Content uploaded by Marcelo Leandro Bueno
Author content
All content in this area was uploaded by Marcelo Leandro Bueno on May 03, 2018
Content may be subject to copyright.
Content uploaded by Danilo Neves
Author content
All content in this area was uploaded by Danilo Neves on Oct 07, 2017
Content may be subject to copyright.
403
tropical forests. e savannas, generally referred to as ‘cer-
rado’ (note that throughout we use lower case to refer to the
vegetation type), are the subject of this study.
e Pleistocene refuge theory postulated that South
American savannas, especially cerrado vegetation, expanded
into the Amazon during the Last Glacial Maximum (LGM;
20 000–13 000 yr before present, 20–13 Ka), where it
fragmented the distribution of rain forest vegetation (Haffer
1969). However, paleoecological studies from the Amazon
have suggested continuous presence of rain forest there
through the LGM (Colinvaux et al. 2000, Mayle et al.
2000, Colinvaux and De Oliveira 2001, Bush et al. 2002,
2011, Mayle and Beerling 2004, Urrego et al. 2005, Bush
and de Oliveira 2006). Paleoecological studies focused on
the Cerrado Domain itself have suggested a retraction of
the cerrado tree flora during the LGM, probably caused by
considerable declines in both precipitation and temperature
(Salgado-Labouriau 1973, 1984, 2001, Ledru 1993, 2002,
Behling 1995, Ferraz-Vicentini and Salgado-Labouriau 1996,
Ecography 40: 403–414, 2017
doi: 10.1111/ecog.01860
© 2016 e Authors. Ecography © 2016 Nordic Society Oikos
Subject Editor: John Williams. Editor-in-Chief: Miguel Araújo. Accepted 20 January 2016
ere is strong evidence that global climate fluctuations,
and Pleistocene glacial/interglacial cycles in particular, have
played a key role in determining both the origin and dis-
tribution of living organisms (Hewitt 2000). While at one
time, tropical regions were considered to have been more
stable than temperate regions during Pleistocene climatic
cycles, a great number of studies have suggested otherwise
(Haffer 1969, 1982, Prance 1982). Savanna is one of the
main Neotropical biomes (Bourlière and Hadley 1983), and it
is thought to have shifted its distribution significantly during
the Pleistocene (Ledru 2002). e Cerrado Phytogeographical
Domain contains the largest expanse of the savanna in the
Neotropics, and there has been extensive research aimed at
understanding its distribution during the Pleistocene (Ledru
1993, 2002, Oliveira-Filho and Ratter 1995, 2002, Ledru
et al. 1996, Salgado-Labouriau 1997, Salgado-Labouriau
et al. 1998, Werneck et al. 2012). In addition to savannas,
which are the main vegetation type, the Cerrado Domain
also contains grasslands, semideciduous and seasonally dry
Effects of Quaternary climatic fluctuations on the distribution of
Neotropical savanna tree species
Marcelo Leandro Bueno, R. Toby Pennington, Kyle G. Dexter, Luciana H. Yoshino Kamino,
Vanessa Pontara, Danilo Mesquita Neves, James Alexander Ratter and Ary Teixeira de Oliveira-Filho
M. L. Bueno (buenotanica@gmail.com), V. Pontara, D. M. Neves and A. T. de Oliveira-Filho, Programa de Pós-graduação em Biologia Vegetal,
Univ. Federal de Minas Gerais – UFMG, Campus Pampulha, Cep 31270-090, Belo Horizonte, Brazil. DMN also at: Royal Botanic Gardens,
Kew, TW9 3AB, UK. – R. T. Pennington, K. G. Dexter and J. A. Ratter, Royal Botanic Garden Edinburgh, 20a Inverleith row, EH3 5LR,
Edinburgh, UK. KGD also at: School of GeoSciences, Univ. of Edinburgh, 201 Crew Builing, King’s Buildings, EH9 3JN, Edinburgh, UK. – L.
H. Y. Kamino, Inst. Prístino, Rua Santa Maria Goretti, 86, Cep 30642-020, Belo Horizonte, Brazil.
In order to develop niche models for tree species characteristic of the cerrado vegetation (woody savannas) of central South
America, and to hindcast their distributions during the Last Glacial Maximum and Last Inter-Glacial, we compiled a
dataset of tree species checklists for typical cerrado vegetation (n 282) and other geographically co-occurring vegetation
types, e.g. seasonally dry tropical forest (n 355). We then performed an indicator species analysis to select ten species
that best characterize typical cerrado vegetation and developed niche models for them using the Maxent algorithm. We
used these models to assess the probability of occurrence of each species across South America at the following time slices:
Current (0 ka pre-industrial), Holocene (6 ka BP), Last Glacial Maximum (LGM – 21 ka BP), and Last Interglacial (LIG –
130 ka BP). e niche models were robust for all species and showed the highest probability of occurrence in the core area
of the Cerrado Domain. e palaeomodels suggested changes in the distributions of cerrado tree species throughout the
Quaternary, with expansion during the LIG into the adjacent Amazonian and Atlantic moist forests, as well as connections
with other South American savannas. e LGM models suggested a retraction of cerrado vegetation to inter-tableland
depressions and slopes of the Central Brazilian Highlands. Contrary to previous hypotheses, such as the Pleistocene refuge
theory, we found that the widest expansion of cerrado tree species seems to have occurred during the LIG, most prob-
ably due to its warmer climate. On the other hand, the postulated retractions during the LGM were likely related to both
decreased precipitation and temperature. ese results are congruent with palynological and phylogeographic studies in
the Cerrado Domain.
404
Ledru et al. 1996, Salgado-Labouriau et al. 1997, Barberi
et al. 2000, Lima-Ribeiro et al. 2004). Understanding the
nature of any LGM retraction of cerrado vegetation in the
Cerrado Domain has important implications, because refu-
gial areas may contain higher overall species richness and
higher genetic diversity within individual species (Collevatti
et al. 2012, Lima et al. 2014), and therefore should be
priorities for conservation. Understanding whether any
refugial areas were numerous and scattered micro-refugia, or
fewer, larger areas is therefore of great relevance (Rull 2009,
2011, Vegas-Vilarrubia et al. 2011).
e key question of whether cerrado vegetation may
have expanded into Amazonia or contracted during the
LGM can be addressed by modelling species distributions.
Recent investigations, based on modelling species distribu-
tions and patterns of species richness, endemism and genetic
variation, have provided increased evidence that climati-
cally stable areas could have played the role of refugia for
moist forest species in the Neotropics during Quaternary
climatic fluctuations (Graham et al. 2006, Carnaval and
Moritz 2008, Carnaval et al. 2009, 2014, Keppel et al.
2012, Werneck et al. 2011, 2012, Montade et al. 2014).
Most of these recent studies have focused on moist forests
and the existence of such refugia for cerrado vegetation has
not been sufficiently tested using newer approaches, such
as species distribution modelling. In addition to this, there
has been little investigation of the distribution of savannas,
and the cerrado in particular, before the LGM (although see
Werneck et al. 2012). During the Last Interglacial (LIG,
which began ∼ 130 000 to 116 000 BP (130–116 Ka), the
climate was significantly warmer than during the Holocene
maximum, registering globally higher temperatures
(ca 2 °C) and higher summer insolation (Otto-Bliesner
et al. 2006). us, expansion of the cerrado vegetation and
contraction of moist vegetation may have actually occurred
during the LIG.
Species distribution modelling can be used to comple-
ment palynological studies and enhance our capacity to
hindcast and forecast changes in population and vegetation
dynamics (Scoble and Lowe 2010, Mellick et al. 2012).
is study is the first to hindcast the distributions of tree
species characteristic of the cerrado vegetation to the Last
Inter-Glacial (LIG) and Last Glacial Maximum (LGM).
Werneck et al. (2012) modelled the distribution of cerrado
vegetation based both on a map of the Cerrado Domain
from Brazilian Inst. of Geography and Statistics (IBGE
1998) and a broader spatial definition, as geographically
defined by Olson et al. (2001). However, this approach
is less realistic biologically than studying the responses
of individual species (Collevatti et al. 2013), which is the
approach that we use here.
Our main questions were: a) was there an expansion or
contraction of the cerrado vegetation during the LGM and/
or LIG; b) if cerrado vegetation contracted during one of
these time periods, were there areas of higher environmental
suitability that could have operated as refugia; and c) if and
when cerrado vegetation expanded, was it extensive enough
to fragment Amazonian forest and/or establish connections
between the cerrado and the savannas of northern South
America?
Methods
Study area
e Cerrado Domain spreads across the Central Brazilian
Highlands, which comprise 1/4 of Brazil’s surface, and to
smaller areas in northwestern Paraguay and eastern Bolivia
(Olson et al. 2001, Oliveira-Filho and Ratter 2002) (Fig. 1).
It is the second largest Phytogeographical Domain in South
America, surpassed in area only by the Amazon (Ribeiro and
Walter 2008). e Cerrado Domain extends over 20 degrees
of latitude and from altitudes of 100 m in the Pantanal
(western floodplains) to 1500 m in the highest tablelands of
the Central Brazilian Highlands (Ribeiro and Walter 2008).
ere is remarkable variation in mean annual temperatures
across the region, ranging from 18 to 28°C. Rainfall also
varies widely, from 800 to 2000 mm yr–1, with a long-lasting
dry season during the austral winter (approximately April–
September) (Ab’Saber 2003).
e prevalent vegetation type of the Cerrado Domain
bears the same name, cerrado. It is a woody savanna that var-
ies from fairly open grasslands to forests with a nearly closed
canopy called cerradão (Ribeiro and Walter 2008). e typi-
cal cerrado vegetation grows on acidic, dystrophic soils and
is one of the richest savanna floras of the world, with over
12 000 species of vascular plants (Mendonça et al. 2008).
Dataset
e floristic dataset was extracted from NeoTropTree
(Oliveira-Filho 2014), a database that consists of tree
(defined as free-standing woody plants 3 m in height)
species checklists for 2000 geo-referenced sites compiled
from the literature and herbarium specimen records. We
extracted all 638 sites and 2155 species from the Cerrado
Domain, representing 288 sites of typical cerrado vegeta-
tion, 112 sites of semideciduous forest, 116 of seasonally dry
tropical forest and 122 of mesotrophic cerradão.
e cerrado vegetation is essentially a vegetation of poor
dystrophic soils, and where more fertile soils occur in the
Cerrado Domain, they are occupied by seasonally try tropi-
cal forests or mesotrophic cerradão, which is transitional
between seasonally dry tropical forests and typical cer-
rado vegetation (Ratter 1973, Ratter et al. 1977, 1978a, b,
Oliveira-Filho and Ratter 2002, Ribeiro and Walter 2008,
Bueno et al. 2013). Seasonally dry tropical forests are nota-
ble for experiencing little fire and are thus occupied by a
different set of plant lineages (e.g. Cactaceae) than those
in typical cerrado vegetation, which experiences frequent
and more intense fires (Pennington et al. 2000, 2009).
Meanwhile, mesotrophic cerradão is an almost closed for-
est with a canopy cover of 50–90%, with trees often grow-
ing to 8–12 m (casting a considerable shade so that the
ground layer is much reduced), and including a blend of
species from both typical cerrado vegetation and seasonally
dry tropical forests (Ratter 1973, 1992, Ratter et al. 1977,
1978a, Furley and Ratter 1988, Oliveira-Filho and Fontes
2000, Oliveira-Filho and Ratter 2002, Bueno et al. 2013).
Semideciduous forests are found in more humid areas
than typical cerrado vegetation, such as along river courses
405
(i.e. gallery forest) or in transition zones with moist forests
of the Amazon or the Mata Atlantica. Semideciduous forests
tend to be richer in species than the other vegetation types
(Oliveira-Filho and Ratter 1995, 2000, 2002).
We then performed an indicator species analysis, ISA
(Dufrêne and Legendre 1997), of the same matrix from
the Cerrado Domain to extract the species that indicate
typical cerrado vegetation communities. e ISA produces
an IV (indicator value) obtained by a combination of a spe-
cies’ frequency within a group compared with other groups
(specificity) and the species’ presence in most sites of that
group (fidelity). We performed the analysis using the labdsv
package (Roberts 2013) in the R Statistical Software (R Core
Team). We then selected the ten species with the top IVs in
typical cerrado and extracted the geographical coordinates
of floristic lists in which the species were present from
NeoTropTree (Table 1).
Bioclimatic variables
For all sites, we obtained the value, at 2.5 arc-min (approxi-
mately 5 km) resolution, of the 19 standard BIOCLIM
variables, which reflect various aspects of temperature,
precipitation, and seasonality and which are likely to be
important in determining species distributions (Hijmans
et al. 2005). We cropped the bioclimatic layers to span
from 12°47′N to 34°46′S and from 78°31′W to 35°00′W,
following Werneck et al. (2012) and which represents a
much larger spatial range than that of the Cerrado Domain.
After assessing correlations between the bioclimatic variables,
we retained 10 of 19 variables, eliminating those with less
biological relevance from groups of strongly interrelated
variables (r 0.9). is procedure was done to avoid over-
parametrization of our modelling with redundant variables.
e final selected variables were: annual mean temperature,
mean diurnal range, isothermality, temperature annual
range, mean temperature of wettest quarter, mean tempera-
ture of the driest quarter, mean temperature of warmest
quarter, annual precipitation, precipitation of wettest month
and precipitation of the driest month.
Model construction
We modelled the ecological niche of the ten selected indicator
tree species (Table 1) using Maxent ver. 3.3 (Phillips
et al. 2006). It has been demonstrated that Maxent often
Figure 1. Geographic distribution of the Cerrado Domain and savannas in South America (Olson et al. 2001), with the location and veg-
etation type of floristic checklists used in this study (typical cerrado vegetation: blue circles, mesotrophic cerradão: red circles, seasonally
dry tropical forest: orange circles, and semideciduous forests: green cirlces), following (Oliveira-Filho 2009). Brazilian states are labelled as
follows: Acre (AC), Alagoas (AL), Amazonas (AM), Bahia (BA), Ceará (CE), Distro Federal (DF), Espírito Santo (ES), Goiás (GO), Mara-
nhão (MA), Minas Gerias (MG), Mato Grosso (MT), Mato Grosso do Sul (MS), Pará (PA), Paraíba (PB), Paraná (PR), Rio de Janeiro (RJ),
Rio Grande do Norte (RN), Rondônia (RO), São Paulo (SP), Sergipe (SE), Tocantins (TO).
406
Neotropical Domains (Carnaval and Moritz 2008, Werneck
et al. 2011, 2012). Spatial models were converted from con-
tinuous outputs into presence/absence maps by applying the
lowest presence threshold for each model. is approach
maximizes agreement between observed and modelled dis-
tributions, balancing the cost arising from an incorrect pre-
diction against the benefit gained from a correct prediction
(Pearson et al. 2007). By summing up the presence/absence
maps obtained under Current, Holocene (6 ka BP), LGM
(21 ka BP) and LIG (130 ka BP) projections, we generated
a map of areas showing historical stability. is combined
map depicted areas that were potentially occupied by typi-
cal cerrado species during the climatic oscillations of the
Quaternary. ese historically stable areas, which we consid-
ered to be potential refugia, were defined as those grid cells
for which the presence of all indicator species was inferred
across all time projections.
Species distribution modelling validation
We calculated the sensitivity (the proportion of observed
presences in relation to those that were predicted, which
quantifies omission errors), the specificity (the proportion
of observed absences compared to those that were pre-
dicted, which quantifies commission errors) and the TSS
(true skill statistic), following Allouche et al. (2006). e
TSS test corrects the overall accuracy of the model predic-
tion by the accuracy expected by chance. is test provides
a score between –1 and 1, with values 0.6 considered to
be good, 0.2–0.6 to be fair to moderate and 0.2 to be
poor (Jones et al. 2010). e TSS is a threshold-dependent
measure that is appropriate for evaluating predictive accu-
racy in cases where the model prediction is formulated as
a presence–absence map (Allouche et al. 2006, Jones et al.
2010). ese analyses used the occurrence points of the ten
indicator tree species and 202 occurrences of Eugenia uru-
guayensis obtained from NeoTropTree (Oliveira-Filho 2014),
which has a restricted niche that differs from the typical cer-
rado species and is ideal for simulating absence points for
typical cerrado vegetation.
We also used a threshold-independent method of model
validation, the receiver operating characteristic (ROC) curve
analysis. e ROC curve is obtained by plotting sensitivity
values (the true positive fraction) on the y-axis against their
outperforms other modelling techniques to identify areas
critical to the maintenance of species populations (Elith
et al. 2006, 2011, Pearson et al. 2007, Phillips and Dudík
2008, Pena et al. 2014). In addition, an important reason for
choosing Maxent was that it allowed us to use presence-only
species data, which is of great utility because the vast majority
of the biotic data available, including those used here, come
in this form (Elith et al. 2006, Phillips and Dudík 2008).
To calibrate and evaluate the quality of the models, we
divided the data for each species into a training set (75% of
occurrences) and a test or validation set (25% of occurrences).
We constructed models five times and averaged the output to
produce the final results used in downstream analyses. Next,
for each species, we defined a threshold value above which
grid cells were considered to have environmental characteris-
tics suitable for the maintenance of viable populations of the
species (Pearson et al. 2007). We used the ‘minimum train-
ing presence’ as the threshold selection method because it
assumes that the species presence is restricted to sites at least
as suitable as those at which the species has been observed so
far (Pearson et al. 2007).
In order to produce models to infer the palaeodistribu-
tion of the cerrado indicator tree species, we produced pro-
jections of the suitability of occurrence during the Current
(0 ka pre-industrial), Mid-Holocene (6 ka BP), Last Glacial
Maximum (LGM – 21 ka BP), and Last Interglacial (LIG
– 130 ka BP) time periods based on climatic simulations
(< www.worldclim.org >; Hijmans et al. 2005). For the Last
Glacial Maximum (21 ka, LGM), Holocene (6 ka) and
Current (0 ka pre-industrial) time periods, we employed
the Community Climate System Model – CCSM4 (Gent
et al. 2011), which represents downscaled climate data from
simulations with Global Climate Models (GCMs) based
on the Coupled Model Intercomparison Project Phase 5
(CMIP5; Taylor et al. 2012). e paleo-climatic model for
the Last Interglacial (120 ka, LIG) used the approach of
Otto-Bliesner et al. (2006). We summed the projections of
the ten indicator tree species for each time period, which
together represent the probability of occurrence of typical
cerrado vegetation during that time period. We performed
all geographic information system (GIS) analyses in ArcGIS
ver. 10 (ESRI 2011).
To indicate potential areas of climatic stability for cerrado
tree species during the whole of the Quaternary, we adopted
protocols similar to those used in recent studies for other
Table 1. The ten tree species selected as indicators of typical cerrado vegetation. IV – indicator value; IV-p – the probability of obtaining as
high an indicator value as that observed for typical cerrado vegetation.
Relative frequency in main vegetation types
Species
Species
records
Typical
cerrado
Mesotrophic
cerradão
Seasonally dry
tropical forests
Semideciduous
forests IV IV-p
Connarus suberosus Planch. 296 0.85 0.59 0.02 0.00 0.50 0.001
Erythroxylum suberosum St. Hil. 234 0.84 0.58 0.01 0.00 0.50 0.001
Palicourea rigida Kunth 182 0.65 0.20 0.00 0.01 0.50 0.001
Kielmeyera coriacea Mart. & Zucc. 244 0.87 0.58 0.03 0.09 0.49 0.001
Annona crassiflora Mart. 194 0.69 0.26 0.02 0.04 0.48 0.001
Caryocar brasiliense Cambess. 263 0.75 0.40 0.03 0.06 0.47 0.001
Couepia grandiflora (Mart. & Zucc.) Benth. 242 0.70 0.33 0.02 0.02 0.46 0.001
Qualea parviflora Mart. 254 0.91 0.84 0.04 0.03 0.45 0.001
Byrsonima coccolobifolia Kunth 218 0.78 0.48 0.03 0.06 0.45 0.001
Qualea grandiflora Mart. 266 0.95 0.92 0.12 0.08 0.43 0.001
407
coupled with a contraction toward central Brazil and eastern
Bolivia (Fig. 2B). Further, there was a notable retraction of
typical cerrado vegetation to inter-tableland depressions and
the slopes of the Central Brazilian Highlands, as well as a low
suitability at higher altitudes (Fig. 3).
In the Mid-Holocene (Fig. 2C), the modelled species
expanded their distributions to approach those of the cur-
rent distribution of the typical cerrado vegetation. Lastly, the
results obtained for the Current projection (Fig. 2D) showed
a distribution similar to that of the Cerrado Domain, as
delimited by Olson et al. (2001). Indeed, a map of under
vs over-prediction of cerrado vegetation with respect to the
map of Olson et al. (2001) shows high congruence, particu-
larly in areas of the central Cerrado Domain. Meanwhile,
we overpredicted typical cerrado vegetation in ecotonal
areas between the Cerrado Domain and adjacent Domains,
i.e. in semideciduous forests that transition to the Amazon
and Mata Altantica moist forests (Supplementary material
Appendix 1, Fig. A1).
Some areas in the central region of the Cerrado Domain
showed a high probability of climatic stability throughout
the Quaternary and are shown in yellow in Fig. 4. ese
postulated refugia for typical cerrado vegetation occur mainly
in Minas Gerais and São Paulo states and the Federal District,
with smaller areas scattered across the Cerrado Domain in
other states, such as Tocantins, Goiás and Mato Grosso. e
distribution of Brazilian conservation units shows a low level
of coincidence with these postulated refugia (Fig. 4).
Discussion
Modelling cerrado indicator species
Our results demonstrated the greatest extent of typical
cerrado tree species in the LIG, the greatest contraction in
the LGM, and a subsequent re-expansion in the Holocene.
e values obtained by TSS and AUC modelling suggest that
the environmental variables used in our models provided
important information on the distribution of the tree species
selected as indicators of typical cerrado vegetation, and were
higher than those obtained by other studies modelling the
climatic distribution of neotropical vegetation (Carnaval and
Moritz 2008, Werneck et al. 2011, 2012, Pena et al. 2014),
but which did not model individual indicator species. e
indicator species method has been widely and effectively
equivalent specificity values (1 – specificity, the false positive
fraction) on the x-axis for all possible thresholds (Fielding
and Bell 1997). e ROC analysis characterizes the predic-
tive performance of a model at all possible thresholds by a
single number, the area under the curve (AUC) (Fawcett
2003, Phillips et al. 2006). A single AUC value was calcu-
lated for each species, representing the average across the five
iterations of model construction. e value of the AUC can
fall between 0.5 and 1.0. If the value is 0.5, the model is no
better than random, while models with values above 0.75
are generally considered potentially useful and models with
a value near one are considered to be strongly supported
(Fielding and Bell 1997, Elith 2002, Rushton et al. 2004,
Phillips et al. 2006).
Results
e ten tree species identified by our indicator species
analyses as the most important indicators of typical cerrado
and therefore chosen to generate ecological niche models
and predict current and past distributions are given in Table
1. e quality of the models, according to AUC and TSS
values computed for the ten indicator tree species, showed
that sample and background predictions generated by
Maxent were generally in agreement (Table 2). at is, the
Maxent model performance in this study is much better than
random. is was confirmed by the correct assignment of
the test data using the models, indicating that the models
showed a good performance in predicting species occurrences
with bioclimatic variables.
e palaeomodels suggest significant changes in the
distributions of typical cerrado tree species during the
Quaternary (Fig. 2). e cerrado tree flora experienced its
maximum expansion during the Last Inter-Glacial (LIG),
when the modelled species spread toward the south and east
of the Amazon basin as well as toward the Atlantic coast in
both southeastern and northeastern Brazil (Fig. 2A). All of
these areas shelter current-day cerrado enclaves within moist
forests of both the Amazonian and Atlantic Forest Domains,
as well as within the semi-arid thorn-woodlands of the
Caatinga Domain.
In contrast with the LIG, the models suggest a maximum
retraction of the modelled cerrado species during the Last
Glacial Maximum (LGM), with an almost entire withdrawal
from both eastern Amazonia and Atlantic coastal areas
Table 2. Evaluation of the model performance for cerrado indicator tree species of cerrado vegetation by Maxent. True skill statistic (TSS) and
area under the curve (AUC).
Species
Training
samples
External test
presence points
External test
absence points Sensitivity Specificity TSS AUC
Connarus suberosus Planch. 237 59 202 0.95 1.00 0.95 0.92
Erythroxylum suberosum St.Hil. 176 58 202 0.78 1.00 0.78 0.92
Palicoria rigida Kunth 137 45 202 0.90 1.00 0.90 0.92
Kielmyera coriacea Mart. & Zucc. 183 61 202 0.61 0.99 0.60 0.92
Annona crassiflora Mart.146 48 202 0.95 1.00 0.94 0.93
Caryocar brasiliensis Cambess.211 52 202 0.85 1.00 0.85 0.93
Couepia grandiflora (Mart. & Zucc.) Benth. 194 48 202 0.64 0.99 0.63 0.92
Qualea parviflora Mart.191 63 202 0.93 1.00 0.93 0.91
Byrsonima coccolobifolia Kunth 164 54 202 0.75 1.00 0.75 0.93
Qualea grandiflora Mart. 200 66 202 0.93 1.00 0.93 0.91
408
2005, 2007a, b). Because of this uncertainty, wherever
possible we discuss our results in light of palaeoclimatic recon-
structions based upon other proxies, including fossil pollen
and speleotherms (Van der Hammen 1991, Ledru 1993,
Van der Hammen and Absy 1994, Ledru et al. 1996, Ferraz-
Vicentini and Salgado-Labouriau 1996, Salgado-Labouriau
1997, Salgado-Labouriau et al. 1997, 1998, Barberi et al.
2000, Saniotti 2002, Cruz et al. 2005, 2006, 2009, Cheng
used to determine ecological indicators of community types,
habitat conditions, and environmental changes (Dufrêne
and Legendre 1997, Carignan and Villard 2002, Niemi
and McDonald 2004, De Cáceres and Legendre 2009, De
Cáceres et al. 2010, 2012).
Our conclusions are based upon palaeoclimate simulations
derived from GCMs, which are known to be inaccurate,
particularly in simulating precipitation (Stainforth et al.
Figure 2. Predicted occupancy across northern South America of ten tree species that are indicators of typical cerrado vegetation during:
(A) – the Last Interglacial (LIG 130 ka BP); (B) – the Last Glacial Maximum (LGM 21 ka BP); (C) – the Mid-Holocene (6 ka BP); and
(D) – under Current climate (0 ka pre-industrial). Predictions were based on ecological niche models of climatic preference developed
separately for each species using the MaxEnt algorithm (Phillips et al. 2006). Predicted occupancy was then summed across all ten species.
Warmer colours (red/yellow) of the logistic output correspond to regions with a higher probability of occurrence. Black lines represent the
borders of Brazilian states and South America countries.
409
Santarém (Haffer 1969, Figueroa and Nobre 1990, Van Der
Hammen and Absy 1994). Although most of this region
is forested, numerous isolated savannas are found there,
and it connects the savannas of central Venezuela with the
savannas of central and northeastern Brazil (Haffer 1969).
Ab’Saber (2003) suggested the existence of savanna cor-
ridors in Amazonia during the Quaternary, though he was
not certain about the period when such corridors may have
existed. He also hypothesised that the corridors probably
linked present-day disjunct patches of Amazonian savannas.
Our models provide some corroboration for this idea, show-
ing the expansion of cerrado species toward many of these
currently disjunct savannas (Sanaiotti et al. 2002), such as
Alter do Chão, Amapá, Redenção, Roraima, Humaitá and
the Beni in Bolivia.
Our results are congruent with those of Werneck et al.
(2012) who also suggested past connections of the cerrado
to other areas of savanna in South America during the LIG
and a lack of significant savanna areas or corridors across
central Amazonia during the LGM. Baker and Fritz (2015)
discussed the importance of applying a salinity and tempera-
ture correction to d18O isotopic records. When these cor-
rections are applied to the mean value of d18O during the
LGM in Amazonia (Cheng et al. 2013), this substantially
alters previous climatic interpretations that the Amazon was
‘severely dry’ during the LGM. ese findings all contrast
with one assumption of the Pleistocene Refuge eory
(Haffer 1969, Prance 1982, Whitmore and Prance 1987),
which implicated an LGM savanna expansion due to drier
climates.
Palaeoecological studies from localities across lowland
tropical South America support a decrease in temperature
et al. 2013, Baker and Fritz 2015). One important point
that is clear from these studies is that climatic changes were
probably not synchronised across lowland tropical South
America; rather, different parts of South America may have
experienced climatic change in different directions at the
same time. Predicting the exact history of dispersal, extinc-
tion and recolonization of any typical cerrado tree species
across lowland tropical South America is therefore challeng-
ing. us, here we try and focus on general patterns that
can be inferred from our multi-species, palaeodistribution
modelling approach.
Cerrado distribution during the LIG, LGM and
Holocene
Climate models suggest a warmer and slightly drier climate
during the LIG in those areas of the present-day Atlantic
and Amazon rain forests into which the cerrado tree spe-
cies modelled here are suggested to have expanded. Seasonal
climates expanded toward the Atlantic coast in southeast-
ern Brazil, and the palaeomodel indicated suitable areas for
typical cerrado species near the coast, e.g. in the Paraíba
river valley, in Rio de Janeiro and São Paulo. In fact, there
were small remnants of cerrado in this region, most of which
have disappeared due to habitat alteration in the last century
(Matsumoto and Bittencourt 2001).
e modelled expansion of typical cerrado tree species
into the Amazon during the LIG is particularly notable
within the ‘Amazonian Dry Corridor’, a transverse zone with
mean annual precipitation below 1750 mm extending in a
northwest-southeast direction near the cities of Óbidos and
Figure 3. Predicted occupancy of ten tree species that are indicators of typical cerrado vegetation during the Last Glacial Maximum (LGM
21 ka BP) showing the main highland systems of central Brazil and the low suitability of higher altitudes. Black lines are stat borders.
Warmer colours (red/yellow) correspond to regions with a higher probability of occurrence for all ten species. See Fig. 2 and text for further
details.
410
in the Cerrado Domain, particularly in the Central Brazilian
Highlands. For example, LGM palaeorecords of Salgado-
Labouriau et al. (1997) and Barberi et al. (2000) infer a
prevalence of cold and semi-arid climates in those highlands,
with strong winds, partial soil exposure and concomitant
increased erosion, based on the almost complete absence
of arboreal pollen elements. Our LGM models suggest
that cerrado tree species persisted at lower altitudes, prob-
ably favoured by a warmer climate, deeper soil and higher
ground water storage than at higher altitudes. us, the
inter-tableland depressions and highland slopes of central
Brazil may have been refugia for cerrado species, rather than
the highlands where climates were too cold and dry, as
suggested by Ab’Saber (2003) and Werneck et al. (2012).
Our model, showing a retraction of major areas of typi-
cal cerrado vegetation during the LGM, is also corrobo-
rated by recent studies of population genetics in cerrado
tree species (Ramos 2007, Novaes 2010, Lima et al.
2014). Phylogeographic studies of Hymenaea stignocarpa,
Plathymenia reticulata, Tabebuia aurea and Mauritia flexuosa
found greater genetic diversity in the central region of the
Cerrado Domain, which is indicative that this area could
have been more stable during the LGM. A study of the
during the LGM, indicated by the expansion of cold-
adapted taxa, which are currently either relictual elements
in Amazonia and the Central Brazilian Highlands, such as
Podocarpus, Ilex, Myrsine and Hedyosmum (Colinvaux et al.
1996, Cardenas et al. 2011), or have vanished completely,
like Araucaria (Ledru 1993). On the other hand, during
the LGM there was a drastic retraction in the occurrence
of the tropical palm Mauritia, which has been considered
as an indicator of higher temperatures (Barberi et al. 2000),
as well as the disappearance of tree species characteristic of
seasonally dry tropical forest in eastern Bolivia (Whitney
et al. 2013).
Many authors agree that climate in the central area of
the Cerrado Domain during the LGM was characterized by
a decrease in both precipitation and temperature (Ferraz-
Vicentini and Salgado-Labouriau 1996, Barberi et al. 2000,
Lima-Ribeiro et al. 2004). However, according to Salgado-
Labouriau et al. (1998), there was no synchronicity of LGM
climates inferred from palynological studies in the Cerrado
Domain, which they attributed to differences in latitude and
regional topography.
Our models emphasize low climatic suitability during the
LGM for cerrado tree species at high altitudes (above ∼ 800 m)
0 model
Pacific
Ocean
Atlantic
Ocean
0360 720 1,440
Km
Datum WGS, 1984
1 model 2 models 3 models 4 models
(A)
(B)
(C)
Figure 4. Predicted regions of historical stability for typical cerrado vegetation across the Quaternary, based on summing the predicted
occupancy of ten indicator tree species across Current (0 ka pre-industrial), Mid-Holocene (6 ka BP), Last Glacial Maximum (LGM 21 ka
BP) and Last Interglacial (LIG 130 ka BP) climatic scenarios. Areas in yellow are those where all ten species are predicted to occur at all four
time periods, and represent postulated refugial areas for typical cerrado vegetation. Areas outlined in red are Brazilian conservation units,
while black lines represent the limits of Brazilian states and South America countries. Maps are given for (A) northern South America,
(B) the central area of the Cerrado Domain, and (C) the Federal District.
411
of a vegetation mosaic with a predominance of typical cer-
rado species, interspersed with forest and wet grassland for-
mations, is suggested by numerous authors (Oliveira-Filho
and Ratter 1995, 2002, Ab’Saber 2003, Ribeiro and Walter
2008). In the modelled current potential distribution, areas
of the central Cerrado Domain are maximally suitable for
the occurrence of typical cerrado tree species, particularly
on the central Brazilian tablelands. is finding is corrobo-
rated by Ratter et al. (2003) and Bridgewater et al. (2004),
who demonstrated that areas of the central Cerrado Domain
show the highest species richness of cerrado tree species.
is high diversity may reflect the stability of the central
Cerrado Domain throughout the timespan of our climate
models. Other studies have indicated that the stability of cli-
mate through time facilitates the accumulation and mainte-
nance of diversity in Neotropical vegetation (Graham et al.
2006, Carnaval et al. 2009, Werneck et al. 2012). e persis-
tence of some species in multiple refugia located throughout
their present distribution indicates that these species might
have persisted through multiple climatic cycles in heteroge-
neous environments (Keppel et al. 2012, Turchetto-Zolet
2013). e microrefugia (yellow areas in Fig. 4) are small
areas with favourable environmental features within which
small populations could have survived when their main dis-
tribution area contracted (Rull 2009, 2011, Vegas-Vilarrubia
et al. 2011). ese areas of historical climate stability likely
allowed a number species to persist through time, whereas
extinction took place in areas that experienced the most
severe climate changes. is then likely resulted in greater
diversity in more stable areas (Rull 2008, 2011, Collevatti
et al. 2012, Keppel et al. 2012).
ere has been a great loss in species diversity and
endemism in important areas of the cerrado, as a result of
disturbance, and total clearance, by humans, especially due
to the expansion of agriculture, cattle ranching, and charcoal
production (Ratter et al. 1997, Silva and Bates 2002). ere
are estimates that less than 20% of the Cerrado Domain veg-
etation remains undisturbed while only 7.44% is legally pro-
tected in conservation units. Meanwhile, many threatened
species remain outside any of the region’s parks and reserves
(MMA 2011) contributing to the status of the Cerrado
Domain as one of the world’s biodiversity hotspots, deserv-
ing urgent conservation intervention (Myers et al. 2000).
In our model, climatically stable areas are mostly outside
the existing protected areas. e few exceptions are those
located in the Federal District and a number of Environmental
Protection Areas (APA; a lesser protection level) in Tocantins
state (Fig. 4). Larger climatically stable areas in Minas Gerais
and São Paulo states have no conservation units (Fig. 4). We
suggest that the areas identified as climatically stable in our
analyses should be incorporated into systematic conservation
planning to preserve the cerrado tree flora, as they represent
probable refugial areas with potentially high species and
genetic diversity.
Conclusion
Palaeodistribution modelling of tree species representative of
typical cerrado vegetation showed expansions and contrac-
tions related to the climatic fluctuations of the Quaternary,
phylogeography of the tree species Caryocar brasiliense by
Collevatti et al. (2003, 2012) showed that multiple lineages
may have contributed to the present-day populations of
Caryocar brasiliense in the Cerrado Domain, and that popu-
lations restricted to refugia in the central region during the
LGM may have spread and dispersed to favourable areas in
the last 7000 yr. Moreover, in his revision of Neotropical
Andira, Pennington (2003) highlighted a north to south
parapatric distribution of Andira cuyabensis and Andira
cordata across the centre of the Cerrado Domain, perhaps
related to a prior separation of the currently continuous
typical cerrado vegetation during the LGM, as also suggested
by the palynological data of Ledru (1993).
At the end of the LGM, between 17 000 and 11 000 BP,
the climate became progressively more humid. However,
permanent polar fronts remained at 10°S–20°S latitude
(∼ 8500 BP), inferred from the presence of Araucaria forests
(now confined to southern and southeastern Brazil) and the
association of temperate-adapted Podocarpus with Caryocar
in areas presently covered by typical cerrado (Ledru 1993,
Ledru et al. 1996). With increasing temperatures, the cold
weather elements were probably confined to higher altitudes,
principally in gallery forests, a hypothesis supported by the
presence of Podocarpus in the higher plateaux of the Federal
District and Chapada dos Veadeiros (Barberi et al. 2000).
e return of warmer, humid conditions in the Mid-
Holocene would have favoured the expansion of typical
cerrado vegetation in the core area of the Cerrado Domain
(Oliveira-Filho and Ratter 2000). e appearance of
Mauritia (Barberi et al. 2000, Ledru 2002, Lima et al. 2014)
and the increasing concentration of charcoal particles, are
both associated with increasing temperatures and the re-
expansion of cerrado vegetation (Salgado-Labouriau 1997).
ese changes are corroborated by palaeoecological studies
from various localities and supported by our palaeomodels
for 6000 BP. During this period, Behling (1995) recorded
an increase of species typical of cerrado vegetation, such
as Curatella americana, in the Lagoa do Pires between the
Cerrado and Atlantic Forest Domains in Minas Gerais state.
In the state of Rondônia, there was an isotopic enrichment
related to the replacement of forest vegetation by typical cer-
rado vegetation (Pessenda 1998a), as also observed in the
region of Humaitá, in the south of Amazonas state (Gouveia
et al. 1997, Pessenda 1998b, De Freitas et al. 2001). At
the Bolivian border with Brazil, in an area now covered
by Amazonian forest, Mayle et al. (2000) and Mayle and
Whitney (2012) also recorded the presence of Curatella and
Mauritia during the same period. is find is compatible
with a trend of continuously increasing pollen deposition of
typical cerrado taxa in the period (Barbieri et al. 2000, Ledru
2002). e patterns may have been accentuated by soil
leaching and acidification, which would also have favoured
the expansion of typical cerrado vegetation (Oliveira-Filho
and Ratter 2000).
Current distribution of cerrado, stable areas and
conservation
Around 2000 yr BP, palaeoecological studies suggest the
onset of present-day climatic conditions. e establishment
412
Carnaval, A. C. et al. 2014. Prediction of phylogeographic ende-
mism in an environmentally complex biome. – Proc. R. Soc.
B 281: 1–10.
Cheng, H. et al. 2013. Climate change patterns in Amazonia and
biodiversity. – Nature 4: 1–6.
Colinvaux, P. and De Oliveira, P. E. 2001. Amazon plant diversity
and climate through the Cenozoic. – Palaeogeogr. Palaeoclima-
tol. Palaeoecol. 166: 51–63.
Colinvaux, P. A. et al. 1996. Temperature depression in the lowland
tropics in glacial times. – Clim. Change 32: 19–33.
Colinvaux, P. A. et al. 2000. Amazonian and neotropical plant
communities on glacial time-scales: the failure of the aridity
and refuge hypotheses. – Quat. Sci. Rev. 19: 141–169.
Collevatti, R. G. et al. 2003. Evidences for multiple maternal
lineages of Caryocar brasiliense populations in the Brazilian
Cerrado based on the analysis of chloroplast DNA sequences
and microsatellite haplotype variation. – Mol. Ecol. 12:
105–115.
Collevatti, R. G. et al. 2012. Recovering the demographical history
of a Brazilian Cerrado tree species Caryocar brasiliense: coupling
ecological niche modeling and coalescent analyses. – Nat.
Conserv. 10: 169–176.
Collevatti, R. G. et al. 2013. Drawbacks to palaeodistribution
modelling: the case of South American seasonally dry forests.
– J. Biogeogr. 40: 345–358.
Cruz, F. W. et al. 2006. A stalagmite record of changes in
atmospheric circulation and soil processes in the Brazilian
subtropics during the Late Pleistocene. – Quat. Sci. Rev. 25:
2749–2761.
Cruz, F. W. et al. 2009. Orbitally driven east-west anti-phasing of
South American precipitation. – Nat. Geosci. 2: 210–214.
Cruz, F. W. J. et al. 2005. Insolation-driven changes in atmospheric
circulation over the past 116,000 years in subtropical Brazil.
– Nature 434: 63–66.
De Cáceres, M. and Legendre, P. 2009. Associations between
species and groups of sites: indices and statistical inference.
– Ecology 90: 3566–3574.
De Cáceres, M. et al. 2010. Improving indicator species analysis
by combining groups of sites. – Oikos 119: 1674–1684.
De Cáceres, M. et al. 2012. Using species combinations in indica-
tor analyses. – Methods Ecol. Evol. 3: 973–982.
De Freitas, H. A. et al. 2001. Late Quaternary vegetation dynam-
ics in the southern Amazon basin inferred from carbon isotopes
in soil organic matter. – Quat. Res. 55: 39–46.
Dufrêne, M. and Legendre, P. 1997. Species assemblages and indi-
cator species: the need for a flexible asymmetrical approach.
– Ecol. Monogr. 67: 345–366.
Elith, J. 2002. Quantitative methods for modeling species habitat:
Comparative performance and an application to Australian
plants. – In: Ferson, S. and Burgman, M. (eds), Quantitative
methods for conservation biology. Springer, pp. 39–58.
Elith, J. et al. 2006. Novel methods improve prediction of species’
distributions from occurrence data. – Ecography 29: 129–151.
Elith, J. et al. 2011. A statistical explanation of MaxEnt for ecolo-
gists. – Divers. Distrib. 17: 43–57.
ESRI 2011. ArcGIS Desktop: release 10. – Environmental Systems
Research Inst., Redlands, CA.
Fawcett, T. 2003. ROC graphs: notes and practical considerations
for data mining researchers. – HP Laboratories Palo Alto,
HPL-2003-4.
Ferraz-Vicentini, K. R. and Salgado-Labouriau, M. L. 1996. Paly-
nological analysis of a palm swamp in central Brazil. – J. S.
Am. Earth Sci. 9: 207–219.
Fielding, A. H. and Bell, J. F. 1997. A review of methods for the
assessment of prediction errors in conservation presence/
absence models. – Environ. Conserv. 24: 38–49.
Figueroa, S. N. and Nobre, C. A. 1990. Precipitation distribution
over central and western tropical South America. – Climanálise
5: 36–45.
with the widest expansion during the LIG, related to a
warmer, more seasonal climate. e inter-tableland depres-
sions and the highland slopes in the central region of the
Cerrado Domain probably operated as refugia for the
Cerrado flora during its major retraction in the LGM, a
conclusion that is highly congruent with palynological and
phylogeographic studies. is central region is indicated as
the most species-rich and most stable throughout the cli-
mate fluctuations of the Quaternary, and the conservation
of such high-diversity and climatically stable areas should be
prioritized.
Acknowledgements – is study was in partial fulfilment of the
Doctoral requirements of MLB who thanks CNPq for supporting
a 12-month study period at the Royal Botanic Garden Edinburgh
(grant SWE-202096/2011-4) and Postdoctoral scholarship in
UFMG (151002/2014-2). MLB thanks the Royal Botanic Garden
Edinburgh for support during the time this research was conducted.
References
Ab’Saber, A. N. 2003. Os domínios de natureza no Brasil: poten-
cialidades paisagísticas. – Ateliê Editorial.
Allouche, O. et al. 2006. Assessing the accuracy of species
distribution models: prevalence, kappa and the true skill
statistic (TSS). – J. Appl. Ecol. 43: 1223–1232.
Baker, P. A. and Fritz, S. C. 2015. Nature and causes of Quaternary
climate variation of tropical South America. – Quat. Sci. Rev.
124: 31–47.
Barberi, M. et al. 2000. Paleovegetation and paleoclimate of vereda
de Águas emendadas, central Brazil. – J. S. Am. Earth Sci. 13:
241–254.
Behling, H. 1995. A high resolution Holocene pollen record from
Lago do Pires, SE Brazil: vegetation, climate and fire history.
– J. Paleolimnol. 14: 253–268.
Bourlière, F. and Hadley, M. 1983. Present-day savannas: an
overview. – In: Goodall, W. D. (ed.), Ecosystems of the world
– tropical savannas. Elsevier, pp. 1–17.
Bridgewater, S. et al. 2004. Biogeographic patterns, ß-diversity and
dominance in the cerrado biome of Brazil. – Biodivers. Conserv.
13: 2295–2318.
Bueno, M. L. et al. 2013. A study in an area of transition between
seasonally dry tropical forest and mesotrophic cerradão, in
Mato Grosso do Sul, southwestern Brazil. – Edinb. J. Bot. 70:
469–486.
Bush, M. B. and de Oliveira, P. E. 2006. e rise and fall of the
refugial hypothesis of Amazonian speciation: a paleoecological
perspective. – Biot. Neotrop. 6: 1–17.
Bush, M. B. et al. 2002. Orbital-forcing signal in sediments of two
Amazonian lakes. – J. Paleolimnol. 27: 341–352.
Bush, M. B. et al. 2011. Climate and vegetation change in the
lowlands of the Amazon Basin. – In: Bush, M. B. et al. (eds),
Tropical rainforest response to climate change. Springer, pp.
61–84.
Cardenas, M. L. et al. 2011. e response of vegetation at the
Andean flank in western Amazonia to Pleistocene climate
change. – Science 331: 1055–1058.
Carignan, V. and Villard, M. A. 2002. Selecting indicator species
to monitor ecological integrity: a review. – Environ. Monit.
Assess. 78: 45–61.
Carnaval, A. C. and Moritz, C. 2008. Historical climate modelling
predicts patterns of current biodiversity in the Brazilian Atlan-
tic forest. – J. Biogeogr. 35: 1187–1201.
Carnaval, A. C. et al. 2009. Stability predicts genetic diversity in
the Brazilian Atlantic forest hotspot. – Science 323: 785–789.
413
MMA 2011. Relatório técnico de monitoramento do desmata-
mento no bioma Cerrado, 2009 a 2010: dados revisados.
– Ministério do Meio Ambiente, < www.mma.gov.br/estruturas/
sbf_chm_rbbio/_arquivos/relatoriofinal_cerrado_2010_final_
72_1.pdf >.
Montade, V. et al. 2014. Stability of a Neotropical microrefugium
during climatic instability. – J. Biogeogr. 41: 1215–1226.
Myers, N. et al. 2000. Biodiversity hotspots for conservation
priorities. – Nature 403: 853–858.
Niemi, G. J. and McDonald, M. E. 2004. Application of ecologi-
cal indicators. – Annu. Rev. Ecol. Evol. Syst. 35: 89–111.
Novaes, R. M. L. 2010. Phylogeography of Plathymenia reticulata
(Leguminosae) reveals patterns of recent range expansion
towards northeastern Brazil and southern Cerrados in eastern
tropical South America. – Mol. Ecol. 19: 985–998.
Oliveira-Filho, A. T. 2009. Classificação das fitofisionomias da
América do Sul cisandina tropical e subtropical: proposta de
um novo sistema – prático e flexível – ou uma injeção a mais
de caos? – Rodriguéia 60: 237–258.
Oliveira-Filho, A. T. 2014. NeoTropTree, Flora arbórea da Região
Neotropical: um banco de dados envolvendo biogeografia,
diversidade e conservação. – < www.icb.ufmg.br/treeatlan/ >.
Oliveira-Filho, A. T. and Ratter, J. A. 1995. A study of the origin
of central Brazilian forests by the analysis of plant species dis-
tribution patterns. – Edinb. J. Bot. 52: 141–194.
Oliveira-Filho, A. T. and Fontes, M. A. L. 2000. Patterns of floristic
differentiation among Atlantic forests in south-eastern Brazil,
and the influence of climate. – Biotropica 32: 793–810.
Oliveira-Filho, A. T. and Ratter, J. A. 2000. Padrões florísticos das
matas ciliares da região do cerrado e a evolução das paisagens
do Brasil Central durante o Quaternário Tardio. – In: Rodrigues,
R. R. and Leitão Filho, H. F. (eds), Matas ciliares: conservação
e recuperação. EDUSP, pp. 73–89.
Oliveira-Filho, A. T. and Ratter, J. A. 2002. Vegetation physiogno-
mies and woody flora of the cerrado biome. – In: Oliveira, P.
S. and Marquis, R. J. (eds), e cerrado of Brazil: ecology and
natural history of a Neotropical savanna. Columbia Univ.
Press, pp. 91–120.
Olson, D. M. et al. 2001. Terrestrial ecoregions of the world: a new
map of life on Earth. – BioScience 51: 933–938.
Otto-Bliesner, B. L. et al. 2006. Simulating Arctic climate warmth
and icefield retreat in the last interglaciation. – Science 311:
1751–1753.
Pearson, R. G. et al. 2007. Predicting species distributions from
small numbers of occurrence records: a test case using cryptic
geckos in Madagascar. – J. Biogeogr. 34: 102–117.
Pena, J. C. C. et al. 2014. Assessing the conservation status of
species with limited available data and disjunct distribution.
– Biol. Conserv. 170: 130–136.
Pennington, R. T. 2003. A monograph of Andira (Leguminosae-
Papilionoideae). – Syst. Bot. Monogr. 64.
Pennington, R. T. et al. 2000. Neotropical seasonally dry forests and
Quaternary vegetation changes. – J. Biogeogr. 27: 261–273.
Pennington, R. T. et al. 2009. Woody plant diversity, evolution,
and ecology in the tropics: perspectives from seasonally dry
tropical forests. – Annu. Rev. Ecol. Evol. Syst. 40: 437–457.
Pessenda, L. C. R. 1998a. e carbon isotope record in soils along
a forest-cerrado ecosystem transect: implications for vegetation
changes in the Rondonia state, southwestern Brazilian Amazon
region. – Holocene 8: 631–635.
Pessenda, L. C. R. 1998b. Radiocarbon dating and stable carbon
isotopes of soil organic matter in forest-savanna boundary areas
in the southern Brazilian Amazon region. – Radiocarbon 20:
1013–1022.
Phillips, S. J. and Dudík, M. 2008. Modeling of species distribu-
tions with Maxent: new extensions and a comprehensive
evaluation. – Ecography 31: 161–175.
Phillips, S. J. et al. 2006. Maximum entropy modelling of species
geographic distributions. – Ecol. Model. 190: 231–259.
Furley, P. A. and Ratter, J. A. 1988. Soil resources and plant com-
munities of the central Brazilian cerrado and their develop-
ment. – J. Biogeogr. 15: 97–108.
Gent, P. R. et al. 2011. e community climate system model
version 4. – J. Clim. 24: 4973–4991.
Gouveia, S. E. M. et al. 1997. Dinâmica de vegetações durante o
Quaternário recente no sul do Amazonas indicada pelos
isótopos do carbono (12C, 13C e 14C). – Geoch. Brasil 11:
355–367.
Graham, C. H. et al. 2006. Habitat history improves prediction of
biodiversity in rainforest fauna. – Proc. Natl Acad. Sci. USA
103: 632–636.
Haffer, J. 1969. Speciation in Amazonian forest birds. – Science
168: 131–137.
Haffer, J. 1982. General aspects of the refuge theory. – In: Prance,
G. T. (ed.), Biological diversification in the tropics. Columbia
Univ. Press, pp. 6–24.
Hewitt, G. 2000. e genetic legacy of the quaternary ice ages.
– Nature 405: 907–913.
Hijmans, R. J. et al. 2005. Very high resolution interpolated
climate surfaces for global land areas. – Int. J. Climatol. 25:
1965–1978.
IBGE 1998. Mapa de vegetação do Brasil. – Fundação Inst.
Brasileiro de Geografia e Estatística, Rio de Janeiro.
Jones, C. C. et al. 2010. Combining local- and large-scale models
to predict the distributions of invasive plant species. – Ecol.
Appl. 20: 311–326.
Keppel, G. et al. 2012. Refugia: identifying and understanding safe
havens for biodiversity under climate change. – Global Ecol.
Biogeogr. 21: 393–404.
Ledru, M. P. 1993. Late Quaternary environmental and climatic
changes in central Brazil. – Quat. Res. 39: 90–98.
Ledru, M. P. 2002. Late Quaternary history and evolution of the
Cerrados as revealed by palynological records. – In: Oliveira,
P. S. and Marquis, R. J. (eds), e Cerrados of Brazil: ecology
and natural history of a Neotropical savanna. Columbia Univ.
Press, pp. 33–50.
Ledru, M. P. et al. 1996. e last 50,000 years in the Neotropics
(southern Brazil): evolution of vegetation and climate.
– Palaeogeogr. Palaeoclimatol. Palaeoecol. 123: 239–257.
Lima, N. E. et al. 2014. Phylogeography and ecological niche
modelling, coupled with the fossil pollen record, unravel the
demographic history of a Neotropical swamp palm through the
Quaternary. – J. Biogeogr. 41: 673–686.
Lima-Ribeiro, M. S. et al. 2004. Mudanças na Vegetação na Região
do Rio Meia Ponte, Goiás, Durante o Quaternário Tardio.
– Estudos: vida e saúde 31: 89–105.
Matsumoto, M. H. and Bittencourt, M. D. 2001. Correlação entre
algumas fisionomias de cerrado, no Vale do Paraíba, e classes
de Índices de Vegetação obtidos espectralmente. – In: Inst.
Nacional de Pesquisas Espaciais (ed.), Anais da X SBSR,
pp. 1665–1672.
Mayle, F. E. and Beerling, D. J. 2004. Late Quaternary changes in
Amazonian ecosystems and their implications for global
carbon cycling. – Palaeogeogr. Palaeoclimatol. Palaeoecol. 214:
11–25.
Mayle, F. E. and Whitney, B. S. 2012. Ecotones between forest and
grassland. – In: Myster, R. W. (ed.), Long-term perspectives on
tropical forest–savanna dynamics in lowland Bolivia from the
Last Ice Age until present. Springer, pp. 189–208.
Mayle, F. E. et al. 2000. Millennial-scale dynamics of southern
Amazonian rain forests. – Science 290: 2291–2294.
Mellick, R. et al. 2012. Palaeodistribution modelling and genetic
evidence highlight differential post-glacial range shifts of a rain
forest conifer distributed across a broad latitudinal gradient.
– J. Biogeogr. 39: 2292–2302.
Mendonça, R. C. et al. 2008. Flora Vascular do Cerrado. – In:
Sano, S. M. and Almeida S. P. (eds), Cerrado: ambiente e flora.
EMBRAPA-CPAC, pp. 289–556.
414
Salgado-Labouriau, M. L. et al. 1998. A dry climatic event during
the Late Quaternary of tropical Brazil. – Rev. Paleobot. Palynol.
99: 115–129.
Sanaiotti, T. M. et al. 2002. Past vegetation changes in Amazon
Savannas determined using carbon isotopes of soil organic
matter. – Biotropica 34: 2–16.
Scoble, J. and Lowe, A. J. 2010. A case for incorporating phyloge-
ography and landscape genetics into species distribution
modelling approaches to improve climate adaptation and
conservation planning. – Divers. Distrib. 16: 343–353.
Silva, J. M. C. and Bates, J. M. 2002. Biogeographic patterns and
conservation in the South American Cerrado: a tropical savanna
hotspot. – Bioscience 52: 225–233.
Stainforth, D. A. et al. 2005. Uncertainty in predictions of the
climate response to rising levels of greenhouse gases. – Nature
433: 403–406.
Stainforth, D. A. et al. 2007a. Confidence, uncertainty and
decision-support relevance in climate predictions. – Phil. Trans.
R. Soc. A 365: 2145–2161.
Stainforth, D. A. et al. 2007b. Issues in the interpretation of climate
model ensembles to inform decisions. – Phil. Trans. R. Soc. A
365: 2163–2177.
Taylor, K. E. et al. 2012. An overview of CMIP5 and the
experiment design. – Bull. Am. Meteorol. Soc. 93: 485–498.
Turchetto-Zolet, A. C. 2013. Phylogeographical patterns shed light
on evolutionary process in South America. – Mol. Ecol. 22:
1193–1213.
Urrego, D. H. et al. 2005. e last glacial maximum: stability and
change in a western Amazonian cloud forest. – J. Quat. Sci.
20: 693–701.
Van Der Hammen, T. 1991. Palaeoecology of the Neotropics: an
overview of the state of affairs. – Boletim IG-USP 8:
35–56.
Van Der Hammen, T. and Absy, M. L. 1994. Amazonia during the
last glacial. – Palaeogeogr. Palaeoclimatol. Palaeoecol. 109:
247–261.
Vegas-Vilarrubia, T. et al. 2011. Quaternary palaeoecology and
nature conservation: a general review with examples from the
neotropics. – Quat. Sci. Rev. 30: 2361–2388.
Werneck, F. P. et al. 2011. e diversification of eastern South
American open vegetation biomes: historical biogeography and
perspectives. – Quat. Sci. Rev. 30: 1630–1648.
Werneck, F. P. et al. 2012. Climatic stability in the Brazilian
Cerrado: implications for biogeographical connections of
South American savannas, species richness and conservation in
a biodiversity hotspot. – J. Biogeogr. 39: 1695–1706.
Whitmore, T. C. and Prance, G. T. 1987. Biogeography and
Quaternary history in tropical America. – Oarendon Press.
Whitney, B. S. et al. 2013. Sensitivity of Bolivian seasonally-dry
tropical forest to precipitation and temperature changes over
glacial–interglacial timescales. – Veg. Hist. Archaeobot. 23:
1–14.
Prance, G. T. 1982. Forest refuges: evidence from woody
angiosperms. – In: Prance, G. T. (ed.), Biological diversification
in the tropics. Columbia Univ. Press, pp. 137–158.
Ramos, A. C. S. 2007. Phylogeography of the tree Hymenaea
stigonocarpa (Fabaceae: Caesalpinioideae) and the influence of
Quaternary climate changes in the Brazilian Cerrado. – Ann.
Bot. 100: 1219–1228.
Ratter, J. A. 1973. Observations on the vegetation of northeastern
Mato Grosso. – Phil. Trans. R. Soc. B 226: 449–492.
Ratter, J. A. 1992. Transitions between cerrado and forest vegetation
in Brasil. – In: Furley, P. A. et al. (eds), Nature and dynamics
of forest-savanna boundaries. Chapman and Hall, pp. 51–76.
Ratter, J. A. et al. 1977. Observações adicionais sobre o cerradão
de solos mesotróficos no Brasil central. – In: Ferri, M. G. (ed.),
IV Simpósio sobre o Cerrado: Bases para a Utilização
Agropecuária. EDUSP, pp. 303–316.
Ratter, J. A. et al. 1978a. Observations on forests of some
mesotrophic soils in central Brazil. – Braz. J. Bot. 1: 47–58.
Ratter, J. A. et al. 1978b. Observations on the vegetation of
northeastern Mato Grosso: II. Forests and soils of the Rio
Suiá-Missui areas. – Proc. R. Soc. B 203: 191–208.
Ratter, J. A. et al. 1997. e brazilian cerrado vegetation and
threats to its biodiversity. – Ann. Bot. 80: 223–230.
Ratter, J. A. et al. 2003. Analysis of the floristic composition of
the Brazilian cerrado vegetation iii: comparison of the woody
vegetation of 376 areas. – Edinb. J. Bot. 60: 57–109.
Ribeiro, J. F. and Walter, B. M. T. 2008. As Principais Fitofisiono-
mias de Cerrado. – In: Sano, S. M. and Almeida, S. P. (eds),
Cerrado: ambiente e flora. EMBRAPA-CPAC, pp. 289–556.
Roberts, D. W. 2013. labdsv: ordination and multivariate analysis
for ecology. – R package ver. 14-1, < http://CRAN.R-project.
org/package = labdsv >.
Rull, V. 2008. Speciation timing and neotropical biodiversity: the
Tertiary–Quaternary debate in the light of molecular phyloge-
netic evidence. – Mol. Ecol. 17: 2722–2729.
Rull, V. 2009. Microrefugia. – J. Biogeogr. 36: 481–484.
Rull, V. 2011. Neotropical biodiversity: timing and potential
drivers. – Trends Ecol. Evol. 26: 508–513.
Rushton, S. P. et al. 2004. New paradigms for modelling species
distributions? – J. Appl. Ecol. 41: 193–200.
Salgado-Labouriau, M. L. 1973. Contribuição à Palinologia dos
Cerrados. – Acad. Brasil de Cienc.
Salgado-Labouriau, M. L. 1984. Reconstrucion de los Ambientes
Atraves de los Granos de Pólen. – Invest. Cienc. Sci. Am. 3:
6–17.
Salgado-Labouriau, M. L. 1997. Late Quaternary paleoclimte in
the savannas of South América. – J. Quat. Sci. 12: 371–379.
Salgado-Labouriau, M. L. 2001. Reconstruindo as Comunidades
Vegetais e o Clima no Passado. – Humanidades 48: 24–40.
Salgado-Labouriau, M. L. et al. 1997. Late Quaternary vegetational
and climatic changes in cerrado and palm swamp from central
Brazil. – Palaeogeogr. Palaeoclimatol. Palaeoecol. 128: 215–226.
Supplementary material (Appendix ECOG-01860 at
< www.ecography.org/appendix/ecog-01860 >). Appendix 1.
