Odonatologica37(2):131-144 June I,2008 Populationgenetic differentiationin three sympatric damselfly species ina highly fragmentedurbanlandscape (Zygoptera: Coenagrionidae) M.Sato¹*Y.Kohmatsu²,M.Yuma³ and Y.Tsubaki¹ 1CenterforEcologicalResearch,Kyoto University,509-3,Hirano 2,Otsu,Shiga,520-2113,Japan 2Research InstituteforHumanityand Nature,457-4,Motoyama,Kamigao,Kitaku, Kyoto,603-8047,Japan 3DepartmentofEnvironmental Solution Technology,FacultyofScienceandTechnology, Ryukoku University,Seta-Oe,Otsu,Shiga,520-2194,Japan ReceivedJuly24,2007/RevisedandAcceptedNovember 12,2007 TheAFLP (AmplifiedFragment LengthPolymorphism) techniquewasused to comparethelevelsofgeneticdiversityanddifferentiation amongParacercion calam- orum, Ischnura senegalensisand I. asiaticaand tocompare thegeneticstructureof populationsfound in highlyfragmentedurban habitatstopopulationsin relatively continuousruralhabitats. Forall3spp., highgeneticdiversitywasfoundinbothareas. However,populationgeneticdifferentiation amongurban populationswasapproxi- matelytwice that ofrural populations,indicatingthat movementsbetween habitat patchesaremorerestrictedin urbanareas,probablyduetohuman disturbancesthat mayfunctionasbarriers. Inter-specificdifferencesregardinggeneticdiversityanddif- ferentiation arefurtherdiscussed intermsofhabitat specificity. INTRODUCTION Habitatdestructionandfragmentation due to humanactivity are a common phenomenon occurring throughout theworldandisarguably thegreatestthreat to biodiversity (PRIMACK, 1998;WILCOVE et al., 1998). Thusroads, build- ings and towns often provide barriers that interferewith anorganisms’ move- ment betweentheremaining habitatpatches, causing population fragmentation (PRIMACK, 1998). Fragmented populations directlyimpactgeneflowwithinaspecies, leading toa *Correspondingauthor:MayumiSato,[email protected] 132 M.Sato,Y.Kohmatsu,M.Yuma& Y.Tsubaki lossofgenetic diversitywithineachpopulation andanincreasein genetic differ- entiationbetween populations (FRANKHAM, 2006). Thisgenetic force, along withdemographicandenvironmentalforces, coulddrivefragmented populations toextinction(GILPIN &SOULE,1986;CROOKS&SANJAYAN,2006).How- ever, despite a growing concern over theinfluenceof habitatfragmentation on genetic structure, fewpopulation genetic studieshave been conducted in urban landscapes, wherehabitats are expected to bemorefragmented andindividual movements between fragmented patches may bemore restricted than thosein rural landscapes. Ideally, population genetic structure presumably resulting fromhabitatfrag- mentationshouldbecompared to thatofacontrol site(i.e., unfragmentedhabi- tat)(VAN DONGEN etah, 1998) withthesamespatial scale attheintra-specific level.However,theuseofsuch acontrolsiteisoftenignored (WILLIAMSetah, 2003; SOFIAetah,2006). Furthermore,bothsitesshouldbelocatedinrelatively closeproximityinordertoeliminateany possible geographiceffectson dispersal such as climaticconditionsandaltitude.Moreover, inorderto detectlowgenetic differentiationsresulting fromasmallspatialscale,molecularmarkers withhigh- resolutionpowershould beemployed (OUBORGet ah, 1999). Therefore,find- ing aspecies inhabiting bothfragmentedandunfragmented habitatsoverasmall geographical rangeandusingamolecularmarkerwithhigh-resolutionpowerare important issues forexamining theeffectsof habitatfragmentation onpopula- tiongenetic structure. AmplifiedFragmentLength Polymorphism (AFLP) (VOSetah, 1995) isamo- lecular markerwhichhasbeenusedinthe fieldof molecularecology andevolu- tion since 1995. Although themethodhasbeenwidely employed inthestudy of plants, fungi, and bacteria, theadvantages of the method(short start-up time, nopriorsequenceknowledge andgood repeatability) shouldmakeita valuable tool to investigate thepopulation genetics ofwild animalspecies (BENSCH & AkESSON, 2005). Amajorconcernaboutthetechnique isthatmarkersaredom- inant, and it is impossible to distinguish homozygous (1/1) from heterozygous (1/0) genotypes(BENSCH& AKESSON,2005). Thisrequires previous know- ledgeoftheinbreeding coefficient(HOLSINGERetah,2002)and,therefore,the Hardy-Weinberg equilibrium is sometimesassumed to estimate allele frequen- cies. However, thelarge numberoflocicounterbalancesthis drawbackof domi- nant markers (PARSONS& SHAW,2001)as ifcanreveal subtlebutsignificant population structures (BENSCH& AKESSON,2005). We usedthethreeZygoptera species, Paracercioncalamorum, Ischnurasenega- lensis, and I. asiatica to compare levelsof genetic diversityand differentiation amongpopulations foundinhighlyurbanizedfragmented habitatrelativetopop- ulationsin continuousrural habitatandthereby determineany effectofhabitat fragmentation ontheirgenetic structure.Adultodonatesareknowntobesensi- tiveto thedestructionand modificationoftheiraquatic habitats(e.g. CARLE, Populationgeneticdifferentiationin sympatricZygoplera 133 1979;BROWN. 1991;CLARK & SAMWAYS, 1996;SUHLING etal„ 2006; SATO & RIDDIFORD, 2008), and this, along withtheirease ofcapture and wide-spread distributionmakethesethreespecies idealforsuch astudy.Allthree damselfly species occursympatrically intheKantoregion in Japananduselentic habitats, mainly occurring in agricultural pondsoftenconnectedby ditchesand rice paddies in rural areas,andin distantly spaced city park ponds in urbanar- eas. However, habitatspecificity seemsto differamong them.P. calamorumhas anarrowerselectionofhabitatwithrespect totherelativestructureandarrange- mentofvegetation atthewatersidecompared to theothertwo species, andofthe two congeneric species, I. senegalensis selects amore confinedrange ofhabitat typescompared to I. asiatica. Thesampling designallowed usto conduct both intra-specificcomparison betweentwocontrasting areas andinter-specific com- parisonamongthethreespecies since differencesinhabitatspecificitycouldhave profound influencesonthegenetic structure. MATERIALAND METHODS STUDYSITESANDSAMPLECOLLECTION — Wecollected adultP. calamorum,I. senega- lensis,and I. asiaticafromatotal of23different sites (11,10,and 13sitesforeach,respectively) in twocontrastingareasontheKantoPlain in Japan(Fig. I).Sincethe mainaim ofourstudy wasto examine theeffectsofhabitat fragmentationondamselflydispersal,wechose the most intensively urbanized areain Japan. 11 ofthe study sites werepondsincity parks in theTokyometropolitan area,andexceptforsite U11,theywerewithin the23wardsofTokyo,occupying muchofthecenter oftheTokyometropolitanarea.This areaisvery built-upwith business,commercial,and high-rise apartment buildings,which arepotentialphysical barriers for dispersaland hence resultin habitat fragmentation.The other 12siteswereselected inthe surroundingcountrysideasacontrol against the urban data; 10 ofthese werelocated in Ibarakiprefecture, the other twoontheborder ofIba- rakiandChibaprefectures.These studysiteswereeitherponds(natural,agriculturalorpark ponds) orrice paddiesand werewithin an areaapproximately 50to60km NE ofTokyo andmainlysur- rounded byrelativelycontinuousagriculturalland. Althoughthereissomeurbanization inthisarea, much oftheareahas still been used forrice paddies,lotus-cultivation,andarable lands due tothe closeproximityofLake Kasumigaura,which isthesecond largestlakein Japan.Irrigatedricepad- dies,withirrigation ditchesand shallow holdingponds,sustain manyspecies ofaquaticorganisms andareconsidered tobe artificialshallow wetlandsinthisrespect(NAKAMURA&SHORT,2001). Landcovertypeswithinapproximately 1500km2ofeachrural andurbanareacoveringall studysites aresummarized inTableI.Thespatialscales ofthetwostudyareasarecomparable.Theminimum and maximumlinear geographicaldistances betweenstudysiteswere4.5kmand 25.7kmintheur- ban areacomparedto4.6km and 27.4kmin therural area,respectively,for P.calamorum,8.6km and 27.0kminthe urbanareacomparedto7.0kmand25.1 km intherural area,respectively,forI. senegalensisand5.3kmand 38.9km in theurbanareacomparedto4.5km and38.2kmintherural area, respectively,forI.asiatica. Samplingwasconducted in JulyandAugust2004. Adult damsel- flies werecollected with netsand immediatelystored in99%ethanol inthefield. Anaverage of 15 individuals foreachsiteforeachspecieswerecaughtand usedfortheanalyses. DNA EXTRACTIONAND AFLPPROCEDURE — The head and thorax weregroundin liquidnitrogenasa basematerial fromwhich toextract totalDNA followingtheslightly modified cetyltrimethylammoniumbromide (CTAB)method suggestedspecificallyforAFLPanalysisforin- sect material(REINEKEetal., 1998).DNA wasextracted withchloroforrmisoamylalcohol(24:1), 134 M.Sato,Y.Kohmatsu,M.Yuma&Y.Tsubaki and theDNApelletswereethanol-precipitatedandthensuspendedinTEbuffer. DNAextractswere furtherpurifiedusingaQIAquick PCR Purification Kit (QIAGEN K.K.,Tokyo)accordingtothe manufacturer’s protocols. AFLPreactions wereconducted usingtheAFLPPlant MappingKit(AppliedBiosystems,Foster City, CA)foraregulargenomefollowingtheprocedures outlined inthemanufacturer’sinstructions with minormodifications. Tento 150ngofinitial DNAper individual weredigestedwith restriction enzymes (Mseland EcoRl;New England Biolabs,Beverly,MA)before ligationofrestriction site- Fig. 1.Location ofthe samplingsitesin urban (Tokyo: Ul-UII) and rural (Ibaraki and Chiba: R1-R12)areas. Populationgeneticdifferentiation insympatric Zygoptera 135 specificadaptors.Thepre-selective and selectiveamplifications werecarried outin aPerkin Elmer GeneAmpPCRSystem9600thermalcycler(AppliedBiosystems).Modifications included reduction ofthetotalvolume ofthesolution toahalfofthevolume suggestedin themanufacturer’sinstruc- tions,which was5.5pLforrestriction and adaptor-ligationand lOpLforpre-selective and selective amplifications.Inordertoremovetheinhibitorysubstancesforamplifications, 1.6pL ofArapdirect (ShimadzuCorporation,Kyoto, Japan)was added tothepre-selectiveamplificationreaction mix- ture.Twoprimercombinations (EcoRI-ACA/Msel-CTG,EcoRI-ACT/Msel-CAC)thatamplified areasonable number ofclearbandswereused. The fluorescence-labeled selectiveamplifiedproductswereresolved with aGene-Scan 500 ROX internal sizestandard(AppliedBiosystems),usinganABI3100automatedsequencer (AppliedBio- systems). Theelectrophoresisimageswereobtainedfrom each run,and peakswereassignedbase- pair sizes usingtheGENESCAN programversion 3.1 (AppliedBiosystems). OnlyDNA fragments ranginginsizefrom 50to500base pairswerediscerned andmanuallyscoredas 1(presence)or0(ab- sence)toassemble abinarydata matrix consistingofatotalof189,215,and203 bands foreach of theP.calamorum,I.senegalensis,and I.asiaticaindividuals sampled,respectively.TenDNAsamples from each specieswereamplifiedatleast twicetoverifythereproducibility ofthe AFLPtechnique, and onlyintense and unambiguousbands werescored. DATAANALYSIS —We assumed theHardy-Weinbergequilibriumtoestimategeneticdiversity and differentiation asnopriorinformation onthematingsystem ofthethree damselflieswasavail- able.Allele frequencieswereestimatedaccordingtoaBayesianmethod withnon-uniform priordis- tribution ofallelefrequencies(ZHIVOTOVSKY, 1999),usingthecomputerprogram AFLP-SURV 1.0(VEKEMANS,2002).Basedontheestimatesofallele frequencies,wedetermined geneticdiver- sityasNei’s unbiased expectedheterozygosity(NEI, 1973).We estimatedthegeneticdifferentiation amongpopulationswithin eachareausingtwo different methods. Wefirstcalculated Wright’sFsp thefixationindex,which isaparameterindicatinghowpopulationsaregeneticallydifferentiated ac- cordingtotheLynch andMilliganmethod(LYNCH&MILLIGAN,1994),usingAFLP-SURV 1.0 (VEKEMANS,2002).Themethodwasspecificallydevelopedforcalculatingdifferentiation andsuit- ableforusewithdominantdatatypes.ThesignificanceofFSTwastestedbycomparingtheobserved F with thedistribution ofthesetofF generatedby 1000randompermutationsfromindividuals. sr ST We alsoused the method thatdoesnotassumethe Hardy-Weinbergequilibrium,but instead,uses the Euclidean distance toestimatethe componentsof variance attributable to differencesamong populationsand among individuals within populations(EXCOFFIER etal., 1992).TheAnalysis ofMOlecular VAriance(AMOVA)(EXCOFFIER etal., 1992)wasperformedtocalculate 4> (F ST ST Table I Percentageofmajorlandcover typeswithin about 1500knr foreachurban (Tokyo)and rural (Iba- raki andChiba)area.Vegetationand landusedata derivedfrom 15 gridcells,each ofwhich isap- proximately10x10km oftheNational SurveyontheNatural Environment(BiodiversityCenterof Japan,1999),werere-categorizedintofourlandcover types.Urbanlandscapesincluderesidential ar- eas,factories,and buildings.AG-testshowed thatthecompositionoflandcover typesinthe urban areaissignificantlydifferentfromthat oftherural area(G=921.5,/?<0.001) Urban area Ruralarea Landcovertype Area(km2) %oftotal Area(km2) %oftotal Paddyfield/Wetlandvegetation 50.5 3.4 433.2 32.1 Arableland/Meadow 217.1 14.6 413.9 30.6 Urban landscapes 1002.0 67.3 233.8 17.3 Others 218.9 14.7 267.4 19.8 Total 1488.6 100 1348.4 100 136 M.Sato,Y.Kohmatsu,M.Yuma &Y.Tsubaki TableII The number ofrural and urban populationssampled(n), estimatesofgeneticdiversity (Flw), and geneticdifferentiation(Fsr<1>ST) in P.calamorumI. senegalensis,and I.asiaticaamongpopulations ineachurban and rural area.—indicate significantdifferencesfrom0atthep < 0.05levelbased on 1000 and 1023randompermutationsforF and rl> respectively ST s[, UUrrbbaann aarreeaa RRuurraall aarreeaa nn HHww FFs„t <<1I>>sStI nn HHww FfSsTt <<1t>>SS|T PP..ccaallaammoorruumm 66 00..227744 00..007799** 00..114455** 55 00..330000 00..003366** 00..007766** II..sseenneeggaalleennssiiss 55 00..332244 00,,003311** 00..006655** 55 00..330088 00..001188** 00..003366** II..aassiiaattiiccaa 77 00..330044 00..000066** 00..003311** 66 00..331133 00..000033** 00..001155 analogue),usingtheprogram Arlequin3.1 (EXCOFFIERetal., 2005).This <t>STissimilar toWeir andCockerham’s 6(WEIR&COCKERHAM, 1984),which iscalculated within ananalysisofvari- anceframeworkandsupposedtobeabetterestimatewhensamplesizesarefewer than30individuals (LOWEetal.,2004).Arlequinalsocalculated pairwised> valuesforallpairsofpopulationswithin ST eacharea.The significanceof<1>S1 andpairwise<I>STweretested usingnon-parametricpermutation ofthedata setwith 1023permutations. Forallpairsofpopulationsineach area,geneticdistanceswerecalculated (NEI,1972)usingAFLP- SURV 1.0(VEKEMANS,2002),andgeographicaldistanceswereobtainedassimplelinear distances, usingKashmir 3D (SUGIMOTO,2002).Wethen testedwhetherthere wasany correlation between thegeneticand geographicaldistancefor each species.Since theindependenceofpairwiseelements ofdistancematricesviolatesthebasicassumptionsforusingasimplecorrelation coefficient,weused the Manteltest (e.g.MANTEL, 1967;MANLY,1986;SOKAL & ROHLF, 1995). RESULTS Nostatisticaldifferenceswerefoundinthegeneticdiversity,asmeasuredby the meanexpected heterozygosity, betweenurbanand ruralareas forP.calamorum, I. senegalensis and I. asiatica(Unpaired t-test; p = 0.130,/? =0.120,/? = 0.201, respectively) (Tab. II).However,incomparisons betweenspecies, asignificant dif- ference was foundin the totalmean expected heterozygosity (combined urban and ruralpopulations) (ANOVA; F = 6.781,/? =0.003). Apost hoc test revealed that P. calamorumhad significantly less genetic diversity thanthatof I.senega- lensisandI. asiatica(LSD test;p=0.001 andp =0.009, respectively), butno dif- ferencewas foundbetween I. senegalensis andI. asiatica(LSD test;p=0.357). Significant population differentiationswithan F and <t> were foundforall SJ s[ three species except a3> valueofI. asiaticaamongruralpopulations thatwas ST not significantly differentfromzero(Tab. II). Therewas a consistent trend for all species withFsxand<I>STvaluesabouttwiceas high in urbancompared toru- ralpopulations (Tab. II), demonstrating thatpopulations intheurbanareawere more differentiatedthan intheruralones.Thepairwise4> valuesrevealedthat ST significant geneticdifferentiationwas morecommonamongurbancompared to ruralpopulations and thepattern was consistentfor allthreespecies (Tab. III). For P. calamorum, of the 15 pairwise comparisons withinthe urban area, all Populationgeneticdifferentiation insympatricZygoptera 137 pairwise 4> values(100 %)were significant,compared to seven(70 %) outof 10 ST pairwise comparisons withintherural area. For I. senegalensis, 10 (100 %) out of 10 pairwise <h valueswithinthe urbanareawere significant, andseven (70 ST %) outof 10withinthe ruralarea. For I. asiatica, ofthe21 pairwise <hSTvalues TableIII Pairwise estimates of<1> in(a)P. calamorum,(b)I. senegalensis,and (c) I. asiatica forurban and S1 rural populations.- indicate significantdifferences from0atthep<0.05 level ((aa))PP..ccaallaammoorruumm UU11 UU22 UU33 UU44 UU55 RRll RR77 RR88 RR1IO0 UU22 00..006600** UU33 00..118899** 00..220000** UU44 00..008888** 00..009966** 00..225500** UU55 00..009966** 00..006699** 00..119922** 00..009966** UU77 00..110011** 00..117755** 00..332200** 00..114477** 00..117711** RR77 00,,009944** RR88 00..008888** 00..007722** RR1IO0 00..006600** 00..008800** 00,,006655** RR1122 00..111177** 00..005533 00..007700 00..005500 ((bb))II..sseenneeggaalleennssiiss UU11 UU33 UU77 UU99 RR33 RR44 RR55 RR66 UU33 00..113377** UU77 00..004433** 00..111133** UU99 00..005522** 00..008822** 00..002266** UU1100 00..005511** 00..110044** 00..003322** 00..002211** RR44 00..007788** RR55 00..004411** 00..004411** RR66 00..004444** 00..004433** 00..001111 RR88 00..002288** 00..004455** 00..001177 00..001100 ((cc))II..aassiiaatliiccaa UU22 UU66 UU77 UU88 UU99 UU1lO0 RR22 RR33 RR44 RR88 RR99 UU66 00..001133 UU77 00..008833** 00..005533** UU88 00.,002299 00..110077** 00..004400 UU99 00..000000 00..000066 00..002277 00..004444 UU1lO0 00..000000 00..001133 00..008833** 00..008855 00..000000 UUllll 00..001100 00,,002222 00..005500** 00..006600** 00..000077 00..000000 RR33 00..000000 RR44 00..000000 00..000000 RR88 00..001177 00..000000 00..005522** RR99 00..000066 00..000000 00..002266 00..002255 RRlllI 00..002244 00,.001144 00..005599** 00,.008855** 00..003399 138 M.Sato, Y.Kohmatsu, M.Yuma &Y. Tsubaki withinthe urban area, six (29 %) were significant, andof 15 pairwise <I> val- ST ues withinthe ruralarea, three(20 %)were significant. There was also a trend that,in bothareas, P. calamorumshowedthehighest F and 4> valuesamong ST S| thethree species, andthe F and(1> valuesofI. senegalensis were higher than ST ST thoseofI. asiatica(Tab. II). TheManteltestshowedno significant relationship betweengenetic distanceand lineargeographical distance ineach areaforeach ofthethreespecies (p >0.1). DISCUSSION Inour study, wefoundrelatively high genetic diversity, butthepatternsof in- creasedgeneticstructuring (as estimatedby F /fU ),amongpopulations from ST ST urbanhabitatscompared to thepopulations fromruralhabitats,were consistent withthe predicted effects of habitatfragmentation on genetic structure forall threespecies (Tab. II). DITCHINGS & BEEBEE (1997), CAIZERGUES et al. (2003), and WIL- LIAMS(2003) all founddecreasedgenetic diversityandincreasedgeneticdiffer- entiationin either naturally or anthropogenically fragmented landscapes when compared to thoseincontrol sites. Interestingly, however, no decreasedgenetic diversity was detectedin our study andthoseby GIBBS (1998), KNUTSEN et al. (2000) and ARNOUD et al. (2003) although increased genetic differentia- tion was found. We propose the following two non-exclusive explanations for thehigh genetic diversity foundin thepresentstudy. Habitatfragmentationmay result inareducedgenetic diversitythrough randomgenetic drift(BAHL etal., 1996;ARNAUD et al., 2003). As randomgenetic driftis knownto depend on effectivepopulation size(BEEBEE &ROWE, 2004;LOWEetal., 2004),thatof eachpopulation intheurbanareaislargeenough to maintainhighgeneticdiver- sity. Analternativeexplanation is thatthereis sufficientgeneflowamong urban populations to maintainhighgenetic diversity. We didnotfindanycorrelationbetweenthegeneticdistanceandthelineargeo- graphical distanceeitherintheurbanorin theruralareas. Wheneversuch acor- relationis found, thefactor limiting dispersal amongpopulations is considered tobegeographical distance, andthis isthe basic premiseof thetheory of ‘isola- tion-by-distance’ (WRIGHT, 1946). Theabsence of such an associationin the presentstudymaysimply reflectinsufficienttimeforanequilibrium betweengene flowanddrifttobereachedandhencetheexpected geneticpatternsofisolation- -by-distance to build up(HUTCHISON & TEMPLETON, 1999;KNUTSEN et al.,2000). However,pairwise <h valuesbetween siteU3 andothersites were s| consistently higher forboth P. calamorum(range0.189-0.320) andI. senegalensis (range0.104-0.137) compared tothepairwise<1> valuesbetweensitesisolatedby ST equalor greaterdistances.Considering thatsiteU3islocatedinthecentreofthe Tokyo metropolitanarea, wesuspect thathumandisturbancessuch as buildings Populationgeneticdifferentiation insympatricZygoptera 139 androadscouldbelimiting geneflowandhavethus createdtheobserved genetic differentiationwhichcannot beexplained by geographical distance. Although all three species showed the samepattern,with urbanpopulations being moregenetically differentiatedthanruralpopulations, the F /<I> values SJ S| of P. calamorumwere twiceas high as thoseofI. senegalensis, whichwere also morethan twiceas high as those ofI. asiatica. Inaddition,thegeneticdiversity of P. calamorumwas significantly lowerthanthatofI. senegalensis and/. asiat- ica. To ourknowledge, thisis one ofonly afewstudies thathavedemonstrated howtheeffects ofhabitatfragmentation onpopulation genetics differedamong closely-related, sympatricspecies. Habitatspecificity couldbeoneexplanation for the observed differencesingenetic diversity anddifferentiationamongthe three species. P. calamorumhasspecifichabitatrequirements, generallyinhabiting pondswith submerged orfloating vegetation butwith relatively littleemergent aquatic veg- etation(UEDA, 1985). They alsorequire trees incloseproximityto thewater as maleroosting sites(UEDA, 1976).On theotherhand, bothI. senegalensis and I. asiaticauseawiderangeofwater typesincluding ponds,smallchannels, sew- agedrainsestuarieswithemergentaquatic vegetation andrice paddies (INOUE & TANI, 1999;WATANABE& MATSU'URA,2006). Dueto the higher habi- tatspecificityof P calamorumcompared tothatofI.senegalensis and I.asiatica, P. calamorumutilizedasmallerrangeof habitats,whichmayresultin restricted gene flow betweenavailable patches, especially in the highly fragmented urban landscape. Whencomparing thetwo Ischnuraspecies, thewiderrangeofhabitatrequire- ments forI. asiatica mightalso explain the observed differencesin F /<I> In ST sT additionto thehabitattypesusedby both species described above, /. asiaticais also seenaroundponds withgrassy typesofvegetation (e.g. Poasceae, Cyperace- ae) thatgrowinwet areaseventhough thepond itselfdoesnot haveany aquatic vegetation (pers. obs.). Theseeurytopiccharacteristics ofI. asiaticamayenable themtofindavailablepatches ofhabitatandcolonizesuccessfully inotherpatch- es, eveninfragmented areas. Although the three damselfly species investigated in this study are neither threatenednor endangered, the factthatthe population genetic differentiation differedamongclosely-related,sympatricspecies hasimportantmanagementim- plications. Habitatspecificity is awell-knownecologicalattributeofspecies that increases vulnerability toextinction(WALDRON et al.,2000). Thus, if habitat fragmentation hasamoresevere impacton habitatspecialists, as this study indi- cates,thismaybeone ofthe mechanismsexplaining why habitatspecialists are more vulnerableto extinction. It is worth noting that I. senegalensis has a larger body size than I. asiatica (adult body length (mm): I. senegalens■=23-25; I. asiatica = 20-25, SUGIMU- RA etal.,2001). Contraryto our intuitionthatlarger species arebetterdispers- 140 M.Sato,Y.Kohmatsu,M.Yuma &Y.Tsubaki ers, the smallerbody size of I. asiatica showed a lowergenetic differentiation. CORBET(1999)consideredthesetwo Ischnuraspeciesas migrantsbasedonthe reports thatadultsofI. asiaticawere caught more thanoccasionally over water far fromland(ASAHINA & TURUOKA, 1970),and larvaeof I.senegalensis inhabitedtemporary pools in seasonal rainfallareas (WEIR, 1968; WATSON, 1980as citedby CORBET, 1999;DUMONT, 1981as citedby CORBET 1999). If being small is aprerequisite for migrant Zygoptera to movea long distance by exposing themselves toaircurrents as CORBET (1999) inferred, I. asiatica, which hasa smaller body than I. senegalensis, should havean advantage indis- persing. An F valueof 1 indicatesperfect genetic differentiationand 0 indicatesno sx differentiation, withmoderate differentiationat values between 0.05 and 0.15 (WRIGHT, 1978). Themagnitude ofthe F / d> valuesfoundforboth urban ST ST and rural populations suggests arelatively high dispersal capability ofallthree damselfly species. Thisfinding is not inaccordance withthegeneralization that damselfliesare weak fliersthatcan disperse only betweenwater bodiesless than I kmapart(CONRAD etal.,2002). Since moderategenetic differentiationwas detectedfor P. calamorumin theurbanarea, theresolutionpowerofthegenetic markerusedin thisstudy was consideredtobe high enough to detectdifference in dispersal ability. Oneshould bearinmindthat dispersal ability measuredby direct observations (i.e. Mark-Release-Recapture) and indirect measurements (i.e.F /4> values) havesubstantially differentmeanings. Theformerisasnap- st st shot ofa one-timedispersal event. Not all the marked individualsare located andlong-distance andrare movements tendto beparticularly difficultto detect because of the spatio-temporal constraints of the method(e.g. RIECKEN & RATHS, 1996;NATHAN,2001;OSBORNE et al.. 2002). On theotherhand, dispersalmeasuredas F /<1> valuesis acumulativeeffect ofgeneflow.Individ- ST S| uals mightsuccessfully disperse to new habitatpatches, eitherby flying directly or in astepping-stone way, or evenby being passively dispersed by aircurrents. Thiscolonizationeventmighthappenwithinageneration insomecasesandover several generations inothercases,andmayresult ingeneflow.Therefore, whilst the resultsof this study suggest thaturbanpopulations are experiencing theef- fectsofhabitatfragmentation, relatively highdispersal events areexpected tobe occurring among the populations. Caution should betaken as this cumulative effectofgeneflowexpressed as F /ffi Imaynotbe reflecting thecurrent status ST S of dispersal where habitatdestructionandmodificationare stilllikely tobeoc- curring. Inconclusion, humandisturbancesinurbanareasare likely torestrictthedis- persal ofindividuals, andtheextentofhabitatfragmentationeffectsonthepopu- lationgeneticdifferentiationdifferedamongthethreedamselflyspecies probably becausetheyhavedifferenthabitatspecificity. However,thisstudyis onlythefirst step towardsdetermininghowweshouldconserve Odonataspecies andmaintain