ARTICLE IN PRESS

BasicandAppliedEcologyxxx(2016)xxx–xxx

Contrasting

predation

services

of

predator

and

omnivore

diversity

mediated

by

invasive

ants

in

a

tropical

agroecosystem

Maxime

Jacquot

,

Philippe

Tixier

,

Olivier

Flores

,

David

Muru

,

Franc¸ois

Massol

,

Brice

Derepas

,

Frédéric

Chiroleu

,

Jean-Philippe

Deguine

a_{CIRAD,UMRPVBMT,F-97410Saint-Pierre,Réunion,France}

b_{UniversitédeLaRéunion,UMRPVBMT,F-97410Saint-Pierre,Réunion,France} c_{CIRAD,UPRGECO,F-34398MontpellierCedex5,France}

d_{CATIE,DepartamentodeAgriculturayAgroforesteria,7170,Cartago,Turrialba30501,CostaRica} e_{CNRS,UniversitédeLille–SciencesetTechnologies,UMR8198Evo-Eco-Paleo,SPICIGroup,F-59655}

Villeneuved’Ascq,France

Received30April2016;accepted13September2016

Abstract

Invasivenaturalenemiesareknowntoeitherstrengthenorweakenthesuppressionofherbivorousarthropods.However,the impactofinvasivespeciesonthepredationserviceprovidedbynaturalenemydiversityremainslargelyunexplored.Here,we teaseaparttherolesofinvasiveantsasprovidersofapredationserviceandapotentialdisservice,i.e.reducingthediversityof naturalenemies.InmangoorchardsonReunionIsland,weevaluatedthepredationservicein20openfieldsbysimultaneously monitoringthepredationonbaiteggsandarthropodcommunitiesintwostrata:thegroundsurfaceandthemangotreecanopy. Ourresultsshowthatthepredationonbaiteggswaslimitedtothegroundsurface.Thisstratumisdominatedbythreeinvasive omnivorousants:PheidolemegacephalaandSolenopsisgeminatastronglyincreasedthepredationrateofbaiteggs,whereas

Brachymyrmexcordemoyiwasresponsibleforonlyasmall decreaseinpredationrate.Predationrate waspositivelyrelated

topredatorspeciesrichness,andwasnegativelyrelatedtoomnivorespeciesrichness.Thenegativerelationshipbetweenthe predationrate andomnivorespeciesrichnessiscausedbythe mostdominant invasiveant,P.megacephala, whichreduces omnivorerichnessandseemstostronglypreyoneggs.Thisstudydemonstrates,forthefirsttime,thedistinctinfluenceofthe diversityoftwotrophicgroupsonthepredationserviceandhowtheseeffectscanbemediatedbyinvasiveantspecies.

Zusammenfassung

InvasivenatürlicheGegenspielerkönnendieKontrollevonherbivorenArthropodenstärken,aberauchschwächen.Der Ein-flussderinvasivenArtenaufdieKontrollleistungderVielfaltdernatürlichenGegenspieleristindessenweitgehendunerforscht. HieranalysierenwirdieRollevoninvasivenAmeisenalsErbringereinerKontrollleistungaberauchalspotentielleVerursacher einesSchadensdurchReduzierungderDiversitätdernatürlichenGegenspieler.InMangoplantagenaufLaRéunionbestimmten wirdiepotentielleKontrollleistungauf20Freiflächen,indemwirparalleldieAbnahmevonausgelegtenFliegeneiernunddie lokalenArthropodengemeinschaftenerfassten.DiesgeschahsowohlaufderBodenoberfläche alsauch indenBaumkronen.

∗_{Correspondingauthorat:CIRAD,UMRPVBMT,F-97410Saint-Pierre,Réunion,France.Fax:+262262499293.}

E-mailaddress:jacquot.maxime.a@gmail.com(M.Jacquot).

http://dx.doi.org/10.1016/j.baae.2016.09.005

UnsereErgebnissezeigen,dassdieVerlusteanKödereiernnuramBodenauftraten.HierdominiertendreiinvasiveAmeisenarten:

PheidolemegacephalaundSolenopsisgeminataerhöhtendieAbnahmederKödereierstark,währendBrachymyrmexcordemoyi

eine geringe Abnahme der Prädationsrate verursachte. DiePrädationsrate war positiv mit dem Artenreichtum der Räuber korreliertundnegativmitdemderOmnivoren.LetzteresgehtaufdieamstärkstendominanteinvasiveAmeise,P.megacephala,

zurück,diedieArtenzahlderOmnivorenstarkreduziertundkräftigvondenKödereiernzufressenscheint.DieseUntersuchung zeigtdenklarenEinflussderDiversitätvonzweitrophischenGruppenaufdieKontrollleistungundwiedieseEffektedurch invasiveAmeisenartenbeeinflusstwerdenkönnen.

©2016Gesellschaftf¨ur ¨Okologie.PublishedbyElsevierGmbH.Allrightsreserved.

Keywords: Agroecosystem;Biologicalcontrol;Biologicalinvasion;Ecosystemservice;Mango;ReunionIsland

Introduction

Infoodwebs,invasivenaturalenemiescanstrengthenor weakenthesuppressionofherbivorousarthropods,providing aservicebyfeedingdirectlyonkeyherbivoresand/ora dis-servicebynegativelyinteractingwithkeynaturalenemiesof herbivores(Snyder&Evans2006;Crowder&Snyder2010). Ininvadedareas,invasivenaturalenemiesnotonlyinteract withkeyspecies,butalsoreducetheabundanceand diver-sityof species(Snyder &Evans2006; Kenis etal.2009). Interferenceandexploitativecompetitionaretwomain mech-anismsunderlyingthesuppressionofpre-establishedspecies, especiallythosewithsimilarnicherequirements(Crowder& Snyder2010).However,fewstudiesfocusonthe simultane-ouseffectsofinvasivenaturalenemiesonherbivorouspests andoncommunitiesof natural enemies.Forinstance,ina Californiavineyard,anexoticspiderincreasedthe suppres-sionofherbivores,whilesimultaneouslyreducingthenumber of nativespiders(Hogg &Daane2011).Invasiveants can alsoprovide services anddisservices, as theyare omnivo-rous(Holway,Lach,Suarez,Tsutsui,&Case2002),interact withmanyspecies,andcansometimescontrolherbivorous pests (Offenberg2015). The lattereffect can be complex. Forinstance,theimportedredfireants(Solenopsisinvicta)

isadirectpredatorofpests,butisalsoinvolvedinintraguild predation, i.e. its local abundanceis negatively correlated withthatof naturalenemiesofherbivorespecies(Eubanks 2001).In amanipulative experiment,the introductionof a dominantinvasiveantwasassociated withreducedspecies richnessandevennessoftheantcommunity,andreducedcrop yield,comparedwithcontrolcommunitiesorant communi-tiesdominatedbyanativespecies(Wielgossetal.2014).The dominantinvasiveprovidedthisdisservicebyincreasing phy-topathogendisseminationandreducingtop–downcontrolof twomajorpests(Wielgossetal.2014).Apartfromthesefew seminalstudies,theimpactofinvasivearthropodsonthe rela-tionshipbetweendiversityofnaturalenemiesandpredation serviceremainslargelyunexplored.

Biological control of herbivorouspests by the diversity of natural enemiesisacrucial servicetoensure crop pro-tection.Meta-analyses showedthat the diversityof natural enemieshasagenerallypositiveeffectonthesuppressionof

herbivorous arthropods (Straub, Finke, & Snyder 2008;

Letourneau,Jedlicka,Bothwell,&Moreno2009)andmore generally on the suppression of preys (Griffin, Byrnes, & Cardinale2013),butthiseffectcanalsobeneutralor neg-ative. Griffin et al. (2013) reported that diverse predator communitiesarenotmoreefficientinsuppressingpreythan their single most efficientspecies. The variable effects of natural enemydiversityonherbivorescanbeexplainedby three mechanisms.Firstly,negativeenemy–enemy interac-tions,suchasmutualintraguildpredation(Vance-Chalcraft, Rosenheim,Vonesh,&Craig2012)andinterference compe-tition (Schmitz, 2007), can reduce pest control. Secondly, higher species diversity can maintain a wider variation of phenotypic traits among enemies. This trait variability canenhancethecollectiveperformanceof naturalenemies throughacomplementarityeffect,orpromotethedominance ofspecieswithextremetraitvalues,suchasspeciesableto controlparticularherbivorespecies(selectioneffect)(Loreau 2000). Thirdly,interspecificfacilitation of thesuppression of herbivores may occur when a predator guild modifies preybehaviour,therebyfacilitatingcapturebyanother preda-tor guild (Losey & Denno 1998, 1999). This variety of mechanismssuggeststhatintegrativestudiesareneededto understandtherolesof differentnativeandinvasive exotic guildsinagivenagroecosystem.

The aims of the present work are: (1) to quantify the relationships between biodiversity of natural enemies and predation service inoursystem;and(2) toassessthe role ofinvasivenaturalenemiesonthisbiodiversity-service rela-tionship.OurstudywasconductedonReunionIsland(Indian Ocean), wherewidespread invasive ant species are found (Blard, Dorow, & Delabie 2003). Invasive ant species are the dominant naturalenemies inour study.Thestudy sys-temismangoorchards(MangiferaindicaL.,Anacardiaceae), mangosbeingoneofthemaintropicalfruitcropsworldwide (FAO2015).Amongseveralinsectpeststhatcausedamage tomangoinflorescences(Amouroux&Normand2013),the mangoblossomgallmidge,Procontariniamangiferae(Felt), andvariousthripsspeciesspendapartoftheirlifecyclein the soil(Lewis 1997; Amouroux, Normand, Nibouche,& Delatte2013).Itisthereforeassumedthatthepredation ser-vice occursintwostrata:onthe cropwherepestsfeedon

inflorescencesandon theground surface where, for some species,larvaeburyandadultsemerge.

Tounderstandtheeffectsofinvasiveantsonthe relation-shipbetweenbiodiversityandpredationservice,weevaluated the predation service in20 open fields by simultaneously monitoringthepredationrateofbaiteggsandthearthropod communitiesinthetwostrata.Wethenbuiltahierarchical Bayesianmodeltoanalyse(1)theeffectsoftheabundance ofeachdominantinvasiveantspecies,predatorrichnessand omnivorerichnessonthepredationservice,and(2)theeffect oftheabundanceofdominantinvasiveantsontheomnivore andpredatorrichness,respectively.

Materials

and

methods

Study

area

and

sampling

The study was conducted in 10 mango orchards (here-after “sites”) onthe west coastof Reunion Island(Indian Ocean) belonging to the BIOPHYTO project network (www.biophyto.org).Wemonitored two plots ateach site. In each site, there were two different farming practices betweenthetwoplots;inoneplot,customarypracticeswere used(organicorconventional),whileconservation biologi-calcontrol practiceswereusedintheother(groundcover, insecticide-andherbicide-free).Farmingpracticesalso var-iedacrosssites,resultinginavarietyofecologicalsituations atplot level.The averagesize of plots was1404m2,with anaverageintra-sitedistanceof89.3m,andanaverage inter-sitedistanceof46km.ThefielddatawerecollectedinAugust 2014duringthemangofloweringseason.

Predation

rate

Weusedabait-removalexperimenttoassessthegeneral predationpotentialinoursystem,bothonthegroundsurface andinthemangotreecanopy.Observationsofbaitremoval arecommonlyusedtoprovideastandardizedmeasurementof predationpressureinecosystems(Aikens,Timms,&Buddle 2013;Monteiroetal.2013;Marliacetal.2015).Abaitwasa cardofgreensandpaperwitheightdeadeggsofthe Cucur-bitfruitfly(Zeugodacuscucurbitae(Coquillett))gluedonit. Wedidnotuseactualpestsofmangoinflorescencesinorder toprevent any outbreak inmonitored farms. Whilebeing slightlydifferentfrompesteggs,webelievethatbaitsarea goodindicatorofthepredationpotentialsinceeggsoffruit fliesaresoft,andcanbe assumedtobepalatableformost predatorandomnivorespecies.Theseeggswerechosenfor theirsize,similartolarvaeandpupaeofmangoblossomgall midgeandthripsspecies.Cucurbitfruitflieswererearedon anartificialdiet(CIRAD,Saint-Pierre,ReunionIsland),eggs were harvested and immediately frozen at −20◦C. Three hourslater,baiteggunitswerepreparedbygluingeighteggs ona4cm×5cmcardofgreensandpaper.Thecardswere

keptfrozenuntiltheywereplacedintheplots,eggsthaws withinminutesafterplacement.Ineachplot,fivecardswere placedineachstratumonthe13thofAugust,2014(atotalof 100eggcardsperstratum).Thenumberofbaiteggsremoved ordamagedwasrecorded24hafterplacementinthefield.

Arthropod

sampling

Arthropods weresampled inthreedifferent ways. First, arthropods on the ground surface were sampled with pit-falltraps.Thetraps(diameter:12cm;depth:11.5cm)were filled with 250ml of a mixture of 1:1 glycerine and salt water(25%).Glasscovers(diameter:25cm)werepositioned approximately 20cm above the traps to keep of the rain. Eightpitfalltrapswereplacedineachplotandleftinplace for oneweek,betweenthe12thand19thofAugust,2014. Second, we used suction sampling (modified leaf blower, STIHL BG56, withan ovalnozzle: 14.5×10cm) to sam-ple the ground surface (with or without weeds). Suction samplesweretakenalongtransects.Transectswerelocated betweentwo rowsof mangotrees andwere perpendicular to them,the length of transects was equal to the distance betweenrows (6.8±1.0m). The totaldistancesampledin eachplotwasproportionaltothesizeoftheplot.Onaverage, 25.3±1.1mweresampledper1000m2,whilethenumberof transectsdependedontheinter-rowdistance.Foreach tran-sect, thenumberof individuals perspecies wascalculated permetre.Third,arthropodswhosehabitatisthemangotree werecollectedusingsuctionsampling.Oneachtree,a sam-ple corresponded tosuction of 1m2 of canopy ateach of thefourcardinalpoints. Thenumberof sampledtreeswas proportionaltotheplotarea,onaverage3.75±0.2treesper 1000m2.Because of the distance between sites, we were onlyable tosample four sitesaday sosuction samplings wereconductedbetweenthe20thand22ndofAugust,2014. Allarthropodscollectedwereidentifiedtothespecieslevel (morphospeciesormorphotypes).Lastly,weassigned arthro-pod species to trophic groupsaccording to data from the literature(seeAppendixAofSupplementarymaterial:Tables 1 and2for detailsonpredatorsandonomnivores).If lar-vaeandadultshavedifferentfeedingecology,weconsidered theecologyofthe stagesampled.Inthisstudy,we consid-eredtwotrophicgroups:predators,includingstrictpredators, whichfeedonlyonotherarthropods(ofanytrophiclevel), andomnivores,whichfeedonotherarthropodsandonplants (includingnectarandpollen)and/ordetritus.

Community

metrics

Numerical dominance occurs when a species reaches greater abundanceand/or biomass at baits and/or in traps thanotherco-occurringspecies(Parr&Gibb2010).In the presentcase,omnivorousspecieswereclassifiedas numer-ically dominant when they representedmore than50% of all omnivorousspecies inmorethan10% of samples (see

AppendixAofSupplementarymaterial:Fig.1).We identi-fieddominanceintwocompartments:(1)thegroundsurface withsamplestakenfrompitfalltrapsandcollectedbysuction sampling;(2)themangotreecanopywithsamplescollected bysuctionsamplingonly.

Foreachplot,thecommunitymetricsfortheground sur-facewerecalculatedusing afulldataset composedof data frompitfalltrapsanddatafromsuctionsampling.Themethod consisted in randomly selecting two samples for each of the two sampling techniquesto form asingle subset with both types of data. This process was repeated 100 times (permutations) to form 100 subsets. For each subset, we evaluated(1) species richness withthe second-order jack-knifeestimator,using the specpool functionof R package

vegan(Oksanenetal.2015)and(2)abundance,asthesum

ofallarthropodsinthesubset.Finallywecalculatedthe aver-agemetricsforthe100subsets:speciesrichnessoftrophic groupsandabundanceofdominantantspecies.Exploratory trialsdemonstratedthattheindicatorvaluesstabilisedbefore 100permutations.TheRcodeforthisprocedureisprovided inAppendixBofSupplementarymaterial:1.The commu-nity metricsinthe mangotree canopy were notestimated becauseofthenearlytotalabsenceofpredationofbaiteggs inthemangotreecanopy(seeAppendixAofSupplementary material:Fig.2).

Model

We built a hierarchical Bayesian model (HBM, Wikle 2003)tosimultaneouslytesttheeffectsoftheabundanceof dominantinvasiveants,predatorrichnessandomnivore rich-nessonthepredationservice,andtheabundanceofdominant invasiveantsonomnivoreandpredatorrichness.Briefly,the datawasstructuredasfollows:ateachofthetensites(noted withsubscriptkinequations),we monitoredtwo plots(j), ineachofwhichweobtaineddataonpredationconcerning the fivebait eggs cards(i),oneestimateof abundancefor eachdominantinvasiveantspeciesandoneestimatedvalue ofspecies richnessforeach trophicgroup– theselasttwo estimatescomingfromthecombinationofbothsuctionand pitfalltrapsampling.

AtthefirstleveloftheHBM,wemodelledtheprobability (pjk)of a bait egg ona sample cardbeing eatenwithin a

givenplot(indexj)withinagivensite(indexk)asafunction ofpredatorspeciesrichness(Pjk),omnivorespeciesrichness

(Ojk)andtheabundanceofeachdominantinvasiveant(Bcjk:

Brachymyrmexcordemoyi;Sgjk:Solenopsisgeminata;Pmjk:

Pheidolemegacephala),following:

logit(pjk)=α1+α2Pjk+α3Ojk+α4Bcjk

+α5Pmjk+α6Sgjk+ξ1k (1a) whereα1isaninterceptandα2–α6aretheeffectsofpredator speciesrichness,omnivorespeciesrichnessandabundance ofthethreedominantinvasiveantspeciesrespectively.ξ1is arandomsiteeffectaccountingforvariationamongsitesand

definedasanormallydistributederror∼N(0,σ1).Weuseda binomialdistributiontolinkthisprobabilitytotheobserved numberofeatenbaiteggs(yijk)amongtheeightbaiteggson

eachcard(indexi): yijk∼Binomial

 8,pjk



(1b) At the second level of the model,we hypothesized that omnivorespeciesrichness(Ojk)andpredatorspecies

rich-ness(Pjk)withinagivenplot(indexj)within agivensite

(indexk)arenormallydistributedobservations,themeansof theirdistributionsbeingmodelledaslinearcombinationsof theabundanceofdominantinvasiveants:

Ojk∼Normal  μOjk,σ2O  (2a) μOjk =β1+β2Bcjk+β3Pmjk+β4Sgjk+ξ2k (2b) Pjk∼Normal  μPjk,σ2P  (3a) μPjk=γ1+γ2Bcjk+γ3Pmjk+γ4Sgjk+ξ3k (3b)

whereμ andσ are themeans andstandarddeviationsand

β1 andγ1 are intercepts,β2–β4 and γ2–γ4 are regression coefficientsmeasuringtheeffects ofdominant invasiveant abundanceonomnivorespeciesrichnessandpredatorspecies richness,respectively.ξ2 andξ3aresiteeffectsaccounting for randomvariation amongsitesanddefinedas normally distributederrors,withazeromeanandestimatedstandard deviationsofσ2andσ3,respectively.Weassigned uninfor-mative priors toall model parameters(i.e., α1–α6, β1–β4,

γ1–γ4andξ1–ξ3):∼N(0,103).

TheMetropolis–Hastingsalgorithmwasusedtosimulate theposteriordistributionsofthemodelparameters.Weused JAGSsoftware(Plummer2003)toperformGibbssampling and the functionjags inR package R2jags (Su& Yajima 2012)asaninterfacetoR.ModelswererunwithfiveMCMC chains of 40,000 iterations and with a burn-in period of 20,000runs.TomonitortheconvergenceofMCMCchains, chaintraceswereinspectedvisually,andwecheckedthatthe potentialscalereductionfactorwasbetween1.0and1.1for eachparameter(Gelman&Shirley2011).Datausedtotest theabovementionedeffectsispresentedinAppendixAof Supplementarymaterial:Fig.3.Allthevariableswere stan-dardisedsothattheeffectscouldbecompareddirectlyusing standardised“path”(regression)coefficients.Antabundance valueswerelog-transformed.

Weconductedmodelaveragingforeachequation.Firstly, weranallpossiblemodelsnestedwithinthethreefullmodels (Eqs.(1)–(3))withallpossiblecombinationsoffixedeffects (namely abundance of each dominant invasive ant, omni-vorerichnessandpredatorrichness)consideredasexplicative variablesforthegivenequation.Thepossiblecombinations of fixed effects corresponded to equations with 0–5 fixed effects for Eq. (1), or with 0–3 fixed effect for Eqs. (2) and(3),whichresultsin48testedmodels.Foreachmodel,

we obtained the devianceandthe posterior distributionof parameters.Secondly,wecalculatedtheBayesian informa-tioncriterion(BIC)andBICweightsofeachmodel.Thirdly, wecalculatedtheposteriormeanand95%credibilityinterval (CI)foreachparameter,basedontheoverallposterior distri-butionofparameter,weightedbytherespectiveBICweight oftheoriginalmodels.Aparameterisconsideredtobe sig-nificantlydifferentfromzeroifitsweighted95%CIdoesnot overlapzero.Wealsosummarisedmodelfitateachlevelwith aBayesianR2 (Gelman&Hill2007),basedontheoverall posteriordistributionof theresponsevariableweightedby modelBICweight.

WeusedR3.1.0(RDevelopmentCoreTeam2014)and JAGS 3.4.0 for all our analyses. The JAGS code for our completemodelisgiveninAppendixBof Supplementary material:2.

Results

Predation

occurs

at

ground

level

Amongallsites,predationofbaiteggsontheground sur-facewasobservedon53outof100cardsinitiallysetup.The predationrateofbaiteggsinthemangocanopystratumwas non-zeroononlysevencardsoutof100(seeAppendixAof Supplementarymaterial:Fig.2).Intherestofthestudy,we consequentlyfocusedonpredationatgroundlevelbecause predationinthetreecanopywasvirtuallyundetected.

Communities

are

dominated

by

three

ant

species

Overall, we sampled and identified 42,485 arthropods, including 63 predator species (1654 individuals) and 22 omnivorousspecies(19,783 individuals).Spiderswerethe mostdiversegroup amongall predators, with52different species sampled(Appendix A of Supplementary material: Table 1). Omnivores were mainly composed of insects, including 16 ant species (Appendix A of Supplementary material:Table2).Themeanspeciesrichnessofomnivores andpredatorswaslowerinsamplesobtainedbysuctiononthe mangotreecanopythaninsamplesobtainedbysuctionfrom thegroundsurface(AppendixAofSupplementarymaterial: Fig.4).Mean abundanceof bothomnivores andpredators waslowerinsamplesobtainedbysuctiononthemangotree canopythaninpitfalltraps(AppendixAofSupplementary material:Fig.4).

On the ground surface, B. cordemoyi (Forel), P.

mega-cephala (Fabricius) and S. geminata (Fabricius) were

identifiedasthenumericallydominantspecies(seeAppendix AofSupplementarymaterial:Fig.1).Estimatedabundances of the three dominant ant species among the 10 sites are shownin Appendix A of Supplementary material: Fig.5. Amongdominantants,P.megacephalawasthemost abun-dantspecies,andrepresentedthenumericalmajorityin44%

Fig.1. PathrepresentationofthehierarchicalBayesianmodel pre-dictingthepredationrateofbaiteggsperplotgiventheabundanceof eachdominantinvasiveant,omnivorespeciesrichness,andpredator speciesrichness(Eqs.(1)–(3)).Significantandnon-significant path-ways(with95%CI)areincolourandingrey,respectively.Green and redlinesshowpositiveand negativesignificantinteractions, respectively.Thenumbernexttothearrowsisthemodel-averaged standardised mean of the parameters. Model-averaged posterior meanand95%credibilityinterval(CI)forallparametersasreported inTable1.ModelselectiontablereportedinAppendixCof Sup-plementarymaterial:Table1.

ofpitfalltrapsand31%ofthegroundsurfacesuction sam-ples.Inthemangotreecanopy,weconsideredthattherewas nodominantspecies,becausenospeciesshowednumerical majorityinasufficientnumberofsamples(seeAppendixA ofSupplementarymaterial:Fig.1).

The

three

dominant

omnivore

ants,

predator

and

omnivore

richness

affect

predation

rate

of

bait

eggs

TheresultsforthemodelinEq.(1)showedthatthe abun-danceofP.megacephalaandS.geminatapositivelyaffected the predation rate (95% CI of α5 and α6: 1.28–5.07 and 0.91–3.80,respectively, Fig.1),whereas theabundanceof

B. cordemoyi negatively affected the predation rate (95%

CIof α4:−1.59 to−0.23,Fig.1).Predatorandomnivore speciesrichnesshaddifferenteffectsonbaiteggpredation. Omnivorespeciesrichnesshadanegativeeffectonthe pre-dationrate(95%CIofα3:−2.15to−0.05,Fig.1).Although predator richnesswas notinfluenced by any dominant ant species,itpositivelyaffectedthepredationrate(95%CIof

α2: 1.07–2.37,Fig.1).Site-dependentrandomeffects also affectedpredation(95%CIofσ2:1.04–4.98;Table1),with effectsofthesameorderofmagnitudeasthoseofP.

mega-cephala or S. geminata. The abundanceof dominant ants,

ran-Table1. Model-averagedmeanparameterestimatesandposteriordistributionforthehierarchicalBayesianmodels(Eqs.(1)–(3)).

Parameter Mean 2.5% 50% 97.5%

Eq.(1)

Interceptα1 −0.71 −2.43 −0.71 0.94

Effectofpredatorspeciesrichnessonpredationα2 1.66 1.07 1.64 2.37

Effectofomnivorespeciesrichnessonpredationα3 −1.21 −2.15 −1.32 −0.05

EffectofB.cordemoyiabundanceonpredationα4 −0.87 −1.59 −0.86 −0.23

EffectofP.megacephalaabundanceonpredationα5 3.27 1.28 3.30 5.07

EffectofS.geminataabundanceonpredationα6 2.30 0.91 2.28 3.80

Siteeffectσ1 2.36 1.04 2.14 4.98

Eq.(2)

Interceptβ1 0.00 −0.61 0.00 0.62

EffectofB.cordemoyiabundanceonomnivorespeciesrichnessβ2 0.29 −0.19 0.30 0.74

EffectofP.megacephalaabundanceonomnivorespeciesrichnessβ3 −0.82 −1.77 −0.75 −0.19 EffectofS.geminataabundanceonomnivorespeciesrichnessβ4 −0.38 −1.18 −0.43 0.57

Siteeffectσ2 0.80 0.19 0.75 1.62

Eq.(3)

Interceptγ1 0.00 −0.60 0.00 0.61

EffectofB.cordemoyiabundanceonpredatorspeciesrichnessγ2 −0.08 −0.64 −0.08 0.53 EffectofP.megacephalaabundanceonpredatorspeciesrichnessγ3 −0.32 −0.92 −0.35 0.47

EffectofS.geminataabundanceonpredatorspeciesrichnessγ4 0.26 −0.47 0.29 0.87

Siteeffectσ3 0.55 0.03 0.50 1.43

domeffectsexplained99%ofthevarianceinthepredation rateofbaiteggs.

Omnivore

species

richness

is

negatively

affected

by

only

one

dominant

invasive

ant

species

Modelcalibrationallowedustoestimatetheeffectsof inva-sive ant abundanceon species diversity. Amongthe three dominant invasive ants, only the abundance of P.

mega-cephalahadanegativeeffectonomnivorespeciesrichness

(95%CIofβ3:−1.77to−0.19,Fig.1).Site-dependent ran-domeffects also affectedomnivore species richness(95% CIofσ2:0.19–1.62;Table1).Theabundanceofdominant invasive ants andsite-dependent effects explained 69% of thevarianceofomnivorespeciesrichness.Noneofthethree dominantantspeciesaffectedpredatorspeciesrichness. Site-dependentrandomeffectsaffectedomnivorespeciesrichness (95%CIofσ3:0.03–1.43;Table1).Only19%ofthevariance ofpredatorspeciesrichnesswasexplainedoverall.

Discussion

Dominant

omnivore

ants

affect

their

own

trophic

group

Inoursystem,S.geminatadidnotaffectthespecies rich-ness of predator or omnivore trophic groups, although S.

geminataisknowntoreducethediversityofotherarthropods

inareastowhichithasbeenintroduced(Wetterer2010).This apparentlackofeffectcouldresultfrommutualinterference

with P. megacephala (Wetterer 2010). To our knowledge,

suchaneffecthasneverbeenreportedforB.cordemoyi.Our results also showed that, onthe ground surface,omnivore species richness decreased with an increase in P.

mega-cephalaabundance.P.megacephala,oneoftheworld’s100

mostinvasiveantspecies(Lowe,Browne,Boudjelas,&De Poorter 2000), is known to reduce the diversity of other arthropodsinareaswhereithasbeenintroduced(Kenisetal. 2009;Wetterer2012).

Bycontrast,wefoundthatnodominantantspecies signif-icantlyaffectedpredatorspeciesrichness.Thisisconsistent withthefactthatinvasivearthropodpredatorsandomnivores haveastrongeffectonlyonnativespecieswithsimilarniche requirements (Snyder & Evans 2006; Crowder & Snyder 2010).Inoursystem,invasivenaturalenemieswere omniv-orous and shared more trophic similarity with omnivores thanwithpredators.Thesimilaritybetweendominant inva-siveantsandomnivoreswasincreasedbythefactthatants constituted alargemajorityof omnivores, whilepredators were mainlycomposedofspiders. Therefore,inthe omni-voretrophicgroup,competitionforfoodandnestsitescan explainthereductioninspeciesrichnessbytheomnivorous ant speciesP. megacephala(Holwayetal.2002),whereas differences in niche requirements and foraging strategies between spiderspecies anddominant invasive antsappear toallowcoexistence.

Dominant

ants

have

different

predatory

abilities

Theabundanceofdominantinvasiveantspeciesaffected the predation rate of bait eggs differently. P.megacephala

andS.geminatahadpositiveeffectsonpredationrate,

Javier,&Heong2002;Dejean,Moreau,Uzac,LeBreton,& Kenne2007),whereasB.cordemoyihadanegativeeffecton thepredationrate.Littleisknownaboutthisspecies,which belongstoanomnivorousgenus(Weiser&Kaspari2006). Theobservednegativeeffectmaybeexplainedby(i) interfer-enceduetocompetitionwithotherantsfortheexploitation of bait eggs; (ii) a low harvesting capacity of B. corde-moyi workers (size=2.27±0.16mm), which are smaller

thanP.megacephalaandS.geminataworkers(2.50±0.08

and3.29±0.23mm,respectively);and(iii)alarge propor-tionofherbivoryintheomnivorousdietofthisspecies.

Predator

and

omnivore

species

richness

have

contrasting

effects

on

the

predation

rate

of

bait

eggs

On the ground surface, the predation rate of bait eggs decreased with omnivore species richness. This negative relationshipwas apparently causedby the most dominant invasive ant, P. megacephala, through a reduction in the species richness of their own trophic group and by their intense preying on bait eggs. Our results are consistent withthe correlationbetween food resource discovery and dominance known in ant ecology (Parr & Gibb 2012).

P. megacephala is the fastest explorer and most efficient

exploiteramongfourof themost problematic invasiveant species(notfoundonReunionIsland):Linepithemahumile

(Mayr), Lasius neglectusVan Loon etal. andWasmannia

auropunctata(Roger)(Bertelsmeier,Avril,Blight,Jourdan,

&Courchamp2015).Inaddition,inthisomnivore-poor com-munity,decreasingtherichnessofintraguildpredatorscould enhancebaiteggremoval.Indeed,onepredatormaybemore efficientinpreysuppressionthanmultiplepredatorsifmutual intraguildpredationoccurs(Vance-Chalcraftetal.2012).

Thepredationrateincreasedwithpredatorspeciesrichness inlinewithexistingmeta-analyses(Letourneauetal.2009; Griffinetal.2013).Inourstudy,thepredatortrophicgroupis mainlycomposedofspiders,anorderwellknownforitsrole inpest control(Marc,Canard, &Ysnel1999).Ourresults areinaccordancewiththeliteraturebecausetheabundance ofspidersisknown tobe positivelycorrelatedwith preda-tionrateofbaiteggs(Werling,Harmon,Straub,&Gratton 2012;Mitchell,Bennett,&Gonzalez2014),andsomespiders (including non-web-and web-buildingspiders) are known topreyoneggs(Pfannenstiel2008;Morrison,Mathews,& Leskey2016).Thepredation service providedby predator diversityappeared notto be affectedby the abundanceof dominantinvasiveants.

Predation

occurs

at

ground

level

Our results showed that predation was very low in the mangocanopy.Thisfindingisconsistentwiththeresultsof

Lemessa,Hambäck,andHylander(2015),whofoundthatthe meanpredationrateofcaterpillarsbyarthropodswas1.6% ontheleavesofcoffeeandavocadoshrubs.Inatemperate

forestcanopy,adecreasinggradientofpredationpressurehas beenfoundfromtheunderstory(composedofsaplings)tothe uppercanopy(Aikensetal.2013).Inmangosystems,thelow predationratemeasuredinthecanopycanbeexplainedby the lowabundanceanddiversityofnatural enemiesinthis stratumcomparedwiththegroundsurface(seeAppendixA ofSupplementarymaterial:Fig.4).Thisdifferencein com-munitycompositionbetweenstratahasalsobeenobserved in vineyardsandapple orchards (Frank, Wratten,Sandhu, &Shrewsbury2007;Simon,Defrance,&Sauphanor2007). However,inthecanopyaboveavineyard,predationbykey natural enemies (e.g. earwigs) compensated for the lower predator activity andspecies richness.Such compensating effectwasnotobservedinthestudiedmangoorchards.

Thefactthatpredationwasstrongeronthegroundsurface than in the tree canopy has implications for the biologi-cal control ofpests.Indeed, insectpestslivingonlyinthe mangocanopy or pests immigratingfrom the surrounding landscape will be regulated very weakly. Pests spending their pre-imaginalstageontheground(such asthemango blossom gall midge and thrips species) canonly be regu-lated in the long term by disrupting their life cycle. Our findings suggest that dominant invasive antscan be effec-tivebiocontrol agents. However,theycanalso havemajor negative effects.For example,S. geminata stings are very painful(Wetterer2010)andcanaffectfarmwork.Another effectthatrequiresevaluationistheinteractionbetween inva-sive antsandhoneydew-producinghemipterans(Offenberg 2015),whichcanfacilitatethespreadofbothinsects(Holway etal.2002).

The methods used inthe present study didnot identify whichspeciesdirectlypreyedonbaiteggs.Inthefuture,video recordingsof baitprey couldbeusedtoconfirmpredation by dominant invasive ants and predator species. Molecu-laranalysesofthegutcontentsofpredatorsandomnivores couldalsobeusedtoidentifythespeciesthatpreyonmango pests (Mollotet al.2014).As antshave central-place for-aging strategies(Krushelnycky, Holway,&LeBrun2010), resourceexploitationbyantscanbeinfluencedbydistance tothenests(Detrain&Deneubourg2009).Tobetter under-standpredation inoursystemandtotakeintoaccount the spatialvariabilityofthecommunity,aperspectivecouldbe toincludethedistancetothenestsasapredictorinmodels. Althoughweshowedthateggpredationratedependsmainly oneffectsatthecommunitylevel(naturalenemydiversityand invasive ant abundance), the site-dependent randomeffect was ofthe sameorderof magnitudeas any ofthe consid-eredcommunityeffects.Futureresearchshouldinvestigate whichfarming practices andlandscape features canaffect predatorandomnivorespeciesrichness,andtheabundance ofdominantinvasiveants.

In conclusion, dominant invasive ants affect the rela-tionship between the diversity of natural enemies and the predationservice.P.megacephalaseemedtocausethe neg-ative relationshipbetweenthepredationrate andomnivore speciesrichness,throughareductioninthespeciesrichness

oftheirowntrophicgroupand,hence,inthepredationofbait eggs.Predatorspeciesrichnesswasnotaffectedbydominant invasiveantsandhasapositiverelationshipwithpredation.

Acknowledgements

WethankMickaëlTenailleau,C.AjaguinSoleyen,M.-L. Moutoussamy,C.BaltzerandL.Müllerfortheirassistance withdatacollection.Wethankthefollowingtaxonomistsfor theirassistanceinspeciesidentification:J.-C.Ledouxfor spi-ders,J.-C.StreitoforHeteroptera,O.Levouxformites,and J.Pousserauforbeetles.Wethankthefarmersandpartners oftheBiophytoproject,especiallyC.Gloanec,D.Vincenot and E. Lucas. We greatly acknowledge the Plant Protec-tion Platform (3P,IBISA).We extendourgratitude tothe MinistryofAgriculture,Food,Fisheries,RuralAffairsand Spatial Planning, which funded the Biophyto project via theTrustAccountforAgriculturalandRuralDevelopment (CASDAR).WethanktheRegionalCouncilofReunion,the DepartementalCounciloftheRegionReunion,theEuropean Union(ERDF,EAFRD),andtheCIRAD.

Appendix

A.

Supplementary

data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.baae.2016.09.005.

References

Aikens, K.R.,Timms, L.L.,& Buddle, C.M. (2013). Vertical heterogeneityinpredationpressureinatemperateforestcanopy. PeerJ,1,e138.http://dx.doi.org/10.7717/peerj.138

Amouroux,P.,&Normand,F.(2013).Surveyofmangopestsin ReunionIsland,withafocusonpestsaffectingflowering.Acta Horticulturae,992,459–466.

Amouroux, P., Normand, F., Nibouche, S., & Delatte, H. (2013). Invasive mango blossom gall midge, Procon-tarinia mangiferae (Felt) (Diptera: Cecidomyiidae) in Reunion Island: Ecological plasticity, permanent and struc-tured populations. Biological Invasions, 15(8), 1677–1693. http://dx.doi.org/10.1007/s10530-012-0400-0

Bertelsmeier,C.,Avril,A.,Blight,O.,Jourdan,H.,&Courchamp, F.(2015). Discovery-dominance trade-off amongwidespread invasiveantspecies.EcologyandEvolution,5(13),2673–2683. http://dx.doi.org/10.1002/ece3.1542

Blard,F.,Dorow,W.H.O.,&Delabie,J.H.C.(2003).Lesfourmis del’îledelaRéunion(Hymenoptera,Formicidae).Bulletinde laSociétéEntomologiquedeFrance,108,127–137.

Crowder, D. W., & Snyder, W. E. (2010). Eating their way to thetop?Mechanismsunderlyingthesuccessofinvasiveinsect generalist predators. BiologicalInvasions, 12(9), 2857–2876. http://dx.doi.org/10.1007/s10530-010-9733-8

Dejean, A., Moreau, C. S., Uzac, P., Le Breton, J., & Kenne, M. (2007). The predatory behavior of Pheidole megacephala. Comptes Rendus Biologies, 330(9), 701–709. http://dx.doi.org/10.1016/j.crvi.2007.06.005

Detrain,C.,&Deneubourg,J.-L.(2009).Socialcuesandadaptive foragingstrategiesinants.InS.Jarau,&M.Hrncir(Eds.),Food exploitationbysocialinsects.Florida,USA:CRCPress(p.40).

Eubanks, M. D. (2001). Estimates of the direct and indi-rect effects of red imported fire ants on biological control in field crops. Biological Control, 21(1), 35–43. http://dx.doi.org/10.1006/bcon.2001.0923

Food andAgriculture Organization(2015).FAOSTAT.Retrieved fromhttp://faostat.fao.org/.

Frank,S.D.,Wratten,S.D.,Sandhu,H.S.,&Shrewsbury,P.M. (2007).Videoanalysistodeterminehowhabitatstrataaffects predatordiversityandpredationofEpiphyaspostvittana (Lep-idoptera:Tortricidae)inavineyard.BiologicalControl,41(2), 230–236.http://dx.doi.org/10.1016/j.biocontrol.2007.01.012 Gelman,A.,&Hill,J.(2007).Dataanalysisusingregressionand

multilevel/hierarchicalmodels.England:CambridgeUniversity Press.

Gelman, A., & Shirley, K. (2011). Inference from simulations and monitoring convergence. In S. Brooks, A. Gelman, G. L.Jones, & X.-L.Meng (Eds.),Handbookof MarkovChain MonteCarlo (pp.163–174).Boston:Chapman&Hall/CRC. http://dx.doi.org/10.1201/b10905-7

Griffin,J.N.,Byrnes,J.E.K.,&Cardinale,B.J.(2013).Effectsof predatorrichnessonpreysuppression:Ameta-analysis. Ecol-ogy,94(10),2180–2187.http://dx.doi.org/10.1890/13-0179.1 Hogg, B. N., & Daane, K. M. (2011). Diversity and

inva-sion within a predator community: Impacts on herbivore suppression. Journal of Applied Ecology, 48(2), 453–461. http://dx.doi.org/10.1111/j.1365-2664.2010.01940.x

Holway,D.A.,Lach, L.,Suarez,A.V.,Tsutsui,N.D.,& Case, T. J.(2002). The causes and consequences of ant invasions. Annual Review of Ecology and Systematics,33(1), 181–233. http://dx.doi.org/10.1146/annurev.ecolsys.33.010802.150444 Kenis,M.,Auger-Rozenberg,M.-A.,Roques,A.,Timms,L.,Péré,

C.,Cock,M.J.W.,...&Lopez-Vaamonde,C.(2009). Ecologi-caleffectsofinvasivealieninsects.BiologicalInvasions,11(1), 21–45.http://dx.doi.org/10.1007/s10530-008-9318-y

Krushelnycky,P.D.,Holway,D.A.,&LeBrun,E.G.(2010). Inva-sionprocessesandcausesofsuccess.InL.Lach,C.L.Parr,&K. L.Abbott(Eds.),Antecology(pp.245–286).NewYork,USA: OxfordUniversityPress.

Lemessa,D.,Hambäck,P.A.,&Hylander,K.(2015).Arthropodbut notbirdpredationinEthiopianhomegardensishigherin tree-poorthanintree-richlandscapes.PLoSOne,10(5),e0126639. http://dx.doi.org/10.1371/journal.pone.0126639

Letourneau,D.K.,Jedlicka,J.A.,Bothwell,S.G.,&Moreno,C. R.(2009).Effectsofnaturalenemybiodiversityonthe suppres-sionofarthropodherbivoresin terrestrialecosystems.Annual ReviewofEcology,Evolution,andSystematics,40(1),573–592. http://dx.doi.org/10.1146/annurev.ecolsys.110308.120320 Lewis,T.(1997).Thripsascroppests.England:CABInternational.

Loreau, M. (2000). Biodiversity and ecosystem function-ing: Recent theoretical advances. Oikos, 91(1), 3–17. http://dx.doi.org/10.1034/j.1600-0706.2000.910101.x

Losey, J. E., & Denno, R. F. (1998). Positivepredator-predator interactions: Enhanced predation rates and synergistic sup-pression of aphid populations. Ecology, 79(6), 2143–2152. http://dx.doi.org/10.1890/0012-9658(1998)079[2143:PPPIEP] 2.0.CO;2

Losey, J. E., & Denno, R. F. (1999). Factors facil-itating synergistic predation: The central role of

synchrony. Ecological Applications, 9(2), 378–386. http://dx.doi.org/10.1890/1051-0761(1999)009[0378:FFSPTC] 2.0.CO;2

Lowe,S.,Browne,M.,Boudjelas,S.,& DePoorter, M.(2000). 100oftheworld’sworstinvasivealienspecies:aselectionfrom theglobalinvasivespeciesdatabase.Auckland:TheInvasive SpeciesSpecialistGroup.

Marc, P., Canard, A., & Ysnel, F. (1999). Spiders (Araneae) useful for pest limitation and bioindication. Agricul-ture, Ecosystems and Environment, 74(1–3), 229–273. http://dx.doi.org/10.1016/S0167-8809(99)00038-9

Marliac, G., Simon, S., Mazzia, C., Penvern, S., Lescour-ret, F., & Capowiez, Y. (2015). Increased grass cover height in the alleys of apple orchards does not promote Cydia pomonella biocontrol. BioControl, 60(6), 805–815. http://dx.doi.org/10.1007/s10526-015-9687-y

Mitchell, M. G. E., Bennett, E. M., & Gonzalez, A. (2014). Agricultural landscape structure affects arthro-pod diversity and arthropod-derived ecosystem services. Agriculture, Ecosystems and Environment, 192, 144–151. http://dx.doi.org/10.1016/j.agee.2014.04.015

Mollot, G., Duyck, P.-F., Lefeuvre, P., Lescourret, F., Mar-tin, J.-F., Piry, S., & Tixier, P. (2014). Cover cropping alters the diet of arthropods in a banana plantation: A metabarcoding approach. PLoS One, 9(4), e93740. http://dx.doi.org/10.1371/journal.pone.0093740

Monteiro, L.B., Lavigne,C., Ricci,B., Franck,P.,Toubon, J.-F.,&Sauphanor,B.(2013).Predationofcodlingmotheggsis affectedbypestmanagementpracticesatorchardandlandscape levels.Agriculture,EcosystemsandEnvironment,166,86–93. http://dx.doi.org/10.1016/j.agee.2011.10.012

Morrison, W. R., Mathews, C. R., & Leskey, T. C. (2016). Frequency, efficiency, and physical characteristics of pre-dation by generalist predators of brown marmorated stink bug(Hemiptera: Pentatomidae)eggs.BiologicalControl, 97, 120–130.http://dx.doi.org/10.1016/j.biocontrol.2016.03.008 Offenberg, J. (2015). Ants as tools in sustainable

agri-culture. Journal of Applied Ecology, 52(5), 1197–1205. http://dx.doi.org/10.1111/1365-2664.12496

Oksanen,J.,Blanchet,F.G.,Kindt,R.,Legendre,P.,Minchin,P. R.,O’Hara,R.B.,...&Wagner,H.(2015).vegan:Community ecologypackage.CRANRPackage.

Parr,C.L.,&Gibb,H.(2010).Competitionandtheroleofdominant ants.InL.Lach,C.L.Parr,&K.L.Abbott(Eds.),Antecology

(pp.77–96).NewYork,USA:OxfordUniversityPress.

Parr, C. L., & Gibb, H. (2012). The discovery-dominance trade-off is the exception, rather than the rule. Journal of Animal Ecology, 81(1), 233–241. http://dx.doi.org/10.1111/j.1365-2656.2011.01899.x

Pfannenstiel,R.S.(2008).Spiderpredatorsoflepidopteraneggs insouthTexasfieldcrops.BiologicalControl,46(2),202–208. http://dx.doi.org/10.1016/j.biocontrol.2008.03.011

Plummer,M.(2003).JAGS:Aprogramforanalysisofbayesian graphical modelsusing gibbs sampling. JAGS:Just Another GibbsSampler.InProceedingsofthe3rdinternational work-shop on distributed statistical computing (DSC 2003).ISSN 1609-395X.

RDevelopmentCoreTeam.(2014).R:ALanguageand Environ-ment for Statistical Computing. R Foundation for Statistical Computing,http://dx.doi.org/10.1038/sj.hdy.6800737

Schmitz,O.J.(2007).Concepts&synthesispredatordiversityand trophicinteractions.Ecology,88(10),2415–2426.

Simon, S., Defrance, H., & Sauphanor, B. (2007). Effect of codling moth management on orchard arthropods. Agri-culture, Ecosystems and Environment, 122(3), 340–348. http://dx.doi.org/10.1016/j.agee.2007.01.020

Snyder, W. E., & Evans, E. W. (2006). Ecological effects of invasive arthropod generalist predators. Annual Review of Ecology, Evolution, and Systematics, 37(1), 95–122. http://dx.doi.org/10.1146/annurev.ecolsys.37.091305.110107 Straub, C. S., Finke, D. L., & Snyder, W. E. (2008). Are the

conservationofnaturalenemybiodiversityandbiological con-trol compatible goals? Biological Control, 45(2), 225–237. http://dx.doi.org/10.1016/j.biocontrol.2007.05.013

Su,Y.S.,&Yajima,M.(2012).R2jags:APackageforRunningjags fromR.http://CRAN.R-Project.Org/package=R2jags.

Vance-Chalcraft,H.D.,Rosenheim,J.A.,Vonesh,J.R.,&Craig,W. (2012).Theinfluenceofintraguildpredationonpreysuppression andpreyrelease:Ameta-analysis.Ecology,88(11),2689–2696.

Way, M. J., Javier, G., & Heong, K. L. (2002). The role of ants,especiallythefireant,Solenopsisgeminata(Hymenoptera: Formicidae), in thebiological controlof tropical uplandrice pests. Bulletin of Entomological Research, 92(5), 431–437. http://dx.doi.org/10.1079/ber2002185

Weiser, M. D., & Kaspari, M. (2006). Ecologicalmorphospace of NewWorldants. EcologicalEntomology, 31(2),131–142. http://dx.doi.org/10.1111/j.0307-6946.2006.00759.x

Werling,B.P.,Harmon,J.,Straub,C.,&Gratton,C.(2012). Influ-enceofnativeNorthAmericanprairiegrassesonpredationof aninsectherbivoreofpotato.BiologicalControl,61(1),15–25. http://dx.doi.org/10.1016/j.biocontrol.2011.11.009

Wetterer,J.K.(2010).Worldwidespreadofthetropicalfireant:

Solenopsisgeminata(Hymenoptera:Formicidae). Myrmecolog-icalNews,14(January),21–35.

Wetterer, J. K. (2012). Worldwide spread of the African big-headedant,Pheidolemegacephala(Hymenoptera:Formicidae). Myrmecological News, 17(August), 51–62. http://dx.doi.org/ 10.2984/1534-6188(2007)61[437:BAIOPI2.0.CO;2]

Wielgoss, A., Tscharntke, T., Rumede, A., Fiala, B., Seidel, H., Shahabuddin, S., & Clough, Y. (2014). Interaction com-plexity matters: Disentangling services and disservices of ant communities driving yield in tropical agroecosystems. Proceedings of the Royal Society B, 281(1775), 20132144. http://dx.doi.org/10.1098/rspb.2013.2144

Wikle,C.K.(2003).HierarchicalBayesianmodelsforpredicting thespreadofecologicalprocesses.Ecology,84(6),1382–1394. http://dx.doi.org/10.1890/0012-9658(2003)084[1382:

HBMFPT2.0.CO;2]

Yusa, Y. (2001). Predation of eggs of the apple snail Pomacea canaliculata(Gastropoda:Ampullariidae)byfireant Solenop-sis geminata. Journal Molluscan Studies, 67(3), 275–279. http://dx.doi.org/10.1093/mollus/67.3.275

Availableonlineatwww.sciencedirect.com