Consistencyinthehabitat degreeofinvasionforthree invasiveplant speciesacrossMediterraneanislands
LaurenceAffre• Carey M.Suehs• Ste´phane Charpentier • MontserratVila`• Giuseppe Brundu • PhilipLambdon • AnnaTraveset • PhilipE.Hulme
Abstract Habitats areknowntovaryintheir vulnerabilitytoinvasion byalienplantsanddifferent speciesoftencolonise distincthabitats. Toassessthe consistency inthedegreeofinvasionofparticular habitats,thisstudyexaminedthefrequency ofoccur- rence andlocal abundanceofthree invasiveplant taxa:Ailanthus altissima,Carpobrotusspp.,and Oxalispes-caprae acrossdifferenthabitattypeson
L.Affre() C.M.Suehs S.Charpentier
InstitutMe´diterrane´end’EcologieetdePale´oe´cologie (IMEP,UMRCNRS-IRD6116),Universite´Paul Ce´zanne,case462,13397Marseillecedex20,France e-mail:
M.Vila`
Estacio´nBiolo´gicadeDon˜ana(EBD-CSIC),Avda. Ame´ricoVespucios/n,IsladelaCartuja,41092Sevilla, Spain
G.Brundu
DipartimentodiBotanicaedEcologiavegetale,Universita`
degliStudidiSassari,ViaF.Muroni25,07100Sassari, Italy
P.Lambdon
NERCCentreforEcologyandHydrology,Hillof
Brathens,Banchory,AberdeenshireAB314BW,UK
A.Traveset
InstitutMediterranid’EstudisAvanc¸ats(CSIC-UIB), MiquelMarque´s21,Esporles,BalearicIslands,Spain
P.E.Hulme
TheBio-ProtectionResearchCentre,LincolnUniversity, POBox84,Canterbury,NewZealand
fourrepresentativeMediterranean islands.Wecon- ductedsystematicfield surveysrecordingthepres- ence-absenceandcoverofthesetaxaontheislandsof Mallorca,Corsica,SardiniaandCrete.Drawingonthe resultsof5,285samplepoints,thefrequency of occurrence ofthethreeinvaderstendstobehigher thanexpectedinurban,ruderal androadsidehabitats. Incontrast,scrubhabitatsrarelycontainanyofthe threeinvaders, indicatingthattheymaybemore resistanttoinvasion.Thedegreeofinvasion,deter- minedbythelocalabundanceofaninvasive plantin anyonehabitat, variesaccordingtothe identity ofthe invaderandtheisland.However, basedonaverage abundance, Oxalispes-capraeexhibitsthehighest degreeofinvasion, andCarpobrotusspp.theleast. Thereisno indication thatany oneofthefourislands ismorepronetoeitherhigherfrequenciesorabun- dancesofthethreeinvaders. Thesepatternssuggest thatanthropogenic changesinMediterranean islands willincreasethevulnerability ofcertainhabitatsto invasionandincreasethedistributionofthesethree invasivetaxaatanyofthefourislands.
Keywords Alien Exotic Habitatinvasibility
Riskassessment Speciesabundance Weeds
Introduction
Islandsareoftenperceivedashighlyvulnerabletothe invasionsofalienmammalsandplants(Elton1958;
LoopeandMueller-Dombois1989;MacDonaldand Cooper1995).Incontrasttothehighprofile impacts ofmammalsonislands thatincludeextinctionof endemicspecies(Courchamp etal.2003),thecon- sequencesofalienplantinvasionshaverecentlybeen portrayedascomparativelybenign(SaxandGaines
2008).Although theproportion ofislandfloras comprising alienspeciescanbehigh,e.g.,*50% (Hulme 2004), thereisscantevidencethatnative species havebeenlostfromislands asaresultof competitionfromalienspecies.Thismay,inpart,be a consequence that alien plant driven extinctions occuroveramuchlongertimescalethanalienanimal driven extinctionsandthatthemajorimpactofalien plantsonislandbiodiversity hasyettoberealised. Alternatively, alienplantspeciesmaybelargely restrictedtohuman dominatedhabitats ofislands and theirimpactonnativecommunities, andhencethe threattoendemics,mightbelimited.
Anthropogenichabitatstendtoexhibitthehigh- est levels of plant invasion across continental Europe, asaresultofbothhigherdisturbanceand propagulepressure(Chytry´etal.2009;Pysˇeketal. inpress). Thispatternisexpectedtoholdtruefor islandsaswell, anditissupported bystudies examining islands and neighbouring continental areas(Gimenoetal.2006).However,whilehuman dominated habitatsaremorefrequentlyinvadedon bothislandsandcontinents, therangeofhabitats occupiedbyalienplantsonislandshasbeenfound to be broader (Gimeno et al. 2006). This may indicate a potentially greater threat of invasive plantsonislands.Toassesssuchathreatweneed to knowthe rangeofhabitatsinvadedonislands, theproportionofavailablehabitats occupiedby invaders,andtheaverageabundanceoftheinvader inspecifichabitats.
Inthiscontext,theMediterranean Basinisagood exampleofanisland-rich,historicalrefugezoneof high conservation value. Estimates of local or regional endemism vary from 37.4 to 50%, with manyoftheseplants beingislandormountain top endemics(Heywood 1995;Andre´sandOjeda2002). Furthermore,thefragileislandhabitatsintheMed- iterraneanBasin havebeen associated with human interventionforthousands ofyears(Cody1986)and areincreasinglydisruptedbyever-expanding urban development, tourism,andtransportinfrastructures (Me´dailandQue´zel1997;Hulme2004),aswellas
invasive plantsthatinvariablyaccompanythispro- cess(e.g.,Lloretetal.2004).
Todeterminethevulnerability ofdifferentMed- iterraneanhabitats toinvasion, weconducted an extensivefieldsurveyofthree‘‘highrisk’’invasive plantsacrossdifferent habitattypes onfourlarge islandsspanningtheMediterranean Basin(Mallorca, Corsica,Sardinia andCrete).Thespeciesselected werethreeofthemostwidespread alienplantson Mediterraneanislands(Hulmeetal.2008),knownto havesignificantimpactsonnativeplantdiversity and soilstructure(Vila`etal.2006)andabletoreachhigh localabundance(Travesetetal.2008).Thesespecies werethefast-growingtreeAilanthus altissima,the scramblingsucculentchamaephyteCarpobrotusspp., andtheannualgeophyte Oxalispes-caprae.Our specificaimsinthisstudyare(1)todeterminewhich habitats aremoreinvaded bythethreealienspecies and(2)totestifthereareconsistencies inhabitat invasionamongspeciesandamongislands.
Materialsandmethods
Invasivetaxa
Thethreealientaxa(hereafter called‘‘invaders’’) werechosenfortheirwidedistributionacrossmost Mediterranean islands,theiroccurrenceinarangeof different habitats and,whileallthreeareclonal,the diversity oflife-formstheyrepresent. Allthreeare includedintheonehundred ofthemostinvasive speciesinEurope(DAISIE2009).
Ailanthusaltissima(P.Mill.)Swingle(Simaroub- aceae) is a fast-growingtree native to China and NorthVietnam(BasnouandVila`2009).Ithasbeen introducedtomostcontinentsasanornamentaltree andforthelandscapingofroadsides.Seedsarewind- andwater-dispersed (meanproductionof300,000 seeds/individual/year),andthespeciesexhibitsstrong resprouting and clonal growth of stems (Kowarik
1995).Exudates fromallpartsoftheplanthavebeen shown tobeallelopathic(DeFeoetal.2003) and herbicidal(HeiseyandHeisey2003).
Carpobrotusspp. L.(Aizoaceae)arerobust, scrambling,mat-formingsucculentsnativetoSouth Africa(WisuraandGlen1993;Delipetrou 2009), havebeenintroducedasornamentals withinpublic andprivategardens,andtostabilisesoilsalongcoastal
dunes,rockyslopes,andcliffs(e.g.,D’Antonioetal.
1993;Suehsetal.2001). IntheMediterraneanBasin, thepresenceoftwohybridisingtaxa(C.edulisand C.affineacinaciformis)iswelldocumented(Suehs etal.2004a, b).Thesetwostronglyclonaltaxa producefleshyfruits(meanproductionof1,000seeds/ fruitand 25fruits/m2)and seeds aredispersed endozoochorously by small mammals (Bourgeois etal.2005)orsecondarymyrmecochory.Inaddition, stemfragmentsarecapableofrootingandestablishing anewclone(Vila`andD’Antonio1998).
Oxalispes-capraeL.(Oxalidaceae)isageophyte nativetoSouthAfricaandintheMediterranean is functionally sterile (Lambdon 2009). The species doesnotproduceseedsbutspreadsasexually viathe production ofundergroundbulbs(Pu¨tz1994) thatare oftendispersed onagriculturalmachinery.Itis primarilyacolonist ofdisturbedgroundandagricul- tural fields from the coast to around 600m in elevation(Rossetal.2008).
Studyislandsandfieldsurvey
Fieldsurveyswereundertakenon fourMediterranean Basin islands (Fig.1): Mallorca (Spain), Corsica (France),Sardinia(Italy)andCrete(Greece).These islandsarefourofthelargestintheMediterranean, supporting ahighhuman populationdensity and encompass astrongEast-Westclimaticgradient (Table1).Thebasicgeographic andclimaticparam- etersoftheislandsarepresented inTravesetetal. (2008).
Fiverandomly chosen10910kmUTMcellsper invaderweresurveyed oneachisland.Sampling reflected the geographic distribution of taxa and
whereallthreetaxacoincidedgeographically,sam- plingcouldbelimitedtothesamefiveUTMcells (e.g.,Crete,Sardinia) but,wheregreatergeographic heterogeneity occurred,UTMsamplesweredesigned totargetlandscapes knowntobeinvadedbypartic- ulartaxa.Insomecases,different UTMcellswere surveyed separatelyforeachinvader(e.g.,Mallorca) whilethesamplingstrategyinCorsicausedamixture ofunique andshared UTMcellsfordifferenttaxa (Table1). For all islands,there exist considerable montaneareasinwhichinvaders areabsent,our samplingstrategythustargeted‘‘invaded’’ land- scapes,thoughthedegreeofinvasionineachcase wasstillrelativelylow.
Tostratifythesampling, eachUTMcellwas dividedinto 100191km squares,withinwhichone randompointwasselectedandsurveyed. Ateach point,thefollowingdatawere recorded:(1) theUTM EastandNorthcoordinates (East,North),(2)the habitat typeswithina 50m radiusof the random point(i.e.,agricultural,coastal,fen,forest, grassland, railroad,river,road,rocky,ruderal, scrub,orurban), (3)thepresence/absence ofeachinvaderineach habitatfoundwithinthe50mradius area,and (4)the abundance (%cover)ofeachinvader ineachhabitat accordingtoaDAFOR scale(Richetal.2005): Dominant([75%),Abundant(50–75%), Frequent (25–50%),Occasional(5–25%)andRare(\5%).For analysis, theDAFOR semi-quantitativecategories (i.e., 1–5) were considered. Therefore, at each randompoint,wecouldobtainthefrequency of occurrenceandcoverabundancevaluesoftheinvader aswell astheoverallavailabilityofeachhabitat. Variationinthenumberofsamplingpointsperinvader andislandoccurredduetotheinaccessibilityofcertain
Fig.1 Mapofthe MediterraneanBasin includingthefoursampled islands(Mallorca,Corsica, SardiniaandCrete)forthe presenceoftheinvaders Ailanthusaltissima, Carpobrotusspp.and Oxalispes-caprae
Table1 Thesurfacearea,distance fromthenearestmainland,populationdensity,annualtouristdensity,number ofUTMcells,and totalnumberofpointssampledintheseUTMcellsforeachofthefourstudyislandssurveyed
Island / Surface area(km2) / Distance(km) / Populationdensity
(persons/km2) / Touristdensity
(persons/km2/year) / UTM
cells / Sample points
Mallorca(Spain) / 3,656 / 170 / 174 / 1,712 / 15 / 1,078
Corsica(France) / 8,682 / 82 / 30 / 709 / 11 / 1,306
Sardinia(Italy) / 24,090 / 230 / 68 / 104 / 5 / 1,485
Crete(Greece) / 8,700 / 100 / 72 / 241 / 5 / 1,416
points (e.g.,cliffs,militaryareas,privateproperty) resulting inatotalof5,285samplingpoints (outofa maximumof6,000)collectedacross thefourislands (Table1).
Degreeofinvasion
Fourvariableswerederivedtoassessthedegreeof invasionat boththe islandandhabitat scales:(1)
calculatedforeachhabitat inwhich aninvader occurred. Afterwhichanaveragewasthenderived acrossallthehabitatsforanindividualisland.
Habitatapparency(HA)whichmeasureswhether
aninvadertendstobeabundantinrareorcommon habitatswasalsocalculated:
i¼h
HA¼XðHi=NÞDAFORi ð2Þ
i¼1
proportionpresence(PP),theproportionofsampling
points(e.g.,frequencyofoccurrence)inwhichan
here,histhetotalnumberofhabitats,Hi
thenumber
invaderwasfound,(2)themeaninvaderabundance (DAFOR) scoredacrossallsamplingpointswhereit wasfound (i.e.,excludingzeros), (3)invasion saturation (IS);cf.PysˇekandPysˇek(1995),and(4) habitatapparency(HA).Invasionsaturation (IS) compares theaverageabundanceofaspeciesacross allsampling points whereitwasfound(i.e.,exclud- ingzeros)totheaverageabundanceacross all sampling pointsthatweresampled (i.e.,including zeros). Values near 0 indicate that the invaded habitatsintheregionarefarfromsaturated (e.g., manysamplingpointshavenoinvader), whereas values approaching 1 indicate that the invaded habitatsintheregionareclosetosaturation and, therefore,mostsamplingpointsareoccupied. ISwas assessed foreachtaxon andforeachhabitat inwhich itwasfound,andthenaveraged acrossallhabitats to obtainascoreforeachisland:
ofsamplingpointsinwhich habitatiwasrecorded,N
thetotalnumberofsamplingpointsacrossallhabitats andDAFORiisthemeanDAFORscoreforhabitati excludingzeros.Foreachhabitatonanyoneisland, Hi andNaretheconstantsandthusHAprovidesa relativemeasureoftheapparency ofaninvaderand whetherittendstobeabundantincommonorrare habitats.HAvariesfrom0(i.e.,nohabitatisinvaded) to5(i.e.,theinvaders can,onaverage,reach dominanceinallhabitats).
Statisticalanalysis
First,wetestedthenullhypothesis thattheobserved frequency ofoccurrenceofeachinvaderina particularhabitatreflected theavailabilityofthe habitatinthelandscape.Statisticalsignificancewas calculatedastheproportionoffrequenciesrandomly
generatedfromavailable-habitat data(basedon5,000
i¼n
IS¼XDAFOR0=DAFORi
i¼1
!ob
nð1Þ
randomizationsplustheonefield observation)that wereasextremeormore extreme(one tailed)than observedinvadedfrequencies(Manly1997).
here,nindicatesthenumber ofdifferent invaded habitats,DAFOR0isthemeanDAFOR scorefor habitatiincludingzeros, andDAFORiisthemean DAFOR scoreforhabitatiexcludingzeros.Invasion saturationcouldonlybecalculatedforhabitatsthat exhibitedsomeoccurrenceoftheinvaderandwas
Second,wealsotestedifdifferencesinthemean
DAFOR scoresacrossindividualhabitatsandtheir among-habitatmeansresulted fromarandomalloca- tionoftheavailableDAFORscorestotheinvaded habitats.Significancewascalculatedasthepropor- tionofhabitat meansversusoverall-mean differences
(basedon5,000randomizationsplustheonefield
Table2 Thepercentageofhabitatssampledperisland
observation)thatwereasextremeormoreextreme(one-tailed)thanfieldobservations(Manly1997).
Third, mean DAFOR scores were compared / Island / Mallorca / Corsica / Sardinia / Crete
Agricultural / 25,79 / 13,86 / 25,79 / 29,94
between islands and between invaders using two- / Coastal / 5,75 / 6,20 / 0,07 / 0,99
grouprandomizationtestsdue toextremeheterosced- / Fen / 0,00 / 1,61 / 0,00 / 0,21
asticitythatwasnotresolvedviadatatransformation. / Forest / 13,91 / 7,73 / 6,20 / 5,58
Significancewasassessedbycomparingthe mean / Grassland / 31,54 / 7,73 / 43,30 / 5,58
differencesbetweentwogroupswiththeone-tailed / Railroad / 0,00 / 0,77 / 0,00 / 0,00
distributiongenerated from 5,000randomsamples / River / 1,67 / 6,20 / 0,07 / 4,03
withoutreplacementofthosetwogroupswithBon- / Road / 0,37 / 10,57 / 0,00 / 8,19
ferroni adjustment(Manly1997). / Rocky / 0,00 / 3,60 / 0,40 / 3,60
Finally,atwo-wayfactorialMultivariateAnalysis / Ruderal / 6,12 / 3,60 / 1,14 / 6,64
of Variance (MANOVA) was used to compare / Scrub / 4,82 / 26,19 / 19,66 / 29,31
invaderandislandeffects,aswellasanislandby
invader interaction on invader occurrence (PP), / Urban / 10,02 / 11,94 / 3,37 / 5,93
invasionsaturation(IS)andhabitatapparency(HA).
PPandISvalueswereBox-Coxtransformed(PP0 =
(PP0.39 -1)/0.39and IS0 =(IS0.37 -1)/0.37)before analysisto eliminate heteroscedasticity(Sokaland
Rohlf 1995).PosthocScheffe´tests werethenused to pinpointpair-wisedifferences betweenislandsand betweeninvaders(SokalandRohlf1995).
Statisticalanalyseswerecomputed usingtheopen sourceRsoftwareprogram,withthestatsandade4 packages(RDevelopmentCoreTeam2008).
Results
Habitatdegreeofinvasion
The percentage of habitats sampled per island is givenin Table2.Forall invaders,urban,ruderal, androadhabitatswereallsignificantly(P\0.05) over-representedon at least one island (Table3). Coastalhabitatsweresignificantlyover-represented onlyforCarpobrotusonallislands. Incontrast, agriculturalhabitatsweresignificantly over-repre- sentedonlyforOxalis inSardiniaandCrete,while river habitats were significantly over-represented onlyforAilanthus inCorsica.Withtheexceptionof Ailanthus inMallorca,scrubhabitatswereconsis- tently (P\0.05) under-representedon all islands. Forests, grasslands androckyhabitatsareeither significantly under-represented, ornon-significantly representedforanyinvaderandisland.Finally,fen habitatswerealwaysnon-significantly under- represented.
Comparisonsbetweenmean DAFORscores for
individualhabitats andtheiroverallhabitatabun- dancerevealsignificantdifferencesinthe abundances of each invader (Fig.2). Significantly (P\0.05) higher-than-average abundancesweredetectedfor Oxalisingrassland habitats inMallorca,agricultural habitatsinCorsicaandCrete,andforesthabitatsin Sardinia, whereas Carpobrotuswasmoreabundant thanexpectedinroadside habitatsinMallorcaand coastalhabitatsinCorsica.Significantly lower-than- averageabundances werefoundforOxalisinruderal habitatsin MallorcaandSardinia,roadsidehabitatsin Corsica,andgrassland, rockyandscrubhabitatsin Crete.Similarly,significantly lower-thanaverage abundances werefoundforCarpobrotusinurban habitatsinCorsica,aswellasforAilanthus inforest andruderalhabitatsinSardinia.
Invaderandislanddifferences
Nosignificant differencesbetweenmeanDAFOR scoreswerefoundbetweentheinvaderswherethey werepresent(Fig.3).Incontrast, allislandpair-wise differences betweenmeanDAFORscores,except Mallorcavs.Sardinia,weresignificant.Corsica(mean DAFOR±SE=1.67±0.11) had significantly lower mean scores than Mallorca (DAFOR=
2.27±0.16;P\0.01),Sardinia(DAFOR=2.31±
0.22;P\0.01)andCrete(DAFOR=3.09±0.27; P\0.001),whileCretehadsignificantly highermean abundance scores than Mallorca (P\0.01) and Sardinia(P\0.05).
Table3 Comparison
MallorcaCorsicaSardiniaCrete
betweenavailableand
occupiedhabitat frequenciesforeach invaderandisland
Significant(P\0.05)over-represented
Fig.2 MeanDAFOR scoresforAilanthus altissima(white), Carpobrotusspp.(grey)and Oxalispes-caprae(black)
indifferenthabitatsin Mallorca,Corsica,Sardinia andCrete.Themean DAFORscoresacross habitatsareindicatedby horizontallines(Ailanthus altissima:dottedlines; Carpobrotusspp.:grey; Oxalispes-caprae:black). Asterisksindicate
significantdifferences relativetotheall-habitat meansattheP\0.05(*), P\0.01(**)and
P\0.001(***)levels
Thetwo-wayfactorialMANOVAonPP,ISand HA revealed significantisland (F=5.56, df=9, P\0.001) and invader (F=89.15, df=6, P\
0.001)effects,aswellasasignificantisland9invader interaction(F=6.72,df =18,P\0.001),indicating
thathabitattrendsofinvasiondegreewerenotthe sameacross islands.
Post hoc Scheffe´ tests pinpointed two signifi-
cantislanddifferencesforPP,withCrete(mean±
SE: PP=0.27±0.05; P\0.01) and Mallorca
Fig.3 Means(±standard errors)fortheproportionpresence (PP),meanDAFORscoreswhenpresent(DAFOR), invasion saturation(IS)andhabitatapparency(HA)fortheinvaders
Ailanthusaltissima,Carpobrotus spp.andOxalispes-caprae ontheislandsofCorsica(Co),Crete(Cr),Mallorca (Ma)and Sardinia(Sa)
(PP=0.24±0.06; P\0.05) having significantly higherfrequencyofoccurrencecomparedtoSardi- nia(PP=0.15±0.04).Corsica(IS=0.14±0.02; HA=2.39±0.09)hadsignificantlylowerinvasion saturationthanCrete(IS=0.31±0.04;P\0.001, Fig.3)andhabitatapparencythanMallorca(HA=
2.53±0.15;P\0.05,Fig.3).
Similarly,posthocScheffe´testshighlightedtwo significantinvaderdifferences(P\0.001)for PP, andISmeans:Oxalismeanfrequencyofoccurrence
and invasion saturation (PP=0.36±0.04; IS=
0.37±0.04) were significantlyhigher than either Carpobrotus(PP=0.09±0.02;IS=0.16±0.02) orAilanthus (PP=0.16±0.04;IS=0.16±0.04). Allthreeinvadercomparisonsweresignificant for habitatapparencywithCarpobrotus (HA=1.76±
0.04)havingthelowest,Oxalis(HA=2.85±0.05) thehighest,andAilanthus(HA=2.75±0.04)an intermediatehabitatapparency. Allremainingpost hocScheffe´testswerenon-significant.
Discussion
Thisstudyattempts,forthefirsttime,tosystemat- icallyandempiricallyassess generaltrendsinthe regional andlocalabundance ofthreewidespread invader plants across four Mediterranean islands. Fourdifferentmeasures ofthehabitatdegreeof invasion wereusedtoexplorethesetrendsandbetter characterise therisks these taxapose toMediterra- neanhabitats.Habitatisconsidered agoodpredictor ofthedegreeofplantinvasionattheregionalscale measuredasalienspeciesrichnessorpercentageof aliensinthetotalflora(Chytry´etal.2008a,b;Pysˇek etal.inpress).Inourstudy, wefoundthatthedegree ofinvasionmeasured asthefrequency ofoccurrence andlocalabundanceofaninvader wasrather consistentacrosssites. Overall,the trendssupport theviewthatinvaders areprimarilyathreatto anthropogenic habitatswhilelessdisturbed habitats appearresistanttoinvasion.Nevertheless, certain patternswerecontextdependentanddiffered among islands.
Anthropogenichabitatsashotspotsofinvasion
Thefrequencyofoccurrenceofinvadersinanthro- pogenichabitats(i.e.,urban,ruderal, roadandrail- road)waseitherinproportiontosignificantly higher compared totheiravailability.Frequency ofoccur- rencewas never less than whatmightbeexpected giventheavailabilityofthesehabitats. Theassoci- ationofinvasive specieswithhuman activitiesand ruderalareashasbeenwidelydemonstrated at different spatial scales (Cadotte and Lovett-Doust
2001;Vila` and Pujadas2001;Deutschewitzet al.
2003;O’FarrellandMilton 2006;Hulme2009a). However, ourstudyhighlightsthatalthoughinvaders frequentlyoccurwithinanthropogenic habitats,they donotnecessarilyattainhighabundance. This suggests that,whilehumanactivitiesmayfacilitate high frequency of invader occurrence in these habitatspossibly duetohigherpropagule pressure, abundancemayitselfbelimitedbyfrequentdistur- bancesand/oralackofsuitableconditions for populationgrowth(Travesetetal.2008;Vila`etal.
2008). Thus,whileoftenperceivedaspotential sourcesforfurtherinvasion, suchanthropogenic habitatsmaynotnecessarily actasimportantsources forspreadtothewiderenvironment(Bothametal.
2009)andlonger human presencedoesnotnecessar- ilyresultinanincreased degreeofinvasion(Planty- Tabacchietal.1996).Given therapidanthropogenic transformation oftheMediterranean,abetterunder- standing oftheroleofthesehabitatsinthespreadof alienspeciesisrequired.
Mediterraneanscrubsasinvasion-resistant habitats
Forcertainhabitats,thefrequency ofinvaderoccur- rencewasconsistently lowerthanexpectedfrom habitat availability, and invaders never reached higherthanaverageabundance.Thisisparticularly thecaseforMediterranean scrublandsthathave previouslybeenshowntohavealowlevelofplant invasion (Pinoetal.2005;Vila`etal.2007).Most alienspeciesintroducedtotheMediterraneantend nottobeastolerantasthenativespecies tothexeric conditions typical of this habitat (Lambdon et al.
2008;Travesetetal.2008;Vila`etal.2008).This
resultisespecially interestinginlightofthefactthat scrubhabitats form alargefractionoftypical Mediterranean islandvegetation,includingourstudy islands.Maintaining naturalscrubhabitatsmaythus serve as a significant barrier to invasive species spread(Marvieretal.2004).
Contextdependenttrendsininvasion
Mostnaturalhabitatsexhibitpatternsthatareinter- mediatebetweenanthropogenic andscrubhabitats. Thedegreetowhichthesehabitats wereconsistently sampledacrosshabitats variedandreflectedtheoften differentlandscapes sampledacrosstheMediterra- nean.Riparian androckoutcrop/cliff habitatswere encounteredrelativelyinfrequentlyandingeneralthe frequency ofinvaderoccurrence waswhatmightbe expectedgiven theiravailabilityinthelandscape. Riparianareasarefoundtobeparticularlyvulnerable toplantinvasions (Pysˇeketal.inpress)butthismay notbetrueforthetaxaselectedforourstudy. Other habitatsdidnotrevealconsistenttrendswhencom- paredacrossthefourislands, notablyforest,grass- land,andcoastalhabitats. Noneofthesehabitats revealedlowerorhigherthanexpected frequenciesof invaderoccurrence, andtheyneverpresentedconsis- tentlyhigherorlower thanaveragelocalabundances. Amoredetailedanalysisthat examinesthe role ofthe
landscape matrixandthedegreetowhichsuch habitatsarefragmentedorclosetopropagulesources isneeded tofullyassesssuchvariabilityinhabitat vulnerabilitytoinvasion(Ohlemu¨lleretal.2006).
Invaderdifferencesonthedegreeofinvasion
Therewerenoconsistent trendsacross islandsinthe fourmeasuresofinvasion wecalculatedforeach taxon.However,boththefrequencyofoccurrence andinvasionsaturation highlightthattherewere markeddifferences amongislandsinthedegreeof invasionforAilanthus (particularlyinCrete)and Oxalis(particularlyinMallorca) butnosuchinter- islandvariationforCarpobrotus spp.Nevertheless, thefrequency ofinvaderoccurrenceandinvasion saturationwas low for all invaderson all islands (withtheexceptionofOxalisinMallorca),indicat- ingthattheinvadersremainrelativelyrare.Given that the three invaders were introduced over
100yearsago(Travesetetal.2008),averagerates ofspreadacross theseislandsdonotappear tobe particularlyhighsinceasignificant proportionof suitable habitatremainsavailableforcolonisation. Othermeasuresofthedegreeofinvasion suchas meanlocalabundanceandhabitatapparencypresent adifferentpicture.Meanlocalabundance was consistentlyhigherforallthreeinvadersonCrete, buttrendsinabundance weremoreindividualistic on theotherislands. Habitatapparency variedleast amongislandsbutmostamongspecies, withCarpo- brotusrevealing particularlylow habitatapparency. Furthermore,thedegreeofinvasionofAilanthusis highestonCreteandlowestonCorsica,whereas for Carpobotusitishighest onCorsicaandloweston Sardinia, whileforOxalisitishighestonMallorca and lowest on Corsica. Thus, at the level of an island,thedegreeofinvasion isdependentonthe invader. This indicates that island level attributes that are knownto influencealien speciesrichness suchas area, humanpopulationsize, GDPand/or leveloftrade(Hulme2009b)maybelessimportant inthedegreeofinvasionofindividualspecies.On the other hand, the comparison across species enablesthethreeinvaderstoberankedfromhighest to lowest degree of invasion as: Oxalis[Ailan- thus[Carpobrotus. Thismirrorstherankinggen- eratedbyaninvasionsuccess indexthatcombines both regional and local abundance (Hulme et al.
2008).Previousstudiesonthesethreeinvaders have alreadydemonstratedthatestablishmentandseedling survival werehigher forOxalisandlowerfor CarpobrotusandespeciallyforAilanthusinabroad rangeofdifferenthabitattypes(Vila`etal.2008). In contrast,clonalperformance washigherforAilan- thusinalargenumber ofhabitattypes,and CarpobrotusandOxalis spread onlyincoastaland agricultural/disturbedhabitatsrespectively(Traveset etal.2008).However,speciesrankingscanbealso affected by propagule pressure (e.g., Ross et al.
2008), minimum residence time (e.g., Pysˇek and
Jaros´ık 2005),andnotnecessarily bypopulation performanceofeachtaxon.
Conclusions
Thisstudy shedslightintowhyislandstendto accumulatealieninvaderswithoutaconcomitantloss ofnativespecies (SaxandGaines 2008).First,alien species, evensomeoftheworstinvaders,arelargely restrictedtoanthropogenic habitatswherelevelsof native plant endemism are often low (Vila` and Mun˜oz1999).Nevertheless,nativeruderalannuals thatevolvedintheMediterranean (Blondeland Aronson 1999) occur in varying associations in fields, pasturesandonroadsides:habitatstypically invadedbyalienplantspecies. Manyofthesenative colonisershaverestricteddistributions andcould represent the elements of the Mediterranean flora mostatriskfrominvasions (Hulmeetal.2008). Second, evenwherenativehabitats arecolonised, saturationislowandpatternsofabundancearehighly variableindicatingthatnativespecies maybe threatenedinsomelocalitiesbutnotinalltheirarea ofdistribution evenifaliensarepresent.However, some caution is required when interpreting the current patterns of distribution and abundance of thesethreeinvaders,particularlyduetotheabsence ofhistoricaldataontherateofspread. Therelatively recentanthropogenic changesinMediterranean islands(Hulme2004)mayacceleratethespread of theseinvaders,alteringlocalabundanceandsatura- tion.Similarly, futureclimatechangemayleadtoa widerrangeofhabitats threatenedbyalienspecies (e.g.,Grittietal.2006).Thusanystatementofrisk mustalsolooktowards futuretrendsinthedynamics ofthesetaxa.
Acknowledgments We thank S. Pacciardi, E. Vidal, F.Me´dail,I.Gimeno,M.C.DelaBandera,andE.Moragues forfieldassistance,andtwoanonymousreviewersforcomments onanearlierversionofthemanuscript.Thisstudyispartof EPIDEMIE (ExoticPlantInvasions: DeleteriousEffectson MediterraneanIslandEcosystems)aresearchprojectsupported bytheEuropeanCommission underthe5thFramework, contributing totheimplementationofKeyAction2.2.1 (Ecosystem Vulnerability) within the Energy, Environment andSustainableDevelopmentthematicprogramme(Contract no.EVK2-CT-2000-00074).Furtherdetailsoftheprojectand thedatabankcanbefoundat This workisalsosupportedbyINVABIO(Ministe`redel’Ecologieet duDe´veloppementDurable,subventionno.01113),theNational ParkofPort-Cros(contractno.97.029.83400), theNational ScienceFoundationGraduateResearchFellowshiptoC.M.S. andtheSpanishMinistryofScienceandInnovationprojects MONTESandRIXFUTURto M.V.
References
Andre´sC,OjedaF(2002)Effectsofafforestationwithpineson woodyplantdiversityofMediterranean heathlandsin southernSpain.BiodiversConserv11:1511–1520
BasnouC,Vila`M(2009)Ailanthus altissima(Mill.)Swingle, tree of heaven (Simaroubaceae, Magnoliophyta). In:
DAISIE (ed) Handbook of alien species in Europe. Springer,Dordrecht,p342
BlondelJ,AronsonJ(1999) Biologyandwildlife ofthe mediterraneanregion.OxfordUniversityPress,Oxford
BothamMS, RotheryP,HulmePE, HillMO, PrestonCD, Roy DB(2009)Dourbanareasactasfociforthespreadof alienplantspecies?Anassessment oftemporaltrendsin theUK.DiversDistrib15:338–345
BourgeoisK,SuehsCM,VidalE,Me´dailF(2005)Invasional
meltdownpotential:facilitation betweenintroducedplants and mammals on French Mediterranean islands. E´co-
science12:248–256
CadotteMW,Lovett-DoustJ(2001)Ecologicalandtaxonomic differences between native and introduced plants of
southwesternOntario.E´coscience8:230–238
Chytry´ M, MaskelLC, PinoJ, PysekP, Vila` M, FontX, SmartSM(2008a)Habitatinvasionsbyalienplants:a
quantitative comparison between Mediterranean, sub-
continentalandoceanicregionsofEurope.JApplEcol
45:448–458
Chytry´M,Jarosˇ´ıkM,Pysˇek P,Ha´jekO,Knollova´O,Tichy´L, DanihelkaJ (2008b) Separatinghabitat invasibility by
alienplantsfromtheactuallevelofinvasion.Ecology
89:1541–1553
Chytry´M, PysˇekP,WildJ,PinoJ,MaskellLC,Vila`M(2009) European map of alien plant invasions based on the
quantitative assessment across habitats. Divers Distrib
15:98–107
CodyML(1986)Diversity,rarity,andconservation inMedi- terranean-climateregions. In:SouleME(ed)Conserva- tion biology: the science of scarcity and diversity. Sinauer,Sunderland,pp122–152
CourchampF,ChapuisJ-L,PascalM(2003)Mammalinvaders onislands:impact,controlandcontrolimpact.BiolRev
78:347–383
D’Antonio CM,OdionDC,TylerCM(1993)Invasion of maritimechaparralbytheintroduced succulentCarpo- brotusedulis.Oecologia95:14–21
DAISIE(2009)Handbookofalien speciesinEurope.Springer, Dordrecht
DeFeo V,DeMartinoL,QuarantaE,Pizza C(2003)Isolation
ofphytotoxiccompoundsfromTree-of-Heaven(Ailan- thusaltissima Swingle).JAgricFoodChem51:1177–
1180
DelipetrouP(2009)Carpobrotusedulis(L.)N.E.Br,freeway iceplant(Aizoaceae,Magnoliophyta). In:DAISIE(ed) HandbookofalienspeciesinEurope.Springer,Dordrecht, p345
DeutschewitzK,Lausch A,Ku¨hnI,KlotzS(2003)Nativeand
alienplantspeciesrichnessinrelationtospatialhetero- geneityonaregionalscaleinGermany. GlobEcolBio- geogr12:299–311
EltonC(1958)Theecologyofinvasionsbyanimalsandplants.
Methuen,London
Gimeno I,Vila`M,HulmePE(2006) Areislands moresus- ceptibletoplantinvasionthancontinents?Atestusing
Oxalispes-caprae inthewesternMediterranean.JBio- geogr33:1559–1565
GrittiES,SmithB,SykesMT(2006)Vulnerability ofMedi- terraneanbasinecosystemstoclimatechangeandinva- sionbyexoticplantspecies.JBiogeogr33:145–157
HeiseyRM,HeiseyTK(2003)Herbicidal effectsunderfield conditionsofAilanthus altissimabarkextract,which containsailanthone.PlantSoil256:85–99
HeywoodVH(1995)TheMediterraneanflorainthecontextof worldbiodiversity.EcologiaMediterranea21:11–18
HulmePE(2004)Islands,invasionsandimpacts:aMediter- raneanperspective.In:Ferna´ndez-Palacios JM,MoriciC (eds) Ecologia insular/island ecology, Associacio`n Es- panoladeEcologiaTerrestre(AEET)CabildoInsularde laPalma,Spain,pp359–383
HulmePE(2009a)Relativerolesoflife-form, landuseand climateinrecentdynamicsofalienplantdistributions in theBritishIsles.WeedRes49:19–28
HulmePE (2009b)Trade, transportand trouble: managing
invasive species pathways in an era of globalisation. JApplEcol46:10–18
HulmePE, BrunduG, CamardaI,DaliasP,LambdonP,Lloret F,Me´dailF,Moragues E,SuehsCM,TravesetA, TroumbisA,Vila`M(2008)AssessingtheriskstoMed- iterraneanislandsecosystemsfromalienplantintroduc- tions.In:Tokarska-Guzik B,BrockJH,BrunduG,Child L,DaehlerCC,PysˇekP(eds)Plantinvasions: human perception,ecologicalimpactsandmanagement.Back-
huysPublishers,Leiden,pp39–56
KowarikI(1995)ClonalgrowthinAilanthusaltissimaona naturalsiteinWestVirginia.JVegSci6:853–856
LambdonPW(2009)Oxalispes-capraeL.Bermudabuttercup (Oxalidaceae,Magnoliophyta).In:DAISIE(ed)Handbook ofalienspeciesinEurope.Springer,Dordrecht,p353
Lambdon PW,LloretF,HulmePE(2008)Doalienplantson Mediterraneanislandstendtoinvadedifferentnichesfrom nativespecies?BiolInvasions10:703–716
LloretF,Me´dailF,Brundu G,HulmePE(2004) Localand regionalabundance ofexoticplantspeciesonMediterra- neanislands: arespeciestraitsimportant?GlobEcol Biogeogr13:37–45
Loope LL, Mueller-Dombois D (1989) Characteristics of
invaded islands, withspecialreferencetoHawaii. In: DrakeJA,MooneyHA,diCastriF,GrovesRH,KrugerFJ, RejmanekM,WilliamsonM(eds)Biologicalinvasions:a globalperspective.Wiley,Chichester,pp257–280
MacDonaldIAW,CooperJ(1995)Insularlessonsforglobal biodiversity conservation with particular reference to alieninvasions. In:Vitousek PM,LoopeLL,AdsersenH (eds) Islands:biologicaldiversityandecologicalfunction. Springer,Berlin,pp189–203
ManlyBFJ(1997)Randomization, bootstrapandMonteCarlo methodsinbiology,2ndedn.ChapmanHall,London
Marvier M, Kareiva PM, Neubert MG (2004) Habitat
destruction,fragmentation, anddisturbancepromote invasionbyhabitatgeneralists inamultispeciesmeta- population.RiskAnal24:869–878
Me´dailF,Que´zelP(1997)Hot-spots analysisforconservation
of plant biodiversityin the MediterraneanBasin.Ann
MissouriBotGard84:112–127
O’FarrellPJ, MiltonSJ (2006)Roadvergeandrangelandplant communities inthesouthernKaroo:exploringwhat influences diversity,dominanceandcover.Biodivers Conserv15(3):921–938
Ohlemu¨llerR,WalkerS,WilsonJB(2006)Localvsregional
factorsasdeterminantsoftheinvasibilityofindigenous forestfragmentsbyalienplantspecies.Oikos112:493–501
Pino J,FontX,Carbo´J,Jove´M,Pallare`sL(2005)Large-scale
correlatesof alien plant invasionin Catalonia (NE of
Spain).BiolConserv122:339–350
Planty-TabacchiA-M,TabacchiE,NaimanRJ,DeferrariC, DecampsH(1996)Invasibilityofspecies-rich commu- nitiesinriparianzones.ConservBiol10:598–607
Pu¨tz N (1994) Vegetative spreading of Oxalis pes-caprae
(Oxalidaceae).PlantSystEvol191:57–67
PysˇekP,Jaros´ık V(2005)Residencetimedeterminesthe distribution ofalienplants.In:InderjitS(ed)Invasive plants: ecologicalandagriculturalaspects. Birkha¨user Verlag,Switzerland,pp77–96
Pysˇek P, Pysˇek A (1995) Invasion by Heracleum mante-
gazzianumindifferenthabitatsintheCzechRepublic. JVegSci6:711–718
PysˇekP,BacherS,Chytry´ M, Jarosˇ´ıkV, Wild J,Celesti-
GrapowL,GassoN,KenisM,LambdonPW,NentwigW, PerglJ,Roques A,Sa´dlo J,SolarzW,Vila`M,Hulme PE (inpress)ContrastingpatternsintheinvasionsofEuro- peanterrestrialandfreshwater habitatsbyalienplants, insectsandvertebrates.GlobEcolBiogeogr
RDevelopment CoreTeam(2008)R:alanguageandenvi- ronmentforstatisticalcomputing. RFoundationforSta- tisticalComputing,Vienna.ISBN3-900051-07-0, URL
RichT,RabaneM,FashamM,McMeechan F,DobsonD (2005)Groundandshrubvegetation.In:HillD,Fasham
M,TuckerG,ShewryM,ShawP(eds)Handbook of biodiversitymethods—survey,evaluation andmonitoring. CambridgeUniversityPress,Cambridge,pp201–221
RossLC,LambdonPW,HulmePE(2008)Disentangling the rolesofclimate,propagule pressureandlanduseonthe currentandpotentialelevationaldistribution oftheinva- siveweedOxalispes-capraeL.onCrete.PerspectPlant EcolEvolSyst10:251–258
SaxDF,GainesSD(2008)Speciesinvasionsandextinction:
the futureof native biodiversityon islands.ProcNatl