Patterns of Floristic Differentiation among Atlantic Forests in Southeastern Brazil and the...

7/27/2019 Patterns of Floristic Differentiation among Atlantic Forests in Southeastern Brazil and the Influence of Climate

1/18

793

BIOTROPICA 32(4b): 793810 2000

Patterns of Floristic Differentiation among Atlantic Forests inSoutheastern Brazil and the Influence of Climate1

Ary T. Oliveira-Filho2,and Marco Aurelio L. Fontes

Departamento de Ciencias Florestais, Universidade Federal de Lavras, 37200-000, Lavras, MG, Brazil

ABSTRACT

The tree flora of southeastern Brazilian Atlantic forests was investigated according to two main aspects: (a) thevariationsin floristic composit ion of both rain and semi-deciduous forests were analyzed in terms of geographic and climaticvariables by performing multi variate analyses on 125 existing floristic checklists; and (b) the links of both rain and semi-deciduous forests to Amazonian forests and Cerrados(woody savanna) were assessed. All analyses were performed at thespecies, genus, and family levels. The information obtained for the 125 forest areaswas organized into an environmentaldatabase containing geographic and climatic records, and a floristic database containing binary presencerecordsfor 2532species, 520 genera, and 106 famil ies. Canonical correspondence analyses (CCA) were uti li zed to assess the relationshipbetween geographic and climati c variables, and tree flora composit ion. Venn diagrams and cluster analyses were used toassess the floristic links to Amazonian forests and Cerrados. The following patterns were detected at all t hree taxonomiclevels: (a) the differentiation between rain and semi-deciduous forests is floristically consistent and strongly correlatedwith rainfall regime, although transit ions may be abrupt to gradual; (b) a northsouth differentiation existsfor both rain

and semi-deciduousforests, probably caused by variationsin both temperature and rainfall regime; (c) The flora of semi-deciduous forests also changes with increasing distance from the ocean and the associated increasing rainfall seasonali ty;and (d) elevation and associated temperatures are strongly correlated with the internal differentiation of both rain andsemi-deciduous forests. To a considerable extent, the tree flora of semi-deciduous forests is a subset of the rain forestflora, probably extracting species that are able to cope with a longer dry season. There is greater floristic similarity at thespecies level between Atlanti c rain and semi-deciduous forests than between any of these and either Amazonian rainforests or Cerrados. Nevertheless, semi-deciduous forests and Cerrados show stronger links, particularly at the genericand familial levels. Therefore, there is lit tle floristic ground for viewing Atlantic rain forests as being closer to theirAmazonian counterparts t han to the adjacent semi-deciduous forests. The most appropriate view of rain and semi-deciduousforests in southeastern Brazil is that of a continuum in treespecies distribution. We suggest t hat t hedefinitionof Atlantic forests should be as comprehensive as that of Amazonian forests.

RESUMO

A flora arborea das florestas Atlanticas do sudeste do Brasil e investigada sob dois aspectos principais: (a) as variacoes emcomposicao florstica de florestas ombrofilas e semidecduas sao analisadas sob a otica de variaveis geograficase climaticaspor meio de analises multivariadas de 125 listagens florsticas existentesna literatura; e (b) os lacos das florestasombrofilas

e semidecduascom asflorestasAmazonicase cerradossao avaliados. Todasanalisesforam feitasnosnveis de especie, generoe famlia. A informacao obtida para as 125 areas de floresta foi organizada em um banco de dados ambientais, contendoregistrosgeograficoseclimaticos, eum banco dedadosflorsticos, contendo registrosbinariosdepresencapara 2532 especies,520 generos e 106 famlias. Analises de correspondencia canonica (ACC) foram uti lizadas para avaliar as relacoes entrevariaveis geograficas e climaticas e a composicao da flora arborea. Diagramas de Venn e analises de agrupamento foramutil izadospara avaliar oslacos florsticoscom florestas Amazonicas e cerrados. Os seguintes padroes foram detectados paratodos os nveis taxonomicos: (a) A diferenciacao entre florestas ombrofilas e semidecduas e floristicamente consistente efortemente correlacionada com o regime de chuvas, embora as transicoes possam ser abruptas a graduais; (b) Ha umadiferenciacao norte-sul tanto para florestasombrofilas como semidecduas, provavelmente causada por variacoesem temper-atura eregime dechuvas; (c) A flora dasflorestassemidecduastambem mudacom adistanciado oceano eo correspondenteaumento da duracao da estacao seca; e (d) Alt itude e suas correspondentes variacoes de temperatura sao fortemente corre-lacionadas com a diferenciacao interna tanto das florestas ombrofilas como semidecduas. A flora arborea das florestassemidecduase, em boamedida, um sub-conjunto da flora dasflorestasombrofilas, provavelmente extraindo especiescapazesde enfrentar uma estacao secamaisprolongada. Ha maissimilaridadeflorstica, no nvel deespecies, entre florestasAtlanticasombrofilas e semidecduas do que entre qualquer destas e as florestas Amazonicas, ou mesmo os cerrados. No entanto,florestassemidecduase cerradosmostram lacosflorsticosmais fortes, particularmente nosnveisdegeneroefamlia. Portanto,ha pouco fundamento florstico para sepensar nasflorestasombrofilasAtlanticascomomaisproximasde suascorrespondentesAmazonicasdo que de suas vizinhassemidecduas. A abordagem mais correta para florestasombrofilas e semidecduas do

sudeste brasileiro e de um contnuo de distribuicao de especies. Sugerimos, portanto, que a definicao de florestasAtlanticasdeve ser tao abrangente quanto a das florestasAmazonicas.

Key words: Atl anti c forests; Brazil ; climate; mult ivariate analysis; phytogeography; tree flora; t ropical rain forests, tropicalseasonal forests.

1 Received 13 August 1998; revision accepted 20 March 1999.2 Correponding author.


2/18

794 Oliveira-Filho and Fontes

THE BRAZILIAN ATLANTIC FORESTS ORIGINALLY COV-ERED an area of ca1.1 million km2, correspondingto 12 percent of the land surface of the country,

stretching for

3300 km along theeastern Brazil iancoast between the latitudes of 6 and 30S (SOSMata Atlantica & IN PE 1993). Atlantic forestsmake up the second largest tropical moist forest areaof South America, after the vast Amazonian domain.The two forests are separated by the so-called di-agonal of open formations, a corridor of seasonaland open formationsthat includesthe semiarid Caa-tingas of northeastern Brazil, the Cerrado (woodysavanna) of central Brazil, and the Chaco of Para-guayArgentinaBolivia (Prado & Gibbs1993). At-lantic forests are amongthe most threatened tropicalforests in the world, because their range largely co-incides with the most populated areas of Brazilwhere the settlement of European pioneers and Af-

rican slavesstarted four centuriesago. They now arereduced to onlyca5 percent of their original coverand most remnants are either small and disturbedfragments or wider areas sheltered on steep moun-tain slopes (Viana& Tabanez 1996, SOS Mata At-lantica1998).

As might be expected for a vast and diversifiedvegetation province, the Atlantic forests have beenlabeled with several names and their classificati onand geographic distribution are sti ll controversial(Leit ao Filho 1987, 1993; Camara 1990). The ac-ademic discord about the inland extent of Atlanticforests became a bitter polit ical dispute after theywere granted strict legal protection by the1988 Bra-zil ian Constitution and declared aBiosphereReserveby UNESCO. Fixing i nland l imits to the Atlanticforests, however, is no easy task, sincetheir transit ionto the hinterland open formations is very complexand more or less gradual. This transition can beclassified into threeregionsaccording to theadjacentopen formation.

A relatively abrupt transition to the semiaridCaati ngas occursin northeastern Brazil where anar-row strip (50 km) of coastal rain forests is bor-dered by an equally narrow inland belt of seasonalsemi-deciduous forests, which also occurs as hinter-land montane forest enclaves, the brejos (Andrade-Lima 1982). The transition between coastal rain for-ests and Cerrados in southeastern Brazil involves a

much larger extent of semi-deciduous forests thatbecomes increasingly wider toward the south andforms complex mosaics with Cerrado vegetation tothe west (Fig. 1). These semi-deciduous forests alsostretch southward along the Parana River basin intoeastern Paraguay and northeastern Argentina wherethey aretransitional to theChaco. In this subtropical

region, large extents of Araucariamixed forests sep-arate the coastal rain forests and the western semi-deciduous forests.

The many definit ions of Atlantic forests foundin theliteraturecan beclassified into two main viewsthat we call here sensu stricto(ss) and sensu lato (sl).Strictly speaking, Atlantic forests (ss) comprise onlythe coastal rain forests up to 300 km inland, whererainfall is locally boosted by oceanic winds and sea-side mountain ranges, particularly in thesouth. Thisis the most traditional and widespread definition ofAtlantic forests since Azevedo (1950) coined theterm Mata Atlanti ca. For instance, the I BGE clas-sification system for Brazilian vegetation (Velosoetal.1991) includes all Atlantic forests in thecategorydense rain forest, along with most Amazonian for-ests. Some authors have suggested that the physi-ognomic similarity between Amazonian and Atlantic

rain forests would show corresponding strong floris-tic links reinforced by numerous disjunct species(Rizzini 1963, Andrade-Lima 1964, Mori et al.1981).

For Atlantic forests (ss) supporters, the neigh-boring semi-deciduous forests are a distinct vegeta-tion formation, often called matas de planalto (pla-teau forests), due to their distribution on the hin-terland highlands. Cabrera and Wil link (1973)raised this differentiation to the biogeographic level,merging the semi-deciduous forests of the ParanaRiver basin with theAraucaria mixed forests to con-sti tute the Paranense Biogeographic Province. Thefloraof semi-deciduousforests is consideredaseithertransit ional between that of rain forests and Cerra-dos(e.g., Leitao Filho 1987) or part of a continuumof forest species distribution that includes centralBrazil ian gallery forests that eventually linksAtlanticto Amazonian forests (Oliveira-Filho & Ratter1995).

The broad view of Atlantic forests (sl) attachessemi-deciduous forests and Araucariamixed foreststo coastal rain forests, pushing the limits of the do-main up to 700 km inward from thecoast (Fernan-des & Bezerra 1990). This definit ion largely corre-sponds to the first attempt to classify Brazil ian veg-etation (von Martius 1840), which named all easternextra-Amazonian forests Dryads. To avoid misun-derstandingscaused by the long-term associati on be-

tween rain forests and Atlantic forests, authorsadopting the broad view usually added theadjectivemoist to encompassboth rain and semi-deciduousforests (e.g., Vianaet al. 1997). This definition ofAtlantic forests has recently become widespread inthe literature, although it wascaused lessby achang-ing scientific approach than by the current policies


3/18

Flora and Climate of Atlantic Forests 795

FIGURE 1. Map of southeastern Brazil showing the predominant vegetation formations, adapted from IBGE (1993),and Walt er climati c diagrams for selected locali ti es.

of both environmental conservati on and researchfunding. Sound quantitative assessments of floristi cvariation in Atlantic forests (sl), in association withunderlying environmental factors, are important to

understanding the geographic patterns of these for-ests for both scientific and conservation purposes.We sought patterns of floristic differentiation

among Atlantic forests (sl) that could be associatedwith geographic and climatic variables, and assessedthe floristic links among Atlantic forests (both rainand semi-deciduous), Amazonian forests, and Cer-

rados. We chose the arboreal component and fo-cused on southeastern Brazil (transition to Cerrados)dueto data availability, and becausewebelieved thatthe threescenarios of transit ion to open formations

required separate analyses. We addressed the follow-ing questions: (a) is the traditional treatment ofcoastal Atlantic rain forests as the only Atlantic for-ests consistent in termsof floristic composition; i.e.,how strongly differentiated are semi-deciduous andrain forests?; (b) to what extent is the tree speciescomposit ion of semi-deciduous forests transit ional


4/18


between thoseof rain forests and Cerrados?; (c) arecoastal Atlantic rain forests closer to Amazonian rainforests in terms of floristic composition than they

are to neighboring semi-deciduous forests?; (d) doboth rain and semi-deciduous forests change theirfloristic composition with the climatic variationsoc-curring within their range?; are rainfall regime andelevation as important for floristic patterns as theyare for tradit ional physiognomic classificati on sys-tems?; and (e) how are the abovequesti onsansweredat the species, genus and family levels?

The major patterns emerging from our analysesadd an important contribution to discussion of thefloristic composition of Atlantic forests. In addition,the results also contribute information on the pat-terns of plant species diversity and distribution inthe region, which isimportant for planning researchand conservation policies.

METHODS

FLORISTIC, GEOGRAPHI C, AND CLIMATIC DATA.Wese-lected from the literature a total of 144 floristicchecklists produced by surveys of the tree flora in125 areas of the Atlantic Forest (sl) in southeasternBrazil (contact authors for data). The geographicrange was defined between the eastern Brazili ancoast and 55 longitude, and between 1400 and2630S latitude (Fig. 2). Vegetation physiognomiesincluded coastal rain forests and semi-deciduousfor-ests, both coastal and hinterland. Forest areas ofedaphic climax were not included (e.g., those oncoastal sand dunes, swamps, and calcareous out-crops). Survey methods varied widely: most authorsused plot or plotless (PCQ) sampling; only six sur-veyswere based on intensivecollecting of plant spec-imens; and many combined both procedures. Forestareas were defined arbitrarily within a maximumrange of 20 km width and 150 m elevation, includ-ing sections of large continuous forest tracts (e.g.,Uba, Prd), assemblagesof forest fragments(e.g., Vic,Ter), and areasseparated by alt itude(e.g., P1-P2-P3,It t-Vma). A few areas with greater elevational range(e.g., Mci) required consulting the authorsabout thepredominant altitude of plant collecting activities.We obtained the following geographic information

for each area: latitudeand longitudeat the center ofthe area, median alt itude, and shortest distancefromthe ocean. We also extracted from the literature theannual and monthly meansfor the temperature andrainfall of each forest area or the nearest meteoro-logical stations. When the samesourceof thecheck-list did not provide the climatic records, they were

obtained from DNMet (1992). Most areas requiredinterpolation and/or standard correction for altitude(Thorntwaite 1948). We obtained the mean dura-

tion of the dry season for each area from thenum-ber of days of water shortage given by Walter dia-grams (Walter 1985). We also calculated a rainfalldistributi on ratio from the proportion between themean precipitation of the dry (JuneAugust) andrainy (DecemberFebruary) seasons. Summarizedinformation for the 125 areas is available from au-thor upon request.

CLASSIFICATION OF TH E FOREST AREAS.Traditionalclassification systems for Brazilian vegetation (e.g.,IBGE System; Veloso et al. 1991) usually classifyforest physiognomiesbased essentially on rainfall re-gime and temperature, the latter inferred from lati-tude and altitude. To assess this approach from the

floristic perspective, weclassified the 125 forest areasas either rain forests or semi-deciduous forests, anddivided them into four elevati onal classes, result ingin eight main forest formations(Fig. 2). Areasnorthof 2320S (tropical climates) or above 700 m(mountain climates) were classified as rain forests inwhich the dry season lasts for 30 daysor less; others(40160 days) were semi-deciduous forests, follow-ing Borhidi (1991). Areas situated south of 2320Sand below 700 m (non-montane subtropical cli-mates) were classified as semi-deciduous and rainforests with total annual rainfall between 1500 and2000 mm and 2000 and 3600 mm, respectively(moist and wet subtropical forests of Holdridge etal.[1971]). Theelevational categorieswere: lowland,300 m; submontane, 300700 m; lower montane,7001100 m; and upper montane, 1100 m. Theclassification criteriawere arbitrary and not intendedto represent any objective limit between forest for-mations; however, the geographic distributi on of theclassified areas coincided almost perfectly with thevegetation units of the map produced by IBGE(1993; compare Figs. 1 and 2).

PREPARATION AND REVISION OF T HE DATABASES.Weentered the information from the 125 checklistsonto spreadsheets using Microsoft Excel 97 in orderto producetwo databases. The first consisted of ba-sic information about each areaforest formation,

geographic coordinates, median alt itude, distancefrom the ocean, annual and monthly meansfor tem-perature and rainfall, mean durati on of dry season,rainfall distribution ratio, and literature source. Thesecond database was essentially a matrix of tree spe-cies presencein the 125 forest areas. Before reachingits final form, the information contained in this da-


5/18


FIGURE 2. Map of southeastern Brazil showing the location of the 125 areas of Atlantic Forest used in t hefloristicanalyses. The forest areas are identi fied by three-lett er codes (contact authors for details), and classified as eit her rainor semi-deciduous forest, and into four elevati onal categories: lowland, 300 m; submontane, 300700 m; lowermontane, 7001100 m; and upper montane, 1100 m.

tabaseunderwent adetailed revision to check all spe-cies names cited in the checklistsfor synonymy andgeographic distributi on. This task required consult-ing 337 monographs and revisions, as well as spe-cialists from the Universit ies of Sao Paulo, Campi-

nas, Rio de Janeiro, Belo Horizonte, and Braslia,and the Botanical Gardens of Rio de Janeiro, SaoPaulo, Kew, New York, and Missouri. When themonographs referred to herbarium specimens un-equivocally collected in any of the 125 areas, thespecieswasadded to the database. Thefinal databasecontained 2532 species, 520 genera, and 106 fami-

lies. To compare the tree flora of southeastern Bra-zilian Atlantic forests to that of Amazonian rain for-ests and Cerrados, we analyzed two addit ional da-tabases. The first was extracted from Oliveira-Filhoand Ratter (1994) and consisted of 22 areasof terra

firme (upland) Amazonian rain forests that con-tained 1530 species. The second consisted of 98 ar-easof Cerrado (Ratteret al.1996) that included 528species. We also derived databases containing thenumber of species per genus and family in each ofthe 125 forest areas, as well as in the Amazonianforests and Cerrado checklists.


6/18


MULT IVARIAT E ANALYSES.We used canonical corre-spondence analysis, CCA (ter Braak 1987, 1995)processed by the program PC-ORD 3.0 (McCune

& Mefford 1997) to investigate therelationshipsbe-tween tree flora and geographic and climatic (here-after geo-climati c) variables of the 125 forest areas.CCA seeks patterns of data structure from a matrixof floristic records correlated to a set of variablesprovided by an environmental matrix. We preparedflori stic matrices at three taxonomic levelsspecies,genus, and family. The species matrix consisted ofbinary data of species presence per forest area andincluded the 1002 species present on five or moreareas. The generic matrix was comprised of thenumber of species per genus in the 125 forest areasand contained all 520 genera. The familial matrixconsisted of the number of speciesper family in the125 forest areas and incorporated all 106 famil ies.

The patterns emerging from each of these floristicmatri ces were related by CCA to 13 variables con-tained in a geo-climatic matri x. These were the geo-graphic variables: (a) lati tude; (b) longitude; (c) me-dian altitude; (d) distance from the ocean and theclimatic variables (e) mean annual temperature andmean monthly temperaturesin (f ) July and (g) Jan-uary; (h) mean temperature range(JanuaryJuly); (i)mean annual rainfall and mean monthly rainfall ofthe (j) dry (JuneAugust) and (k) rainy (DecemberFebruary) seasons; (l) rainfall distribution ratio; and(m) mean duration of the dry season. After prelim-inary analyses, six variables were eliminated due toeither high redundancy (variance inflation factor20) or poor correlation (intra-set correlationswithaxes1, 2, or 3 0.4). Theeli minated variableswere:(a) longitude, highly redundant with distance fromthe ocean; (b) latitude, poorly correlated with theaxes; and themonthly temperatures(f, g) and seasonrainfall (j, k), all redundant with the respectivean-nual means. The Monte Carlo permutation test wasperformed to assess the significance of the correla-tions found.

CONDENSED FLORISTIC DATA.Because the CCAsdemonstrated that the forest classification systemadopted was highly consistent, we eventually con-densed the floristic information contained in theda-tabases by clustering the records within main forest

formations; however, as the CCAs also detected anadditional northsouth differentiation for rain andsemi-deciduous forests of lower altitudes and aneastwest pattern for semi-deciduous forests, the re-spective subcategories were spli t accordingly. A con-secutive upper level clustering operation, with bothrain and semi-deciduous forests, merged lowland

and submontane forests as low altitude forests,and lower and upper montane forests as simplyhigh altitude forests. Weused the condensed data

to perform a direct quantitative assessment of thefloristic linksbetween theforest formationsby plot-ting the number of shared and exclusive species inVenn diagrams. Wealso performed hierarchical clas-sifications (Kent & Coker 1992) of the condensedmatrices using the program PC-ORD 3.0. Clusteranalyses used Sorensen distancesfor speciespresencedata and euclidian distances for genera and families(number of species as abundance data); the linkagemethod was group average. A TWI NSPAN analysisof the condensed species matrix was performed toidentify indicator species for the forest formations.The genera and familieswith the highest number ofspecies in each forest formation were extracted fromthe matrices.

RESULTS

MULT IVARIAT E ANALYSES.CCA results for the threetaxonomic levels are summarized in Table 1. Eigen-values were relatively high for species (0.3), indi-cati ng considerable species turnover along the gra-dients summarized in axes 1 and 2. On the otherhand, they were increasingly small for genera andfamilies (0.2), indicating short gradients (i.e,there were large proportions of genera and familiesoccurri ng throughout the gradients, varying essen-tially i n abundance; ter Braak 1995). The cumula-ti vepercentagevariancesaccounted for by CCA axeswere also small, indicating that considerablenoiseremained unexplained. As might be expected forhigher and more widely distri buted taxa, the per-centage variance was higher for families (i.e., thenoise was lower). Small eigenvalues and low ex-plained variances are normal in vegetation data anddo not impair the significanceof taxaenvironmentrelations (ter Braak 1988). In fact, Pearsons corre-lationsfor taxaenvironment were consistently high.In addition, MonteCarlo permutation tests indicat-ed that taxa data and geo-climatic variablesweresig-nificantly correlated for the first threecanonical axesof all ordinations.

The first canonical axis in the speciesordinationwas correlated best with distance from the ocean,

followed by rainfall distribution ratio, altitude andduration of the dry season (Table 2). Also, durationof the dry season washighly correlated with distancefrom the ocean (posit ively) and rainfall distri butionratio (negatively). This indicatesthat thedata struc-ture summarized by the first axis primarily reflecteda geographic gradient based on penetration into the


7/18


TABLE 1. Summary of canoni cal correspondence analyses (CCAs) and M onte Carl o permutat ion test of the tree flora andgeo-cli matic vari ables for 125 areas of t he Atlanti c Forest performed for binary data of species presence, andfor genera and famil ies using the number of species as abundance data.

Axis 1 Axis 2 Axis 3

Eigenvalue (species)Cumulative percentage variance of speciespresence dataPearsons correlation for speciesenvironmentSignificance of speciesenvironment correlation (Monte Carlo test)

0.4056.40.9540.01

0.29010.9

0.9340.01

0.18013.7

0.8290.01

Eigenvalue (genera)Cumulative percentage variance of genera abundance dataPearsons correlation for generaenvironmentSignificance of generaenvironment correlation (Monte Carlo test)

0.2006.40.9330.01

0.13810.9

0.9170.01

0.09213.8

0.7840.05

Eigenvalue (families)Cumulative percentage variance of families abundance dataPearsons correlation for familiesenvironmentSignificance of familiesenvironment correlati on (Monte Carlo test)

0.08110.5

0.8860.01

0.04616.4

0.8220.01

0.02820.0

0.7410.01

continents interior which corresponded to a majorclimatic gradient (i.e.,an increasingly longer dry sea-son). Altitude, another important geographic gradi-ent in the first canonical axis, was highly and neg-atively correlated with both annual temperature andrainfall seasonality. The second canonical axis hadstronger correlations with annual temperature andtemperature range. The CCA for genera showedcorrelation patterns very similar to those found forspecies, although inverting the relative weight ofsome variables on the first two axes. Unlike the for-mer ordinations, the CCA for families related thedistance from the ocean and rainfall variablesto axis2, and altitude and temperature variables to axis 1.

The relationship between the geo-climatic vari-ables and the species composit ion of the 125 forestareas is illustrated by the CCA biplot in Figure 3.The applicati on of the forest classification categoriesto the ordinated areas helped the interpretation ofthe biplot. T he dichotomy between rain and semi-deciduous forests appears as a diagonal perpendicu-lar to the arrow of distance from theocean. Towardthe right side of the diagram, the pattern summa-rized for rain forests (closed symbols) involves de-creasing alt itude and temperature range, and increas-ing annual temperature and rainfall distribution uni-formity. Lower and upper montane rain forests areclearly discriminated at the top of the diagram. The

areas of low altitude rain forests are found in twodistinct groups: the northern group (bottom-right)composed of the areas in Bahia and Esprito Santo,and the southern group (top right), which includesthe areas in Rio de Janeiro, Sao Paulo, and Parana.The rainiest areas of coastal Sao Paulo and Parana,all with annual rainfall 2000 mm, are moreclear-

ly discriminated in axis 3 (not shown), which i sstrongly correlated with annual rainfall.

Three main patterns are summarized for semi-deciduousforests (open symbols) on theCCA biplot(Fig. 3). The areas of coastal lowland semi-decidu-ous forests of northern Rio de Janeiro (Cfr, Cgo,Imb) appear on the bottom right, not far from thenorthern lowland rain forests, and are highly corre-lated with elevated temperatures. They are followedtoward the center left by the submontanesemi-de-ciduous forests of the Rio Doce basin in easternMi nas Gerais (Prd, Crt, Cng, Vic, Pnv, Imd, Sbr),all of northern distri bution. Most submontanesemi-deciduousforests of southern distribution (Sao Pau-

lo and Parana) appear on the center-left of the dia-gram and are followed toward the left by lower andupper montanesemi-deciduousforests. The stronglyseasonal areas of western distribution (Goias, Distri-to Federal, northern Sao Paulo, and western MinasGerais) are accommodated on the bottom left of thediagram and also show an altitudinal gradient.

Ashappened with the CCA for species, the sameforest groupsare distinguished in thebiplotsyieldedby the CCAs for genera and families (Fig. 4). Thismade clear that the same geographic and climaticvariables underlie the floristic patterns at all threetaxonomic levels. The ordinati on of species, genera,and families by CCA i s not shown in Figures3 and4 becausethe biplotswould becrowded by countlesstaxa dots.

ANALYSES OF CONDENSED FLORISTIC IN FORMATI ON.As they were extracted from floristic checklists forparticular forest areas, the condensed informationmust be regarded as a means of quanti tatively as-


8/18

TABLE 2. Canonical corr espondence analyses (CCAs) of the tr ee flora and geo-cli mati c variablesfor 125 areas of the Atl anti c Forest performed fofor genera and famil ies using the number of species asabundance data: i ntra-set and i nter-set correlati onsfor the first two ordinati on for the geo-cli matic variables supplied. Correlati onswit h absolute values0.5 are enhanced in bold.

Geo-climatic variables

Intra-setcorrelations

Axis 1 Axis 2

Inter-setcorrelations

Axis 1 Axis 2

Geo-climati c vari

Altit.Dist.ocean

Ann.temp.

Tera

CCA for species

AltitudeDi stance from the oceanAnnual temperatureTemperature rangeAnnual rainfallRainfall distribution ratioDuration of dry season

0.760.82

0.420.060.280.81

0.73

0.310.500.750.750.43

0.290.41

0.720.78

0.400.060.270.78

0.70

0.290.470.700.710.40

0.270.38

0.41

0.770.08

0.170.62

0.35

0.06

0.350.270.470.72

0.490.11

0.400.07

CCA for genera:

AltitudeDi stance from the oceanAnnual temperatureTemperature range

Annual rainfallRainfall distribution ratioDuration of dry season

0.880.530.76

0.34

0.090.830.46

0.160.740.57

0.44

0.560.180.68

0.820.500.71

0.32

0.080.770.43

0.150.680.52

0.40

0.510.160.62

0.42

0.800.13

0.150.630.35

0.02

0.31

0.270.480.73

0.51

0.110.440.04

CCA for families:

AltitudeDi stance from the oceanAnnual temperatureTemperature rangeAnnual rainfallRainfall distribution ratioDuration of dry season

0.820.100.95

0.410.26

0.390.12

0.490.78

0.170.030.580.71

0.78

0.720.090.84

0.370.23

03.50.11

0.400.66

0.140.03

0.480.59

0.64

0.42

0.800.13

0.150.63

0.35

0.02

0.310.270.480.73

0.510.11

0.440.04


9/18


FIGURE 3. Biplot yielded by canonical correspondence analysis showing t he ordination of 125 areas of AtlanticForest on the first two axes, based on the presence of 1002 tree species, and their correlation with geo-climaticvariables, shown as arrows (scaled 2.5 for clarity). Forest areas are identified by their three-lett er codes (Fig. 2) andforest f ormati ons are indicated by symbols.

sessing the floristic links between the main forestformations and not as actual figures for number ofspecies, either total or in common.

The Venn diagrams in Figure 5 give a direct

assessment of the composition and patterns for thetree flora in some main Atlantic Forest formations.It is evident from the first diagram that the flora ofrain forests is much richer in tree species than thatof semi-deciduous forests. Rain forests had 31 per-cent more species than semi-deciduous forests, al-though they were represented by a smaller number

of surveyed areas. It is also clear that rain and semi-deciduous forests shared a high proportion of treespecies: 50 and 66 percent, respectively. Further-more, the proportion is higher on the semi-decidu-

ousside, suggesting that their arboreal flora is, to agreat extent, afraction of the much richer rain forestflora.

The second and third Venn diagrams in Figure5 show the distribution of the number of speciesontwo altitudinal classes of both rain and semi-decid-uousforests, the latter also showing the relationship


10/18


FIGURE 4. Biplots yielded by canonical correspon-dence analyses showing the ordinati on of 125 areas ofAtlantic Forest on thefirst two axesbased on the numberof tree species per genus (A) and family (B), and theircorrelation with geo-climati c variables, shown as arrows(scaled 2.5 for clarity). Forest areas are identi fied bytheir three-lett er codes (Fig. 2) and forest formati onsareindicated by symbols (see Fig. 3).

FIGURE 5. Venn diagramsshowing the number of treespeciesshared by 125 areasof Atlantic forest, merged intomain Atlanti c Forest formations, 22 areas of AmazonianRain Forests, and 98 areas of Cerrado. N number ofchecklists; S number of species.

to the western forests. Although the total numberof species decreased with alt itude in both cases, thenumber of surveyed areas also decreased, and thismay have accentuated the decline. The proportionof species shared by the two altitudinal classes wassimilar: 37 and 35 percent for rain and semi-decid-uous forests, respectively. Western semi-deciduousforests shared 63 and 59 percent of specieswith lowand high alt itude semi-deciduous forests.

The fourth and fifth Venn diagramsstrongly re-inforcethe differentiation between thenorthern (Ba-

hia, Esprito Santo, and eastern Minas Gerais) andsouthern (Rio de Janeiro, Sao Paulo and Parana)sections of both low altitude rain and semi-decidu-ous forests. T hey shared only 26 and 33 percent ofthe species, while northern rain and semi-deciduousforests shared 29 percent (sixth diagram) and south-ern rain and semi-deciduous forests shared 24 per-

cent (seventh diagram) of the species. Therefore, thedif ferenti ati on between northern and southerngroupsof the same forest formation were apparentlyof a similar magnitudein termsof thedifferentiationbetween different formations within the same geo-graphic region.

Cerrados shared a much larger proportion(55%) of their flora with Atlantic forests than didAmazonian forests (20%; eighth and ninth dia-grams). Also, the Cerrado flora was much moreclosely related to Atlantic semi-deciduous foreststhan to the Atlantic rain forests, while the species

shared with Amazonian forests were distributedmore evenly.The classificati on dendrograms (Fig. 6) show

somecommon patterns for thethreetaxonomic lev-els. Most rain and semi-deciduous Atlantic foreststend to cluster together. The flora of both Cerradoand Amazonian forests were highly differentiated


11/18


FIGURE 6. Dendrograms from group averaging ofSrensensfloristic similarity for speciesand squared euclid-ian distances for genera and families of the arboreal flora

of 125 areas of Atlantic Forests, merged into nine mainforest formations, 23 areas of Amazonian Forests, and 98areas of Cerrado. RF rain forests; SF Semi-deciduousForests. Distance on the dendrogram scale is Wisharts ob-

jective function, not a raw dissimilarity measure.

from the Atlantic forests at the species and genericlevels. At the familial level, however, Cerradomergedat a low level with montanesemi-deciduousforests,while Amazonian forests remained drasti cally sepa-rated.

The most important genera and familiesof eachmain forest formation are given i n Tables 3 and 4,and a selection of indicator speciesgiven by TWIN-

SPAN for some formations is provided in Table 5.The frequency of these species or the relative abun-dance of somegenera and familiesin a survey mayhelp classify the forest area and allow inferencesabout the local climate. Someinteresting trendscanbeobserved with increasing altitudein thefivemainAtlantic Forest formations (Table 3). The relative

importance decreased for some genera such asMar-li erea, Pouteri a, Tr ichili a, Erythroxylum, Swartzia, Fi-cus,andM achaeri um, and increased for others such

asM iconia, M olli nedia, Nectandra, Myrsine, Ti bouch-ina, Solanum, I lex, and Gomidesia. The same hap-pened with somefamilies: therelativeimportanceofSapotaceae, Chrysobalanaceae, Rutaceae, and Mor-aceae decreased, while that of Melastomataceae,Compositae, Solanaceae, and Myrsinaceae increasedwith increasing altitude. Western semi-deciduousforests apparently combineelementsof both low andhigh altitude.

The rank of genera and familieswith the highestnumber of species given in Table 4 reinforces thepatterns shown in the dendrograms. A few generaand famil iesshowed marked differencesin their rankposit ion between Atlanti c rain and semi-deciduousforests (e.g., Pouteria, Gomidesia, Ilex, M achaeri um,

Sapotaceae, and Chrysobalanaceae). Many generaranked high in both Atlantic forests were less im-portant either in Cerrado (e.g., Ocotea, Inga, Fi cus,andTrichili a) or Amazonian rain forests (e.g., Myr-cia, Inga, M achaeri um,andSolanum). Likewise, im-portant families in Atlantic forests were ranked in alower position in Cerrados(e.g., Lauraceae and Eu-phorbiaceae) and Amazonian rain forests (e.g.,M e-lastomataceae and Myrtaceae). In addit ion, Cerradohad particularly important genera (e.g., Byrsonima,Ki elmeyera, Vochysia,and Alibertia) and families(e.g.,Malpiphiaceae and Vochysiaceae). T he same is ob-served in Amazonian forests for some genera (e.g.,Li cania, Proti um, Eschweilera, Virola, andMouriri)and families (e.g., Sapotaceae, Chrysobalanaceae,Burseraceae, and Lecythidaceae). An important linkbetween Amazonian and Atlantic rain forests, how-ever, li es in the numerous species of SapotaceaeandChrysobalanaceae, and in the higher diversity of Le-cythidaceae, compared to semi-deciduous forests.

DISCUSSION

Most floristic patterns and their correlation withgeo-climati c variables, particularly rainfall seasonali tyand temperature, were consistent at all three taxo-nomic levels, suggesting that these factors have hada long time influence on the evolution and specia-tion of treetaxa in southeastern Brazil . Thiswasnot

a surpri se because rainfall and temperature are thechief factors determining the distribution of theworlds vegetation formations, and the history ofvegetation and climate in southeastern Brazil duringthe Quaternary shows dramatic shifts in both tem-perature and rainfall regime (Ledru 1993, Ledruetal.1998).


12/18

TABLE 3. Genera and famil ies wi th the highest number of species (S) on the tree flora for 102 areas of the Atl anti c Forest classified i nto five maareas.

Rain forests (N 48)

Low alt itude(N 27)

S1475

High altitude(N 21)

S1280

Semi-deciduous forests (N 77

Low alti tude(N 36)

S1124

High altitude(N 16) 7

EugeniaMyrciaOcoteaMiconiaPouteriaM arli ereaErythroxylumIngaLicaniaFicus

70373635262221201919

EugeniaMiconiaOcoteaMyrciaMollinediaIngaSolanumGomi desiaTibouchinaPsychotr ia

73484032272421171616

Eugeni aOcoteaMyrciaMiconiaM achaeri umFicusIngaCaseari aM aytenusErythroxylum

44322625211915141212

MiconiaMyrciaOcoteaEugeni aIngaIlexNectandraTabebuiaTibouchinaTrichilia

222

TibouchinaPsychotr iaTabebui aTrichiliaGomi desiaM achaeri umCaseari aGuatteria

CordiaMollinediaNectandra

1817161616151413

131312

M aytenusM arli ereaMyrsineM yrceugeni aPouteriaCaseari aM achaeri umTrichilia

RudgeaSympl ocosFicus

1515131313121212

121111

TrichiliaMollinediaSolanumGuatteriaTabebuiaSwartziaPouteriaCordia

RudgeaNectandraCampomanesia

12121211111010

9

999

SolanumM achaeri umFicusPsychotr iaGuatteriaAspidospermaM aytenusErythroxylum

Caseari aSennaMollinedia

MyrtaceaeFabaceaeRubiaceaeLauraceaeMelastomataceaeCaesalpiniaceaeMimosaceaeSapot aceaeChrysobalanaceaeEuphorbiaceae

203918478765849494342

MyrtaceaeMelastomataceaeLauraceaeRubi aceaeFabaceaeMimosaceaeEuphorbiaceaeCaesalpiniaceaeMonimiaceaeSolanaceae

203828176584935353531

MyrtaceaeFabaceaeRubiaceaeLauraceaeEuphorbiaceaeMimosaceaeCaesalpiniaceaeMoraceaeMelastomataceaeAnnonaceae

129836661474640353430

MyrtaceaeLauraceaeMelastomataceaeFabaceaeRubiaceaeMimosaceaeEuphorbiaceaeCaesalpiniaceaeAsteraceaeAnnonaceae

843332222


13/18


TABLE3.

Continued.

Rainforests(N

48)

Lowaltitude

(N

27)

S1475

Highaltitude

(N

21)

S1280

Semi-deciduousforests(N

77)

Lowaltitude

(N

36)

S1124

Highaltitude

(N

16)

S709

Western

(N

25)

S699

Moraceae

Annonaceae

Clusiaceae

Meliaceae

Apocynaceae

4138272625

Asteraceae

Sapotaceae

Annonaceae

Clusiaceae

Sapindaceae

3

0

2

9

2

5

2

3

2

3

Rutaceae

Flacourtiaceae

Sapotaceae

Sapindaceae

Bignoniaceae

2824232119

B

ignoniaceae

C

lusiaceae

M

eliaceae

R

utaceae

S

apindaceae

1414141414

Apocynaceae

Flacourtiaceae

Vochysiaceae

Meliaceae

Arecaceae

1515151413

Sapindaceae

Bignoniaceae

Rutaceae

Flacourtiaceae

Erythroxylaceae

Arecaceae

Monimiaceae

25242322212017

Flacourtiaceae

Meliaceae

Myrsinaceae

Moraceae

Vochysiaceae

Bignoniaceae

Chrysobalanaceae

1

9

1

9

1

9

1

8

1

8

1

7

1

7

Meliaceae

Solanaceae

Clusiaceae

Monimiaceae

Nyctaginaceae

Apocynaceae

Myrsinaceae

18181717171614

F

lacourtiaceae

M

oraceae

S

olanaceae

M

onimiaceae

V

erbenaceae

A

quifoliaceae

V

ochysiaceae

13131311111010

Bignoniaceae

Asteraceae

Chrysobalanaceae

Clusiaceae

Rutaceae

Sapindaceae

Anacardiaceae

13121211111110

The tree species composit ion of both rain andsemi-deciduousforests also washighly influenced byalt itude and associated temperaturesat all threetax-

onomic levels, a well-known fact for mountain veg-etation worldwide (Hugget 1995). Some floristicpatterns found with increasing altitudealso coincid-ed with those cited by Gentry (1995) for Andeanand Central American forests (i.e., the increasingcontribution of Asteraceae, Melastomataceae, andSolanaceae to the tree flora and the decrease of Le-guminosae, with the exception of Inga species).Many genera strongly correlated with high altitudesin southeastern Brazil such as Drymis, Hedyosmum,Weinmannia, Clethra, Podocarpus, M eli osma, M eri -ania, Ilex, Clusia, M yrsine, M iconia, Prunus, Roupala,and Trichipteris, are also considered diagnosti c ofNeotropical cloud forests (Webster 1995). At thespecies level, many listed as high alti tude indicators

such as Drymis brasili ensis, Weinmannia di scolor, W.paulli ni foli a, Podocarpus lambertii , P. sellowii , andHedyosmum brasiliense, are already known to followan upper montane distributi on pattern along Bra-zilian mountain ranges (Giulietti & Pirani 1988).

The importance of altitude in the floristic dif-ferentiati on of semi-deciduous forests has been doc-umented for the stateof Sao Paulo (Saliset al.1995,Torres et al. 1997), and southeastern Brazil in gen-eral (Oliveira-Filho et al. 1994). Occasional frostshavebeen mentioned by Oliveira-Filho et al.(1994)as an important factor limiting species distributiontoward both higher elevati ons and latitudes. Resis-tanceto frosts is akey factor determining treespeciesdistribution in Australian forests(Read & Hi ll 1989,Read & Hope 1989). T he influence of altitudeonclimate, however, is far morecomplex than creatingtemperature gradients and frost events. Rising ele-vation also decreases atmospheri c pressure, increasessolar radiation, accelerates air masses, promoteshigher cloudiness, and boosts rainfall (Jones 1992).

The well established differentiation between rainand semi-deciduous Atlantic forests based on phys-iognomy was floristically consistent at all threetax-onomic levels, and to some extent correlates withthe coastlandhinterland dichotomy. This has beendemonstrated for the state of Sao Paulo (Torres etal.1997) where there is a strong floristic separationbetween coastal hygrophyllous forests (annual

rainfall

2000 mm and no dry season) and hinter-land mesophyllousforests (annual rainfall ca1400mm and a dry season). Sao Paulo and Parana arewhere this dichotomy is most pronounced in south-eastern Brazil due to the relati vely abrupt vegeta-tional transition at the Serra do Mar, the coastalmountain range that helps create two sharply dis-


14/18


TABLE 4. Genera and fami li es wit h the hi ghest number of species (S) on t he tree flora of Atlant ic rai n forests, semi-deciduous forests, Amazonian rai n forests, and Cerrados. N number of areas.

Atlanticrain forests(N 48)

S2012

Atlanticsemi-deciduous

forests(N 77)

S1533

Amazonianrain forests(N 22)

S1530

Cerrado(N 98)

S528

Eugeni aMiconiaMyrciaOcoteaMollinediaIngaTibouchinaErythroxylumM arli ereaPouteriaGomidesiaSolanumPsychotr iaLicania

FicusM achaeri umTrichiliaM aytenusRudgeaTabebuiaCaseari a

10859574930282827272725252321

21201918181716

Eugeni aMiconiaMyrciaOcoteaM achaeri umFicusIlexErythroxylumIngaCaseari aSolanumGuatteriaM aytenusSenna

TibouchinaMollinediaAspidospermaNectandraTrichiliaCampomanesiaPouteria

5946463724222019191515141414

14141313131212

PouteriaLicaniaIngaProtiumMiconiaSwartziaOcoteaEschweileraSloaneaCaseari aEugeni aFicusVirolaDuguetia

MouririCordiaPouroumaHi rtellaAnibaM icropholi sIryanthera

4438373526252419181818171715

15141414141413

MyrciaMiconiaByrsonimaEugeni aAspidospermaPsidiumKielmeyeraVochysiaCordiaBauhiniaM achaeri umAlibertiaSymplocosLicania

ErythroxylumCaseari aVernoniaOcoteaQualeaM yrsineSolanum

2420181211

988777776

6655555

MyrtaceaeMelastomataceaeRubiaceaeFabaceaeLauraceaeCaesalpiniaceaeMimosaceaeEuphorbiaceaeSapotaceaeChrysobalanaceaeAnnonaceaeMoraceaeMonimiaceaeClusiaceaeSolanaceaeAsteraceaeSapindaceaeRutaceaeMeliaceaeBignoniaceaeErythroxylaceaeFlacourtiaceae

308117116111106

6867535049484340393837333129282727

MyrtaceaeFabaceaeRubiaceaeLauraceaeMelastomataceaeMimosaceaeCaesalpiniaceaeEuphorbiaceaeMoraceaeAnnonaceaeRutaceaeAsteraceaeSapindaceaeFlacourtiaceaeSapotaceaeClusiaceaeSolanaceaeApocynaceaeBignoniaceaeNyctaginaceaeMonimiaceaeMyrsinaceae

187100

8476675755543837302929262625242323222121

FabaceaeSapotaceaeLauraceaeMimosaceaeAnnonaceaeCaesalpiniaceaeChrysobalanaceaeMoraceaeRubiaceaeBurseraceaeEuphorbiaceaeLecythidaceaeApocynaceaeMyrtaceaeArecaceaeMelastomataceaeMyristicaceaeClusiaceaeFlacourtiaceaeMeliaceaeSapindaceaeBombacaceae

93887976747368625251494746403635333232302927

MyrtaceaeRubiaceaeMelastomataceaeCaesalpiniaceaeAsteraceaeFabaceaeMalpighiaceaeVochysiaceaeApocynaceaeAnnonaceaeMimosaceaeClusiaceaeEuphorbiaceaeAnacardiaceaeArecaceaeChrysobalanaceaeVerbenaceaeBignoniaceaeLauraceaeBoraginaceaeCombretaceaeMyrsinaceae

51302725242422181716161312101010

988777

tinct climates. Theseaward side of the Serra do Marhas the highest mean annual rainfall (up to 3600

mm) of the entire Atlantic Forest range, while theinland side has typical seasonal climateswith annualrainfall between 1300 and 1600 mm. This reinforc-esthe inaccurate imageof two sharply distinct forestformations.

The transit ion from rain to semi-deciduousfor-ests may be gradual and complex, and not necessar-

ily linked to the coastalhinterland climatic gradi-ents. Coastal climates vary dramatically in south-

eastern Brazil . Annual rainfall declines from south-ern Sao Paulo toward the north of Rio de Janeirostate where semi-deciduous forests reach the coastnear Campos dosGoitacazes and give rise to a gapin rain forest distributi on (Fig. 1). T he drier coastalclimate of this region is caused by the cold oceanicupwelling of Cabo Frio (Araujo 1997). Coastal rain


15/18


TABLE 5. Selecti on of tree species associated wi th major groups of Atl anti c Forest formati ons based on TWI N SPANclassificati on of 125 forest areas and 1002 species.

Northern lowaltituderain forests: Bactr is setosa, Bathysa nicholsoni i, Brosimum guianensis, Byrsonima seri cea, Cam-pomanesia guavir oba, Carpotroche brasi li ensis, Car yocar edule, Caseari a ulmifoli a, Cedrela odorata, Chr ysophyll um lucen-ti folium, Cupania emarginata, Dalbergia nigra, Dipl otropis i ncexis, Ecclinusa ramiflora, Esenbeckia leiocarpa, Eugeni abrasi li ensis, E. moraviana, E. sti ctosepala, E. sulcata, E. umbell iflora, E. verr ucosa, Fi cus organensis, Gall esia i ntegrifoli a,Geissospermum laeve, Heli ocarpus ameri canus, H imatanthus lancifoli us, Horti a arborea, Hydrogaster tr inervi s, Hymenolob-ium janeirense, Inga edul is, I . flagell iformi s, I . stri ata, Ixora gardneri ana, Jacaranda puberula, Jacarati a heptaphylla, Joan-nesia pri nceps, L ecythi s pisoni s, L onchocarpus campestr is, M achaeri um scleroxylum, M arl ierea obscura, M . sylvati ca, M ay-tenus robusta, M elanoxylon brauna, M etrodorea nigra, M icropholi s gardneri anum, M ourir i chamissoana, M yrocarpusfrondosus, Ocotea elegans, O. indecora, O. puberula, O. veluti na, Paratecoma peroba, Platymiscium floribundum, Pogon-ophora schomburgkiana, Pourouma guianensis, Proti um warmingianum, Pseudobombax grandiflorum, Pseudopiptadeni acontorta, P. leptostachya, Pterocarpus rohri i , Pterygota brasi li ensis, Rinorea bahiensis, Rusti a formosa, Schizolobium parahyba,Simarouba amara, Sloanea guianensis, Solanum swartzi anum, Sparatt osperma leucanthum, Str yphnodendron pulcherr imum,Swartzia acuti folia, S. flaemingi i , Tabebuia roseo-alba, Tetrastyli dium grandifolium, Thyrsodium spruceanum, Tri chil iacasarett i , T. elegans, T. lepidota, Vataireopsis araroba, Vi rola gardneri , V. officinali s, V. olei fera, Vi smia baccifera, Vi texmegapotamica, Zanthoxylum monogynum, Zollernia ilicifolia.

Southernlowaltituderainforests:All ophyluspeti olul atus, Alseisfloribunda, Aniba firmula, Astrocaryum aculeati ssimum,Attalea dubia, Bali zia pedicell ari s, Bathysa meridi onali s, Brosimum glazioui , Calyptranthes grandifolia, Chrysophyllumflexuosum, Ci tr onella megaphyll a, Coussapoa microcarpa, Cryptocarya moschata, Cupania racemosa, Ecclinusa rami flora,

Eri otheca pentaphyll a, Eugeni a cerasiflora, Euterpe edul is, Fi cus organensis, Gal ipea multi flora, Gomi desia anacardi aefolia,G. spectabi li s, Guatteria austral is, Heisteri a si lvi anii , Hymenolobium janeirense, Il ex dumosa, I nga capitata, I . eduli s,Jacaranda puberul a, Li caria armeni aca, Lonchocarpus muehlbergianus, M aloueti a arborea, M arli erea suaveolens, M . to-mentosa, M icropholi s crassipedicell ata, M olli nedia schotti ana, M . uleana, M yrcia pubipetala, Nectandra membranacea,Neomytranthes glomerata, Nephelea sternbergi i, Ocotea brachybotra, O. dispersa, O. divari cata, O. puberul a, Pari nariexcelsa, Pil ocarpus pauciflorus, Posoqueri a acuti foli a, Pouteri a caimito, P. venosa, Psychotri a cart hagenensis, Pterocarpusrohrr i, Qui ina glazi ovii , Ruprechtia laxi flora, Sclerolobium denudatum, Sebasti ani a brasili ensis, Stylogyne ambigua, Sweeti afruti cosa, Tabebuia heptaphyll a, Talauma ovata, Tetrastyli dium grandif olium, Tetrorchidi um rubr ivenium, Virola gardneri ,V. oleifera.

High altituderain andsemi-deciduousforests: Bathysa meri dionali s, Byrsonima laxi flora, Calyptranthes clusiaefoli a,Casearia obliqua, Cecropia glazioui, Cinnamomum glaziovii, Clethra scabra, Clusia criuva, Connarusregnellii, Cryptocaryasali gna, Daphnopsis fasciculata, Dri mys brasil iensis, Eremanthus incanus, Eugeni a blastantha, Euplassa incana, Ficus lu-schnatiana, F. mexiae, Geonoma schott iana, Gomidesia eri ocalyx, G. spectabil is, Gordonia fruti cosa, Guatt eri a ni grescens,Hedyosmum brasi li ensis, H eisteri a si lvi anii , Hieronyma ferr uginea, L eucochlorum incuriale, M aytenusglazioviana, M . sal-ici folia, M eli osma sell owii , M iconia brunnea, M . chartacea, M . cinnamomifoli a, M . peperi carpa, M. theaezans, Molli nediaargyrogyna, M yrcia l aruotteana, M yrsine lancifoli a, N ectandra grandiflora, N . lanceolata, N . ni ti dula, N. puberul a, Ocoteabrachybotra, O. si lvestr is, Ouratea semiserrata, Picramnia glazioviana, Pimenta pseudocaryophyll us, Protium wi dgreni i,Psychotri a suterella, Qual ea jundi ahy, Qui ina glazi ovii , Schefflera angusti ssima, S. calva, Salacia ell ipti ca, Siphoneugenawi dgreni ana, Solanum bul latum, Symplocoscelastr inea, Tabebuia chrysotri cha, Ti bouchi na stenocarpa, Tr embleyaparvi flora,Tri chil ia emarginata, Tri chipteri s corcovadensis, Vani ll osmopsiserythropappa, Vismia brasil iensis, Weinmania pauliniaefolia,Xylosma ciliatifolium.

Eastern lowaltitudesemi-deciduousforests: Acacia glomerosa, A. polyphylla, Albizia niopoides, Aloysia virgata,Apuleia leiocarpa, A spidosperma polyneuron, Astr onium fraxi ni foli um, Balfourodendron r i edeli anum, Bastardi opsi sdensi flora, Chrysophyll um gononocarpum, Copaifera l angsdorffii, Cordi a tr i chotoma, Cu pani a oblongifol i a, Du gueti alanceolata, Enterolobium contortisiliquum, Esenbeckia leiocarpa, Eugenia involucrata, E. moraviana, E. sulcata, Ficusguarani ti ca, F. i nsi pida, F. obtusi foli a, Guarea gui donea, G. kunt hi ana, Hi matanthus lancei foli us, H olocalyx balansae,Lonchocarpus cult ratus, M achaeri um paraguari ensi s, M aytenus aquifoli a, M etr odorea sti pular i s, M yrci a mu lt i flora,M . r ostr ata, M . t oment osa, N ectandra cissi flora, O cotea puberu la, Ourat ea castaneifoli a, Parapiptadenia ri gida, Pa-tagonul a ameri cana, Pi cramnia sell owii , Pi sonia ambi gua, Procki a crucis, Pru nus sell owii , Pterogynen i tens, Si parunagui anensi s, Solanum granul oso-leprosum, Swart zia apetal a, Sweeti a fr uti cosa, Tabernaemontana hystr i x, Tri chili acasarett i , T. cati gua, T. clausseni i , T. elegans, T. hi rt a, Xylopia seri cea, Z anthoxylum cari baeum, Z . r i edeli anum,Zeyheri a tubercul osa.

Western montaneandsubmontanesemi-deciduousforests:Acosmi um dasycarpum, Acrocomia acul eata, Aegi -phila lhotzkiana, Agonandra brasiliensis, Albizia niopoides, Alibertia concolor, A. macrophylla, Anadenanthera per-egri na, Apei ba ti bourbou, Astr onium fraxi ni foli um, Bastardi opsi s densi flora, Cal li sthene major, D albergia mi scolob-ium, Diospyros hispida, Eugenia punicifolia, Faramea cyanea, Genipa americana, Gomidesia lindeniana, Guatteriasell owiana, H edyosmum brasi li ense, I nga alba, I xora warmi ngii , L uehea paniculata, M achaeri um acuti foli um, M ar-gari tari a nobil i s, M i coni a chami ssois, M . sell owiana, M yrci a tomentosa, Nectandra cissi flora, Platypodi um elegans,Pouteri a gardnerii , Proti um spruceanum, Pseudobombax tomentosum, Pseudolmedi a guarani ti ca, Pterogyne ni tens,Qualea di chotoma, Rudgea vi bur noides, Salacia ell i pti ca, Si paruna guianensi s, Si phoneugena densi flora, Styrax cam-porum, Sweeti a f rut i cosa, Symplocus ni tens, Termi nali a ar gentea, T i bouchi na candoll eana, V i rola sebi fera, V i smi aguianensi s, Xylopia aromati ca.


16/18


TABLE 5. Continued.

Supertramp species:Aegiphila sellowiana, Alchornea glandulosa, A. triplinervea, Amaioua guianensis, Andira fraxini-folia, Aspidosperma parvifolium, Cabralea canjerana, Calophyllum brasiliense, Cariniana estrellensis, Casearia decandra, C.sylvestr is, Cecropia pachystachya, Cedrella fissi li s, Celt is i guanaea, Copaifera langsdorffii, Cordi a sell owiana, Croton flori -bundus, Cupania vernal is, Dendropanax cuneatum, Endli cheri a paniculata, Erythroxylum ci tr if olium, Eugeni a florida,Guapira opposit a, Guarea guidonia, G. macrophyll a, Guazuma ulmi folia, H ymenaea courbari l, Inga vera, Jacaratia spinosa,Luehea divari cata, M abea fistul i fera, M aclura ti nctori a, M atayba elaeagnoides, M . guianensis, M aytenus communis, M yrciarostrata, M yrciaria flori bunda, Myrsine umbell ata, Nectandr a oppositi foli a, Ocotea corymbosa, Pera glabrata, Piptadeniagonoacantha, Protium heptaphyll um, Roupala brasi li ensis, Sapium glandulatum, Sorocea bonplandi i , Tabebuia serrat i folia,Tapirira guianensis, Trema micrantha, Trichilia catigua, Zanthoxylum rhoifolium.

forests reappear in Esprito Santo state, as annualrainfall increases and seasonali ty decreases, untilreaching the warm and hyper-humid hylaea ofsouthern Bahia state. As demonstrated by the floris-tic analyses, these climatic variationscorrelated witha strong floristic differentiation between northern

and southern coastal rain forests, a fact already de-tected by Siqueira (1994) and Oliveira-Filho andRatter (1995). Thesetwo forest blocksrepresent dif-ferent and gradual floristic transitions from rain tosemi-deciduous forests as rainfall amounts decrease,both ending at the CamposdosGoitacazesgap. Thenorthern transit ion occurs in warmer climate andincludes declining mean temperatures toward thesouth. The southern transit ion starts in cooler sub-tropical climate and includes rising temperaturesto-ward the north. Mean temperature is probably thechief factor determining the northsouth floristicdifferentiation of coastal rain forests, while bothtemperature and rainfall regime account for muchof the internal variation within thetwo forest blocks.

Another important change toward the north i sthat the mountain ranges are progressively fartherfrom the coast and become lower in altitude, par-ticularly north of Rio Doce. This opens spacefor awider band of coastal lowlands known as tabuleirosthat extend from northern Rio de Janeiro to north-eastern Brazil . Changesin rainfall regimetoward theinterior are more gradual and have a smaller rangein Esprito Santo, eastern Minas Gerais, and south-ern Bahia than in Sao Paulo and neighboring states.Likewise, the transiti on between rain and semi-de-ciduous forests is relatively gentler in the north. Ad-ditionally, low altitudespenetratedeep into thecon-tinent along the valleys of the Rio Doce, Mucuri,

and Jequitinhonha, allowing typical lowland rainforest speciesto expand their distri bution toward theinterior. This is known for many rain forest species,both Amazonian and Atlantic, that are able to ex-pand their distribution into areasof strongly season-al climatesvia riverine forests (Oliveira-Filho & Rat-ter 1995). This may explain why the areas of semi-

deciduous forest in the Rio Doce River basin showstrong floristic l inks with the tabuleiro rain forestsof Esprito Santo and southern Bahia.

Rainfall seasonali ty was apparently more impor-tant than annual rainfall in distinguishing rain andsemi-deciduous forests, and the 30-day duration of

the dry season wasan effectivelimit ing criterion thatproduced a geographic distri bution largely coinci-dent with that of IBGE (1993). Rain forests arefound in areas with annual rainfall as low as 1173mm (Rio de Janeiro), provided that the rains arewell distri buted. Only among subtropical forests oflow alt it udes does annual rainfall prevail over sea-sonali ty in separating rain and semi-deciduous for-ests, probably because low winter temperaturesplayan addit ional role in forest deciduousness as sug-gested by Holdridge et al. (1971). Increasing rainfallseasonality with increasing distance from the oceanalso was a leading factor determining floristic differ-entiati on among hinterland semi-deciduous forests.The forest enclaves occurring within the Cerradodomain (e.g., Braslia), where dry seasons may lastfor up to 160 days, were the most differentiatedsemi-deciduousforests; they also had greater flori sticlinks with the Cerrado tree flora. These facts mayreinforcethe view of rain and semi-deciduousforestsin southeastern Brazil asa continuum of treespeciesdistributi on determined basically by rainfall regimethat also leadsto the Cerrado treeflora, assuggestedby Leitao Filho (1987); however, one must bear inmind that other important factors also intervene inforestsavanna transitions, particularly fire and soilfertility (Furleyet al. 1992).

To a considerable extent, the tree flora of semi-deciduous forests is a fraction of the much richer

rain forest flora, and probably is composed of speciesable to cope with relatively longer dry seasons. Treespecies diversity is highly correlated with water con-sumption and energy uptake, resourcesthat are par-ti tioned among species and limit their number inforest communities (Hugget 1995). Water shortageprobably playsthechief role in reducing speciesri ch-


17/18


ness of semi-deciduousforests compared to rain for-ests. The simpler structure of semi-deciduousforestsalso favors a comparatively reduced number of un-

derstory species (Gentry & Emmons 1987). Therewas greater floristic similarity at all taxonomic levelsbetween Atlantic rain and semi-deciduous foreststhan between any of the latter and Amazonian rainforests. Therefore, there is little floristic ground forviewing Atlantic rain forests as being closer to Am-azonian rain forests than to their adjacent semi-de-ciduous forests. Amazonian and Atlantic rain forestsare more similar in physiognomic characteristicsthan floristic aspects (Silva & Shepherd 1986). It isnot incorrect to describe rain and semi-deciduousforests as physiognomic and floristic expressions ofa single great Atlantic Forest domain i f the conceptof the Amazonian forest domain includesclosed rainforests (both upland and floodplain), open rain for-

ests (monsoon forests), heath forests (campinarana),and incidentally, semi-deciduous forests (Veloso etal.1991). T he definit ion of Atlantic forests shouldbe as comprehensiveasthat of Amazonian forests inorder to encompass all forest physiognomieseast ofthe dry corridor, from northeastern Brazil to south-ern Brazil, eastern Paraguay, and northeastern Ar-

gentina. This addsnot only the semi-deciduousfor-ests but also the southern subtropical Araucariafor-ests and the northeastern enclavesof brejo forests to

the South American Atlantic forests.

ACKNOWLEDGMENTS

ATOF thanks the CNPq for research grant no. 301644/88-8. We were helped during the t axonomic revision ofthe database by Carolyn Proenca (Myrtaceae) from theUniversity of Braslia; Jose Rubens Pirani (Simaroubaceaeand Rutaceae), Marina Tereza Campos (Rubiaceae), andLucia Lohman (Bignoniaceae) from the University of SaoPaulo; Maria Lucia Kawazaki (Myrtaceae) and Ines Cor-deiro (Euphorbiaceae) from the Sao Paulo Botanic Garden;Carlos Reynel (Zanthoxylum) from the Missouri BotanicalGarden; Washington Marcondes-Ferreira (Aspidosperma)from t he University of Campinas; D ouglas Daly (Burser-aceae) from the New York Botanical Garden; and GwilLewis (Leguminosae), Terry Pennington (Ingaand Sapo-taceae), and Daniela Zappi (Cactaceae and Rubiaceae)

from the Royal Botanic Gardens, Kew. Tereza Sposit o, Ro-sangela Tri stao Borem, Maryland Sanchez, Fernando Ped-roni, Jeanini Felfili, D orothy SueDunn deAraujo, WagnerLopes, Ariane Peixoto, Alexandre Francisco daSilva, MariaTeresa Zugli ani Toniato, Bruno Senna Correa, VeronicaUlup Andersen, Sebasti ao do Amaral, and Marcelo Trin-dade Nascimento helped to obtain checklists that were dif-ficult to access. We thank Brenda Mackey and GarryHartshorn for their valuable comments on the manuscript.

LITERATURE CITED

ANDRADE-LIMA, D. 1964. Contribuicao a dinamica da flora do Brasil .Arq. Inst. Cien. Terra2: 1520.. 1982. Present-day forest refuges in northeastern Brazil. I nG. T. Prance (Ed.). Biological diversification in thetropics, pp. 220245. Plenum Press, New York, New York.

ARAUJO, D. S. D. 1997. Mata Atlantica: CPD sit e SA14, Cabo Frio Region, south-eastern Brazil .I nS. D. Davis, V. H.Heywood, O. Herrera-MacBryde, J. Vi lla-Lobosand A. C. H amilton. (Eds.). Centresof plant diversit y: a guide

and strategy for their conservation, pp. 373375. World Wildlife Fund and The World Conservation Union,London, England.AZEVEDO, A. 1950. Regioes climato-botanicas do Brasil. Estudo fit ogeografico e florestal. Anuario Brasileiro deEconomia

Florestal 11: 201232.BORHIDI, A. 1991. Phytogeography and vegetati on ecology of Cuba. Akademiai Kiado, Budapest, H ungary.CABRERA, A. L., AND A. W ILLINK. 1973. Biogeografia de America Latina. Secretaria General de la Organizacion de los

EstadosAmericanos, Washington, D.C.CAMARA, I. G. 1990. Plano de acao para a Mata Atlanti ca. Fundacao SOS Mata Atlanti ca, Sao Paulo, Brasil.DNM ET. 1992. Normais climatologicas (19611990). Departamento Nacional de Meteorologia, Mi nisterio da Agricul-

tura. Braslia, Brasil.FERNANDES, A., ANDP. BEZERRA. 1990. Estudo fitogeografico do Brasil. Stylus Com., Fortaleza, Brasil . 205 pp.FURLEY, P. A., J. PROCTOR, ANDJ. A. RATTER. 1992. Nature and dynamics of forestsavanna boundaries. Chapman and

Hall, London, England.GENTRY, A. H. 1995. Patterns of diversit y and floristic composit ion in Neotropical montane forests.I nS. P. Churchill ,

H. Balslev, E. Forero, and J. L. Luteyn (Eds.). Biodiversity and conservation of Neotropical montane forests.Neotropical Montane Forest Biodiversity and Conservation Symposium 1, pp. 103126. N ew York BotanicalGarden, New York, New York., AND L. H. EMMONS. 1987. Geographical variation in fertil it y, phenology, and composit ion of the understory of

Neotropical forests. Biotropica 19: 216227.GIULIETTI, A. M., AND J. R. PIRANI. 1988. Patterns of geographic distribution of somespeciesfrom the Espinhaco Range,Mi nas Gerais and Bahia, Brazil. InP. E. Vanzolini and W. R. Heyer (Eds.) Proceedings of a Workshop onNeotropical Distributi on Patterns, pp. 3969. Academia Brasileira de Ciencias, Rio de Janeiro, Brasil .

HOLDRIDGE, L. R., W. C. GRENKE, W. H. HATHEWAY, T. LIANG, AND J. A. TOSIJR. 1971. Forest environmentsin tropicalli fe zones: a pilot study. Pergamon, Oxford, England.

HUGGET, R. J. 1995. Geoecology, an evolutionary approach. Routledge, London, England.IBGE. 1993. Mapa de vegetacao do Brasil . Instituto Brasileiro de Geografia e Estatstica. Rio de Janeiro, Brasil.


18/18


JONES, H. G. 1992. Plants and microclimate: a quantitative approach to environmental plant physiology 2. CambridgeUniversity Press, Cambridge, England.

KENT, M., AND P. COKER. 1992. Vegetation description and analysis: a practi cal approach. Belhaven Press, London,England.

LEDRU, M. P. 1993. Late Quaternary environmental and climatic changes in central Brazil. Quat. Res. 39: 9098., ANDM. L. SALGADO-LABOURIAU, ANDM. L. LORSCHEIT ER. 1998. Vegetation dynamics in southern and centralBrazil during the last 10,000 yr. B. P. Rev. Palaeobot. Palyn. 99: 131142.

LEITAO FILHO, H . F. 1987. Consideracoes sobre a florsticade florestas tropicais e sub-tropicais do Brasil. IPEF 35: 4146.. 1993. Ecologia da Mata Atlanti ca em Cubatao. Ed. UNESP/UN ICAMP, Sao Paulo, Brasil.

MCCUNE, B., AND M. J. MEFFORD. 1997. PC-ORD. M ultivariate analysis of ecological data, version 3.0. MjM SoftwareDesign, Glaneden Beach, Oregon.

MORI, S. A., B. M. BOOM, AND G. T. PRANCE. 1981. Di stribution patterns and conservation of eastern Brazilian coastalforest tree species. Brittonia 33: 233245.

OLIVEIRA-FILHO, A. T., AND J. A. RATTER. 1994. D atabase: woody flora of 106 forest areas of eastern tropical SouthAmerica. Royal Botanic Garden Edinburg, Edinburgh, Scotland., AND . 1995. A study of the origin of central Brazilian forests by the analysis of plant speciesdistributionpatterns. Edinb. J. Bot. 52: 141194., E. A. VILELA, M. L. GAVILANES, AND D. A. CARVALHO. 1994. Comparison of t he woody flora and soils of sixareas of montane semi-deciduous forest in southern Mi nas Gerais, Brazil. Edinb. J. Bot. 51: 355389.

PRADO, D. E., AND P. E. GIBBS. 1993. Patterns of species distribution in the dry seasonal forests of South America. Ann.Mo. Bot. Gard. 80: 902927.

RATTER, J. A., S. BRIDGEWATER, R. ATKIN SON, AND J. F. RIBEIRO. 1996. Analysis of the floristic composit ion of theBrazilian cerrado vegetation. Edinb. J. Bot. 53: 153180.READ, J., ANDR. S. HILL. 1989. The responseof someAustralian temperaterain forest treespeciesto freezingtemperatures

and i ts biological significance. J. Biogeogr. 16: 2127., ANDG. S. H OPE. 1989. Foliar resistance of some evergreeen tropical and extratropical Australasian Nothofagusspecies. Aust. J. Bot. 37: 361373.

RIZZINI, C. T. 1963. Nota previa sobre a divisao fitogeografica do Brasil. Revta. Brasil. Geogr. 25: 164.SALIS, S. M., G. J. SHEPHERD, ANDC. A. JOLY. 1995. Floristic comparison of mesophytic semi-deciduous forests of the

interior of the state of Sao Paulo, southeast Brazil. Vegetatio 119: 155164.SILVA, A. F., AND G. J. SHEPHERD. 1986. Comparacoes florsticas entre algumas matas brasileiras utilizando anal ise de

agrupamento. Revta. Brasil. Bot. 9: 8186.SIQUEIRA, M. F. 1994. Anal ise florsti ca e ordenacao de especies arboreas da Mata Atlantica atraves de dados binarios.

MSc. thesis, UniversidadeEstadual de Campinas, Campinas, Brasil .SOS MATAATLANTICA. 1998. Atlas da evolucao dos remanescentes florestais e ecossistemas associados no domnio da

Mata Atlantica no perodo 19901995. Fundacao SOS Mata Atlantica, Sao Paulo, Brasil., AND IN PE. 1993. Evolucao dosremanescentesflorestaise ecossistemasassociadosdo domnio daMataAtlantica.Fundacao SOS Mata Atlantica, I nstituto Nacional de PesquisasEspaciais, Sao Paulo, Brasil.

TERBRAAK, C. J. F. 1987. The analysis of vegetationenvironment relati onships by canonical correspondence analysis.

Vegetatio 69: 6977.. 1988. CANOCOa FORTRAN program for canonical community ordination by (partial) (detrended) (ca-nonical) correspondence analysis, principal components analysis, and redundancy analysis, version 2.1. Technicalreport LWA-88-02, TNOInsti tute of Applied Computer Science, Wageningen, Germany.. 1995. Ordination. In R. H . G. Jongman, C. J. F. ter Braak, and O. F. R. van Tongeren. Data analysis incommunity and landscape ecology, pp. 91173. CambridgeUniversity Press, Cambridge, England.

THORNTWAITE, C. W. 1948. An approach toward a rational classification of climate. Geogr. Rev. 38: 5594.TORRES, R. B., F. R. MARTINS, ANDL. S. K. GOUVEA. 1997. Climate, soil, and tree flora relationships in forests in the

state of Sao Paulo, southeastern Brazil . Revta. Brasil . Bot. 20: 4149.VELOSO, H . P., A. L. R. RANGELFILHO, AND J. C. A. LIMA. 1991. Classificacao da vegetacao brasileira, adaptada a um

sistema universal. Instituto Brasileiro de Geografia e Estatstica, Rio de Janeiro, Brasil .VIANA, V. M., ANDA. A. J. TABANEZ. 1996. Biology and conservation of forest fragments in the Brazilian Atlantic moist

forest. I nJ. Schelhas and R. Greenberg (Eds.). Forest patchesin tropical landscapes, pp. 151167. Island Press,Washington, DC., , AND J. L. F. BATI STA. 1997. Dynamics and restoration of forest fragments in the Brazilian Atlanticmoist forest. I nW. F. Laurance and R. O. Bierregaard (Eds.). Tropical forest remnantsEcology, management,and conservation of fragmented communities, pp. 351365. T he University of Chicago Press, Chicago, Ill inois.

VON MARTIUS, C. F. P. 1840. Tabulphysiognomie. Brasili regionis iconibus express. I nC. F. P. von Martius, S.

Endlicher, A. G. Eichler, and J. Urban (Eds.). Flora brasiliensis, vol. 1, pp. 1110, Tomus1. Lipsaea apud Frid.Fleischer in Comm., Munich, Germany.

WALTER, H. 1985. Vegetation of the earth and ecological systems of the geo-biosphere. Third edition Springer-Verlag,Berlin, Germany.

WEBSTER, G. L. 1995. T he panorama of Neotropical cloud forests.I nS. P. Churchill, H. Balslev, E. Forero, and J. L.Luteyn (Eds.). Biodiversity and conservation of Neotropical montane forests. Neotropical Montane Forest Bio-diversit y and Conservation Symposium 1, pp. 5377. New York Botanical Garden, New York, New York.

Patterns of Floristic Differentiation among Atlantic Forests in Southeastern Brazil and the...

Documents

Transcript of Patterns of Floristic Differentiation among Atlantic Forests in Southeastern Brazil and the...