AbstractQuestions WhataretheecologicalprocessesthatdeterminethespatialdistributionofspeciesandspeciesdiversityPartitioningbetadiversitycanprovidefundamentalinsights into the processes that determine the spatial variation of species assem-blages However studying beta diversity is conventionally based only on speciescompositiondataignoringthestructuralcomponentofcommunitiesStudy site TemperatemixedbroadleafndashconiferforestinJiaoheJilinProvincenorth-easternChinaMethods Wecharacterizedthevariationofcommunityassemblagesintermsofspe-ciescompositionsizestructureorconsideringbothcomponentsWethenemployedenvironmentalandspatialvariablesasexplanatoryfactorstopartitionthevariationinbothcompositionalandstructuralcomponentsofcommunityassemblageandas-sess the relative contributions of the niche and neutral processes to communityassemblyResults Thevaluesofoverallbetadiversity(BDstatistics)andtherelativecontribu-tion of individual sampling units to beta diversity (LCBD indices) depended onwhether thespeciescompositionsizestructureorbothtogetherhadbeentakenintoaccountThevalueofcompositionalndashstructuralbetadiversitywasthelargestfollowedbytraditionalcompositionalbetadiversitythesmallestwasthestructuralbetadiversityThesiteswithhighcontributionstobetadiversity(LCBDvalues)var-iedamongstructuralandcompositionalcomponentsTheexplanatorypoweroftheenvironmentalvariablesand thespatialvariablesalsovariedwidelywithdifferentcomponentsofacommunityThecombinationofenvironmentalandspatialvariablesexplainedthehighestproportionofvariation(438)inthecompositionalcompo-nentandexplainedthelowestproportionofvariation(254)inthestructuralcom-ponentofcommunityassemblageConclusion Both deterministic and stochastic processes are acting to determinecommunityassemblagesintermsofspeciescompositionandstructureinourtem-perateforestsiteOurstudyhighlightstheimportanceofconsideringthestructuralcomponentofforestcommunitiesinadditiontocompositionaldatawhenstudyingbetadiversity
258emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
1emsp |emspINTRODUCTION
Understandingthemechanismsthatdeterminethespatialdistri-butionofspeciesandspeciesdiversityisacentralthemeinecol-ogy (Chave 2004 Chesson 2000 Hutchinson 1961 Ricklefs1990 Vellend 2017) Deterministic niche-based and stochasticneutral processes have beenwidely discussed as potential driv-ers of community assembly (Chesson 2000 HilleRisLambersAdler Harpole Levine ampMayfield 2012Hubbell 2001 2006MayfieldampLevine2010)butthefactorsunderlyingtherelativecontributionof the twoprocessesarestillunresolved (Legendreetal2009Punchi-Manageetal2014vanderPlasetal2015)Niche and neutral theories emphasize different mechanisms assources of species diversity Niche theory predicts that deter-ministicprocessessuchashabitatfilteringandcompetitionshapespeciesassemblagesNeutraltheoryincontrastassumesthatallspecies are essentially functionally equivalent (HilleRisLambersetal 2012Hubbell 20012006Keddy1992) andemphasizesthe importance of stochastic processes in community assemblysuchasrandombirthdeathdispersaleventsspeciationandsto-chasticextinction(Caswell1976Hubbell2001)Itisnowgener-allyacceptedthatboththedeterministicandstochasticprocessesarepotentiallyimportantdeterminantsofthespatialdistributionobserved in community assemblagesAt present however theirrelative importance in shapingdifferent componentsof commu-nityorganization(iethestructuralcompositionalorbothcom-ponentstogether)isnotclear(DeCaacuteceresetal2012Legendreetal2009Punchi-Manageetal2014)Inthepresentstudywedefinedtheldquocompositionalrdquotermasthespeciescompositiondata(egspeciesabundancevalues)Weconstrainedthedefinitionofldquostructuralrdquotorefertothediameteratbreastheightoftheindivid-ualtreesmakingupthecommunity
Thevariationinspeciescompositionobservedamongasetofsamplingunitswithinaregionisoftendescribedasbetadiversity(Whittaker19601972)Theinterestofcommunityecologistsforbetadiversitystemsnotonlyfromthefactthatitlinkslocal(iealphadiversity) and regionaldiversity (ie gammadiversity) (DeCaacuteceresetal2012)butalsobecauseitcanprovidefundamentalinsights into theprocesses thatdetermine the spatialpatternofspecies assemblages (Anderson etal 2011 Chase 2010 Kraftetal 2011 Legendre amp De Caacuteceres 2013Myers etal 2013)Beta diversity can bemeasured inmany different ways (KoleffGastonampLennon2003LegendreBorcardampPeres-Neto2005Legendre amp Legendre 2012 Legendre etal 2009) Beta diver-sityestimatesaremostoftenbasedonspeciescompositionaldata(egspeciesabundancevaluesorspecies incidence)whichtakethe formof a site-by-species datamatrixwith sites in rows andspecies abundances in columns Although species compositiondataarefundamentallyimportanttheyalonemaybeinsufficientfor describing community organization and may neglect othervaluable information to study community assembly processessuchasthestructuralcomponent(egthesizestructureofcon-stituentindividuals)ofacommunity(DeCaacuteceresLegendreampHe
2013FaithAustinBelbinampMargules1985Fangetal2012)The phenomenon of competition asymmetry emphasizes thatlarge individuals usually compete disproportionately with theirsmaller-sized neighbors (Weiner 1990) Big trees control moreabove- and below-ground resources (eg light and mineral nu-trients)thansmalltrees(SchwinningampWeiner1998)ThereforelargerindividualstendtohavegreaterimpactonthefunctionanddynamicsofforestecosystemsthansmallonesMoreovernaturalmulti-species communities may exhibit similar compositions butdifferinotherfeaturessuchasthesizestructureoftheirindividu-als(DeCaacuteceresetal2013)Thedistributionofindividualsizesisalsoan importantcomponenttorepresentandunderstandcom-munity assembly therefore using species abundances only (iethe compositional component) to describe forest beta diversitymay be an oversimplification of the spatial variation of commu-nities In order to get comprehensive insight into the processesthat determine the spatial pattern of species assemblages it isnecessarytoensurefirstthatwehavetheabilitytodescribebetadiversity inacomprehensivewayWhetherornotthestructuralcomponent shouldor couldbeconsideredaltogetherwithotherbeta diversity components has never been investigated and re-mainstobeexplored
In thisstudywegeneralizedtheconventionalapproachto thestudy of beta diversity by considering structural data in additiontocompositionaldataWefirstmeasurethespatialvariationofas-semblagesonthebasisofspeciescompositionandsizestructureofconstituentsWethenusetheenvironmentalandspatialvariablesas explanatory factors to partition the variation in compositionalandstructuralcomponentsofcommunityassemblageWespecifi-callyaddressthefollowingquestions(a)Canwetakeboththespe-ciescompositionalandsizestructuralcomponentsofacommunityintoaccountwhendescribingbetadiversity Isthereacorrelationbetween thesebetadiversity components (b)How is the assess-mentofthesebetadiversitycomponentsaffectedbythesizeofthesampling units (c)When considering both the compositional andstructural components together towhatextentarebetadiversityassessmentsaffectedbytherelativeimportanceaccordedtostruc-tural vs compositional differences (d)What is the relative contri-bution of the environmental and spatial variables to communityassemblyintermsofspeciescompositionsizestructureorconsid-eringbothcomponents
Allindividualswithadiameteratbreastheight(dbh)of1cmormoreintheplotwereidentifiedmeasuredandspatiallymappedin2010Atotalof49684individualtreesbelongingto20familiesand47speciesintheplotwereusedinthepresentstudyTheplotwasdividedinto120(50mtimes50m)750(20mtimes20m)and3000(10mtimes10m)subplotshereaftercalledquadratsTopographicandsoilvari-ableswerealsoavailable foreachquadratFour topographicvari-ables(altitudequadratconvexityslopeandaspect)werecalculatedforeachquadratfollowingtherecommendationofHarmsConditHubbellandFoster(2001)andYamakuraetal(1995)Eightsoilen-vironmentalandnutrientvariablesweremeasuredpHtheamountof organic matter and the total amounts as well as the availablenutrients of nitrogen (N) phosphorus (P) and potassium (K) (ggYanZhangWangZhaoampGadow2015)All laboratoryanalyseswereconductedfollowingtheproceduresrecommendedbytheSoilScienceSocietyofChina(1999)
22emsp|emspStatistical analyses
221emsp|emspCumulative abundance profiles
Theconceptofcumulativeabundanceprofile (CAP)developedbyDeCaacuteceresetal(2013)isdefinedasafunctionthattakestheval-uesofastructuralvariable (egheightdbhetc)as inputandre-turnsthecumulativeabundanceofindividualswhosevaluesofthestructuralvariableareequal toor largerthanthe inputvalueTheCAP framework generalizes traditional species abundance valuesand allows researchers to describe the structural component of acommunityInthepresentstudythestructuralvariablewasdiam-eteratbreastheight(dbh)AccordingtothischoicethevalueofCAPforagivendbhvalueisthecumulativeabundanceoftreeindividualsasbigasorbiggerthantheinputvalueFunctionCAPinfactreplacestheabundancevalueofadbhclassbythesumofabundancesinthisandlargerdbhclasses
222emsp|emspCommunity tables
Following theconventionalmethods speciescomposition tables(ie quadrats in rows species in column and the table contain-ing individual counts)wereassembled in this studywecall thistablethetraditional species composition matrix(YCOMP)InordertogeneralizeatraditionalspeciesabundancevalueanddescribethesizestructurecomponentofthecommunitytheCAPsconsider-ingspeciesidentitywerecalculatedtoobtainthespecies composi-tion combined with structural data matrix(YCOMPndashSTR)TheYCOMPndashSTR isamatrixwithasmanyrowsasplotrecordsandwherecolumnsareorganized inblocksandthereareasmanyblocksasspecies
andeachblockhasasmanycolumnsassizeclassesDisregardingspeciesidentityofthedifferentindividualsCAPswerealsocalcu-latedtoobtainthecommunity structural matrix (YSTR)TheYSTR is amatrixwithasmanyrowsasplotrecordsandasmanycolumnsas size classes
FunctionsldquostratifyvegdatardquoandldquoCAPrdquointhevegclustRpack-age(DeCaacuteceresFontampOliva2010)availableonCRAN(httpsCRANR-projectorgpackage=vegclust) were applied to calculatetheCAPsFunctionsldquostratifyvegdatardquoandldquoCAPrdquorequirediscretiz-ingthestructuralvariableandthenumberofsizebinsaffectstheimportanceaccordedtostructuraldifferencesThustherearede-cisions tobemadewhencreatingYSTR and YCOMPndashSTR particularlyhowwedefine thebinsof thestructuralvariables (egdbhbins)Inthisstudywetestedfrom1-cmbinsizeto15-cmbinsizetodis-cretizedbhintoclassesThat is1cmbinsleadtodbhclasses1ndash22ndash33ndash4andsoonwhereas5cmbinsleadtodbhclasses1ndash56ndash1010ndash15andsoonThesmallerthesizeofdbhbinthemorecolumnswillbeproducedineachblockinthetableYCOMPndashSTRindicatingthatmoreweightisaccordedtodifferencesinstructureandviceversaIfthebinsizewasbigenoughsothatthenumberofcolumnsineachblock in the tableYCOMPndashSTRwasonewewouldhave thatYCOMPndashSTR=YCOMPGenerallythelargerthesizeofdbhbinsthemoresimi-lar will YCOMP and YCOMPndashSTR be
223emsp|emspPairwise dissimilarity in terms of community composition and structure
Wecalculateddissimilaritymatrices between all pairs of quadratsusingthepercentagedifferenceindex(akaBrayndashCurtisdissimilar-ity)oncommunitymatricesYCOMPYSTRandYCOMPndashSTRtoobtainthe compositional dissimilarity matrix (DCOMP) the structural dissimilar-ity matrix (DSTR)andthe compositionalndashstructural dissimilarity matrix (DCOMPndashSTR) respectively In order to explore the pairwise covari-ation between the three kinds of dissimilarity assessments (ieDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTR)wefirstcomputed principal coordinates of each dissimilarity matrix usingprincipalcoordinatesanalysis (PCoA) thencomparedtheresultingmatricesofprincipalcoordinateskeepingallaxesusingtheRVco-efficientWeexpectedthatDCOMPndashSTRwouldbecorrelatedtobothDCOMP and DSTRbutthestrengthofthecorrelationdoesdependonthechosensizeofdiameterbins(ieontheweightgiventostruc-turalvscompositionalinformation)
Function ldquovegdistrdquo with the dissimilarity index ldquobrayrdquo in theveganRpackage(Oksanenetal2018)wasusedtocalculatethedissimilarity matrices D Function ldquopcoardquo in the ape R package(ParadisClaudeampStrimmer2004)wasusedtocomputeprincipalcoordinates of eachdissimilaritymatrixD Thedissimilarities inD matricesweresquare-rootedbeforePCoAinordertomakethema-tricesEuclideanandpreventthegenerationofnegativeeigenvaluesandcomplexPCoAaxes (DeCaacuteceresetal2013)Function ldquocoef-fRVrdquo in the FactoMineR R package (Husson Josse LeampMazet2015)wasusedtocalculatetheRVcoefficientsbetweenthematri-cesofprincipalcoordinates
260emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
224emsp|emspBeta diversity components (BDCOMP BDSTR and BDCOMPndashSTR)
Conventionallybetadiversity (abbreviatedBD) is assessed fromasite-by-speciesdatamatrixotherbasiccharacteristics(egsizeofindividuals)ofthecommunityareignoredInordertogeneral-izetheconceptoftraditionalbetadiversitytoCAPdataweap-pliedtheindexproposedbyLegendreetal (2005)andLegendreandDeCaacuteceres(2013)tocomputebetadiversityasthevarianceofthecommunitydataLegendreandDeCaacuteceres(2013)showedhow to compute the total variance of the community composi-tion datamatrix from a dissimilaritymatrixD The total sum ofsquaresSS(Y)canbeobtainedfromadissimilaritymatrixDusingEquation1(LegendreampDeCaacuteceres2013LegendreampLegendre2012) Dividing SS(Y) by (n - 1) produces the classical unbiasedestimateofthetotalvarianceofYcomputedfromauser-selectedEuclidean dissimilarity matrixD (ie Equation 2)We used thatapproach to calculate the traditional compositional beta diversity (BDCOMP) the structural beta diversity (BDSTR) and the composi-tionalndashstructural beta diversity(BDCOMPndashSTR)respectivelyusingthefollowingequations
D =(Dhi)isanntimesnsymmetricdissimilaritymatrix(eitherDCOMP
DSTRorDCOMPndashSTR) i and h representthesamplingunitsn is thenumberofthesamplingunits If thecalculationsstartwithaper-centagedifferenceDmatrixwhichisnon-Euclideanonecomputesthesquare-rootsoftheDvaluesintheDmatrixtomakeitEuclideanbeforeusingthetransformedDvaluesinEquations1and2
225emsp|emspLocal contributions to beta diversity in terms of community composition and structure
Legendre andDe Caacuteceres (2013) suggested that total beta diver-sity can be partitioned into Local Contributions toBetaDiversity(LCBDwhicharecomparative indicatorsoftheecologicalunique-nessofthesites)TheLocalContributionstoBetaDiversity(LCBDi)represent the relative contributionsof the samplingunit i to betadiversityLCBDi indicateshowexceptional thecompositionofsitei iswhencomparedtothecentroidofallpointswhichwouldrep-resent a theoretical site with the average species composition ofall the sampling units In the present study the LCBD representsthe degree of uniqueness of each sampling unit in terms of com-position andor structure of community assemblages LCBDi indi-cescanbecalculatedfromthedissimilaritymatricesD(LegendreampDeCaacuteceres2013)OnefirsttransformsthedistancematrixDintomatrixA =(ahi)=(ndash05D2
hi) then centers thematrix as proposedbyGower(1966)
where Iisanidentitymatrixofsizen1isavectorofones(oflengthn)and1primeisitstranspose(LegendreampLegendre2012)HereeachdiagonalelementofmatrixGistheSSivalues(iethesquareddis-tancetothecentroidoftheithsamplingunit)Hencethevectoroflocalcontributionsofthesitestobetadiversity(LCBDi)is
The LCBD indices are scaled to sum to 1 We used functionldquoLCBDcomprdquointheadespatialRpackage(Drayetal2018)avail-ableonCRAN(httpsCRANR-projectorgpackage=adespatial)tocalculatetheLCBDindices
Wecheckedwhetherthere isacorrelationbetweentheLCBDcoefficientscalculatedfromspeciescompositionsizestructureorusingthetwocomponentstogetherHencewecalculatedSpearmanrank correlations pairwise between the three typesof LCBDvec-tors (ieLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRandLCBDSTRvsLCBDCOMPndashSTR)SincetheLCBDindicesindicatethede-greeofuniquenessof thesamplingunits in termsof their speciescompositionandorsizestructureweplottedtheLCBDvaluesonmapsof the30-haplotLargeLCBDvalues indicate thesites thathaveuniquespeciesassemblagesandsmallLCBDvaluesindicatethesites thathaveassemblagesthatareverysimilar to those inothersites Againwe expected LCBDCOMPndashSTR to be correlated to bothLCBDCOMP and LCBDSTR butwith the strength of the correlationdependingontheweightgiventostructuralvscompositionalinfor-mationWethusshowedthetwoextremecasesoftheLCBDmapaccordinga largestweighttothestructuralcomponentandcorre-spondinglythesmallestrelativeweighttothecompositionalcompo-nent(ie1-cmbinsize)andgivingthelargestrelativeweighttothecompositionalcomponent(ie15-cmbinsize)
226emsp|emspSets of explanatory variables environmental and spatial variables
Following Legendre etal (2009)we used altitude convexity andslope to construct third-degreepolynomial functions (ie yieldingninevariables)Themonomialswithexponentsallowthemodelingof nonlinear relationships between the topographic variables andthe response variablesWe calculated the aspect of a quadrat astheaverageangleofthefourtriangularplanesthatdeviatefromthenorthdirectionWe thusused the sin (aspect) and cos (aspect) inorder to include it in a linear regressionmodelWe thereforeob-tained11expandedtopographicvariablesWethencombinedthese11 expanded topographic variables with the eight soil variables(described insection21Studysitesanddatacollection) toobtainthe environmental variables data table (ie 19 variables) for eachquadratWe computed eigenfunctions of distance-basedMoranrsquoseigenvector maps (dbMEM also called Principal Coordinates ofNeighbour Matrices PCNM Borcard Legendre Avois-Jacquet amp
(1)SS(Y) =1
n
nminus1sum
h=1
nsum
i=h+1
D2hi
(2)BD = SS(Y)∕(nminus1)
(3)G =
(
Iminus11
n
)
A
(
Iminus11
n
)
(4)(LCBDi) = (SSi)∕SS(Y) = diag(G)∕SS(Y)
emspensp emsp | emsp261Journal of Vegetation Science
YAO et Al
Tuomisto2004LegendreampLegendre2012Legendreetal2009)acrossthe3000(10mtimes10m)750(20mtimes20m)and120(50mtimes50m)quadratsThedbMEMeigenfunctionswithpositiveeigenval-uesonlywereusedasspatialvariablesWeappliedforwardmodelselection(withpermutationtestsatthe5significanceleveloftheincrease in R2 ateachstep) toextract thesignificantenvironmentvariablesandeigenfunctionsofdbMEMusingthefunctionldquoforwardselrdquointhepackageadespatial(Drayetal2018)
227emsp|emspVariation partitioning of DCOMP DSTR and DCOMPndashSTR
Tocompare the influenceofniche-basedandspatialprocessesoncommunityassembly representedbycommunitycompositionsizestructure or the two components together distance-based re-dundancyanalysis (dbRDALegendreampAnderson1999Legendreamp Legendre 2012)was used to partition the variation of each ofthree community matrices (Borcard Legendre amp Drapeau 1992Legendre etal 2009 Peres-Neto Legendre Dray amp Borcard2006) Specificallyweused the two setsof explanatoryvariables(after forwardmodelselection) topartitionvariation in theprinci-palcoordinatetablesextractedfromDCOMPDSTRDCOMPndashSTRsepa-rately into fractions explained by the four different components(a)purehabitat (b)spatiallystructuredhabitat (c)purespaceand(d)undetermined (BorcardampLegendre1994Borcardetal1992DeCaacuteceresetal 2012Legendreetal 2009Myersetal 2013Punchi-Manage etal 2014) We hypothesized that the niche
processesareresponsiblefortheproportionofvariationexplainedbythepurehabitatandthespatially-structuredhabitatcomponents(a+b)(LaliberteacutePaquetteLegendreampBouchard2009Legendreetal2009)Whilewehypothesizedthattheproportionofvariationexplainedbythepurespatialcomponent(c)isrelatedtoindepend-ent biological processes (eg dispersal limitation competition fa-cilitationhistoricaleventsandJanzenndashConnelleffects)(LegendreampLegendre2012Legendreetal2009Punchi-Manageetal2014)Theundeterminedproportionofvariation(d)mayberelatedtosto-chastic processes or undefined non-spatially-structured biologicalor environmental variables (Dumbrell Nelson Helgason DythamampFitter2010)Thatallowedustoassesstherelativecontributionsoftheenvironmentalandspatialvariablestocommunityassemblyin termsof composition structureor taking the twocomponentstogetherAllanalyseswereperformedusingR(RCoreTeam2017)
3emsp |emspRESULTS
31emsp|emspPairwise dissimilarity in terms of community composition and structure
Wefound thatdissimilaritymatricescomputed fromspeciescom-position(DCOMP)sizestructure(DSTR)andconsideringbothcompo-nentstogether(DCOMPndashSTR)werecorrelatedHoweverthestrengthof the correlationdependedon the sizeofbinsused todiscretizethestructuralvariableandonthesizeof thequadrats (Figure1andashc) Overall the correlation between DCOMP vs DCOMPndashSTR was
F IGURE 1emspThecorrelationsbetweenthepairwisedissimilarityintermsofspeciescomposition(DCOMP)sizestructure(DSTR)andbothcomponentstogether(DCOMPndashSTR)ThecorrelationsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRand DSTR vs DCOMPndashSTRvarywithdbhbinsatthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofRVcoefficientsof1ndash15cmdbhbinsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTRrespectively(d)BoxplotsforRVcoefficientsofthethreepairwisedissimilaritycomparisons(aggregatedoverall1ndash15cmdbhbinsizes)foreachofthethreequadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
262emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
substantially stronger than that ofDCOMP vs DSTR andofDSTR vs DCOMPndashSTR
The correlation ofDCOMP vs DCOMPndashSTR increasedwith the in-crease of bin size Correspondingly the correlations of DSTR vs DCOMPndashSTR showed the opposite trend (Figure1andashc) As to the ef-fectofthesizeofthesamplingunitsthestrengthofcorrelationsin-creasedwiththequadratsize(Figure1d)exceptforthecorrelationbetweenDCOMP and DCOMPndashSTRwhichexhibitsnosignificantdiffer-encebetweenthe10mtimes10mand20mtimes20mquadrats(p = 023Figure1d)
32emsp|emspThe three components of beta diversity (BD) BDCOMP BDSTR and BDCOMPndashSTR
Thebetadiversity(BD)valueswerecloselyrelatedtowhetherthespecies composition size structure or both components togetherhad been taken into account Among these three components ofbetadiversityBDCOMPndashSTRwasgreatestcloselyfollowedbyBDCOMPandthesmallestwasBDSTR(Figure2)Sincethesizestructureofin-dividualswasnot consideredwhen calculatingBDCOMP this indexwasnotaffectedbythesizeofdbhbins(Figure2andashc)ThevaluesofBDCOMPndashSTRandBDSTRhoweverdecreasedslightlywithanincreaseofbinsizeWhenincreasingdbhbinsizethevaluesofBDCOMPndashSTR graduallyapproachedthevaluesofBDCOMP (Figure2andashc)BDalso
varied as a function of quadrat size (Figure2) values of BDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossbinsizes)systemati-callydecreasedwithincreasingquadratsize(Figure2d)
33emsp|emspLocal contributions to beta diversity in terms of community composition and structure
LocalContributionstoBetaDiversitycalculatedusingspeciescom-positionsizestructureorbothcomponentswerecorrelatedAgainthe strengthof correlationsdependedon the sizeofdbhbins andon the size of quadrats (Figure3andashc) In the case of LCBDCOMP vs LCBDCOMPndashSTR the strength of the correlation increased with anincreaseofbinsizeCorrespondinglythecorrelationofLCBDSTR vs LCBDCOMPndashSTRshowedtheoppositetrend(Figure3andashc)Thecorrela-tionsofLCBDCOMPvsLCBDSTRandLCBDSTRvsLCBDCOMPndashSTR were significantly different for differentquadrat sizesA striking findingwasthatthestrengthofcorrelationswasweakeratthescaleof20mtimes20m than that at the scalesof10mtimes10mor50mtimes50m(Figure3d)HowevercorrelationsbetweenLCBDCOMPvsLCBDCOMPndashSTRwerenotsubstantiallyaffectedbythesizeofquadrats(Figure3d)
F IGURE 2emspTheBetaDiversity(BD)intermsofspeciescomposition(BDCOMP)sizestructure(BDSTR)andbothcomponentstogether(BDCOMPndashSTR)ThevaluesofBDCOMPndashSTRandBDSTRvarywiththesizeofbinsofthestructuralvariable(dbhbinsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mThesizestructureofindividuals(iethedbh)isnotconsideredwhencalculatingtheBDCOMPthusthevaluesofBDCOMPwerenotaffectedbythebinsizeIngraphs(a)(b)and(c)thehorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofBDCOMPndashSTRandBDSTRacross1ndash15cmbinsizesrespectively(d)ValuesofBDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossdbhbinsizes)varywiththesamplingunitsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
emspensp emsp | emsp263Journal of Vegetation Science
YAO et Al
themeantobetadiversity)arevariedamongthreecomponentsofacommunity (Figure4)Specifically342(456)290(387)and331 (441)outof750quadratscontributedmorethanthemeanto beta diversity in term of species composition (ie LCBDCOMPFigure4a)sizestructure(ieLCBDSTRFigure4g)andbothcompo-nentstogether(ieLCBDCOMPndashSTRFigure4f)respectively
34emsp|emspVariation partitioning of matrices DCOMP DSTR and DCOMPndashSTR
Theexplanatorypoweroftheenvironmentalvariablesandthespa-tialvariablesvariedforthethreetypesofmatricesandwithquadratsizes (Table 1) The variation explained by the environmental vari-ables(a+b)andbythespatialvariables(b+c)increasedsystemati-callywith increasingscale (Table1)Averagingacrossquadratsizeshabitatandspacejointlyexplained438254and341ofthevariation in compositional component structural component andthe two components together of community assemblage respec-tivelyHoweverthecontributionofthepurehabitatcomponent (a)was negligible The combination of environmental and spatial vari-ablesexplained the lowestproportionofvariation in the structuralcomponentaloneandexplainedthehighestproportionofvariationinthecompositionalcomponentalone(Table1)Boththeenvironmen-talvariables(a+b)andthepurespatialvariables(c)explainedmore
Forest ecosystems can be characterized and evaluated in terms ofboththeirstructureandcomposition(Peet1992)Inpreviousstud-ies the compositional and structural components of a communityassemblagewereusuallyanalyzedseparately(egFangetal2012)Howeverthenatureofspeciesassemblagesindicatesthateitherspe-cies composition or size structure of constituent individuals alonemayoversimplifycommunityorganization (DeCaacuteceresetal2013)Changes in structure and compositionmay be onlyweakly related(egArsenaultampBradfield1995)thereforeassessmentofbothsi-multaneouslyisimportantwhenevaluatingcommunityassemblyInthepresentstudywegeneralizedtheconventionalapproachtocom-munityassemblagebyincorporatingstructuraldataofacommunityinadditiontocompositionaldatausingtheCAPframeworkToourknowledge this is the firstpaper that investigates inasinglestudythevariationinboththecompositionalandstructuralcomponentsofcommunityassemblagessimultaneouslyaswellasitsdeterminants
F IGURE 3emspThecorrelationsbetweentheLocalContributionstoBetaDiversity(LCBD)intermsofcommunitycomposition(LCBDCOMP)structure(LCBDSTR)andbothcomponentstogether(LCBDCOMPndashSTR)ThecorrelationsofLCBDCOMPvsLCBDSTRLCBDCOMP vsLCBDCOMPndashSTRandLCBDSTR vs LCBDCOMPndashSTRwiththesizeofbinsofthestructuralvariable(binsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofSpearmanrsquosrankcorrelationcoefficientr across 1ndash15 cm binsizeofLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRLCBDSTR vs LCBDCOMPndashSTRrespectively(d)BoxplotsfortheSpearmanrsquosrankcorrelationcoefficientrbetweenthepairwiseofthethreekindsofLCBDof1ndash15cmbinsizesatdifferentquadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
258emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
1emsp |emspINTRODUCTION
Understandingthemechanismsthatdeterminethespatialdistri-butionofspeciesandspeciesdiversityisacentralthemeinecol-ogy (Chave 2004 Chesson 2000 Hutchinson 1961 Ricklefs1990 Vellend 2017) Deterministic niche-based and stochasticneutral processes have beenwidely discussed as potential driv-ers of community assembly (Chesson 2000 HilleRisLambersAdler Harpole Levine ampMayfield 2012Hubbell 2001 2006MayfieldampLevine2010)butthefactorsunderlyingtherelativecontributionof the twoprocessesarestillunresolved (Legendreetal2009Punchi-Manageetal2014vanderPlasetal2015)Niche and neutral theories emphasize different mechanisms assources of species diversity Niche theory predicts that deter-ministicprocessessuchashabitatfilteringandcompetitionshapespeciesassemblagesNeutraltheoryincontrastassumesthatallspecies are essentially functionally equivalent (HilleRisLambersetal 2012Hubbell 20012006Keddy1992) andemphasizesthe importance of stochastic processes in community assemblysuchasrandombirthdeathdispersaleventsspeciationandsto-chasticextinction(Caswell1976Hubbell2001)Itisnowgener-allyacceptedthatboththedeterministicandstochasticprocessesarepotentiallyimportantdeterminantsofthespatialdistributionobserved in community assemblagesAt present however theirrelative importance in shapingdifferent componentsof commu-nityorganization(iethestructuralcompositionalorbothcom-ponentstogether)isnotclear(DeCaacuteceresetal2012Legendreetal2009Punchi-Manageetal2014)Inthepresentstudywedefinedtheldquocompositionalrdquotermasthespeciescompositiondata(egspeciesabundancevalues)Weconstrainedthedefinitionofldquostructuralrdquotorefertothediameteratbreastheightoftheindivid-ualtreesmakingupthecommunity
Thevariationinspeciescompositionobservedamongasetofsamplingunitswithinaregionisoftendescribedasbetadiversity(Whittaker19601972)Theinterestofcommunityecologistsforbetadiversitystemsnotonlyfromthefactthatitlinkslocal(iealphadiversity) and regionaldiversity (ie gammadiversity) (DeCaacuteceresetal2012)butalsobecauseitcanprovidefundamentalinsights into theprocesses thatdetermine the spatialpatternofspecies assemblages (Anderson etal 2011 Chase 2010 Kraftetal 2011 Legendre amp De Caacuteceres 2013Myers etal 2013)Beta diversity can bemeasured inmany different ways (KoleffGastonampLennon2003LegendreBorcardampPeres-Neto2005Legendre amp Legendre 2012 Legendre etal 2009) Beta diver-sityestimatesaremostoftenbasedonspeciescompositionaldata(egspeciesabundancevaluesorspecies incidence)whichtakethe formof a site-by-species datamatrixwith sites in rows andspecies abundances in columns Although species compositiondataarefundamentallyimportanttheyalonemaybeinsufficientfor describing community organization and may neglect othervaluable information to study community assembly processessuchasthestructuralcomponent(egthesizestructureofcon-stituentindividuals)ofacommunity(DeCaacuteceresLegendreampHe
2013FaithAustinBelbinampMargules1985Fangetal2012)The phenomenon of competition asymmetry emphasizes thatlarge individuals usually compete disproportionately with theirsmaller-sized neighbors (Weiner 1990) Big trees control moreabove- and below-ground resources (eg light and mineral nu-trients)thansmalltrees(SchwinningampWeiner1998)ThereforelargerindividualstendtohavegreaterimpactonthefunctionanddynamicsofforestecosystemsthansmallonesMoreovernaturalmulti-species communities may exhibit similar compositions butdifferinotherfeaturessuchasthesizestructureoftheirindividu-als(DeCaacuteceresetal2013)Thedistributionofindividualsizesisalsoan importantcomponenttorepresentandunderstandcom-munity assembly therefore using species abundances only (iethe compositional component) to describe forest beta diversitymay be an oversimplification of the spatial variation of commu-nities In order to get comprehensive insight into the processesthat determine the spatial pattern of species assemblages it isnecessarytoensurefirstthatwehavetheabilitytodescribebetadiversity inacomprehensivewayWhetherornotthestructuralcomponent shouldor couldbeconsideredaltogetherwithotherbeta diversity components has never been investigated and re-mainstobeexplored
In thisstudywegeneralizedtheconventionalapproachto thestudy of beta diversity by considering structural data in additiontocompositionaldataWefirstmeasurethespatialvariationofas-semblagesonthebasisofspeciescompositionandsizestructureofconstituentsWethenusetheenvironmentalandspatialvariablesas explanatory factors to partition the variation in compositionalandstructuralcomponentsofcommunityassemblageWespecifi-callyaddressthefollowingquestions(a)Canwetakeboththespe-ciescompositionalandsizestructuralcomponentsofacommunityintoaccountwhendescribingbetadiversity Isthereacorrelationbetween thesebetadiversity components (b)How is the assess-mentofthesebetadiversitycomponentsaffectedbythesizeofthesampling units (c)When considering both the compositional andstructural components together towhatextentarebetadiversityassessmentsaffectedbytherelativeimportanceaccordedtostruc-tural vs compositional differences (d)What is the relative contri-bution of the environmental and spatial variables to communityassemblyintermsofspeciescompositionsizestructureorconsid-eringbothcomponents
Allindividualswithadiameteratbreastheight(dbh)of1cmormoreintheplotwereidentifiedmeasuredandspatiallymappedin2010Atotalof49684individualtreesbelongingto20familiesand47speciesintheplotwereusedinthepresentstudyTheplotwasdividedinto120(50mtimes50m)750(20mtimes20m)and3000(10mtimes10m)subplotshereaftercalledquadratsTopographicandsoilvari-ableswerealsoavailable foreachquadratFour topographicvari-ables(altitudequadratconvexityslopeandaspect)werecalculatedforeachquadratfollowingtherecommendationofHarmsConditHubbellandFoster(2001)andYamakuraetal(1995)Eightsoilen-vironmentalandnutrientvariablesweremeasuredpHtheamountof organic matter and the total amounts as well as the availablenutrients of nitrogen (N) phosphorus (P) and potassium (K) (ggYanZhangWangZhaoampGadow2015)All laboratoryanalyseswereconductedfollowingtheproceduresrecommendedbytheSoilScienceSocietyofChina(1999)
22emsp|emspStatistical analyses
221emsp|emspCumulative abundance profiles
Theconceptofcumulativeabundanceprofile (CAP)developedbyDeCaacuteceresetal(2013)isdefinedasafunctionthattakestheval-uesofastructuralvariable (egheightdbhetc)as inputandre-turnsthecumulativeabundanceofindividualswhosevaluesofthestructuralvariableareequal toor largerthanthe inputvalueTheCAP framework generalizes traditional species abundance valuesand allows researchers to describe the structural component of acommunityInthepresentstudythestructuralvariablewasdiam-eteratbreastheight(dbh)AccordingtothischoicethevalueofCAPforagivendbhvalueisthecumulativeabundanceoftreeindividualsasbigasorbiggerthantheinputvalueFunctionCAPinfactreplacestheabundancevalueofadbhclassbythesumofabundancesinthisandlargerdbhclasses
222emsp|emspCommunity tables
Following theconventionalmethods speciescomposition tables(ie quadrats in rows species in column and the table contain-ing individual counts)wereassembled in this studywecall thistablethetraditional species composition matrix(YCOMP)InordertogeneralizeatraditionalspeciesabundancevalueanddescribethesizestructurecomponentofthecommunitytheCAPsconsider-ingspeciesidentitywerecalculatedtoobtainthespecies composi-tion combined with structural data matrix(YCOMPndashSTR)TheYCOMPndashSTR isamatrixwithasmanyrowsasplotrecordsandwherecolumnsareorganized inblocksandthereareasmanyblocksasspecies
andeachblockhasasmanycolumnsassizeclassesDisregardingspeciesidentityofthedifferentindividualsCAPswerealsocalcu-latedtoobtainthecommunity structural matrix (YSTR)TheYSTR is amatrixwithasmanyrowsasplotrecordsandasmanycolumnsas size classes
FunctionsldquostratifyvegdatardquoandldquoCAPrdquointhevegclustRpack-age(DeCaacuteceresFontampOliva2010)availableonCRAN(httpsCRANR-projectorgpackage=vegclust) were applied to calculatetheCAPsFunctionsldquostratifyvegdatardquoandldquoCAPrdquorequirediscretiz-ingthestructuralvariableandthenumberofsizebinsaffectstheimportanceaccordedtostructuraldifferencesThustherearede-cisions tobemadewhencreatingYSTR and YCOMPndashSTR particularlyhowwedefine thebinsof thestructuralvariables (egdbhbins)Inthisstudywetestedfrom1-cmbinsizeto15-cmbinsizetodis-cretizedbhintoclassesThat is1cmbinsleadtodbhclasses1ndash22ndash33ndash4andsoonwhereas5cmbinsleadtodbhclasses1ndash56ndash1010ndash15andsoonThesmallerthesizeofdbhbinthemorecolumnswillbeproducedineachblockinthetableYCOMPndashSTRindicatingthatmoreweightisaccordedtodifferencesinstructureandviceversaIfthebinsizewasbigenoughsothatthenumberofcolumnsineachblock in the tableYCOMPndashSTRwasonewewouldhave thatYCOMPndashSTR=YCOMPGenerallythelargerthesizeofdbhbinsthemoresimi-lar will YCOMP and YCOMPndashSTR be
223emsp|emspPairwise dissimilarity in terms of community composition and structure
Wecalculateddissimilaritymatrices between all pairs of quadratsusingthepercentagedifferenceindex(akaBrayndashCurtisdissimilar-ity)oncommunitymatricesYCOMPYSTRandYCOMPndashSTRtoobtainthe compositional dissimilarity matrix (DCOMP) the structural dissimilar-ity matrix (DSTR)andthe compositionalndashstructural dissimilarity matrix (DCOMPndashSTR) respectively In order to explore the pairwise covari-ation between the three kinds of dissimilarity assessments (ieDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTR)wefirstcomputed principal coordinates of each dissimilarity matrix usingprincipalcoordinatesanalysis (PCoA) thencomparedtheresultingmatricesofprincipalcoordinateskeepingallaxesusingtheRVco-efficientWeexpectedthatDCOMPndashSTRwouldbecorrelatedtobothDCOMP and DSTRbutthestrengthofthecorrelationdoesdependonthechosensizeofdiameterbins(ieontheweightgiventostruc-turalvscompositionalinformation)
Function ldquovegdistrdquo with the dissimilarity index ldquobrayrdquo in theveganRpackage(Oksanenetal2018)wasusedtocalculatethedissimilarity matrices D Function ldquopcoardquo in the ape R package(ParadisClaudeampStrimmer2004)wasusedtocomputeprincipalcoordinates of eachdissimilaritymatrixD Thedissimilarities inD matricesweresquare-rootedbeforePCoAinordertomakethema-tricesEuclideanandpreventthegenerationofnegativeeigenvaluesandcomplexPCoAaxes (DeCaacuteceresetal2013)Function ldquocoef-fRVrdquo in the FactoMineR R package (Husson Josse LeampMazet2015)wasusedtocalculatetheRVcoefficientsbetweenthematri-cesofprincipalcoordinates
260emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
224emsp|emspBeta diversity components (BDCOMP BDSTR and BDCOMPndashSTR)
Conventionallybetadiversity (abbreviatedBD) is assessed fromasite-by-speciesdatamatrixotherbasiccharacteristics(egsizeofindividuals)ofthecommunityareignoredInordertogeneral-izetheconceptoftraditionalbetadiversitytoCAPdataweap-pliedtheindexproposedbyLegendreetal (2005)andLegendreandDeCaacuteceres(2013)tocomputebetadiversityasthevarianceofthecommunitydataLegendreandDeCaacuteceres(2013)showedhow to compute the total variance of the community composi-tion datamatrix from a dissimilaritymatrixD The total sum ofsquaresSS(Y)canbeobtainedfromadissimilaritymatrixDusingEquation1(LegendreampDeCaacuteceres2013LegendreampLegendre2012) Dividing SS(Y) by (n - 1) produces the classical unbiasedestimateofthetotalvarianceofYcomputedfromauser-selectedEuclidean dissimilarity matrixD (ie Equation 2)We used thatapproach to calculate the traditional compositional beta diversity (BDCOMP) the structural beta diversity (BDSTR) and the composi-tionalndashstructural beta diversity(BDCOMPndashSTR)respectivelyusingthefollowingequations
D =(Dhi)isanntimesnsymmetricdissimilaritymatrix(eitherDCOMP
DSTRorDCOMPndashSTR) i and h representthesamplingunitsn is thenumberofthesamplingunits If thecalculationsstartwithaper-centagedifferenceDmatrixwhichisnon-Euclideanonecomputesthesquare-rootsoftheDvaluesintheDmatrixtomakeitEuclideanbeforeusingthetransformedDvaluesinEquations1and2
225emsp|emspLocal contributions to beta diversity in terms of community composition and structure
Legendre andDe Caacuteceres (2013) suggested that total beta diver-sity can be partitioned into Local Contributions toBetaDiversity(LCBDwhicharecomparative indicatorsoftheecologicalunique-nessofthesites)TheLocalContributionstoBetaDiversity(LCBDi)represent the relative contributionsof the samplingunit i to betadiversityLCBDi indicateshowexceptional thecompositionofsitei iswhencomparedtothecentroidofallpointswhichwouldrep-resent a theoretical site with the average species composition ofall the sampling units In the present study the LCBD representsthe degree of uniqueness of each sampling unit in terms of com-position andor structure of community assemblages LCBDi indi-cescanbecalculatedfromthedissimilaritymatricesD(LegendreampDeCaacuteceres2013)OnefirsttransformsthedistancematrixDintomatrixA =(ahi)=(ndash05D2
hi) then centers thematrix as proposedbyGower(1966)
where Iisanidentitymatrixofsizen1isavectorofones(oflengthn)and1primeisitstranspose(LegendreampLegendre2012)HereeachdiagonalelementofmatrixGistheSSivalues(iethesquareddis-tancetothecentroidoftheithsamplingunit)Hencethevectoroflocalcontributionsofthesitestobetadiversity(LCBDi)is
The LCBD indices are scaled to sum to 1 We used functionldquoLCBDcomprdquointheadespatialRpackage(Drayetal2018)avail-ableonCRAN(httpsCRANR-projectorgpackage=adespatial)tocalculatetheLCBDindices
Wecheckedwhetherthere isacorrelationbetweentheLCBDcoefficientscalculatedfromspeciescompositionsizestructureorusingthetwocomponentstogetherHencewecalculatedSpearmanrank correlations pairwise between the three typesof LCBDvec-tors (ieLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRandLCBDSTRvsLCBDCOMPndashSTR)SincetheLCBDindicesindicatethede-greeofuniquenessof thesamplingunits in termsof their speciescompositionandorsizestructureweplottedtheLCBDvaluesonmapsof the30-haplotLargeLCBDvalues indicate thesites thathaveuniquespeciesassemblagesandsmallLCBDvaluesindicatethesites thathaveassemblagesthatareverysimilar to those inothersites Againwe expected LCBDCOMPndashSTR to be correlated to bothLCBDCOMP and LCBDSTR butwith the strength of the correlationdependingontheweightgiventostructuralvscompositionalinfor-mationWethusshowedthetwoextremecasesoftheLCBDmapaccordinga largestweighttothestructuralcomponentandcorre-spondinglythesmallestrelativeweighttothecompositionalcompo-nent(ie1-cmbinsize)andgivingthelargestrelativeweighttothecompositionalcomponent(ie15-cmbinsize)
226emsp|emspSets of explanatory variables environmental and spatial variables
Following Legendre etal (2009)we used altitude convexity andslope to construct third-degreepolynomial functions (ie yieldingninevariables)Themonomialswithexponentsallowthemodelingof nonlinear relationships between the topographic variables andthe response variablesWe calculated the aspect of a quadrat astheaverageangleofthefourtriangularplanesthatdeviatefromthenorthdirectionWe thusused the sin (aspect) and cos (aspect) inorder to include it in a linear regressionmodelWe thereforeob-tained11expandedtopographicvariablesWethencombinedthese11 expanded topographic variables with the eight soil variables(described insection21Studysitesanddatacollection) toobtainthe environmental variables data table (ie 19 variables) for eachquadratWe computed eigenfunctions of distance-basedMoranrsquoseigenvector maps (dbMEM also called Principal Coordinates ofNeighbour Matrices PCNM Borcard Legendre Avois-Jacquet amp
(1)SS(Y) =1
n
nminus1sum
h=1
nsum
i=h+1
D2hi
(2)BD = SS(Y)∕(nminus1)
(3)G =
(
Iminus11
n
)
A
(
Iminus11
n
)
(4)(LCBDi) = (SSi)∕SS(Y) = diag(G)∕SS(Y)
emspensp emsp | emsp261Journal of Vegetation Science
YAO et Al
Tuomisto2004LegendreampLegendre2012Legendreetal2009)acrossthe3000(10mtimes10m)750(20mtimes20m)and120(50mtimes50m)quadratsThedbMEMeigenfunctionswithpositiveeigenval-uesonlywereusedasspatialvariablesWeappliedforwardmodelselection(withpermutationtestsatthe5significanceleveloftheincrease in R2 ateachstep) toextract thesignificantenvironmentvariablesandeigenfunctionsofdbMEMusingthefunctionldquoforwardselrdquointhepackageadespatial(Drayetal2018)
227emsp|emspVariation partitioning of DCOMP DSTR and DCOMPndashSTR
Tocompare the influenceofniche-basedandspatialprocessesoncommunityassembly representedbycommunitycompositionsizestructure or the two components together distance-based re-dundancyanalysis (dbRDALegendreampAnderson1999Legendreamp Legendre 2012)was used to partition the variation of each ofthree community matrices (Borcard Legendre amp Drapeau 1992Legendre etal 2009 Peres-Neto Legendre Dray amp Borcard2006) Specificallyweused the two setsof explanatoryvariables(after forwardmodelselection) topartitionvariation in theprinci-palcoordinatetablesextractedfromDCOMPDSTRDCOMPndashSTRsepa-rately into fractions explained by the four different components(a)purehabitat (b)spatiallystructuredhabitat (c)purespaceand(d)undetermined (BorcardampLegendre1994Borcardetal1992DeCaacuteceresetal 2012Legendreetal 2009Myersetal 2013Punchi-Manage etal 2014) We hypothesized that the niche
processesareresponsiblefortheproportionofvariationexplainedbythepurehabitatandthespatially-structuredhabitatcomponents(a+b)(LaliberteacutePaquetteLegendreampBouchard2009Legendreetal2009)Whilewehypothesizedthattheproportionofvariationexplainedbythepurespatialcomponent(c)isrelatedtoindepend-ent biological processes (eg dispersal limitation competition fa-cilitationhistoricaleventsandJanzenndashConnelleffects)(LegendreampLegendre2012Legendreetal2009Punchi-Manageetal2014)Theundeterminedproportionofvariation(d)mayberelatedtosto-chastic processes or undefined non-spatially-structured biologicalor environmental variables (Dumbrell Nelson Helgason DythamampFitter2010)Thatallowedustoassesstherelativecontributionsoftheenvironmentalandspatialvariablestocommunityassemblyin termsof composition structureor taking the twocomponentstogetherAllanalyseswereperformedusingR(RCoreTeam2017)
3emsp |emspRESULTS
31emsp|emspPairwise dissimilarity in terms of community composition and structure
Wefound thatdissimilaritymatricescomputed fromspeciescom-position(DCOMP)sizestructure(DSTR)andconsideringbothcompo-nentstogether(DCOMPndashSTR)werecorrelatedHoweverthestrengthof the correlationdependedon the sizeofbinsused todiscretizethestructuralvariableandonthesizeof thequadrats (Figure1andashc) Overall the correlation between DCOMP vs DCOMPndashSTR was
F IGURE 1emspThecorrelationsbetweenthepairwisedissimilarityintermsofspeciescomposition(DCOMP)sizestructure(DSTR)andbothcomponentstogether(DCOMPndashSTR)ThecorrelationsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRand DSTR vs DCOMPndashSTRvarywithdbhbinsatthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofRVcoefficientsof1ndash15cmdbhbinsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTRrespectively(d)BoxplotsforRVcoefficientsofthethreepairwisedissimilaritycomparisons(aggregatedoverall1ndash15cmdbhbinsizes)foreachofthethreequadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
262emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
substantially stronger than that ofDCOMP vs DSTR andofDSTR vs DCOMPndashSTR
The correlation ofDCOMP vs DCOMPndashSTR increasedwith the in-crease of bin size Correspondingly the correlations of DSTR vs DCOMPndashSTR showed the opposite trend (Figure1andashc) As to the ef-fectofthesizeofthesamplingunitsthestrengthofcorrelationsin-creasedwiththequadratsize(Figure1d)exceptforthecorrelationbetweenDCOMP and DCOMPndashSTRwhichexhibitsnosignificantdiffer-encebetweenthe10mtimes10mand20mtimes20mquadrats(p = 023Figure1d)
32emsp|emspThe three components of beta diversity (BD) BDCOMP BDSTR and BDCOMPndashSTR
Thebetadiversity(BD)valueswerecloselyrelatedtowhetherthespecies composition size structure or both components togetherhad been taken into account Among these three components ofbetadiversityBDCOMPndashSTRwasgreatestcloselyfollowedbyBDCOMPandthesmallestwasBDSTR(Figure2)Sincethesizestructureofin-dividualswasnot consideredwhen calculatingBDCOMP this indexwasnotaffectedbythesizeofdbhbins(Figure2andashc)ThevaluesofBDCOMPndashSTRandBDSTRhoweverdecreasedslightlywithanincreaseofbinsizeWhenincreasingdbhbinsizethevaluesofBDCOMPndashSTR graduallyapproachedthevaluesofBDCOMP (Figure2andashc)BDalso
varied as a function of quadrat size (Figure2) values of BDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossbinsizes)systemati-callydecreasedwithincreasingquadratsize(Figure2d)
33emsp|emspLocal contributions to beta diversity in terms of community composition and structure
LocalContributionstoBetaDiversitycalculatedusingspeciescom-positionsizestructureorbothcomponentswerecorrelatedAgainthe strengthof correlationsdependedon the sizeofdbhbins andon the size of quadrats (Figure3andashc) In the case of LCBDCOMP vs LCBDCOMPndashSTR the strength of the correlation increased with anincreaseofbinsizeCorrespondinglythecorrelationofLCBDSTR vs LCBDCOMPndashSTRshowedtheoppositetrend(Figure3andashc)Thecorrela-tionsofLCBDCOMPvsLCBDSTRandLCBDSTRvsLCBDCOMPndashSTR were significantly different for differentquadrat sizesA striking findingwasthatthestrengthofcorrelationswasweakeratthescaleof20mtimes20m than that at the scalesof10mtimes10mor50mtimes50m(Figure3d)HowevercorrelationsbetweenLCBDCOMPvsLCBDCOMPndashSTRwerenotsubstantiallyaffectedbythesizeofquadrats(Figure3d)
F IGURE 2emspTheBetaDiversity(BD)intermsofspeciescomposition(BDCOMP)sizestructure(BDSTR)andbothcomponentstogether(BDCOMPndashSTR)ThevaluesofBDCOMPndashSTRandBDSTRvarywiththesizeofbinsofthestructuralvariable(dbhbinsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mThesizestructureofindividuals(iethedbh)isnotconsideredwhencalculatingtheBDCOMPthusthevaluesofBDCOMPwerenotaffectedbythebinsizeIngraphs(a)(b)and(c)thehorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofBDCOMPndashSTRandBDSTRacross1ndash15cmbinsizesrespectively(d)ValuesofBDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossdbhbinsizes)varywiththesamplingunitsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
emspensp emsp | emsp263Journal of Vegetation Science
YAO et Al
themeantobetadiversity)arevariedamongthreecomponentsofacommunity (Figure4)Specifically342(456)290(387)and331 (441)outof750quadratscontributedmorethanthemeanto beta diversity in term of species composition (ie LCBDCOMPFigure4a)sizestructure(ieLCBDSTRFigure4g)andbothcompo-nentstogether(ieLCBDCOMPndashSTRFigure4f)respectively
34emsp|emspVariation partitioning of matrices DCOMP DSTR and DCOMPndashSTR
Theexplanatorypoweroftheenvironmentalvariablesandthespa-tialvariablesvariedforthethreetypesofmatricesandwithquadratsizes (Table 1) The variation explained by the environmental vari-ables(a+b)andbythespatialvariables(b+c)increasedsystemati-callywith increasingscale (Table1)Averagingacrossquadratsizeshabitatandspacejointlyexplained438254and341ofthevariation in compositional component structural component andthe two components together of community assemblage respec-tivelyHoweverthecontributionofthepurehabitatcomponent (a)was negligible The combination of environmental and spatial vari-ablesexplained the lowestproportionofvariation in the structuralcomponentaloneandexplainedthehighestproportionofvariationinthecompositionalcomponentalone(Table1)Boththeenvironmen-talvariables(a+b)andthepurespatialvariables(c)explainedmore
Forest ecosystems can be characterized and evaluated in terms ofboththeirstructureandcomposition(Peet1992)Inpreviousstud-ies the compositional and structural components of a communityassemblagewereusuallyanalyzedseparately(egFangetal2012)Howeverthenatureofspeciesassemblagesindicatesthateitherspe-cies composition or size structure of constituent individuals alonemayoversimplifycommunityorganization (DeCaacuteceresetal2013)Changes in structure and compositionmay be onlyweakly related(egArsenaultampBradfield1995)thereforeassessmentofbothsi-multaneouslyisimportantwhenevaluatingcommunityassemblyInthepresentstudywegeneralizedtheconventionalapproachtocom-munityassemblagebyincorporatingstructuraldataofacommunityinadditiontocompositionaldatausingtheCAPframeworkToourknowledge this is the firstpaper that investigates inasinglestudythevariationinboththecompositionalandstructuralcomponentsofcommunityassemblagessimultaneouslyaswellasitsdeterminants
F IGURE 3emspThecorrelationsbetweentheLocalContributionstoBetaDiversity(LCBD)intermsofcommunitycomposition(LCBDCOMP)structure(LCBDSTR)andbothcomponentstogether(LCBDCOMPndashSTR)ThecorrelationsofLCBDCOMPvsLCBDSTRLCBDCOMP vsLCBDCOMPndashSTRandLCBDSTR vs LCBDCOMPndashSTRwiththesizeofbinsofthestructuralvariable(binsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofSpearmanrsquosrankcorrelationcoefficientr across 1ndash15 cm binsizeofLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRLCBDSTR vs LCBDCOMPndashSTRrespectively(d)BoxplotsfortheSpearmanrsquosrankcorrelationcoefficientrbetweenthepairwiseofthethreekindsofLCBDof1ndash15cmbinsizesatdifferentquadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
emspensp emsp | emsp259Journal of Vegetation Science
Allindividualswithadiameteratbreastheight(dbh)of1cmormoreintheplotwereidentifiedmeasuredandspatiallymappedin2010Atotalof49684individualtreesbelongingto20familiesand47speciesintheplotwereusedinthepresentstudyTheplotwasdividedinto120(50mtimes50m)750(20mtimes20m)and3000(10mtimes10m)subplotshereaftercalledquadratsTopographicandsoilvari-ableswerealsoavailable foreachquadratFour topographicvari-ables(altitudequadratconvexityslopeandaspect)werecalculatedforeachquadratfollowingtherecommendationofHarmsConditHubbellandFoster(2001)andYamakuraetal(1995)Eightsoilen-vironmentalandnutrientvariablesweremeasuredpHtheamountof organic matter and the total amounts as well as the availablenutrients of nitrogen (N) phosphorus (P) and potassium (K) (ggYanZhangWangZhaoampGadow2015)All laboratoryanalyseswereconductedfollowingtheproceduresrecommendedbytheSoilScienceSocietyofChina(1999)
22emsp|emspStatistical analyses
221emsp|emspCumulative abundance profiles
Theconceptofcumulativeabundanceprofile (CAP)developedbyDeCaacuteceresetal(2013)isdefinedasafunctionthattakestheval-uesofastructuralvariable (egheightdbhetc)as inputandre-turnsthecumulativeabundanceofindividualswhosevaluesofthestructuralvariableareequal toor largerthanthe inputvalueTheCAP framework generalizes traditional species abundance valuesand allows researchers to describe the structural component of acommunityInthepresentstudythestructuralvariablewasdiam-eteratbreastheight(dbh)AccordingtothischoicethevalueofCAPforagivendbhvalueisthecumulativeabundanceoftreeindividualsasbigasorbiggerthantheinputvalueFunctionCAPinfactreplacestheabundancevalueofadbhclassbythesumofabundancesinthisandlargerdbhclasses
222emsp|emspCommunity tables
Following theconventionalmethods speciescomposition tables(ie quadrats in rows species in column and the table contain-ing individual counts)wereassembled in this studywecall thistablethetraditional species composition matrix(YCOMP)InordertogeneralizeatraditionalspeciesabundancevalueanddescribethesizestructurecomponentofthecommunitytheCAPsconsider-ingspeciesidentitywerecalculatedtoobtainthespecies composi-tion combined with structural data matrix(YCOMPndashSTR)TheYCOMPndashSTR isamatrixwithasmanyrowsasplotrecordsandwherecolumnsareorganized inblocksandthereareasmanyblocksasspecies
andeachblockhasasmanycolumnsassizeclassesDisregardingspeciesidentityofthedifferentindividualsCAPswerealsocalcu-latedtoobtainthecommunity structural matrix (YSTR)TheYSTR is amatrixwithasmanyrowsasplotrecordsandasmanycolumnsas size classes
FunctionsldquostratifyvegdatardquoandldquoCAPrdquointhevegclustRpack-age(DeCaacuteceresFontampOliva2010)availableonCRAN(httpsCRANR-projectorgpackage=vegclust) were applied to calculatetheCAPsFunctionsldquostratifyvegdatardquoandldquoCAPrdquorequirediscretiz-ingthestructuralvariableandthenumberofsizebinsaffectstheimportanceaccordedtostructuraldifferencesThustherearede-cisions tobemadewhencreatingYSTR and YCOMPndashSTR particularlyhowwedefine thebinsof thestructuralvariables (egdbhbins)Inthisstudywetestedfrom1-cmbinsizeto15-cmbinsizetodis-cretizedbhintoclassesThat is1cmbinsleadtodbhclasses1ndash22ndash33ndash4andsoonwhereas5cmbinsleadtodbhclasses1ndash56ndash1010ndash15andsoonThesmallerthesizeofdbhbinthemorecolumnswillbeproducedineachblockinthetableYCOMPndashSTRindicatingthatmoreweightisaccordedtodifferencesinstructureandviceversaIfthebinsizewasbigenoughsothatthenumberofcolumnsineachblock in the tableYCOMPndashSTRwasonewewouldhave thatYCOMPndashSTR=YCOMPGenerallythelargerthesizeofdbhbinsthemoresimi-lar will YCOMP and YCOMPndashSTR be
223emsp|emspPairwise dissimilarity in terms of community composition and structure
Wecalculateddissimilaritymatrices between all pairs of quadratsusingthepercentagedifferenceindex(akaBrayndashCurtisdissimilar-ity)oncommunitymatricesYCOMPYSTRandYCOMPndashSTRtoobtainthe compositional dissimilarity matrix (DCOMP) the structural dissimilar-ity matrix (DSTR)andthe compositionalndashstructural dissimilarity matrix (DCOMPndashSTR) respectively In order to explore the pairwise covari-ation between the three kinds of dissimilarity assessments (ieDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTR)wefirstcomputed principal coordinates of each dissimilarity matrix usingprincipalcoordinatesanalysis (PCoA) thencomparedtheresultingmatricesofprincipalcoordinateskeepingallaxesusingtheRVco-efficientWeexpectedthatDCOMPndashSTRwouldbecorrelatedtobothDCOMP and DSTRbutthestrengthofthecorrelationdoesdependonthechosensizeofdiameterbins(ieontheweightgiventostruc-turalvscompositionalinformation)
Function ldquovegdistrdquo with the dissimilarity index ldquobrayrdquo in theveganRpackage(Oksanenetal2018)wasusedtocalculatethedissimilarity matrices D Function ldquopcoardquo in the ape R package(ParadisClaudeampStrimmer2004)wasusedtocomputeprincipalcoordinates of eachdissimilaritymatrixD Thedissimilarities inD matricesweresquare-rootedbeforePCoAinordertomakethema-tricesEuclideanandpreventthegenerationofnegativeeigenvaluesandcomplexPCoAaxes (DeCaacuteceresetal2013)Function ldquocoef-fRVrdquo in the FactoMineR R package (Husson Josse LeampMazet2015)wasusedtocalculatetheRVcoefficientsbetweenthematri-cesofprincipalcoordinates
260emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
224emsp|emspBeta diversity components (BDCOMP BDSTR and BDCOMPndashSTR)
Conventionallybetadiversity (abbreviatedBD) is assessed fromasite-by-speciesdatamatrixotherbasiccharacteristics(egsizeofindividuals)ofthecommunityareignoredInordertogeneral-izetheconceptoftraditionalbetadiversitytoCAPdataweap-pliedtheindexproposedbyLegendreetal (2005)andLegendreandDeCaacuteceres(2013)tocomputebetadiversityasthevarianceofthecommunitydataLegendreandDeCaacuteceres(2013)showedhow to compute the total variance of the community composi-tion datamatrix from a dissimilaritymatrixD The total sum ofsquaresSS(Y)canbeobtainedfromadissimilaritymatrixDusingEquation1(LegendreampDeCaacuteceres2013LegendreampLegendre2012) Dividing SS(Y) by (n - 1) produces the classical unbiasedestimateofthetotalvarianceofYcomputedfromauser-selectedEuclidean dissimilarity matrixD (ie Equation 2)We used thatapproach to calculate the traditional compositional beta diversity (BDCOMP) the structural beta diversity (BDSTR) and the composi-tionalndashstructural beta diversity(BDCOMPndashSTR)respectivelyusingthefollowingequations
D =(Dhi)isanntimesnsymmetricdissimilaritymatrix(eitherDCOMP
DSTRorDCOMPndashSTR) i and h representthesamplingunitsn is thenumberofthesamplingunits If thecalculationsstartwithaper-centagedifferenceDmatrixwhichisnon-Euclideanonecomputesthesquare-rootsoftheDvaluesintheDmatrixtomakeitEuclideanbeforeusingthetransformedDvaluesinEquations1and2
225emsp|emspLocal contributions to beta diversity in terms of community composition and structure
Legendre andDe Caacuteceres (2013) suggested that total beta diver-sity can be partitioned into Local Contributions toBetaDiversity(LCBDwhicharecomparative indicatorsoftheecologicalunique-nessofthesites)TheLocalContributionstoBetaDiversity(LCBDi)represent the relative contributionsof the samplingunit i to betadiversityLCBDi indicateshowexceptional thecompositionofsitei iswhencomparedtothecentroidofallpointswhichwouldrep-resent a theoretical site with the average species composition ofall the sampling units In the present study the LCBD representsthe degree of uniqueness of each sampling unit in terms of com-position andor structure of community assemblages LCBDi indi-cescanbecalculatedfromthedissimilaritymatricesD(LegendreampDeCaacuteceres2013)OnefirsttransformsthedistancematrixDintomatrixA =(ahi)=(ndash05D2
hi) then centers thematrix as proposedbyGower(1966)
where Iisanidentitymatrixofsizen1isavectorofones(oflengthn)and1primeisitstranspose(LegendreampLegendre2012)HereeachdiagonalelementofmatrixGistheSSivalues(iethesquareddis-tancetothecentroidoftheithsamplingunit)Hencethevectoroflocalcontributionsofthesitestobetadiversity(LCBDi)is
The LCBD indices are scaled to sum to 1 We used functionldquoLCBDcomprdquointheadespatialRpackage(Drayetal2018)avail-ableonCRAN(httpsCRANR-projectorgpackage=adespatial)tocalculatetheLCBDindices
Wecheckedwhetherthere isacorrelationbetweentheLCBDcoefficientscalculatedfromspeciescompositionsizestructureorusingthetwocomponentstogetherHencewecalculatedSpearmanrank correlations pairwise between the three typesof LCBDvec-tors (ieLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRandLCBDSTRvsLCBDCOMPndashSTR)SincetheLCBDindicesindicatethede-greeofuniquenessof thesamplingunits in termsof their speciescompositionandorsizestructureweplottedtheLCBDvaluesonmapsof the30-haplotLargeLCBDvalues indicate thesites thathaveuniquespeciesassemblagesandsmallLCBDvaluesindicatethesites thathaveassemblagesthatareverysimilar to those inothersites Againwe expected LCBDCOMPndashSTR to be correlated to bothLCBDCOMP and LCBDSTR butwith the strength of the correlationdependingontheweightgiventostructuralvscompositionalinfor-mationWethusshowedthetwoextremecasesoftheLCBDmapaccordinga largestweighttothestructuralcomponentandcorre-spondinglythesmallestrelativeweighttothecompositionalcompo-nent(ie1-cmbinsize)andgivingthelargestrelativeweighttothecompositionalcomponent(ie15-cmbinsize)
226emsp|emspSets of explanatory variables environmental and spatial variables
Following Legendre etal (2009)we used altitude convexity andslope to construct third-degreepolynomial functions (ie yieldingninevariables)Themonomialswithexponentsallowthemodelingof nonlinear relationships between the topographic variables andthe response variablesWe calculated the aspect of a quadrat astheaverageangleofthefourtriangularplanesthatdeviatefromthenorthdirectionWe thusused the sin (aspect) and cos (aspect) inorder to include it in a linear regressionmodelWe thereforeob-tained11expandedtopographicvariablesWethencombinedthese11 expanded topographic variables with the eight soil variables(described insection21Studysitesanddatacollection) toobtainthe environmental variables data table (ie 19 variables) for eachquadratWe computed eigenfunctions of distance-basedMoranrsquoseigenvector maps (dbMEM also called Principal Coordinates ofNeighbour Matrices PCNM Borcard Legendre Avois-Jacquet amp
(1)SS(Y) =1
n
nminus1sum
h=1
nsum
i=h+1
D2hi
(2)BD = SS(Y)∕(nminus1)
(3)G =
(
Iminus11
n
)
A
(
Iminus11
n
)
(4)(LCBDi) = (SSi)∕SS(Y) = diag(G)∕SS(Y)
emspensp emsp | emsp261Journal of Vegetation Science
YAO et Al
Tuomisto2004LegendreampLegendre2012Legendreetal2009)acrossthe3000(10mtimes10m)750(20mtimes20m)and120(50mtimes50m)quadratsThedbMEMeigenfunctionswithpositiveeigenval-uesonlywereusedasspatialvariablesWeappliedforwardmodelselection(withpermutationtestsatthe5significanceleveloftheincrease in R2 ateachstep) toextract thesignificantenvironmentvariablesandeigenfunctionsofdbMEMusingthefunctionldquoforwardselrdquointhepackageadespatial(Drayetal2018)
227emsp|emspVariation partitioning of DCOMP DSTR and DCOMPndashSTR
Tocompare the influenceofniche-basedandspatialprocessesoncommunityassembly representedbycommunitycompositionsizestructure or the two components together distance-based re-dundancyanalysis (dbRDALegendreampAnderson1999Legendreamp Legendre 2012)was used to partition the variation of each ofthree community matrices (Borcard Legendre amp Drapeau 1992Legendre etal 2009 Peres-Neto Legendre Dray amp Borcard2006) Specificallyweused the two setsof explanatoryvariables(after forwardmodelselection) topartitionvariation in theprinci-palcoordinatetablesextractedfromDCOMPDSTRDCOMPndashSTRsepa-rately into fractions explained by the four different components(a)purehabitat (b)spatiallystructuredhabitat (c)purespaceand(d)undetermined (BorcardampLegendre1994Borcardetal1992DeCaacuteceresetal 2012Legendreetal 2009Myersetal 2013Punchi-Manage etal 2014) We hypothesized that the niche
processesareresponsiblefortheproportionofvariationexplainedbythepurehabitatandthespatially-structuredhabitatcomponents(a+b)(LaliberteacutePaquetteLegendreampBouchard2009Legendreetal2009)Whilewehypothesizedthattheproportionofvariationexplainedbythepurespatialcomponent(c)isrelatedtoindepend-ent biological processes (eg dispersal limitation competition fa-cilitationhistoricaleventsandJanzenndashConnelleffects)(LegendreampLegendre2012Legendreetal2009Punchi-Manageetal2014)Theundeterminedproportionofvariation(d)mayberelatedtosto-chastic processes or undefined non-spatially-structured biologicalor environmental variables (Dumbrell Nelson Helgason DythamampFitter2010)Thatallowedustoassesstherelativecontributionsoftheenvironmentalandspatialvariablestocommunityassemblyin termsof composition structureor taking the twocomponentstogetherAllanalyseswereperformedusingR(RCoreTeam2017)
3emsp |emspRESULTS
31emsp|emspPairwise dissimilarity in terms of community composition and structure
Wefound thatdissimilaritymatricescomputed fromspeciescom-position(DCOMP)sizestructure(DSTR)andconsideringbothcompo-nentstogether(DCOMPndashSTR)werecorrelatedHoweverthestrengthof the correlationdependedon the sizeofbinsused todiscretizethestructuralvariableandonthesizeof thequadrats (Figure1andashc) Overall the correlation between DCOMP vs DCOMPndashSTR was
F IGURE 1emspThecorrelationsbetweenthepairwisedissimilarityintermsofspeciescomposition(DCOMP)sizestructure(DSTR)andbothcomponentstogether(DCOMPndashSTR)ThecorrelationsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRand DSTR vs DCOMPndashSTRvarywithdbhbinsatthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofRVcoefficientsof1ndash15cmdbhbinsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTRrespectively(d)BoxplotsforRVcoefficientsofthethreepairwisedissimilaritycomparisons(aggregatedoverall1ndash15cmdbhbinsizes)foreachofthethreequadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
262emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
substantially stronger than that ofDCOMP vs DSTR andofDSTR vs DCOMPndashSTR
The correlation ofDCOMP vs DCOMPndashSTR increasedwith the in-crease of bin size Correspondingly the correlations of DSTR vs DCOMPndashSTR showed the opposite trend (Figure1andashc) As to the ef-fectofthesizeofthesamplingunitsthestrengthofcorrelationsin-creasedwiththequadratsize(Figure1d)exceptforthecorrelationbetweenDCOMP and DCOMPndashSTRwhichexhibitsnosignificantdiffer-encebetweenthe10mtimes10mand20mtimes20mquadrats(p = 023Figure1d)
32emsp|emspThe three components of beta diversity (BD) BDCOMP BDSTR and BDCOMPndashSTR
Thebetadiversity(BD)valueswerecloselyrelatedtowhetherthespecies composition size structure or both components togetherhad been taken into account Among these three components ofbetadiversityBDCOMPndashSTRwasgreatestcloselyfollowedbyBDCOMPandthesmallestwasBDSTR(Figure2)Sincethesizestructureofin-dividualswasnot consideredwhen calculatingBDCOMP this indexwasnotaffectedbythesizeofdbhbins(Figure2andashc)ThevaluesofBDCOMPndashSTRandBDSTRhoweverdecreasedslightlywithanincreaseofbinsizeWhenincreasingdbhbinsizethevaluesofBDCOMPndashSTR graduallyapproachedthevaluesofBDCOMP (Figure2andashc)BDalso
varied as a function of quadrat size (Figure2) values of BDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossbinsizes)systemati-callydecreasedwithincreasingquadratsize(Figure2d)
33emsp|emspLocal contributions to beta diversity in terms of community composition and structure
LocalContributionstoBetaDiversitycalculatedusingspeciescom-positionsizestructureorbothcomponentswerecorrelatedAgainthe strengthof correlationsdependedon the sizeofdbhbins andon the size of quadrats (Figure3andashc) In the case of LCBDCOMP vs LCBDCOMPndashSTR the strength of the correlation increased with anincreaseofbinsizeCorrespondinglythecorrelationofLCBDSTR vs LCBDCOMPndashSTRshowedtheoppositetrend(Figure3andashc)Thecorrela-tionsofLCBDCOMPvsLCBDSTRandLCBDSTRvsLCBDCOMPndashSTR were significantly different for differentquadrat sizesA striking findingwasthatthestrengthofcorrelationswasweakeratthescaleof20mtimes20m than that at the scalesof10mtimes10mor50mtimes50m(Figure3d)HowevercorrelationsbetweenLCBDCOMPvsLCBDCOMPndashSTRwerenotsubstantiallyaffectedbythesizeofquadrats(Figure3d)
F IGURE 2emspTheBetaDiversity(BD)intermsofspeciescomposition(BDCOMP)sizestructure(BDSTR)andbothcomponentstogether(BDCOMPndashSTR)ThevaluesofBDCOMPndashSTRandBDSTRvarywiththesizeofbinsofthestructuralvariable(dbhbinsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mThesizestructureofindividuals(iethedbh)isnotconsideredwhencalculatingtheBDCOMPthusthevaluesofBDCOMPwerenotaffectedbythebinsizeIngraphs(a)(b)and(c)thehorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofBDCOMPndashSTRandBDSTRacross1ndash15cmbinsizesrespectively(d)ValuesofBDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossdbhbinsizes)varywiththesamplingunitsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
emspensp emsp | emsp263Journal of Vegetation Science
YAO et Al
themeantobetadiversity)arevariedamongthreecomponentsofacommunity (Figure4)Specifically342(456)290(387)and331 (441)outof750quadratscontributedmorethanthemeanto beta diversity in term of species composition (ie LCBDCOMPFigure4a)sizestructure(ieLCBDSTRFigure4g)andbothcompo-nentstogether(ieLCBDCOMPndashSTRFigure4f)respectively
34emsp|emspVariation partitioning of matrices DCOMP DSTR and DCOMPndashSTR
Theexplanatorypoweroftheenvironmentalvariablesandthespa-tialvariablesvariedforthethreetypesofmatricesandwithquadratsizes (Table 1) The variation explained by the environmental vari-ables(a+b)andbythespatialvariables(b+c)increasedsystemati-callywith increasingscale (Table1)Averagingacrossquadratsizeshabitatandspacejointlyexplained438254and341ofthevariation in compositional component structural component andthe two components together of community assemblage respec-tivelyHoweverthecontributionofthepurehabitatcomponent (a)was negligible The combination of environmental and spatial vari-ablesexplained the lowestproportionofvariation in the structuralcomponentaloneandexplainedthehighestproportionofvariationinthecompositionalcomponentalone(Table1)Boththeenvironmen-talvariables(a+b)andthepurespatialvariables(c)explainedmore
Forest ecosystems can be characterized and evaluated in terms ofboththeirstructureandcomposition(Peet1992)Inpreviousstud-ies the compositional and structural components of a communityassemblagewereusuallyanalyzedseparately(egFangetal2012)Howeverthenatureofspeciesassemblagesindicatesthateitherspe-cies composition or size structure of constituent individuals alonemayoversimplifycommunityorganization (DeCaacuteceresetal2013)Changes in structure and compositionmay be onlyweakly related(egArsenaultampBradfield1995)thereforeassessmentofbothsi-multaneouslyisimportantwhenevaluatingcommunityassemblyInthepresentstudywegeneralizedtheconventionalapproachtocom-munityassemblagebyincorporatingstructuraldataofacommunityinadditiontocompositionaldatausingtheCAPframeworkToourknowledge this is the firstpaper that investigates inasinglestudythevariationinboththecompositionalandstructuralcomponentsofcommunityassemblagessimultaneouslyaswellasitsdeterminants
F IGURE 3emspThecorrelationsbetweentheLocalContributionstoBetaDiversity(LCBD)intermsofcommunitycomposition(LCBDCOMP)structure(LCBDSTR)andbothcomponentstogether(LCBDCOMPndashSTR)ThecorrelationsofLCBDCOMPvsLCBDSTRLCBDCOMP vsLCBDCOMPndashSTRandLCBDSTR vs LCBDCOMPndashSTRwiththesizeofbinsofthestructuralvariable(binsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofSpearmanrsquosrankcorrelationcoefficientr across 1ndash15 cm binsizeofLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRLCBDSTR vs LCBDCOMPndashSTRrespectively(d)BoxplotsfortheSpearmanrsquosrankcorrelationcoefficientrbetweenthepairwiseofthethreekindsofLCBDof1ndash15cmbinsizesatdifferentquadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
260emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
224emsp|emspBeta diversity components (BDCOMP BDSTR and BDCOMPndashSTR)
Conventionallybetadiversity (abbreviatedBD) is assessed fromasite-by-speciesdatamatrixotherbasiccharacteristics(egsizeofindividuals)ofthecommunityareignoredInordertogeneral-izetheconceptoftraditionalbetadiversitytoCAPdataweap-pliedtheindexproposedbyLegendreetal (2005)andLegendreandDeCaacuteceres(2013)tocomputebetadiversityasthevarianceofthecommunitydataLegendreandDeCaacuteceres(2013)showedhow to compute the total variance of the community composi-tion datamatrix from a dissimilaritymatrixD The total sum ofsquaresSS(Y)canbeobtainedfromadissimilaritymatrixDusingEquation1(LegendreampDeCaacuteceres2013LegendreampLegendre2012) Dividing SS(Y) by (n - 1) produces the classical unbiasedestimateofthetotalvarianceofYcomputedfromauser-selectedEuclidean dissimilarity matrixD (ie Equation 2)We used thatapproach to calculate the traditional compositional beta diversity (BDCOMP) the structural beta diversity (BDSTR) and the composi-tionalndashstructural beta diversity(BDCOMPndashSTR)respectivelyusingthefollowingequations
D =(Dhi)isanntimesnsymmetricdissimilaritymatrix(eitherDCOMP
DSTRorDCOMPndashSTR) i and h representthesamplingunitsn is thenumberofthesamplingunits If thecalculationsstartwithaper-centagedifferenceDmatrixwhichisnon-Euclideanonecomputesthesquare-rootsoftheDvaluesintheDmatrixtomakeitEuclideanbeforeusingthetransformedDvaluesinEquations1and2
225emsp|emspLocal contributions to beta diversity in terms of community composition and structure
Legendre andDe Caacuteceres (2013) suggested that total beta diver-sity can be partitioned into Local Contributions toBetaDiversity(LCBDwhicharecomparative indicatorsoftheecologicalunique-nessofthesites)TheLocalContributionstoBetaDiversity(LCBDi)represent the relative contributionsof the samplingunit i to betadiversityLCBDi indicateshowexceptional thecompositionofsitei iswhencomparedtothecentroidofallpointswhichwouldrep-resent a theoretical site with the average species composition ofall the sampling units In the present study the LCBD representsthe degree of uniqueness of each sampling unit in terms of com-position andor structure of community assemblages LCBDi indi-cescanbecalculatedfromthedissimilaritymatricesD(LegendreampDeCaacuteceres2013)OnefirsttransformsthedistancematrixDintomatrixA =(ahi)=(ndash05D2
hi) then centers thematrix as proposedbyGower(1966)
where Iisanidentitymatrixofsizen1isavectorofones(oflengthn)and1primeisitstranspose(LegendreampLegendre2012)HereeachdiagonalelementofmatrixGistheSSivalues(iethesquareddis-tancetothecentroidoftheithsamplingunit)Hencethevectoroflocalcontributionsofthesitestobetadiversity(LCBDi)is
The LCBD indices are scaled to sum to 1 We used functionldquoLCBDcomprdquointheadespatialRpackage(Drayetal2018)avail-ableonCRAN(httpsCRANR-projectorgpackage=adespatial)tocalculatetheLCBDindices
Wecheckedwhetherthere isacorrelationbetweentheLCBDcoefficientscalculatedfromspeciescompositionsizestructureorusingthetwocomponentstogetherHencewecalculatedSpearmanrank correlations pairwise between the three typesof LCBDvec-tors (ieLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRandLCBDSTRvsLCBDCOMPndashSTR)SincetheLCBDindicesindicatethede-greeofuniquenessof thesamplingunits in termsof their speciescompositionandorsizestructureweplottedtheLCBDvaluesonmapsof the30-haplotLargeLCBDvalues indicate thesites thathaveuniquespeciesassemblagesandsmallLCBDvaluesindicatethesites thathaveassemblagesthatareverysimilar to those inothersites Againwe expected LCBDCOMPndashSTR to be correlated to bothLCBDCOMP and LCBDSTR butwith the strength of the correlationdependingontheweightgiventostructuralvscompositionalinfor-mationWethusshowedthetwoextremecasesoftheLCBDmapaccordinga largestweighttothestructuralcomponentandcorre-spondinglythesmallestrelativeweighttothecompositionalcompo-nent(ie1-cmbinsize)andgivingthelargestrelativeweighttothecompositionalcomponent(ie15-cmbinsize)
226emsp|emspSets of explanatory variables environmental and spatial variables
Following Legendre etal (2009)we used altitude convexity andslope to construct third-degreepolynomial functions (ie yieldingninevariables)Themonomialswithexponentsallowthemodelingof nonlinear relationships between the topographic variables andthe response variablesWe calculated the aspect of a quadrat astheaverageangleofthefourtriangularplanesthatdeviatefromthenorthdirectionWe thusused the sin (aspect) and cos (aspect) inorder to include it in a linear regressionmodelWe thereforeob-tained11expandedtopographicvariablesWethencombinedthese11 expanded topographic variables with the eight soil variables(described insection21Studysitesanddatacollection) toobtainthe environmental variables data table (ie 19 variables) for eachquadratWe computed eigenfunctions of distance-basedMoranrsquoseigenvector maps (dbMEM also called Principal Coordinates ofNeighbour Matrices PCNM Borcard Legendre Avois-Jacquet amp
(1)SS(Y) =1
n
nminus1sum
h=1
nsum
i=h+1
D2hi
(2)BD = SS(Y)∕(nminus1)
(3)G =
(
Iminus11
n
)
A
(
Iminus11
n
)
(4)(LCBDi) = (SSi)∕SS(Y) = diag(G)∕SS(Y)
emspensp emsp | emsp261Journal of Vegetation Science
YAO et Al
Tuomisto2004LegendreampLegendre2012Legendreetal2009)acrossthe3000(10mtimes10m)750(20mtimes20m)and120(50mtimes50m)quadratsThedbMEMeigenfunctionswithpositiveeigenval-uesonlywereusedasspatialvariablesWeappliedforwardmodelselection(withpermutationtestsatthe5significanceleveloftheincrease in R2 ateachstep) toextract thesignificantenvironmentvariablesandeigenfunctionsofdbMEMusingthefunctionldquoforwardselrdquointhepackageadespatial(Drayetal2018)
227emsp|emspVariation partitioning of DCOMP DSTR and DCOMPndashSTR
Tocompare the influenceofniche-basedandspatialprocessesoncommunityassembly representedbycommunitycompositionsizestructure or the two components together distance-based re-dundancyanalysis (dbRDALegendreampAnderson1999Legendreamp Legendre 2012)was used to partition the variation of each ofthree community matrices (Borcard Legendre amp Drapeau 1992Legendre etal 2009 Peres-Neto Legendre Dray amp Borcard2006) Specificallyweused the two setsof explanatoryvariables(after forwardmodelselection) topartitionvariation in theprinci-palcoordinatetablesextractedfromDCOMPDSTRDCOMPndashSTRsepa-rately into fractions explained by the four different components(a)purehabitat (b)spatiallystructuredhabitat (c)purespaceand(d)undetermined (BorcardampLegendre1994Borcardetal1992DeCaacuteceresetal 2012Legendreetal 2009Myersetal 2013Punchi-Manage etal 2014) We hypothesized that the niche
processesareresponsiblefortheproportionofvariationexplainedbythepurehabitatandthespatially-structuredhabitatcomponents(a+b)(LaliberteacutePaquetteLegendreampBouchard2009Legendreetal2009)Whilewehypothesizedthattheproportionofvariationexplainedbythepurespatialcomponent(c)isrelatedtoindepend-ent biological processes (eg dispersal limitation competition fa-cilitationhistoricaleventsandJanzenndashConnelleffects)(LegendreampLegendre2012Legendreetal2009Punchi-Manageetal2014)Theundeterminedproportionofvariation(d)mayberelatedtosto-chastic processes or undefined non-spatially-structured biologicalor environmental variables (Dumbrell Nelson Helgason DythamampFitter2010)Thatallowedustoassesstherelativecontributionsoftheenvironmentalandspatialvariablestocommunityassemblyin termsof composition structureor taking the twocomponentstogetherAllanalyseswereperformedusingR(RCoreTeam2017)
3emsp |emspRESULTS
31emsp|emspPairwise dissimilarity in terms of community composition and structure
Wefound thatdissimilaritymatricescomputed fromspeciescom-position(DCOMP)sizestructure(DSTR)andconsideringbothcompo-nentstogether(DCOMPndashSTR)werecorrelatedHoweverthestrengthof the correlationdependedon the sizeofbinsused todiscretizethestructuralvariableandonthesizeof thequadrats (Figure1andashc) Overall the correlation between DCOMP vs DCOMPndashSTR was
F IGURE 1emspThecorrelationsbetweenthepairwisedissimilarityintermsofspeciescomposition(DCOMP)sizestructure(DSTR)andbothcomponentstogether(DCOMPndashSTR)ThecorrelationsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRand DSTR vs DCOMPndashSTRvarywithdbhbinsatthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofRVcoefficientsof1ndash15cmdbhbinsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTRrespectively(d)BoxplotsforRVcoefficientsofthethreepairwisedissimilaritycomparisons(aggregatedoverall1ndash15cmdbhbinsizes)foreachofthethreequadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
262emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
substantially stronger than that ofDCOMP vs DSTR andofDSTR vs DCOMPndashSTR
The correlation ofDCOMP vs DCOMPndashSTR increasedwith the in-crease of bin size Correspondingly the correlations of DSTR vs DCOMPndashSTR showed the opposite trend (Figure1andashc) As to the ef-fectofthesizeofthesamplingunitsthestrengthofcorrelationsin-creasedwiththequadratsize(Figure1d)exceptforthecorrelationbetweenDCOMP and DCOMPndashSTRwhichexhibitsnosignificantdiffer-encebetweenthe10mtimes10mand20mtimes20mquadrats(p = 023Figure1d)
32emsp|emspThe three components of beta diversity (BD) BDCOMP BDSTR and BDCOMPndashSTR
Thebetadiversity(BD)valueswerecloselyrelatedtowhetherthespecies composition size structure or both components togetherhad been taken into account Among these three components ofbetadiversityBDCOMPndashSTRwasgreatestcloselyfollowedbyBDCOMPandthesmallestwasBDSTR(Figure2)Sincethesizestructureofin-dividualswasnot consideredwhen calculatingBDCOMP this indexwasnotaffectedbythesizeofdbhbins(Figure2andashc)ThevaluesofBDCOMPndashSTRandBDSTRhoweverdecreasedslightlywithanincreaseofbinsizeWhenincreasingdbhbinsizethevaluesofBDCOMPndashSTR graduallyapproachedthevaluesofBDCOMP (Figure2andashc)BDalso
varied as a function of quadrat size (Figure2) values of BDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossbinsizes)systemati-callydecreasedwithincreasingquadratsize(Figure2d)
33emsp|emspLocal contributions to beta diversity in terms of community composition and structure
LocalContributionstoBetaDiversitycalculatedusingspeciescom-positionsizestructureorbothcomponentswerecorrelatedAgainthe strengthof correlationsdependedon the sizeofdbhbins andon the size of quadrats (Figure3andashc) In the case of LCBDCOMP vs LCBDCOMPndashSTR the strength of the correlation increased with anincreaseofbinsizeCorrespondinglythecorrelationofLCBDSTR vs LCBDCOMPndashSTRshowedtheoppositetrend(Figure3andashc)Thecorrela-tionsofLCBDCOMPvsLCBDSTRandLCBDSTRvsLCBDCOMPndashSTR were significantly different for differentquadrat sizesA striking findingwasthatthestrengthofcorrelationswasweakeratthescaleof20mtimes20m than that at the scalesof10mtimes10mor50mtimes50m(Figure3d)HowevercorrelationsbetweenLCBDCOMPvsLCBDCOMPndashSTRwerenotsubstantiallyaffectedbythesizeofquadrats(Figure3d)
F IGURE 2emspTheBetaDiversity(BD)intermsofspeciescomposition(BDCOMP)sizestructure(BDSTR)andbothcomponentstogether(BDCOMPndashSTR)ThevaluesofBDCOMPndashSTRandBDSTRvarywiththesizeofbinsofthestructuralvariable(dbhbinsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mThesizestructureofindividuals(iethedbh)isnotconsideredwhencalculatingtheBDCOMPthusthevaluesofBDCOMPwerenotaffectedbythebinsizeIngraphs(a)(b)and(c)thehorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofBDCOMPndashSTRandBDSTRacross1ndash15cmbinsizesrespectively(d)ValuesofBDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossdbhbinsizes)varywiththesamplingunitsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
emspensp emsp | emsp263Journal of Vegetation Science
YAO et Al
themeantobetadiversity)arevariedamongthreecomponentsofacommunity (Figure4)Specifically342(456)290(387)and331 (441)outof750quadratscontributedmorethanthemeanto beta diversity in term of species composition (ie LCBDCOMPFigure4a)sizestructure(ieLCBDSTRFigure4g)andbothcompo-nentstogether(ieLCBDCOMPndashSTRFigure4f)respectively
34emsp|emspVariation partitioning of matrices DCOMP DSTR and DCOMPndashSTR
Theexplanatorypoweroftheenvironmentalvariablesandthespa-tialvariablesvariedforthethreetypesofmatricesandwithquadratsizes (Table 1) The variation explained by the environmental vari-ables(a+b)andbythespatialvariables(b+c)increasedsystemati-callywith increasingscale (Table1)Averagingacrossquadratsizeshabitatandspacejointlyexplained438254and341ofthevariation in compositional component structural component andthe two components together of community assemblage respec-tivelyHoweverthecontributionofthepurehabitatcomponent (a)was negligible The combination of environmental and spatial vari-ablesexplained the lowestproportionofvariation in the structuralcomponentaloneandexplainedthehighestproportionofvariationinthecompositionalcomponentalone(Table1)Boththeenvironmen-talvariables(a+b)andthepurespatialvariables(c)explainedmore
Forest ecosystems can be characterized and evaluated in terms ofboththeirstructureandcomposition(Peet1992)Inpreviousstud-ies the compositional and structural components of a communityassemblagewereusuallyanalyzedseparately(egFangetal2012)Howeverthenatureofspeciesassemblagesindicatesthateitherspe-cies composition or size structure of constituent individuals alonemayoversimplifycommunityorganization (DeCaacuteceresetal2013)Changes in structure and compositionmay be onlyweakly related(egArsenaultampBradfield1995)thereforeassessmentofbothsi-multaneouslyisimportantwhenevaluatingcommunityassemblyInthepresentstudywegeneralizedtheconventionalapproachtocom-munityassemblagebyincorporatingstructuraldataofacommunityinadditiontocompositionaldatausingtheCAPframeworkToourknowledge this is the firstpaper that investigates inasinglestudythevariationinboththecompositionalandstructuralcomponentsofcommunityassemblagessimultaneouslyaswellasitsdeterminants
F IGURE 3emspThecorrelationsbetweentheLocalContributionstoBetaDiversity(LCBD)intermsofcommunitycomposition(LCBDCOMP)structure(LCBDSTR)andbothcomponentstogether(LCBDCOMPndashSTR)ThecorrelationsofLCBDCOMPvsLCBDSTRLCBDCOMP vsLCBDCOMPndashSTRandLCBDSTR vs LCBDCOMPndashSTRwiththesizeofbinsofthestructuralvariable(binsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofSpearmanrsquosrankcorrelationcoefficientr across 1ndash15 cm binsizeofLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRLCBDSTR vs LCBDCOMPndashSTRrespectively(d)BoxplotsfortheSpearmanrsquosrankcorrelationcoefficientrbetweenthepairwiseofthethreekindsofLCBDof1ndash15cmbinsizesatdifferentquadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
emspensp emsp | emsp261Journal of Vegetation Science
YAO et Al
Tuomisto2004LegendreampLegendre2012Legendreetal2009)acrossthe3000(10mtimes10m)750(20mtimes20m)and120(50mtimes50m)quadratsThedbMEMeigenfunctionswithpositiveeigenval-uesonlywereusedasspatialvariablesWeappliedforwardmodelselection(withpermutationtestsatthe5significanceleveloftheincrease in R2 ateachstep) toextract thesignificantenvironmentvariablesandeigenfunctionsofdbMEMusingthefunctionldquoforwardselrdquointhepackageadespatial(Drayetal2018)
227emsp|emspVariation partitioning of DCOMP DSTR and DCOMPndashSTR
Tocompare the influenceofniche-basedandspatialprocessesoncommunityassembly representedbycommunitycompositionsizestructure or the two components together distance-based re-dundancyanalysis (dbRDALegendreampAnderson1999Legendreamp Legendre 2012)was used to partition the variation of each ofthree community matrices (Borcard Legendre amp Drapeau 1992Legendre etal 2009 Peres-Neto Legendre Dray amp Borcard2006) Specificallyweused the two setsof explanatoryvariables(after forwardmodelselection) topartitionvariation in theprinci-palcoordinatetablesextractedfromDCOMPDSTRDCOMPndashSTRsepa-rately into fractions explained by the four different components(a)purehabitat (b)spatiallystructuredhabitat (c)purespaceand(d)undetermined (BorcardampLegendre1994Borcardetal1992DeCaacuteceresetal 2012Legendreetal 2009Myersetal 2013Punchi-Manage etal 2014) We hypothesized that the niche
processesareresponsiblefortheproportionofvariationexplainedbythepurehabitatandthespatially-structuredhabitatcomponents(a+b)(LaliberteacutePaquetteLegendreampBouchard2009Legendreetal2009)Whilewehypothesizedthattheproportionofvariationexplainedbythepurespatialcomponent(c)isrelatedtoindepend-ent biological processes (eg dispersal limitation competition fa-cilitationhistoricaleventsandJanzenndashConnelleffects)(LegendreampLegendre2012Legendreetal2009Punchi-Manageetal2014)Theundeterminedproportionofvariation(d)mayberelatedtosto-chastic processes or undefined non-spatially-structured biologicalor environmental variables (Dumbrell Nelson Helgason DythamampFitter2010)Thatallowedustoassesstherelativecontributionsoftheenvironmentalandspatialvariablestocommunityassemblyin termsof composition structureor taking the twocomponentstogetherAllanalyseswereperformedusingR(RCoreTeam2017)
3emsp |emspRESULTS
31emsp|emspPairwise dissimilarity in terms of community composition and structure
Wefound thatdissimilaritymatricescomputed fromspeciescom-position(DCOMP)sizestructure(DSTR)andconsideringbothcompo-nentstogether(DCOMPndashSTR)werecorrelatedHoweverthestrengthof the correlationdependedon the sizeofbinsused todiscretizethestructuralvariableandonthesizeof thequadrats (Figure1andashc) Overall the correlation between DCOMP vs DCOMPndashSTR was
F IGURE 1emspThecorrelationsbetweenthepairwisedissimilarityintermsofspeciescomposition(DCOMP)sizestructure(DSTR)andbothcomponentstogether(DCOMPndashSTR)ThecorrelationsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRand DSTR vs DCOMPndashSTRvarywithdbhbinsatthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofRVcoefficientsof1ndash15cmdbhbinsofDCOMP vs DSTRDCOMP vs DCOMPndashSTRandDSTR vs DCOMPndashSTRrespectively(d)BoxplotsforRVcoefficientsofthethreepairwisedissimilaritycomparisons(aggregatedoverall1ndash15cmdbhbinsizes)foreachofthethreequadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
262emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
substantially stronger than that ofDCOMP vs DSTR andofDSTR vs DCOMPndashSTR
The correlation ofDCOMP vs DCOMPndashSTR increasedwith the in-crease of bin size Correspondingly the correlations of DSTR vs DCOMPndashSTR showed the opposite trend (Figure1andashc) As to the ef-fectofthesizeofthesamplingunitsthestrengthofcorrelationsin-creasedwiththequadratsize(Figure1d)exceptforthecorrelationbetweenDCOMP and DCOMPndashSTRwhichexhibitsnosignificantdiffer-encebetweenthe10mtimes10mand20mtimes20mquadrats(p = 023Figure1d)
32emsp|emspThe three components of beta diversity (BD) BDCOMP BDSTR and BDCOMPndashSTR
Thebetadiversity(BD)valueswerecloselyrelatedtowhetherthespecies composition size structure or both components togetherhad been taken into account Among these three components ofbetadiversityBDCOMPndashSTRwasgreatestcloselyfollowedbyBDCOMPandthesmallestwasBDSTR(Figure2)Sincethesizestructureofin-dividualswasnot consideredwhen calculatingBDCOMP this indexwasnotaffectedbythesizeofdbhbins(Figure2andashc)ThevaluesofBDCOMPndashSTRandBDSTRhoweverdecreasedslightlywithanincreaseofbinsizeWhenincreasingdbhbinsizethevaluesofBDCOMPndashSTR graduallyapproachedthevaluesofBDCOMP (Figure2andashc)BDalso
varied as a function of quadrat size (Figure2) values of BDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossbinsizes)systemati-callydecreasedwithincreasingquadratsize(Figure2d)
33emsp|emspLocal contributions to beta diversity in terms of community composition and structure
LocalContributionstoBetaDiversitycalculatedusingspeciescom-positionsizestructureorbothcomponentswerecorrelatedAgainthe strengthof correlationsdependedon the sizeofdbhbins andon the size of quadrats (Figure3andashc) In the case of LCBDCOMP vs LCBDCOMPndashSTR the strength of the correlation increased with anincreaseofbinsizeCorrespondinglythecorrelationofLCBDSTR vs LCBDCOMPndashSTRshowedtheoppositetrend(Figure3andashc)Thecorrela-tionsofLCBDCOMPvsLCBDSTRandLCBDSTRvsLCBDCOMPndashSTR were significantly different for differentquadrat sizesA striking findingwasthatthestrengthofcorrelationswasweakeratthescaleof20mtimes20m than that at the scalesof10mtimes10mor50mtimes50m(Figure3d)HowevercorrelationsbetweenLCBDCOMPvsLCBDCOMPndashSTRwerenotsubstantiallyaffectedbythesizeofquadrats(Figure3d)
F IGURE 2emspTheBetaDiversity(BD)intermsofspeciescomposition(BDCOMP)sizestructure(BDSTR)andbothcomponentstogether(BDCOMPndashSTR)ThevaluesofBDCOMPndashSTRandBDSTRvarywiththesizeofbinsofthestructuralvariable(dbhbinsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mThesizestructureofindividuals(iethedbh)isnotconsideredwhencalculatingtheBDCOMPthusthevaluesofBDCOMPwerenotaffectedbythebinsizeIngraphs(a)(b)and(c)thehorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofBDCOMPndashSTRandBDSTRacross1ndash15cmbinsizesrespectively(d)ValuesofBDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossdbhbinsizes)varywiththesamplingunitsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
emspensp emsp | emsp263Journal of Vegetation Science
YAO et Al
themeantobetadiversity)arevariedamongthreecomponentsofacommunity (Figure4)Specifically342(456)290(387)and331 (441)outof750quadratscontributedmorethanthemeanto beta diversity in term of species composition (ie LCBDCOMPFigure4a)sizestructure(ieLCBDSTRFigure4g)andbothcompo-nentstogether(ieLCBDCOMPndashSTRFigure4f)respectively
34emsp|emspVariation partitioning of matrices DCOMP DSTR and DCOMPndashSTR
Theexplanatorypoweroftheenvironmentalvariablesandthespa-tialvariablesvariedforthethreetypesofmatricesandwithquadratsizes (Table 1) The variation explained by the environmental vari-ables(a+b)andbythespatialvariables(b+c)increasedsystemati-callywith increasingscale (Table1)Averagingacrossquadratsizeshabitatandspacejointlyexplained438254and341ofthevariation in compositional component structural component andthe two components together of community assemblage respec-tivelyHoweverthecontributionofthepurehabitatcomponent (a)was negligible The combination of environmental and spatial vari-ablesexplained the lowestproportionofvariation in the structuralcomponentaloneandexplainedthehighestproportionofvariationinthecompositionalcomponentalone(Table1)Boththeenvironmen-talvariables(a+b)andthepurespatialvariables(c)explainedmore
Forest ecosystems can be characterized and evaluated in terms ofboththeirstructureandcomposition(Peet1992)Inpreviousstud-ies the compositional and structural components of a communityassemblagewereusuallyanalyzedseparately(egFangetal2012)Howeverthenatureofspeciesassemblagesindicatesthateitherspe-cies composition or size structure of constituent individuals alonemayoversimplifycommunityorganization (DeCaacuteceresetal2013)Changes in structure and compositionmay be onlyweakly related(egArsenaultampBradfield1995)thereforeassessmentofbothsi-multaneouslyisimportantwhenevaluatingcommunityassemblyInthepresentstudywegeneralizedtheconventionalapproachtocom-munityassemblagebyincorporatingstructuraldataofacommunityinadditiontocompositionaldatausingtheCAPframeworkToourknowledge this is the firstpaper that investigates inasinglestudythevariationinboththecompositionalandstructuralcomponentsofcommunityassemblagessimultaneouslyaswellasitsdeterminants
F IGURE 3emspThecorrelationsbetweentheLocalContributionstoBetaDiversity(LCBD)intermsofcommunitycomposition(LCBDCOMP)structure(LCBDSTR)andbothcomponentstogether(LCBDCOMPndashSTR)ThecorrelationsofLCBDCOMPvsLCBDSTRLCBDCOMP vsLCBDCOMPndashSTRandLCBDSTR vs LCBDCOMPndashSTRwiththesizeofbinsofthestructuralvariable(binsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofSpearmanrsquosrankcorrelationcoefficientr across 1ndash15 cm binsizeofLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRLCBDSTR vs LCBDCOMPndashSTRrespectively(d)BoxplotsfortheSpearmanrsquosrankcorrelationcoefficientrbetweenthepairwiseofthethreekindsofLCBDof1ndash15cmbinsizesatdifferentquadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
262emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
substantially stronger than that ofDCOMP vs DSTR andofDSTR vs DCOMPndashSTR
The correlation ofDCOMP vs DCOMPndashSTR increasedwith the in-crease of bin size Correspondingly the correlations of DSTR vs DCOMPndashSTR showed the opposite trend (Figure1andashc) As to the ef-fectofthesizeofthesamplingunitsthestrengthofcorrelationsin-creasedwiththequadratsize(Figure1d)exceptforthecorrelationbetweenDCOMP and DCOMPndashSTRwhichexhibitsnosignificantdiffer-encebetweenthe10mtimes10mand20mtimes20mquadrats(p = 023Figure1d)
32emsp|emspThe three components of beta diversity (BD) BDCOMP BDSTR and BDCOMPndashSTR
Thebetadiversity(BD)valueswerecloselyrelatedtowhetherthespecies composition size structure or both components togetherhad been taken into account Among these three components ofbetadiversityBDCOMPndashSTRwasgreatestcloselyfollowedbyBDCOMPandthesmallestwasBDSTR(Figure2)Sincethesizestructureofin-dividualswasnot consideredwhen calculatingBDCOMP this indexwasnotaffectedbythesizeofdbhbins(Figure2andashc)ThevaluesofBDCOMPndashSTRandBDSTRhoweverdecreasedslightlywithanincreaseofbinsizeWhenincreasingdbhbinsizethevaluesofBDCOMPndashSTR graduallyapproachedthevaluesofBDCOMP (Figure2andashc)BDalso
varied as a function of quadrat size (Figure2) values of BDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossbinsizes)systemati-callydecreasedwithincreasingquadratsize(Figure2d)
33emsp|emspLocal contributions to beta diversity in terms of community composition and structure
LocalContributionstoBetaDiversitycalculatedusingspeciescom-positionsizestructureorbothcomponentswerecorrelatedAgainthe strengthof correlationsdependedon the sizeofdbhbins andon the size of quadrats (Figure3andashc) In the case of LCBDCOMP vs LCBDCOMPndashSTR the strength of the correlation increased with anincreaseofbinsizeCorrespondinglythecorrelationofLCBDSTR vs LCBDCOMPndashSTRshowedtheoppositetrend(Figure3andashc)Thecorrela-tionsofLCBDCOMPvsLCBDSTRandLCBDSTRvsLCBDCOMPndashSTR were significantly different for differentquadrat sizesA striking findingwasthatthestrengthofcorrelationswasweakeratthescaleof20mtimes20m than that at the scalesof10mtimes10mor50mtimes50m(Figure3d)HowevercorrelationsbetweenLCBDCOMPvsLCBDCOMPndashSTRwerenotsubstantiallyaffectedbythesizeofquadrats(Figure3d)
F IGURE 2emspTheBetaDiversity(BD)intermsofspeciescomposition(BDCOMP)sizestructure(BDSTR)andbothcomponentstogether(BDCOMPndashSTR)ThevaluesofBDCOMPndashSTRandBDSTRvarywiththesizeofbinsofthestructuralvariable(dbhbinsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mThesizestructureofindividuals(iethedbh)isnotconsideredwhencalculatingtheBDCOMPthusthevaluesofBDCOMPwerenotaffectedbythebinsizeIngraphs(a)(b)and(c)thehorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofBDCOMPndashSTRandBDSTRacross1ndash15cmbinsizesrespectively(d)ValuesofBDCOMPBDSTRandBDCOMPndashSTR(afteraveragingacrossdbhbinsizes)varywiththesamplingunitsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
emspensp emsp | emsp263Journal of Vegetation Science
YAO et Al
themeantobetadiversity)arevariedamongthreecomponentsofacommunity (Figure4)Specifically342(456)290(387)and331 (441)outof750quadratscontributedmorethanthemeanto beta diversity in term of species composition (ie LCBDCOMPFigure4a)sizestructure(ieLCBDSTRFigure4g)andbothcompo-nentstogether(ieLCBDCOMPndashSTRFigure4f)respectively
34emsp|emspVariation partitioning of matrices DCOMP DSTR and DCOMPndashSTR
Theexplanatorypoweroftheenvironmentalvariablesandthespa-tialvariablesvariedforthethreetypesofmatricesandwithquadratsizes (Table 1) The variation explained by the environmental vari-ables(a+b)andbythespatialvariables(b+c)increasedsystemati-callywith increasingscale (Table1)Averagingacrossquadratsizeshabitatandspacejointlyexplained438254and341ofthevariation in compositional component structural component andthe two components together of community assemblage respec-tivelyHoweverthecontributionofthepurehabitatcomponent (a)was negligible The combination of environmental and spatial vari-ablesexplained the lowestproportionofvariation in the structuralcomponentaloneandexplainedthehighestproportionofvariationinthecompositionalcomponentalone(Table1)Boththeenvironmen-talvariables(a+b)andthepurespatialvariables(c)explainedmore
Forest ecosystems can be characterized and evaluated in terms ofboththeirstructureandcomposition(Peet1992)Inpreviousstud-ies the compositional and structural components of a communityassemblagewereusuallyanalyzedseparately(egFangetal2012)Howeverthenatureofspeciesassemblagesindicatesthateitherspe-cies composition or size structure of constituent individuals alonemayoversimplifycommunityorganization (DeCaacuteceresetal2013)Changes in structure and compositionmay be onlyweakly related(egArsenaultampBradfield1995)thereforeassessmentofbothsi-multaneouslyisimportantwhenevaluatingcommunityassemblyInthepresentstudywegeneralizedtheconventionalapproachtocom-munityassemblagebyincorporatingstructuraldataofacommunityinadditiontocompositionaldatausingtheCAPframeworkToourknowledge this is the firstpaper that investigates inasinglestudythevariationinboththecompositionalandstructuralcomponentsofcommunityassemblagessimultaneouslyaswellasitsdeterminants
F IGURE 3emspThecorrelationsbetweentheLocalContributionstoBetaDiversity(LCBD)intermsofcommunitycomposition(LCBDCOMP)structure(LCBDSTR)andbothcomponentstogether(LCBDCOMPndashSTR)ThecorrelationsofLCBDCOMPvsLCBDSTRLCBDCOMP vsLCBDCOMPndashSTRandLCBDSTR vs LCBDCOMPndashSTRwiththesizeofbinsofthestructuralvariable(binsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofSpearmanrsquosrankcorrelationcoefficientr across 1ndash15 cm binsizeofLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRLCBDSTR vs LCBDCOMPndashSTRrespectively(d)BoxplotsfortheSpearmanrsquosrankcorrelationcoefficientrbetweenthepairwiseofthethreekindsofLCBDof1ndash15cmbinsizesatdifferentquadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
emspensp emsp | emsp263Journal of Vegetation Science
YAO et Al
themeantobetadiversity)arevariedamongthreecomponentsofacommunity (Figure4)Specifically342(456)290(387)and331 (441)outof750quadratscontributedmorethanthemeanto beta diversity in term of species composition (ie LCBDCOMPFigure4a)sizestructure(ieLCBDSTRFigure4g)andbothcompo-nentstogether(ieLCBDCOMPndashSTRFigure4f)respectively
34emsp|emspVariation partitioning of matrices DCOMP DSTR and DCOMPndashSTR
Theexplanatorypoweroftheenvironmentalvariablesandthespa-tialvariablesvariedforthethreetypesofmatricesandwithquadratsizes (Table 1) The variation explained by the environmental vari-ables(a+b)andbythespatialvariables(b+c)increasedsystemati-callywith increasingscale (Table1)Averagingacrossquadratsizeshabitatandspacejointlyexplained438254and341ofthevariation in compositional component structural component andthe two components together of community assemblage respec-tivelyHoweverthecontributionofthepurehabitatcomponent (a)was negligible The combination of environmental and spatial vari-ablesexplained the lowestproportionofvariation in the structuralcomponentaloneandexplainedthehighestproportionofvariationinthecompositionalcomponentalone(Table1)Boththeenvironmen-talvariables(a+b)andthepurespatialvariables(c)explainedmore
Forest ecosystems can be characterized and evaluated in terms ofboththeirstructureandcomposition(Peet1992)Inpreviousstud-ies the compositional and structural components of a communityassemblagewereusuallyanalyzedseparately(egFangetal2012)Howeverthenatureofspeciesassemblagesindicatesthateitherspe-cies composition or size structure of constituent individuals alonemayoversimplifycommunityorganization (DeCaacuteceresetal2013)Changes in structure and compositionmay be onlyweakly related(egArsenaultampBradfield1995)thereforeassessmentofbothsi-multaneouslyisimportantwhenevaluatingcommunityassemblyInthepresentstudywegeneralizedtheconventionalapproachtocom-munityassemblagebyincorporatingstructuraldataofacommunityinadditiontocompositionaldatausingtheCAPframeworkToourknowledge this is the firstpaper that investigates inasinglestudythevariationinboththecompositionalandstructuralcomponentsofcommunityassemblagessimultaneouslyaswellasitsdeterminants
F IGURE 3emspThecorrelationsbetweentheLocalContributionstoBetaDiversity(LCBD)intermsofcommunitycomposition(LCBDCOMP)structure(LCBDSTR)andbothcomponentstogether(LCBDCOMPndashSTR)ThecorrelationsofLCBDCOMPvsLCBDSTRLCBDCOMP vsLCBDCOMPndashSTRandLCBDSTR vs LCBDCOMPndashSTRwiththesizeofbinsofthestructuralvariable(binsizes=1ndash15cm)atthescaleof(a)10mtimes10m(b)20mtimes20mand(c)50mtimes50mIngraphs(a)(b)and(c)thehorizontalreddottedlineshorizontalbluelong-dashlinesandhorizontalgreensolidlinesrepresentthemeanvaluesofSpearmanrsquosrankcorrelationcoefficientr across 1ndash15 cm binsizeofLCBDCOMPvsLCBDSTRLCBDCOMPvsLCBDCOMPndashSTRLCBDSTR vs LCBDCOMPndashSTRrespectively(d)BoxplotsfortheSpearmanrsquosrankcorrelationcoefficientrbetweenthepairwiseofthethreekindsofLCBDof1ndash15cmbinsizesatdifferentquadratsizes[Colourfigurecanbeviewedatwileyonlinelibrarycom]
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
264emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
We found thatbothoverall betadiversity (BD) and the rela-tive contribution of sampling units to beta diversity (LCBD) de-pended on whether the species composition size structure orbothcomponentstogetherhadbeentakenintoaccountBetadi-versity partitioning indicated that the explanatory power of the
environmental and the spatial variables also varied widely withdifferentcomponentsofacommunityOur resultshighlight thatconsideringboth species compositional and size structural com-ponentsmaybeamorecomprehensivewaytodescribethecom-munityorganization
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
emspensp emsp | emsp265Journal of Vegetation Science
YAO et Al
41emsp|emspStructural and compositional components of forest variation
The framework of CAP allowed us to incorporate the distribu-tionof individualtreesize intotheanalysisofcommunityassem-blage thusmaking it possible toquantify the spatial variationofcommunitystructurebetadiversityEvensosuchstructuralbetadiversitycanbequantified independentlyor incombinationwithspecies composition TheBDCOMPndashSTR is the largest among thesethreecomponentsofbetadiversity indicating thatapplyingspe-ciescompositionaloneorsizestructurealonetoassessthebetadi-versitymayunderestimatethevariationofassemblages(Figure2)ThevaluesofBDCOMPareclosertotheBDCOMPndashSTRvaluesthanthatofBDSTR(theBDSTRvaluesarerelativelysmallFigure2)ThusasfarasourCAPframework isconcerned it seemsmoreappropri-atetoquantifybetadiversityusingthespeciescomposition indi-viduallythanusingthesizestructureindividuallyNeverthelessifstructureprovidesindependentinformationandisdeemedimpor-tantoneshouldincorporateitinBDassessmentAsbetadiversityindiceswerecalculatedfromdissimilaritymatrices thestructuralcomponent of beta diversity depended on the weight given to
structural vs compositional informationwhencalculatingdissimi-larity(Figure2andashc)Thelargerthebinsizes(iethesmallerweightgiventospeciesstructuralinformation)thecloserBDCOMPndashSTR val-uesapproachedthevaluesofBDCOMP(Figure2andashc)Ifthebinsizesare big enough theBDCOMPndashSTR value and theBDCOMP value are expected to converge at a certain size of dbh binNeverthelessconsideringthenecessityofcomprehensiveassessmentofbetadi-versityweadvocateforsmallbinsizesastheyprovidemoreinde-pendentstructuralinformationFinallyitisimportanttonotethatthisforestplotincludes47differenttreespecieswhichresultsinastrongrelativeweightofthecompositionalcomponentofBDCOMPndashSTRwhenusingtheCAPframeworkRepeatingourstudyinforestswith lower species richness or in this forest but using a coarsercompositional resolution (eg at the family level)would result inlargerrelativeweightofthestructuralcomponent
42emsp|emspLocal contributions to beta diversity in terms of community composition and structure
F IGURE 4emspMapsof30-ha(500mtimes600m)plotshowingthelocalcontributionstobetadiversity(LCBD)intermsofcommunitycompositionandstructurefor750quadrats(20mtimes20m)ThesolidcirclesrepresentthevaluesofLCBDiforeachithquadrat(i =[1750])(a)ThemapofLCBDsonlyintermsofspeciescompositionNotethatthesizestructureofindividuals(iedbh)isnotconsideredwhencalculatingtheLCBDCOMPthusthevaluesofLCBDCOMPwerenotaffectedbythesizeofthebinsofthestructuralvariable(b)ndash(e)ThetwoextremecasesoftheLCBDmap(b)and(c)givingthemostweighttothestructuralcomponentandcorrespondinglytheleastweighttothecompositionalcomponent(ie1-cmbinsize)and(d)and(e)givingthemostweighttothecompositionalcomponentandcorrespondinglytheleastweighttothestructuralcomponent(ie15-cmbinsize)(f)and(g)MapsofLCBDsafteraveragingacrossdbhbinsizesSizeofthecirclesisproportionaltotheLCBDivaluesTheblackandgreysolidcirclesrepresentthesiteswithLCBDvalueshigherandlowerthanthemeanrespectively
Fractions(a)ndash(d)(adjustedR2statistics)(a)variationexplainedbytheenvironmentalvariablesaftercontrollingforthespatialstructure(b)variationexplainedbythespatiallystructuredenvironmentalvariables(c)spatiallystructuredvariationexplainedbypurespaceaftercontrollingforenviron-mentalvariation(d)residualvariationEnvironmentalvariablesusedtocomputefraction(a+b)dbMEMeigenfunctionsweretheexplanatoryvaria-blesusedtocomputefraction(b+c)Only5-cm-diameterclasses(iebinsize=5cm)asthestructuralvariablewereusedtocalculatetheYSTR and YCOMPndashSTR
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
266emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
amp De Caacuteceres 2013) However natural communities may exhibitsimilarspeciescompositionsbutdifferinotherfeaturessuchasthesizestructureof individuals Inthepresentstudyweassessedthedegreeofuniquenessofthequadratsintermsnotonlyoftheirspe-ciescompositionbutalsooftheirsizestructureandbyusingbothcomponentstogetherThedegreeofuniquenessofquadratsintermsofcommunitycompositionindividuallyhadaveryweakcorrelationwithuniqueness in termsof size structure individually (LCBDCOMP vsLCBDSTRFigure3) indicating that sites that areunique in spe-cies composition are not necessarily unique in size structure andvice versa Additionally we found that the spatial distribution ofsiteswithhighLCBDvaluesisdifferentforthetwocomponentsofacommunitywith siteswithhighstructuraluniquenessoccurringinsmallforestpatches(Figure4)Alltheseresultsreinforcetheideathatconsideringbothspeciescompositionalandsizestructuralcom-ponentsmaybeamorecomprehensivewaytodescribethecommu-nityorganizationHoweverifweacceptthefactthatsitesthatareuniqueintermsofsizestructurearetheresultofgapdynamics(seebelow) the non-zero correlation between LCBDCOMP vs LCBDSTR mayindicatethatforestgapsmaybecolonizedbyspeciesthatmaylaterbesuppressedastheforestgrowssothatrecentforestgapshaveadifferentspeciescompositionthanclosedforeststructures(Comitaetal2009)
43emsp|emspPartitioning the structural and compositional components of beta diversity
About344ofthevariationincommunityassemblagewasdeter-minedbyenvironmentalandspatialvariablesdependingonthescale(quadratsize)andonwhichcomponentsofcommunityassemblage(ie compositional component structural component and takingbothcomponents together)were taken intoaccountThispropor-tion isslightly lowerthanthevaluesfound instudiesbyLegendreetal (2009)andPunchi-Manageetal (2014)Areasonforthisre-sult is thatwe incorporateddifferencesboth insizestructureandspeciescomposition intocommunityassemblagesratherthanonlyusingtheconventionalspeciescompositiondataWhenthespeciescomposition and size structure of the constituent individuals areincorporatedintothecommunityatthesametimemorevariationwilloccurincommunityassemblagesInourstudyhabitat(a+b)ex-plainedmorevariationinthecompositionalcomponent(190)thaninthestructuralcomponent(117)ofthecommunityassemblagesBetadiversitypartitioningindicatedthatthevariationinthestruc-turalcomponentislessdictatedbyenvironmentthanvariationinthecompositional componentWe here hypothesize that canopy gapdynamicswillbethepotentialdriversofstructuralvariationWinterinourstudyareaiscoldandlongwithalongsnowfallperiodThesnowfallperiodlastsforhalfayearandthesnowcoverthicknessinmountainousareasreaches40ndash50cmWehypothesizethatpulsesofmoderate-severity disturbancesmay be caused by snowstormsin our site In the absence of stand-replacing disturbances forestcanopiesareopenedperiodicallybythedeathofsinglebigtreesorsmallgroupsofadulttreescreatingcanopygapsSnowstormsmay
havealteredforeststructurebyselectivelyremovinglargercanopytrees Environmental selection of individuals shapes compositionbydeterminingthefitnessof individualswhereasstructuralvaria-tionmayhavesomerelationshipwithenvironmentalconditions(ielargertreesinsiteswherelargersizesaresupportedforenergyorwateravailability)butingeneralisthereflectionofdifferentstagesaroundgapdynamicsPreviousstudiesareconsistentwithourfind-ingsFraverandWhite(2005)forinstancefoundthattherepeatedmoderate-severity disturbances (iewindstorms) causeddramaticstructural changes they caused no significant change in speciescomposition
Becausetherelativeimportanceofbothnicheandneutralthe-oryinstructuringcommunitiesvarieswithspatialscale(Legendreetal 2009 Punchi-Manage etal 2014) we conducted scale-dependentanalyses InsharpcontrasttothefindingbyLegendreetal(2009)forabroad-leavedforestinChinawefoundthattheproportionofundeterminedvariation incompositionalandstruc-turalcomponentsofcommunityassemblageswasveryhighatfinespatialscales(upto945forthestructuralcomponent780forthecompositionalcomponentand844forbothcomponentsto-gether)butdecreasedsystematicallywith increasingspatial scale(up to a minimum of 373 for compositional component at the50-mscale)TheseresultsareinlinewiththefindingsbyPunchi-Manageetal(2014)inaSriLankandipterocarpforestandbyDeCaacuteceresetal(2012)inacomparisonofseveralforestsOntheonehandthehighproportionofunexplainedvariationmayberelatedtounmeasuredandnotspatially-structuredbiologicalorenviron-mentalvariablesXuetal(2016)showedthatthesoilnutrientsintheupper(0ndash10cmconsideredinourstudy)andlowersoillayers(10ndash20cm)andtheheavymetalelements(CuNiCdAsPbZnMoCrMnandMg)inthesoilshowastrongcorrelationwiththespeciesspatialdistributionsatJiaoheThismaypartlyexplainwhythepureenvironmentalvariable(a)explainedsuchlittlevariationinthecommunityassemblagesAnotherexplanationforthehighpro-portionofunexplainedvariationisthatitmaybeduetostochasticprocesseswhich related to theneutral theoryassuming that thedynamicsofpopulationsareprimarilydrivenbyecologicaldriftanddispersal(Legendreetal2009)Ontheotherhandtheproportionofundeterminedvariationincompositionalandstructuralcompo-nents of community assemblages decreased systematically withincreasingspatialscaleThismayindicatethatcommunityassem-blageishighlystochasticintermsofspeciescompositionandtreesizedistributionatfinescales (ie10-mscale)butthisfinescalestochasticitytendstosmoothoutatthe50-mscalewheremoreconsistent habitat-driven species assemblages emerged Whenvariance partitioning is conducted on the structural componentalonetheunexplained(d)fractionisdominantWhiletheinfluenceofenvironmental factorsonsizestructuremaybe less importantthan for thecompositionalcomponent theeffectof localdistur-bances (eg appearanceof canopygaps resulting frommortalityof largetrees)results inrandomspatialpatternsofquadratswithrather different structure contributing to a large unexplainedfraction
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
emspensp emsp | emsp267Journal of Vegetation Science
YAO et Al
5emsp |emspCONCLUSIONS
SpeciescompositionandsizestructurearethetwoessentialfeaturesofacommunityOnlyoneofthemindividuallymaybeinsufficienttodescribetheorganizationoftreespeciesassemblagesDefiningandquantifyingbetadiversityusingthespeciescompositionalonemaybesufficienttheninmanyoccasionsNeverthelessspeciescompo-sition is justonedimensionofbiodiversity variation in size struc-tureisalsoimportantIncorporatingstructuraldatainbetadiversityassessmentsallowsecologiststomakeuseofvaluableinformationcollectedduringfieldsurveysIfitisavailablethereisnoreasontoignorethewealthofinformationaboutsizestructurewhencompar-ingspeciesassemblagesOurstudyhighlightstheneedtoincorpo-ratethestructuraldataofacommunityinadditiontocompositionaldatawhenquantifyingandanalyzingbetadiversityFinallyourre-sults suggest thatbothdeterministicandstochasticprocessesarerelevantdeterminantsofcompositionalandstructuralcomponentsof communityassemblages inour temperate forestNeverthelesstheseprocessesarescale-andorresolution-dependent
ACKNOWLEDGEMENTS
WewouldliketothankHeHuaijiangDingShengjianNiRuiqiangandZuoQiangandseveralothersforassistingwiththefielddatacollectionAuthorsaregratefultotheJiaoheManagementBureauof the Forest Experimental Zone for permission to undertake thefieldworkWealsothankthreeanonymousreviewersforprovidingthevaluablecomments
Anderson M J Crist T O Chase J M Vellend M InouyeB D Freestone A L hellip Swenson N G (2011) Navigatingthe multiple meanings of β diversity a roadmap for the prac-ticing ecologist Ecology Letters 14 19ndash28 httpsdoiorg101111j1461-0248201001552x
ArsenaultAampBradfieldGE (1995)Structuralndashcompositionalvaria-tioninthreeage-classesoftemperaterainforestsinsoutherncoastalBritishColumbiaCanadian Journal of Botany7354ndash64httpsdoiorg101139b95-007
Borcard D amp Legendre P (1994) Environmental control and spatialstructureinecologicalcommunitiesanexampleusingoribatidmites(Acari Oribatei) Environmental and Ecological Statistics 1 37ndash61httpsdoiorg101007BF00714196
Borcard D Legendre P Avois-Jacquet C amp Tuomisto H (2004)DissectingthespatialstructureofecologicaldataatmultiplescalesEcology851826ndash1832httpsdoiorg10189003-3111
Chesson P (2000) Mechanisms of maintenance of species diversityAnnual Review of Ecology and Systematics31343ndash366httpsdoiorg101146annurevecolsys311343
ComitaLSUriarteMThompsonJJonckheereICanhamCDampZimmermanJK(2009)Abioticandbioticdriversofseedlingsur-vival inahurricane-impacted tropical forestJournal of Ecology971346ndash1359httpsdoiorg101111j1365-2745200901551x
DeCaacuteceresMFontXampOlivaF(2010)Themanagementofvegeta-tionclassificationswithfuzzyclusteringJournal of Vegetation Science211138ndash1151httpsdoiorg101111j1654-1103201001211x
De Caacuteceres M Legendre P amp He F (2013) Dissimilarity mea-surements and the size structure of ecological communitiesMethods in Ecology and Evolution 4 1167ndash1177 httpsdoiorg1011112041-210X12116
De Caacuteceres M Legendre P Valencia R Cao M Chang L-W Chuyong G hellip He F (2012) The variation of tree betadiversity across a global network of forest plots Global Ecology and Biogeography 21 1191ndash1202 httpsdoiorg101111j1466-8238201200770x
DraySBaumanDBlanchetGBorcardDClappeSGuenardGhellipWagnerHH(2018)adespatial Multivariate multiscale spatial analy-sis R package version 03-2RetrievedfromhttpsCRANR-projectorgpackage=adespatial
Dumbrell A J NelsonM Helgason T Dytham C amp Fitter A H(2010)Relativerolesofnicheandneutralprocesses instructuringa soil microbial community ISME Journal 4 337ndash345 httpsdoiorg101038ismej2009122
FaithDAustinMBelbinLampMargulesC (1985)Numericalclas-sification of profile attributes in environmental studies Journal of Environmental Management2073ndash85
Fang J Shen Z Tang ZWang XWang Z Feng J hellip Zheng C(2012) Forest community survey and the structural character-istics of forests in China Ecography 35 1059ndash1071 httpsdoiorg101111j1600-0587201300161x
FraverSampWhiteAS(2005)Disturbancedynamicsofold-growthPicea rubens forests of northern Maine Journal of Vegetation Science 16 597ndash610 httpsdoiorg101111j1654-11032005tb02401x
Gower J C (1966) Some distance properties of latent root and vec-tormethodsusedinmultivariateanalysisBiometrika53325ndash338httpsdoiorg101093biomet533-4325
Harms K E Condit R Hubbell S P amp Foster R B (2001)Habitat associations of trees and shrubs in a 50-ha neotrop-ical forest plot Journal of Ecology 89 947ndash959 httpsdoiorg101111j1365-2745200100615x
HilleRisLambers J Adler P B Harpole W S Levine J M ampMayfield M M (2012) Rethinking community assembly
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
268emsp |emsp emspenspJournal of Vegetation Science
YAO et Al
through the lens of coexistence theory Annual Review of Ecology Evolution and Systematics 43 227ndash249 httpsdoiorg101146annurev-ecolsys-110411-160411
HubbellSP(Ed)(2001)The unified theory of biodiversity and biogeogra-phyPrincetonNJPrincetonUniversityPress
Hubbell S P (2006) Neutral theory and the evolution of eco-logical equivalence Ecology 87 1387ndash1398 httpsdoiorg1018900012-9658(2006)87[1387NTATEO]20CO2
HussonFJosseJLeSampMazetJ(2015)FactoMineR Multivariate exploratory data analysis and data mining Version 1314 RetrievedfromhttpsCRANR-projectorgpackage=FactoMineR
Hutchinson G E (1961) The paradox of the plankton American Naturalist95137ndash145httpsdoiorg101086282171
KeddyPA(1992)Assemblyandresponserulestwogoalsforpredic-tive community ecology Journal of Vegetation Science3 157ndash164httpsdoiorg1023073235676
KoleffPGastonKJampLennonJJ(2003)MeasuringbetadiversityforpresencendashabsencedataJournal of Animal Ecology72367ndash382httpsdoiorg101046j1365-2656200300710x
KraftN JBComita L SChase JM SandersN J SwensonNGCrist TOhellipMyers JA (2011)Disentangling thedriversofβdiversityalong latitudinalandelevationalgradientsScience3331755ndash1758httpsdoiorg101126science1208584
Legendre P amp Anderson M J (1999) Distance-based redundancyanalysis testing multispecies responses in multifactorial ecolog-ical experiments Ecological Monographs 69 1ndash24 httpsdoiorg1018900012-9615(1999)069[0001DBRATM]20CO2
Legendre P Borcard D amp Peres-Neto P R (2005) Analyzing betadiversity partitioning the spatial variation of community com-position data Ecological Monographs 75 435ndash450 httpsdoiorg10189005-0549
OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH(2018)vegan Community ecology package Version 25-2Retrievedfromhttpscranr-projectorgpackage=vegan
Paradis EClaude JampStrimmerK (2004)APEAnalysesof phylo-genetics and evolution inR languageBioinformatics20 289ndash290httpsdoiorg101093bioinformaticsbtg412
PeetRK (1992)Community structureand function InDCGlenn-LewinRKPeetampTTVeblen(Eds)Plant succession theory and prediction(pp103ndash140)NewYorkNYChapmanampHall
Punchi-Manage R Wiegand T Wiegand K Getzin S SavitriGunatillekeCV ampNimalGunatilleke IAUN(2014)EffectofspatialprocessesandtopographyonstructuringspeciesassemblagesinaSriLankandipterocarpforestEcology95376ndash386httpsdoiorg10189012-21021
RCoreTeam (2017)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpswwwR-projectorg
Ricklefs R E (1990) Seabird life histories and the marine environ-mentsomespeculationsColonial Waterbirds13(1)1ndash6httpsdoiorg1023071521414
SoilScienceSocietyofChina(Ed)(1999)Soil agricultural chemical analy-sis procedureBeijingChinaChineseAgriculturalSciencePress
van der Plas F Janzen T Ordonez A FokkemaW Reinders JEtienneRSampOlffH(2015)Anewmodelingapproachestimatesthe relative importance of different community assembly pro-cesses Ecology961502ndash1515httpsdoiorg10189014-04541
VellendM(Ed)(2017)The theory of ecological communitiesPrincetonNJPrincetonUniversityPresshttpsdoiorg1015159781400883790
Weiner J (1990) Asymmetric competition in plant popula-tions Trends in Ecology and Evolution 5 360ndash364 httpsdoiorg1010160169-5347(90)90095-U
WhittakerRH (1960)VegetationoftheSiskiyoumountainsOregonand California Ecological Monographs 30 279ndash338 httpsdoiorg1023071943563
YanYZhangCWangYZhaoXampvonGadowK(2015)Driversofseedlingsurvivalinatemperateforestandtheirrelativeimportanceat threestagesofsuccessionEcology and Evolution54287ndash4299httpsdoiorg101002ece31688
How to cite this articleYaoJZhangCDeCaacuteceresMLegendrePZhaoXVariationincompositionalandstructuralcomponentsofcommunityassemblageanditsdeterminantsJ Veg Sci 201930257ndash268 httpsdoiorg101111jvs12708
