Multi-scale comparisons of tree composition in Amazonian terra firme forests

13 Pages·2009·1.68 MB·English

Biogeosciences,6,2719–2731,2009
Biogeosciences
www.biogeosciences.net/6/2719/2009/
©Author(s)2009. Thisworkisdistributedunder
theCreativeCommonsAttribution3.0License.
Multi-scale comparisons of tree composition in Amazonian terra
firme forests
E.N.HonorioCoronado1,4,T.R.Baker2,O.L.Phillips2,N.C.A.Pitman3,R.T.Pennington4,R.Va´squezMart´ınez5,
A.Monteagudo2,H.Mogollo´n6,N.Da´vilaCardozo7,M.R´ıos7,R.Garc´ıa-Villacorta7,E.Valderrama7,M.Ahuite7,
I.Huamantupa5,D.A.Neill8,W.F.Laurance9,H.E.M.Nascimento9,10,S.SoaresdeAlmeida11,T.J.Killeen12,
L.Arroyo13,P.Nu´n˜ez14,andL.FreitasAlvarado1
1InstitutodeInvestigacionesdelaAmazon´ıaPeruana,Av. A.Jose´ Quin˜oneskm2.5,Iquitos,Peru
2EarthandBiosphereInstitute,SchoolofGeography,UniversityofLeeds,Leeds,UK
3CenterforTropicalConservation,DukeUniversity,Durham,USA
4RoyalBotanicGardenEdinburgh,20aInverleithRow,EdinburghEH35LR,UK
5ProyectoFloradelPeru´,Jard´ınBota´nicodeMissouri,Oxapampa,Peru
6FindingSpecies,6930CarrollAve.,Suite600,P.O.Box5289,TakomaPark,MD20912,USA
7FacultaddeCienciasBiolo´gicas,UniversidadNacionaldelaAmazon´ıaPeruana,Iquitos,Peru
8MissouriBotanicalGarden,c/oNaturalezayCulturaInternacional,Loja,Ecuador
9SmithsonianTropicalResearchInstitute,Balboa,Panama
10BiologicalDynamicsofForestFragmentsProject,Manaus,Brazil
11MuseuParaenseEmilioGoeldi,66.040170Bele´m,Para´,Brazil
12CenterforAppliedBiodiversityScience,ConservationInternational,WashingtonD.C.,USA
13MuseoNoelKempffMercado,SantaCruz,Bolivia
14HerbarioVargas,UniversidadNacionalSanAntonioAbaddelCusco,Cusco,Peru
Received: 9December2008–PublishedinBiogeosciencesDiscuss.: 29January2009
Revised: 3November2009–Accepted: 3November2009–Published: 30November2009
Abstract. We explored the floristic composition of terra whereareaswithsimilarsoilconditionsweremorelikelyto
firme forests across Amazonia using 55 plots. Firstly, we shareahighnumberofspecies.Overall,theseresultssuggest
examinedthefloristicpatternsusingbothgenus-andspecies- thatregional-scalevariationinfloristiccompositioncanrival
level data and found that the species-level analysis more continental-scale differences within Amazonian terra firme
clearly distinguishes among forests. Next, we compared forests, and that variation in floristic composition at both
the variation in plot floristic composition at regional- and scales is influenced by geographical distance and environ-
continental-scales,andfoundthataverageamong-pairfloris- mental factors, such as climate and soil fertility. To fully
ticsimilarityanditsdecaywithdistancebehavesimilarlyat accountforregional-scalevariationincontinentalstudiesof
regional-andcontinental-scales. Nevertheless,geographical floristiccomposition,futurefloristicstudiesshouldfocuson
distancehaddifferenteffectsonfloristicsimilaritywithinre- forest types poorly represented at regional scales in current
gionsatdistances<100km,wherenorth-westernandsouth- datasets,suchasterrafirmeforestswithhighsoilfertilityin
western Amazonian regions showed greater floristic varia- north-westernAmazonia.
tionthanplotsofcentralandeasternAmazonia. Finally,we
quantifiedtheroleofenvironmentalfactorsandgeographical
distancefordeterminingvariationinfloristiccomposition.A
partialManteltestindicatedthatwhilegeographicaldistance 1 Introduction
appearedtobemoreimportantatcontinentalscales,soilfer-
tilitywascrucialatregionalscaleswithinwesternAmazonia, One of the scientific challenges in tropical forest ecology
is to map and understand the patterns of floristic composi-
Correspondenceto: tion and diversity (Prance et al., 2000; Phillips and Miller,
E.N.HonorioCoronado 2002). Describing these patterns is important for predict-
([email protected]) ingthemechanismsthatdeterminespeciesdistributionsand
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
2720 E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests
developingeffectiveconservationstrategiesinthefaceofde- notknownhowthisvariabilitycomparestocontinental-scale
forestation and climate change. Although progress is being patterns of floristic composition, or to what extent different
madeinassemblingthelargedatasetsthatarerequiredtoun- forestsinthisregionresemblecommunitiesinotherpartsof
derstand patterns of tropical forest diversity (e.g. ter Steege Amazonia. It is also important to consider floristic patterns
etal.,2006;Pitmanetal.,2008),largegapsremain,andthe at different scales, because the mechanisms that determine
role of different processes determining these patterns at lo- these patterns may differ. For example, at a continental-
cal,regional,andcontinentalscalesispoorlyunderstood. It scale,broadgradientsintreecompositioninAmazoniahave
isthereforeimportanttodevelopprinciplesfromexistingin- beenrelatedtovariationinenvironmentalconditionssuchas
formation that can both inform current conservation policy soilfertilityanddryseasonlength(TerborghandAndresen,
anddirectfutureresearch. 1998;terSteegeetal.,2000,2006).Theprincipalgradientin
While there are practical challenges with species-based floristiccompositioncontraststheeasternregionsofAmazo-
analyses in hyper-diverse Amazonian forests (e.g. Phillips nia(GuianaShieldandBrazil)thataregeologicallyolderand
et al., 2003), ideally it is preferable to study variation at havepoorersoils,withwesternareaswheresedimentsfrom
the species, rather than family or genus level, because it is theAndeshavebeendepositedmorerecently(Quesadaetal.,
speciesthattypicallyshowrestricteddistributionsandhence 2009a). Inaddition,asecondgradientincompositionisas-
should best define floristic patterns. For example, phyto- sociatedwiththegradientinclimateseasonalityfromsouth-
geographic patterns of Brazilian savannah woodland were eastern(southernBoliviaandcentralBrazil)tonorth-western
only clarified using species-level data on floristic composi- (Colombia, Ecuador, and northern Peru) Amazonia. How-
tion (Ratter et al., 2003) and gradients in the floristic com- ever, at a regional scale within western Amazonia, floristic
positionofplotsoftheAmazonianfloodplainwererecently patternshavebeenrelatedtodispersallimitationduetogeo-
resolvedbyspecies-levelanalysis(Wittmannetal.,2006).In graphicaldistance,thecapacityofafewgroupsofspeciesto
addition,foraseriesofforestinventoriesinwesternAmazo- dominate large areas (“oligarchies”), large geological units,
nia, Higgins and Ruokolainen (2004) showed that a reduc- as well as fine-scale soil heterogeneity (e.g. Pitman et al.,
tionintaxonomicresolution,fromspeciestogenustofamily 2001;Phillipsetal.,2003;Tuomistoetal.,2003a;Vormisto
level, resulted in a decrease in the mean floristic difference et al., 2004; Fine et al., 2005; Mac´ıa and Svenning, 2005;
betweensites(0.88,0.58and0.32respectively).Thisdecline Montufar and Pintaud, 2006; Ruokolainen et al., 2007; Pit-
suggested that species-level analysis best resolved floristic manetal.,2008;Duqueetal.,2009). Differentfactorsmay
differences between sites. However, in terms of describing well be important at different scales but the relative impor-
the patterns of floristic variation in Amazonia, most studies tanceofgeographicaldistanceandenvironmentalconditions,
have typically focused on either genus or family level com- atbothregionalandcontinentalscales,hasnotbeenstudied
parisons (e.g. Terborgh and Andresen, 1998; ter Steege et inAmazonia.
al., 2000, 2006). In contrast, studies at a species level have Inthisstudyweaddressthreequestionsrelatedtofloristic
usually either focused on restricted areas in western Ama- patternswithinAmazoniaterrafirmeforests: (1)Dogenus-
zonia (e.g. Higgins and Ruokolainen, 2004; Phillips et al., andspecies-leveldatagivesimilarpatternsoffloristiccom-
2003; Duque et al., 2009) or on a few taxa (e.g. Tuomisto position? (2) Is regional- and continental-scale variation in
etal.,2003a;Vormistoetal.,2004). Fullspecies-levelstud- floristiccompositionsimilarinmagnitude? (3)Doenviron-
ies of floristic composition remain difficult to carry out at mentalfactorsandgeographicaldistancehaveasimilarrole
a continental scale in Amazonia because of the high diver- inexplainingfloristicdissimilarityatregionalandcontinen-
sity and difficulties of developing datasets with consistent talscalesinAmazonianforests?
identificationsandnomenclature(HigginsandRuokolainen,
2004). However,largelybasedonrecenttaxonomicpublica-
tions(e.g.Va´squez,1997;JørgensenandLeo´n-Ya´nez,1999; 2 Materialsandmethods
Ribero et al., 1999), current ecological datasets do contain
reliablespecies-levelinformationformanytaxa. Thesedata 2.1 Treefloristicplotdata
couldofferinsightsintowhetherresultsfromcurrentfamily-
andgenus-levelanalysesoffloristiccompositionarelikelyto We compiled 55 floristic tree inventories of terra
resemblefuturefullspecies-levelanalyses. firme Amazonian forests (supplementary mate-
Asecondfeatureofcurrentpublishedanalysesisthatthey rial (http://www.biogeosciences.net/6/2719/2009/
typically focus on a single spatial scale. However, under- bg-6-2719-2009-supplement.pdf), Fig. 1), of which 30
standing the relative magnitude of regional and continental plots are in north-western (NWA; Ecuador and Peru), 13
variation in species composition is important for assessing in south-western (SWA; Peru and Bolivia), ten in central
thesensitivityofcontinental-scalecompositionalpatternsto (CA; Brazil), and two in eastern regions of Amazonia (EA;
restrictedsamplingofregionalfloristicvariation. Forexam- Brazil). These plots represent the broad gradients in soil
ple, north-western Amazonia is known for its high beta di- fertilityanddryseasonlengthacrossAmazonia(Sombroek,
versity at a regional scale (Tuomisto et al., 2003a), but it is 2000; 2001). None of the plots are believed to have had
Biogeosciences,6,2719–2731,2009 www.biogeosciences.net/6/2719/2009/
E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests 2721
Figure1.
Fig.1.LocationofAmazonianterrafirmeplotsinSouthAmerica,highelevationareasarerepresentedinblack.Outsetshowsnorth-western
Amazonianplots.
recent direct human impact (Freitas, 1996; Phillips et al., matrixoffloristicdistancebetweensiteswascalculatedsep-
2003; Pitman et al., 2008). Unusual formations of terra arately for genera and species using the Bray-Curtis index
firmeforest,suchaswhitesandforestandlianaforest,were basedontherelativeabundanceofthesetaxa(Pitmanetal.,
excluded from these analyses. We restricted the dataset 2008). Pearson’s correlation coefficient was used to exam-
to plots that have one hectare of inventoried trees with a ine the similarity of the two distance matrices (cf. Higgins
diameter ≥10cm (diameter at breast height, DBH), with andRuokolainen,2004). Next,thevariationinfloristiccom-
information on the number of individuals and species, and position at the genus and species level was examined using
with voucher collections in herbaria. Prior to analysis, DetrendedCorrespondenceAnalysis(DCA)withPASTver-
every scientific name was checked to validate its existence sion 1.82b (Hammer et al., 2001). The species-level DCA
and to detect synonymy. Over 130 references including analysiswasrunusingallspeciesthatoccurintwoormore
monographs, floras, checklists and revisions were used plots,whilethegenus-levelanalysisusedallgenerathatoc-
duringthisprocess. cur at one or more plots. Finally, the seven most abun-
dant families and genera were used to examine whether the
2.2 Comparison of floristic analyses using genus- and patterns of floristic composition along the main axis of the
species-leveldata species-level DCA were consistent within different families
1
andgenera. Withineachfamily,onegenus,andwithineach
Two floristic analyses (distance matrices and floristic ordi- genus, two species were selected (the genus/species with
nation) were used to compare patterns of floristic composi- thehighestabundanceandthehighestcorrelation(Kendall’s
tion obtained using genus- and species-level data. First, a tau value) between the axis 1 scores and species relative
www.biogeosciences.net/6/2719/2009/ Biogeosciences,6,2719–2731,2009
2722 E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests
abundance). Forthesetwotaxawithineachfamilyorgenus, plementarymaterial:http://www.biogeosciences.net/6/2719/
thedirectionofthecorrelationbetweentheaxis1scoresand 2009/bg-6-2719-2009-supplement.pdf).Inaddition,wealso
relativeabundancewascompared. used plot-level soil data from two databases, one related to
plots of NWA (Pitman et al., 2008) and the other related to
2.3 Variationoffloristiccompositionatcontinentaland plotsofRAINFOR(Quesadaetal.,2009b).Thesedataquan-
regionalscales tify the nutrient content of surface soils (0–20cm) per each
plot based on at least five random sampling points. Sam-
Theaveragefloristicsimilarityandthedecayinfloristicsim- pleswereanalyzedatLaMolinaNationalAgrarianUniver-
ilarity with distance were used to compare continental- and sity, Peru (Pitman et al., 2008) or the University of Leeds,
regional-floristicpatterns. Asabove,floristicsimilaritywas UK (Quesada et al., 2009b). Soil fertility was quantified in
calculatedusingtheBray-Curtisindexbasedontherelative threewaysusingthesedata. Firstly,usingtheexchangeable
abundance of species. First, the average (±95% confidence cations (Ca++, Mg++, K+, Na+, Al+++) and the sum of
limit)floristicsimilaritywascalculatedatacontinentalscale basecations(SB=Ca+++Mg+++K++Na+)(Huston, 1980),
using all 55 plots and for each region using 30 plots from secondly, including only two exchangeable cations (Ca++
NWA, 13 plots from SWA and 12 plots C&E Amazonia, and Mg++) and thirdly using all the additional available
based on 1000 randomly selected pairs of plots within each soil data (percentage of sand, clay and silt, and pH in both
group.Thedistributionoffloristicsimilaritieswascompared databases; percentage of organic material, phosphorus, and
for all plots, and separately for plots located at distances of potassium – Pitman et al., 2008, and total reserve bases,
<100or≥100km.Next,thedecayoffloristicsimilaritywith total extractable P, total P, and total N and C – Quesada
distancewascomparedbetweenregionsandatacontinental et al., 2009b). Because of the differences in protocols to
scale. Values of geographical distance between plots were analyse soil fertility, the plots associated with each detailed
transformedusingnaturallog[ln(x)]torepresenttheneutral soil database were analysed separately. Soil data were log-
theory(Hubbell,2001),thatpredictsnon-lineardistancede- transformedfollowingPhillipsetal.(2003). Differencesbe-
cayinfloristicsimilarity. Overlapping95%confidencelim- tweenplotswereexpressedinEuclideandistancescomputed
its of the intercepts and slopes of the relationship between separatelyforeachfactor.
floristicdistanceandln(distance)wasusedtotestforsignif-
icant differences in the decay of floristic similarity between
regionsandatacontinentalscale. 3 Results
2.4 Roleofenvironmentalfactorsandgeographicaldis- 3.1 Floristicdata
tanceinexplainingfloristicdissimilarityatregional
andcontinentalscales The 55 floristic inventories (plots) compiled from Amazo-
nia have a total of 32515 trees with diameter ≥10cm, of
PartialManteltestswereusedtotesttherelativeinfluenceof which55%werefromthethirtyplotsofnorth-westernAma-
geographicaldistanceandenvironmentalfactors,suchascli- zonia(NWA),22%fromthe13plotsofsouth-westernAma-
mate(dryseasonlength-DSL)andsoilfertility,asdetermi- zonia(SWA)andtherestfromthetwelveplotsofcentraland
nantsofthefloristicdissimilarityatcontinentalandregional easternAmazonia(C&EA).Acrossalltheplots,onaverage,
scales (Tuomisto et al., 2003a; Ruokolainen et al., 2007). 99.0%oftreeswereidentifiedtofamily,95.8%togenus,and
ThepartialManteltestinvolvescomputingthePearsoncor- 73.2%to species. Aftertheexclusion ofthe26.8% oftrees
relation coefficient or standardized Mantel statistic (r) as a with no reliable species-level scientific name, 93 families,
measureofthestrengthofrelationshipbetweentwodistance 473 genera and 1661 species remained in the entire dataset
matrices, while controlling for correlations with a third dis- ofwhich512speciesoccurredonlyinoneplot.
tancematrix. SignificancewasassessedusingaMonteCarlo
randomization procedure to estimate the probability of er- 3.2 Comparison of floristic analyses using genus- and
rorbycomparingobserveddistributionsofragainstthedis- species-leveldata
tribution of random values generated by permuting one of
thematricesandrecalculatingr 999times(p<0.001). The Thecorrelationbetweendistancematricesdemonstratedthat
floristic dissimilarity between two plots was calculated us- only 57% of the floristic variation at the species level was
ing the Bray-Curtis index based on the relative abundance explainedbythegenus-leveldata. Overall,thespecies-level
ofspecies. Dryseasonlengthwascalculatedastheaverage analysis detected greater floristic differences between plots
numberofmonthsperyearwithrainfall<100mm.Avariety than the genus-level analysis as the average floristic dif-
ofmethodsforquantifyingsoilfertilitywereuseddepending ference between sites increased when the level of analysis
on the available data. First, for all plots, we used a sim- changed from genus (0.63) to species (0.87). However, or-
pleclassificationdevelopedbyMalhietal.(2004),inwhich dinationusingDCAshowedthattheaffinitiesbetweenplots
soilswereclassifiedineightsoilfertilitycategories(seesup- atthegenusandspecieslevelwerebroadlysimilar(Fig.2).
Biogeosciences,6,2719–2731,2009 www.biogeosciences.net/6/2719/2009/
Figure2a.
E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests 2723
(a)
2
s
xi
A
Figure2b.
Axis1
(b)
2
s
xi
A
Axis1
Fig.2. DCAordinationsoftherelativeabundanceof(a)473genera, and(b)1149species(excludingthosethatoccuratonlyoneplot)
occurringin55Amazonianterrafirmeplots.Symbols:(1)CandE;((cid:13))NW,and((cid:3))SWAmazonia.
The relatively low correlation between species- and genus- analysis (Fig. 2). If these plots are excluded, the axis 1
level distance matrices was caused by a specific group of DCA scores for the two ordinations calculated using genus
sites:plotsfromManausandCaxiuana(C&EA),JenaroHer- and species data, were very similar and closely correlated
rera,QuebradaBlancoandsomeoftheplotsfromtheNapo (slope=1.06±0.16,r2=0.84).
River (NWA) were rather similar in terms of genus-level
Additional results also suggested that the species-
composition,andwereonlydistinguishedusingspecies-level
level data were more effective than the genus level at
1
www.biogeosciences.net/6/2719/2009/ Biogeosciences,6,2719–2731,2009
1
2724 E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests
distinguishing between plots. For example, in the DCA at ences when only plots less than 100km apart were consid-
the genus level, the first axis explained only 42.0% of the eredseparately(Fig.4). Atthisscale,allplotswithinC&EA
floristic variation, while at the species level, the equivalent have relatively high similarity values. In contrast, in NWA
valuewas72.5%. Inaddition,thepercentageofspecieswith and SWA some pairs of plots were floristically very differ-
astrongcorrelationwiththefirstaxis(6.4%ofspecieswith ent(<10%ofsimilarity)evenatsuchsmallscales(e.g.Su-
tau≥0.5)washigherthanthepercentageofgenera(3.2%of cusari(plot23)andRioOrosa(plot82)inLoretoandplotsof
generawithtau≥0.5). CuzcoAmazonico(plots17and19)andTambopata(plot28)
In general, along the first axis, plots were distributed inMadredeDios).
according to the regions where they occurred, along an Floristic similarity between plot pairs declined with
east/west axis: plots with low scores on axis 1 were found increasing distance at continental and regional scales as
in C&E Amazonia and plots with high axis 1 scores oc- a natural-log function of distance. Distance explained
cur in WA (Fig. 2). Lecythidaceae, Sapotaceae and Burs- a lower proportion of the variation in floristic simi-
eraceae and the genera Eschweilera, Pouteria and Protium larity in NWA than in other regions (r2: ALL=0.662;
increasedinabundancefromEAtoWA,whiletheMoraceae NWA=0.274; SWA=0.577; C&EA=0.889). Never-
and Arecaceae and their genera Pseudolmedia and Iriartea theless, neither the slopes nor the intercepts of the
showed the opposite trend; the only well-distributed family linear regressions were significantly different between
was the Fabaceae. The Myristicaceae and its genus Iryan- any region or with the continental-scale pattern (slope
theraweremoreabundantinmostofthenorth-westernAma- ±95% CI: ALL=−0.053±0.002; NWA=−0.042±0.007;
zonianplots. Whilethemostabundantspeciesshowedsimi- SWA=−0.046±0.009; C&EA=−0.051±0.004; intercepts
larrelationshipsbetweentheirrelativeabundanceandaxis1 ±95% CI: ALL=0.834±0.026; NWA=0.704±0.081;
scores as their respective genus and family, other species, SWA=0.759±0.102;C&EA=0.844±0.041).
such as Pouteria engleri and Inga capitata (Fig. 3) demon-
strated that rather different patterns could be found within 3.4 Roleofenvironmentalfactorsandgeographicaldis-
some groups, highlighting the value of species-level analy- tanceinexplainingfloristicdissimilarityatregional
ses. andcontinentalscales
The partial Mantel test showed that geographical distance,
3.3 Variation of floristic composition at regional and
soil fertility and climate all explained significant parts of
continentalscales
floristicdissimilaritybothatcontinentalandregionalscales.
Usingthesoilclassesasameasureofsoilfertilityataconti-
Atacontinentalscale,floristicsimilarityvaluesrangedfrom
nental scale, geographical distance (41.1%) explained more
0.3% (plot 41 vs. plot 78) to 70.3% (plot 42 vs. plot 43)
of the floristic dissimilarity than soils (22.9%) and climate
showing that some plots contained almost entirely different
(5.8%).AsimilarpatternwasobservedinSWA,althoughthe
species, while others were almost identical. Even though
contributionofclimatewasnotsignificant(0.5%,p>0.05).
high similarity values were common between plots located
In contrast, in NWA, the most important factor was soils
closed to each other, high values were also found between
(11.3%) followed by geographical distance (7.1%) and cli-
plots separated by great distances. For example, a similar-
ityvalueof17.0%±1.5wasfoundbetweenplotsfromTam- mate(6.7%). Whenplot-levelsoildatawereused,thiscon-
trastbetweentherelativeimportanceofsoilsanddistancein
bopataandCuzcoAmazonicoinSWA,andplotsfromYana-
NWAandatacontinentalscaleremained. However,thepat-
mono, Buenavista, RioOrosaandJatunSachainNWAthat
are930–1600kmapart. Highsimilarityvalues(13.2%±1.2) ternsforSWAwerelessrobust,withaswitchfromahigher
relativeimportanceofdistancetoahighervalueforsoilfer-
werealsofoundbetweenplotsseparatedbysimilarlongdis-
tility (10.3%). When only Ca and Mg were considered, the
tances (approx. 1550km), between plots from Jenaro Her-
percentageofthevariationexplainedbyeachfactorwasusu-
rera in NWA and plots from Manaus in CA. At a regional
ally higher than when all cations or soil data were included
scale,floristicsimilarityvaluesrangedfrom1.6%(plot4vs.
atbothscales(Table1).
plot 34) to 61.5% (plot 77 vs. plot 78) in NWA, from 4.8%
(plot29vs.plot41)to70.3%(plot42vs.plot43)inSWA,
and from 13.1% (plot 35 vs. plot 54) to 57.9% (plot 56 vs.
4 Discussion
plot57)inC&EA.
The average floristic similarity between plots at a con- 4.1 Do genus- and species-level data provide similar
tinental scale was 0.135 – lower than the values for the patternsoffloristiccomposition?
different regions – 0.198 in NWA, 0.224 in SWA, and
0.366 in C&E Amazonia. However, none of these val- Overall,genus-andspecies-levelinformationproducedsim-
ues were significantly different (95% Confidence limits: ilarfloristicpatterns(Fig.2). However,thedecreaseinaver-
ALL=0.021–0.475; NWA=0.054–0.417; SWA=0.059– age floristic dissimilarity (0.87 to 0.63), caused by a reduc-
0.603; C&EA=0.138–0.562). There were greater differ- tion in taxonomic resolution from species to genus, clearly
Biogeosciences,6,2719–2731,2009 www.biogeosciences.net/6/2719/2009/
E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests 2725
Figure3.
%)
(
n.
u
b
A
el.
R
%)
(
n.
u
b
A
el.
R
%)
(
n.
u
b
A
el.
R
%)
(
n.
u
b
A
el.
R
%)
(
n.
u
b
A
el.
R
%)
(
n.
u
b
A
el.
R
%)
(
n.
u
b
A
el.
R
Axis1 Axis1 Axis1
Family Genus Species
Fig.3.Distributionofthesevenmostabundantfamilies,generaandspeciesalongthemainaxisoftheDCAusingspecies-leveldata.Species
withthehighestvalueofrelativeabundancewithinthegenusareinblack(left)andwiththehighesttauvalueareingrey(right). Minimum
andmaximumvaluesoftheverticalaxesofthegraphsvaryasfollows,startingfromthetop:Families(1)3–23.7;(2)0–23.5;(3)0.4–19.0;
(4)0.4–16.8;(5)1.0–21.6;(6)0–30.8;(7)0–17;Genera(1)0.2–7.7;(2)0–19.2;(3)0–10.8;(4)0.2–10.5;(5)0–17.4;(6)0–23.4;(7)0–16.3;
Species(1A)0–0.9;(1B)0-0.5;(2A)0-10.5;(2B)0-1.4;(3A)0-6.4;(3B)0-1;(4A)0-4.8;(4B)0–1.7;(5)0–12.7;(6)0–23.4;(7A)0–8.1;
(7B)0–2.4.
www.biogeosciences.net/6/2719/2009/ Biogeosciences,6,2719–2731,2009
Figure4.
2726 E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests
80
)
%
(
y
c
en 40
u
q
e
r
F
0
80
)
%
(
y
c
n 40
e
u
q
e
r
F
0
80
)
%
(
y
c
n 40
e
u
q
e
r
F
0
80
)
%
(
y
c
n 40
e
u
q
e
r
F
0
Fig.4.Frequencyofsimilarityindexforallregionsandeveryregionconsideringpairsofplotslocatedlessthan100kmapartandmorethan
100kmapart.
demonstrated that the species-level analysis was able to de- Blanco,andsomeplotsfromtheNapoRiver-NWA)thatwere
tect a finer level of detail in the floristic patterns. This floristically similar at the genus level, but clearly differen-
decrease in floristic dissimilarity was very similar to that tiated by their species (Fig. 2). At the genus level, their
foundforaseriesofforestinventoryplotsinwesternAma- similarity emerged due to the high relative abundance of
zoniathatincludedtrees≥2.5cmdiameter(dissimilarityde- Eschweilera (Lecythidaceae), Pouteria (Sapotaceae), Lica-
creased from 0.88 to 0.58 using species- and genus-level nia (Chrysobalanaceae) and Protium (Burseraceae). How-
analyses respectively; Higgins and Ruokolainen, 2004) and ever, at the species level, even though all of these plots
isconsistentwiththeexpectationthatmostgenerashouldbe sharedahighabundanceofEschweileracoriacea,therewere
morewidelydistributedthanmostspecies. clear differences among the sites. For example, plots at Je-
naro Herrera contain many rarely collected species of the
Within the overall similarity of floristic patterns at
familiesSabiaceae(Ophiocaryumheterophyllum),Theaceae
the species and genus level, there was a group of
(Gordoniafructicosa),Styracaceae(Styraxheteroclitus)and
plots(Caxiuana-EA,Manaus-CA,JenaroHerrera,Quebrada
Biogeosciences,6,2719–2731,2009 www.biogeosciences.net/6/2719/2009/
E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests 2727
Anisophyllaceae (Anisophyllea guianensis; Spichiger et al.,
Table1.Relativeimportanceofgeographicaldistance,soilfertility
1989, 1990). In the 55 plots assembled, these species were
andclimateinexplainingvariationinfloristicdissimilarityatconti-
only present in up to four plots from the Napo River, Que- nentalandregionalscales. Valuesfromtoptobottomrepresentthe
brada Blanco, Yanamono, and Sucusari in Peru and Huan- Mantelcoefficient(r),variationexplainedbyafactorcontrolledby
chacaDosinBolivia.Incontrast,plotsatManauswerechar- theothertwofactors(%)andp-value.
acterizedbythehighspeciesrichnessofthegeneraEschweil-
era and Pouteria (de Oliveira and Mori, 2001): at least ten Distance Soils Climate
newspeciesofthefamilySapotaceaehavebeenrecentlyde-
Continental
scribedfromtheoriginalcollectionsinManaus(Pennington, Generalsoilclassification
2006). N=30plots 0.641 0.478 0.241
41.1% 22.9% 5.8%
These different floristic relationships could be an artefact
(0.001) (0.001) (0.001)
ofvaryinglevelsoftaxonomicresolution. Forexample,ifa
Cationsandsumofbases
speciesthatgrowsinbothsites,butiscorrectlyidentifiedin
N=18plotsa 0.554 0.181 0.266
oneplot(includedintheanalysis)butmisidentifiedoridenti-
30.6% 3.3% 7.1%
fiedtoanunknownspeciesinanotherplot(excludedfromthe (0.001) (0.060) (0.002)
analysis),thiswilltendtodecreasethespecies-levelfloristic
MgandCa
similarity between plots. However, both the sites at Jenaro N=18plotsa 0.581 0.360 0.261
Herrera and Manaus have a long history of botanical work 33.7% 12.9% 6.8%
(0.001) (0.001) (0.002)
andmanyofthesametaxonomistshavebeeninvolvedinthe
identification process. In general, even though these errors Availablesoildata
may have occurred, plots established by different research N=18plotsa 0.543 0.247 0.311
29.5% 6.1% 9.7%
teamsdosometimesgrouptogetherintheordinationbecause
(0.001) (0.016) (0.001)
ofthepresenceandhighabundanceofspecificspecies(e.g.
Regional–NWA
the cluster containing Jatun Sacha – D. Neill; Orosa, Bue-
Generalsoilclassification
navista–N.Pitman;Yanamono–R.Va´squez;Fig.2b): dif- N=30plots 0.259 0.336 0.266
ferent research teams working in high diversity forests can 6.7% 11.3% 7.1%
reportverysimilarfloristicresults. Alternatively,thediffer- (0.001) (0.001) (0.001)
entpatternsatthespeciesandgenuslevelmayreflectfunda- Cationsandsumofbases
mental differences in the underlying mechanisms that have N=14plotsb 0.240 0.389 0.068
5.7% 15.1% 0.5%
determinedthepatterns.Wesuspectthatthegenus-levelsim-
(0.026) (0.004) (0.263)
ilaritymaybeduetothesimilarlowfertilityofthetwosites,
whilethespecies-leveldifferencesmayreflectastrongerrole MgandCa
N=14plotsb 0.259 0.440 0.065
for dispersal limitation on the distribution of more recently
6.7% 19.4% 0.4%
evolvedspecies. (0.020) (0.001) (0.288)
Availablesoildata
4.2 Isregional-andcontinental-scalevariationinfloris- N=14plotsb 0.163 0.433 −0.052
ticcompositionsimilarinmagnitude? 2.7% 18.7% 0.3%
(0.080) (0.001) (0.633)
Inthisstudy, boththeaveragefloristicsimilarityanditsde- Regional–SWA
Generalsoilclassification
caywithdistancewerenotsignificantlydifferentamongre-
N=13plots 0.569 0.337 −0.074
gionsorbetweenregionalandcontinentalscales. Thedecay
32.4% 11.4% 0.5%
model using individuals that occur in two or more plots in (0.001) (0.002) (0.769)
this analysis decreased significantly with distance, and the
Cationsandsumofbases
distance decay was similar to that demonstrated by Con- N=10plotsa 0.056 0.223 0.165
dit et al. (2002) using all individuals of trees from Yasuni 0.3% 5.0% 2.7%
(0.357) (0.077) (0.139)
(Ecuador) and Manu (Peru), by Duque et al. (2009) using
all individuals of trees from the Colombian Amazonia and MgandCa
byTuomistoetal.(2003a)usingspecifictaxa(Pteridophytes N=10plotsa 0.201 0.371 0.273
4.1% 13.7% 7.5%
andMelastomataceae)inwesternAmazonia. Ourresultsare
(0.104) (0.017) (0.045)
consistent with a significant role for dispersal limitation, a
Availablesoildata
keyprocessthatdeterminesspeciesturnoverinspaceunder N=10plotsa 0.166 0.321 0.273
neutralmodels(Hubbell,2001;Conditetal.,2002;Duqueet 2.8% 10.3% 7.5%
al.,2009). (0.149) (0.028) (0.041)
However, the comparisons here also emphasise that very
dissimilarforestscanalsooccurnearby,andthatfloristically GeneralsoilclassificationfollowingMalhietal.(2004),aQuesada
etal.(2009b),bPitmanetal.(2008)
www.biogeosciences.net/6/2719/2009/ Biogeosciences,6,2719–2731,2009
2728 E.N.HonorioCoronadoetal.: TreecompositioncomparisonsofAmazonianforests
similar pairs of plots can be found much further away. In soils. UsingpteridophytesandthefamilyMelastomataceae,
these cases, floristic similarity cannot be predicted by the Tuomistoetal.(2003a)demonstratedthatenvironmentalfac-
distance decay model. For example, a wide range of floris- tors,especiallysoiltype,werealsoimportantforspeciesdis-
ticsimilaritywasfoundforplotswithin100kminNWAre- tribution and abundance patterns within terra firme forests
sembling the continental-scale patterns (Fig. 4). Such high inwesternAmazonia. Similarresultswereshownusingthe
betadiversityofthenorth-westernAmazoniaregionhasbeen same taxa in a one-hectare plot in Ecuador (Poulsen et al.,
notedpreviously(Tuomistoetal.,2003a;Ruokolainenetal., 2006), a 43-km long transect in northern Peru (Tuomisto et
2007). Conversely, our results also supported the observa- al., 2003b), palms in north-western Amazonia (Vormisto et
tionthatwithinAmazonia,somespeciesremaineddominant al.,2004;Normandetal.,2006),speciesoftreeswithdiam-
at large scales and therefore distant forests can be floristi- eter ≥10cm in specific areas in Colombia (Duivenvoorden,
cally rather similar. Pitman et al. (2001) showed this pat- 1995),anetworkof0.1haplotsinsouth-westernAmazonia
tern with Iriartea deltoidea, the most common tree species (Phillipsetal.,2003)andatabroaderscale,usinggeneraof
inbothYasuniinNWA(Ecuador)andManuinSWA(Peru; treesfromthewholeofAmazonia(terSteegeetal.,2006).
45–49 individuals/ha) and similar results were obtained by Cation concentrations, and particularly Mg++ and Ca++,
Mac´ıaandSvenning(2005)foraseriesof0.1haplotsinMa- may play a key role in the process of determining domi-
didiinBoliviaandYasuniinEcuador. Althoughtheidentity nant species. For plants, magnesium plays a critical role
of abundant species was often different in our set of forest in many physiological processes such as seed germination
plots, this pattern was very similar to our findings in terra andtheproductionofchlorophyllandfruitswhilecalciumis
firmeforestsatcontinentalandregionalscales. Forexample, used to regulate physiological processes that influence both
atacontinentalscale, speciessuchasEschweileracoriacea growthandresponsestoenvironmentalstresses(McLaugh-
(Lecythidaceae)wasoneofthemostabundantspeciesinJe- lin and Wimmer, 1999). Concentrations of soil cations are
naro Herrera and Manaus/Caxiuana located 1550–2500km also associated with various aspects of forest structure and
apart, and Iriartea deltoidea (Arecaceae) dominates forests function: cations can affect seedling growth rates of tropi-
in Jatun Sacha and Tambopata/Cusco Amazonico, 1600km caltrees(e.g.Denslowetal., 1987)andareassociatedwith
apart. At a regional scale, regardless of the geographi- floristic patterns and habitat preferences in Asian tropical
caldistancewithinnorth-westernAmazonia,distinctgroups forests(e.g.Baillieetal.,1987;Paolietal.,2006). However,
of species dominated nutrient-rich soils such as those of where concentrations of soil cations are low, the concentra-
Yanamono, Buenavista, Rio Orosa and Jatun Sacha (Otoba tions of aluminium may also be important: low concentra-
parvifolia, O. glycycarpa, Iryanthera juruensis, I. paraen- tionsofMgandCaareassociatedwithveryhighaluminium
sis), which were located 70–720km apart and the nutrient- contents in Amazonian soils (Quesada et al., 2009a). This
poor soils of Jenaro Herrera and Sucusari (Eschweilera co- might be particularly important for groups of trees, such as
riacea, Micropholis guyanensis, Minquartia guianensis and some genera of Melastomataceae, which are aluminium ac-
Osteophloeum platyspermum), which were located 200km cumulators(Jansensetal.,2002). Forexample,Tuomistoet
apart. al. (2003a) and Ruokolainen et al. (2007) using full inven-
tories of Melastomataceae found that Ca and Mg explained
4.3 Doenvironmentalfactorsandgeographicaldistance partofthefloristicpatternsinterrafirmeforestsinAmazo-
haveasimilarroleinexplainingfloristicdissimilar- nia while aluminium concentrations correlated with species
ityatregionalandcontinentalscalesinAmazonian diversity and richness (Tuomisto and Ruokolainen, 2005).
forests? Further studies of soil properties, such as phosphorus frac-
tionation,arerequiredinordertounderstandmorefullythe
The geographical distance and the environmental factors, roleofsoilsindeterminingfloristicpatterns.
suchassoilfertilityanddryseasonlength,explainedpartof Alargepercentageofthefloristicvariationthatwefound
the floristic dissimilarity at continental and regional scales. remained unexplained by the combination of factors exam-
While geographical distance appears to be more important ined in this paper (ALL: 30–60%; NWA: 75–80%; SWA:
at a continental scale, soil fertility was more important at a 55–95%). Therefore it is likely that there were additional,
regionalscalewithinwesternAmazonia(Table1). as yet unmeasured factors that may affect the floristic pat-
Our results are consistent with evidence that adaptation terns. These may include the effects of historical climate
to different edaphic conditions plays a key role in deter- and geological changes on species distribution and diversi-
mining spatial variation in floristic composition (Gentry, fication. For example, de Oliveira and Nelson (2001), us-
1988; Tuomisto et al., 1995). In this study, we showed ing the abundance of genera in different sites in the Brazil-
that the abundance of groups of species is related to the ian Amazon, showed that factors such as past disturbance
gradients in soil fertility at both regional and continental history may be important in determining floristic dissimi-
scales. SpeciesofLecythidaceaeandSapotaceaewerechar- larities. Pitman et al. (2008) discussed a range of possible
acteristicallyfoundonpoorersoilsandspeciesofArecaceae historical explanations for the disjunction in species com-
and Myristicaceae were more commonly found on richer position that occurs at the Peruvian and Ecuadorian border.
Biogeosciences,6,2719–2731,2009 www.biogeosciences.net/6/2719/2009/