Diversity and Distributions. 2019;25:1846–1856.
Received: 22 Augus t 2018
Revised: 14 June 2019
Accepted: 2 August 2019
DOI : 10.1111/ddi .12984
Habitat‐specific impacts of climate change in the Mata
Atlântica biodiversity hotspot
Luíz Fernando Esser1 | Danilo M. Neves2,3 | João André Jarenkow1
This is an op en access arti cle under the ter ms of the Creative Commons Attribution L icense, which pe rmits use, dis tribu tion and reprod uction in any med ium,
provide d the original wor k is properly cited.
© 2019 The Auth ors. Diversity and Distributions published by John Wiley & Sons Ltd.
1Federal Univer sity of Rio Gran de do Sul ,
Porto Alegre, B razil
2University of Arizona, Tucs on, AZ, USA
3Federal Univer sity of Minas Ger ais, Be lo
Luíz Fern ando Es ser, Federal University of
Rio Gra nde do Sul, Porto Alegre, Brazil.
Editor: Janet Franklin
Aim: Elucidate the potential impacts of climate changes on the distribution and con-
servation of the multiple habitats of the Mata Atlântica biodiversity hotspot, which
are often treated as a unique entity in ecological studies.
Location: The whole extension of the South American Atlantic Forest Domain plus
forest intrusions into the neighbouring Cerrado and Pampa Domains, which com-
prises rain forest (‘core’ habitat) and five environmentally marginal habitats, namely
high elevation/latitude forest, rock outcrop habitats, riverine forest, semideciduous
forest and restinga woodlands.
Time period: Current (2000) and future scenarios (2050 and 2070).
Major taxa studied: Tree species.
Methods: We modelled the responses of 282 diagnostic tree species, using multiple
algorithms and distinct scenarios of climate change (828,234 projections).
Results: Potential loss of suitable environment summed 50.4% in semideciduous for-
est, 58.6% in riverine forest and 66% in rock outcrop habitats. Predictions for rain
forest (12.2%), restinga woodlands (7.6%) and high elevation/latitude forest (5.2%)
showed that overall loss of suitable environment will be relatively less severe for
these habitats. Habitats that are confined to narrow edaphic conditions, namely rock
outcrop habitats and riverine forest, are less studied and will likely suffer the greatest
loss of biodiversity because their species are more dispersal limited.
Main conclusions: Because these habitats occupy distinct environmental conditions,
lumping them in ecological analyses might lead to erroneous interpretations in stud-
ies aiming to evaluate the impacts of global change in the Mata Atlântica biodiversity
hotspot. This reinforces the importance of our approach and urges for conserva-
tion strategies that account for habitat heterogeneity in the Mata Atlântica and other
biodiversity conservation, communities' distribution models, habitat conservation,
macroecology, tree communities, vegetation
ESSER Et a l.
1 | INTRODUCTION
The Mata Atlântica of South America is renowned worldwide for
being one of the 36 biodiversity hotspots for conservation prioriti-
zation (Mit termeier, Turner, Larsen, Brooks, & Gascon, 2011; Myers,
Mittermeier, Mittermeier, da Fonseca, & Kent, 20 00; Williams et al.,
2011). Less known facts, however, are that (a) the hotspot status
is specifically referring to its core vegetation type, the rain forest,
and that (b) the Mata Atlântica also houses a diverse and complex
mosaic of vegetation types, with their occurrence and distribution
determined by the harshest extremes of five key environmental fac-
tors (Figure 1; Neves et al., 2017; Scarano, 2009). Thus, vegetation
types are defined here as a plant assemblage and its associated en-
vironmental conditions (hereafter ‘habitat’). Following Walter (1971),
these factors can be classified into azonal (non‐climatic) and zonal
(climatic). The distribution of azonal habitats in the Mata Atlântica
is determined by rocky substrates (rock outcrop dwarf‐forests and
savannas, hencefor th rock outcrop habitats), salinity (white‐sand
woodlands, henceforth restinga woodlands) or waterlogged soils
(tropical riverine forests, henceforth riverine forest), while the
distribution of zonal habitats is determined by frost (montane and
subtropical riverine forests, henceforth high elevation/latitude for-
est), drought stress (semideciduous forests) or high levels of rainfall
(cloud and rain forest, henceforth rain forest).
In a seminal article, Scarano (20 09) argued that environmentally
marginal habitat s in the Mata Atlântica comprise an impoverished
subset of rain forest species that can tolerate the harshest extremes
of their environmental conditions. A recent study, however, showed
that all Mata Atlântica habitats are strikingly distinct both floristi-
cally and environmentally (Neves et al., 2017), suggesting that mar-
ginal habitats are not simply a nested subset of the more diverse
Mata Atlântica rain forest. For conservation purposes, a pertinent
takeaway message in Neves et al. (2017) is that a substantial por-
tion of the plant diversity in the Mata Atlântica might be neglected
if the spatial design for new protected areas is solely based upon
studies that places these multiple habitats together (e.g., Zwiener et
Currently, marginal habitats receive much less protection com-
pared with the rain forest (Neves et al., 2017), despite harbouring
3,160 tree species that are not found anywhere else in the world,
including in the rain forest of the Mata Atlântica. Yet, current levels
of fragmentation and the continuous habitat loss are high through-
out the Mata Atlântica, raising several concerns in the scientific com-
munity (Galindo‐Leal, Jacobsen, Langhammer, & Olivieri, 2003; Joly,
Metzger, & Tabarelli, 2014; Neves et al., 2017; Tabarelli, Cardoso
Da Silva, & Gascon, 2004; Tabarelli, Pinto, Silva, Hirota, & Bede,
2005). In addition to these impacts associated with land use change
in Mata Atlântica habitats, human‐induced climate change (IPCC,
2013) will have widespread effects on Mata Atlântica's ecosystems
(Ferro, Lemes, Melo, Loyolo, & Fenton, 2014; Lemes, Loyolaet, &
Flammini, 2013; Loyola, Lemes, Brum, Provete, & Duarte, 2014).
The persistence of biodiversit y through such global change will
demand biogeographic shifts at all levels of biological organization
(e.g. from populations to communities to functional groups, Bhatta,
Grytnes, & Vetaas, 2018; Frainer et al., 2017; McLachlan, Hellmann,
& Schwar tz, 2007, respectively. See also Barnosk y et al., 2017, for a
In the last decades, ecological niche modelling became a major
tool to predict the impacts of climate changes on biodiversity, aiding
conservation planning in future, dynamic scenarios (Peterson, 2001;
Peterson, Egber t, Sánchez‐Cordero, & Price, 2000; Peterson et al.,
2002). With the development of novel learning machine algorithms
(Guisan & Thuiller, 20 05) and more accurate climate change predic-
tions (Moss et al., 2010), we are now capable to reduce analy tical
FIGURE 1 Distribution of Mata Atlântica habitat s in South America (sensu Scarano, 2009) and main environmental factors (arrows)
sorting species across these habitats (adapted from Neves et al., 2017). Ellipses indicate zonal habitats, and rectangles indicate azonal
ESSER Et al.
uncertainties and provide the much‐needed information to support
conservation prioritization while accounting for global change sce-
narios (Elith et al., 2006). This is of particular relevance for biodiver-
sity hotspots, where species are likely to be more susceptible due to
its reduced population sizes caused by habitat fragmentation.
Our goal here is to elucidate the potential impacts of climate
changes in Mata Atlântica habitats' distribution and conservation.
Because Mata Atlântica habitats occupy distinct climatic and geo-
graphic space, our hypothesis is that climate changes will severely
impact all habitats, though to different degrees. In addition, because
South America will experience increasing temperatures with re-
duced water availability (IPCC, 2013), we predict that future climate
changes will have less severe impacts in restinga, rock outcrop hab-
itats and semideciduous habitats, and more severe impacts in plant
communities found at high elevation/latitude and in riverine and rain
2 | METHODS
2.1 | The dataset
We conducted environmental niche modelling for Mata Atlântica
habitat s using diagnostic species obtained from Neves et al. (2017),
with their presence points available in NeoTropTree (Oliveira‐Filho,
2017). Using diagnostic species to model the climatic distribution of
neotropical vegetation has proven a more efficient approach, given
its higher TSS and AUC values (Bueno et al., 2017) compared with
previous studies (Carnaval & Moritz, 2008; Pena, Kamino, Rodrigues,
Mariano‐Neto, & de Siqueira, 2014; Werneck, 2011; Werneck,
Nogueira, Colli, Sites, & Costa, 2012), and has been effectively used
to determine ecological indicators of community types, habitat con-
ditions and environmental changes (Carignan & Villard, 2002; De
Cáceres & Legendre, 2009; De Cáceres, Legendre, & Moretti, 2010;
De Cáceres, Legendre, Wiser, & Brotons, 2012; Dufrêne & Legendre,
1997; Niemi & McDonald, 2004). To avoid overparameterization
(SDMs in this study have three climatic variables as input data;
see Section 2.2 below), we first excluded species with <20 records
(Thuiller, Guéguen, Renaud, Karger, & Zimmermann, 2019), summing
a tot al of 282 species (see Table S1). These species were classified
in Neves et al. (2017) as diagnostic (see Tichy & Chytry, 2006) of six
Mata Atlântica habitats, with each habitat being distributed across
limited ranges of six environmental gradients: rain forest (warm and
wet climates), high elevation/latitude forest (environments associ-
ated with seasonal cold), semideciduous forest (seasonal drought),
restinga (salinity), rock outcrop habitats (seasonal fire and shallow
soils) and riverine forests (seasonal soil waterlogging). In order to
reflect these limiting environmental conditions in the analyses, we
modelled the species of each habitat using distinct geographic de-
limitations, detailed below (see Figure S2).
Spatial scope for species from high elevation/latitude and rain
forests comprised the whole extent of the Mata Atlântica and the
biogeographical Domains found in the neighbouring South American
dry diagonal, namely Caatinga, Cerrado and Chaco. Because species
from semideciduous forests are widely distributed across the dry
diagonal, their spatial scope comprised the Mata Atlântica, dry di-
agonal Domains and the neighbouring lowland Amazon (warmer cli-
mates). Restingas, riverine and rock outcrop habitats are constrained
within conditions that are primarily related to soil. Therefore, despite
species from restingas, riverine and rock outcrop habitats having cli-
matic suitability in other habitats (e.g. rain forests), these species
are restricted to specific edaphic conditions (e.g. soil waterlogging
in riverine forests). Thus, we modelled the potential distribution of
these species within their edaphically suitable areas, which we es-
tablished as the current distribution of restingas, riverine and rock
outcrop habitats, respectively. We defined the distribution of the
Mata Atlântica habitats, dry diagonal Domains and lowland Amazon
in geographic space by creating polygons from a set of points. The
6,243 NeoTropTree sites (points) were previously classified into one
of the South American biogeographic Domains and into one of the
Mata Atlântica habitats where applicable. The size of each polygon
was then estimated based on the distance between a given site and
the other sites around it (wall‐to‐wall map).
Bioclimatic variables were obtained from Wor ldClim v.1.4
(Hijmans, Cameron, Parra, Jones, & Jarvis, 2005). Climatic layers
were obtained at a 5‐arcmin grain size (~10 km). This spatial resolu-
tion is particularly appropriate for this study because species check-
lists (sites) in NeoTropTree are defined by a single habitat, following
the classification system proposed by Oliveira‐Filho (2017), con-
tained in a circular area with a 10 km diameter. NeoTropTree data
were originally compiled from an extensive survey of published and
unpublished (e.g. PhD theses) literature, particularly those on woody
plant community surveys and floristic inventories. New species oc-
currence records obtained from both major herbaria and taxonomic
monographs were then added to the checklists when they were col-
lected within the 10 km diameter of the original NeoTropTree site
and within the same habitat. The habitat delimitation was conducted
using the package ‘dismo’ (Hijmans & Elith, 2015) in R Statistical
Environment (R Development Core Team, 2011).
2.2 | Variable selection
Variable selection was very conservative in order to build under-
standable and ecologically meaningful models (Figure 2). We fol-
lowed a multiple‐step variable selection routine, consisting of the
following: (a) using variance inflation factors (VIF) to identify highly
collinear variables, which were progressively excluded through a
stepwise procedure. VIFs were computed using two methods: VIFcor
(threshold = 0.5) and VIFstep (threshold = 10; see Marquardt, 1970,
for meth od details). We th en extracte d bioclimatic values from pre s-
ence points and (b) performed a principal components analyses (PCA)
to visualize which variables were more effective in segregating the
climatic space of each habitat relative to the climatic space of all other
habitat s. We also (c) per formed PCA s for each habitat separately to
assess which climatic variables showed higher correlations with the
first three principal components (there was a negligible increase in
constrained variance by adding a four th component). Lastly, we (d)
ESSER Et a l.
used Pearson's correlation to test whether all variables selected for
a given habitat showed low correlation (cut‐off = − 0.5 < p < .5). We
also legitimated the variable selection with literature review, which
allowed us to select variables that better represented the climatic
space occupied by the species of each habitat, while taking into ac-
count their ecological relevance (see Table S3).
2.3 | Environmental niche modelling
Models were calculated in three independent cross‐validation runs
with 30% of data kept to evaluate the model and two evaluation
met ho ds (true sk ill st atistic, TSS, and are a un de r the recei ver operat-
ing charac teristic, AUC) for every algorithm available in ‘biom od2’ r
Package (Thuiller, Georges, Engler, Georges, & Thuiller, 2014; gen-
eralized linear models, generalized additive models, boosted regres-
sion trees, classification tree analysis, ar tificial neural net works,
Bioclim, flexible discriminant analysis, multiple adaptive regression
splines, random forest and MaxEnt). We only kept ensemble mod-
els with TSS higher than 0.7. We generated 1,000 pseudoabsences
through different background areas for each habitat, since they have
distinc t spatial scopes in our analyses (see section 2.1 and Figure
S2 for more details). A caveat to this approach is the recommenda-
tions of Barbet‐Massin, Jiguet, Albert, and Thuiller (2012) regarding
the use of lower pseudoabsences in some algorithms. Nonetheless,
here we followed Thuiller (2014), which points out that the main ad-
vantage of biomod2 lies in the capability to compare and combine
mul tiple algor it hms using the sam e se t of init ial data and paramete ri-
zation. We controlled for spatial autocorrelation in our models using
a generalized least squares framework (Zuur, Ieno, Walker, Saveliev,
& Smith, 2009), wh ic h consists in modelling alpha diver sity as a func-
tion of predicting variables using different spatial correlation struc-
tures (exponential, gaussian, spherical, linear and rational quadratics)
and then selecting the best model (highest delta AIC relative to the
null model; i.e. no spatial autocorrelation). We then built a raster with
cell sizes as values weighted by presence probabilities to provide a
more conser vative measure of the potential area occupied by each
habitat (Figure 2). Models were projected to CMIP5 data (Coupled
Model Intercomparison Project Phase 5; downscaled at 5‐arc‐min-
ute spatial resolution) using all General Circulation Models available
in WorldCli m v.1.4 (Hijmans et al., 2005) to the four Representative
Concentration Pathways (RCP2.6, 4.5, 6.0 and 8.5) to the years of
2050 and 2070 (BCC‐CSM1‐1, CCSM4, GISS‐E2‐R, HadGEM2‐AO,
HadGEM2‐ES, IPSL‐CM5A‐LR, MIROC‐ESM‐CHEM, MIROC‐ESM,
MIROC5, MRI‐CGCM3 and NorESM1‐M), summing 88 scenarios and
a total of 828,234 projections. Species projections were summed
into an alpha diversity raster for each habitat and weighted by the
FIGURE 2 Methods summary.
Environmental niche models (ENM) were
projected to 11 Atmosphere Ocean
General Circulation Models (AOGCM)
and four represent ative concentration
pathways (RCP) to 2050 and 2070.
To calculate potential occupied area,
the presence probability rasters were
multiplied by cell area rasters, generating
a weighted area raster, following two
approaches: (i) considering a presence–
absence map with a threshold = 0.5, that
is, each cell with presence probabilit y
>0.5 sum 100 km2 (grid cell size) of the
total potential area. (ii) Considering that
cells could be partially occupied, that
is, occupancy models that are either
gradually fading (a) or abruptly changing
ESSER Et al.
maximum number of species before generating habitat suitability
maps (Figure 2). Ensemble models were generated for each habitat
by first summing their diagnostic species distribution maps and then
dividing the resulting map by the number of diagnostic species in a
given habitat. This generated a final suitability map (ranging from
zero to one) for each habitat.
Finally, to assess the potential conservation status of Mata
Atlântica habitats, we overlaid the current and future distributions
of each habitat on to the coverage of protected areas in the World
Database on Protected Areas (IUCN & UNEP‐WCMC , 2015).
3 | RESULTS
3.1 | Potential area and conservation status
Our models showed that in current climatic conditions, the existing
network of protected areas is more effective in protecting the po-
tential distribution of azonal habitat s (17.4% compared to only 9.0%
in zonal habitats, Figure 3). Semideciduous forest is the least pro-
tected habitat, with only 7.1% of its potential area (537,640.29 km2)
occurring within protected areas (39,320.39 km2). Amongst azonal
habitat s, riverine forest was the least protected, with only 8.5% of
its potential area (91,492.64 km2) occurring within protec ted areas
(7,816 km2). On average, 13.8% of the potential distributions of mar-
ginal habitats are found within protec ted areas, which is higher than
the potential distribution of rain forest occurring within protec ted
areas (10.2%; 41,203.47 km2).
From current conditions to the worst climate change scenario,
the high elevation/latitude forest was the least affected, with 5.2%
of potential area shrinkage, followed by restinga (7.6%), and rain for-
est (12.2%). In contrast, future scenarios for semideciduous, riverine
and rock outcrop habitats were worrisome. Potential area shrinkage
in future climatic scenarios can be as high as 50.4% in semideciduous
forest, 58.6% in riverine forest and 66% in rock outcrop habitats.
This loss of climatically suit able areas across all habit ats is also re-
flected in their levels of protection. From current to worst scenario,
restinga woodlands are predicted to lose climatic suitability in 6.6%
of its currently protected area, followed by high elevation/latitude
(8.0%), rain (13.6%) and riverine (55.3%) forests. The current net-
work of protected areas in rock outcrop habitats is predicted to
undergo the most severe impacts of climate change, with 60.1% of
shrinkage in areas of climatic suitability for species of rock outcrop
habitat s in these protected areas. Conversely, shrinkage in areas of
climatic suitability for species of semideciduous forest (50.4%) will
mainly occur outside protec ted areas (19.0% of protected area loss).
3.2 | Distribution of azonal habitats
Riverine forests, which are mainly found in Central Brazil, and rock
outcrop habitats, which are mainly found in the transition between
Mata Atlântica and Cerr ado, are predicted to lose higher levels of cli-
matic suitability in lower latitudes (see Figures S4 and S5). Restinga is
predicted to lose lower amounts of suitable climatic space relative to
the other Mata Atlântica habitats, suggesting higher climatic stability
across coastal white‐sand environments in eastern South America
(see Figures S6 and S7).
3.3 | Distribution of zonal habitats
Our results showed a substantial degree of overlap in the climatic
spaces occupied by species from high elevation/latitude, semidecid-
uous and rain forests (for decoupled maps check Figures S8, S9 and
S10). This suggests that the abrupt contours that are currently used
for delimiting the distribution of these three habitats might be too
simplistic (Figure 4). Under current climatic conditions, for instance,
our models showed that for 3.7% of the geographic space covered
FIGURE 3 Potential area (in km2) of
Mata Atlântica habitats (total area in a
and b; protec ted area in c and d) through
scenarios of increase in CO2 concentration
ESSER Et a l.
by zonal habitats, there is an equivalent probability that a given area
(~100 km2) is suitable for species from high elevation/latitude, sem-
ideciduous and rain forests. This intercept increases through scenar-
ios, varying from 6.2% of overlap in RCP2.6/2070 and RCP6.0/2070
to 7.1% in RCP8.5/2070.
Climatic overlap between two habitats is even higher. Species
from high elevation/latitude and rain forest showed the highest de-
gre e of overlap in climati c su it abil it y, ranging from 14% in cur re nt cli-
matic conditions to 24.3% in RCP8.5/2070. In contr ast, species fro m
se mid e cidu ous an d ra in fo res ts sh owe d a much lo w er deg ree of ove r-
lap in climatic suitability (6.7% in current climate), which decreases
over time (1.2% in RCP8.5/2070). Unique climatic space (i.e., suitable
for species of a single habitat) is highly variable across high eleva-
tion/latitude, semideciduous and rain forests, and unstable through
time. Potential climatic uniqueness for rain forest species ranges
from 1.1% in current climate conditions to 0.5% in RCP8.5/2050,
reaching a minimum of 0.2% in RCP6.0/2050. Semideciduous forest
showed both the highest degree of climatic uniqueness and future
instability, ranging from 25.1% in current climatic conditions to only
8% in RCP8.5/2070. Species from high elevation/latitude forest
showed 18.9% of potential climatic uniqueness, which decreases to
18.1% in RCP8.5/2070 and 13.8% in RCP4.5/2050 (Figure 5).
3.4 | Climatically stable areas
Areas in southeastern Brazil showed a high probability of climatic
stability for species of all three zonal habitats (Figure 6). These po-
tential refugia occur mainly in Rio de Janeiro and São Paulo states.
Potential refugia for species of high elevation/latitude forest are also
found in southern Brazil and Uruguay. Potential refugia for species
of semideciduous forests are scattered across central and south-
eastern Brazil, with larger areas in Minas Gerais state. The distribu-
tion of protected areas shows a low level of coincidence with these
postulated refugia (Figure 6), ranging from 13.4% in high elevation/
latitude and semideciduous forest to 32.8% in rain forests.
Areas in eastern and central‐western Brazil showed a high prob-
ability of climatic stability for species of azonal habitat s. Existing
protected areas in the Federal District and across Minas Gerais state
(e.g., Canastra National Park) are potential refugia for species of riv-
erine forest. Potential refugia for species of rock outcrop habitats
FIGURE 4 Distribution of zonal
habitats through climate change
scenarios yielded by environmental niche
modelling of their diagnostic species (see
Table S1 and Section 2). Habitats were
plotted using a red‐green‐blue colour
scheme. The brightest shades of red,
green and blue represent the highest
probability of occurrence of species from
semideciduous, rain and high elevation/
latitude forests, respectively. If a grid cell
is potentially occupied by two habitats,
its colour will represent an intermediate
palette between the colours for these
two habit ats. White indicates presence
of all habitats, while black indicates full
absence. Pearson's correlations between
current and each of the future distribution
maps are given for semideciduous, rain
and high elevation/latitude forests (S, R
and H, respectively)
ESSER Et al.
are scattered in Minas Gerais state (Gandarela and Caparaó National
Parks, and Brigadeiro State Park). Large areas of climatic stability
areas for species of restinga woodlands are found in northeastern
Brazil, across the coastline of Bahia, Alagoas and Pernambuco states.
However, these climatically stable restinga woodlands are mostly
found out side existing protected areas (only 19.9% within protected
areas; Figure 6).
4 | DISCUSSION
Here, we showed that both core and marginal habitats of the Mata
Atlântica will be severely impacted by human‐induced climate
change, though to different, uneven degrees. For instance, consider-
ing variation from current conditions to the most pessimistic sce-
nario of climate change in our models (RCP8.5/2070), rain forest is
likely to be more climatically stable relative to semideciduous, river-
ine and rock outcrop habitats, but more impacted than high eleva-
tion/latitude forest and restinga woodlands. These findings are of
relevance for conser vation planning predicated on the protection of
biodiversity under climate change scenarios. Because there is a con-
siderable level of plant endemism in both core and marginal habitats
(Neves et al., 2017), a portion of such species could be neglec ted if
future conservation strategies prioritise regions of highest climatic
stability regardless of habitat heterogeneity (e.g., Lemes et al., 2013;
Ferro et al., 2014; Loyola et al., 2014; Zwiener et al., 2017, Sobral‐
Souza, Vancine, Ribeiro, & Lima‐Ribeiro, 2018), but core and mar-
ginal habitats are unevenly distributed across these stable regions.
4.1 | Potential area and conservation status
Through the scenarios, protected areas in riverine forest will have
more stable climates across the southeastern portion of its current
distribution, highlighting the importance of these areas for protect-
in g via b le pop u lat ion si ze s of riv erin e spec i es. Co ngr uen t wit h the re -
sults for riverine forest, our future scenario models showed that rock
outcrop habitats will lose more climatically suitable areas in their
lower latitudes, suggesting that southernmost sites may function
as climatic refugia for this hyperdiverse habitat (Neves et al., 2018).
However, given the scattered spatial configuration of these rock
outcrop sites, dispersal is likely to be very limited, which suggests
that conservation strategies might need to consider new protected
areas that connect these outcrop islands through the lowlands. In
fact, previous studies (Mews, Pinto, Eisenlohr, & Lenza, 2014; Neves
et al., 2018) provided evidence that rock outcrop habitats and their
surrounding lowland savannas are likely to form a continuous meta-
communit y with spatial variation in woody plant population sizes
being mainly driven by source–sink dynamics (Pulliam & Danielson,
FIGURE 5 Number of grid cells per
zonal habitat and their intercepts through
climate change scenarios. Bottom groups
represent the total number of grid cells
of semideciduous (S.Set), rain (R. Set)
and high elevation/latitude forests
(H.Set). Upper‐left groups represent
grid cells that in our models are uniquely
covered by semideciduous (S), rain (R)
and (H) high elevation/latitude forest s.
Upper‐right groups represent grid cells
where three (high elevation/latitude‐
rain‐semideciduous, HRS) or two habitats
overlap (semideciduous‐rain, SR; high
high elevation/latitude‐rain, HR). Habitat s
with climatic suitability ≥0.33 in a grid cell
were considered present. Chord diagrams
were made using circlize R package (Gu,
Gu, Eils, Schlesner, & Brors, 2014)
ESSER Et a l.
1991). Therefore, here we stress that protected areas aiming to se-
cure biodiversity of rock outcrop habit ats should not be limited to
rock outcrop areas. Rather, effective protected areas should func-
tion as ecological corridors connecting multiple rock outcrop sites
through lowland environments.
Our models showed that while climate in restinga woodlands are
expected to be more stable over time when compared to other hab-
itats, this level of stability is highly variable within its distribution,
with central and southern restingas being relatively more stable.
In addition to this uneven impact of climate change across restinga
woodlands, coastal environments are also expected to be affec ted
by erosion and sea level raise (EUROSION, 2004; IPCC , 2013). This
suggests that conservation planning for restinga woodlands will
require a high degree of complexity, with its effectivity depending
on strategies that account for geomorphological variation changes
associated with both climate and land use change. Restinga has
suffered massive fragmentation due to high human occupation in
coastal areas and a rapidly developing, unplanned tourism industry.
Amongst zonal habitats, semideciduous forest is predicted
to be the most impacted, losing 64% of its current potential dis-
tribution under the most pessimistic scenario (RCP8.5/2070).
Mo re ove r, whi le ou r mode ls pre di c t cli mat ic st ability fo r spec ies of
semideciduous forest in southeastern Brazil, there is a high degree
of potential area shrinkage for species of semideciduous forests
in northeastern Brazil (see Figure S10). These results therefore
suggest that conservation strategies aiming to protect suitable
climatic space for these northern species would have to consider
corridors that could potentially link their current and future suit-
able climates. Conversely, high elevation/latitude and rain forests
are relatively stable over time, indicating the need for tailor‐made
conservation strategies for each habitat of the Mata Atlântica.
Nonetheless, biodiversity in these forests is poorly and unequally
captured by the current net work of protected areas, especially in
southern Brazil (Saraiva, dos Santos, Overbeck, Giehl & Jarenkow,
2018). Here, we suggest that accounting for climate change sce-
narios, in addition to multi‐dimensional biodiversity assessments
FIGURE 6 Climatic st abilit y in Mata
Atlântica habit ats yielded by ecological
niche models of 269 diagnostic tree
species. Coloured grid cells (stable sites)
represent areas where all diagnostic
species of a given habitat are predicted
to occur in all 89 scenarios of current and
future climates (four concentrations of
atmospheric carbon for the years 2050
and 2070, 11 AOGCMs). Red contours
indicate the current network of protec ted
areas in South America. Black contours
represent the national borders and state
limits in Brazil. Values in parentheses
indicate the amount of climatic ally st able
areas in square kilometres for each
habitat. Acronyms represent Brazilian
states mentioned in the Results: Alagoas
(AL), Bahia (BA), Federal District (FD),
Minas Gerais (MG), Pernambuco (PE), Rio
de Janeiro (RJ) and São Paulo (SP)
ESSER Et al.
as in Saraiva et al. (2018), might improve current and future con-
servation strategies for these neglected high elevation/latitude
and rain forests.
4.2 | Climate change and compositional complexity
Previous studies (Neves et al., 2017; Oliveira‐Filho, Budke, Jarenkow,
Eisenlohr, & Neves, 2015; Oliveira‐Filho & Fontes, 2000) that ad-
dressed climatic differentiation amongst Mata Atlântica habitats
showed that while these habitats are floristically distinct, such com-
positional dif ferentiation is only partially explained by variation in
current climatic conditions. Our models not only supported the idea
that delimiting the distribution of Mata Atlântica habitats is no easy
task, but also showed that such complexity will likely increase under
climate change, that is, because we currently lack a complete un-
derstanding of the factors that control the distribution of species
through space and climatic gradients, predicting climate‐driven bio-
geographical shifts is inherently uncer tain.
There are many potentially important factors in determining the
distribution of species that we have not accounted for adequately and
that should be considered/addressed in future studies (see Neves et
al. , 2018; Ti teux, Duf re ne, Jac ob, Paq uay, & Defo ur ny, 20 04). Amongst
these factors, the importance of biotic processes (e.g. competition,
natural enemies) to species distributions and community composition
is the most neglec ted in the literature, especially in studies address-
ing compositional turnover under climate change scenarios. Here, we
highlight that accounting for biotic processes and assessing how they
may potentially var y through time is not trivial for studies aiming to
accurately predict the impacts of global change on biodiversity.
4.3 | Climatic stability and protected areas
Biodiversity loss from climate change arises because species move to
track suitable climate, and agricultural lands, urban development or
transportation corridors may stop their movement (Hannah, Midgley,
Hughes, & Bomhard, 2005; Heller & Zavaleta, 2009). Protected areas
and biodiversity‐friendly land uses lessen these barriers to movement
(U r b an, 20 15) , but th e d a t a nee ded to inf o r m la nd us e man age r s req uir e
ins ight s from ecol og is ts in which the movements of var ious species ar e
modelled under multiple climate scenarios. In our models, climatically
stable areas are mostly outside the existing protected areas (83.8%).
We, therefore, suggest that the areas identified as climatically stable
in our analyses should be incorporated into systematic conservation
planning and restoration project s to preser ve Mata Atlântica habitats.
Altogeth er, these areas function as probable refugial are as and climati-
cally stable corridors connecting unstable protected areas to currently
protected refugial areas.
5 | CONCLUSIONS
Our study showed that ‘lumping’ the natural heterogeneity of
the Mata Atlântica can bring great havoc for future conser vation
strategies, and highlighted three additional factors to be consid-
ered in conservation planning for this biodiversity hotspot: (a)
we still have little understanding of how climate controls species
distribution across the Mata Atlântica, and therefore, the future
distri bution of spec ies fro m zonal habitats, namel y high elevation/
latitude, semideciduous and rain forests, is highly uncertain. New
conservation strategies will need to account for such uncertainty
when estimating which areas in geographic space are more likely
to protect species from a given habitat and which areas are likely
to represent climatic overlaps that are suitable for species from
two o r more habit ats. (b) The maintenance of ha bit at area through
tim e wi ll like ly depend on major bio geographi ca l sh if ts (see result s
for semideciduous forests). Thus, new conservation strategies
will need to account for the climatic space that will likely facili-
tate gradual migration under a changing environment. (c) Under
climate change scenarios, spatial rearrangements for species of
azonal habitats can only occur within the range that comprises
their edaphic requirements, namely rock outcrops (rock outcrop
habitats), seasonally waterlogged soils (riverine forest) and white‐
sand saline soils (restinga woodlands). This leads to a more lim-
ited array of conser vation strategies for these habitats. Thus, for
azonal habitats, considering conser vation strategies that prevent
the cu rrent ly high level s of fr ag menta ti on asso ci at ed with land use
change is a must.
Further studies assessing climate changes impacts in habitats
may trace how areas might change in (diagnostic) species compo-
sition and richness over time, culminating in the emergence of new
habitat s. In terms of azonal habitats, plant–soil relationship should
be addressed carefully, considering influences of climate on sub-
strates, as well as the suitability for plants under new combination of
climate and edaphic conditions.
We thank Marinez Siqueira, Gerhard Overbeck and Demétrio
Guadagnin for their insightful considerations on an earlier version of
th i s manu scr i pt. We al so tha nk tw o an ony mou s refer ees fo r thei r val -
uable contributions to this manuscript. D.M.N. was funded by a col-
laborative research grant from the US National Science Foundation
(DEB‐1556651) during the time this research was completed.
DATA AVA ILAB ILITY STATE MEN T
All environmental layers are available in the WorldCl im data-
base (http://www.world clim.org). Species data are available in the
NeoTropTree website (http://www.neotr optree.info).
Luíz Fernando Esser https://orcid.org/0000‐0003‐2982‐7223
Danilo M. Neves https://orcid.org/0000‐0002‐0855‐4169
João André Jarenkow https://orcid.org/0000‐0003‐2747‐3468
ESSER Et a l.
Barbet‐Massin, M., Jiguet, F., Albert , C. H., & Thuiller, W. (2012). Selecting
pseudo‐absences for species distribution models: How, where and
how many? Methods in Ecology and Evolution, 3(2), 327–338. https ://
doi .org/10.1111/j. 2041‐210X .2011 .0 0172.x
Bar nosky, A. D., Had ly, E. A., Gonz al ez, P., Hea d, J., Polly, P. D., La wi ng ,
A. M., … Zhang, Z. (2017). Merging paleobiology with conservation
biology to guide the future of terrestrial ecosystems. Science, 355,
eaah4787. https ://doi.org/10.1126/scien ce.aah4787
Bhat ta , K . P., Gr ytne s, J. A ., & Veta as , O. R. (2 018 ). Dow nh ill shif t of alpine
plant assemblages under contemporary climate and land‐use changes.
Ecosphere, 9(1), e02084. https ://doi.org/10.1002/ecs2.2084
Bueno, M. L ., Pennington, R . T., Dex ter, K. G., Kamino, L. H. Y., Pontara,
V., Neves, D. M., … de Oliveira‐Filho, A . T. (2017). Effects of quater-
nary climatic fluctuations on the distribution of Neotropical savanna
tree species. Ecography, 40(3), 4 03–414. ht tp s ://doi. org/10 .1111/
Carignan, V., & Villard, M.‐A. (2002). Selecting indicator species to mon-
itor ecological integrit y: A review. Environmental Monitoring and
Assessment, 78, 45–61.
Carnaval, A. C., & Moritz, C. (2008). Historical climate modelling pre-
dicts pat terns of current biodiversit y in the Brazilian Atlantic
forest . Journal of Biogeography, 35, 1187–1201. https ://doi.
org /10.1111/j .136 5‐269 9.20 07.01870 .x
De Cáceres, M., & Legendre, P. (20 09). Associations between species and
groups of sites: Indices and statis tical inference. Ecology, 90, 3566–
3574. https ://doi.org/10.1890/08‐1823.1
De Cáceres, M., Legendre, P., & Moretti, M. (2010). Improving indicator
species analysis by combining groups of sites. Oikos, 119, 1674–1684.
https ://doi.o rg /10.1111 /j.160 0‐070 6. 20 10.1833 4. x
De C ácer es , M., Leg endr e, P., Wis er, S. K. , & Bro ton s, L . (2 012 ). Us i n g sp ec ies
combinations in indicator value analyses. Methods in Ecol ogy a nd Evoluti on,
3, 973–982. https ://doi.org/10.1111/j.2041‐210X.2012.00246.x
Dufrêne, M., & Legendre, P. (1997). Species assemblages and indicator
species: The need for a flexible asymmetrical approach. Ecological
Monographs, 67, 345–366. https ://doi.org/10.2307/2963459
Elith, J., H. Graham, C., P. Anderson , R., Dudík, M., Ferrier, S., Guisan, A .,
… E. Zimmermann, N. (20 06). Novel methods improve prediction of
species’ distributions from occurrence data. Ecography, 29(2), 129–
151. https ://doi.org/10.1111/j.2006.0906‐7590.04596.x
EUROSION (2004). Living with coastal erosion in Europe – Sediment and
space for sustainability. 40 p, Luxemburg: Europ ean Commission.
Ferro V. G., Lemes P., Melo A. S., Loyola R., Fenton B. (2014). The Reduced
Effectiveness of Protected Areas under Climate Change Threatens
Atlantic Forest Tiger Moths. PLoS ONE, 9(9), e107792.
Frainer, A., Primicerio, R ., Kor tsch, S., Aune, M., Dolgov, A. V., Fossheim ,
M., & Aschan, M. M. (2017). Climate‐driven changes in functional
biogeogr aphy of Arctic marine fish communities. Proceedings of the
National Academy of Sciences of the United States of America, 114 (46),
12202–12207. https ://doi.org/10.1073/pnas.17060 80114
Galindo‐Leal, C., Jacobsen, T. R., Langhammer, P. F., & Olivier i, S.
(2003). State of t he hotspots: The dynamics of biodiversity loss.
In C. Galindo‐Leal, & I. G. de Câmara (Eds.), The Atlantic Forest of
South America: Biodiversity status, threats, and outlook (pp. 12–23).
Washington, DC: Center for Applied Biodiversity Science and
Gu, Z., Gu, L., Eils, R., Schlesner, M., & Brors, B. (2014). circlize imple-
ments and enhances circular visualization in R. Bioinformatics, 30,
Guisan A., Thuiller W. (20 05). Predicting species distribution: offering
more than simple habitat models. Ecology Letters, 8(9), 993–100 9.
Hannah, L., Midgley, G., Hughes, G., & Bomhard, B. (20 05). The view
from the Cape: extinction risk , protected areas and climate change.
BioScience, 55(3), 231–242.
Heller, N. E., & Zavaleta, E. S. (20 09). Biodiversit y management in the
face of climate change: a review of 22 years of recommendations.
Biological conservation, 142(1), 14–32.
Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A.
(2005). Very high resolution interpolated climate surfaces for global la nd
areas. International Journal of Climatology, 25, 1965‐1978.
Hijmans, R. J., & Elith, J. (2015). Species distribution modeling with R . R
CRAN Project, 79 pp. Retrieved from http://www.idg.pl/mirro rs/
CRAN/web/packa ges/dismo/ vigne ttes/sdm.pdf%5Cnpa pers://
e 2 1 e 3 1 4 0 ‐ c c c 7 ‐ 4 1 4 2 ‐ 9 1 1 c ‐ e 7 4 f c 5 c e e c f 7 / P a p e r / p 1 0 1 6 2
IPCC (2013). IPCC Fifth Assessment Report (AR5). IPCC , s. 10–12.
Cambridge: Cambridge University Pre ss.
IUCN and UNEP‐WCMC (2015). The World Database on Protected Areas
(WDPA). Retrieved from ww w.prote ctedp lanet.net
Joly, C. A., Metzger, J. P., & Tabarelli, M. (2014). Experiences from the
Brazilian Atlantic Forest: Ecological findings and conservation ini-
tiatives. New Phytologist, 204(3), 459–473. https ://doi.o rg /10.1111 /
Lemes, P., Loyola, R. D., & Flammini, A. (2013). Accommodating Species
Climate‐Forced Dispersal and Uncertainties in Spatial Conservation
Planning. PLoS ONE, 8(1), e54323.
Loyola, R. D., Lemes, P., Brum, F. T., Provete, D. B., & Duarte, L. D. S.
(2014). Clade‐specific consequences of climate change to amphibian-
sin Atlantic Forest protected areas. Ecography, 37(1), 65–72.
Marquardt, D. W. (1970). Generalized inverses, ridge regression, biased
linear estimation, and nonlinear estimation. Technometrics, 12(3),
M c La ch la n , J. S. , He ll m an n, J . J. , & S c hw ar t z , M . W. (2 0 0 7 ). A fr a m ew or k fo r de -
bate of assisted migration in an era of climate change. Conservation Biology,
21(2), 297–302. https ://doi.org/10.1111/j.1523‐1739.20 07.00676.x
Mews, H. A., Pinto, J. R. R., Eisenlohr, P. V., & Lenza, E. (2014). Does size
matter? Conservation implications of differing woody population
sizes with equivalent occurrence and diversit y of species for threat-
ened savanna habitats. Biodiversity and Conservation, 23(5), 1119–
1131. ht tps ://doi.org/10.10 07/s10531‐014‐0 651‐4
Mittermeier, R. A., Turner, W. R., Larsen, F. W., Brooks, T. M., & Gascon, C.
(2011). Global biodiversity conser vation: The critical role of hotspots.
In F., Zachos & J., Habel (Eds), Biodiversity hotspots (p p. 3–2 2). Sp r i nger,
Berlin, Heidelberg. https ://doi.org/10.1007/978‐3‐642‐20992‐5_1
Moss, R. H., Edmonds, J. A., Hibbard, K. A., Ma nning, M. R., Rose, S. K., Van
Vuu ren , D. P., … & Meeh l, G. A. (2010 ). The nex t g en era ti on of sc en ari os
for climate change research and assessment. Nature, 463(7282), 747.
Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B.,
& Kent, J. (2000). Biodiversity hotspots for conservation priorities.
Nature, 403(6772), 853–858. ht tps ://doi.org/10.1038/350 02501
Neves, D. M., Dexter, K. G., Pennington, R. T., Bueno, M. L., de Miranda, P.
L. S., & Oliveira‐Filho, A. T. (2018). Lack of floristic identity in campos
rupestres—A hyperdiverse mosaic of rocky montane savannas in South
America. Flora, 238, 24–31. https ://doi.org/10.1016/j.flora.2017.03.011
Neves, D. M., Dex ter, K. G., Pennington, R . T., Valente, A. S. M., Bueno,
M. L., Eisenlohr, P. V., … Oliveira‐Filho, A. T. (2017). Dissecting a bio-
diversity hotspot: The importance of environmentally marginal hab-
itats in the Atlantic Forest Domain of S outh America. Diversity and
Distributions, 23(8), 1–12. https ://doi.org/10.1111/ddi.12581
Niemi, G. J., & McDonald, M. E. (2004). Application of ecological indica-
tors. Annual Review of Ecology, Evolution, and Systematic s, 35, 89–111.
https ://doi.org/10.1146/annur ev.ecols ys.35.112202.130132
Oliveira‐filho, A. T. (2017). NeoTropTree, Flora arbórea da Região
Neotropical: Um banco de dados envolvendo biogeografia, diversidade e
conservação. Retrieved from http://www.neotr optree.info
Oliveira‐Filho, A. T., Budke, J. C., Jarenkow, J. A., Eisenlohr, P. V., & Neves,
D. R . M. (2015). Delving into the variations in tree species compo-
sition and richness across South American subtropical Atlantic and
Pampean forests. Journal of Plant Ecology, 8(3), 242–260. https ://doi.
org /10.1093/jp e/rtt058
ESSER Et al.
Oliveira‐Filho, A. T., & Fontes, M . A. L. (2000). Patterns of floris tic dif-
ferentiation among Atlantic Forests in Southeas tern Br azil and
the influence of climate1. Biotropica, 32(4b), 793–810. https ://doi.
org /10.1111/j .1744 ‐7429.200 0. tb 006 19.x
Pena, J. C . D. C., Kamino, L . H . Y., Rodrigues, M., Mariano‐Neto, E., & de
Siqueir a , M. F. (2 0 14) . As s e ssing the conser v a t ion st a t u s of sp e c i e s wi t h
limited available data and disjunct distribution. Biological Conservation,
170, 130–136. https ://doi.org/10.1016/j.biocon.2013.12.015
Peterson, A . T. (20 01). Predicting species ’ geographic distributions
based on ecologic al niche m odeling. The Condor, 103, 599–605. https
Peterson, A. T., Egbert, S. L., Sánchez‐Cordero, V., & Price, K. P. (2000).
Geographic analysis of conservation priority: Endemic birds and
mammals in Veracruz, Mexico. Biological Conservation, 93 (1), 85–94.
Peterson, A. T., Ortega‐Huerta, M. A., Bartley, J., Sanchez‐Cordero, V.,
Soberon, J., Buddemeier, R. H., & Stockwell, D. R . B. (2002). Future
projections for Mexican faunas under global climate change scenar-
ios. Nature, 416 (6881), 626–629. https ://doi.org/10.1038/416626a
Pull iam , H. R ., & Da nie lso n, B. J. (1991). Sour ces , sinks, and habit at selec-
tion: A landscape perspective on population dynamics. The American
Naturalist, 137, S50–S66. https ://doi.org/10.1086/285139
R Development Core Team (2011). R: A language and environment for sta-
tistical computing. R Foundation for Statistical Computing.
Saraiva, D. D., Santos, A. S. D., Overbeck , G. E., Giehl, E. L. H., &
Jarenkow, J. A. (2018). How effective are protected areas in con-
serving tree taxonomic and phylogenetic diversity in subtropical
Brazilian Atlantic Forests? Journal for Nature Conservation, 42, 28–35.
Scarano, F. R. (2009). Plant communities at the periphery of the Atlantic
rain fores t: Rare‐species bias and its risk s for conser vation . Biological
Conservation, 142(6), 1201–1208. https ://doi.org/10.1016/j.biocon.
Sobral‐Souza, T., Vancine, M. H., Ribeiro, M. C., & Lima‐Ribeiro, M. S.
(2018). Ef ficiency of protected areas in Amazon and Atlantic Forest
conser vation: A spatio‐temporal view. Acta Oecologica, 87, 1–7.
Tabarelli, M., Cardoso Da Silva, J. M., & Gascon, C. (2004). Forest frag-
mentation, synergisms and the impoverishment of neotropical for-
ests. Biodiversity and Conservation, 13 (7), 1419–1425. https ://doi.
Tabarelli, M., Pinto, L. P., Silva, J. M. C., Hirota, M., & Bede, L. (2005).
Challenges and opportunities for biodiversity conservation in the
Brazilian Atlantic Forest. Conservation Biology, 19(3), 695–700. https
Thuiller, A. W., Georges, D., Engler, R., Georges, M. D., & Thuiller, C. W.
(2014). Package ‘biomod2’ . Species distribution modeling within an en-
semble forecasting framework. https://CRAN. R-project. org/package=
Th ui ll er, W. (2014) . Ed it or ial comm en tar y on ‘B IO MOD – Opt im iz ing pre -
dictions of species distributions and projecting potential future shifts
under global change’. Global Change Biology, 20, 3591–3592. https ://
doi .org/10.1111/gcb.1272 8
Thuiller, W., Guéguen, M ., Renaud, J., Karger, D. N., & Zimmermann , N.
E. (2019). Uncertainty in ensembles of global biodiversity scenar-
ios. Nature Communications, 10(1), 1446. https ://d oi.or g/10 .103 8/
Ti ch y, L., & Ch ytr y, M. (2 006 ). Stat is tical det erm ina ti on of dia gno sti c sp e-
cies for site groups of unequal size. Journal of Vegetation Science, 17,
809–818 . ht tps ://doi.o rg /10.1111 /j.1654‐1103 .2 006.tb 02 5 04.x
Titeux, N., Dufrene, M., Jacob, J., Paquay, M., & D efourny, P. (2004).
Multivariate analysis of a fine‐scale breeding bird atlas using a geo-
graphical information system and partial canonical correspondence
analysis: environmental and spatial effects. Journal of Biogeography,
31(11), 1841–1856 .
Urban, M. C. (2015). Accelerating extinction risk from climate change.
Science, 348(6234), 571–573.
Walter, H. (1971). Ecology of tro pical and subtro pical vegetation. Edinburgh,
UK: Oliver & Boyd.
Werneck, F. P. (2011). The diversification of eastern South American
open vegetation biomes: Historical biogeogr aphy and perspec-
tives. Quaternary Science Reviews, 30, 16 30–1648. https ://doi.
Werneck, F. P., Nogueira, C., Colli, G. R., Sites, J. W., & Costa, G. C. (2012).
Climatic stability in the Brazilian Cerrado: Implications for biogeo-
graphical connections of South American savannas, species richness
and conservation in a biodiversity hotspot. Journal of Biogeography,
39, 1695–1706. https ://doi.org /10.1111/j.1365‐2699.2012.02715.x
Williams, K. J., Ford, A., Rosauer, D. F., De Silva, N., Mittermeier, R.,
Bruce, C., … Margules, C. (2011). Forests of East Australia: The 35th
biodiversity hotspot. In Biodiversity hotspots (pp. 295–310). Springer,
Berlin, Heidelberg. https ://doi.org/10.1007/978‐3‐642‐20992‐5_16
Zuur, A. F., Ieno, E. N., Walker, N., Saveliev, A. A., & Smith, G. M.
(2009). Mixed effects models and extensions in ecology with R.
New York: Springer Science & Busine ss Media. https ://doi.
Zwiener, V. P., Padial, A. A., Marques, M. C . M., Faleiro, F. V., Loyola, R .,
& Peterson , A. T. (2017). Planning for conser vation and restor ation
under climate and land use change in the Brazilian Atlantic Forest.
Diversity and Distributions, 23(8), 955–966. https ://doi.org/10.1111/
Luíz Fernando Esser is a PhD student in Botany at the Federal
University of Rio Grande do Sul, Brazil. He is interested in biodi-
versity evolution and environmental niche evolution, focusing on
what influences the establishment of plant communities in space
and time. Danilo M. Neves is a professor of macroecolog y at the
Federal Universit y of Minas Gerais, Brazil. He is interested in the
evolutionary dimension of community ecology, with an empha-
sis on historical biogeography of terrestrial biomes. João André
Jarenkow is a research professor at the Federal University of Rio
Grande do Sul. His main research interests include phytogeogra-
phy and plant community ecology of southern Brazilian Atlantic
Author contributions: L .F.E. conceived the ideas, applied the
methodology and led the writing; D.M.N. contributed for the im-
provement of writing and the methods; J.A .J. improved the writ-
ing and supervised the first author.
Additional supporting information may be found online in the
Suppor ting Information section at the end of the article.
How to cite this article: Esser LF, Neves D, Jarenkow JA.
Habitat‐specific impacts of climate change in the Mata
Atlântica biodiversity hotspot. Divers Distrib. 2019;25:1846–
1856. htt ps ://doi.or g/10.1111/ddi .12984