ORIGINAL ARTICLE
Presence,richnessandextinctionofbirds ofpreyintheMediterraneanand Macaronesian islands
J.A.Dona´zar1,L.Gangoso1*,M.G.Forero1 andJ.Juste2
Departmentsof1AppliedBiologyand
2Evolutionary Biology,Estacio´nBiolo´gicade
Don˜ana,C.S.I.C.,Sevilla,Spain
*Correspondence:LauraGangoso, Department ofAppliedBiology,Estacio´nBiolo´gicade
Don˜ana,C.S.I.C.,AvdaMa Luisas/n,41013
Sevilla,Spain.
E-mail:
ABSTRACT
AimWeanalysemodernpatternsofrichness,presenceandextinctionofbirdsof prey(AccipitriformsandFalconiforms)intheMediterraneanandMacaronesian islands,usinganintegratedapproachinvolvingbothbiogeographicalandhuman- inducedfactors.
Location Forty-three islands grouped into nine Mediterranean and
Macaronesianarchipelagos.
Methods Information about25speciesbreedingduringthepastcenturyand theirfate(permanenceorextinction)was compiledfromtheliteratureand regionalreports.Jaccard’ssimilarityindexandclusteranalyseswereappliedto defineislandassemblages.Inordertodetectthefactorsdrivingrichness,presence andextinction,generalizedlinearmodels(GLM)wereappliedto32explanatory variables, evaluatinglocation,physiography,isolationofisland,taxonomic affinitiesandlife-historypatternsoftheraptorspecies.
ResultsIslandsbelongingtothesamearchipelago clusteredwhenraptor assemblageswerecompared,revealingamarkedbiogeographicalsignal.Species richnesswasinfluenced by island area and accessibilityfrom the continent (explaineddevianceof51%intheGLM).Modelsoftheprobabilityofpresence (explaineddevianceof32%)revealedpositiveinfluencesofmigratorypatterns (maximumforpartialmigrants),sizeofdistributionareasandproximitytomain migrationroutes.Themodelforprobabilityofextinctionexplainedonly8%of thedeviance.Itrevealedthatpopulationslivingonislandswithahighdensityof human population weremorepronetodisappear.Also, raptorsdependingon humanresourceshadmoreriskofextinction.
MainconclusionsBasicpredictionsof island biogeographycanexplaincurrent patternsofraptorrichness inthestudyareadespitemillenniaofintense humanizationprocesses.Colonizationsuccessappearstodependonlife-history traitslinkedtomigratoryanddispersalstrategies,whereasbody-sizeconstraints arenot influential.Additionally,our resultsrevealtheimportance ofspecies- basedanalysesinstudiesofislandbiogeography.
Keywords
Extinction,generalized linearmodelling,islandbiogeography,Macaronesia, Mediterranean,raptors,species richness.
INTRODUCTION
Islandbiogeography(MacArthurWilson,1963,1967)has been one ofthe most widelydiscussedaspectsofmodern ecologicaltheory. Classically, its core centres on species richness,whichisdependentonpatchsize(islandarea)and
degreeofisolation.Thusdistancetothecontinentdetermines probabilityofarrival,whereas islandareainfluencesprob- ability of extinction. Subsequent predictions derived from thesecentralargumentshavebeenbroadlydocumented and provenforverydifferenttaxonomic groups(seeRicklefs& Lovette,1999;Johnsonetal.,2000forreviews).Ontheother
hand,manyaspectsofthetheoryofislandbiogeographyand, inparticular,theconceptofequilibrium,have beendeeply criticised(seeBrownLomolino,1998; Walter,2004for reviews).Ithasalsobeenarguedthat approachesbasedon comparisonsinvolvingdifferenttaxonomicgroupshaveover- lookedfine-tuningmechanismsdeterminingcolonizationand extinctionevents. Inparticular,theprobabilityofisland colonizationmaybedependent on aspecies’ life-history characteristics, particularly dispersal capacity (Ferna´ndez- Palacios &Andersson,1993).Ontheotherhand,the permanenceorextinctionofalocal populationcanbe influencedbyspecies-specificfactorssuchasbodysizeand degreeofecologicalspecialization:largeorganismsatthetopof foodchainsaremorepronetosufferdeclineswithenviron- mentalchanges(PeltonenHanski,1991;AlcoverMcMinn,
1994;RosenzweigClark,1994; Millien-ParraJaeger, 1999). Thusanalysesfocusingonphylogenetically closely related groupsshouldbepreferredtoavoidthelargely irrelevant varianceinherentinjointtreatmentofdifferentgroupsoftaxa (CarrascalPalomino,2002).In addition,whenfactorsrelated tothelifehistoryofeachspeciesareconsidered,itispossibleto determinespecies-basedcolonizationandextinctionpatterns which,intheend,maymodelthecompositionandassemblyof insularcommunities(Lomolino,2000).
Anthropogenicdisturbancemayhave adeepimpacton speciesrichnessthrough the modification ofsomeintrinsic islandcharacteristics suchashabitatabundanceanddiversity (Russelletal.,2004).Itseffects,however,mayvarydepend- ingon scale.It iswellknown that humanization ofisland systemshaswidelyalteredtheinitialstructure and richness ofaviancommunitiesworldwide.Directpersecution,habitat destruction and facilitation of biologicalinvasionsare the main consequencesfollowinghuman settlementon islands. Theybringabout processesofextinctionoflocalendemism andtheexpansionofalienspecies (Chownetal.,1998).For instance,inthePacifictropicsasmanyas2000birdspecies were lostafterhumancolonization(Steadman,1995), and temperate insularsystemshavesufferedaparallelimpover- ishment (Milberg Tyrberg, 1993; Alcoveretal., 1998; Palmeretal.,1999).Birds ofpreyliveonmostislandsofthe world,withthesole exceptionoftheAntarcticcontinent (WhiteKiff,2000).Frequently,diversityinbirdsofpreyis high in comparison with other taxonomic groups because birdsofpreytypicallyhavegoodflightabilities,permitting manyspeciesto reachoceanicislandsfarfrom continental regions. Unlike other taxonomic groups which may have beenconstrained bystrictecological requirements and/or specific dispersal constraints, birds of prey, at least in temperateregions,toleratemoderatehabitattransformations drivenbyhuman activities(Rodrı´guez-Estrella etal.,1998), although speciesoflargebodysizemaybemore prone to suffernegativeeffects becauseoftheirconservativelife- history traits (Alcover McMinn, 1994; White Kiff,
2000).Thisscenariomakesbirdsofpreyanappropriate tool forthestudyofbiogeographicalpatternsandtheinteraction betweenintrinsicfactorslinkedtoislandcharacteristicsand
those relatedto transformations derivedfrom human activities.
Inthispaperweexaminethefactorsconditioning species composition andrichnessofbirdsofpreyintheMediterra- neanislandsandMacaronesia.Theseregionsareconsidered
‘biodiversityhot-spots’duetotheirhighratesofendemism (CovasBlondel,1998;Juan etal.,2000).Thisvarietyis particularlyoutstandinginbirdsofprey:upto26speciescan befoundintheregion,withsomeislandshavinghadmore than15breedingspecies. Islandsoftheregionshowvery differentdegreesof humantransformation,whichallowsusto dealwiththeabove-mentionedquestions.Two groupsof hypothesesare tested in this study. The first group is in relationtothebasicprinciplesofislandbiogeographytheory. Weexpectthatislandsize,habitatcomplexityanddegreeof isolationfrom the mainland willinfluencespeciesrichness. Accordingly,richnessisexpectedtoincreasewithislandarea andproximitytothemainland.Withinthisgeneralscenario wealsotestwhether,aspredicted,therearemoresimilaritiesin thecompositionof communitiesofbirdsofprey between islandsbelongingtothesamearchipelago thanbetween archipelagos.Second,wetesthowaspecies’ life-history characteristicsaffecttheprobabilityofpresenceandextinction aftercontrollingforislandsizeandisolation.Wepredictthat: (1)smallbody-sizedraptorsshouldbefoundmorefrequently ininsularsystemsduetotheirrelatively highpopulation turnover compared withlargebody-sizedraptors (Newton,
1979);(2) becauseof the ‘rescue’effectthere isa greater probability offindingspecieson islandsifthere isalarge populationonaneighbouringarea of mainland(Brown & Kodrick-Brown,1977;Sara` Morand,2002); (3)probability ofextinctionshouldbehigherforlarge-bodiedspecieswith lowpopulation turnover; and (4) speciesdirectlybenefiting fromhuman-induced transformations willshowlowerprob- abilitiesofextinction.
MATERIALSANDMETHODSStudyarea
Thestudyencompasses43islandsofatleast100km2 inthe Mediterranean Sea and Macaronesia (Appendix1; Fig.1). Theselection comprisesallmajorislandgroupsand archipelagos with a broad range of geologicalage, from
20Myr (late Miocene) to 0.8Myr (Pleistocene). Both progressive isolation and divergent evolution have taken place, mainlyaftertheglaciationsoftheNorthernHemi- sphereand desertificationoftheAfricanContinent. Conse- quentlytheseislandshavedifferentclimatesand vegetation. Macaronesianarchipelagosexhibitthelargestenvironmental gradientandvegetationranges.Laurel forestformations dominateintheAzores,Madeiraandonthenorthern slopes ofthewesternCanaryIslands,whereassemi-desert,succulent scruboccupiesthesouthernpartsofthewesternislandsand allthe easternislandsofthe latterarchipelago(Ferna´ndez- Palacios Andersson, 2000). The Mediterranean islands
Azores
Corsica-Sardinia
30°N
Madeira
Balearic Islands
Sicily-Malta
Aegeum Cyprus
Canary Islands
Figure1Geographicalsetting ofstudyarea, andgroupingof43islandsstudiedintonine archipelagos.Macaronesia:Azores,Madeira, CanaryIslandsandCapeVerde;Mediterra- nean:BalearicIslands,Corsica–Sardinia, Sicily–Malta,AegeumandCyprus.
Cape Verde
havea relativelyhomogeneous climatewith scarcerainfall and more xeric vegetation; the observed variations are mainly determined by altitudinal gradients (Covas & Blondel,1998).Humanization hasbeen veryimportant all over, but whereas the islands of the Mediterranean were settledbyhumans between12,000and 9000yrbp(but see Ramisetal.,2002),Macaronesianarchipelagoswerecolon- izedonlyfrom2500yrbp(Canaries)to500yrbp(Azores, MadeiraandCapeVerde).
Birdsofprey
Our main objectivewas to determine factors influencing raptorrichness,presenceandextinctionontheislandsofthe MediterraneanBasin andMacaronesia.Toaccomplishthis goal,wecompileddatageneratedduringthetwentiethcentury. Literatureaboutornithologicalsurveysinthestudyareaprior to thisperiod isextremelyscarce,soitwasnot feasibleto extendouranalysestofactorsoperatingbeforethetwentieth century.Thepossibility ofusingsub-fossildatatolookat factorsoperatingonalargertime-scalewasdiscardedbecause reliablepalaeontologicalinformation isscarce andvaries amongislands.Inaddition,boneremainsoflarge species appear more frequently(AlcoverMcMinn,1994;Rando,
2002),thereforesub-fossil datawereconsideredtobe misleadingandbiased.
Information wasgatheredfromgeneralsources(Cramp& Simmons,1980; HagemeigerBlair, 1997),monographic revisions(MuntanerMayol,1996)andunpublishedregional reports.Weconsideredonlythoseislandsforwhichthereisa goodamount ofornithologicalliterature.All thepotential breedingspeciesareincludedintheanalyses.Weconsidered potential breederstobethosespeciesthat havebeenfound duringthepastcenturybreedinginanareaofaround1500km from the island.The onlyexceptionwasEleonora’sfalcon (Falcoeleonorae),becauseof itsuniqueecologyandthevirtual
absence of continental populations (Cramp Simmons,
1980).
Geographical pattern
Theconvenienceofincluding geographicalvariablesin the studywascheckedusingfaunisticsimilaritytests.Islands were grouped accordingto(1) ‘true’Mediterranean and Atlantic archipelagos,or(2)‘groups’ofislandsthataregeographically close(<100km).
Associationpatternsamongislandsandarchipelagoswere inspectedfirst,usingJaccard’ssimilarityindex (Jaccard,1901). Thisindexconsiderstheproportion betweenlocalitieswhere twospeciesarefoundandthetotalnumberoflocalitieswhereat leastoneofthemisfound.Thisindexiscommonlyusedin ecologicalstudiesandwasselectedforitseasyinterpretation, independenceofnegativematches,symmetryandhomogeneity properties(JansonVegelius,1981).Dice’sindex(Dice,1945), withsimilarmathematicalproperties(butinwhichmatchesare double-weighted)wasalsocalculatedsincetheuseofalternative measurementsofassociationisrecommended(Huba´lek,1982). Similaritiesweretransformedtodistancesandtherelationships among islandswerevisualizedthrough aneighbour-joining reconstruction(Saitou& Nei,1987)inwhichthebranch lengthsaremeaningful.Robustnessofthenodesofthetopology wasassessedthrough bootstrapping (Felsenstein,1985)after
1000resamplings.Theanalysiswasperformedinmatlaband thescriptsareavailablefromtheauthorsonrequest.
Responsevariables
Threeresponsevariablesweredefined,asfollows.
1.Speciesrichness:thenumber ofspeciesthatbredoneach islandduringthepastcentury.Aspecieswascomputedasifit wasrecordedbreedingonanislandforaperiodofatleast
10continuousyears.
2.Probabilityofpresence:apriori,weconsideredpotential colonizerstobethosespecieshavingcontinentalpopulations withinaradiusof1500kmofeachisland.Thisisthelargest distancefoundbetweenneighbouringpopulationsof thesame speciesin the studyarea(betweenpopulations ofred kite, MilvusmilvusinthearchipelagosofCanaryIslandsandCape Verde;CrampSimmons,1980).Oneachislandapotential colonizingspecies wascharacterizedas1ifithadbeen recordedbreedingintheislandduringthepastcentury,or0if not.Forthisanalysis,Africanspecieslinkedtothetropicalrain forestwereexcluded.
3.Probabilityofextinction:foreachspeciesandforagiven island,anextinctioneventwascomputedas1whenthespecies wasextinctinthatisland.Value0wasassignedtospeciesthat havemaintainedpopulationstothepresent.
Explanatory variables
Islandcharacteristics
Weusedtotalarea,geographical location(latitude)and maximumaltitudeasgeographicalvariables thatcould influencerichness,presenceandextinctionevents(Table1). Thelattervariable isagoodpredictorofhabitatdiversity (CarrascalPalomino,2002).Weevaluatedtheaccessibility of theislandtothebirdsofprey throughaset ofvariables measuringthedistanceonebirdhastotravel toreachthe island.Wefirstconsideredthepossibility ofreachingtheisland fromthemainlandbytheshortest,mostdirectflight or, alternatively,bythesameroutebutfollowingnarrowerstraits. Additionally,wemeasuredthedistancefromtheislandtothe
Table1Explanatoryvariablesusedtocharacterize MediterraneanandMacaronesianIslandsandraptorspecies(seeMaterialsand methodsforinformationsources)
VariableUnits
Island
1Islandcode(Appendix1)
2Archipelagocode(Fig. 1;Appendix1)
3Biogeographicalregion:(1)Mediterranean;(2)Macaronesia(Fig.1)
4LatitudeDegrees
5Numberofislands(>100 km2)inthearchipelago
6Islandareakm2
7HumandensityInhabitantskm)2
8Maximumaltitudem
9Minimumdistancetocontinentkm
10Minimumdistancetonearestcoast(islandorcontinent)km
11Minimumsumofwidthofmarinechannelsnecessarytocrosskm toreachislandfollowingroutewithlessflightabovesea
12Maximumwidthofmarinechannelstocrosstoreachisland,followingrouteasvariable(10)km
13Numberofmarinechannelstocrosstoreachisland,followingrouteasvariable(11)
14Landoccupiedby‘open’habitats(cultures,grazing,shrubs)%
15Landoccupiedbyurbanareas%
16Geographicalpositionwithrespecttomigrationroutes:(1)withinaprincipalroute; (2)withinasecondaryroute;(3)farfrommainandsecondaryroutes
17Presenceoftop-predator birdsofprey(goldeneagle,Bonelli’seagle,goshawk):(1)present;(2)absent
Species
18Speciescode(Appendix1)
19Taxonomicposition:(1)Accipitridae;(2)Falconidae
20Distributionarea(numberof5·5kmsquares)occupiedbyspeciesina500 kmradius
21Generaldistribution:(1)Palaearctic;(2)African;(3)both
22Migratorypattern:(1)resident;(2)long-distancemigrant;(3)partialmigrant
23Adultmeanbodyweightg
24Nestsitehabitat:(1)tree;(2)cliff;(3)ground
25Reproductivestrategy:(1)solitary;(2)colonial
26Medianclutchsize
27LengthofincubationperiodDays
28LengthofnestlingperiodDays
29SexualmaturityYears
30Foraging habitat:(1)open;(2)semi-open;(3)woods
31Affinitytohumanactivities:(1)breedinginbuildings; (2)feedingonlivestockcarcasses; (3)withoutdirectaffinity
32PopulationtrendinWesternPalaearcticduringtwentiethcentury:(1)withoutreductioninbreedingarea; (2)reductionofbreedingarea50%;(3)reductionofbreedingarea50%
threemainmigratoryroutescrossingtheWesternPalaearctic: Gibraltar–WesternAfricancoast,Messina,andtheBosphorus (Berthold,2001). Wealsoevaluatedthepresenceoftop- predatorraptorsasanindicationofwell-structuredfoodwebs (Newton,1979).Finally thosevariablesrelatedtohuman pressureonislandecosystems,suchas populationdensityand land use, wereevaluatedfrom sources that had compiled information from the pastdecade(Table1),asthere isan almosttotallackofinformationforprevioustimes.Although potentiallybiased(someislandsaresuffering accelerated transformations),itseemed bettertohaveanestimatorof human-induced transformationsthantoignoretheirpossible effects.Statisticsonhuman populations wereobtainedfrom stategeneral sources.Landuseswereobtainedfromthe CORINE Land Cover database (CEC 1999) by using theXToolsextension(available at us/DIVISIONS/management/state_forests/GIShome.asp) of ArcView GIS3.2.
Speciescharacteristics
Thetaxonomiccategoriesfamily(Accipitridaevs.Falconidae) andspecieswerecodified.Bodysize,breedingperformance, andforagingandmigratoryhabitsdescribedthelifehistory ofeachspecies.Wealsoquantifiedthe abundance ofeach speciesinsurrounding areasofdifferentradiiasanestimate ofthe probabilityof‘rescue’events.Finally,weconsidered whetherthespeciesusesbuildingsfornestingand/or usually consumesby-productsofhuman activities (e.g.livestock carcasses)toevaluatetheeffectsofassociationwithhuman activities.
Statisticalanalyses
Initially,weanalysedapossiblephylogeneticeffect and investigatedwhether familyor speciescategoriesshould be includedinfurtheranalyses.WeperformedaGLMusingsas macroglimmix(Littelletal.,1996)toidentify thetaxonomic level(familyorspecies)atwhichmostofthevariationoccurs intheresponsevariablesprobability ofpresenceandextinction (Figuerola,2000).Theanalysisincluded familyand species nestedwithinfamilyasrandomeffects.Bothmodelsshowed thatphylogeneticrelationshipexplainedonly 28%and24%of theoriginaldeviance intheprobabilitiesofpresenceand extinctionrespectively.Inbothanalysesmostofthevariation wasexplainedbyspecies(accountingfor28%and26%ofthe originaldevianceintheanalysisofpresenceandextinction, respectively),whilefamilyaccounted foronly4%and 11% respectively.Consequentlyonlyspeciesidentitywasincluded infurtheranalyses asarandom variabletocontrolfor phylogeneticeffects.
Themainanalyses wereperformedapplyinggeneralized linearmixedmodel(GLMM),whichalloweddissectionofthe relativecontribution ofislandintrinsiccharacteristics and species-specificlife-history traits from a joint analysisof geographical,ecologicalandhuman-relatedfactors.Thiskind
ofapproachisuncommoninbiogeographicalresearch(Covas
Blondel,1998; BurbidgeManly,2002).Wefittedthe mixedmodelsbyPROCMIXEDinsas whentheresponse variablewasnormallydistributed.Appropriatelinkfunctions and error structures fornormal and binomiallydistributed datawereimplemented.Whennecessary,species,islandand archipelagoweretreatedasrandom terms.Eachexplanatory variable(Table1) and their interactions werefitted to the observeddatafollowingaforwardstepwiseprocedure,which resultsinthemostadequatemodelforexplainingvariationin theresponsevariablewhereonlysignificanteffectsare retained (Foreroetal.,2002).Specialattention waspaidtopotential interactions betweenbiogeographicalregion (Mediterranean andMacaronesia)andtheexplanatoryvariables,todetermine whethertherewasadifferentialresponsebetweenthesetwo zones.Finally,higherpolynomialmodelswerealsofittedto accountforpotentialnonlinearrelationships.
The specificcharacteristicsofeachfitted model wereas follows.
1.Speciesrichness:aGLMMwithnormalerrorandidentity linkfunctionwasfittedforthe17variablesrelatedtoisland characteristics(Table1).Archipelagowasfittedasarandom term.
2.Presence:aGLMMwithbinomialerror and logisticlink functionwasfittedforthepresence(1)orabsence(0)ofeach species in each island (Appendix1). All 32 explanatory variableswereincluded (Table1). Species,archipelagoand island nestedinarchipelagowereconsideredasrandomterms.
3.Extinction:aGLMMwithbinomialerrorandlogisticlink functionwasfittedfortheextinction(1)orpermanence(0) of the breeding populations (Appendix1). Variables and statistical treatmentswere similartothoseoftheformer analysis.
RESULTSRaptorrichness
Atotalof25speciesofbirdsofpreywererecordedbreeding in the 43 different islands studied (Appendix1). If we consideratotalof49speciesbreedinginaradiusof1500km ofthe study island,only51%ofthesecontinental species havemanagedtoestablish breedingpopulationsinthe MediterraneanandMacaronesianislands duringthepast century.Twentyoneofthespecies haveadistribution restrictedtoEuropeandtheAfricanNorth Sahara,andfour are distributed through Europe and Africa(including the sub-Saharanregion).Noneis distributedexclusivelyinsub- SaharanAfrica:24speciesbreedinginthatregionhavenever beenfoundnestinginislandsofthestudyarea.Consequently onlyspecies withpartialortotalcircummediterranean distributionwereconsideredinfurtheranalyses.Thenumber of breeding species by island ranged from 1 (only the Europeanbuzzard,Buteobuteoispresentintheislands ofthe Azores)to18inSicily,withamedianvalueoffourspecies perisland(Appendix1).
Geographical pattern
TheislandassemblagesbasedonJaccard’sindexand Dice’s indexwerealmostidentical.ResultsderivedfromJacccard’s index (havingahigherbootstrapsupport)areshowninFig.2. IntheresultingtopologyallthreeAtlanticarchipelagos(the Azores,Canaryand CapeVerdeIslands)wererecoveredin clearlydistinguishedclusters.ButwhereastheAzores Islands hadnostructure,theCanarianandCapeVerdearchipelagos showedstronginternalstructure.Ontheotherhand,onlytwo main groups could be defined among the Mediterranean islands (easternvs.western).Interestingly,theBalearicIslands clusterclosertotheAtlanticarchipelagos(CanariesandCape Verde)thantootherMediterraneanislands.Bootstrapvalues supportonlythosenodesinthetopologydefiningeachofthe Atlanticarchipelagos(althoughtheCanaryislandsappearin two groups), the two largestislandsof the Balearicsand, finally,agroupmadeupofSicilytogetherwithCorsicaand Sardinia(Fig.2).
Speciesrichness
Aftercontrolling by archipelago(Z¼1.20,P¼0.11),the GLMM explainedupto51.5%oftheinitialdevianceand includedtwomainexplanatoryvariables:islandareaandwidth ofthelargestmarinechannelnecessarytocrosstoreachthe island.Species richnessincreasedwithislandarea,and decreasedwithdistancetothemainland(Table2).
Probabilityofpresence
Thetworandom effectsspecies(Z¼2.83,P¼0.002)and island nested in archipelago (Z¼1.95, P¼0.002) had significant effects, but not archipelago alone (Z¼1.28, P¼0.1). Aftercontrolling by these variables,the best-fit modelexplainedupto32%oftheinitialdeviance(Table3). Thisindicatesthattheprobabilityofpresence onanisland increaseswithislandareaandwithproximitytoamigration route.Theeffectofthemigratoryhabitsof aspecieswasmore significant:partialmigrantsweremorepronethanmigrantor sedentary species to maintain populations on islands. In
S.Vicente(CV)
66 S.Nicolau(CV) BoaVista(CV)
Table2Resultsofgeneralizedlinearmixedmodelonspecies richnessofbirdsofpreyintheislandsoftheMediterraneanand Macaronesia(n¼43)
41
56
Fogo(CV)
Sal(CV) Maio(CV)
Parameter estimate
Standard
errorFP
S.Antao(CV) Santiago(CV)
Hierro(CA)
67 Gomera(CA) Tenerife(CA)
G.Canaria(CA)
56Fuerteventura(CA) LaPalma(CA)
Lanzarote(CA)
64Mallorca(BA)
Menorca(BA)
Malta
Corsica
62Sardinia
Sicily
Ibiza(BA)
Intercept7.010.98
Islandarea0.00050.000141.730.0001
Widthofseachannel*)0.00620.00218.430.007
Explaineddeviance(%)51.5
*Widthoflargestseachannelnecessarytocrosstoreachisland.
Table3Resultsofgeneralizedlinearmixedmodelfor probabilityofpresenceonMediterraneanandMacaronesian
Crete
Rhodas
Cyprus
islandsbybirdsofpreywithpopulationsina1500kmradius
Thasos
Nasos
60Lesbos
Eubea
Parameter estimate
Standard
errorFP
Samos
Cefalonia
Intercept)2.971.00
Migratorystatus22.71 0.0001
Quios
Samotracia
Andros
Sedentary)3.570.55
Corfu
Lemnos
Migrant)2.840.87
Partialmigrant0.00
Madeira
Terceira(AZ) S.Jorge(AZ)
Distributionarea
(500kmradius)
1.500.1212.11 0.0001
80 Fallal(AZ)
Pico(AZ) Flores(AZ) S.Miguel(AZ)
Figure2Similaritiesinbirdofpreyguildcompositionamong43 islandsstudiedfromtheMediterraneanandMacaronesia. Topologyobtainedafteraneighbour-joiningdistancerecon- structionbasedonJaccard’ssimilarityindex.Bootstrapvalues after1000replicates.Archipelagos:CV,CapeVerde;CA,Canary Islands;BA,BalearicIslands;AZ,Azores.
Islandarea1.5·10)4 0.003.830.0001
Migratoryways6.090.0023
Principalway1.470.96
Secondaryway)0.570.98
Outofway0.00
Presenceoflargeraptors5.350.0209
Absent)1.080.47
Present0.00
Explaineddeviance(%)32.0
addition,specieswithbroaddistributionareasintheWestern Palaearctic (measured in 500km radius) showed higher probabilitiesofpresence.Finally,theprobabilityofpresence waspositivelyrelatedtotheexistenceoftop-predator raptors.
Probabilityofextinction
In 53%ofthe islandsstudied, at leastonespeciesbecame extinctorsuffered50%declineinpopulation size(and/or breeding area) during the past century (Appendix1). The medianvalue wastwoextinctionsperisland.CyprusandSan Vicente (Cape Verde) were the islands with the greatest numberofextinctspecies(four).Meanratesofextinctionfor theCanarianandCapeVerdearchipelagoswere29%and43% respectively.Mediterranean islandsshowedhigherrates:the meanvaluewas 23%,butreaching50%inCyprusand100%in Malta(Appendix1).
Aftercontrollingbyspecies(Z¼1.74,P¼0.04),archipel- ago (Z¼0.51,P¼0.3) and island nested in archipelago (Z¼1.38,P¼0.08),the model explainedonly8%ofthe originaldeviance(Table4).TheGLMMshowedthatraptor populationshadahigherprobabilityofextinctionwhentheir distributionareainneighbouringwesternPalaearctic areas (radius 500km) wassmall.In addition, islandswithdense human populations were more prone to suffer extinction events.Finally, speciesdependingonhuman resourcesfor feeding(livestockcarcasses)and breeding(buildings)hada higherriskofextinction.
DISCUSSION
Thesimilarityindexesshowedafairlyimportantbiogeograph- icalsignalinthe compositionandstructureofguildsofbirds of preyontheMediterraneanandMacaronesian islands(Fig.2). Thisresultjustifiestheconsiderationofgeographicalvariables infurtheranalyses.Nevertheless,thelengthofbranchesand bootstrapvaluessupportsuchastructureonlyfortheAtlantic archipelagosandtheBalearics,andagroupmadeupofSicily, CorsicaandSardiniaintheWesternMediterraneanSea.The associationamong the other Mediterranean islands,partic-
Table4Resultsofgeneralizedlinearmixedmodelfor probabilityofextinctionofbirdsofpreyinislandsofthe MediterraneanandMacaronesiaduringthetwentiethcentury
ularlyintheEasternMediterranean,wasmuchweaker (Fig.2). This patternindicates therelativeimportanceofhistoricaland species-specificcharacteristics (such as dispersal capacity) over linear measures of distance (Ferna´ndez-Palacios & Andersson,1993)inunderstandingislandsimilarities,partic- ularlyintheMediterraneanSea. TheAtlanticarchipelagos showhigherstabilityintheirspecies composition(lower turnover) whichmaybeclearlyrelatedtothemuch higher speciationratefoundonthem.As manyas11endemic subspeciesofbirdsofpreyhavebeenrecognizedinMacaro- nesia,contrastingwithonlythree(Accipitergentilisarragoni, AccipiternisuswolterstorffiandButeob.arragoni)foundinthe Mediterranean(allonCorsica).
Relatively highspeciesturnover in the Mediterranean islands,promotedbyshorterdistancestothemainland,would havehamperedanystructuringoftheguild,asshownbythe topology.Thusarelativelyhighturnoverrateofcolonization/ extinctionepisodes wouldhavebeenanimportant factor determiningfaunalcompositionontheMediterraneanislands. Therearemanyreportsofoccasionalbreeders(e.g.Bannerman
Bannerman,1983)ontheseislands,eventsthatareprobably underestimatedsincetheyaredifficulttodetect.Theshallow differentiationbetween EasternandWesternMediterranean islands(althoughnotsupportedbybootstrap)couldbedueto thepresenceintheformergroupofspeciesshowingonlyan easterndistributionthatdonotreachthewesternsideofthe Mediterranean.
Manystudieshaveshownaclose relationshipbetween speciesrichnessandislandarea (Temple,1981;Thiollay,1997; BrownLomolino,1998;Whittaker,1998;Millien-Parra& Jaeger,1999).Thisrelationship maybedue to thepositive relationshipsofislandarea,eitherwithtotalhabitatavailability orwithhabitatstructureanddiversity (Williams,1964; Carrascal& Palomino,2002).Distinguishingtheserelated factorsisdifficult, butpossible throughourmodelling approach.Intermediatestepsintheconstructionofthemodel showedasignificantandpositiveeffectofmaximumaltitude (an indicator of habitat diversity: Ferna´ndez-Palacios & Andersson,1993; CarrascalPalomino,2002)onspecies richness.However,this effectdisappearedwhenislandareawas includedinthemodel.Birdsofpreylivingintemperateareas show coarse-grained habitat selection (Sa´nchez-Zapata & Calvo, 1999)becausetheirnumbersarelargelydependenton thetotalextension ofsuitableland.Thusisland-dwellingbirds
ofpreyintemperateareasappeartobehavedifferentlyfrom
Parameter
estimate
Standard
errorFP
thoselivingintropicalzones,whicharemoredependenton small-scalehabitatfeatures(Thiollay,1997).
Food(livestockcarcasses)3.601.05
Nodependence0.00
ontheseamaylimitthedisplacementoflargeandmedium- sized species(Pennycuick, 1972). Migrating raptors travel
Distributionarea
(500kmradius)
)0.710.267.290.0079
fromEuropetoAfricaacrosstheMediterraneanSeamainly
Explaineddeviance(%)8.0
throughthenarrowerstraits(GibraltarandtheBosphorus)to
minimize flyingoveropen sea(Berthold, 2001).However,
partofthemigratoryflowreachesAfricathrough theItalian PeninsulaandthestraitofSicily,whichis140kmwide. This indicatesthatraptorsarealsoabletocrossquitelarge bodies ofwater.Theregularrecordingofnon-breeding individuals andmigratorygroupsonislandssituatedfarawayfromthe continent isevidencesupporting thefactthat watermasses do not representabsolutegeographicalbarriersforraptors. Asmanyas18speciesofnon-breeding birdsofpreyhave beenrecorded on theCanaryIslands;sixareconsideredas regular visitors (Milvusmigrans, Circusaeruginosus,Circus cyaneus,Circuspygargus,Hieraaetus pennatus, Falcosubbuteo: Martı´n Lorenzo,2001).Asimilarpictureisfound inthe Balearics,whereuptoseven non-breedingraptorspeciescan be observed regularly on these islands, mainly during migration(BannermanBannerman,1968).Itislikelythat waterstraitsactassemi-permeable barriers,creatinga metapopulation structure that maydeterminelowprobabil- itiesofcolonizationofappropriatehabitatsandlandsbeyond thebarrier(SerranoTella,2003).
Thelackofasignificanteffectofbodysizeintheprobabilityof presencemodeldoesnotsupportthearguments thatrestrictthe distributionoflarge-bodyraptorstoislandsclosetomainland (AlcoverMcMinn,1994).SpeciessuchastheEgyptianvulture (Neophronpercnopterus)areabletoreachislandsasdistantfrom thecontinentasCapeVerde(600km).However,theprobability ofpresencewashigherforspecies withpartialmigration patterns. Thesespeciesmayhavemore flexiblephylopatric behaviouronthebasisoftheiradaptationtounstableenviron- ments(Hamilton May,1977).Suchflexibilitymayfavour colonizationofislands byindividualsarrivinghazardously duringtheirmigratorytrips.Thelinkbetweenmigrationand probabilityofpresenceisshownagainbyourresults,indicating theroleoftheisland’svicinitytothemainmigratoryroutes crossingthe studyarea.In addition,probabilityofpresencewas positivelyinfluencedbytheexistenceofthespeciesinanareaof
500km radius around the island.Largerareas(and conse- quentlypopulations)may determinenotonlygreaterpossibil- itiesofreachingtheisland(Sara`Morand,2002), butalsoa higherprobabilityofpersistence(post-colonization)througha
‘rescueeffect’(Brown Kodrick-Brown,1977).Finally,a positiveinfluenceofthepresenceoftop-predatorbirdsofprey wasfound,indicatingthatislandswithcomplexfoodwebsare moreabletosustainbreedingraptors.
Extinctionratesonsomeislandshavebeenhighduringthe past century, especiallyfor islandsin some ofthe Macar- onesianarchipelagossuchasCapeVerde.However,thelow devianceexplainedbyour model (8%) indicatesthat these eventsmayeitherdependonotherfactors,differentfromthose analysed,orbemodelledbystochasticevents(demographics, environmental,genetic,etc.)asistheruleforsmall,isolated populations(seeMeffe &Carroll,1994forreview).Itis interestingtonotethatextinctionsappeartobemoreprobable onislands withdensehumanpopulations,which probably reflectsdirectpersecution.
Furthermore, and contrary to our prediction, species directlydependent on human activitiesaremore prone to
sufferpopulationdeclines,andextinctionprobabilitydidnot dependonbody size.Theseresultscanbederivedfromthefact thatspeciesincontactwithhumansaremoresusceptibleto persecution.Scavengersareoftenaffectedbyindirectpoison- ingderivedfrompersecution oflivestockpredators. Thisis consideredtobetheprincipalcauseoftheincreasingrarityof red kites(M.milvus)on CapeVerdeand the Canaryand BalearicIslands,Egyptianvultures on CapeVerdeand the CanaryIslands, andcinereous(Aegypiusmonachus),griffon (Gyps fulvus)and bearded (Gypaetus barbatus) vultures on Sardinia,SicilyandCrete(Hiraldo etal.,1979;Muntaner & Mayol,1996;HilleThiollay,2000;Xirouchakisetal.,2001; Dona´zaretal.,2002).Ourresultsunderlinethatgeneralizable conservationstrategiesarenotadequateforendangeredbirds ofpreyoftheMediterraneanandMacaronesianislands.An increasingnumber ofprojects,suchasthosefunded bythe EuropeanUnion,havebeendirectedtowardsconservationof someofthelastinsularpopulationsofcinereous, beardedand Egyptianvultures (European LIFENature Database; europa.eu.int/comm/life/home.htm). Itseemsclearthatthese projectsmustbebasedoncarefulevaluationofecologicaland human-related factorspotentiallylimitingeachlocalpopula- tion(Dona´zaretal.,2002).
Inconclusion,ourresultsrevealthatlong-termandintense humanizationofMediterraneanandMacaronesianislandshas notalteredthegeneral patternspredictedbytheisland biogeographytheory:therichnessofbirdsofpreydependson islandareaanddistancetocontinentalsourcepopulations.Itis alsoremarkablethatthe tworegionsdidnotdifferin the trends observed(alltheinteractionswerenotsignificant). This suggeststhatourfindingscanprobablybegeneralizedtoother insularsystems. Habitatrequirementsofbirdsofpreyin temperateinsularbiomesappeartobelow, whichshould increase thesurvival oftheirpopulations.Themodelshighlight theimportance ofspecies-basedanalysesinstudiesofisland biogeography:speciesidentity explainedan important part oftheobservedvariability intheprobabilityofpresence (Lomolino,2000).Species’inherentfeaturesaredetermining, butdonotappear tobelinkedtofixedlife-historytraitssuchas thoseimposedbybodysize.Flexibilityinbehaviouralpatterns appearstobeofgreaterimportance.Inparticular,interspecific variabilityintheflexibilityofdispersalandmigratorystrategies maybeimportant inexplaining arrivalandpermanencein insularsystems(Rosenzweig,1995;Owensetal.,1999).
ACKNOWLEDGEMENTS
Wethank R.Barone,C.J.Palacios,V.Penteriani, P.Beja, P.J.PaesdeFaria,A.C.Pereira,F.dePablo,A.Corsoand S.Xirouchakisforunpublishedinformation onraptor status on severalislandsofthe study area,and R.E.Straussfor matlabprogrammes.O.Ceballoscontributedtoformulating the work,and J.L.Tella,D.Serrano,M.Palmerand two anonymousrefereesdiscussedandimprovedthemanuscript. Rachel AtkinsandLaurenBortolottigentlycorrectedthe Englishstyle.TheConsejerı´adeMedioAmbientedelCabildo
InsulardeFuerteventura,theViceconsejerı´adeMedioAmbi- entedelGobiernodeCanarias,andProjectsREN2000-1556
GLOand CGL2004-00270funded thisresearch.M.G.F.was supportedbyaRyCcontractoftheSpanishMCyT.
REFERENCES
Alcover,J.A.McMinn,M.(1994)Predatorsofvertebrateson islands.BioScience,44,12–18.
Alcover,J.A.,Sans,A.Palmer,M.(1998)Theextent of extinctionsofmammalsonislands.JournalofBiogeography,
25,913–918.
Baldacchino,A.E.(1981)LestatutdesrapacesaMalte.Rapaces Mediterraneens (ed. by G. Cheylan and J.C. Thibault), pp.17–18.CROP,Aix-en-Provence,France.
Bannerman,D.A.(1963)BirdsoftheAtlantic Islands.VolI.A historyofthebirdsoftheCanaryIslandsandoftheSalvages. Oliver& Boyd,Edinburgh,UK.
Bannerman,D.A.(1965)BirdsoftheAtlantic Islands.VolII.A historyofbirdsofMadeira, theDesertas,andthePortoSanto Islands.Oliver& Boyd,Edinburgh,UK.
Bannerman,D.A.Bannerman,W.M.(1958)BirdsofCyprus. Oliver& Boyd,Edinburgh,UK.
Bannerman, D.A.Bannerman, W.M.(1966)Birdsofthe AtlanticIslands. VolIII. Ahistory ofbirdsoftheAzores. Oliver& Boyd,Edinburgh,UK.
Bannerman, D.A.Bannerman, W.M.(1968)Birdsofthe AtlanticIslands.VolIV.HistoryofbirdsoftheCapeVerde Islands.Oliver& Boyd,Edinburgh,UK.
Bannerman,D.A.Bannerman,W.M.(1983)Thebirdsofthe
Balearics.CroomHelm,Beckenham,UK.
Berthold,P.(2001)Birdmigration.Ageneralsurvey.Oxford
UniversityPress,Oxford,UK.
Brown,J.H.Kodrick-Brown,A.(1977)Turnoverratesin insularbiogeography:Effectofimmigrationonextinction. Ecology,58,445–449.
Brown, J.H.Lomolino,M.V.(1998)Biogeography,2ndedn. SinauerAssociates,Sunderland,MA,USA.
Burbidge,A.A.Manly,B.F.J.(2002)Mammalextinctionson Australianislands:causes andconservationimplications. JournalofBiogeography,29,465–473.
Carrascal,L.M.Palomino,D.(2002)Determinantesdela riquezadeespeciesdeavesenlasIslasSelvagemyCanarias. Ardeola,49,211–221.
Chown,S.L.,Gremmen,N.J.M.Gaston,K.J.(1998)Ecolo- gicalbiogeographyofsouthern oceanislands:species–area relationships,humanimpacts, andconservation.The AmericanNaturalist, 152,562–575.
Cosson,F.(1985)SomerecentdataoftheraptorsofRhodes (Greece).BulletinoftheWorldWorkingGrouponBirdsof Prey,2,57–60.
Covas,R.BlondelJ.(1998) Biogeographyandhistoryof the
Mediterraneanbirdfauna.TheIbis,140,395–407.
Cramp,S.Simmons,K.E.L.(1980)ThebirdsoftheWestern
Palearctic,Vol.2.OxfordUniversity Press,Oxford,UK.
Dice,L.R.(1945)Measuresoftheamount ofecologicasso- ciationbetweenspecies.Ecology,26,297–302.
Dona´zar, J.A., Palacios, C.J., Gangoso, L., Ceballos, O.,
Gonza´lez,M.J.Hiraldo,F.(2002)Conservationstatusand limitingfactorsintheendangeredpopulation ofEgyptian Vulture(Neophronpercnopterus)intheCanary Islands. BiologicalConservation,107,89–97.
Felsenstein,J.(1985)Confidencelimitson phylogenies:an approachusingthebootstrap.Evolution,39,783–791.
Ferna´ndez-Palacios, J.M. Andersson, C. (1993) Species
compositionandwithinarchipelagoco-occurrencepatterns intheCanaryIslands.Ecography,16,31–36.
Ferna´ndez-Palacios,J.M.Andersson,C.(2000)Geographical
determinantsinthebiologicalrichnessintheMacaronesian region.ActaPhytogeographicaSuecica,85,41–50.
Figuerola,J.(2000)Ecologicalcorrelatesoffeathermitepre- valenceinpasserines.JournalofAvianBiology,31,489–494. Forero, M.G., Tella,J.L.,Hobson, K.A.,Bertellotti, M. Blanco,G.(2002)Conspecificfood competition explains variabilityin colonysize:a test in Magellanicpenguins.
Ecology,83,3466–3475.
Hagemeiger,W.J.M.Blair,M.J.(eds)(1997)TheEBCCatlas ofEuropeanbreedingbirds:theirdistribution andabundance. T&ADPoyser,London.
Hamilton,W.D.May,R.D.(1977)Dispersal instable habitats.Nature,269,578–580.
Handrinos, G.Demetropoulos, A.(1983)Birdsofpreyof
Greece.Efstathiadis& Sons,Athens.
Hille,S.Thiollay,J.M.(2000)Theimminentextinctionof thekitesMilvus milvusfascicaudaandMilvusm.migranson theCapeVerdeIslands.BirdConservationInternational, 10,
361–369.
Hiraldo,F.,Delibes,M.Caldero´n,J.(1979)Elquebranta- huesosGypaetus barbatus. Instituto NacionalparalaCon- servacio´ndelaNaturaleza,Madrid.
Huba´lek,Z.(1982)Coefficientsofassociationandsimilarity
based on binary (presence–absence)data: an evaluation.
BiologicalReview,57,669–689.
Jaccard,P.(1901)Distributiondelaflorealpine danslebassin desDransesetdansquelquesre´gions vo´ısines. Bulletindela Socie´t´eVaudoise desSciences Naturelles,37,241–272.
Janson,S.Vegelius,J.(1981)Measuresofecologicalasso- ciation.Oecologia,49,371–376.
Johnson,K.P.,Adler,F.R.Cherry,J.L.(2000)Geneticand phylogeneticconsequencesofislandbiogeography.Evolu- tion,54,387–396.
Juan,C.,Emerson,B.C.,Orom´ı, P.Hewitt,G.M.(2000)
Colonizationanddiversification:towardsa phylogeographic synthesis fortheCanaryIslands.TrendsinEcologyand Evolution,15,104–109.
Littell,R.C.,Milliken,G.A.,Stroup,W.W.Wolfinger,R.S. (1996)SASsystemformixedmodels.SASInstitute,Cary,NC, USA.
Lomolino,M.V.(2000)Aspecies-based theoryofinsular zoogeography.GlobalEcologyandBiogeography,9,39–58.