Geometric morphometric analysis of wing shape variation in ten European populations of Calopteryx splendens (Harris, 1782) (Zygoptera: Calopterygidae)

2009·3.9 MB·English

by  Sadeghi, S

Odonatologica38(4):341-357 December 1,2009
Geometric morphometricanalysis of wing
shape variationin ten Europeanpopulationsof
Calopteryxsplendens(Harris, 1782)
(Zygoptera: Calopterygidae)
S.Sadeghi D.Adriaens and H.J. Dumont
¹,
BiologyDepartment,FacultyofSciences,StateUniversityofGent,Ledeganckstraat35,
B-9000Gent,Belgium
ReceivedApril7,2009 /ReviewedandAcceptedMay6, 2009
Thewingsof 10C.splendenspopulationswereexaminedbylandmark-based geo-
metricmorphometricanalysis. Subspecifictaxain thisgroup arecurrently basedon
wingspotsizein <S6.Here,thevariation inwingshapeandsizeisevaluated,totest
whethershapeisdifferentatapopulationlevel,and whetherthis has implicationsat
ataxonomiclevel.ItwasfoundthatGeometricMorphometriessuccessfullydiscrimi-
natespopulations:overallwingshapesignificantlydiffered betweenpopulationsbut
theresultswereonlypartlycompatiblewithtaxonomicstudiesbasedonwingspotsize.
Irrespectiveofwingspot, allpopulationsshowed differentiation inwingshapeeven
thoughnotinwingsize;4groupswererecognizedbased onwingshape:(1)Turkish!
population;(2)Spanish,Finnish,RussianandTurkish2populations;(3)Italian,Ger-
manandFrenchpopulations;(4)GreekandAlbanian populations.Ordination ofthe
populationsbasedonconsensusdataandclusteranalysisphenogramconfirmed such
apattern.The Spanishpopulation(C. xanthostoma),didnotshow astrong identity,
whiletheTurkish 1(C.s. waterstoni)wasquiteisolated.The Italianpopulation(C. s.
caprai)showed morerelation totheFrench (C.s. faivrei)and German populations
than toAlbanian and Greekpopulations.
INTRODUCTION
Calopteryxsplendens isawidespread damselfly, foundinmostofEurope, large
partsofSiberiaandmuchof westernandcentralAsia(MERTENS etal., 1992).
Thereisgreatvariationamongmalesinwingcoloration(seeSIVA-JOTHY, 1999).
Atsexualmaturity,theamountofpigmentation becomesfixed(HOOPER etal.,
1 Present address: Biology Department,Faculty ofSciences, Shiraz University, Shiraz, Iran; —
[email protected]
342 S.Sadeghi,D.Adriaens & H.J.Dumont
1999;CORDOBA-AGUILAR, 1993).Traditionally subspecific taxa havebeen
distinguished by thesizeandposition ofthe pigmented wing spot,and by(mat-
ing) behavior (MERTENS et al„ 1992; DUMONT etah, 2005). About a doz-
en ofsubspecies havebeen recognized. C. s. splendens occurs fromBritainand
southernScandinaviaovermostofnorthernandwestern Europe; C. s. xanthos-
toma(often consideredagood species) livesinsouthernFrance, northernItaly,
western andsouthernSwitzerland,the IberianPeninsulaand NorthAfrica; C.
s. caprai is in centralItaly and possibly MediterraneanFrance; C. s. balcanica
inhabitseastern MediterraneanEurope (ASKEW, 2004). Variousothernames,
such as 'C. s. intermedia, C. s.faivrei, C. s. taurica, C. s. tschaldirica, C. s. water-
stoni, C. s. cartvelica, C. s. amasina, C. s. erevanense, C. s. mingrelica refertopu-
tativesubspecies, all ofwhichare moreorlessgeographically confined, butoften
withoverlapping rangesand strongvariationinwing spotsize.
Fewauthorsdenyacorrelationbetweendegreeofwingpigmentation andmale
mating orterritorialsuccess (HOPEMAN& ABRAMSON,2005); mostresults
suggestthat wing pigmentation is areliable signal ofmalequalityand plays a
rolein materecognition by females(GRETHER. 1996;TYNKKYNEN,2004;
HOPEMAN, 2005). Thus, visual discriminationbasedon wing pigmentation is
a majorcomponent ofreproductive isolationin Calopteryx species (WAAGE,
1975).Size and density of wing pigmentation isalso correlatedwithresistance
against disease and immunological condition(RANTALA et ah, 2000; SIVA-
-JOTHY,2000;CORDOBA-AGUILAR,2002; KOSKIMAKI etah, 2004).
Inspiteoftheseindications, the questionarises whethervariationinwingspot
isreally ataxonomically validdiscriminator.Thedelimitationofmany subspe-
cies is indeedfuzzy, anddinesare common(DUMONT etah, 1993). Moreover,
even ifvaluable,itwouldstill be meaningful totry andreinforceitby additional
markers. Here,weattempt translating morphological traitsin unambiguous nu-
merical data,andconfronttheresults withtraditionaltaxonomy.
Geometricmorphometries isarelatively newtechnique thathasgenerated val-
uableresults in manyfields of classic morphometry. A major advantage ofthe
geometric frameworkis acomprehensive use ofinformationabout shape, avail-
able from a set of landmarks(BOOKSTEIN, 1996). Variationsin body shape
(and wings as part of the body) haveimportant fitness consequencesbecause
they canaffecttheability tooccupy habitatssuccessfully (GATZ, 1979;LOSOS
& S1NERVO, 1989), to prevail in predator-prey interactions(WALKER. 1997;
NAGEL&SCHLUTER, 1998),andtoreproduce successfully. So, wecanexpect
selection to actuponwingphenotype.
Insect wings havebeenthe subject of geometricmorphometricanalysis inthe
past (ROHLF& SLICE, 1990; BAYLAC&DAUFRESNE. 1996); they are es-
pecially attractive because they can be treated with biological realism in only
two dimensions. Wing morphometries can help to characterize populations
withinaspecies, asshownby theanalysis of geographic variationin populations
Wingshapevariation in EuropeanCalopteryxsplendens 343
ofDrosophila lummei(HAAS & TOLLEY, 1998), Drosophila serrata (HOFF-
MAN& SHIRRIFS, 2002) andScythrisobscurella(Lepidoptera) (ROGGERO
& d’ENTREVES, 2005). Wings also proved useful to study complexes of spe-
cies, forexample in Diptera (DE LARIVA et al.,2001), orexamine theeffects
of hybridization, such as inApis meliferasubspecies (SMITH etah, 1997).The
wingvenationpatternofOdonata(like thatofmostotherflyinginsects) haslong
provided studentsofOdonatawitha rich source of diagnostic characters atall
taxonomiclevels(REHN, 2003).
In this study we use landmark-basedgeometric morphometries methodto
quantifyand analyze wingmorphological featuresinten European andAnato-
lianC. splendens populations. Thecentralaimof ourstudywas toevaluate wing
shape variation, testing thepossible useof wingshape patternsforintraspecific
taxonomy, andattemptingto distinguisheffectsoflocal adaptation fromtaxon-
linkedmorphological differences.
MATERIALANDMETHODS
Specimensweredriedorpreservedin70%ethyl alcohol.Sincethenumber ofspecimensfromone
location wasnotenoughforastrongstatisticalresult,in somecasesweused twogeographicallyvery
close adjacentpopulationswith nosignificant dilferencebetween the means(F&T test,p>0.1)as
asinglepopulation(Tab.I,Fig. 2).Inall,weused 344specimens from 10Europeanlocalities. For
simplicitywenamedthepopulationsaftertheircountryoforigin.
TableI
Number ofspecimens andsamplinglocalities
Population No. Localities
(country)
Albania 29 - MesopotamneartoDelvine,5-VI-1993,(39:59N, 20:04E)
27 -Tirana,Albania,6-VI-1993,(41:20N, 19:49 E)
France 16 - Sissonne,9-V-1993,(49:34N,03:53 E)
20 -Canal aubord duRhone, Gard, 12-VII-2004
Greece 25 - Saulopoulo,loaninna,13-VI-1993,(39:44N,20:53 E)
13 - Eleftheri,Thesprotia,Greece,25-V-1998,(39:18N,20:25E)
Spain 15 - Cuenca,14-VII-1988,(40:04N,02:08W)
9 -Alcaniz,20-VI1I-1991,(41:03N,00:09 W)
Turkey 1 22 - Derecik,nearTrabzon,Turkey, 19-VIII-1988,(41:00N,39:43 E)
7 - Besikduzu,WTrabzon,Turkey,20-VI1I-1988,(41:02N, 39:13 E)
Finland 27 - Kitee,SEFinland,13-VIII-2004,(62:10N, 30:08 E)
Germany 32 - Gerlenhofen,Ulm,Germany,31-VIII-1993,(48:20N, 10:04E)
Italy 16 - Casteldi Sangero,Italy,30-V-1993,(41:47N, 14:07 E)
16 - Gildone-Campobasso,Italy,01-VI-1993,(41:33N, 14:39E)
Russia 31 - IsmaylowskyPark,Moscow, 12-VII-1989,(57:44N, 37:37E)
Turkey 2 19 - Golderesi-kemer,Fethiye,Mugla,21-VII-87, (36:39N, 29:22 E)
20 - 9km.WestofSeki,Mugla,21-VII-87,(36:24N,29:13 E)
344 S.Sadeghi, D.Adriaens &H.J.Dumont
Theanteriorleftwingofeachmalespecimenwasscannedonaflatbedtablescanner(AgfaSNAP-
SCAN 1236)asadigitalRGBcolorimagewith400dpiresolution. Beforescanning,driedspecimens
weresoaked in 70%ethylalcohol for 20minutes andwereleft to dryatroom temperatureto gain
flexibilityforhandling.Damagedwingswereexcluded fromanalysis.
WINGSTRUCTURE. - Wingvenationincalopterygidsisdense,withnumerousantenodal and
postnodalcross-veins and ananal fieldthat iscomposedofvariable number ofcells. A dense and
variable venation,alongwith darkandshinymetalliccoloronwingsin Calopteryxsplendens(male),
restrictsourchoices forlandmarkingandmakesitdifficulttofindidentical landmark placesonwing
pictures.
We collected 19homologouslandmarks on the nodes of wingvenation (Fig. 1) usingtpsDig2
(ROHLF,2006).Thelandmarks representedwingshape,and included allthose thatcould reliably
beidentified.Theywerechosenattheintersection ofwingveins inthehyalinepart ofthewingand/
oratthe wingedge.Hence, 18landmarks canbeconsidered astypeIlandmarks and one(LM4)as
atypeII(BOOKSTEIN, 1991).Thefollowinglandmarks wereused(seeFig. 1):(1)costa- subcosta
connection,(2)nodus,(3)radius2 - wingmarginconnection,(4)distal tipofthewing,(5)medius-
wingmarginconnection,(6)cubitus1 -wingmarginconnection,(7)cubitus2-wingmarginconnec-
tion,(8)ventraltipofanaltriangle,(9)proximalapexofanaltriangle,(10, 11& 12)distalanglesof
arculus, (13)originofmedius,(14)originofcubitusl,(15)distal angleofanaltriangle,(16)origin
ofcubitus2,(17)Ru+Rsand R+M connection,(18) originofIR3 (thirdinterradial),(19)originof
radiusl. Thenomenclature ofthewingvenationusedfollows DUMONT(1991).
Toestimatethedigitizationerroranddefiningthoselandmarks thatcanbedigitizedwiththehigh-
estaccuracy, weusedprotocolbyAdriaens,http://www.fun-morph.ugent.be/Research/Methodology/
Morphometrics.pdf). Thisprotocol allows quantifyingtheextent ofdigitization (andorientation)
errorwith respecttothevariation observed inthepopulations(basedonasubsample).Aftertesting,
approximately11.6%oftheobservedvariation was duetodigitization(andorientation)error (due
tonatureofthewingsandtheapproachfollowed here,orientationerrorcanbeconsidered tobe ab-
Fig. 1.C. splendenswaterstoni:(a)nomenclature ofwingveins; — (b)landmarksposition
Wingshapevariation inEuropeanCalopleryxsplendens 345
sent).We alsocompareddifferentcombinations of24and20landmarks in twoseparatepilotstud-
iestoselectthe combinationwith thehighestnumber oflandmarks thatshowedthelowest amount
ofnoise. Assuch, these 19landmarks werethebestcombination.
AGeneralized Procrustesanalysis(GPA)wasperformedtosuperimposelandmark configurations
usingleast-squaresestimatesfortranslation androtation parameters(ADAMSetal.,2004).GPA is
animportantprocedurebecauseitremovesvariation duetodifferences intranslation,orientation,
and size, and superimposestheobjectsin acommoncoordinate system.Shapedistances between
GPAalignedspecimensin Kendall shapespacearesubsequentlyprojectedintoaEuclidean space
that istangenttothisKendall’s shapespace(ROHLF, 1999;SLICE,2001). Tovisualize wingshape
differenceswegeneratedthin-platesplinedeformationgrids(ROHLF,2004).
STATISTICALANALYSIS.— Correlationsbetween theProcrustesandtangentshapedistances
werecalculated usingtpsSmallsoftware (ROHLF, 1998),toensurethatthe amountofshapevari-
ation intheoriginaldata setwasadequatelyrepresentedafterprojectioninthe tangentspace. The
sampleshowedperfectlycorrelated distances (RA2= 1.000),allowingfurtherstatistical testingusing
theprojecteddataset (seeROHLF, 1998).
Asameasureofoverallsizevariation ofthewings,thecentroid size (thesquarerootofthesum
ofthesquaredinterlandmark distances)wascalculated foreachpopulation(Fig.3)(BOOKSTEIN,
1991; 1996, SLICE et al., 2007;ZELDITCH, 2004). Centroid size wascalculated usingtpsRelw
(ROHLF,2007)andtestedfornormalityusingtheShapiro-Wilktest.Allpopulationswerenormally
distributed (p>0.05).Leven’stest wasused totest forhomogeneityofthevariance (MILLIKEN&
JOHNSON, 1984).Aone-wayANOVAwasconductedonwhole datasettotestsignificantcentroid
size differences between and within populationsand apost hoctest (Tukey’s test)defined pairwise
differencesincentroidsizeofpopulations(SOKAL&ROHLF, 1995).
Foranalyzingwingshapevariation within andamongpopulations,principalcomponentanalysis
Fig. 2.Distribution mapofthepopulationsstudied andthumbnailpicturesoftheiranteriorleftwing.
Mapsource:http://z.about.com/d/geography/l/0/2/L/eurasia.jpg
346 S.Sadeghi,D.Adriaens&H.J. Dumont
(PCA)and canonical variates analysis(CVA) wereconducted onthe landmark coordinates dataset,
PCAasatool forexploringpatternsofvariation within populationusingvariance-covariance ma-
trixand CVAforanalyzingand testingdifferencesbetween populations.
AMANOVAandtwo-grouppermutationtests(2000permutations)wereperformed,withsquared
Mahalanobis distancecalculated onthelandmark coordinates data setusingPAST(HAMMER&
HARPER,2007)todeterminewhether geographicallyseparatedpopulationsfrom different coun-
triesdifferin wingshape.Inthisanalysisthecriteria Wilk'slambdaisusedand whenthe MANOVA
showed significantoveralldifference between groups, theanalysis proceededbypair-wisecompari-
sons(post-hoc)bypairwise Hotelling’stests. We alsoperformedtwo-grouppermutationtest (2000
permutations)andextracted squaredMahalanobis distance foreverytwopopulations.
We also measured consensus shapedata (mean shape)of the separatepopulationstoillustrate
ordination ofthe shapes’consensusby arelativewarp ordination plotusingtpsSmaland tpsRelw
(ROHLF,2003,2007). Inall figures,theconsensuslandmark configurations(i.e. theconfiguration
thatisobtainedbyaveragingspecimenlandmarkcoordinates inthegeneralizedprocrustesanalysis)
areusedasareference,with thesubtledifferences betweenpopulationsshownasathreetimesexag-
geratedtransformation grid.
Pheneticrelationshipsbetweenthetenpopulationswerealsoinvestigatedthroughaclusteranalysis
usingthe matrixofprocrustes distancesbetween pair-wise populationconsensusconfigurations. A
UPGMA(Unweightedpair-groupmethodwith arithmetic means)waschosenastheyled toaphe-
nogramwith thelargestcopheneticcorrelationstotheoriginalprocrustes distancematrix (ROHLF,
2002).
All statistical analyseswereperformedin PAST(PaleontologicalStatistics) version 1.57(HAM-
MER,2007)and SPSS(version 15.0.1,2006).Graphicaldepictionsofwing-shapetransformations
in tpsSplin(ROHLF, 2004)and IMP(SHEETS,2000)and of thephenogramsweregeneratedin
NTSYSpc(Version2.1,ROHLF,2000).
Fig. 3.MeancentroidsizeoftenEuropeanpopulations.Abbreviation ofpopulations:Al:Albanian;
Fi:Finnish;Fr:French;Ge: German;Gr:Greek;It:Italian;Ru: Russian;Sp:Spanish;TI:TurkishI:
T2:Turkish2 populations.
Wingshapevariation in EuropeanCalopteryx splendens 347
RESULTS
WINGSIZEVARIATION
Themeancentroidsizeofthepopulations singled outtheGreekandAlbanian
populations as thosewiththe largest, and the Spanish population as thatwith
smallest wingsize (Fig. 3).
AShapiro-Wilk testrevealedanormaldistributionofallpopulations (p>0.05)
and Levene’s testshowed significant homogeneity of variances (p<0.05) based
on means. A one-way ANOVA ofmean centroidsizes showedsignificant dif-
ferences betweeninter-andintera-population variations(F =63.47,p=0.000).
TheresultsofTukey HSD as apost hoc teston centroidsizes are summarizedin
TableIIas pair-wisedifferences.Theresultof homogeneous subsetsof centroid
sizesextracted fromTukey HSD identifiedfourgroups withsignificantly differ-
ent centroidsize; theGreekandAlbanianpopulations (first group) hadthebig-
gestsize, theSpanish, Turkish1 andTurkish2populations werethesmallest(sec-
ondgroup). TheItalian, FrenchandGermanpopulations (group 3)hadmedium
wingcentroidsize, andtheFinnishandRussianpopulations (fourth group)were
situatedbetweenthe largest andmediumsize groups(Fig. 3).
Nosignificant differenceincentroidsize was foundamongmembersofeach
group.Wefoundthatgroupone(Albania andGreece) was highly significant in
differencewithotherpopulations in meancentroidsize.
WINGSHAPEVARIATION
PCAorrelativewarpanalysisofallspecimens explained 65.7%ofshape vari-
ation withinsamples by thetwo firstPCAaxes extractedfromthe variance-cov-
TableII
ResultsofTukeyHSD(post-hoc)test onwingcentroidsize, insignificantvaluesbolded — *<0.05;
- **<0,01
Country A1 Fr Gr Sp T1 Fi Ge It Ru T2
A1 -
Fr 48.54" -
Gr 3.26 51.80" -
Sp 71.06" 22.52" 74.33" -
T1 62.40" 13.86 65.67" 8.66 -
Fi 40.07” 8.46 43.34** 30.98” 22.32" -
Ge 52.82" 4.28 56.09" 18.24** 9.58 12.74 -
It 51.04" 2.50 54.31" 20.02* 11.36 10.96 1.78 -
Ru 36.98" 11.56 40.24" 34.08" 25.42** 3.10 15.84 14.07 -
T2 65.6" 17.10* 68.91** 5.42 3.24 25.56** 12.82 14.60 28.67** -
348 S.Sadeghi,D.Adriaens&H.J.Dumont
ariancematrix(PC1 explains 50.5%andPC2,15.2%). Atotalofuptoseven axes
wererequiredto covermorethan90%oftheshape variation.AtfirstglancePC1
vsPC2scatterplot showedanoverlap ofmostpopulations andtheirdistribution
onthis plot (Fig. 4) did not correspond with geographical differencesbetween
populations. However, PCA weakly showed some population shape variation
(Turkish 1 population from Albanian, Spanish, Russian, andTurkish2 popula-
tions). Furthermore, through PCI the vectors on landmarksshows inclination
of landmarks3and 4to posteriorand landmarks5,6, 7 toanteriorpartof the
wingwhichleadsto increasedistancebetweenlandmarks2and3inonesideand
between4and 5 inotherside. Through PC2the landmarks3to8 inclineto the
centralpartofthewingwhichleadsto adeclineinlength andwidthofthewing
(see Appendix 2). Theselandmarks referto connecting point of Radius2 with
posterior wing margin, distaltipof thewing,andconnectionof Medius, Cubi-
tus1 andCubitus2 and anal triangle tip with ventral wing margin respectively
(Fig. 1).
MANOVAfoundasignificant overalldifferencebetweenpopulations [Wilk's
lambda. 0.0076, F: 6.47, p =0.0000). Hotelling’s pair-wise comparisons [post-
hoc)showedhighly significantdifferencesbetweenallpopulations exceptbetween
FrenchandGermanpopulations whichwere atalowsignificance level(p<0.05).
Squared Mahalanobisdistanceofpopulations confirmedtheseresults(Tab. III).
Fig. 4.PCIvs PC2screenplot.For abbreviations seeFigure 3.
Wingshapevariation in EuropeanCalopteryxsplendens 349
TableIII
Hotelling’spair-wisecomparisons(F-scores andpvalues)and Mahalanobis distances andpvalues
for 10populationsonupperand lowerdiagonalrespectively - *<0.05; - ** <0.01
Country Al Fr Or Sp Tl Fi Ge It Ru T2
Al 0 11.4** 3.0** 14.8** 30.6** od «* 10.6** 9.7** 8.3** 13.1**
Fr 0.408** 0 13.8** 6.8*» 14.7** 2.6** 1.8* 3.5** 3.5** 5.9**
Or 0.302** 0.586** 0 12.8** 38.1** 14.7** 11.7** 16.7** 10.1** 14.3**
Sp 0.661** 0.828** 0.821** 0 15.7** 9.4** 7.1** 9.2** 8.8** 6.6**
Tl 0.952** 1.129** 1.397** 1.548** 0 10.3** 8.5** 13.7** 15.6** 15.1**
Fi 0.696** 0.539** 1.01** 1.237** 1.271** 0 3.1** 6.0** 3.7** 6.5**
Ge 0.519** 0.415* 0.703** 0.763** 0.973** 0.583* 0 2.9** 8.0** 4.8**
It 0.537** 0.466** 0.860** 1.313** 1.19** 0.804** 0.498** 0 7.2** 6.9**
Ru 0.455** 0.552** 0.598** 1.014** 1.28** 0.680* 0.815** 0.801** 0 5.8**
T2 0.519** 0.562** 0.467*• 0.718** 0.971** 0.707** 0.547** 0.617** 0.558** 0
Everycoupleofpopulations significantly differed(/?<0.05) insquared Mahalano-
bis distances.
Differences betweenpopulations were not well illustrated by a CVA plot. A
scatter plot ofCV1 (eigenvalue 3.206) vs. CV2(eigenvalue 1.455)showedagen-
eralpatternsimilarto PCAbutwithlessoverlap. CV1 revealedTurkish 1 promi-
nently, AlbanianandGreek(together),and Italianpopulations moreor lessdif-
fered inwingshape fromothers(Fig. 5).
Moreover,considering CV2, ItalianandGermanpopulations separated inshape
fromRussian, Spanish andTurkish2(together) populations. Otherpopulations
moreor lessoverlapped eachother(Fig. 5). Therewere seven distinctCVs (Ap-
pendix 1)andotherCVAplotsalso showedgood separation fromotherpopula-
tions. Cvl vs CV4welldefinedshape differencesbetweenSpanish andTurkish2
and CV2 vs. CV3 (eigenvalue 0.8595) and CV4 (eigenvalue 0.6239) plots (not
shownhere)showedgooddifferencesbetweenFinnishandRussianpopulations,
and between Turkish2 and Finnish, Spanish and French, Italianand Russian
populations.
Therewas a high correlationbetweenCV1 and PC2 (Spearman’s rho corre-
lation=0.65). Thiscorrelationrefers to landmarkssituated in themiddlepart
of the wing, that show anexpansion between landmarks2, 7,8, and 16 anda
contractionbetweenlandmarks 10-18intheanteriorpart, apartofthewing in-
formativein taxonomicstudies.
UPGMAclusteranalyses andalso relativewarpordinationofshape consensus
showedthesamepopulation relationships (Figs6,7).Inarelativewarpordination
plot,GreekandAlbanian(together), Italian,Turkish1 andTurkish2populations
showed thelargest differences, whileSpanish, Russian and, Finnishpopulations
on theone handand, FrenchandGermanpopulations ontheotherhandwere
close toeach other. TheAlbanianpopulation was closerto Greek population,
350 S.Sadeghi, D.Adriaens& H.J,Dumont
and theFrench population fell betweenthe Spanish and Germanpopulations.
Russian, Finnish, Turkish2 andSpanish populations andFrench, Germanand
Italianpopulations areclosely related. AlbanianandGreek populations as third
andTurkish1 population as fourthgroupsettledat twoextremesofthe X-axis.
These relationships are quitecompatible withthe UPGMA dendrogram, esti-
matedfromProcrustesdistances.
The wing shape transformationgrids adjacent to the dendrogram depicted
shape similaritiesand differences between populations. Theseillustrated trans-
formationsofthewing landmarkswhichled to such aclustering.
DISCUSSION
Geographic variationinwide-rangingspecies isubiquitous(MAYR, 1963)and
oftenreflectsadaptations ofapopulation tolocalenvironmentsand bioticfactors
(RICKLEFS& MILES, 1994;McPEEK, 1990);suchadaptations areexpected to
endupin(sub)speciation, butitis not alwayseasy to determinewhenthisproc-
esshastakenplace. One difficultyis thechoiceof charactersuponwhichtobase
sucha decision. Manyaspects offlightperformance correlatedto fitnessdirectly
or indirectly, are determinedbywing form.Apropercomprehensive analysis of
wing shapemay thusprovideaninsight in phenotypic variationrelated toflight
Fig,5.CVA plot,CV1 (eigenvalue3.434)vsCV2(eigenvalue1.925),For abbreviations seeFigure3,