The tolerance of island plant-pollinator networks to alien plants

JournalofEcology2011,99,202–213doi:

10.1111/j.1365-2745.2010.01732.x

Thetoleranceofislandplant–pollinatornetworkstoalienplants

ChristopherN.Kaiser-Bunbury1*?,TerenceValentin2,JamesMougal2,DenisMatatiken2andJabouryGhazoul1

EcosystemManagement,InstituteofTerrestrialEcosystems,FederalInstituteofTechnologyZurich(ETH),

¨tstrasse16,8092Zurich,Switzerland;and2NationalParkAuthority,POBOX1240,LaurierRoad,Victoria,Universita

´,SeychellesMahe

1

Summary

1.Invasivealienplantspeciesposeaseverethreattonativeplantcommunitiesworld-wide,especiallyonislands.Whilemanystudiesfocusonthedirectimpactofalienplantsonnativesystems,indirecteffectsofplantinvadersonco-?oweringnatives,forexamplethroughcompetitionforpollinationservices,arelesswellstudiedandtheresultsarevariable.

2.Weusedsixtemporallyandtaxonomicallyhighlyresolvedplant–pollinatornetworksfromthe

,Seychelles,toinvestigatetheindirectimpactofinvasivealienplantspeciesonislandofMahe

remnantnativeplantcommunitiesmediatedbysharedpollinators.Weemployedfullyquantitativenetworkparametersandinformationonplantreproductivesuccess,andpollinatordiversityandbehaviour,todetectchangesinplant–pollinatornetworksalonganinvasiongradient.

3.Thenumberofvisitstoandfruitsetofnativeplantsdidnotchangewithinvasionintensity.Weightedplantlinkageandinteractionevenness,however,wasloweratinvadedsitesthanatlessinvadedsites.Thesepatternswereprimarilydrivenbyshiftsininteractionsofthemostcommonpollinator,theintroducedhoneybeeApismellifera,whileweakinteractionsandstrongnativeinteractionsremainedunchanged.

4.Synthesis.Theimplicationsofthese?ndingsaretwofold:?rst,quantitativenetworkparametersareimportanttoolsfordetectingunderlyingbiologicalpatterns.Secondly,alienplantsandpollinatorsmayplayagreaterroleinshapingnetworkstructureathighthanlowlevelsofinvasion.Wecouldnotshow,however,whetherthepresenceofinvasiveplantsresultinasimpli?cationofplant–pollinatornetworksthatisdetrimentaltonativeplantsandpollinatorsalike.

Key-words:IndianOcean,indirectinteractions,inselbergs,interactionconnectance,invasionecology,invasivealienspecies,plantcommunities,pollinationwebs,SeychellesIslands

Introduction

Theintroductionofalienspeciesintonativeecosystemsconsti-tutesoneofthemajorthreatsforthepersistenceandintegrityofecosystemfunctioning(Macketal.2000;Salaetal.2000).Pollinationinteractionsplayavitalroleinthereproductionofmostplantspecies,andempiricalstudieshaveshownthatalienplantspeciesalterpollinationofneighbouring,mostlynative,plantsthroughsharedpollinators(reviewedinTraveset&Richardson2006;Bjerknesetal.2007).Theeffectsofindirectinteractionsbetweenplantspeciesrangefrommutuallybene?-

*Correspondenceauthor.E-mail:c.kaiser-bunbury@biology.au.dk?Presentaddress:DepartmentofBiologicalSciences,EcologyandGenetics,AarhusUniversity,NYMunkegade114,DK-8000AarhusC,Denmark.

cial(Moeller2004;Moragues&Traveset2005;Molina-Mon-tenegro,Badano&Cavieres2008),toneutral(Aigner2004;Jones2004;Kaiser-Bunbury&Muller2009)andcompetitive(reviewedinMitchelletal.2009).Alienplantscancompeteforpollinationwithnativeplantsthroughchangesinpollinatorvisitationandshiftsinquantityandqualityofpollendeliveredto?owers,thusaffectingplantreproductivesuccessofnativespecies(Chittka&Schurkens2001;Brown,Mitchell&Gra-ham2002;Kandorietal.2009)orcausingshiftsinforagingbehaviourofnativepollinators(Ghazoul,Liston&Boyle1998;Cunningham2000;Ghazoul2004).Sofar,however,noclearandconsistentgeneralpatternhasemerged(Traveset&Richardson2006)andthusitisdif?culttopredicttheconse-quencesofinvasiononnative?oraandfauna(Strayeretal.2006).Arecentsynthesisfoundanoverallslightlynegativeimpactonplant?tnessandpollinatorvisitationrateresulting

Ó2010TheAuthors.JournalofEcologyÓ2010BritishEcologicalSociety

Plantinvasionandpollinationnetworks203

fromsharedpollinatorsbetweennativeandalienplantspecies(Morales&Traveset2009).Mostofthepresentstudiesinvesti-gatesmallsystemsconsistingoffewplantspeciesandtheirpollinators,providingdetailedinsightsontheunderlyingbio-logicalprocessesofpair-wiseinteractions.Here,weuseanetworkapproachtostudytheeffectsofplantinvasiononplant–pollinatorinteractionsandnativeplant?tnessonthecommunitylevel.

Inrecentyears,thestudyofpollinationinteractionsatthecommunitylevelhas?ourishedandavarietyoftoolshavebeendevelopedthathasgreatlyfurtheredourunderstandingofcomplexnetworkstructureanddynamics(Bascompte&Jordano2007).Onecharacteristicpropertyofpollinationnetworksisthenestedstructureoftheirinteractions,whichmeansthatmostpollinationnetworksconsistof(i)acoreofgeneralistspeciesthatinteractcloselywitheachotherand(ii)afewspecializedspeciesthattendtointeractwiththemostgeneralizedspecies(Bascompteetal.2003).Sinceasymmetryofinteractions,whichwaspreviouslynotedbyPetanidou&Ellis(1996),createsahighdegreeofredundancyinpollinationnetworks,ithasbeensuggestedthatthisfeaturecontributessubstantiallytothepersistenceofspecializedspeciesandtherobustnessofnetworks(Memmott,Waser&Price2004;Kaiser-Bunburyetal.2010).Whilethelevelofsophisticationofnetworktoolsandtheirusagehasadvancedrapidly,frompurelyqualitativenetworkparametersbasedonpres-zquez2005;Jordano,Bascompte&ence?absencelinks(e.g.Va

Olesen2006)toquantitativeparametersbasedonameasureofinteractionstrength(e.g.Bersier,Banasek-Richter&Cattin2002;Bluthgen,Menzel&Bluthgen2006),theirdescriptiveandpredictivepowerofreal-worldprocessesandbiological

zquez,Chac-mechanismslagsbehind(Bluthgenetal.2008;Va

off&Cagnolo2009).Forinstance,studiesontheimpactofalieninvasivespeciesonplant–pollinatornetworkshavepro-vidednovelinsightintothemechanismsbywhichthesespeciesintegrateintonativepollinationsystems(Memmott&Waser2002;Olesen,Eskildsen&Venkatasamy2002;Morales&Aizen2006;Aizen,Morales&Morales2008).Atthecommu-nitylevel,however,welackinformationontheimpactofalienspeciesonnativeplant?tnessandpollinatorfauna.Whilesomesmaller-scaleexperimentalstudiesexist(Lopezaraiza-Mikeletal.2007),thereis,toourknowledge,nosuchworkthatcombinesabroadpollinationnetworkapproachwiththeevaluationofbiologicalprocessesunderlyingthechangesinnetworktopology.Ourstudyusedquantitativenetworkparametersasatooltodetectchangesinplant–pollinatorcom-munitiescausedbyplantinvasionandlinkedthesechangestopollinatorvisitationrateandplantreproductivesuccess.

Islandsareparticularlywellsuitedforstudyingtheeffectsofinvasivespeciesonplant–pollinatorinteractions.Mostislandshaveexperiencedmanyintroductionssincehumancoloniza-tion(Kuefferetal.2010)and,togetherwithhabitatdestruc-tionandexploitation,invasivealienspecieshavedevastatednativeisland?orasandfaunasworld-wide(Whittaker&Fern-ndez-Palacios2006).Duetotheirsmallsizeandisolation,a

biotasofoceanicislandsaretypicallydepauperatecomparedtocontinentalones(MacArthur&Wilson1967).Incommunity

studies,smallsystemsizesfacilitateboththeinclusionofallspeciesanddatacollectionoverprolongedperiods,mini-mizingtheriskofsamplingbias.Moreover,islandplant–pollinatornetworksarecharacterizedbypredominantlygen-eralizedinteractions(Bernardelloetal.2001;Rathcke2001;Dupont,Hansen&Olesen2003;Kaiser-Bunbury,Memmott&Muller2009).Inislandsystemswheremanyplantspeciesinteractwithmanypollinatorspeciesandviceversa,most?owersarewhite,greenorgenerallydullincolourandmor-phologicallyunspecializedwithbowl-shapedcorollasoffer-ing?oralrewardstoawiderangeof?owervisitors(Carlquist1974;Webb&Kelly1993;Andersonetal.2001;Kaiser2006).Aliencolonizingplantspeciesthathavesimi-lar?owershapesandcolourstothenativeplantsarelikelytoattractthenativepollinatorcommunityandtherebyinterferewithnativeplant–pollinatornetworks.Collectively,theseconditionsinherenttoislandsprovideanidealbasisforourwork,asalienplantshavethepotentialtointerferewithmutualisticinteractions,andrecommendationsforcon-servationmaybedirectlyapplicabletohabitatrestoration.Weusedsixfullyquantitativepollinationnetworkstoinves-tigatetheimpactofinvasiveplantspeciesonco-?oweringnativeplantsandtheirassociatedpollinatorcommunitiesinanislandsetting.Wehypothesizedthatalienplantscompetewithnativespeciesforpollinationandnegativelyaffecttheirrepro-ductivesuccess.Weasked:(i)Doespollinatorvisitationfre-quencyofnativeplantsdecreasewithincreasinginvasion?(ii)Dochangesinpollinatorforagingbehaviourresultinadeclinein?tnessamongnativeplants?Moregenerally,weinvestigatedthemechanismsbywhichalienplantsimpactuponplant–polli-natornetworksandaffectthetoleranceofnativecommunitiestofutureinvasion.

Materialsandmethods

STUDYSITES

,thelargestgraniticislandoftheOurstudywasconductedonMahe

belongstoSeychelles,IndianOcean(4°40¢S,55°26¢E;Fig.1).Mahe

theoldestoceanicislandgroupintheworld,beingaremnantofGondwanalandfromwhichitsplitc.70millionyearsago.Theislands’greatage,coupledwiththeirgeographicalisolation,hasresultedinauniquebiotawhichhasarisenmorebyevolutionthanimmigration(Dalziel1995),andc.35%ofthenativeSeychelles?oraareendemictotheislands(Stoddart1984).

Whilehabitatclearanceandalienspecieshavepermanentlychan-gedthecharacterofmostnativecommunitiesontheSeychelles,thevegetationongraniticinselbergs–steep-sidedmonolithicoutcrops–hasremainedlargelyintact.Inselbergshavestrikinglydifferentmi-croclimaticandedaphicconditionsfromtheirsurroundings.Conse-quently,theyharbouruniqueplantcommunitiescharacterizedbyanoutstandingdegreeofendemism,which,locally,canbeashighas96%ofthespecies(Fleischmannetal.1996).Inselbergplantcommu-nitiesarecharacterizedbyshrubsandsmalltrees,includingpalmsandscrewpalms.Averagecanopyheightis1–2mwithafewtreesthatattainamaximumheightof4–5m.DominantplantspeciesincludenativePandanusspp.,Memecyloneleagni,Paragenipawrightii,Canthiumbibracteatum,andfourspeciesofendemicpalmsDeckenianobilis,Nephrospermavanhoutteanum,Phoenicophorium

Ó2010TheAuthors.JournalofEcologyÓ2010BritishEcologicalSociety,JournalofEcology,99,202–213

204C.N.Kaiser-Bunburyetal.

Fig.1.Mapofthestudysitesonthemain

,Seychelles.Allsixsitesrepre-islandofMahe

sentmid-altitudeinselbergvegetationwithahighdegreeofplantendemism.

borsigianumandRoscheriamelanochaetes(Robertson1989;Fried-man1994).Thereareno?oweringherbsinthecommunities.Theinselbergscreaterefugesforendemicplantsformerlyfoundinhabi-tatsnowdominatedbyalienspecies.Inselbergs,nevertheless,areexposedtotheencroachmentofalienplants,resultinginaseriesofcommunitiesscatteredacrosstheislandwithavaryingdegreeofinva-issimilarsion.AsthedisturbancehistoryforinselbergsacrossMahe

(L.Chong-Seng,http://wendang.chazidian.comm.),weassumethatthevaryingdegreeofinvasiononinselbergsisprimarilyafunctionofintensityanddura-tionofexposuretoalienplantsfromthesurroundinghabitat.Never-theless,itisimportanttonotethatweanalyseddifferentinselbergswithpotentiallydifferentdisturbancehistoriesinsteadofapurelytemporalinvasiongradient.

Wecollectedfullyquantitativepollinationnetworksfromsixmid-altitudeinselbergcommunities(Fig.1inset;Table1)betweenSep-tember2007andApril2008.Full-seasonnetworkscoveredtheentire?oweringseason(eightconsecutivemonths)andtemporalnetworksreferto1-monthnetworks.Temporalmonthlynetworkscanbetrea-tedassemi-independentunitsbecausethemethodologyandtech-niquesofdatacollectionwerethesameforeachnetworkandthe?owercommunitydifferedbetweenmonths.Thesiteswerelocatedonaverage4.47km(±3.41SD)apart,andthehabitatateachsitewasclearlyseparatedfromthesurroundinghabitatbysteeprockfaces.Noneofthesiteswereactivelymanagedormodi?ed.Weselectedthesitesbasedonthefollowingcriteria:(i)theyrepresentedmid-altitudeinselbergcommunitieswithbroadlythesamemixofnativeplantspecies;(ii)thedistancebetweenthesiteswasconsideredsuf?cientfortheirpollinatorcommunitiestobelargelyindependent;and(iii)thedegreeofplantinvasionvaried,rangingfromsparselyinvadedsitestothosedominatedbyalienplants.Althoughitispossi-blethatafewlong-distanceforagerssuchasbirds,hawkmothsandlargebeescrossedbetweensites,itisrelativelyunlikelythatthese?owervisitorscarriedlargeamountsofpollenbetweenconspeci?csatdifferentsites.Atallstudysites,weestablishedbetweentwoandfourparallel100-mtransectswhichcoveredc.80%ofeachinselberghabi-tat(Table1).Transectsweredividedinto25msectionsforastrati-?ed,randomsamplingofplant,seedlingand?oralabundance.Pollinatorobservationswerenotrestrictedtotransectsbutscatteredacrosstheentireinselbergdependingontheavailabilityof?oweringplants.

INSELBERGPLANTANDPOLLINATORCOMMUNITIES

Wedeterminedthedegreeofplantinvasionandnativeplantspeciescompositionofeachinselbergcommunityby:(i)surveyingeachcom-munitypriortothestudybetween2and8September2007;and(ii)recordingmonthly?oralabundancebetweenSeptember2007andApril2008,theperiodduringwhichonaverage79%(range74–82%)ofinselbergspecies?owered.Fortheplantsurvey,everythreepacesalongeachtransect(walkedbythesameresearcher)werecordedallindividuals>1minverticalheightthatoccurredonbothsidesofthetransects.Thus,ateachpointatotalareaof3m2wassurveyed(3mlengthoneachside·0.5mwidth).Seedlingsandsaplingsatallsiteswererecordedbetween21and27April2008in1·1mquadratseverythreepacesonalternatingsidesofthetransect.

Monthly?oralabundancewasdeterminedbyconducting?owercountsincubicmetresrandomlylocatedas?ve1-m3cubesonalter-natingsitesalongeach25-mtransectsection(n=20cubesper100-mtransect).Thecubeswereplacedattheclosestpatchof?owerstothetransect,and?oweringplantswithinthedistanceof2mperpen-diculartothetransectwereincludedin?owercounts.Wecountedthenumberof?oralunitsin?ower(referredtoas‘?owers’hereafter)ineachcube.One?oralunitwasde?nedasoneindividual?oweroraclusterof?owersinthecaseofpalmin?orescences.Floralabun-dance,themeannumberof?owerspercubicmetre,wascalculatedforeachspeciesbydividingthetotalnumberof?owersbythetotalnumberofcubessampledateachsite(forfurtherdetailsonsamplingmethodsseeKaiser2006;Kaiser-Bunbury,Memmott&Muller2009).

Wede?nedtwomeasuresofinvasionintensityateachsite:(i)theratioofalientonativeplantabundance;and(ii)theratioofalientonative?oralabundance(Table1).Theformerwasusedtoinvestigatechangesinplant-speci?ctraitsalongtheinvasiongradient,http://wendang.chazidian.comworkparameters,however,arestronglyassociatedwithpollinatorbehaviour,whichinturnismostlikelytobein?uencedbytheavailabilityofnativeandalien?oralresourcesinthecommunity(e.g.Hegland&Totland2005;Stang,Klinkhamer&vanderMeijden2006;Bjerknesetal.2007).Thusweusedtheratioofalientonative?oralabundancetoinvestigatetheimpactofinvasiononpollinationnetwork

structure.

Ó2010TheAuthors.JournalofEcologyÓ2010BritishEcologicalSociety,JournalofEcology,99,202–213

Plantinvasionandpollinationnetworks205

,Seychelles,withinformationontotal?owerobservationtimes,nativeandalienplantabundancesandinvasiongradients(ratioofalien:nativeplantTable1.CharacteristicsofthesixstudysitesonMaheand?oralabundances)acrosssites

Pollinatorobservationswereconductedoneachplantspecieswhich?oweredbetweenSeptember2007andApril2008.Afteranini-tialphaseduringwhichobservationsessionslastedfor15min,wechangedthestandardobservationunitto30min(plus2minsettlingtime)toreducethelevelofdisturbance.Weaimedtoobservepollina-torsfor3h?plantspecies?site?month,butbadweatherperiodsandsuddenstopsof?oweringreducedthetotalobservationtime?spe-cies?monthto2.83±0.87h(totalnumberofobservationsessionsatallsitesn=2476resultedin1169hofobservation;Table1).Werecordedtheidentityofall?owervisitorswhichtouchedtherepro-ductivepartsof?owersandthenumberof?owersobserved.Eachvis-itorapproachingtheobservedplantwasconsideredasanewindividualandwasthusrecordedasaseparatevisit.Weusedthetotalnumberofvisitsofeachanimalspeciesasameasureofabundanceofapollinatorspecies.

NETWORKPARAMETERS

Ratio?oralabundancealiens:nativesRatioplantabundancealiens:nativesPlantabundancenatives(aliens)[?m2]

2.512.652.411.571.431.04(0.08)(0.30)(0.42)(0.48)(0.93)(1.01)0.030.110.170.310.650.970.0560.0490.0580.2550.1540.553

Weconstructedplant–pollinatorinteractionmatricesforfull-seasonnetworksateachsiteandcalculatedquantitativeparametersforeachnetwork.Forthenetworkparameters,weusedvisitationfrequencies

zquez,Morris&Jordanoasasurrogateforinteractionstrength(Va

z-2005;Bascompte,Jordano&Olesen2006;Sahli&Conner2007;Vaquezetal.2007)andexpressedmeanvisitationfrequencyperhour(‘interactionfrequency’)asthetotalnumberofvisits??ower?hourofanimalspeciesimultipliedbythe?oralabundanceoftheplantspeciesjvisitedbyi(Kaiser-Bunbury,Memmott&Muller2009).Thatis,eachvisitwasquanti?edbasedonthe?oralabundanceoftheinterac-tionpartnerineachfull-seasonortemporalnetwork.

Theapproachof‘timedobservations’usedinthisstudycontrastthemethodsof‘transectwalks’usedinmostpollinationnetworkstudies.Duringtransectwalks,dataonvisitationarecollectedaccordingtotherelativeabundanceofplantspecies(or?owers)inthesystem,i.e.visitstoabundantplantspeciesaredisproportionatelymorelikelytoberecordedthanvisitstorareplantspecies.Thissam-plingdesignmightincreasetheriskofoverestimatingthedegreeofspecializationofrareplantspecies,andabundantplantspeciesmayappearmoregeneralizedinrelationtoother,lessabundantspeciesinthesystem.Standardizingobservationtimecouldreducethisriskaslongastheobservationtimeofeachplantspeciesissuf?cientlylongtorecordthemajorityofitspollinatorspecies.

Quantitativeparametersareweightedmeasuresthatincorporatetheinteractionfrequencyofindividualmutualists.Thesemetricsaremorerobusttosamplingdifferencesthantheirqualitativecounter-parts(Banasek-Richter,Cattin&Bersier2004),makingthemmoreconservativewhencomparinghabitats.Wecalculatedfourquantita-tiveparametersofwell-describedqualitativeequivalents(seebelowandTable2),i.e.weightedlinkageforplants(lwp;alsocalledweightedvulnerability)andpollinators(lwa;weightedgenerality),quantitativeconnectance(Cq;weightedconnectance),andinteractionevenness(IE)(Bersier,Banasek-Richter&Cattin2002;Tylianakis,Tscharntke&Lewis2007;Kaiser-Bunbury,Memmott&Muller2009).Weightedlinkageandquantitativeconnectance,incontrasttoqualitativelinkage(thenumberofinteractionpartnersofeachspe-cies)andconnectance(thenumberofrealizedinteractionsoverthenumberofpossibleinteractionsinanetwork),givesindividualweighttoeachtaxonrespectiveoftheirtotalinteractionfrequencyandthere-forebettercapturesthefunctionalimportanceofataxoninthecom-munity.InteractionevennesswasderivedfromtheShannonindexandisde?nedasIE=pijlog2pij?log2S,whereSisthetotalnumberofplant–pollinatorinteractionsinthenetworkandpijisthepropor-tionofinteractionsbetweenplantiandpollinatorj(Tylianakis,

Observationtime(h)Totallengthtransects(m)No.transectsAltitude(ma.s.l.)Size(ha)LocationSites

Ó2010TheAuthors.JournalofEcologyÓ2010BritishEcologicalSociety,JournalofEcology,99,202–213

BernicaCopoliaCasseDentresTroisFreLaReserveTeaPlantation55°26¢51–53¢¢55°27¢23–28¢¢55°26¢11–15¢¢55°26¢48–53¢¢55°30¢13–17¢¢55°26¢21–27¢¢E,E,E,E,E,E,4°40¢8–16¢¢S4°40¢7–12¢¢S4°39¢16–17¢¢S4°38¢10–14¢¢S4°42¢35–37¢¢S4°39¢40–42¢¢S$1.7$1.3$0.7$1.3$1.0$1.4320570490460300420442443400400175400400400181.5263.5158.75213.75164187.5

206C.N.Kaiser-Bunburyetal.

Table2.Summarystatisticsofplant–pollinatornetworksatthesixstudysites.Qualitativenetworkparametersincluderatioofanimaltoplantspecies,fullnetworksize,connectance,andmeanandmaximalplantandanimallinkage.Overallconnectance(C)isbasedontheconnectanceofeachmonthlynetwork

CasseDent17(3)4763811522.3564026.42414

8.9±1.393.2±0.39

TroisresFre17(4)5566741782.62115523.02117

8.9±1.163.2±0.44

La

Reserve10(6)49103231533.0678429.82114

10.2±1.453.1±0.43

Tea

Plantation18(2)5469681632.70108026.72214

8.5±1.383.0±0.38

Statistics

Numberofnativeplantspecies(aliens)(P)Numberofpollinatortaxa(A)Numberofvisits(V)

Numberofinteractions(I)RatioA?P

Networksize(S)Connectance(C)

Maximalplantlinkage(lmax)Maximalanimallinkage(lmax)Meanplantlinkage(lp)Meananimallinkage(la)

Bernica16(2)57121171933.17102627.43616

10.7±1.753.3±0.42

Copolia21(2)6572832132.83149518.52217

9.2±1.183.3±0.42

zquez,Chacoff&Cagnolo2009).Inter-Tscharntke&Lewis2007;Va

actionevennessisameasureoftheuniformityofinteractionsbetweendifferentmutualistsinthenetwork,itshowshowwell-distributedinteractionsareamongspecieswithinthecommunity.Unevennet-works,forexample,showhighskewnessinthedistributionsofinter-actionfrequencies.

Foreaseofcomparisontootherpublishedpollinationnetworks,wealsopresentqualitativenetworkparameters(seeTable2).Wecal-culatednumberofplantspecies(P),numberof?owervisitortaxa(A),websize(S=P·A),totalnumberofvisits(V)andinteractionsrecorded(I),andmeanandmaximal(lmax)linkageforanimal(la)andplantspecies(lp).Wealsopresentnetworkconnectance,C=100·(I?S)basedonthemeanconnectanceforeachtimeper-iod(Medanetal.2006).

DATAANALYSIS

Allanalyseswererunwiththestatisticalsoftwarer2.9.1(RDevelop-mentCoreTeam2005),mixedeffectmodelswere?ttedwiththefunc-tionlmer(library:lme4;Bates&Maechler2009),andmodeldesignandselectionprocedurefollowedrecommendationsbyZuuretal.(2009).

Wecomparedtheabundancepatternsofnativeandalienadults,seedlingsandsaplingsbetweensiteswithalinearmixedmodel.Seed-lingandsaplingabundance(log-transformed)wereenteredasresponsevariablesinseparatemodelsandtheexplanatoryvariablesadultabundance(log-transformed)andoriginofthespecieswereenteredasmainandinteraction?xedeffects.Themodelsincludedspeciesandsiteascrossedrandomeffects.

POLLINATORDIVERSITYANDINTERACTIONFREQUENCY

accountforpotentialtemporalandspatialnon-independenceweincludedsiteandtimeperiodascrossedrandomfactorsintothemod-els.(ii)Weappliedalog(Y+0.001)transformationtointeractionfrequencyandusedthedataasresponsevariableinthemodel.Floralabundanceandinvasionlevelwere?ttedas?xedeffects,andonlymaineffectswereincludedintothe?nalmodel.Interactionfrequencyforeachplantspecieswaspooledfortimeperiodstoreducethenum-berofobservationswithnovisits.We?ttedonecompletemodelwithallnativeplantspeciesandindividualmodelsforeachofthesevenmostcommonnativeplants.Inthecompletemodel,plantspecies(n=30)andsite(n=6)wereenteredascrossedrandomeffects.Inthemodelsforindividualplantspecies,timeperiod(n=8)andsitewereincludedascrossedrandomeffects.ToobtainP-valuesformixedmodelsweranMarkovChainMonteCarlo(MCMC)for10000iterationsusingthefunction‘pvals.fnc’inthelibrary‘languag-eR’(Baayen2008).MCMCisatypeofBayesianGibbsSamplingapproachwhichvariesparametervaluesrandomlyandestimateslike-lihoodtobuildupadistributionfromwhichpointestimatesandcred-ibleintervalsarederived.

POLLINATIONSUCCESS

Toassesstheeffectofalien?oweringplantsonpollinatordiversityandthedegreeofcompetitionforpollinatorvisitationbetweennativeandalienplants,wedeterminedtheratioin?oralabundanceofalientonativeplants(invasionlevel)foreachtimeperiodandsite.We?t-tedtwolinearmixedmodelswithaGaussiandistribution:(i)thenumberofpollinatorspeciesateachsiteandtimeperiod(n=48)wasenteredasresponsevariableand?oralabundanceandinvasionlevelwereenteredas?xedeffects.Floralabundancewasincludedinthemodelasitwaspreviouslyshowntoaffectpollinatorspeciesrich-ness(Rathcke1983;Thompson2001;Ebelingetal.2008).To

Werecordedpollinationsuccessofnativeplantsastheproportionoffertilized?owers,i.e.fruitset.Werecordedfruitsetfromatotalof31nativeplantspecies.Ofthose,16speciesoccurredattwoormoresitesinsuf?cientnumberofindividuals(>3?oweringfemales).Sevennativeplantspeciesoccurredatfourormoresitesinsuf?cientnum-bersforrandomization,soindividualbetween-sitecomparisonswererestrictedtotheseplantspecies.Ofthe16species,12(75%)areeitherself-incompatiblehermaphrodite,dioecious,orconsecutivelymonoe-cious,andofthesevenmostcommonspecies,?ve(71.5%)aredepen-dentonavectortodeliverpollenbetween?owersforreproduction.Wemarkedatotalof29107budson594branchesorin?orescenc-esnestedin406plants.Duetothedifferentnumberoftrees(orshrubs)perspeciespersite,thedataonfruitsetwereunbalanced.Also,forthecalculationandanalysesoffruitsetweneededtoaccountforthenumberof?owersoneachplant.Wethereforeanaly-sedfruitsetpatternswithageneralizedmixedeffectmodel,usingabinomialdistribution,with?oralabundanceandinvasionintensityas?xedeffects,andspecies,treeandin?orescenceidentityasnestedandsiteascrossedrandomeffects.Thismethodisrobustforunbal-anceddataand,byusingthe‘cbind’commandtocalculatefruitsetweweightedthesamplesizesbythenumberof?owersrecordedon

Ó2010TheAuthors.JournalofEcologyÓ2010BritishEcologicalSociety,JournalofEcology,99,202–213