ArticlePDF Available

Delving into the variations in tree species composition and richness across South American subtropical Atlantic and Pampean forests

Abstract and Figures

Aims: We analyse here the variations in species composition and richness and the geographic ranges of the tree species occurring in South American subtropical Atlantic and Pampean forests. Our goals were to assess (i) the floristic consistency of usual classifications based on vegetation physiognomy, climate and elevation; (ii) the leading role of temperature-related variables on the variations in species composition and richness; (iii) the predominance of species with tropical–subtropical ranges, possibly as a result of forest expansion over grasslands after the Last Glacial Maximum (LGM); (iv) the restriction of most subtropical endemics to stressful habitats as a possible result of past forest refuges during the LGM. Methods: The region was defined by the Tropic of Capricorn to the north, the Rio de la Plata to the south, the Atlantic shoreline to the east and the catchment areas of the upper Paraná and Uruguay Rivers to the west. Multivariate analyses, multiple regression modelling and variance partition analyses were performed on a database containing 63 994 occurrence records of 1555 tree species in 491 forest sites and 48 environmental variables. All species were also classified according to their known geographic range. Important Findings: A main differentiation in species composition and richness was observed between the eastern windward coastlands (rain and cloud forests) and western leeward hinterlands (Araucaria and semi-deciduous forests). Pre-defined forest types on both sides were consistent with variations in tree species composition, which were significantly related to both environmental variables and spatial proximity, with extremes of low temperature playing a chief role. Tree species richness declined substantially towards the south and also from rain to seasonal forests and towards the highland summits and sandy shores. Species richness was significantly correlated with both minimum temperature and actual evapotranspiration. About 91% of the subtropical flora is shared with the much richer tropical flora, probably extracting species that can cope with frost outbreaks. The 145 subtropical endemics were not concentrated in harsher habitats.
Content may be subject to copyright.
Journal of
Plant Ecology
VOLUME 8, NUMBER 3,
PAGES 242–260
JUNE 2015
doi:10.1093/jpe/rtt058
Advance Access publication
2 December 2013
available online at
www.jpe.oxfordjournals.org
© The Author 2013. Published by Oxford University Press on behalf of the Institute of Botany, Chinese Academy of Sciences and the Botanical Society of China.
All rights reserved. For permissions, please email: journals.permissions@oup.com
Delving into the variations in tree
species composition and richness
across South American subtropical
Atlantic and Pampean forests
Ary T.Oliveira-Filho1,*, Jean C.Budke2, João A.Jarenkow3,
Pedro V.Eisenlohr1 and Danilo R.M.Neves4
1 Departamento de Botânica, Instituto de Ciências Biológicas, Universidade Federal de Minas Gerais, Av. Antônio Carlos,
6627, Belo Horizonte, Minas Gerais, 31270–901, Brazil
2 Departamento de Ciências Biológicas, Universidade Regional Integrada do Alto Uruguai e das Missões, Av. Sete de
Setembro, 1621, Erechim, Rio Grande do Sul, 99700–000, Brazil
3 Departamento de Botânica, Instituto de Biociências, Universidade Federal do Rio Grande do Sul, Av. Bento Gonçalves,
9500, Porto Alegre, Rio Grande do Sul, 91501–970, Brazil
4 Royal Botanic Garden Edinburgh, Inverleith Row, Edinburgh, EH3 5LR, UK
*Correspondence address. Laboratório de Ecologia Vegetal, Departamento de Botânica, Instituto de Ciências
Biológicas, Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, Belo Horizonte, Minas Gerais,
31270–901, Brazil. Tel:+55-31-3409-2688; Fax: +55-31-3409-2671; E-mail: ary.oliveira.filho@gmail.com
Abstract
Aims
We analyse here the variations in species composition and richness
and the geographic ranges of the tree species occurring in South
American subtropical Atlantic and Pampean forests. Our goals were
to assess (i) the floristic consistency of usual classifications based on
vegetation physiognomy, climate and elevation; (ii) the leading role
of temperature-related variables on the variations in species com-
position and richness; (iii) the predominance of species with tropi-
cal–subtropical ranges, possibly as a result of forest expansion over
grasslands after the Last Glacial Maximum (LGM); (iv) the restriction
of most subtropical endemics to stressful habitats as a possible result
of past forest refuges during the LGM.
Methods
The region was defined by the Tropic of Capricorn to the north, the
Rio de la Plata to the south, the Atlantic shoreline to the east and
the catchment areas of the upper Paraná and Uruguay Rivers to the
west. Multivariate analyses, multiple regression modelling and vari-
ance partition analyses were performed on a database containing
63994 occurrence records of 1555 tree species in 491 forest sites
and 48 environmental variables. All species were also classified
according to their known geographic range.
Important Findings
A main differentiation in species composition and richness was
observed between the eastern windward coastlands (rain and cloud
forests) and western leeward hinterlands (Araucaria and semi-decid-
uous forests). Pre-defined forest types on both sides were consistent
with variations in tree species composition, which were significantly
related to both environmental variables and spatial proximity, with
extremes of low temperature playing a chief role. Tree species rich-
ness declined substantially towards the south and also from rain
to seasonal forests and towards the highland summits and sandy
shores. Species richness was significantly correlated with both mini-
mum temperature and actual evapotranspiration. About 91% of the
subtropical flora is shared with the much richer tropical flora, prob-
ably extracting species that can cope with frost outbreaks. The 145
subtropical endemics were not concentrated in harsher habitats.
Keywords: distribution patterns, endemic species, floristic
connections, macro-scale species richness, spatial and climatic
predictors, subtropical forests
Received: 30 April 2013, Revised: 11 September 2013,
Accepted: 12 October 2013
INTRODUCTION
The highly threatened remnants of South American Atlantic
forests, one of the World’s biodiversity hotspots (Galindo-Leal
and Câmara 2003; Myers etal. 2000), are presently a focus
of an increasing number of studies on the spatial distribu-
tion of biological traits across scales that vary from whole
geographic ranges to fragmented habitats. The variety of
June
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 243
biological traits includes biodiversity, species endemism and
rarity, and selected groups of organisms circumscribed either
taxonomically or functionally (Carnaval and Moritz 2008;
Metzger etal. 2009; Pardini etal. 2009). Recent contributions
include birds (Harris et al. 2005; Silva et al. 2004), mammals
(Costa et al. 2000; Galetti etal. 2009; Vieira et al. 2009), rep-
tiles (Grazziotin etal. 2006) and arthropods (Tyler etal. 1994;
Uehara-Prado etal. 2009), as well as plants (Lopes etal. 2009;
Murray-Smith et al. 2008; Oliveira-Filho and Fontes 2000;
Scarano 2009) and plant–animal interactions (Almeida-Neto
etal. 2008). These studies represent a great contribution for
conservation science and policy as they bring more light to
the knowledge on the present-day distribution of biological
variation across the scanty but precious forest remnants that
represent <16% of the primitive extension of Atlantic forests
(Ribeiro etal. 2009).
Many studies of the kind are constrained by the difficulty
of selecting and collecting abundant and reliable informa-
tion from a highly diverse biota and its complex ecological
network, which have been heavily impacted by humans
(Metzger 2009). In the same context, the analysis of meta-
data at geographic scales is seriously affected by the fact that
Atlantic forests have been poorly and irregularly collected
throughout their area, and the known range of many taxa is
therefore liable to constant changes. Comparatively speaking,
tree species are probably a more privileged (or less problem-
atic) group as suggested by the increasing number of wide-
scale floristic analyses produced for the tree flora of Atlantic
Biogeographic Domain as a whole (Oliveira-Filho etal. 2006),
or for some of their geographic sections, such as the south-
east (Marques et al. 2011; Oliveira-Filho and Fontes 2000),
the north-east (Ferraz et al. 2004; Santos et al. 2007), the
southern Araucaria forests (Higuchi et al. 2012; Jarenkow
and Budke 2009), the Paraguay-Paraná Basin (Spichiger etal.
2004) and the Brazilian states of São Paulo (Salis etal. 1995;
Scudeller etal. 2001; Torres etal. 1997) and Rio Grande do Sul
(Gonçalves and Souza 2013). These studies have made impor-
tant contributions to the knowledge on the distribution pat-
terns of tree species and associated environmental variables
and helped assessing the floristic consistency of traditional
classifications of the Atlantic Biogeographic Domain into both
biogeographic units and vegetation types, most of which were
originally produced from descriptions of climatic patterns and
vegetation typesonly.
Tree species distribution patterns of Atlantic forests are like-
wise the study object of this article, and we here bring focus
to a peculiar geographic section of their range: the subtropics.
The original area of ca 1.1 million km2 of the Atlantic Domain
stretches for over 4000 km along the eastern coast of Brazil
between 4° and 32° S (Thomas 2008) and therefore extends
across both the tropical and subtropical realms. The currently
prevalent concept of Atlantic forests is the wide-sense con-
cept that includes not only the narrow band (generally <100
km wide) of coastland tropical and subtropical evergreen rain
forests (the previously dominant strict sense concept) but also
a number of other forest types (Carnaval and Moritz 2008;
Fernandes and Bezerra 1990). Seasonal semi-deciduous for-
ests outstand among the other types because they cover the
largest area and make up most of the hinterland extension
of both tropical and subtropical Atlantic forests that meet the
neighbouring Biogeographic Domains dominated by open and
seasonally dry vegetation types. These hinterland expanses of
semi-deciduous forests become increasingly larger towards
the subtropics where they go as far as eastern Paraguay and
north-eastern Argentina and flank the hyper-seasonal to sem-
iarid vegetation of the Chaco and Espinal Domains to the west
and the hyper-seasonal prairies of the Pampa Domain to the
south (Fig.1). Subtropical Atlantic forests is also home of a
large swathe of an exceptional vegetation type, the Araucaria
forests that cover large extents of the inner highlands of the
three southern Brazilian states and the Misiones province in
Argentina (Fig.1). They are easily recognized for their cano-
pies that are fully outcropped by the chandelier-like crowns
of the gymnosperm Araucaria angustifolia (Bertol.) Kuntze.
At higher altitudes and near shallower soils Araucaria forests
form vegetation mosaics with highland grasslands, the campos,
and meet Atlantic cloud forests along the crests of the coastal
hill-ranges (Jarenkow and Budke 2009; Klein 1990).
Studies of tree species distribution patterns have dem-
onstrated that the traditional dichotomy between Atlantic
coastland rain and hinterland seasonal forests based on physi-
ognomy, and climate is coherently corresponded by a floristic
differentiation though it invariably comes out as a continuum
of species turnover rather than highly distinct tree floras
(Ferraz etal. 2004; Oliveira-Filho and Fontes 2000; Oliveira-
Filho etal. 2005; Santos etal. 2007; Torres etal. 1997). Oliveira-
Filho etal. (2006), when assessing this coastland–hinterland
dichotomy across the whole Atlantic Domain, found evidence
that the two tree floras were less differentiated in the sub-
tropical sector. This is why, we readdress the issue here with
increased emphasis and effort. We considered that the excep-
tional subtropical climate of the southern extreme of Atlantic
forests and the outstanding physiognomic heterogeneity of
their vegetation justified a floristic and environmental analy-
sis focused in the region, particularly when they help in clari-
fying some controversial issues regarding its classification into
biogeographic provinces, the differentiation among forest
types, and the distribution patterns of both species richness
and endemism. In addition, the clarification of these patterns
is of vital importance for conservation issues.
Subtropical Atlantic forests have also been the subject of
a number of palaeo- and actuobiogeophic studies, and spe-
cial attention has been given to the environmental deter-
minants of vegetation physiognomies (e.g. Falkenberg and
Voltolini 1995; Leite 2002; Roderjan etal. 2002; Spichiger etal.
1995), the formation of species blends through main migra-
tion routes (e.g. Leite 2002; Rambo 1951a, 1953, 1961; Smith
1962; Waechter 2002; Waechter and Jarenkow 2003) and the
changes in vegetation cover that would have accompanied
the climatic shifts of the Quaternary caused by the Glacial/
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
244 Journal of Plant Ecology
VEGETATION TYPES OF THE 491 FORESTSITES:
Coastal sandy evergreen rain dwarf-forest
Lowland evergreen rain forest Subtropical Atlantic
Lower hill-land evergreen rain forest rain forests
Upper hill-land evergreen rain forest
Lower hill-range evergreen cloud forest Subtropical Atlantic
Lower hill-range rocky evergreen cloud dwarf-forest cloud dwarf-forests
Upper hill-range rocky evergreen cloud dwarf-forest
Lowertableland evergreen seasonal Araucaria forest Subtropical Atlantic
Upper tablelandevergreen seasonal Araucaria forest evergreen seasonal
Lower hill-range evergreen seasonal Araucaria forest Araucariaforest
Upper hill-range evergreen seasonal Araucaria forest
Lowland semi-deciduous seasonal forest Subtropical Atlantic
Lower tableland semi-deciduous seasonal forest semi-deciduous forests
Upper tableland semi-deciduous seasonal forest
Coastal sandy seasonal evergreen dwarf-forest Subtropical Pampean
Lowland riverine semi-deciduous seasonal forest semi-deciduous forests
Lower hill-land riverine semi-deciduous seasonal forest
24°
26°
28°
30°
32°
34°
54°56°550°458°46°
São Paulo
Paraná
Santa
Rio Grande
Catarina
do Sul
PARAGUAY
URUGUAY
ARGENTINA
BRAZIL
Tropic of Capricorn
Atlantic rain and cloud forests
Atlantic Araucariaforests
Atlantic semi-deciduous forests
Pampaand highland prairies
Tropical savannas (cerrado)
Subtropical savannas (espinal)
Semi-arid woodland (chaco)
Figure1: Map of south-eastern South America showing the distribution of the dominant vegetation types of the subtropical section of the
Atlantic Forest and Pampa Biogeographic Domains, and the location of the 491 forest sites used in this study together classified into forest types
following Oliveira-Filho (2009).
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 245
Interglacial cycles (e.g. Behling 1995; Behling et al. 2005;
Behling and Pillar 2007, 2008; Behling and Negrelle 2001; de
Oliveira et al. 2005; Negrelle 1998; Quadros and Pillar 2002;
Stefenon et al. 2008). Most studies generally propose a post-
glacial process of forest expansion into a grassland-dominated
region during the Holocene, and the formation of species
blends composed mostly of immigrants not only from outer
sources but also from local forest refuges such as riparian
forest (Oliveira etal. 2008). As forests expanded, grasslands
would be confined to the their present-day distribution, as
either the highland enclaves of the Atlantic Domain or the
lowland prairies of the Pampa Domain.
Despite their important contribution to those mainstream
studies and also to conservation initiatives, there are not many
wide-scale quantitative assessments of the present-day floris-
tic variation of subtropical Atlantic forests in association with
underlying environmental factors, and the existing ones had
their focus in different geographic ranges or vegetation types
(Gonçalves and Souza 2013; Higuchi etal. 2012; Jarenkow and
Budke 2009; Oliveira-Filho etal. 2006; Spichiger etal. 2004).
In this study, we investigated patterns of geographic distribu-
tion and endemism of 1555 tree species based on checklists
produced for 173 forest sites spread across the whole sub-
tropical section of the Atlantic Domain and the section of the
neighbouring Pampa Domain that extends southwards until
the Rio de La Plata (Fig. 1). The purpose was to assess the
expanded distribution of Atlantic forest species towards the
south through the forest enclaves of the Pampa Domain. We
not only sought patterns of tree species distribution and floris-
tic differentiation within this range but also investigated the
distribution range of all 1555 species outside the subtropics
and therefore assessed the proportion of strictly subtropical
endemic species. As subtropical Atlantic and Pampean for-
ests include the southern extreme of the latitudinal range of
a great number of tree species, they are also another suitable
object to contribute to the long-standing controversial theme
of biogeography, which is explaining the widely recognized
pattern of the latitudinal diversity gradient (LDG; Hillebrand
2004; Willig etal. 2003). Among the many available hypoth-
eses our data set was suitable to assess those related to energy
(primary productivity) and climate harshness.
This study was therefore based on five questions of crucial
conservation value for subtropical Atlantic and Pampean for-
ests. (i) Is the dichotomy between coastland and hinterland
Atlantic forests also consistent in terms of tree species compo-
sition in the subtropical realm? We expected a positive answer
based on known climatic contrasts and on the supposed post-
glacial expansion of forests into the region in two distinct
fronts. (ii) Classifications of these forests based on vegetation
physiognomy, climate and elevation are consistent with vari-
ations in tree species composition? We expected a positive
answer because of similar results found in the tropical section
of Atlantic forests. (iii) Are extremes of low temperature and
temperature seasonality the main factors underlying varia-
tions in tree species composition and richness in the region?
We expected a positive answer because climate diagrams
show no period of water deficit for the whole region. (iv) Is
the tree flora of subtropical Atlantic forests composed of a vast
majority of wide-range non-endemic species? We expected a
positive answer because of the consistent body of evidence
supporting a massive forest expansion over grasslands after
the Last Glacial Maximum (LGM). (v) Is the present distribu-
tion of subtropical endemic tree species concentrated in more
stressful habitats? We expected a positive answer because it
is also believed that post-LGM forest expansion also occurred
from existing forest refuges.
METHODS
Studyarea
We circumscribed the geographic range of South American
subtropical Atlantic and Pampean forests and classified the
vegetation following the criteria and nomenclature proposed
by Oliveira-Filho (2009) for the vegetation of eastern tropical
and subtropical South America. This system is actually a fur-
ther development of the widely accepted Instituto Brasileiro
de Geografia e Estatística (IBGE) classification system for
Brazilian vegetation (Veloso etal. 1991; reissued in IBGE 2012)
though it seeks to describe physiognomic and environmental
variations at much smaller scales than those covered by the
IBGE and therefore offers additional and optional descrip-
tive tools. The latitudinal range extended from the Tropic of
Capricorn (23°2616S) to the southern margin of the Rio de
la Plata (ca. 35°S), and the longitudinal range extended from
the Atlantic coast to the catchment areas of the Upper Paraná
and Uruguay Rivers (Fig.1). The area encompassed then the
subtropical section of the Atlantic Biogeographic Domain and
the northern section of the Pampa Domain sensu Fernandes
and Bezerra (1990) or, alternatively, sections of the Atlantic,
Paranense and Pampa Biogeographic Provinces sensu Cabrera
and Willink (1973).
Geographic and floristic metadata
We extracted the data set from TreeAtlan 2.1 (Oliveira-Filho
and Eisenlohr 2012), a relational database that contains tree
species checklists compiled from the literature and herbar-
ium specimens for >2000 sites across tropical and subtropical
South America. TreeAtlan contains floristic data only (pres-
ence/absence) because it has been designed to provide for
studies at large geographic scales for which abundance infor-
mation is scarce and, according to Kent (2011), of limited
value compared with species occurrence. ATreeAtlan ‘site’ is
defined by a single vegetation type contained in a circular area
with a 10-km diameter. Therefore, where two or more vege-
tation types co-occur, there may be overlapping sites, each for
a distinct vegetation type. TreeAtlan does not include occur-
rence records without indication or evidence of vegetation
type and sites with poor species lists (based on data set man-
ager’s experience). The latter is an important sieve because
different sample efforts across sites may bias their descriptive
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
246 Journal of Plant Ecology
power. TreeAtlan ‘tree species’ are defined as plants with free-
standing stems potentially growing >3 m high. The extracted
data set contained 63 994 occurrence records of 1555 tree
species in 491 sites (or checklists) of subtropical Atlantic and
Pampean forests and 4 geographic and 45 environmental var-
iables for each site. The geographic variables were latitude,
longitude, altitude and shortest distance to the ocean. The
environmental variables included 27 and 18 related to climate
and substrate, respectively. The climatic variables consisted of
the 19‘bioclimatic parameters’ produced by WorldClim 1.4,
a high resolution (30) set of global climate layers created by
Hijmans et al. (2005); the potential and actual evapotranspi-
ration (mm) and the aridity index, derived from WorldClim
layers by Zomer et al. (2006); the mean duration (days) and
severity (mm) of water deficit periods, both extracted from
Walter diagrams (Walter 1985); and the mean frequency of
frosts (days), cloud cover (%) and cloud interception (mm),
obtained from gridded data sets produced by Jones and Harris
(2008). The substrate-related variables included the sites’ pre-
vailing slope and aspect, obtained from CGIAR-CSI (2006);
the mean, minimum and maximum monthly soil moisture
(%), obtained from the International Soil Moisture Network
(www.ipf.tuwien.ac.at/insitu/); and 13 selected topsoil varia-
bles (0–20 cm in depth), obtained from the Harmonized World
Soil Database (FAO/IIASA/ISRIC/ISSCAS/JRC 2012) followed
by the selection of a single soil class for each site based on
the authors’ experience on soil–vegetation type relationships.
The topsoil variables were drainage class (ranked), rockiness
(% surface), proportion (% volume) of size fractions (clay,
silt, sand and gravel), bulk density (kg·dm−3), organic car-
bon (%weight), pH in water, salinity (ECe in dS·m-1), cation
exchange capacity (cmol·kg−1), total saturation of bases (%)
and total exchangeable bases (cmol·kg−1). As soil variables are
widely known to be very heterogeneous within the scale of
the sites, their inclusion here had the additional purpose of
assessing the power of the aforementioned data sources. We
organized the data set into two matrices, both with the 491
sites as row headings. The species matrix, obtained from the
checklists, contained the 1555 tree species as column headings
and binary occurrence records as entries. The environmental
matrix contained the 49 geographic and environmental vari-
ables as column headings. The geographic distribution of the
491 sites and their classification into forests types following
Oliveira-Filho (2009) are given in Fig.1. In order to assess the
coastland–hinterland dichotomy of the Atlantic tree flora, the
sites were merged into two main groups, hereafter referred
to as ‘Coastland’ forest sites (160 sites of subtropical Atlantic
rain and cloud forests located between the highland ranges
and the coastline) and ‘Hinterland’ forest sites (331 sites of
Pampean and Atlantic seasonal forests located south and west
of the highland ranges, including both Araucaria and semi-
deciduous forests).
We also extracted two additional data sets from TreeAtlan
2.1. The first added to the species matrix all other occurrence
records of its 1555 species across the Neotropics, including
therefore their distribution both inside and outside the geo-
graphic range of the 491 study sites. We then estimated the
contribution of endemic and non-endemic species to the sub-
tropical Atlantic and Pampean tree flora and that of other
Biogeographic Domains, specifically the Atlantic (tropical
section), Amazonian, Cerrado, Caatinga, Chaco and Andean
Piedmont Domains. We also considered as ‘widespread
Neotropical’ all those with distribution extending towards
Central America and the Caribbean. The second additional
data set only added species occurrence records of a narrow
latitude band from 23°S to the Tropic of Capricorn in order
reach an entire number, thereby equalizing graph intervals.
Then, we merged the whole set into five main forest types
and 12 1°-latitude bands (from 23 to 35°S) with the purpose
of describing changes in tree species richness across both lati-
tude and forest type.
Data analyses
We firstly performed an outlier analysis of the species matrix
and removed the top seven outlying sites (standard deviations
of average Sorensen’s distance >2.5), all of them in the Pampa
Domain and with very poor checklists. We then performed
a Detrended Correspondence Analysis (DCA) to reduce the
dimensionality of the outlier-free species matrix into species
composition gradients and then assessed their consistency
with the classification of sites into vegetation types by inspec-
tion of the clusters formed on the ordination diagrams. The
ordinations of all 484 Subtropical sites, 160 Coastland sites
and 324 Hinterland sites are hereafter referred to as DCA1,
DCA2 and DCA3, respectively. In the case of DCA2, however,
the residuals were above tolerance, and we only achieved tol-
erance after eliminating singletons and down-weighting rare
species. We assessed the species replacement along ordination
axes through the length of gradients and the significance of
eigenvalues through randomization tests with 1000 permu-
tations (McCune and Grace 2002).These analyses and tech-
niques are described in McCune and Grace (2002) and were
performed on PC-ORD 6.0 (McCune and Mefford 2006).
We then prepared regression models for all significant axes
of the three DCAs with ordination scores as response variables
representing the main species composition gradients summa-
rized by each axis. During this routine, we avoided the pitfalls
pointed out by Eisenlohr (2013). We performed a pre-selec-
tion of predicting environmental variables for each regression
model maintaining only those with post-hoc R2 > 0.2 with ordi-
nation scores. We then performed an additional and progres-
sive elimination of collinear variables based on their Variance
Inflation Factor (VIF) until maintaining only those with VIF <
10 (Quinn and Keough 2002). We identified sets of collinear
variables with the help of a Pearson correlation matrix and
then chose a single variable for each set based on the highest
correlation with the response variable.
Once defined the final set of candidate predictors of each
model, we assessed the spatial structure of residuals using
Moran’s correlograms followed by sequential Bonferroni
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 247
correction (Fortin and Dale 2005). Because there was spatial
structure in all cases, we added spatial filters as fixed variables
in the model selection process (Diniz-Filho etal. 2008). We
obtained the filters by the Moran’s Eigenvector Map (MEM)
method (Dray etal. 2006), giving preference to the significant
ones (P<0.05 in correlations with DCA axes). When MEMs
were not successful in accounting for the spatial structure
(second axis of DCA1 and first axis of DCA3), we progressively
inserted spatial filters through eigenvectors derived from con-
nectivity matrices (Griffith 2003). We eventually performed
an automatic selection of the best set of predictors using
the Corrected Akaike Information Criterion (Burnham and
Anderson 2002).
We then confirmed the normality of residuals in all models
graphically (residual histograms resembling the normal curve)
and through D’Agostino-Pearson’s and Lilliefors’ tests (at least
one of them should indicate normality). In some cases, it was
necessary to remove some sites that showed extreme values
of standardized residuals. We also addressed the spatial inde-
pendence of residuals (Diniz-Filho and Bini 2005) and the
remaining assumptions of regression models (homocedastic-
ity and linearity) by graphical analysis of residuals, following
Quinn and Keough (2002). We obtained the models using
the package SAM 4.0 (Rangel etal. 2010) though MEMs were
obtained using the package ‘spacemakeR’ (Dray etal. 2006) of
R Statistical Environment (R Development Core Team 2012).
In order to help the interpretation of models, we plotted their
significant predictors as arrows on DCA diagrams using the
graphic resources in PC-ORD6.0.
We obtained the relative contribution of environmental
and spatial components in explaining the variation in spe-
cies composition (variance partitioning) following the rou-
tine proposed by Dray etal. (2012) and Legendre etal. (2012),
using the packages ‘spacemakeR’, ‘packfor’ (Dray etal. 2009),
‘vegan’ (Oksanen etal. 2011) and ‘spdep’ (Bivand 2013) in
the R Statistical Environment. This routine involves (i) the
removal of species singletons; (ii) the Hellinger transforma-
tion of binary occurrence records; (iii) the preparation of two
Redundancy Analyses, one for species and environmental
variables and another for species and spatial filters; and (iv)
progressive selection of environmental and spatial variables,
the latter separated into two components, macro-spatial and
micro-spatial. Collinearities are not removed in variance parti-
tioning (Oksanen et al. 2013). The routine eventually produces
the fractions of the total variance explained exclusively by the
environmental, macro-spatial and micro-spatial components
and their significance, as well as the fractions explained by
inseparable combinations of these three components. We per-
formed the routine separately for all 484 Subtropical sites, the
160 Coastland sites and the 324 Hinterlandsites.
We also prepared a regression model for species richness
as response variable using the same procedure described
above. The species richness values were those obtained for
each of five main groups of forest types at each 1°-latitude
band (N = 42) because of both the high variance of species
richness of individual sites (coefficient of variation=55.8%)
and the main goal of approaching the LDG. The predicting
variables were the 42 means of the same above-mentioned 49
geographic and environmental variables plus three additional
variables: (i) the total area covered by the forest type at each
latitude band obtained from polygons drawn in ArcMapTM
10.0, (ii) the number of forest sites per area (as an estimate
of sampling intensity) and (iii) the standard deviation of for-
est site altitudes (as an estimate of environmental heterogene-
ity). The coordinates used were the central latitude of each
band and the mean longitude of forest sites. The final model
ended up with only two predictors and no spatial filter but
required the elimination of a single outlier with high standard-
ized residual.
In order to characterize the floristic composition of
Coastland and Hinterland forests, we performed an indica-
tor species analysis, obtaining indicator values calculated with
the method of Tichý and Chytrý (2006) available in PC-ORD
6.0 and testing their significance with Monte Carlo tests with
4999 randomizations. As there were 768 and 210 significant
IVs for Coastland and Hinterland forests, respectively, and 577
non-significant IVs, we listed only the 50 species with the top
IVs of the two groups as well as the 50 most frequent species
with non-significant IVs. We then performed the same test for
six main groups of forest types and listed those with the 50
top significant IVs.
RESULTS
Variations in tree species composition and
environmental and spatial predictors
The eigenvalues obtained by DCAs were significant for the first
three ordination axes of DCA1 and the first two ordination axes
of DCA2 and DCA3 (online supplementary Table S1). The coef-
ficients of determination for the correlations between ordina-
tion distances and distances in the original n-dimensional space
(online supplementary Table S1) indicated that DCA2 was the
most efficient in summarizing the higher proportion of the total
variance, reaching a cumulative R2 > 0.7 in the secondaxis.
All eigenvalues were <0.5 (online supplementary Table
S1), suggesting relatively ‘weak’ gradients in terms of spe-
cies turnover (ter Braak 1995). The first eigenvalue of DCA1
was the only one to approach this threshold indicating that
it summarized a particularly stronger gradient. Nevertheless,
the length of a gradient is more precise in assessing species
turnover, and values between one and four represent a scale
from half-change to complete replacement (Hill and Gauch
1980). In a remarkable opposition to what was indicated by
the eigenvalues, most lengths of gradients approached a full
species turnover (> 3); the only exceptions were the third axis
of DCA1 and second of DCA2(online supplementary Table S1).
These are the only gradients then to be considered ‘weak’,
while ‘strong’ gradients may be assumed in all othercases.
All DCA ordination diagrams (Figs 2 and 3) segregated dis-
tinct clusters of sites that were highly coincident with their a
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
248 Journal of Plant Ecology
priori classification into forest types. These results provided
strong evidence that the classification of Oliveira-Filho (2009),
based mainly on physiognomy and climate, was strongly con-
sistent with differences in tree species composition. All seven
regression models prepared for forest site scores on DCA ordi-
nation axes as dependent variables and environmental and
spatial filters as predictors yielded high values of R, R2 and
adjusted R2 and were all highly significant (online supple-
mentary Table S2).
The expected floristic dichotomy between Coastland and
Hinterland forests stood out with great clarity on the first
axis of DCA1 (Fig. 2a and b), although there was an addi-
tional discrimination of most Pampean semi-deciduous for-
ests towards the extreme left. The predictors selected for this
axis (online supplementary Table S3) were rainfall seasonal-
ity and cloud cover, both increasing towards the coast, and
shortest distance to the ocean, total base saturation of topsoil
and temperature annual range, all three increasing towards
the interior, the latter with extreme values at the Pampean
sites (Fig. 2a). Though selected, the aridity index was non-
significant. The model also included 14 significant and 11
non-significant MEMs. Despite the patent floristic dichotomy,
windward rain and cloud forests (Coastland sites) and leeward
seasonal forests (Hinterland sites) shared a considerable num-
ber of species: 816 out of 1541 (Jaccard index=52.9%). This
also meant that Coastland and Hinterland forests contained
86.6% and 66.3% of the total flora, respectively. On the other
hand, 519 of the 1335 Coastland forest species (38.9%) and
206 of the 1022 Hinterland forest species (20.2%) are appar-
ently restricted to them. The floristic dichotomy is therefore
chiefly related to species frequency and secondarily to species
composition (see indicator species in online supplementary
Table S4).
Distinctions among forest types within both Coastland
and Hinterland forests turned up in the second axis of DCA1
(Fig. 2a). On the right side (Coastland), cloud forests and
cloud dwarf-forests were discriminated, at the bottom half,
from rain forests, concentrated at the top. Likewise, on the
left side (Hinterland), Araucaria forests and Atlantic semi-
deciduous forests were concentrated on the bottom and
top halves, respectively, though there was some overlap in
between. Afurther discrimination turned up at the extreme
left between Pampean and Atlantic semi-deciduous forests.
The second axis, therefore, apparently displaced forest sites
of either higher altitude (Araucaria) or latitude (Pampean)
towards the bottom half of the diagram. The predictors
selected for this axis (online supplementary Table S3) were
mean annual temperature and potential evapotranspiration,
both increasing towards lower altitudes and northern lati-
tudes, and soil surface rockiness and altitude itself, increas-
ing towards higher altitudes only (Fig. 2a). The model also
included 15 significant and 10 non-significant spatial filters
(from connectivity matrices).
The third axis of DCA1 (Fig. 2b) provided additional dis-
criminations within the two main groups: on the right side,
Coastland sites were disposed in a clear sequence of eleva-
tion bands, with highland forests at the top, lowland forests
at the middle and coastal dwarf-forests at the bottom; on the
left side, Hinterland sites were disposed with a clear discrimi-
nation between Atlantic (Araucaria and semi-deciduous) and
Pampean (semi-deciduous) forests, and an additional discrim-
ination of Pampean coastal dwarf-forests at the bottom. The
predictors selected for the this axis (online supplementary
Table S3) were increasing altitude and decreasing minimum
monthly temperature away from both lowland coastal and
Pampean sites, and increasing topsoil salinity and sand frac-
tion towards the coastal dwarf-forests of both the Atlantic and
Pampean Domains (Fig. 2b). The model also included nine
significant and two non-significantMEMs.
The ordination diagram yielded by DCA2 for Coastland
forest sites only (Fig.3a) showed a clear discrimination of
forest types based essentially on altitudinal bands. Coastal
dwarf-forests and lowland and lower hill-land rain forests
are concentrated on the left side of the diagram, and only
a combination of the gradients summarized by both axes
allowed some distinction among the three forest types. This
sequence is followed towards the right by a large cluster of
upper hill-land rain forests, concentrated at the bottom cen-
tre of the diagram, and then by a cluster of lower and upper
hill-range cloud forests and dwarf-forests, at the bottom
right. The predictors selected for the first axis (online supple-
mentary Table S3) were frost frequency, soil surface rocki-
ness and soil drainage, all increasing towards the highlands,
and the maximum monthly temperature, the only increas-
ing towards the seashore (Fig.3a). The minimum monthly
temperature, though selected, was non-significant. The
model also included a single MEM. The predictors selected
for the second axis (online supplementary Table S3) were
topsoil salinity and sand fraction, both increasing towards
coastal dwarf-forests, and rainfall of the warmest quarter
and isothermality, both increasing towards hill-land rain for-
ests (Fig.3a). The predictor ‘total exchangeable bases of top-
soil’ was non-significant. Likewise, the model also included
a singleMEM.
The ordination diagram yielded by DCA3 for Hinterland
forest sites only (Fig.3b) also showed a discrimination of for-
est types across gradients. The first axis placed Araucaria for-
ests and semi-deciduous forests at the left and right halves of
the diagram, respectively, though it also showed some blend
at the centre, which suggests a continuum of species turnover
rather than two rather distinct floras. The predictors selected
for this axis (online supplementary Table S3) were frost fre-
quency and cloud cover, both increasing towards Araucaria for-
ests, and potential evapotranspiration and minimum monthly
temperature, both increasing towards semi-deciduous forests
(Fig.3b). The model also included six significant and 57 non-
significant spatial filters (from connectivity matrices). The sec-
ond axis displaced Pampean semi-deciduous forests towards
the top half of the diagram and, likewise, the gradual pattern
also suggests a continuum of species turnover. Contrasting to
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 249
coastal sandy (dwarf)lower hill-range
Evergreen rain lowlandEvergreen lower hill-range (dwarf)
forests: lower hill-landcloud forests: upper hill-range (dwarf)
upperhill-land lowland
lower tablelandSemi- deciduouslower tablelan
dA
tlantic
Evergreen seasonalupper tablelandseasonal forests: upper tableland
Araucaria forests: lower hill-range coastal sandy (dwarf)
upper hill-range riverine lowlan
dP
ampean
riverine lower tableland
0
100
200
Axis 3
0100 200 300 400
Axis 1
(b)
A
ltitude
Rainfall seasonality
Cloudcover
A
ridity
index
Temperature
annual range
Shortest distance
to the ocean
TBS
Topsoil salinity
Minimum mo nthly
temperature
Topsoil sand
fraction
TBS= Total bases saturation of to
p
soil
0
0
100 200 300 400
100
200
300
Axis 1
Axis 2
(a)
Mean annual
temperature
Rainfall seasonality
Cloudcover
PET
PET= Potential evapotranspiration
DIS = Shortest distance to the ocean
TBS= Total bases saturation of to
p
soil
DIS
TBS
Temperature
annual range
Soil surface rockines
Altitude
Figure2: Diagrams yielded by detrended correspondence analysis (DCA1) showing the ordination on the first three axes of the 484 sites of
Subtropical Atlantic and Pampean forests based on binary occurrence records of 1545 tree species. Diagrams show ordinations in the first ×
second axes (a) and first × third axes (b). The classification of sites into forest types is given by symbols and the significant environmental predic-
tors selected by a posteriori multiple regression models are shown as arrows pointing at the prevailing direction of increasing values and lengths
proportional to their correlations with ordination scores.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
250 Journal of Plant Ecology
Coastland forest sites, there was no clear distinction of alti-
tudinal bands among Hinterland forest sites. The predictors
selected for the second axis (online supplementary Table S3)
were cloud cover, actual evapotranspiration, soil drainage and
isothermality, all increasing from Pampean to Atlantic semi-
deciduous forests, and soil water deficit, the only increasing
0100 20
00
0
100
200
300
Axis 1
Axis 2
lower tableland
Evergreen seasonal upper tableland
Araucariaforests: lower hill-range
upper hill-range
lowland
Semi-deciduous lower tablelandAtlantic
seasonal forests: upper tableland
coastal sandy (dwarf)
riverine lowlandPampean
riverine lower tableland
(b)
Minimum mo nthly temperature
Actual evapotranspiration
Soil drainage
Isothermality
Soil water defici t
Frost frequency
Cloudcover
Potential evapotranspiration
0100 200300
0
50
100
150
200
Axis 1
Axis 2
coastal sandy (dwarf)
Evergreenrainlowland
forests: lower hill-land
upperhill-land
lower hill-range
Evergreenlower hill-range (dwarf)
cloud forests: upper hill-range (dwarf)
Topsoil sand fraction
Topsoil salinity
Soil surface rockiness
Soil drainage
Frost frequency
Isothermality
Maximum
monthly
temperature
RWM
RWM = Rainfall of the warmest quarter
Figure3: Diagrams yielded by detrended correspondence analyses showing the ordination on the first two axes of (a) the 160 sites of Coastland
Atlantic rain and cloud forests (DCA1) based on binary occurrence records of 1372 tree species and (b) the 331 sites of Hinterland Atlantic and
Pampean seasonal forests (DCA2) based on the binary occurrence records of 1002 species. The classification of sites into forest types is given by
symbols, and the significant environmental predictors selected by a posteriori multiple regression models are shown as arrows pointing at the
prevailing direction of increasing values and lengths proportional to their correlations with ordination scores.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 251
towards the Pampa (Fig.3b). The model also included seven
significant and five non-significantMEMs.
Bubble maps illustrate the variation across the forest
sites of four chosen environmental predictors among those
selected by the models (online supplementary Fig. S1). Mean
annual temperatures decreased towards both the south and
the hinterland highlands, while mean annual temperature
ranges increased away from the coast reaching the highest
values in the Pampa and western lowlands. The frequency
of frosts was only partially coherent with the mean annual
temperature as it showed a stronger relationship with alti-
tude, as the highest values were clearly coincident with the
ridge tops of the Serra do Mar and Serra Geral. The mean
annual rainfall seasonality increased towards both the north
and the coastland. It is worth stressing here that increased
rainfall seasonality in the whole region is caused by a much
higher precipitation volume in the summer compared with
that of the winter. High rainfall seasonality means, there-
fore, a period of excessive rain instead of a significant rainless
period. In fact, Walter climatic diagrams indicated no period
of effective water deficit in all studysites.
The total variance in tree species composition explained by
environmental and spatial predictors for the three data sets
(Subtropical, Coastland and Hinterland sites) varied between
26% and 34% leaving >65% of undetermined residuals in all
cases (Fig.4). Despite this, the variance explained by envi-
ronmental predictors was significant in all cases (Subtropical:
F22,360 = 1.9204; P = 0.005; Coastland: F13,137 = 3.1155;
P = 0.005; Hinterland: F16,262 = 1.3985; P = 0.005) as was
the variance explained by spatial predictors (Subtropical:
F101,360 = 1.5756; P = 0.005; Coastland: F9,137 = 1.4775;
P=0.005; Hinterland: F45,262=1.7194; P=0.005). In the case
of spatial predictors, the significance was found only for the
macro-spatial and not for the micro-spatial components. The
environmental predictors explained 25.8%, 24.0% and 20.4%
of the variance in Subtropical, Coastland and Hinterland sites,
respectively, while spatial predictors explained 30.6%, 12.6%
and 26.4% of the variance for the same sites. The respective
spatially structured environmental fractions (undifferenti-
ated) were 22.3%, 10.4% and 18.8%. The main contrasts
were then the much higher proportion of variance purely
explained by environmental predictors in Coastland sites and
the much higher proportions of variance explained by spatial
predictors (both purely and coupled with environmental pre-
dictors) in both Subtropical and Hinterlandsites.
Variations in species richness across latitude and
foresttypes
Rain and cloud forests, with their 1324 tree species, were by
far the richest among the five main groups of forest types,
while cloud and coastal dwarf-forests, with 308 and 505 spe-
cies, respectively, were the poorest ones (Fig.5a). Araucaria
forests and semi-deciduous forests produced intermediate
values: 714 and 844 species, respectively. (see online supple-
mentary Table S5, with a selection of indicator species of the
five main groups of forest types).
Subtropical Atlantic and Pampean forests showed a consist-
ent decline in tree species richness with increasing latitude
(Fig.5a), from 1568 species between 23–24°S to 98 species
between 34–35°S. There was a particularly steeper decline for
the much richer rain forests between 23–27°S and for semi-
deciduous forests between 23–25°S. Araucaria forests formed
a plateau with a slight southward decline up to 31°S and a
gap between 23–25°S, certainly because of the lack of sample
sites at these latitudes (see Fig.1). Asimilar pattern was pro-
duced for the considerably poorer cloud dwarf-forests, while
coastal dwarf-forests showed a very gentle decline across the
whole latitudinal range. Although the exclusively Atlantic
forest types (rain, cloud and Araucaria forests) disappear sud-
denly between 30–32°S, the decline in total species richness
showed no increased steepness and proceeded steadily across
the Pampean enclaves of semi-deciduous forests.
The model prepared for the variations in species richness
across 1°-latitude bands and the five main groups of forest
types (adjusted R2= 0.795; F2,41=76.284; P <0.001) ended
up with no spatial dependency and only two significant
environmental predictors: minimum monthly temperature
(β=0.775; standard error=10.599; t=10.287; P<0.001) and
Figure4: Proportion of variance in tree species composition explained
by environmental predictors only (a), spatially structured environ-
mental predictors (b) and spatial predictors (macro-scale) only (c), as
well as the proportion of unexplained residuals (d). Proportions are
given for three sets of forest sites: 484 Subtropical sites, 160 Coastland
sites and 331 Pampean sites.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
252 Journal of Plant Ecology
actual evapotranspiration (β=0.281; standard error=0.182;
t=3.735; P<0.001). It should be noted that variables pro-
duced to represent sampling intensity (total area covered by
the forest type at each latitude band and the number of for-
est sites per area) and environmental heterogeneity (standard
deviation of altitudes) were not selected for this model.
Species distribution ranges and subtropical
endemics
A large proportion of the southward decline of tree species
richness across subtropical Atlantic and Pampean forests was
certainly due to a decreasing number of species with a tropi-
cal–subtropical distribution range, because 1396 out of the
1541 species recorded in the 491 subtropical sites (90.6%)
were non-endemics, i.e. they were also recorded in the trop-
ics. This left the tree flora of subtropical Atlantic and Pampean
forests with 145 alleged subtropical endemic species only
(see online supplementary Table S6). Although the number
of subtropical endemic species was positively proportional to
the species richness of six groups of forest types (Fig.5b), the
proportion of endemics showed an opposite trend (Fig.5c),
with higher values in the species poorest forest types: the
Pampean semi-deciduous forests and the cloud and coastal
dwarf-forests.
The geographic range of the 1396 non-endemic tree spe-
cies shows that 657 are of strictly Atlantic distribution and
that another 228 are distributed across both Atlantic and
Amazonian forests, 47 of which through the caatinga ‘leap’
and 181 through the gallery forest ‘corridors’ of the Cerrado
Domain (Fig.6). This amounts to 885 species (57.4%) with
a predominantly ‘warm-moist’ distribution which is, by far,
the largest group in the species blend of subtropical forests.
A smaller but expressive group is that of the 262 species
(17.0%) with a predominantly ‘warm-seasonal’ distribu-
tion that are shared with the seasonally dry tropical forests
(SDTF) of the South American diagonal of open formations.
Another 49 species (3.2%), with a predominantly ‘cool-sem-
iarid’ distribution, are shared with the Chaco Domain, while
only 19 (1.2%), with a ‘cool-seasonal’ distribution, are shared
with Andean piedmont forests and show a disrupt distribu-
tion with the Chaco in between. The remaining 195 species
(12.7%) are versatile weedy trees occurring throughout the
Neotropics.
DISCUSSION
Forest types, distribution patterns and
environment
The a priori classification of subtropical Atlantic and Pampean
forests based essentially on vegetation physiognomy, climate
and elevation was highly consistent with variations in tree
species composition. As evidenced by the high proportions
of species shared by forest types, the floristic differentiation
among them results mainly from gradual changes of the spe-
cies blend, and species replacements are not concentrated at
Figure 5: (a) Number of tree species with increasing latitude in
tropical and subtropical Atlantic Pampean forests south of 23°, given
for the whole set 484 sites and five main groups of forest types. (b)
Relationship between number of species and number of Subtropical
endemic species in six main groups of forest types. (c) Relationship
between number of species and proportion of Subtropical endemic
species in six main groups of forest types. Contrasting to (a), semi-
deciduous forests were divided into Pampean and Atlantic in (b)
and (c).
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 253
transitional areas. Despite this, variations in species compo-
sition were strongly related to climatic variables even when
accounting for the effect of spatial proximity on both species
and climate. This spatial heterogeneity of both vegetation and
environment is certainly derived from the complex geomor-
phology of the region, centred by the Southern Brazilian pla-
teau, with its gently decreasing altitudes towards the Paraná
and Uruguay River valleys and sharp coastland slopes that rise
from sea level up to 2800 m asl (Bigarella 1991).
The steep ridge along the coast is certainly involved in the
top-level differentiation and sharpest transition between sub-
tropical Atlantic forests, which is that between Coastland (rain
and cloud) and Hinterland forests (seasonal Araucaria and semi-
deciduous), as they are placed on the wind and leeward sides of
the ridge, respectively, with cloud dwarf-forests appearing along
the rocky crests in-between. Orographic clouds are formed by
the forced uplift of humid air coming from the sea and promote
year-round abundant precipitation and water surplus on coast-
lands, though total rainfall is much higher during the summer.
Asmaller share of the oceanic fronts reach the hinterlands on the
leeward side, particularly in the summer, but both sides receive
additional precipitation from cold Antarctic fronts, particularly
in the winter, when oceanic fronts are less active (Roderjan etal.
2002). Because of this, there is no relevant period of water defi-
cit on both sides of the ridge which, in fact, show a wide varia-
tion in precipitation patterns. Nonetheless, water deficit is a key
factor determining the limits of the Atlantic Domain in the sub-
tropics because it meets the Chaco, Espinal and Pampa Domains
precisely where rainfall drops enough to produce at least a few
months of drought. In coherence with this, forest enclaves of
145
47
262
181
49
195
657
Tropical Atlantic forests
(warm-moist)
Seasonally dry forests
(warm-seasonal)
Chaco
(cool-semiarid)
A
ndean
Piedmont
forests
(cool-seasonal)
Neotropical
(generalists)
warm-seasonal (262)
cool-moist (145)
Caatinga
‘leap
Amazonianforests
(warm-moist)
gallery forest
‘corridors’
Chaco
‘leap
Atlantic forests
(sensu lato)
Endemic
(cool-moist)
warm-moist
(885)
cool-seasonal (19)
cool-semiarid (49)
generalists (195)
19
Figure6: Distribution of the 1555 tree species registered in subtropical Atlantic forests in other main Biogeographic Domains and their clas-
sification into six climatic groups.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
254 Journal of Plant Ecology
the Pampa Domain occur on sites where ground water accu-
mulates (Paz and Bassagoda 2002) and grassland enclaves occur
in the Atlantic Domain where ground water storage capacity is
deficient, although fire history interferes largely in both cases
(Behling etal. 2005; Behling and Pillar 2007).
As periods of water shortage in the subtropical section of the
Atlantic Domain are meaningful for vegetation only at forest–
grassland transitions, the primary factor underlying the contrast
between windward rainforests and leeward seasonal forests is
therefore temperature seasonality (Oliveira-Filho 2009), con-
trasting with the tropical section where rainfall seasonality is
the main factor underlying the differentiation between Atlantic
rain and seasonal forests (see Oliveira-Filho et al. 2006). Low
temperature extremes are reckoned as one the most important
factors determining stressful habitats marginal to Atlantic forest
vegetation (Scarano 2002) and should be thought, accordingly,
as a leading force shaping both the species blend and species
richness of Atlantic forests under the strong influence of lati-
tude, altitude and distance from the ocean (Oliveira-Filho and
Fontes 2000). Both daily and year-round temperature oscilla-
tion are lowered on coastlands by the rapid exchange of air
masses between land and sea and by the long overcast peri-
ods of the summer, so that coastlands typically enjoy higher
thermal stability than hinterlands, regardless of both latitude
and altitude (Leite 2002). This probably explains why altitude
was apparently more important than latitude in differentiating
rain forests among themselves, as variations are spatially much
sharper on coastlands than on hinterlands. This was clearly
reflected by the much higher proportion of variance explained
only by spatial predictors for Coastland sites than those yielded
for all sites and Hinterlandsites.
Regardless of altitude, increasing latitudes also increase tem-
perature variation, both along the year and among years, and
this explains why subtropical latitudes are so liable to outbreaks
of freezing temperatures, and the occurrence of frost generally
determines the limits between tropical and subtropical vegetation
worldwide (Box and Fujiwara 2005). In fact, frosts regularly hit
the region during the winter, normally sparing only the coastal
plains, and snowfall varies from erratic to regular on highlands.
Besides, it is not unusual to observe mass killings of particular
tree populations in certain areas struck by exceptionally harsh
frosts (Klein 1984; Rambo 1980). It was then no surprise that
tree species richness in the region showed the worldwide pat-
tern of decreasing richness with declining minimum tempera-
tures and actual evapotranspiration (Rambo 1961; Smith 1962;
Waechter and Jarenkow 2003). The latter, in particular, is one
of the best correlated variables with the primary productivity of
terrestrial ecosystems (Currie etal. 2004). The spatial distribution
of tree species richness in the region is therefore coherent with
both the species-energy and climate harshness hypotheses pro-
posed to explain the LDG (Pimm and Brown 2004).
Low temperature records are particularly important for plant
species distribution where they add to other stressful factors, as
happens in marginal forest habitats near mountain tops, where
the capture of light, water and nutrients is constrained by soil
shallowness and high nebulosity and forests typically show
lower statures (dwarfishness) and species richness (Roderjan
et al. 2002; Scarano 2002). Although cloud forests are often
considered an extension of rain forests worldwide (see Box and
Fujiwara 2005), subtropical Atlantic cloud dwarf-forests shared
a considerable number of species with both highlands rain for-
ests and Araucaria forests, and this is why some authors con-
sider them as a floristic transition between rain and Araucaria
forests (Falkenberg and Voltolini 1995; Rambo 1956).
On the other extreme, coastlines also represent marginal
forest habitats stressed by salt spray, wind exposure and the
low mineral fertility and water storage capacity of the sandy
substrate, although some are also situated on permanently
waterlogged sites and are additionally stressed by anoxia and
acidity (Falkenberg 1999; Waechter 1990). Coastal forests
therefore also show dwarfishness and impoverished species
composition, but their species blend is usually a fraction of
the nearby rain or seasonal forests. Towards their southern
end, in coastal Uruguay, they become additionally stressed
by much lower temperatures and incorporate ‘cool-arid’
elements, such as Prosopis nigra (Griseb.) Hieron., P. affinis
Spreng., Opuntia monacantha (Willd.) How. and Aspidosperma
quebrachoblanco Schltdl. (Paz and Bassagoda 2002).
Important classification systems (e.g. Leite 2002; Veloso
et al. 1991) consider Araucaria forests as a rain forest type
because of their predominantly evergreen character. The pre-
sent results, instead, reinforce the proposal of Oliveira-Filho
(2009) who included them within seasonal forests because
of the aforementioned top-level differentiation related to
higher temperature seasonality in the subtropical hinterlands.
In fact, the apparent homogeneity of these forests is chiefly
determined by the abundance of the evergreen A.angustifo-
lia, although the species blend includes assorted proportions
of evergreen, semi-deciduous and deciduous trees. In fact,
Araucaria forests are certainly better described as an extensive
‘mosaic’ of evergreen seasonal forests with a species composi-
tion highly influenced by latitude, altitude and the vicinity of
other forest types (Higuchi et al. 2012; Jarenkow and Budke
2009; Klein 1984, 1990; Rambo 1953).
The leading factor discriminating two types of seasonal for-
ests, Araucaria (evergreen) and semi-deciduous, probably relies
on the cool and warm extremes of their temperature range,
as suggested by the balanced influence of latitude and altitude
on the geographic limits of Araucaria forests (Backes 1999;
Hueck 1953). Those forests typically occur above much higher
altitudes in the tropics, usually >1000 m asl, than in the sub-
tropics, where they may reach altitudes as low as 426 m asl
at their southernmost reach (Longhi et al. 1996). Increasing
records of maximum temperature are probably a key factor
restricting the distribution of Araucaria forests towards lower
altitudes and latitudes, beyond which they are replaced by
semi-deciduous forests and their predominantly ‘warm-sea-
sonal’ flora. This certainly also explains why semi-deciduous
forests rarely reach higher altitudes in the subtropics as they
usually do in the tropics (see Oliveira-Filho and Fontes 2000).
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 255
In the opposite direction, decreasing records of minimum tem-
perature probably restrict the penetration of ‘warm-seasonal’
species into Araucaria forests and determine, to a great extent,
their predominantly east–west species gradient (Gonçalves and
Souza 2013; Jarenkow and Budke 2009). Another well-known
restriction to the distribution of Araucaria forests is related to
shallow soils, and this is probably why they are commonly
replaced by either grasslands or dwarf-forests in those habitats,
with fire probably playing the role of sharpening the forest–
grassland transition (Backes 1999; Pillar and Quadros 1997).
Semi-deciduous forests show the widest distribution range
among the types of subtropical Atlantic forest as they spread
farther south than rain and Araucaria forests to reach the
coastline and penetrate the Pampa Domain as galleries and
forest enclaves. Many authors classify the forests of the south-
ern extreme as deciduous forests, specifically those of the
Uruguay and Jacuí River basins (e.g. Leite 2002; Oliveira-Filho
etal. 2006; Veloso etal. 1991) and only those to the north as
semi-deciduous. We here classified the whole range as semi-
deciduous because canopy deciduousness very rarely drops
to <30% of the leaf mass, which is the criterion proposed by
Oliveira-Filho (2009) to classify seasonal forests as deciduous.
In addition, there was no evidence of a north–south floristic
differentiation of subtropical semi-deciduous forests across
the Atlantic Domain but only between these and the species-
poor forest enclaves of the Pampa Domain.
Species distribution ranges and endemism
The high proportion of non-endemic tree species (>90%) in
subtropical Atlantic forests reinforces the evidence that this
flora is composed mostly of tropical species capable of extend-
ing their distribution into the subtropics. In addition, the vast
majority of the non-endemics are shared with outer ‘warm-
moist’ and ‘warm-seasonal’ tropical forests; it can be appealing
to associate the fact to a high contribution of ‘immigrants’ to
the subtropical tree flora. Most studies postulate that, between
28000 and 11000year BP, including therefore the whole LGM
(ca. 21000–18 000year BP), the study region was predomi-
nantly covered by semiarid to seasonally dry grasslands and
that forests only started expanding over grasslands between
9000 and 8000 year BP, as the climate became increasingly
warmer and moister and eventually covered most of the region
(e.g. Behling 1995, 1997, 2002; Behling etal. 1997; Bolzon and
Marchiori 2002; de Oliveira etal. 2005). Nevertheless, the pro-
cess apparently was neither spatially synchronic nor unbro-
ken during the Holocene, and grassland fire, for instance, was
probably a main hindrance to forest expansion, particularly
during the dry episodes of the Holocene (Bauermann et al.
2008). There is no evidence however of any generalized for-
est withdrawal but of phases of forest regression and drastic
changes in species composition (Behling et al. 2001; Garcia
etal. 2004; Iriondo and Garcia 1993; Ledru etal. 2008).
The picture of a forest-free area being colonized by immi-
grants from outer sources through the expansion of their dis-
tribution range is, at least, too simplistic, as is the idea that all
endemic species survived past climatic harshness in forest ref-
uges. The most realistic picture is probably that of a continu-
ous spatial rearrangement of species distribution ranges as the
spatial configuration of their potential habitats changed along
with climatic fluctuations. If many wide-range forest species
may have indeed expanded their distribution into the region,
many others may have moved their entire range towards
the south, with a total withdrawal from northern locations,
and this may have occurred to some present-day subtropical
endemics, while others did survive in refuges, such as ripar-
ian forests (Oliveira et al. 2008). If an increasingly warmer
and moister climate is the main cause of the southward for-
est expansion during the Holocene, one could expect that
some species with a present-day predominantly southern dis-
tribution would be more abundant farther north during the
LGM, particularly those classified here into the ‘cool-moist’
group. This is precisely what has been found for species of the
Antarctic element of South American flora, which are true
relics of an austral (and cooler) Gondwanan connection and
are represented here by five genera: Araucaria, Podocarpus,
Weinmannia, Drimys and Gaultheria (Rambo 1951b; Waechter
2002). They all show a typical distribution that becomes
increasingly concentrated on highlands towards thenorth.
Pollen records indicate that, during the LGM, Araucaria
was widely distributed in Western and Southeastern Brazil,
often combined with Podocarpus, Weinmannia and Drimys
(Garcia etal. 2004; de Oliveira etal. 2005; Ledru 1993; Ledru
et al. 1998). In full contrast, the present-day distribution is
massively concentrated in the subtropics farther away from
much smaller tropical populations found only on upper high-
lands. In addition, pollen records from a meteor crater near
São Paulo and, therefore, in the middle of the present-day
distribution gap demonstrated that Araucaria forests estab-
lished and vanished there in a number of occasions through-
out the Quaternary (Ledru et al. 2008). Molecular studies
of Podocarpus also indi-cate a post-glacial massive retreat of
two beforehand widespread species that are now reduced
to small relic populations of P. lambertii Klotzsch ex Endl.,
in some isolated highlands up to Northeastern Brazil, and
P. sellowii Klotzsch ex Endl., scattered as far as Bolivia and
the Amazon, although both species form large populations in
Southern Brazil (Behling 1996; Ledru etal. 2007). Likewise,
a great number of tree species of the ‘cool-moist’ element
of Atlantic forests are presently restricted to highlands in
the tropics but found at various altitudes in the subtropics.
P. sellowii, Weinmannia paulliniifolia Pohl, Drimys brasiliensis
Miers and Laplacea fruticosa (Schrad.) Kobuski, e.g. are found
from mountain tops to near sea level in the subtropics, but
on highlands only in the tropics. Exclusive to highlands in
both realms include P.lambertii, A. angustifolia, Agarista euca-
lyptoides (Cham. & Schltdl.) G.Don and Myrceugenia bracteosa
(DC.) D.Legrand & Kausel. Certainly because of this, many
floristic features of most types of subtropical Atlantic forests
coincided with those cited by Gentry (1995) for tropical mon-
tane forests in the Andes and Central American, such as the
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
256 Journal of Plant Ecology
high contribution of Asteraceae, Melastomataceae, Lauraceae
and Solanaceae, and the abundance of some diagnostic gen-
era of Neotropical cloud forests referred to by Webster (1995),
such as Drymis, Hedyosmum, Weinmannia, Clethra, Podocarpus,
Meliosma, Ilex, Myrsine, Miconia, Prunus, Roupala and Cyathea.
After all this strong body of evidence, it is surprising that most
descriptions of the forest expansion during the Holocene are
focused in species migrations from both tropical rain and sea-
sonal forests (‘warm-moist’ and ‘warm-seasonal’ species) and
give little to no attention to the migration of ‘cool-moist’ or
‘montane’ species. Rambo (1953) mentioned that there was
an ‘old mountain flora’ with species of Myrceugenia, Myrcia,
Siphoneugena, Solanum, Ocotea, Agarista, Miconia, Clethra,
Symplocos, Ilex, Symphyopappus, Piptocarpha, Leandra and
Tibouchina and later described forest expansion on subtropi-
cal highlands as essential part of a further advance of previ-
ously established forests on surrounding lowlands and valleys
(Rambo 1961). Smith (1962), on the other hand, sought for
disrupted Andean connections that were recently pointed out
by Gonçalves and Souza (2013) for the upper montane flora
of eastern Araucaria forests.
In fact, our results strongly suggest that the tree flora of sub-
tropical forests is mostly composed of species with a tropical–
subtropical distribution that are capable of coping with outbursts
of freezing temperatures, and this includes not only the species
of the ‘cool-moist’ and ‘cool-seasonal’ groups but also the much
larger contingent of cool-tolerant species of the ‘warm-moist’
and ‘warm-seasonal’ groups. The latter would include some
wide-range Amazonian and/or Atlantic forest species which are
able to expand their distribution into areas of strongly seasonal
climates via gallery forests (Oliveira-Filho and Ratter 1995), but
a much larger number is apparently associated with coastland
rain forests areas between Rio de Janeiro and coastal Bahia
which are presumed to be historically more stable (Carnaval
and Moritz 2008). The higher thermal stability of subtropical
coastlands probably promoted higher rates of successful immi-
gration of ‘warm-moist’ than of ‘warm-seasonal’ species.
Subtropical Atlantic forests are replaced by open vegeta-
tion types (chaco, espinal or pampas) where total rainfall drops
enough to produce at least a couple of months with actual
drought. This reinforces the hypothetical picture of most tropi-
cal seasonal forest species migrating into the subtropics through
the inner lowlands of the Paraguay, Parana and Uruguay River
basins. The fact that many of them shed their leaves during
the cold season, when there is no meaningful water deficit,
could be no more than a preserved ecological trait (Leite and
Klein 1990; Marchioretto et al. 2007). The deciduousness of
species of the ‘warm-seasonal’ group is probably involved in
their capability to cope with both low temperature extremes
in the subtropics and water deficit periods in the tropics. The
role of temperature is also evident in another relevant element
of the subtropical non-endemic tree flora, the ‘cool-seasonal’,
characterized by an amphi-Chacoan disrupt distribution sensu
Waechter (2002), such as those of Schinus molle L., Ilex para-
guariensis A.St.-Hil., Cordia americana (L.) Gottschling & J.J.
Mill., Vasconcellea quercifolia A. St.-Hil., Cordyline spectabilis
Kunth & Bouché, Cedrela lilloi C.DC., Myrcianthes pseudomato
(D. Legrand) McVaugh, Myrrhinium atropurpureum Schott and
Diatenopteryx sorbifolia Radlk. As postulated by Smith (1962)
and later on by Gonçalves and Souza (2013), a number of
Andean–Atlantic distributions were probably disconnected by
late expansions of aridity in the Chaco Domain.
Both the ‘warm-’ and ‘cool-seasonal’ species groups con-
tributing to subtropical Atlantic forests are also a fraction of
the so-called SDTF flora which is distributed along a circum-
Amazonian arch that connects the caatingas in NE Brazil to
coastal Caribbean dry forests (see Pennington etal. 2009). Their
contribution to subtropical Atlantic forests is naturally concen-
trated in the two types of seasonal forests, i.e. semi-decidu-
ous and Araucaria forests (Oliveira-Filho etal. 2006; Spichiger
etal. 2004, 2006). This probably explains why some authors
decided to split them from coastland rain forests to integrate a
distinct phytogeographic unit (e.g. Cabrera and Willink 1973;
Dinerstein etal. 1995; Morrone 1999, 2001). This view is some-
what coherent with the presumed history of forest expansion
during the Holocene. Rambo (1961) recognized two main
migration corridors for the tropical flora towards the south: the
western, through the Paraná and Uruguay River valleys, with
a mesophyllous (seasonal) character, and the eastern, along
the Atlantic coast, with a hygrophyllous (moist) character. He
argued that the former reached the south long before the lat-
ter, and this is in full agreement with the pollen record that
indicates the presence of dense forests on the Argentine plains
ca. 8000year BP (Iriondo and Garcia 1993), while coastal rain
forests were not found in the subtropics before 7000year BP
(Ledru etal. 2008). The subsequent expansion of rain forests,
and their ‘warm-moist’ species, along the coast could have
pushed back a number of ‘warm-seasonal’ species towards the
interior, and, since then, seasonal forests concentrated in hin-
terlands although some typical ‘moist’ species penetrated deep
into the continent (Behling et al. 2005; Behling and Negrelle
2001; Lorscheitter 2003).
Despite the above-mentioned dichotomy, hinterland sea-
sonal forests and coastland rain forests share a much higher
proportion of species in the subtropics than in the tropics
(Oliveira-Filho etal. 2006), and, because of this, there is lit-
tle ground to restrict the Atlantic Domain to coastland rain
forests. Aspatial rearrangement of species distribution ranges
in connection with the changing shapes of their potential
habitats is probably a more realistic picture of the Holocene
forest expansion than that based on two species migration
flows. The formation of the species blends therefore would
depend on tradeoffs between the chances of successful migra-
tion into potential habitats and local extinction in the face
of environmental harshness and antagonistic interactions.
The ‘arid element’ of subtropical Atlantic forests is illustrative
here. Typical chaco species, such as Jodina rhombifolia (Hook.
& Arn.) Reissek, Sideroxylon obtusifolium (Roem. & Schult.)
T.D. Penn., Scutia buxifolia Reissek, Colletia paradoxa (Spreng.)
Escal. and Schinus polygamus (Cav.) Cabrera, are found in
stressful habitats of the Atlantic Domain, such as rocky out-
crops and coastal sand deposits, and a few others, like Berberis
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 257
laurina Billb., are also found in Araucaria and cloud forests
(Waechter 2002). On the other extreme, typical rain forest
species, as the palmetto Euterpe edulis Mart., may be found as
far inland as the Moconá Park in Argentina, locally favoured
by the high moisture of riparian habitats (Daviña etal. 1999).
SUPPLEMENTARYDATA
Supplementary material is available online at Journal of Plant
Ecology online.
FUNDING
Brazilian government research agencies Conselho Nacional
de Desenvolvimento Científixo e Tecnológico (CNPq; grant
number 301644/88-8); Royal Botanic Gardens Kew for the
Kew Latin America Research Fellowship; New York Botanical
Garden.
ACKNOWLEDGMENTS
The first author thanks the Brazilian government research agencies
CNPq, for research grant no.301644/88-8, and FAPEMIG for funding
the participation in Association for Tropical Biology and Conservation
(ATBC) and Society for Tropical Ecology (gtò) 2009 Joint Meeting
at Philipps-Universität Marburg, Germany, when a lecture was given
on the present paper. We also thank the Royal Botanic Gardens Kew
for the Kew Latin America Research Fellowship that funded the first
revision of the database, in 2006, and both the Conselho Nacional
de Desenvolvimento Científixo e Tecnológico (CNPq) and the New
York Botanical Garden for funding and supporting the second revi-
sion, in 2012. We are indebted to the two anonymous reviewers and
the Associate Editor for their valuable contribution to the manuscript.
Conflict of interest statement. None declared.
REFERENCES
Almeida-Neto M, Campassi F, Galetti M, etal. (2008) Macroecological
correlates of vertebrate-dispersal syndromes along the Atlantic for-
est. Global Ecol Biogeogr 17:503–13.
Backes A (1999) Condicionamento climático e distribuição geográfica
de Araucaria angustifolia (Bertol.) Kuntze no Brasil – II. Iheringia,
Série Botânica 49:31–51.
Bauermann SG, Macedo RB, Behling H, et al. (2008) Dinâmicas
vegetacionais, climáticas e do fogo com base em palinologia e
análise multivariada no Quaternário tardio do sul do Brasil. Revista
Brasileira de Paleontologia 11:87–96.
Behling H (1995) Investigations into the late Pleistocene and
Holocene history of vegetation and climate in Santa Catarina (S.
Brazil). Vegetat Hist Archaeobot 4:127–52.
Behling H (1996) First report on new evidence for the occurrence
of Podocarpus and possible human presence at the mouth of the
Amazon during the Late-glacial. Vegetat Hist Archaeobot 5:241–6.
Behling H (1997) Late Quaternary vegetation, climate and fire his-
tory of the Araucaria forest and campos region from Serra Campos
Gerais, Paraná State (south Brazil). Rev Palaeobot Palynol 97:109–21.
Behling H (2002) South and southeast Brazilian grasslands during
Late Quaternary times: a synthesis. Palaeogeogr, Palaeoclimatol,
Palaeoecol 177:19–27.
Behling H, Negrelle RRB (2001) Tropical rain forest and climate
dynamics of the Atlantic lowland, Southern Brazil, during the late
Quaternary. Quat Res 56:383–9.
Behling H, Pillar VD (2007) Late Quaternary vegetation, biodiver-
sity and fire dynamics on the southern Brazilian highland and
their implication for conservation and management of modern
Araucaria forest and grassland ecosystems. Philos Trans R Soc Lond B
Biol Sci 362:243–51.
Behling H, Pillar VP (2008) Vegetation and fire dynamics in southern
Brazil during the late Quaternary and their implication for con-
servation and management of modern grassland ecosystems. In
Schröder HG (ed). Grasslands: Ecology, Management and Restoration.
Hauppauge, NY: Nova Science Publishers, 181–94.
Behling H, Negrelle RRB, Colinvaux PA (1997) Modern pollen rain
data from the tropical Atlantic rain forest, Reserva Volta Velha,
South Brazil. Rev Palaeobot Palynol 97:287–99.
Behling H, Bauermann SGP, Neves CP (2001) Holocene environmen-
tal changes in the São Francisco de Paula region, southern Brazil. J
South Am Earth Sci 14:631–9.
Behling H, Pillar VP, Bauermann SG (2005) Late Quaternary grass-
land (Campos), gallery forest, fire and climate dynamics, studied
by pollen, charcoal and multivariate analysis of the São Francisco
de Assis core in western Rio Grande do Sul (southern Brazil). Rev
Palaeobot Palynol 133:235–48.
Bigarella JJ (1991) Aspectos físicos da paisagem. In Fundação S.O.S.
Mata Atlântica (ed). Mata Atlântica/Atlantic Rain Forest. São Paulo,
Brazil: Editora Index, 62–93.
Bivand R (2013) Package ‘spdep’: Spatial Dependence: Weighting Schemes,
Statistics and Models. R package version 0.5–56. http://cran.r-project.
org/web/packages/spdep/index.html (15 October 2012, date last
accessed).
Bolzon RT, Marchiori JNC (2002) A vegetação no sul da América –
Perspectiva paleoflorística. Ciência e Ambiente 24:5–24.
Box EO, Fujiwara K (2005) Vegetation types and their broad-scale
distribution. In van der Maarel E (ed). Vegetation Ecology. Maiden,
MA: Blackwell Science, 107–28.
Burnham KP, Anderson DR (2002) Model Selection and Multi-Model
Inference, a Practical Information-Theoretic Approach, 2nd edn. New
York, NY: Springer.
Cabrera AL, Willink A (1973) Biogeografia de America Latina.
Washington, DC: Secretaria General de la Organización de los
Estados Americanos.
Carnaval AC, Moritz C (2008) Historical climate modelling predicts
patterns of current biodiversity in the Brazilian Atlantic forest. J
Biogeogr 35:1187–201.
CGIAR-CSI (2006) NASA Shuttle Radar Topographic Mission (SRTM)
(data set available as 3 arc-second DEMs). http://srtm.csi.cgiar.org/
(12 August 2012, date last accessed).
Costa LP, Leite YLR, Fonseca GAB, et al. (2000) Biogeography of
South American forest mammals: endemism and diversity in the
Atlantic forest. Biotropica 32:872–81.
Currie DJ, Mittelbach GG, Cornell HV, etal. (2004) A critical review of
species-energy theory. Ecol Lett 7:1121–34.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
258 Journal of Plant Ecology
Daviña JR, Rodríguez ME, Honfi AI, etal. (1999) Floristic studies of
the Moconá Park, Misiones, Argentina. Candollea 54:231–49.
de Oliveira PE, Behling H, Ledru M-P, etal. (2005) Paleovegetação
e paleoclimas do Quaternário do Brasil. In Souza CRG, Seguio
K, Oliveira AMS, etal. (eds). Quaternário do Brasil. Ribeirão Preto,
Brazil: Holos, 52–74.
Dinerstein E, Olson DM, Graham DJ, et al. (1995) A Conservation
Assessment of the Terrestrial Ecoregions of Latin America and the
Caribbean. Washington DC: World Wildlife Fund and the World
Bank.
Diniz-Filho JAF, Rangel TFLVB, Bini LM (2008) Model selection and
information theory in geographical ecology. Global Ecol Biogeogr
17:479–88.
Diniz-Filho JAF, Bini LM (2005) Modelling geographical patterns in
species richness using eigenvector-based spatial filters. Global Ecol
Biogeogr 14:177–85.
Dray S, Legendre P, Peres-Neto P (2006) Spatial modeling: a compre-
hensive framework for principal coordinate analysis of neighbor
matrices (PCNM). Ecol Model 196:483–93.
Dray S, Legendre P, Blanchet FG (2009) Packfor: Forward Selection
with Permutation (Canoco p.46) (R package version 0.0–7⁄ r58).
http://R-Forge.R-project.org/projects/sedar/ (15 October 2012,
date last accessed).
Dray S, Pélissier R, Couteron P, et al. (2012) Community ecology
in the age of multivariate multiscale spatial analysis. Ecol Monogr
82:257–5.
Eisenlohr PV (2013) Challenges in data analysis: pitfalls and sug-
gestions for a statistical routine in vegetation ecology. Braz J Bot
36:83–87.
Fortin M-J, Dale MRT (2005) Spatial Analysis. AGuide for Ecologists.
Cambridge, UK: Cambridge University Press.
Falkenberg DB (1999) Aspectos da flora e da vegetação secundária da
restinga de Santa Catarina, Sul do Brasil. Insula 28:1–30.
Falkenberg DB, Voltolini JC (1995) The montane cloud forest in
Southern Brazil. In Hamilton LS, Juvik JO, Scatena FN (eds). Tropical
Montane Cloud Forest. New York, NY: Springer-Verlag, 138–49.
FAO/IIASA/ISRIC/ISSCAS/JRC (2012) Harmonized World Soil Database
(version 1.2). Rome, Italy: FAO.
Fernandes A, Bezerra P (1990) Estudo fitogeográfico do Brasil. Fortaleza,
Brazil: Stylus Comunicações.
Ferraz EMN, Araújo EL, Silva SI (2004) Floristic similarities
between lowland and montane areas of Atlantic Coastal Forest in
Northeastern Brazil. Plant Ecol 174:59–70.
Galetti M, Giacomini HC, Bueno RS, etal. (2009) Priority areas for the con-
servation of Atlantic forest large mammals. Biol Conserv 142:1229–41.
Galindo-Leal C, Câmara IG (2003) Atlantic Forest hotspot status: an
overview. In Galindo-Leal C, Câmara IG (eds). The Atlantic Forest
of South America. Washington, DC: Center for Applied Biodiversity
Science, 3–11.
Garcia MJ, de Oliveira PE, Saraiva R, etal. (2004) A Holocene vegeta-
tional and climatic record from the Atlantic rainforest belt of coastal
State of São Paulo, SE Brazil. Rev Palaeobot Palynol 99:181–99.
Gentry AH (1995) Patterns of diversity and floristic composition in
Neotropical montane forests. In Churchill SP, Balslev H, Forero E,
etal. (eds). Biodiversity and Conservation of Neotropical Montane Forests.
Neotropical Montane Forest Biodiversity and Conservation Symposium, 1.
New York, NY: The New York Botanical Garden, 103–26.
Gonçalves ET, Souza AF (2013) Floristic variation in ecotonal areas:
patterns, determinants and biogeographic origins of subtropical
forests in South America. Austral Ecol 38: doi:10.1111/aec.12051.
Grazziotin FG, Monzel M, Echeverrigaray S, et al. (2006)
Phylogeography of the Bothrops jararaca complex (Serpentes:
Viperidae): past fragmentation and island colonization in the
Brazilian Atlantic Forest. Mol Ecol 15:3969–82.
Griffith D (2003) Spatial Autocorrelation and Spatial Filtering: Gaining
Understanding Through Theory and Scientific Visualization. Berlin,
Germany: Springer-Verlag.
Harris GM, Jenkins CN, Pimm SL (2005) Refining biodiversity conser-
vation priorities. Conserv Biol 19:1957–68.
Higuchi P, Silva AC, Ferreira TS, etal. (2012) Floristic composition and
phytogeography of the tree component of Araucaria Forest frag-
ments in southern Brazil. Braz J Bot 35:145–57.
Hill MO, Gauch HG Jr (1980) Detrended Correspondence analysis: an
improved ordination technique. Vegetatio 42:47–58.
Hillebrand H (2004) On the generality of the latitudinal diversity gra-
dient. Am Nat 163:192–211.
Hijmans RJ, Cameron SE, Parra JL, etal. (2005) Very high resolution
interpolated climate surfaces for global land areas. Inter J Climatol
25:1965–78.
Hueck K (1953) Distribuição e habitat natural do pinheiro-brasileiro
(Araucaria angustifolia). Boletim da Faculdade de Filosofia, Ciências e
Letras da Universidade de São Paulo (Bot. n.10)156:4–24.
IBGE (2012) Manual Técnico da Vegetação Brasileira, 2nd edn. Manuais
Técnicos em Geociências, 1.Rio de Janeiro, RJ, Brasil: Instituto
Brasileiro de Geografia e Estatística (IBGE).
Iriondo MH, Garcia MO (1993) Climatic variations in the Argentine
plains during the last 18,000 years. Palaeogeogr, Palaeoclimatol,
Palaeoecol 101:209–20.
Jarenkow JA, Budke JC (2009) Padrões florísticos e análise estrutural
de remanescentes florestais com Araucaria angustifolia no Brasil.
In Fonseca CSD, Souza AF, Zanchet AML, et al. (eds). Floresta
com Araucária: Ecologia, Conservação e Desenvolvimento Sustentável.
Ribeirão Preto, SP, Brazil: Holos, 54–83.
Jones P, Harris I (2008) CRU Time Series (TS) High Resolution Gridded
Datasets. London: University of East Anglia Climate Research Unit (CRU);
NCAS British Atmospheric Data Centre, UK. http://badc.nerc.ac.uk/
view/badc.nerc.ac.uk__ATOM__dataent_1256223773328276 (23
November 2010, date last accessed).
Kent M (2011) Vegetation Description and Analysis, a Practical Approach.
Oxford, UK: Wiley-Blackwell.
Klein RM (1984) Aspectos dinâmicos da vegetação do sul do Brasil.
Sellowia 36:5–54.
Klein RM (1990) Os tipos florestais com Araucária em Santa
Catarina. In Anais do 36° Congresso Brasileiro de Botânica (Curitiba,
1985). Brasília, DF, Brazil: Instituto Brasileiro do Meio Ambiente
(IBAMA), 101–19.
Ledru MP (1993) Late Quaternary environmental and climatic
changes in central Brazil. Quat Res 39:90–98.
Ledru MP, Salgado-Labouriau ML, Lorscheiter ML (1998) Vegetation
dynamics in southern and central Brazil during the last 10,000 yr.
B.P. Rev Palaeobot Palynol 99:131–42.
Ledru MP, Salatino A, Salatino MLF, etal. (2007) Regional assessment
of the impact of climatic change on the distribution of a tropical
conifer in the lowlands of South America. Divers Distribut 13:767–71.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
Oliveira-Filho etal. | Tree species of subtropical South American forests 259
Ledru MP, Mourguiart P, Riccomini C (2008) Related changes in bio-
diversity, insolation and climate in the Atlantic rainforest since the
last interglacial. Palaeogeogr, Palaeoclimatol, Palaeoecol 271:140–52.
Legendre P, Borcard D, Roberts DW (2012) Variation partitioning involv-
ing orthogonal spatial eigenfunction submodels. Ecology 93:1234–40.
Leite PF (2002) Contribuição ao conhecimento fitoecológico do sul do
Brasil. Ciência Ambiental 24:51–73.
Leite PF, Klein RM (1990) Vegetação. In Instituto Brasileiro de
Geografia e Estatística (ed). Geografia do Brasil: Região Sul, Vol. 2.
Rio de Janeiro, RJ, Brazil: 113–50.
Longhi SJ, Vaccaro S, Coelho MCB, etal. (1996) Análise fitossoci-
ológica de um remanescente de Floresta Ombrófila Mista em
Itaára, Santa Maria-RS. In Simpósio sobre ecossistemas naturais do
Mercosul: O ambiente da Floresta, Vol. 1. Santa Maria, RS, Brazil:
CEPEF/UFSM, 79–89.
Lopes AV, Girão LC, Santos BA, etal. (2009) Long-term erosion of tree
reproductive trait diversity in edge-dominated Atlantic forest frag-
ments. Biol Conserv 142:1154–65.
Lorscheitter ML (2003) Contribution to the Holocene history of
Atlantic rain Forest in the Rio Grande do Sul state, southern Brazil.
Revista del Museo Argentino de Ciencias Naturales 5:261–71.
Marchioretto MS, Mauhs J, Budke JC (2007) Fenologia de espécies
arbóreas zoocóricas em uma floresta psamófila no sul do Brasil.
Acta Botanica Brasilica 21:193–201.
Marques MCM, Swaine MD, Liebsch D (2011) Diversity distribu-
tion and floristic differentiation of the coastal lowland vegetation:
implications for the conservation of the Brazilian Atlantic forest.
Biodivers Conserv 20:153–68.
Metzger JP (2009) Conservation issues in the Brazilian Atlantic for-
est. Biol Conserv 142:1138–40.
Metzger JP, Martensen AC, Dixo M, etal. (2009) Time-lag in biological
responses to landscape changes in a highly dynamic Atlantic forest
region. Biol Conserv 142:1166–77.
McCune B, Grace JB (2002) Analysis of Ecological Communities.
Gleneden Beach, OR: MjM Software Design.
McCune B, Mefford MJ (2006) PC-ORD, Multivariate analysis of ecologi-
cal data, Version 5.10. Gleneden Beach, OR: MjM Software Design.
Morrone JJ (1999) Presentación preliminar de un nuevo esquema
biogeográfico de América del Sur. Biogeographica 75:1–16.
Morrone JJ (2001) The Parana Subregion and Its Provinces. Buenos
Aires, Argentina: Physis.
Murray-Smith C, Brummitt NA, Oliveira-Filho AT, etal. (2008) Plant
diversity hotspots in the Atlantic coastal forests of Brazil. Conserv
Biol 23:151–63.
Myers N, Mittermeier RA, Mittermeier CG, etal. (2000) Biodiversity
hotspots for conservation priorities. Nature 403:853–8.
Negrelle RRB (1998) Fitodiversidade do componente arbóreo da
floresta Atlântica: uma análise preliminar. Revista de Tecnologia e
Ambiente 4:39–4.
Oksanen J, Blanchet FG, Kindt R, et al. (2011) Vegan: Community
Ecology Package. R package version 1.17–11. http://vegan.r-force.r-
project.org/ (15 October 2012, date last accessed).
Oksanen J, Blanchet FG, Kindt R, et al. (2013) Vegan: Community
Ecology Package. http://cran.r-project.org, http://vegan.r-forge.
rproject. org/ (12 October 2012, date last accessed).
Oliveira MAT, Behling H, Pessenda LCR (2008) Late-Pleistocene and
mid-Holocene environmental changes in highland valley head
areas of Santa Catarina state, Southern Brazil. J South Am Earth
Sci 26:55–67.
Oliveira-Filho AT (2009) Classificação das fitofisionomias da América
do Sul extra-Andina: proposta de um novo sistema – prático e
flexível – ou uma injeção a mais de caos? Rodriguésia 60:237–58.
Oliveira-Filho AT, Eisenlohr PV (2012) O banco de dados Treeatlan e
a classificação da vegetação da América do Sul tropical e subtropi-
cal. In Sociedade Botânica do Brasil (ed). 63º Congresso Brasileiro de
Botânica, Botânica Frente às Mudanças Globais. Joinville, SC, Brazil:
35–8.
Oliveira-Filho AT, Fontes MAL (2000) Patterns of floristic differentia-
tion among Atlantic forests in south-eastern Brazil, and the influ-
ence of climate. Biotropica 32:793–810.
Oliveira-Filho AT, Ratter JA (1995) A study of the origin of central
Brazilian forests by the analysis of plant species distribution pat-
terns. Edinb J Bot 52:141–94.
Oliveira-Filho AT, Tameirão-Neto E, Carvalho WAC et al. (2005)
Análise florística do compartimento arbóreo de áreas de Floresta
Atlântica sensu lato na região das Bacias do Leste (Bahia, Minas
Gerais, Espírito Santo e Rio de Janeiro). Rodriguésia 56:185–235.
Oliveira-Filho AT, Jarenkow JA, Rodal MJN (2006) Floristic rela-
tionships of seasonally dry forests of eastern South America
based on tree species distribution patterns. In Pennington RT,
Ratter JA, Lewis GP (eds). Neotropical Savannas and Dry Forests:
Plant Diversity, Biogeography and Conservation. Boca Raton, FL:
CRC Press, 159–92.
Pardini R, Faria D, Accacio GM, Laps RR, etal. (2009) The challenge of
maintaining Atlantic forest biodiversity: a multi-taxa conservation
assessment of specialist and generalist species in an agro-forestry
mosaic in southern Bahia. Biol Conserv 142:1178–90.
Paz EA, Bassagoda MJ (2002) Los bosques naturales del Uruguay -
Tipos y composición. Ciência e Ambiente 24:35–50.
Pennington RT, Lavin M, Oliveira-Filho AT (2009) Woody plant
diversity, evolution and ecology in the tropics: perspectives
from seasonally dry tropical forests. Ann Rev Ecol, Evolut Syst
40:437–57.
Pillar VP, Quadros F (1997) Grasslands-forest boundaries in southern
Brazil. Coenoses 12:119–26.
Pimm SL, Brown JH (2004) Ecology. Domains of diversity. Science
304:831–3.
Quadros FLF, Pillar VP (2002) Transições floresta–campo no Rio
Grande do Sul. Ciência e Ambiente 24:109–18.
Quinn GP, Keough MJ (2002) Experimental Design and Data Analysis for
Biologists. Cambridge, UK: Cambridge University Press.
R Development Core Team (2012) R: ALanguage and Environment for
Statistical Computing. Vienna, Austria: The R Foundation for Statistical
Computing. http://www.R-project.org/ (15 October 2012, date last
accessed).
Rambo B (1951a) A imigração da selva higrófila no Rio Grande do
Sul. Anais de Botânica do Herbário Barbosa Rodrigues 3:55–91.
Rambo B (1951b) O elemento andino no pinhal riograndense. Anais
de Botânica do Herbário Barbosa Rodrigues 3:7–39.
Rambo B (1953) História da flora do planalto riograndense. Anais de
Botânica do Herbário Barbosa Rodrigues 5:185–232.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
260 Journal of Plant Ecology
Rambo B (1956) A flora fanerogâmica dos Aparados riograndenses.
Sellowia, 7:235–98.
Rambo B (1961) Migration routes of the South Brazilian rain forest.
Pesquisas, Série Botânica 12:1–54.
Rambo B (1980) A mata pluvial do Alto Uruguai. Roessléria 3:101–39.
Rangel TF, Diniz-Filho JAF, Bini LM (2010) SAM: Acomprehensive
application for spatial analysis in macroecology. Ecography 33:1–5.
Ribeiro MC, Metzger JP, Martensen AC, et al. (2009) The Brazilian
Atlantic forest: how much is left, and how is the remaining forest dis-
tributed? Implications for conservation. Biol Conserv 142:1141–53.
Roderjan CV, Galvão F, Kuniyoshi YS, etal. (2002) As unidades fitoge-
ográficas do Estado do Paraná. Ciência e Ambiente 24:75–92.
Salis SM, Shepherd GJ, Joly CA (1995) Floristic comparison of mes-
ophytic semideciduous forests of the interior of the state of São
Paulo, Southeast Brazil. Vegetatio 119:155–64.
Santos AMM, Cavalcanti DR, Silva JMC, etal. (2007) Biogeographical
relationships among tropical forests in north-eastern Brazil. J
Biogeogr 34:437–46.
Scarano FR (2002) Structure, function and floristic relationships of
plant communities in stressful habitats marginal to the Brazilian
Atlantic rainforest. Ann Bot 90:517–24.
Scarano FR (2009) Plant communities at the periphery of the Atlantic
rain forest: Rare-species bias and its risks for conservation. Biol
Conserv 142:1201–08.
Scudeller VV, Martins FR, Shepherd GJ (2001) Distribution and abun-
dance of arboreal species in the atlantic ombrophilous dense forest
in southeastern Brazil. Plant Ecol 152:185–99.
Silva JMC, Sousa MC, Casteleti CHM (2004) Areas of endemism for
passerine birds in the Atlantic forest, South America. Global Ecol
Biogeogr 13:85–92.
Smith LB (1962) Origins of the flora of southern Brazil. Cont US Nat
Herbarium 35:215–49.
Spichiger R, Palese R, Chautems A, et al. (1995) Origin, affinities and
diversity hotspot of the Paraguayan dendrofloras. Candollea 50:515–37.
Spichiger R, Calange C, Bise B (2004) Geographical zonation in the
Neotropics of tree species characteristic of the Paraguay-Paraná
Basin. J Biogeogr 31:1489–501.
Spichiger R, Bise B, Calenge C, etal. (2006) Biogeography of the for-
ests of the Paraguay-Paraná Basin. In Pennington RT, Ratter JA,
Lewis GP (eds). Neotropical Savannas and Dry Forests: Plant Diversity,
Biogeography and Conservation. Boca Raton, FL: CRC Press, 185–203.
Stefenon VM, Behling H, Gailing O, etal. (2008) Evidences of delayed size
recovery in Araucaria angustifolia populations after post-glacial coloniza-
tion of highlands in Southeastern Brazil. An Acad Bras Cienc 80:433–43.
ter Braak CJF (1995) Ordination. In Jongman RHG, ter Braak CJF,
van Tongeren OFR (eds). Data Analysis in Community and Landscape
Ecology. Cambridge, UK: Cambridge University Press, 91–173.
Thomas WW (2008) Introduction and acknowledgements. In Thomas
WW (ed). The Atlantic Coastal Forest of Northeastern Brazil. New York,
NY: The New York Botanical Garden Press, 1–5.
Tichý L, Chytrý M (2006) Statistical determination of diagnostic spe-
cies for site groups of unequal size. J Vegetation Sci 17:809–18.
Torres RB, Martins FR, Kinoshita LS (1997) Climate, soil and tree
flora relationships in forests in the state of São Paulo, southeastern
Brazil. Revista Brasileira de Botânica 20:41–9.
Tyler H, Brown KSJ, Wilson K (1994) Swallowtail Butterflies of
the Americas. A Study in Biological Dynamics, Ecological Diversity,
Biosystematics and Conservation. Gainesville, FL: Scientific
Publishers.
Uehara-Prado M, Fernandes JO, Bello AM, etal. (2009) Selecting ter-
restrial arthropods as indicators of small-scale disturbance: afirst
approach in the Brazilian Atlantic forest. Biol Conserv 142:1220–28.
Veloso HP, Rangel Filho ALR, Lima JCA (1991) Classificação da
Vegetação Brasileira, Adaptada a um Sistema Universal. Rio de Janeiro,
RJ, Brazil: Instituto Brasileiro de Geografia e Estatística, IBGE.
Vieira MV, Olifiers N, Delciellos AC, etal. (2009) Land use vs. frag-
ment size and isolation as determinants of small mammal com-
position and richness in Atlantic forest remnants. Biol Conserv
142:1191–200.
Waechter JL (1990) Comunidades vegetais das restingas do Rio
Grande do Sul. In Anais do 2° Simpósio de Ecossistemas da Costa Sul
e Sudeste Brasileira; Estrutura, Função e Manejo, Vol. 3 (Águas de
Lindóia, 1990). São Paulo, SP, Brazil: Academia de Ciências do
Estado de São Paulo (ACIESP), 228–48.
Waechter JL (2002) Padrões geográficos na flora atual do Rio Grande
do Sul. Ciência e Ambiente 24:93–108.
Waechter JL, Jarenkow JA (2003) Padrões geográficos como evidên-
cias de processos dinâmicos em florestas brasileiras. In Claudino-
Sales W (ed). Ecossistemas Brasileiros: Manejo e Conservação. Fortaleza,
CE, Brazil: Expressão Gráfica Editora, 217–26.
Walter H (1985) Vegetation of the Earth and Ecological Systems of the Geo-
biosphere, 3rd edn. Berlin, Germany: Springer-Verlag.
Webster GL (1995) The panorama of Neotropical cloud forests.
1. In Churchill SP, Balslev H, Forero, etal. (eds). Biodiversity and
Conservation of Neotropical Montane Forests. Neotropical Montane Forest
Biodiversity and Conservation Symposium. New York, NY: New York
Botanical Garden, 53–77.
Willig MR, Kaufmann DM, Stevens RD (2003) Latitudinal gradients
of biodiversity: pattern, process, scale and synthesis. Ann Rev Ecol
Syst 34:273–309.
Zomer RJ, Trabucco A, van Straaten O, et al. (2006) Carbon, Land
and Water: AGlobal Analysis of the Hydrologic Dimensions of Climate
Change Mitigation Through Afforestation/Reforestation. IWMI Research
Report 101. Colombo, Sri Lanka: International Water Management
Institute.
by guest on August 26, 2015http://jpe.oxfordjournals.org/Downloaded from
... conditions are small subsets of the flora also present in warmer and moister tropical forests (Oliveira-Filho et al., 2013), which are richer in species and endemisms (Lima et al., 2020). Nested metacommunity structure was previously evidenced for trees either in a similar spatial scale (i.e., including distinct forest types across the Brazilian Atlantic Forest) or in smaller scales (within each forest type), but the nested structure was not perfect as clumped species F I G U R E 3 Niche breadth of 1138 species considering their occurrence range (lower and upper limits) across the studied environmental gradients: (a) annual mean temperature and (b) climatic water deficit. ...
... Forest, under a subtropical climate (Neves et al., 2017;Oliveira-Filho et al., 2013). Therefore, even with larger occurrence breadth, coldtolerant species have higher abundance under lower temperatures, probably because of their adaptive traits and consequent benefits to coexistence in local biotic interactions (Abeli et al., 2014;Dallas et al., 2017;Sporbert et al., 2020). ...
Article
Full-text available
Species under milder climates (e.g., warm and wet) tend to experience lower variability in temperature and rainfall regimes and might occur in narrower climatic ranges than species that tolerate harsher conditions (e.g., cold or dry climates). Thus, tree species that occur under harsh conditions should have a broader climatic range, being a small subset of the flora. Here, we assess the influence of climate on species distribution of 1138 tree species from the Atlantic Forest biodiversity hotspot. We investigate their range (or niche breadth), and the “center of gravity” index (or niche optima), along with gradients of mean annual temperature and climatic water deficit (CWD). We further identified those species associated with conditions on different ends of temperature and moisture gradients. We found a small subset of species occurring under colder temperatures or under drier conditions, and these species had a wider niche breadth. The warm or wet‐affiliated species had narrower ranges along with the temperature and the CWD gradients, respectively. Moreover, species affiliated to warm and those to moister conditions had greater densities near their occurrence limits, thus they may be more susceptible to climate changes. We conclude that global climate changes will affect the incidence and abundance distribution patterns of tree species along this threatened biodiversity hotspot, mainly those with narrow niches and within the limit of its distribution. Abstract in Portuguese is available with online material Here, we assess the influence of climate on species distribution of 1,138 tree species from the Atlantic Forest biodiversity hotspot. We found a small subset of species occurring under colder temperatures or under drier conditions, and these species had wider niche breadth. The warm or wet‐affiliated species had narrower ranges along with the temperature and the CWD gradients, respectively. We conclude that global climate changes will affect the incidence and abundance distribution patterns of tree species along this threatened biodiversity hotspot, mainly those with narrow niches and in the limit of its distribution.
... Belonging to the Atlantic Forest biome in the highlands of southern Brazil, the Mixed Ombrophilous Forest, or Araucaria Forest, is one of the most valuable forest ecosystem from an ecological point of view, as it has rare and endemic species (Oliveira-Filho et al., 2015;Chaves et al., 2013). Araucaria angustifolia (Bertol.) ...
Article
Full-text available
The Araucaria Forest is one of the most threatened ecosystems in Brazil due to selective logging and deforestation, mainly during the 20th century. The great majority of remnants are secondary forests with less than 50 ha, i.e., old-growth fragments are extremely rare. By presenting the characterization of rare old-growth fragments in comparison with secondary ones, the present study delivers unprecedented results about their composition, structure, and dendrometric characterization. In total, 22 clusters were installed: 12 clusters (37 plot units, 37,000 m²) in old-growth forest and 10 clusters (38 plot units, 38,000 m²) in secondary forest. Analyses included a dendrometric characterization (N, d, ddom, h, hdom, G and V), a phytosociological classification and multivariate analysis. Results indicated that old-growth and secondary forests show relevant differences at both dendrometric and phytosociological levels. The volume of old-growth forests, 582 m³ ha⁻¹ (362–893 m³ ha⁻¹), is substantially higher than that observed for secondary ones, 334 m³ ha⁻¹ (193–501 m³ ha⁻¹). A. angustifolia accounts for 55% (6–90%) of the volume in the old-growth forests, yet only 21% (0–52%) in the secondary. Multivariate analysis showed that old-growth forests are positively correlated with d, h, and G but negatively correlated with S and the Shannon index.
... Various prior works describe the different types of vegetation present in southern Brazil and Uruguay (e.g., Klein, 1978Klein, , 1979Boldrini, 2009;Oliveira-Filho et al., 2015), mainly influenced by the climate and topography of the region (Figure 1). The Atlantic Forest occupies the northern part of the southern region and the coastal plain. ...
... For example, many drought-sensitive tree species from Amazonian wetlands are able to track favorable wetland microclimates and expand their distribution ranges into tropical savannas (i.e., Cerrado and Caatinga), particularly along N-S oriented rivers, such as the Xingu and Araguaia . Similar biogeographic exchanges were reported for tree species of the Brazilian Atlantic rainforest that extend their distribution ranges towards the South American Cerrado, and the subtropical Pampas through riparian corridors (i.e., Oliveira-Filho et al., 2015). ...
Chapter
Riparian zones and river wetlands include aquatic environments and the adjacent terrestrial areas they temporarily flood. As highly dynamic systems, river wetlands provide a varied assortment of hydrological, biogeochemical and geomorphic settings that drive landscape patterns and processes. These, in turn, impact natural communities and populations in important ways, and provide valuable ecosystem services for humans. River wetlands generate environmentally heterogeneous landscapes, and thus provide differentiated habitats for a rich variety of biological communities and species with varied life history traits. Indeed, river wetlands are among the most biodiverse ecosystems in many climate zones. On landscapes, river wetlands can function as dispersal corridors along which populations expand, as refugia that permit populations to persist in adverse environments, or as dispersal barriers that divide and isolate populations. These landscape processes have different and important ecological and evolutionary consequences. The outstanding role of river wetlands for landscape ecology and biogeography is resumed in this chapter. Human impacts in wetland degradation are exemplified, and the negative consequences for landscapes are discussed.
... This study was conducted in the northern region of the state of Rio Grande do Sul, southern Brazil (Supplementary Material Fig. S1). This region is part of the Atlantic Forest biome located in the transition zone between the Seasonal Forest and the Araucaria Rain Forest (Oliveira-Filho et al. 2015). In the study region, native forest formations are diverse (Leyser et al. 2012;Loregian et al. 2012) and characterised by the presence of deciduous, semi-deciduous and perennial tree species (Ruschel et al. 2005;Athayde et al. 2009). ...
Article
Context In headwater streams, allochthonous litterfall input is an essential process to maintain the ecosystem functioning. The presence of non-native species in the riparian vegetation, with phenological characteristics distinct from those of most native tree species, can alter this process. Aims We evaluated the effect of the phenology of a non-native species (Hovenia dulcis) on the temporal patterns and biomass of litterfall input into streams by comparing one stream with (Hovenia stream) and another without (Native stream) H. dulcis in their riparian vegetation. Methods We quantified the litterfall input from native tree species and H. dulcis for 1 year by using buckets suspended above the streambeds. Key results The temporal pattern of litterfall input changed between streams, with quantitative differences between them during autumn and winter. In the Native stream, litterfall input was slightly higher in spring (∼79 g m−2 month−1), followed by winter (∼68 g m−2 month−1) and autumn (∼54 g m−2 month−1), whereas in the Hovenia stream, it was concentrated in autumn (∼126 g m−2 month−1). Conclusions and implications Our results indicated that the presence of H. dulcis in the riparian vegetation, when in high density, changes the temporal pattern and biomass of litterfall input into subtropical streams.
... At higher altitudes with shallower soils, the Atlantic forest merges with high elevation grasslands (Scheer and Mocochinski, 2009). The different forest expressions are related to rainfall distribution in the winter and summer regimes, and temperatures in winter or the occurrence of frost ( Fig. 1; Oliveira-Filho et al., 2015). In austral winter, the climate at Colônia's latitude is mainly influenced by northward shifts of polar air masses (e.g. the STF) resulting in permanent drizzle and moisture. ...
Article
The early Pleistocene was driven by 41 kyr glacial cycles that have been rarely characterized in continental records, especially in South America. Most of long-term records derive from marine records (e.g.sea surface temperatures (SST)) and have been widely used to infer past climate dynamics but implications for the continent have rarely been evaluated. We present an early Pleistocene record (COL17c) from the Colônia basin in the Atlantic forest domain in southeastern Brazil. Our aim was to integrate past environmental dynamics and the drivers of change between ca. 1.5 to 1.3 Myr at the latitude of Colônia(23°S; ca. 700 m a.s.l.). We applied a multi-proxy approach including pollen, charcoal, X-Ray fluorescence scanning (XRF), biomarkers and diatoms. We identified three glacial periods and four interglacials containing a continuous cool forest, mainly dominated by Araucaria. The glacial periods were characterized by increases in organic matter input on the lake, semi-deciduous forest, and shore and herbaceous vegetation. In contrast, the interglacials were marked by increases in evergreen forest and reduced organic matter input. We attribute these alternating phases of vegetation and lake productivity to meridional temperature differences that regulated the input of moisture at the latitude of Colônia. After 1.430 Myr, glacial and interglacial periods showed a different dynamism with an increase in Araucaria forests and drops in relative temperature, concomitant with regional long-trend cooling observed in marine records. The observed forest responses inferred from the COL17c record are in phase with regional climate features such as the development of the cold Pacific tongue and the equatorward migration of subpolar fronts, highlighting the strong influence of the Southern Hemisphere at Colônia during the early Pleistocene.
... Paraná has high coverage of this forest type both historically and currently and also presents a high influence of species-rich tropical forests from the north. Going southward in the Araucaria Mixed Forest, in the Misiones Province and the states of Santa Catarina and Rio Grande do Sul, there is a marked reduction in tree species richness, possibly because the subtropical/temperate climate with frequent frosts limit the establishment of tropical species (Oliveira-Filho et al. 2015). ...
Article
Full-text available
Although the Araucaria Mixed Forest has long been recognized for their woods and ecosystem services, we still lack basic information on what tree species occur there. Habitat loss and overexploitation have led several tree species of this forest into an extinction process. Therefore, it is urgent to compile what are the tree species of this forest type, identify if these species are threatened and which were not assessed for their threat category. We aimed to answer: (1) How many tree species occur in the Araucaria Mixed Forest? 2) How many of these species are under a threat category? (3) Does the number of threatened species per state/province mirror the species richness of the state/province through the Araucaria Mixed Forest distribution? We found 1,213 tree species for the forest type. The states of São Paulo, Minas Gerais and Rio de Janeiro presented the highest species richness. The number of assessed species reflected the total number of species per state/province. Of the species listed, 5.3 % were classified as threatened and 72.8 % have not been assessed. We provided the most comprehensive tree species list to date for the Araucaria Mixed Forest and unveiled the conservation status of its tree flora. Keywords: Araucaria angustifolia; arboreal species; Atlantic Forest; ecoregion; extinction risk; species list; species pool; subtropical forest; threatened species; tree flora
Article
The conversion of native forests into pastures is still a common practice in Brazil. Abandoned pastures have great potential for natural regeneration and therefore could play an important role in meeting the enormous demand for forest restoration. Few studies, however, have investigated the extent to which spatially-structured environmental variables and community structure are correlated with the variation in species abundance of regenerating forests on abandoned pastures. Therefore, we aimed to determine whether environmental and spatial variables were capable of explaining the variation in abundance of woody species on abandoned pastures in the subtropical Atlantic Forest. We systematically distributed 45 sample plots with size and inclusion criteria that changed according to the vegetation layer in three different abandoned pastures. In general, most of the variation in species abundance that our models were able to explain was correlated with spatially-unstructured physical-chemical soil properties. A smaller part of the variation was correlated with spatially-structured soil variables and topography-related variables. An even smaller portion of the variation was spatially-structured but was not correlated with spatially-structured environmental variables. Therefore, our results suggest that the variation in species abundance of regenerating subtropical Atlantic forests on abandoned pastures is more closely related to niche-based processes mediated by environmental variables than to stochastic spatial processes.
Article
Protected areas (PAs) represent a powerful refuge for maintaining and safeguarding biodiversity. Generally, PAs are delineated to protect terrestrial taxa, providing incidental protection to the aquatic ecosystems within their borders. Here, we compare water quality within PAs and non-PAs in southern Brazil, encompassing remnants of the Atlantic Forest biome, to assess whether PAs serve as a buffer from external pressures for aquatic ecosystems within their boundaries. In addition to physicochemical and microbiological water parameters, we analysed 147 pesticide and 31 pharmaceutical compounds in water samples from 33 sites within and outside PAs. The water quality did not differ between PAs and non-PAs but indicated clear pollution from sewage discharges. We found 19 pesticides and five pharmaceuticals in streams within the study area. We detected pesticides in all sampling sites, with the herbicide 2,4-dichlorophenoxyacetic acid present in 91% of them. Our data show that PAs are insufficient means to mitigate the impacts stemming from their catchments, and the running water that reaches their domains already shows signs of anthropogenic interference, which may affect aquatic biodiversity. Protection and management measures require consideration of the whole watershed to protect freshwater habitats and biota.
Article
Choosing models to predict volume for individual trees is a relevant step in estimating growing stock volume. When large-area estimates are needed, models should ideally be constructed based on observations acquired across the population. However, due to multiple constraints, models are often constructed using samples collected at one or in a few sites. Therefore, this study aimed to evaluate trade-offs between using models based on a regional dataset (RD) and a model based on a local dataset (LD) and to assess the effects of such models on large-scale estimates of stem volume (V) per unit area. Regional and local V datasets and inventory data acquired throughout the subtropical Brazilian Atlantic Forest were employed. When used to predict V for trees of the RD (n = 1,192), the local model (LM), based on 419 tree observations, presented a mean systematic percentage error (MSPE) of +11%. In turn, when the regional model (RM) (n = 1,192) was used to predict V for trees of the LD (n = 419), it yielded an MSPE of-3%. The estimated mean V ha-1 using the LM was 9.4% greater than the estimated mean using the RM. Moreover, less precise estimates were attained using the LM. Study Implications: This study demonstrated with a subtropical forest illustration that models constructed with a large but local dataset performed poorly across a regional population. Therefore, if models with broader application are required, efforts should be invested in collecting calibration data across the population to capture its inherent allometric variability. Model selection proved to be an important step in the estimation of growing stock volume for a regional population, as differences among estimated means reached 8 m³ ha-1 , a number that could affect forest management, payments for carbon uptake, and upscaling of ground forest volume predictions to remotely sensed data.
Article
Full-text available
A vegetação da região da campanha no Rio Grande do Sul é composta por formações campestres com raras e esparsas ocorrências de florestas de galeria. Uma importante questão discutida entre os botânicos é se no decorrer do Holoceno a paisagem, da região da campanha teve sempre predomínio de vegetação campestre como nos dias atuais ou se ocorreram formações florestais. Estudos palinológicos podem contribuir de maneira significativa para o entendimento desta questão. A análise preliminar da zona transicional entre as florestas e campos no Rio Grande do Sul, permite concluir que no Holoceno superior (3.000 anos A.P.) os elementos campestres dominaram o espectro polínico sobre os demais elementos pertencentes a outros tipos de formações. Significativo intercâmbio deu-se apenas dentro do mesmo hábito campestre, às vezes com prevalência de Poaceae e Baccharis (campos secos); outras vezes de Cyperaceae (campos úmidos).
Article
Full-text available
Aim To use the method of parsimony analysis of endemism to identify areas of endemism for passerine birds in the Atlantic Forest, South America, and to compare the locations of these areas with areas previously identified for birds as well as other taxa. Location The Atlantic Forest, eastern South America. Methods We analysed a matrix composed of the presence (1) or absence (0) of 140 endemic species in 24 quadrats of 1 x 1 degree distributed along the Atlantic Forest to find the most parsimonious area cladogram. Results Fourteen most parsimonious cladograms were found and then summarized in a single consensus tree. Four areas of endemism were identified: Pernambuco, Central Bahia, Coastal Bahia, and Serra do Mar. Main conclusions Avian areas of endemism in the Atlantic Forest have significant generality, as they are highly nonrandom and congruent with those of other groups of organisms. A first hypothesis about the historical relationships among the four areas of avian endemism in the Atlantic Forest is delineated. There is a basal dichotomy among areas of endemism in the Atlantic Forest, with Pernambuco forming a northern cluster and Coastal Bahia, Central Bahia and Serra do Mar comprising a southern cluster. Within the southern cluster, Central Bahia and Serra do Mar are more closely related to each other than to Coastal Bahia.
Article
Full-text available
RESUMO As variações da composição da flora arbórea de 60 áreas de floresta atlântica sensu lato (ombrófilas e semidecíduas) da região das Bacias do Leste, englobando o sul da Bahia, o Espírito Santo, o leste de Minas Gerais e o norte do Rio de Janeiro, são analisadas em articulação com variáveis geográficas e climáticas. Listagens de espécies são fornecidas para 16 destas áreas. Análises multivariadas detectaram três padrões de distribuição. (a) A diferenciação entre florestas ombrófilase semidecíduas na regiãoé floristicamente consistente efortemente correlacionada com a sazonalidade do regime de chuvas. A flora arbórea das florestas semidecíduas é, em boa medida, um subconjunto da flora das florestas ombrófilas, extraindo espécies provavelmente mais eficientes em resistir e competir sob condições de seca mais prolongada. (b) Existe uma diferenciação latitudinal tanto para florestas ombrófilas e semidecíduas, que aproxima floristicamente as duas fisionomias dentro da mesma faixa latitudinal. Este padrão é causado provavelmente por variações térmicas e pluviométricas. As florestas ombrófilas são interrompidas no norte fluminense devido ao clima estacional, mas isto não tem como contrapartida uma disjunção na distribuição de espécies arbóreas. (c) As variações da altitude estão fortemente correlacionadas com a diferenciação interna tanto das florestas ombrófilas como das semidecíduas.
Book
Nowadays, ecologists worldwide recognize the use of spatial analysis as essential. However, because of the fast-growing range of methods available, even an expert might occasionally find it challenging to choose the most appropriate one. Providing the ecological and statistical foundations needed to make the right decision, this second edition builds and expands upon the previous one by: • Encompassing the basic methods for spatial analysis, for both complete census and sample data • Investigating updated treatments of spatial autocorrelation and spatio-temporal analysis • Introducing detailed explanations of currently developing approaches, including spatial and spatio-temporal graph theory, scan statistics, fibre process analysis, and Hierarchical Bayesian analysis • Offering practical advice for specific circumstances, such as how to analyze forest Permanent Sample Plot data and how to proceed with transect data when portions of the data series are missing. Written for graduates, researchers and professionals, this book will be a valuable source of reference for years to come.