Diversity and Distributions. 2019;1–15. | 1wileyonlinelibrary.com/journal/ddi
Received:29October2018 | Revised:19January2019 | Accepted:15February2019DOI: 10.1111/ddi.12913
B I O D I V E R S I T Y R E S E A R C H
Distribution trends of European dragonflies under climate change
Tim Termaat1,2 | Arco J. van Strien3 | Roy H. A. van Grunsven1 | Geert De Knijf4 | Ulf Bjelke5 | Klaus Burbach6 | Klaus‐Jürgen Conze7 | Philippe Goffart8 | David Hepper9 | Vincent J. Kalkman10 | Grégory Motte8 | Marijn D. Prins1,11 | Florent Prunier12 | David Sparrow13 | Gregory G. van den Top1 | Cédric Vanappelghem14,15 | Michael Winterholler16 | Michiel F. WallisDeVries1,17
1DeVlinderstichting/DutchButterflyConservation,Wageningen,TheNetherlands2BosgroepMiddenNederland,Ede,TheNetherlands3StatisticsNetherlands,TheHague,TheNetherlands4ResearchInstituteforNatureandForest,Brussels,Belgium5SwedishBiodiversityCentre,SwedishUniversityofAgriculturalSciences,Uppsala,Sweden6AGLibellenBayern,Marzling,Germany7AKLibellenNRW,Essen,Germany8DirectiongénéraleopérationnelleAgriculture,RessourcesnaturellesetEnvironnement(DGARNE),Départementdel'EtudeduMilieuNatureletAgricole,DirectiondelaNatureetdel'Eau,ServicepublicdeWallonie,Gembloux,Belgium9BritishDragonflySociety,Peterborough,UK10EuropeanInvertebrateSurvey—TheNetherlands,NationaalNatuurhistorischMuseumNaturalis,Leiden,TheNetherlands11NaturalisBiodiversityCentre,Leiden,TheNetherlands12AEAElBosqueAnimado,ValledelGenal,Spain13CyprusDragonflyStudyGroup,Pafos,Cyprus14Sociétéfrançaised'odonatologie,Boisd'Arcy,France15Unité“Evolution,Ecologie,Paléontologie”,UMRCNRS8198,Bat.SN2UniversitédeLille,Villeneuved'Ascq,France16BavarianEnvironmentAgency,Augsburg,Germany17PlantEcologyandNatureConservationGroup,WageningenUniversity,Wageningen,TheNetherlands
ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionandreproductioninanymedium,providedtheoriginalworkisproperlycited.©2019TheAuthors.Diversity and DistributionsPublishedbyJohnWiley&SonsLtd
CorrespondenceTimTermaat,DeVlinderstichting/DutchButterflyConservation,Wageningen,TheNetherlands.Email:[email protected]
Editor:AlanAndersen
AbstractAim: Polewardrangeshiftsofspeciesareamongthemostobviouseffectsofclimatechangeonbiodiversity.Asaconsequenceoftheserangeshifts,speciescommunitiesarepredictedtobecomeincreasinglycomposedofwarm‐dwellingspecies,butthishas only been studied for a limited number of taxa,mainly birds, butterflies andplants.Asspeciesgroupsmayvaryconsiderablyintheiradaptationtoclimatechange,it isdesirabletoexpandthesestudiestoothergroups,fromdifferentecosystems.Freshwater macroinvertebrates, such as dragonflies (Odonata), have been rankedamongthespeciesgroupswithhighestpriority.Inthispaper,weinvestigatehowthe
2 | TERMAAT ET Al.
1 | INTRODUC TION
Climatechangehasaprofound impactontheoccurrenceofmanyspecies of plants and animals (Parmesan & Yohe, 2003; Root etal., 2003;Walther et al., 2002).One of themost distinctive con‐sequencesisthepolewardshiftofspeciesdistributionrangesasaresultof increasingtemperatures,resulting inchanges inthecom‐position of species communities (Chen, Hill, Ohlemüller, Roy, &Thomas,2011;Hickling,Roy,Hill,Fox,&Thomas,2006;Kampichler,VanTurnhout,Devictor,&VanderJeugd,2012;Lindström,Green,Paulson,Smith,&Devictor,2013;Masonetal.,2015).Speciesvaryintheirresponsetoclimatewarming,duetodifferenttemperaturerequirementsanddifferentdispersalandcolonizationcapacities.Ingeneral,warm‐dwellingspeciesandspecieswithgooddispersalca‐pacityaremorelikelytobe“winners”thancold‐dwellingspeciesandspecieswithpoordispersalcapacity(Francoetal.,2006;Rosset&Oertli,2011;Virkkala&Lehikoinen,2014).Asaconsequence,com‐munitiesarepredictedtobecomeincreasinglycomposedofwarm‐dwelling,mobilespecies.
Thismayseemstraightforward,buttheeffectsofclimatechangeon species’ trends and community compositions have only beenstudiedfora limitednumberoftaxa(butseeHicklingetal.,2006;Masonetal.,2015),mainlybirds,butterfliesandplants(Bertrandetal.,2011;Britton,Beale,Towers,&Hewison,2009;Clavero,Villero,&Brotons,2011;Davey,Devictor,Jonzén,Lindström,&Smith,2013;
Devictoretal.,2012a;Jiguetetal.,2010;Roth,Plattner,&Amrhein,2014;Virkkala&Lehikoinen,2014).Togainabetterunderstandingofhowclimatechangeaffectstotaldiversity,moretaxaneedtobecovered,includingtaxafromdifferenthabitats.Freshwatermacroin‐vertebratesshouldberankedamongthefaunalgroupswithhighestpriority,astheyhaveverydifferentlifehistoriesfrombirdsandbut‐terfliesandoccupyverydifferentecosystems.Theyareknowntoreactquicklytoawiderangeofchangesintheirhabitats(Rosenberg&Resh, 1993). Furthermore, freshwater covers only 0.8% of theEarth's surface, while supporting almost 6% of all described spe‐cies,mostofwhichareinsects(Dijkstra,Monaghan,&Pauls,2014;Dudgeonetal.,2006).Atthesametime,theyareamongthemostseverelythreatenedecosystemsintheworld,withaquaticspeciesbeingmorethreatenedthanterrestrialspecies(Collenetal.,2014;Darwalletal.,2018;Dudgeonetal.,2006).Forthesereasons,fresh‐water invertebrateshavebeenindicatedasanessentialfuturead‐ditiontoEurope'sbiodiversitymonitoringprogramme(Feest,2013;Thomas,2005).
Unfortunately, monitoring freshwater invertebrates comeswithdrawbacks.Mostgroupsaresospecies‐rich thatcollecting,sortingandidentifyingsamplestospecies levelrequiremuchef‐fortandexperience.Therefore,thenumberofspecialistsstudyingthesegroupsis,inmostcountries,limited,whichresultsinanin‐completepictureofspecies’distributions.Dragonflies (Odonata)present an exception to this rule. Adult dragonflies are large,
occurrence of dragonflies in Europe has changed in recent decades, and if thesechangesareinparallelwithclimatechange.Location: Europe.Methods: Weuse data from10 European geographical regions to calculate occu‐pancy indicesandtrendsfor99 (69%)of theEuropeanspecies.Next,wecombinetheseregionalindicestocalculateEuropeanindices.Todetermineifchangesinre‐gionaldragonflycommunitiesinEuropereflectclimaticwarming,wecalculateSpeciesTemperatureIndices(STI),Multi‐speciesIndicators(MSI)andCommunityTemperatureIndices(CTI).Results: 55of99consideredspecies increased inoccupancyatEuropean level,32speciesremainedstable,andnonedeclined.Trendsfor12speciesareuncertain.MSIof cold‐dwelling andwarm‐dwelling species differ in some of the regions, but in‐creasedatasimilarrateatEuropeanlevel.CTIincreasedinallregions,exceptCyprus.TheEuropeanCTIincreasedslightly.Main conclusions: Europeandragonflies,ingeneral,haveexpandedtheirdistributioninresponsetoclimatechange,eventhoughtheirCTIlagsbehindtheincreaseintem‐perature.Furthermore,dragonfliesprovedtobeasuitablespeciesgroupformonitor‐ingchangesincommunities,bothatregionalandcontinentallevel.
K E Y W O R D S
citizensciencedata,climatechange,CommunityTemperatureIndex,Multi‐speciesIndicator,Odonata,SpeciesTemperatureIndex
| 3TERMAAT ET Al.
colourful insects, which are easy to spot and relatively easy toidentify at species level,making them attractive to a large pub‐lic.Withamanageable143speciesrecordedinEurope(Kalkmanet al., 2018), they constitute a suitablegroup for citizen scienceprojects.Furthermore,dragonfliesarewellestablishedasusefulorganismstoassessandmonitoraquaticandwetlandecosystemquality(Oertli,2008),andtheyareknowntoreactquicklytocli‐mate change (Bush, Theischinger, Nipperess, Turak, & Hughes,2013;Hassall,2015).
InmostEuropeancountries,dragonfly recordinghas increasedinrecentdecades,mediatedbythepublicationofseveralgoodfieldguidesandnationaldistributionatlases.ThishasresultedinasteepincreaseinavailabledistributiondatafromcitizenscienceprojectsandthepublicationofaEuropeandistributionatlasin2015(Boudot&Kalkman,2015).Themajorityofthesedistributiondatarefertorecordscollectedwithoutstandardization,whichareunsuitableforstraightforwardcalculationofdistributionstrends.However,previ‐ous studieshave shown that these “opportunistic” recordscanbeusedtoderivereliabletrendestimatesofdragonfliesonanationalscale,ifoccupancymodelsareapplied.Thesemodelstaketheimper‐fectdetectionofspeciesintoaccount,andthereby,theymaysimul‐taneouslycorrectforobservationandreportingbiasaswell(Isaac,
Van Strien,August,DeZeeuw,&Roy, 2014;Van Strien, Termaat,Groenendijk, Mensing, & Kéry, 2010; Van Strien, Van Swaay, &Termaat,2013).Moreover,VanStrien,Termaatetal.(2013)showedinapilotstudy,usingrecordsofasinglespeciesfromfivewesternEuropeanregions,thatoccupancyindicesfrommultipleregionscanbecombinedtocalculatesupraregionalindicesandtrends.
Inthispaper,weinvestigatehowtheoccurrenceofdragonfliesinEuropehaschangedinrecentdecades,andifthesechangesareinparallelwithclimatechange.Weusedistributiondatafrom10Europeangeographicalregions—rangingfromSwedentoCyprus—tocalculateoccupancy indicesandtrends forasmanydragonflyspecies as possible.Next,we combine these regional indices tocalculate European indices. To determine if changes in regionaldragonflycommunitiesinEuropereflectclimaticwarming,wecal‐culateSpeciesTemperatureIndices(STI),Multi‐speciesIndicators(MSI)andCommunityTemperatureIndices(CTI).Wehypothesizethat (a) warm‐dwelling species have more positive trends thancold‐dwellingspecies, that,asaconsequence, (b)warm‐dwellingspecies have increased their share in regional communities and(c)thattheseeffectsincreaseonasouth–northgradientthroughEurope,astheratioofwarm‐andcold‐dwellingspeciesdecreaseswithincreasinglatitude.
F I G U R E 1 ParticipatingEuropeangeographicalregions,hereconsideredascountriesorloweradministrativedivisions
4 | TERMAAT ET Al.
2 | METHODS
2.1 | Species records
We gathered dragonfly distribution records, from 1990 onwards,from the following European geographical regions (countries orlower administrative divisions, hereafter referred to as “regions”):Sweden,Britain (UnitedKingdomexcludingNorthern Ireland), theNetherlands,NorthRhine‐Westphalia (aGerman state),Bavaria (aGerman state), Flanders (a Belgian region, including Brussels re‐gion),Wallonia (aBelgian region),France,Andalusia (aSpanishau‐tonomouscommunity)andCyprus(Figure1).Theresultingdatasetincluded recordsof99 species,whichequals69%of all dragonflyspeciesrecordedinEurope(Kalkmanetal.,2018).
Allrecordsusedinthisstudycoveradultdragonfliesonly.Themajority of these records are “opportunistic,” that is, collectedwithout a standardized field protocol and without a design en‐suringthegeographical representativenessofsampledsites.Theperiodofdatacoverageandthenumberofrecordsperunitareavary considerably among regions, depending on data availability(Supporting Information Table S1). All data in each region werevalidated by experts to prevent false‐positive records. To stan‐dardizethegeographicalreferencesystem,allobservationsweremapped in the ETRS89/ETRS‐LAEA (EPSG:3035) reference sys‐tem.Becauseweused1×1kmgridsquaresasthedefinitionofasiteinouranalyses,allobservationswerereferencedto1×1kmETRS‐LAEAsquares.
2.2 | Generating non‐detection data
Occupancy models require detection/non‐detection data col‐lectedduringreplicatedvisits.Validreplicatedvisitsareonlythosevisitsmadeinaperiodofclosurewithintheyear;thisistheperiodduringwhichasiteisconsideredeithertobeoccupiedorunoccu‐piedbythespeciesandnotabandonedorcolonized(MacKenzieetal.,2006).Fordragonflies,weconsideredtheperiodofclosureasthemainflightperiodofaspecies.Closureperiodsweredefinedfor each combination of species and region. For each combina‐tion,approximately5%ofboththeearliestandthelatestrecordswereexcluded,resultinginthespecies’mainflightperiod.Thesemain flightperiodswereexpressed in Juliandates.Forexample,weusedJuliandates125–210astheclosureperiodofPyrrhosoma nymphula(anearlyflyingspecies)inFranceandJuliandates200–240astheclosureperiodofAeshna viridis(alateflyingspecies)inSweden.
Almostalldataobtainedwererecordsofspeciespresence.Thenon‐detectionrecordsofagivenspeciesweregeneratedfromtheinformation of sightings of other dragonfly species, followingVanStrienetal.(2010)andVanStrien,VanSwaayetal.(2013).Anyob‐servationofagivenspecieswastakenas1(detection),whereaswerated0(non‐detection) ifanyspeciesotherthanthegivenspecieshadbeenreportedbyanobserverataparticular1×1kmsiteandonaparticulardatewithinthespecies'closureperiod.
2.3 | Species trend analysis
2.3.1 | Annual occupancy estimates and trends: regional level
First,wecalculatedannualoccupancyperspecies,foreachregionseparately.WeappliedthesamedynamicoccupancymodelasVanStrienet al. (2010) andVanStrien,VanSwaayet al. (2013) toes‐timate annual occupancy ψ, adjusted for detection probability p. Because all parameters in themodelmay differ between regions,theanalyseswereperformedseparatelyforeachregionandthere‐gionalresultswerecombinedinasecondstep.ThedescriptionofthemodelisderivedfromRoyleandKéry(2007)andRoyleandDorazio(2008).Here,ψ istheproportionofsuitable1×1kmsquaresthatisoccupied.Asquareisdefinedassuitableifthespecieshadbeenrecordedthereat leastonce in1990–2008.Theoccupancymodelconsistsoftwohierarchicallycoupledsubmodels,oneforoccupancyandonefordetection,thelatterbeingconditionalontheoccupancysubmodel.Theoccupancysubmodelestimatesannualprobabilityofpersistenceφt andofcolonizationγt andcomputes theannualoc‐cupancyprobabilitypersiterecursivelythrough:
Thus,whethersiteioccupiedinyeart−1isstilloccupiedinyeartisdeterminedbythepersistenceprobability,andwhethersitei un‐occupiedinyeart−1isoccupiedinyeartdependsonthecoloniza‐tionprobability.Alloccupancyprobabilitiespersite togetheryieldtheestimatedannualnumberofoccupied1×1kmsitesperregion.Thesamesiteswereincludedintheanalysisforallyears;estimatesfor sites not surveyedduring someyearswerederived from sitesthatweresurveyedinthoseyears.
Thedetectionsubmodelestimatestheyearlydetectionp,butinaddition,pismadeasafunctionofcovariates.WeusedtheJuliandate as a covariate forp because thedetectionof the species isexpectedtovaryovertheseason,duetochangingpopulationsizeduringthecourseoftheflightperiod.Detectionisalsoreducedifobserversdonotreportalltheirsightings.Hence,weincludetheincompletenessofrecordingasacovariatefordetection.Wedis‐tinguished: (a) single recordsofany speciesonone siteanddatewithoutrecordsofotherspecies,(b)shortday‐lists,thatis,recordsoftwoorthreespeciesmadebyasingleobserverononesiteanddateand(c)comprehensiveday‐lists,thatis,recordsofmorethanthreespeciesperobserver,siteanddate.These listsmayormaynotincludethespeciesinquestion.Thesecategorythresholdsaresufficientlylownottobeconfoundedbyrealdifferencesinspeciesnumberbetweensites.Inmost1×1kmsitesintheregions,therearemore than three species to be found and oftenmanymore.Effectsofbothcovariateswereincludedinthedetectionsubmodelviaalogitlink:
�it=�i,t−1�t−1+ (1−�i,t−1)�t−1
logit(pijt)=�t+�1 ∗dateijt +�2 ∗date2ijt
+�∗1(short day - list)ijt+�∗
2(comprehensive day - list)ijt,
| 5TERMAAT ET Al.
wherepijtistheprobabilitytodetectthespeciesatsiteiduringvisitj in year t,αtistheannualinterceptimplementedasarandomeffect,β1 and β2are the linearandquadraticeffectsof thedateofvisit j and δ1 and δ2aretheeffectsofshortday‐listsandcomprehensiveday‐lists,relativetosinglerecords.
We fitted themodels in aBayesianmodeof inference usingJAGS(Plummer,2017)onthecomputerclusterLISA(https://sub‐trac.sara.nl). We chose uninformative priors for all parameters,using uniform distributionswith values between 0 and 1 for allparametersexceptδ1 and δ2 (valuesbetween−10and10),β1,β2 (valuesbetween−10and10)andαt(valuesbetween0and5)forthestandarddeviationofthenormaldistributionusedaspriorfortherandomyeareffect.
Foreachanalysis,weranthreeMarkovchainswithsufficientit‐erations toensureconvergenceas judged from theGelman‐RubinRhatstatisticandsavedthelast93iterationsforuseatsupraregionallevel.Thisnumberof iterations isanempiricallyobtainedcompro‐mise between the reliability of the estimates and data handlingcapacity.Themodelproducedannualestimatesofoccupancyperre‐gion,whichwereconvertedintoannualindiceswithfirstyear=100.Thetrendinoccupancywasconsideredsignificantifitsconfidenceintervaldidnotincludezero.
2.3.2 | Annual occupancy estimates and trends: European level
In thenext step, theannualoccupancyestimatesper regionwereaggregatedtoobtainEuropeanoccupancyindicesandtrendsfortheperiod 1990–2015.Missing yearly values from a particular regionwereestimated (“imputed”) fromaveragedyear‐to‐yearoccupancyratios in all other regions. For example, 1990 was missing in theSwedishdataset.To imputeoccupancyestimatesofSwedishspe‐cies,we applied the1991/1990 ratios fromall other regionswithdatafrombothyears.Asaconsequenceoftheseimputations,con‐fidenceintervalsincreasedforyearswithlackingdatafromoneormoreregions,especiallywhenthesewerelargeregions(e.g.,France).
Regionsdifferinthenumberofsitessurveyed,soanaiveaggre‐gationhastheriskofbiasedEuropeantrends.Hence,wedevelopedaproceduretoweighregionsaccordingtothesamplingintensityinrelationtotherangeofspeciesineachregion.ThisprocedureisanadaptationofproceduresappliedbyVanSwaay,Plate,andVanStrien(2002)andGregoryetal.(2005).Weightswerecalculatedasthequo‐tientofrelativerangeandrelativesamplingintensitytocompensateforoversamplingandundersampling.Relativerangewasdefinedastherangeofaspeciesinaregion,asapercentageofitstotalrangeinallregionsforwhichanoccupancyindexcouldbeobtained.Relativesampling intensitywasdefinedas thenumberof1×1kmsquaressurveyedatleastonceinthisperiodwithintheregionalrangeofthespecies, relative to the totalnumberof surveyedsquares inall re‐gionswithindices.Weightsperregionweresimilarforeachyearbe‐causethesamesiteswereintheanalysisforallyears.TheweightednumbersofoccupiedsiteswereaddedacrossregionsandconvertedintoEuropeanannualindiceswith1990=100.Wetookintoaccount
theuncertaintyoftheestimatednumberofoccupiedsitesperregionbyaddingthenumberofsitesestimatedperregionforeachofthesaved93iterationsandthencombiningtheresultsofalliterations.
2.4 | Species Temperature Indices
WecalculatedtheSTIforeachdragonflyspeciesoccurringinEurope(Boudot&Kalkman,2015) (Supporting InformationTableS2).TheSTIofagivenspeciesistheaveragetemperature(expressedinde‐greesCelsius)oftheEuropeanpart(excludingRussia)ofthespecies’rangeandistakenasaproxyforspecies’dependenceontempera‐ture.Thesecalculationswerebasedon2,736siteswithspeciesre‐cordsunderlyingtherangemapsof theEuropeanatlasbyBoudotandKalkman(2015;availablethroughKalkmanetal.,2018)andcli‐matedataofWorldClim(http://www.worldclim.org;accessedMarch2017;averagemonthlytemperaturesfor1960–1990).Theanalyseswerecarriedoutata50×50kmgridscale.Foreachgridsquare,wecalculatedtheannualmeantemperaturetoestimatetheSTIasthemean temperature of occupied squares.Although the distributiondatacoveredEuropetoagreatextent,wefounditnecessarytocor‐rectfordifferencesinsamplingintensitybetweenregions.Thiswasachievedbybootstrapping,whichconsistedof100 replicationsofasubsetofrandomlychosen50x50gridsquareswithinanareaof250×250km.STIswereestimatedasthemeantemperatureofalloccupiedsquaresoverallreplications.
The period covered by the temperature data fromWorldClim(1960–1990) differed from the period covered by the atlas’ rangemaps (>1990).However, relative differences in STI among speciesarerobusttothetimewindowconsidered(Devictoretal.,2012b).
2.5 | Multi‐species Indicators
To determinewhether warm‐dwelling species havemore positivetrends than cold‐dwelling species, we calculated Multi‐speciesIndicators (MSI), by combining the trends in occupancy indices ofcold‐dwelling and of warm‐dwelling species, respectively.We didthis for each region separately and for Europe as a whole. Cold‐dwellingspeciesweredefinedasspecieswithSTIlowerthan9.8°C,whichisthemedianSTIofallspeciesincludedinourstudy.Warm‐dwellingspeciesweredefinedasspecieswithSTI>9.8°C.StandarddeviationsofSTIdidnotdifferbetweenthetwogroups (one‐wayANOVA,F(1,97)=0.554,p=0.458).
MSIwerecalculated including theirconfidence intervals,usingtheRscript“MSItool”(Soldaat,Pannekoek,Verweij,VanTurnhout,&VanStrien,2017).Thismethodisdevelopedtoaccountforsam‐pling error of species indices in the calculation of Multi‐speciesIndicators, by calculating confidence intervals using Monte Carlosimulationsofannualspeciesindices.
2.6 | Community Temperature Indices
Ultimately, we calculated a CTI for each region, as the averageSTI of all species in the region, weighted by species occupancies
6 | TERMAAT ET Al.
(probabilities of occurrence). CTI is thus expressed in degreesCelsius.Similarly,wecalculatedEuropeanCTI.AtemporalincreaseinCTIdirectly reflects that the speciesassemblage is increasinglycomposedofspeciesthatoccurathighertemperatures(thatiswithhigh STI). With this approach, we follow Devictor et al. (2012a),withtheprincipaldifferencethatwefocusonregionalcommunitiesbasedonoccupancydatafromkmsquares,insteadoflocalcommu‐nitiesbasedonabundancedatafromtransects(althoughDevictoretal.,2012aalsoincludedananalysisonpresence–absencedatawhichcompareswithourapproach).
3 | RESULTS
3.1 | Occupancy indices and trends
Thenumberofspeciesforwhicharegionaltrendcouldbecalculatedwithsufficientlylowstandarderrors,thatis,standarderrorslowenoughtodetecta5%orhigherannual increaseordecline,rangedfromfiveforCyprusto79forFrance(Table1).Intotal,wewereabletocalculate
trendswithsufficientlylowstandarderrorsfor90of99speciesinourdataset,foratleastoneoftheregions(SupportingInformationDataS1).
In7outof10 regions,morespecies increased thandecreasedtheiroccupiedrange(Table1).TheseregionswereSweden,Britain,theNetherlands,North Rhine‐Westphalia, Flanders,Wallonia andFrance.NosignificantdifferencebetweenincreasinganddecliningspecieswasfoundforBavaria,becausethisregionhadahighnum‐berofstablespecies(36of59speciestrendswithsufficientlylowstandarderrors).ForAndalusiaandCyprus,thenumberofspeciestrendswith sufficiently low standard errorswas too small to findsignificantdifferencesbetweentrendclasses.
For all regions combined, 55 species moderately increased inoccupancy, indicating that they expanded their distribution at aEuropean level,32species remainedstableandnonedeclined.Asan example, indices of Sympecma fusca (a moderately increasingspecies) and Gomphus vulgatissimus (a stable species) are showninFigure2.Trendestimatesof12 specieshad too large standarderrors. European indices and trends of all species are provided inSupportingInformationDataS2.
TA B L E 1 Numberofspeciespertrendclasspergeographicalregion(fromnorthtosouth)andforEurope
Region Trend period N species Increase Stable Decline Uncertain % Increase χ2 p
Sweden 1991–2014 64 47 1 0 16 73.4 47.0 <0.001
Britain 1980–2012 50 26 12 2 10 52.0 20.6 <0.001
Netherlands 1991–2015 68 39 10 7 12 57.4 22.3 <0.001
NorthRhine‐Westphalia 1990–2010 67 21 15 0 31 31.3 21.0 <0.001
Flanders 1990–2015 62 27 17 7 11 43.5 11.8 <0.001
Wallonia 1990–2015 65 26 25 0 14 40.0 26.0 <0.001
Bavaria 1990–2013 73 8 36 15 14 11.0 2.1 0.144
France 1990–2012 87 30 45 4 8 34.5 19.9 <0.001
Andalusia 2006–2015 57 1 5 0 51 1.8 NA NA
Cyprus 2006–2015 35 3 2 0 30 8.6 NA NA
Europe 1990–2015 99 55 32 0 12 55.6 55.0 <0.001
Note. χ2:valueofchi‐squaredtest;p:probabilityvalue. Increase=significantincrease(p<0.05);Stable=nosignificantchange;Decline=significantdecline(p<0.05);Uncertain=nosignificantchangeandstandarderrorstoolargetodetecta5%trendifithadoccurred.
F I G U R E 2 Europeanindex(1990–2015)ofSympecma fusca and Gomphus vulgatissimus.Linearregressionlines(dashedlines)werealignedthroughtheyeareffectstosummarizeoverallchange
80
90
100
110
120
130
140
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
)001 = 0991( xednI
Sympecma fusca
Gomphus vulgatissimus
| 7TERMAAT ET Al.
3.2 | Species Temperature Indices
STIrangedfrom2.0°Cfortheboreo‐alpinespeciesAeshna caerulea to18.3°CfortheMediterranean(andAfrican)speciesTrithemis arte-riosa. (Mean=9.8°C;SD=3.3°C)(SupportingInformationTableS2).
3.3 | Multi‐species Indicators
MSIofwarm‐dwellingspecieswereincreasinginallregions(Figure3).Surprisingly,MSIofcold‐dwellingspeciesalsoincreasedinSweden,Britain, the Netherlands and North Rhine‐Westphalia. In Flanders,
F I G U R E 3 Multi‐speciesIndicators(MSI)ofwarm‐dwellingspecies(SpeciesTemperatureIndex>9.8°C)andcold‐dwellingspecies(SpeciesTemperatureIndex<9.8°C)pergeographicalregion(fromnorthtosouth)andforEurope.Thefirstyearwithdatawassetto100.Smoothedtrendlineswereplottedthroughtheyeareffectstosummarizeoverallchange.Shadedareasrepresentconfidenceintervals.Pleasenotethaty‐axesdiffer
0
20
40
60
80
100
120
140
16019
90
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Britain
MSI_warm (n = 13) MSI_cold (n = 27)
0
50
100
150
200
250
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
Sweden
MSI_warm (n = 12) MSI_cold (n = 42)
050
100150200250300350400450
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
Netherlands
MSI_cold (n = 36) MSI_warm (n = 21)
8 | TERMAAT ET Al.
Wallonia and France, MSI of cold‐dwelling species was stable, inBavariaitdeclined,andinAndalusia,itwasuncertain(Table2).Cyprushas only one cold‐dwelling species (Enallagma cyathigerum), whichincreased.
ComparingMSItrendsofwarm‐dwellingandcold‐dwellingspe‐cies (Table2) shows the formerwas significantlymorepositive inBritain, the Netherlands, Flanders, Wallonia, Bavaria and France.
AtaEuropeanlevel,however,theMSIofwarm‐dwellingandcold‐dwellingspeciesbothincreasedatasimilarrate.
3.4 | Community Temperature Indices
CTI increased in all regions, except Cyprus (Table 3; SupportingInformation Figure S1). Themost significantly increasing CTI was
020406080
100120140160180
North Rhine-Westphalia
MSI_warm (n = 19) MSI_cold (n = 29)
0
50
100
150
200
250
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
Flanders
MSI_warm (n = 23) MSI_cold (n = 34)
0
50
100
150
200
250
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
Wallonia
MSI_warm (n = 22) MSI_cold (n = 33)
F I G U R E 3 (Continued)
| 9TERMAAT ET Al.
found for the Netherlands. The Dutch dragonfly fauna “warmed”at 9.5×10−3°Cyear−1 over the period 1991–2015 (0.23°C overthewholeperiod).Theweakest increasewas found forBritain, at1.2×10−3°Cy−1overtheperiod1990–2015(0.03°Coverthewholeperiod).TheEuropeanCTIincreasedjustasslowly,at1.2×10−3°Cy−1 overtheperiod1990–2015(0.03°Coverthewholeperiod).
4 | DISCUSSION
We found clear effects of climate changeon severalwarm‐dwell‐ingspecies,consistentwithobservedchangesinEuropeandistribu‐tionsinthelastfewdecades(Boudot&Kalkman,2015).Inaddition,thedifferencesinMSIofwarm‐dwellingandcold‐dwellingspecies
0
20
40
60
80
100
120
140
160
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
Bavaria
MSI_warm (n = 23) MSI_cold (n = 40)
80859095
100105110115120125130
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
France
MSI_cold (n = 41) MSI_warm (n = 41)
0
50
100
150
200
250
300
350
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Andalusia
MSI_warm (n = 22) MSI_cold (n = 4)
F I G U R E 3 (Continued)
10 | TERMAAT ET Al.
indicate that climate changehas changeddragonfly occurrence atthecommunitylevelaswell.
4.1 | Testing of hypotheses
4.1.1 | Regional level
Wehypothesizedthat(a)warm‐dwellingspeciesshowmoreposi‐tive trends than cold‐dwelling species and this was confirmedfor 6 of 10 studied regions (Bavaria, Britain, Flanders, France,Netherlands,Wallonia). In Sweden—the most northern region inour study—bothMSI of cold‐dwelling species andMSI ofwarm‐dwelling species were stable between 1990 and 2001 and bothincreasedinacomparablepacefrom2002onward.Thissuggeststhat climatic conditions in the 1990s were probably limiting formostspeciesinSweden,includingcold‐dwellingspecies.Withtheexceptionofsomeextremecold‐tolerantspecies,suchasA. caeru-lea and Somatochlora sahlbergii, all Swedish species reach theirnorthernrangelimit inthisregion.Recenttemperaturerisesthusappeartohaveresultedinimprovedconditionsfornearlyallspe‐cies.Furthermore,weexpectedthat(b)warm‐dwellingspecieshadincreasedtheirshareinregionalcommunities.Thiswasconfirmed
for all regions exceptCyprus,whereonlyone cold‐dwelling spe‐ciesoccurs.However,withanincreasingCTIof1.2×10−3°Cyear−1 onaverage,upto9.5×10−3°Cyear−1fortheNetherlands(Table3),this“warming”ofregionalcommunitiesevolvesmoreslowlythanthe increase in temperature itself (1.1×10−2°Cyear−1, after cor‐rectingforthedifferencein latitudinalgradientbetweenCTIandactualtemperature;Devictoretal.,2012a),butthedifferencefortheNetherlands isminimal.Thus,dragonflies inEuropeareaccu‐mulating a substantial “climatic debt,” that is, the difference be‐tweenshiftsintemperatureandshiftsindistribution(Devictoretal.,2012a;Menéndezetal.,2006),whichvariesbetweenregions.Ultimately,ourhypothesis that (c) trends inregionalCTI increaseon a south–north gradient through Europe is rejected. HighestCTI increaseswerefoundforregionsonamoderate latitude(theNetherlands, Flanders, Wallonia) and for Andalusia (althoughmeasured over a shorter time span), while lowest CTI increaseswere foundforBritain,FranceandBavaria.Regionsdiffer insizeand subsequently in latitudinal gradient. This may, in theory, af‐fect regional occupancy trends (and thus regional CTI trends) tosomeextent,possibly limiting thevalidityofacomparisonat theregionallevel.CalculatingCTIacrossequallysizedlatitudinalbandswouldbeapreferableapproach,butrequiresahigherdatadensity
020406080
100120140160180200
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Cyprus
MSI_warm (n = 15) MSI_cold (n = 1)
020406080
100120140160180
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
Europe
MSI_warm (n = 49) MSI_cold (n = 50)
F I G U R E 3 (Continued)
| 11TERMAAT ET Al.
insomeofourregionsthaniscurrentlyavailable.TheMSItrendsofcold‐dwellingandwarm‐dwellingspecies(Table2)donotshowastructuraldifferencebetweenlargerandsmallerregionsindicat‐ingthatitisunlikelythatCTItrenddifferencesareconfoundedbydifferencesinregionsize.
4.1.2 | European level
AtEuropeanlevel,MSIofwarm‐dwellingandcold‐dwellingspecieswere similar, both having a slightly positive trend. At communitylevel though, the increase in European CTI of 1.2×10−3°Cyear−1
Region Trend period CTI slope SE p
Sweden 1991–2014 2.6×10−3 1.5×10−3 0.110
Britain 1990–2012 1.2×10−3 0.5×10−3 0.017
Netherlands 1991–2015 9.5×10−3 1.1×10−3 <0.001
NorthRhine‐Westphalia 1990–2010 2.0×10−3 0.8×10−3 0.025
Flanders 1990–2015 5.4×10−3 1.0×10−3 <0.001
Wallonia 1990–2015 4.3×10−3 0.8×10−3 <0.001
Bavaria 1990–2013 1.7×10−3 0.7×10−3 0.028
France 1990–2012 1.3×10−3 0.4×10−3 0.011
Andalusia 2006–2015 8.8×10−3 8.3×10−3 0.325
Cyprus 2006–2015 −26.7×10−3 9.5×10−3 0.023
Europe 1990–2015 1.2×10−3 0.5×10−3 0.019
Note. SE:standarderror;p:probabilityvalue.
TA B L E 3 SlopeofCommunityTemperatureindex(CTI)pergeographicalregion(fromnorthtosouth)andforEurope
TA B L E 2 Multi‐speciesIndicator(MSI)trendsofcold‐dwellingspecies(SpeciesTemperatureIndex<9.8°C)andwarm‐dwellingspecies(SpeciesTemperatureIndex>9.8°C)pergeographicalregion(fromnorthtosouth)andforEurope
Region Trend period N species MSI trend ± SE Classification p
Sweden cold 1991–2014 42 1.025 ± 0.002 Moderateincrease 0.177
Sweden warm 1991–2014 12 1.030 ± 0.005 Moderateincrease
Britaincold 1990–2012 27 1.005 ± 0.001 Moderateincrease 0.013
Britainwarm 1990–2012 13 1.012 ± 0.003 Moderateincrease
Netherlandscold 1991–2015 36 1.010 ± 0.003 Moderateincrease <0.001
Netherlandswarm 1991–2015 21 1.044 ± 0.006 Moderateincrease
NorthRhine‐Westphaliacold
1990–2010 29 1.012 ± 0.003 Moderateincrease 0.115
NorthRhine‐Westphaliawarm
1990–2010 19 1.019 ± 0.005 Moderateincrease
Flanderscold 1990–2015 34 1.003 ± 0.002 Stable <0.001
Flanderswarm 1990–2015 23 1.019 ± 0.004 Moderateincrease
Walloniacold 1990–2015 33 1.002 ± 0.002 Stable <0.001
Walloniawarm 1990–2015 22 1.024 ± 0.004 Moderateincrease
Bavariacold 1990–2013 40 0.998 ± 0.001 Moderatedecline <0.001
Bavariawarm 1990–2013 23 1.008 ± 0.003 Moderateincrease
Francecold 1990–2012 41 1.002 ± 0.001 Stable 0.017
Francewarm 1990–2012 41 1.005 ± 0.001 Moderateincrease
Andalusiacold 2006–2015 4 1.029 ± 0.034 Uncertain 0.489
Andalusiawarm 2006–2015 22 1.030 ± 0.011 Moderateincrease
Cypruscold 2006–2015 1 1.072 ± 0.030 Moderateincrease 0.694
Cypruswarm 2006–2015 15 1.055 ± 0.015 Moderateincrease
Europecold 1990–2015 50 1.011 ± 0.002 Moderateincrease 0.362
Europewarm 1990–2015 49 1.012 ± 0.002 Moderateincrease
Note. SE:standarderror;p:probabilityvalue.Moderateincrease=significantincrease≤5%(p<0.05);stable=nosignificantchange;moderatede‐cline=significantdecline≤5%(p<0.05);uncertain=nosignificantchangeandstandarderrorstoolargetodetecta5%trendifithadoccurred.
12 | TERMAAT ET Al.
shows that warm‐dwelling species have slightly increased theirshare.TocomparethisoutcomewiththetrendsinCTIofEuropeanbirdsandbutterflies (providedbyDevictoretal.,2012a,asbasedonpresence–absencedata),we re‐calculated theEuropeanCTIofdragonflies for the sameperiod1990–2008.Over these18years,CTI of European dragonflies increased with 2.4×10−3°Cyear−1,whichiscomparablewiththeincreaseinCTIofEuropeanbutterflies(2.5×10−3°Cyear−1) andconsiderablygreater than the increase inCTIofEuropeanbirds (1.9×10−3°Cyear−1).This is in linewith thewell‐known ability of dragonflies to quickly colonize newhabitats(Corbet,1999).Dragonfliesshouldprobablybeconsideredasmoredispersivethanbutterflies,which,fortheirpart,mayshowaquickercommunityresponseatalocalscale,duetotheirgenerallyshorterlifecycle.ThenetoutcomeoftheseopposingdifferencesmayhaveresultedinasimilarCTItrendbetweendragonfliesandbutterflies.The slower responseofbird communities to climaticwarminghasbeensuggestedbyDevictoretal. (2012a) tobeaconsequenceoftheirslowerpopulationturnover.
Inconclusion,climatechangehasaconsiderablepositive im‐pactontheoccurrenceofdragonfliesinseveralEuropeanregions.However, at a continental scale, CTI's are changing only slowlyso far, due to the relatively positive response of cold‐dwellingspecies.
4.2 | Limitations of CTI
SeveralauthorshavehighlightedtheCTIasausefultoolforassess‐ingtheeffectofclimatechangeonthecompositionofcommunities(Devictor, Julliard,Couvet,& Jiguet,2008; Lindströmet al., 2013;Rothetal.,2014).However,ourresultsshowthatastableCTIdoesnotnecessarilymeanthatclimatechangeisnotaffectingtheoccur‐renceofspecies.InSweden,manydragonflyspecieshavebenefitedfrom climate warming, including species of cool conditions. ThishasledtoincreasingMSItrendsforbothwarm‐dwellingandcold‐dwellingspecies,whileleavingCTIalmostunaffected.WethereforerecommendareviewingofCTIinrelationtoMSIofwarm‐dwellingandcold‐dwellingspecies,especiallyinhigh‐latituderegionswheretemperaturesmayhavelimitedspecieswithlowSTIaswellashighSTI.Inaddition,weknowthatmanydragonflyspecieshavesubstan‐tiallyexpanded their rangenorthwards (Boudot&Kalkman,2015;Hickling,Roy,Hill,&Thomas,2005;Ott,2010).Dragonflycommuni‐tieshavechangedasaresultoftheseexpansions,yetthisismaskedbyanincreaseinotherspeciesresultinginaquitestableCTI.Forex‐ample,itislikelythatthereductioninorganicpollutionandnutrientinput inthe lastquarterofthe20thcenturyhascompensatedtheeffectsof increasing temperature for species that are sensitive tolowoxygenlevels(Ketelaar,2010;Termaat,VanGrunsven,Plate,&VanStrien,2015).TheselimitationsofCTIasanindicatorofclimatechangearealsorelevantwhencalculationsarebasedonlocalabun‐dances instead of regional distributions, even though CTI trendsbasedonabundancesshowastrongerresponsetoclimatechangethanwhenbasedonoccupancy(Lindströmetal.,2013;Virkkala&Lehikoinen,2014).
4.3 | Threats of climate change
WewereabletocalculateEuropeantrendsinoccupancyfor87spe‐cies (88%of speciesoccurring inourdata set). Fifty‐fiveof thesespecieshaveincreasedfrom1990to2015,while32haveremainedstable and none have declined. This is a remarkably positive out‐come, given the fact that the conservation status ofmany fresh‐waterorganisms isknown tohavedeterioratedglobally (Collenetal.,2014;Dudgeonetal.,2006).Althoughwerecognizethatsomespecies with a stable trend in occupancy (distribution) may havedeclinedinabundance(populationsize),weconsideritunlikelythatthiswouldchangetheoverallpictureofrangeexpansion,giventhatoccupancyandpopulationtrendsshowbroadsimilarity(VanStrienetal.,2010).
Nexttothepositiveeffectsofclimatechangeforwarm‐dwellingspecies,recentimprovementsinwaterqualityandtheexecutionofwetland restoration projects are likely to have contributed to therecoveryofdragonflies inat leastsomeof the regions (Parkinson,Goffart,Kever,Motte,&Schott,2017;Termaatetal.,2015).
Jaeschke, Bittner, Reineking, and Beierkuhnlein (2013) com‐binedclimatescenarioswiththeassumeddispersalabilitiesofsixspecies,topredictchangesintheirEuropeandistributionsby2035.Theirmodelpredictedastrongdeclineforfivespecies(Coenagrion mercuriale, −50%;C. ornatum, −67%;Leucorrhinia albifrons, −39%;L. caudalis,−58%;Ophiogomphus cecilia,−24%)andanincreaseforone species (L. pectoralis, +21%). These predictions are in sharpcontrastwiththeresultsofourstudyovertheperiod1990–2015,aswefoundstable trends inEuropeanoccupancy forC. mercuri-ale,L. caudalis and O. cecilia,andincreasingtrendsforC. ornatum,L. albifrons and L. pectoralis (SupportingInformationDataS1).Weexplainthesedifferencesbytheestimationsofmaximumspeciesdispersal abilities applied by Jaeschke et al. They used the ob‐servedmaximumdispersaldistancesmentioned in the literature,whichrefertoobservationsfromcapture–mark–recapturestudies.Thesestudiesmaygiveanestimationofdistancescoveredbythemajorityofthestudiedpopulation,butundoubtedlymissdispersaleventsbyindividualsovermuchlongerdistances(seealsoSuhling,Martens, & Suhling, 2017), leading to a severe underestimationof maximum dispersal abilities. These extreme dispersal eventsmayberareandseldomnoticed,but theydeterminethepaceatwhichspeciesdistributionsmayexpand.Nexttoannualestimatesofoccupancy,occupancymodelsalsoprovideannualestimatesofpersistenceandcolonization.Theseparametersmaybemore in‐formativeforfutureresearchontheeffectofvariationinspecies’dispersalabilities.
ThenotionthatEuropeandragonfliesaregenerallydoingratherwellanddonotappeartobegreatlyharmedbyclimatechange,doesnot apply to all species, nor to all regions. Some species, such asthearcticSomatochlora sahlbergiandthealpineS. alpestris, are in a “deadendstreet,”astheycannotshifttheirrangefurthernorthortoahigheraltitude(DeKnijfetal.,2011).Othercold‐dwellingspe‐cies,suchasCoenagrion hastulatum,aredoingwellinSweden,whilebeing threatened inmore southern regions. Furthermore, indirect
| 13TERMAAT ET Al.
effectsofclimatechangemayaffectdragonflies.Desiccationofhab‐itatssuchassmallstreamsandpondsisathreattoseveralspecies(Kalkmanetal.,2010,2018),especiallyinsouthernEurope,whichisanunderrepresentedregioninourstudy.Also,changesincommu‐nitiesincentralandnorthernEuropemayleadtomoreinterspecificcompetitionbetweendragonfly species, possibly threatening indi‐vidual species in the future.Ultimately, dragonflies across Europefacemultiplethreatsotherthanclimatechange,suchashabitatdeg‐radationanddestruction,eutrophication(Kalkmanetal.,2010)andexposure to pesticides (Jinguji, Thuyet,Uéda,&Watanabe, 2013;VanDijk,VanStaalduinen,&VanderSluijs,2013).Therelativecon‐tributionofthesedifferentenvironmentalchanges largelyremainstobeestablished.
4.4 | Trends from distribution data
WebasedourtrendcalculationsonreadilyavailabledistributiondatafromtenEuropeanregions,usingoccupancymodelstoaccountforimperfectdetection.Theserecordsallowedustoassessoccupancyindiceswithouttheneedforastandardizedfieldworkprogramme.Assuch,ourmethodimmediatelyinformsaboutdistributiontrendsandmayserveasan“earlywarningsystem”forspecieswithadete‐rioratingconservationstatusand,byproxy,thequalityoffreshwaterhabitats(Oertli,2008).However,ourstudylacksdatafromeasternEurope andwe have insufficient data from southern Europe ade‐quatelytorepresentthatarea.Unfortunately,18outof19dragonflyspeciesconsideredtobethreatenedatEuropeanlevelareconfinedto southernoreasternEurope (Kalkmanetal.,2018).Consideringthis, our European indices and trendsmay be biased to somede‐greeatpan‐Europeanlevel.However,dragonflydatasetsarerapidlygrowing inmanycountries, includingseveraleasternandsouthernEuropeancountries(Boudot&Kalkman,2015).Moreover,anetworkofEuropeanodonatologistshasexpandedoverthepastfewyearsand theusefulnessof aEuropeandragonflymonitoring scheme isgainingattention.WearethereforeconfidentthatEuropeanindiceswillbecomemorerobustwithfutureupdatesandwillhaveabettergeographicalcoverage.
4.5 | Future prospects
Overall, this studyhas shown thatdragonfliespresenta suitablespeciesgrouptogainbetterunderstandingofbiodiversitychangesandtheircauses, includingclimatechange,andthatsuitabledataneededfortheseanalysesarebecomingavailable.Dragonfliesmaytherefore satisfy the need for a biodiversity indicator based onfreshwaterinvertebrates(Feest,2013).Theyarelikelytorepresentothertaxawhichareprimarilywarm‐adapted.Usingopportunisticdataanalysedwithoccupancymodelsenablestheassessmentofspecies’distributiontrendsonbothregionalandEuropeanscale.Thesetrendsinformaboutthestateoffreshwaterhabitats,whichisurgentlyrequired(Darwalletal.,2018).Hence,wesuggestadd‐ingdragonfliesasanindicatorgrouptotheEuropeanbiodiversitymonitoring programme (European Environmental Agency, 2012),
toinvestintheextensionofaEuropeandragonflyrecordingnet‐work,andtoencouragethecentralizationofEuropeandragonflydistributiondata.
ACKNOWLEDG EMENTS
We thank all dragonfly observerswho contributed to this studybymakingtheirobservationsavailabletotheirregionaldataman‐aging organizations.Data from Swedenwere obtained from theSwedish SpeciesObservation System; this system ismaintainedbytheSwedishSpeciesInformationCentreatSwedishUniversityof Agricultural Sciences. Data from Britain were obtained fromtheBritishDragonfly Society Recording Scheme.Data from theNetherlands were obtained from the Dutch National DatabaseFlora and Fauna. Most records are currently collected throughthe online platforms Waarneming.nl and Telmee.nl. Data fromNorthRhine‐Westphaliawereobtained from thedatabaseman‐agedbytheWorkingGroupDragonfliesNorthRhine‐Westphalia(AK Libellen NRW). Data from Bavaria were obtained from the“DatenbankArtenschutzkartierung,”maintained by theBavarianStateMinistryoftheEnvironment.DatafromFlanderswereob‐tained from the Flemish Dragonfly Society. Data fromWalloniawere obtained from SPW/DGARNE/DEMNA‐Working GroupGomphusandNatagora‐observations.Mostdragonfly records inboth Flanders andWallonia are currently collected through theonline platformsWaarnemingen.be and Observations.be, whichare managed by Natuurpunt and Natagora. Data from FrancewereobtainedfromthedatabasemanagedbytheFrenchSocietyofOdonatology (SfO).Data fromAndalusiawere obtained fromthe databasemanaged by Red deObservadores de Libélulas deAndalucía(ROLA).DatafromCypruswereobtainedfromtheda‐tabasemanaged by theCyprusDragonfly StudyGroup; this da‐tabase includes records collected through the online platformObservation.org.VincentDevictorkindlyprovided thevaluesofCTItrendslopesforbirdsandbutterfliesfromhis2012paperasbasedonpresence–absencedata.EddieJohnisthankedforproof‐readingthetext.
DATA ACCE SSIBILIT Y
Aggregateddatausedforspecies’occupancymodellingareavailablefromDutchButterflyConservation(http://www.vlinderstichting.nl).ClimatedatausedforcalculationofSpeciesTemperatureIndicesareavailablefromhttp://www.worldclim.org.
ORCID
Tim Termaat https://orcid.org/0000‐0002‐5974‐7294
Arco J. van Strien https://orcid.org/0000‐0003‐0451‐073X
Roy H. A. van Grunsven https://orcid.org/0000‐0001‐8873‐1024
Geert De Knijf https://orcid.org/0000‐0002‐7958‐1420
Michiel F.WallisDeVries https://orcid.org/0000‐0003‐3808‐2999
14 | TERMAAT ET Al.
R E FE R E N C E S
Bertrand, R., Lenoir, J., Piedallu, C., Riofrío‐Dillon, G., de Ruffray, P.,Vidal,C.,…Gégout,J.‐C.(2011).Changesinplantcommunitycom‐positionlagbehindclimatewarminginlowlandforests.Nature,479,517–520.https://doi.org/10.1038/nature10548
Boudot,J.P.,&V.J.Kalkman(Eds.)(2015).Atlas of the European dragon-flies and damselflies.Zeist,TheNetherlands:KNNVPublishing.
Britton, A. J., Beale, C. M., Towers, W., & Hewison, R. L. (2009).Biodiversity gains and losses: Evidence for homogenisation ofScottishalpinevegetation.Biological Conservation,142,1728–1739.https://doi.org/10.1016/j.biocon.2009.03.010
Bush,A.,Theischinger,G.,Nipperess,D.,Turak,E.,&Hughes,L.(2013).Dragonflies: Climate canaries for river management.Diversity and Distributions,19,86–97.https://doi.org/10.1111/ddi.12007
Chen,I.C.,Hill,J.K.,Ohlemüller,R.,Roy,D.B.,&Thomas,C.D.(2011).Rapid range shifts of species associated with high levels of cli‐mate warming. Science, 333, 1024–1026. https://doi.org/10.1126/science.1206432
Clavero,M.,Villero,D.,&Brotons,L.(2011).Climatechangeorlandusedynamics: Do we know what climate change indicators indicate?PLoS ONE,6,e18581.
Collen, B., Whitton, F., Dyer, E. E., Baillie, J. E. M., Cumberlidge, N.,Darwall, W. R. T., … Böhm, M. (2014). Global patterns of fresh‐water species diversity, threat and endemism. Global Ecology and Biogeography,23,40–51.https://doi.org/10.1111/geb.12096
Corbet, P. S. (1999). Dragonflies: Behaviour and ecology of Odonata. Colchester,UK:HarleyBooks.
Darwall,W.,Bremerich,V.,DeWever,A.,Dell,A.I.,Freyhof,J.,Gessner,M.O.,…Weyl,O.(2018).TheAllianceforFreshwaterLife:Aglobalcalltouniteeffortsforfreshwaterbiodiversityscienceandconser‐vation. Aquatic Conservation: Marine and Freshwater Ecosystems,28,1015–1022.
Davey, C. M., Devictor, V., Jonzén, N., Lindström, Å., & Smith, H. G.(2013). Impact of climate change on communities: Revealing spe‐cies'contribution.Journal of Animal Ecology,82,551–561.https://doi.org/10.1111/1365‐2656.12035
DeKnijf,G., Flenker,U.,Vanappelghem,C.,Manci,C.O., Kalkman,V.J.,&Demolder,H. (2011).The statusof twoboreo‐alpine species,Somatochlora alpestris and S. arctica, in Romania and their vulner‐ability to the impact of climate change (Odonata: Corduliidae).International Journal of Odonatology,14,111–126.
Devictor,V.,Julliard,R.,Couvet,D.,&Jiguet,F.(2008).Birdsaretrackingclimatewarming,butnotfastenough.Proceedings of the Royal Society of London B: Biological Sciences,275,2743–2748.
Devictor, V., Van Swaay, C., Brereton, T., Chamberlain, D., Heliölä, J.,Herrando,S.,…Jiguet,F.(2012a).Differencesintheclimaticdebtsofbirdsandbutterfliesatacontinentalscale.NatureClimate Change,2,121–124.
Devictor, V., Van Swaay, C., Brereton, T., Chamberlain, D., Heliölä, J.,Herrando,S.,…Jiguet,F.(2012b).ReplytoRodriguez‐Sanchezetal.Nature Climate Change,2,638–639.
Dijkstra, K.D. B.,Monaghan,M. T., & Pauls, S. U. (2014). Freshwaterbiodiversity and aquatic insect diversification. Annual Review of Entomology,59,143–163.
Dudgeon,D.,Arthington,A.H.,Gessner,M.O.,Kawabata,Z.‐I.,Knowler,D.J.,Lévêque,C.,…Sullivan,C.A.(2006).Freshwaterbiodiversity:Importance, threats, status and conservation challenges.Biological Reviews,81,163–182.https://doi.org/10.1017/S1464793105006950
EuropeanEnvironmentalAgency(2012).Streamlining European biodiver-sity indicators 2020: Building a future on lessons learnt from the SEBI 2010 process.EEATechnicalreportNo.11/2012,Copenhagen.
Feest,A. (2013).Theutilityof theStreamliningEuropeanBiodiversityIndicators2010(SEBI2010).Ecological Indicators,28,16–21.https://doi.org/10.1016/j.ecolind.2012.10.015
Franco, A. M. A., Hill, J. K., Kitschke, C., Collingham, Y. C., Roy, D.B., Fox, R., … Thomas, C. D. (2006). Impacts of climate warm‐ing and habitat loss on extinctions at species' low‐latitude rangeboundaries. Global Change Biology, 12, 1545–1553. https://doi.org/10.1111/j.1365‐2486.2006.01180.x
Gregory,R.D.,VanStrien,A.,Vorisek,P.,GmeligMeyling,A.W.,Noble,D.G.,Foppen,R.P.B.,&Gibbons,D.W.(2005).Developingindica‐torsforEuropeanbirds.Philosophical Transactions of the Royal Society B,360,269–288.https://doi.org/10.1098/rstb.2004.1602
Hassall,C.(2015).Odonataascandidatemacroecologicalbarometersforglobal climate change.Freshwater Science,34, 1040–1049. https://doi.org/10.1086/682210
Hickling,R.,Roy,D.B.,Hill,J.K.,Fox,R.,&Thomas,C.D. (2006).Thedistributions of a wide range of taxonomic groups are expand‐ing polewards. Global Change Biology, 12, 450–455. https://doi.org/10.1111/j.1365‐2486.2006.01116.x
Hickling,R.,Roy,D.B.,Hill,J.K.,&Thomas,C.D.(2005).AnorthwardshiftofrangemarginsinBritishOdonata.Global Change Biology,11,502–506.https://doi.org/10.1111/j.1365‐2486.2005.00904.x
Isaac,N.J.B.,VanStrien,A.J.,August,T.A.,DeZeeuw,M.P.,&Roy,D.B.(2014).Statisticsforcitizenscience:Extractingsignalsofchangefromnoisyecologicaldata.Methods in Ecology and Evolution,5,1052–1060.https://doi.org/10.1111/2041‐210X.12254
Jaeschke,A.,Bittner,T.,Reineking,B.,&Beierkuhnlein,C. (2013).Canthey keep up with climate change? – Integrating specific disper‐sal abilities of protected Odonata in species distribution mod‐elling. Insect Conservation and Diversity, 6, 93–103. https://doi.org/10.1111/j.1752‐4598.2012.00194.x
Jiguet,F.,Devictor,V.,Ottvall,R.,VanTurnhout,C.,VanderJeugd,H.,&Lindström,Å.(2010).Birdpopulationtrendsarelinearlyaffectedbyclimatechangealongspeciesthermalranges.Proceedings of the Royal Society of London B: Biological Sciences,277,3601–3608.
Jinguji,H.,Thuyet,D.Q.,Uéda,T.,&Watanabe,H.(2013).Effectofimi‐daclopridandfipronilpesticideapplicationonSympetrum infuscatum (Libellulidae:Odonata)larvaeandadults.Paddy and Water Environment,11,277–284.https://doi.org/10.1007/s10333‐012‐0317‐3
Kalkman, V. J., Boudot, J.‐P., Bernard, R., Conze, K.‐J., De Knijf, G.,Dyatlova, E., … Sahlén, G. (2010). European Red List of dragonflies. LuxembourgCity,Luxembourg:PublicationsOfficeoftheEuropeanUnion.
Kalkman, V. J., Boudot, J. P., Bernard, R., De Knijf, G., Suhling, F., &Termaat,T.(2018).DiversityandconservationofEuropeandragon‐fliesanddamselflies(Odonata).Hydrobiologia,811,269–282.https://doi.org/10.1007/s10750‐017‐3495‐6
Kampichler,C.,VanTurnhout,C.A.M.,Devictor,V.,&VanderJeugd,H.P.(2012).Large‐scalechangesincommunitycomposition:Determininglanduseandclimatechangesignals.PLoS ONE,7,e35272.https://doi.org/10.1371/journal.pone.0035272
Ketelaar, R. (2010). Recovery and further protection of rheoph‐ilic Odonata in the Netherlands and North Rhine‐Westphalia.Brachytron,12,38–49.
Lindström,A.,Green,M.,Paulson,G.,Smith,H.G.,&Devictor,V.(2013).Rapidchangesinbirdcommunitycompositionatmultipletemporalandspatial scales in response to recentclimatechange.Ecography,36,313–322.https://doi.org/10.1111/j.1600‐0587.2012.07799.x
MacKenzie,D. I.,Nichols,J.D.,Royle,J.A.,Pollock,K.H.,Hines,J.E.,&Bailey, L. L. (2006).Occupancy estimation and modeling: Inferring patterns and dynamics of species occurrence.SanDiego,CA:Elsevier.
Mason,S.C.,Palmer,G.,Fox,R.,Gillings,S.,Hill,J.K.,Thomas,C.D.,&Oliver,T.H.(2015).Geographicalrangemarginsofmanytaxonomicgroupscontinuetoshiftpolewards.Biological Journal of the Linnean Society,115,586–597.https://doi.org/10.1111/bij.12574
Menéndez, R., Megías, A. G., Hill, J. K., Braschler, B., Willis, S. G.,Collingham,Y.,&Thomas,C.D.(2006).Speciesrichnesschangeslag
| 15TERMAAT ET Al.
behindclimatechange.Proceedings of the Royal Society of London B: Biological Sciences,273,1465–1470.
Oertli,B.(2008).Theuseofdragonfliesintheassessmentandmonitor‐ingofaquatichabitats. InA.Córdoba‐Aquilar (Ed.),Dragonflies and damselflies: Model organisms for ecological and evolutionary research (pp.79–95).Oxford,UK:OxfordUniversityPress.
Ott,J.(2010).Thebigtreknorthwards:RecentchangesintheEuropeandragonflyfauna.InJ.Settele,L.D.Penrev,T.A.Geogiev,R.Grabaum,V.Grobelink,S.Klotz,M.Kotarac,&I.Kühn(Eds.),Atlas of biodiversity risk(pp.82–83).Sofia,Bulgaria:PensoftPublishers.
Parkinson, D., Goffart, P., Kever, D., Motte, G., & Schott, O. (2017).Réponsedesodonatesàlarestaurationdestourbièresardennaises.Forêt Nature,142,47–55.
Parmesan,C.,&Yohe,G.A.(2003).Globallycoherentfingerprintofcli‐matechangeimpactsacrossnaturalsystems.Nature,421,37–42.
Plummer, M. (2017). JAGS Version 4.3.0 user manual. Retrieved fromhttp://sourceforge.net/projects/mcmc‐jags/
Root, T. L., Price, J. T.,Hall, K. R., Schneider, S.H., Rosenzweig,C.,&Pounds,J.A.(2003).Fingerprintsofglobalwarmingonwildanimalsandplants.Nature,421,57.https://doi.org/10.1038/nature01333
Rosenberg,D.M.,&V.H.Resh(Eds.)(1993).Freshwater biomonitoring and benthic macroinvertebrates.NewYork,NY:Chapman&Hall.
Rosset, V., & Oertli, B. (2011). Freshwater biodiversity under climatewarming pressure: Identifying the winners and losers in temper‐ate standingwaterbodies.Biological Conservation,144, 2311–2319.https://doi.org/10.1016/j.biocon.2011.06.009
Roth,T.,Plattner,M.,&Amrhein,V.(2014).Plants,birdsandbutterflies:Short‐term responses of species communities to climate warmingvaryby taxon andwith altitude.PLoS ONE,9, e82490. https://doi.org/10.1371/journal.pone.0082490
Royle,J.A.,&Dorazio,R.M.(2008).Hierarchical modeling and inference in ecology.Amsterdam,TheNetherlands:AcademicPress.
Royle, J.A.,&Kéry,M. (2007).ABayesian state‐space formulationofdynamicoccupancymodelling.Ecology,88,813–1823.
Soldaat,L.L.,Pannekoek,J.,Verweij,R.J.T.,VanTurnhout,C.A.M.,&VanStrien,A.J.(2017).AMonteCarlomethodtoaccountforsam‐plingerrorinmultispeciesindicators.Ecological Indicators,81,340–347.https://doi.org/10.1016/j.ecolind.2017.05.033
Suhling,F.,Martens,A.,&Suhling, I. (2017).Long‐distancedispersal inOdonata:ExamplesfromaridNamibia.Austral Ecology,42,544–552.https://doi.org/10.1111/aec.12472
Termaat,T.,VanGrunsven,R.H.,Plate,C.L.,&VanStrien,A.J.(2015).StrongrecoveryofdragonfliesinrecentdecadesinTheNetherlands.Freshwater Science,34,1094–1104.https://doi.org/10.1086/682669
Thomas, J. A. (2005). Monitoring change in the abundance and dis‐tribution of insects using butterflies and other indicator groups.Philosophical Transactions of the Royal Society B: Biological Sciences,360,339–357.https://doi.org/10.1098/rstb.2004.1585
VanDijk, T. C., Van Staalduinen,M. A., & Van der Sluijs, J. P. (2013).Macro‐invertebratedeclineinsurfacewaterpollutedwithimidaclo‐prid.PLoS ONE,8,e62374.
VanStrien,A. J.,Termaat,T.,Groenendijk,D.,Mensing,V.,&Kéry,M.(2010). Site‐occupancy models may offer new opportunities fordragonflymonitoringbasedondailyspecies lists.Basic and Applied Ecology,11,495–503.https://doi.org/10.1016/j.baae.2010.05.003
Van Strien, A. J., Termaat, T., Kalkman, V., Prins, M., De Knijf, G.,Gourmand, A.‐L., … Vanreusel, W. (2013). Occupancy modelling
as a new approach to assess supranational trends using oppor‐tunistic data: A pilot study for the damselflyCalopteryx splendens. Biodiversity and Conservation,22,673–686.https://doi.org/10.1007/s10531‐013‐0436‐1
VanStrien,A.J.,VanSwaay,C.A.M.,&Termaat,T.(2013).Opportunisticcitizensciencedataofanimalspeciesproducereliableestimatesofdis‐tributiontrendsifanalysedwithoccupancymodels.Journal of Applied Ecology,50,1450–1458.https://doi.org/10.1111/1365‐2664.12158
VanSwaay,C.A.M.,Plate,C.L.,&VanStrien,A.(2002).Monitoringbut‐terfliesintheNetherlands:Howtogetunbiasedindices.Proceedings of the Section Experimental and Applied Entomology of the Netherlands Entomological Society, Amsterdam,13,21–27.
Virkkala,R.,&Lehikoinen,A. (2014).Patternsofclimate inducedden‐sityshiftsofspecies:Polewardshiftsfasterinnorthernborealbirdsthan in southern birds.Global Change Biology,20, 1–9. https://doi.org/10.1111/gcb.12573
Walther,G.‐R.,Post,E.,Convey,P.,Menzel,A.,Parmesan,C.,Beebee,T.J.C.,…Bairlein,F.(2002).Ecologicalresponsestorecentclimatechange.Nature,416,389.https://doi.org/10.1038/416389a
BIOSKE TCH
Tim Termaat is an ecologist at Bosgroep Midden Nederland(ForestryOwnersGroup): a cooperation for sustainable forestand nature management in Ede, the Netherlands. Previously,he worked at Dutch Butterfly Conservation (Wageningen, theNetherlands)onprojectsconcerningthemonitoringandconser‐vationofdragonflies(Odonata).Hismaininterestisunderstand‐ingtheoccurrenceofdragonfliesondifferentscales,fromlocaltoecozone,andthefactorsdrivingcurrentchanges.
Authorcontributions:T.T.,G.D.K.,U.B.,K.B.,K.‐J.C.,P.G.,D.H.,V.J.K.,G.M., F.P.,D.S.,C.V. andM.W. contributeddata for thismanuscript.M.P. helpedwith data handling.G.v.d.T. calculatedSpeciesTemperatureIndices.A.J.v.S.principallyconductedsta‐tisticalanalyses,withinputfromT.T.,R.H.A.v.G.andM.F.W.T.T.led the writing with significant input from R.H.A.v.G., M.F.W.,A.J.v.S.andG.D.K.Allauthorshavereviewedthemanuscript.
SUPPORTING INFORMATION
Additional supporting information may be found online in theSupportingInformationsectionattheendofthearticle.
How to cite this article:TermaatT,vanStrienAJ,vanGrunsvenRHA,etal.DistributiontrendsofEuropeandragonfliesunderclimatechange.Divers Distrib. 2019;00:1–15. https://doi.org/10.1111/ddi.12913