institute,Inc.,Redlands,CA.

GIBSON,E.J.,AND R.BERGMAN.1954.Theeffectof

training on absoluteestimation of distanceover

theground.JournalofExperimentalPsychology48:

473–482.

———, AND J. PURDY. 1955. The effectof prior

trainingwithascaleofdistanceonabsoluteand

relativejudgmentsofdistanceoverground.Journal

ofExperimentalPsychology50:97–105.

HASTIE,T., AND R.J.TIBSHIRANI.1990. Generalized

additivemodels.ChapmanandHall,London,UK.

HEINEMANN, D.1981.Arangefinderforpelagicbird

censusing.JournalofWildlifeManagement45:489–

493.

KOMDEUR,J.,J.BERTELSEN,ANDG.CRACKNELL.1992.

Manualfor aeroplaneandship surveysofwaterfowl

andseabirds.IWRBSpecialPublicationNo.19, Na-

tionalEnvironmentalResearchInstitute,Roskilde,

Denmark.

MARQUES,T. A.2004. Predictingand correctingbias

causedbymeasurementerrorinlinetransectsam-

plingusingmultiplicativeerrormodels.Biometrics

60:757–763.

———. 2007. Incorporating measurementerror and

density gradients in Distance Sampling surveys.

Ph.D. dissertation,UniversityofStAndrews,St.

Andrews,Scotland,UK.

———, M. ANDERSEN,S.CHRISTENSEN-DALSGAARD,

S.BELIKOV,A.BOLTUNOV,O.WIIG,S.T.BUCK-

LAND,ANDJ.AARS.2006.TheuseofGPStorecord

distancesinahelicopterline-transectsurvey.Wildlife

SocietyBulletin34:759–763.

MATEOS,M. 2009. Radartechnologyappliedtothestudy

ofseabirdmigrationacrosstheStraitofGibraltar.

Ph.D.dissertation,UniversityofCadiz,Cadiz,Spain.

OIEN,N.,ANDT.SCHWEDER.1992.Estimatesofbiasand

variabilityinvisualdistancemeasurementsmadeby

observersduringshipboardsurveysofnortheastern

Atlanticminkewhales.Reports oftheInternational

WhalingCommission42:407–412.

PETERSEN,I.K.,T.K.CHRISTENSEN, J.KAHLERT,M.

DESHOLM, AND A.D.FOX.2006.Finalresultsof

birdsstudiesattheoffshore windfarmsatNysted

andHornsReef,Denmark.NationalEnvironmental

ResearchInstitute, Ministryofthe Environment,

Roskilde,Denmark.

PROGRAMAMIGRES.2009.Seguimientodelamigracio´n delasavesenelEstrechodeGibraltar:resultadosdel ProgramaMigres2008.Migres,RevistadeEcolog´ıa

1:83–101.

RDEVELOPMENT CORETEAM.2008.R:alanguageand

environmentforstatisticalcomputing.RFoundation

forStatisticalComputing,Vienna,Austria.

RANSOM,D. J.,AND W.E.PINCHAK.2003.Assessing

accuracy ofalaserrangefinderinestimatinggrassland

bird density.WildlifeSocietyBulletin 31: 460–

463.

SCHWEDER,T.1996.Anoteonabuoy-sightingexper-

iment intheNorth Seain1990. Reportsofthe

InternationalWhalingCommission46:383–385.

———.1997.MeasurementerrormodelsfortheNor-

wegianminke whalesurveyin 1995. Reportsof

the International WhalingCommission47:485–

488.

SPEAR,L.B.,ANDD.G.AINLEY.1997.Flightspeedof

seabirdsinrelationtowindspeedanddirection.Ibis

139:234–251.

WILLIAMS,R., R. LEAPER,A.N. ZERBINI, AND P. S.

HAMMOND. 2007.Methodsforinvestigatingmea-

surement error in cetaceanline-transectsurveys.

JournaloftheMarineBiologicalAssociationofthe

UnitedKingdom87:313–320.

YEN,P.P.W.,W.J.SYDEMAN,ANDK.D.HYRENBACH.

2004.Marinebirdcetaceanassociationswithbathy-

metrichabitatsandshallow-watertopographies:im-

plicationsfortropictransferandconservation.Jour-

nalofMarineSystems50:79–99.