Ecological Applications, 26(6), 2016, pp. 1854–1867
© 2016 by the Ecological Society of America
Threshold effect of habitat loss on bat richness in cerrado-­forest
landscapes
Renata L. Muylaert,1,3 Richard D. Stevens,2 and Milton C. Ribeiro1
1
Department
2
of Ecology, Universidade Estadual Paulista (UNESP), 24A Av., 1515, 13506-900, Rio Claro, Brazil
Department of Natural Resources Management, Museum of Texas Tech University, Lubbock, Texas 79409 USA
Abstract.
Understanding how animal groups respond to contemporary habitat loss
and fragmentation is essential for development of strategies for species conservation. Until
now, there has been no consensus about how landscape degradation affects the diversity
and distribution of Neotropical bats. Some studies demonstrate population declines and
species loss in impacted areas, although the magnitude and generality of these effects on
bat community structure are unclear. Empirical fragmentation thresholds predict an accentuated drop in biodiversity, and species richness in particular, when less than 30% of the
original amount of habitat in the landscape remains. In this study, we tested whether bat
species richness demonstrates this threshold response, based on 48 sites distributed across
12 landscapes with 9–88% remaining forest in Brazilian cerrado-­forest formations. We also
examined the degree to which abundance was similarly affected within four different feeding guilds. The threshold value for richness, below which bat diversity declines precipitously,
was estimated at 47% of remaining forest. To verify if the response of bat abundance to
habitat loss differed among feeding guilds, we used a model selection approach based on
Akaike’s information criterion. Models accounted for the amount of riparian forest, semideciduous forest, cerrado, tree plantations, secondary forest, and the total amount of
forest in the landscape. We demonstrate a nonlinear effect of the contribution of tree
plantations to frugivores, and a positive effect of the amount of cerrado to nectarivores
and animalivores, the groups that responded most to decreases in amount of forest. We
suggest that bat assemblages in interior Atlantic Forest and cerrado regions of southeastern
Brazil are impoverished, since we found lower richness and abundance of different groups
in landscapes with lower amounts of forest. The relatively higher threshold value of 47%
suggests that bat communities have a relatively lower resistance to habitat degradation
than other animal groups. Accordingly, conservation and restoration strategies should focus
on increasing the amount of native vegetation of landscapes so as to enhance species
richness of bats.
Key words: cerrado; Chiroptera; conservation; environmental gradient; fragmentation threshold; habitat
amount; interior Atlantic Forest; landscape resilience.
Introduction
Changes in landscape attributes such as the amount
of forest cover can alter species composition (Fahrig
2013, Dirzo et al. 2014) and are thought to be relevant
to the maintenance of biodiversity and integrity of
­ecological processes. Moreover, species can respond idiosyncratically to anthropogenic impacts. Thus, to better
conserve or restore particular aspects of the contemporary biota, it is imperative to determine which species
in different taxonomic groups most respond to anthropogenic alterations and how they persist in fragmented
environments (Sutherland et al. 2013). Information
about ecological aspects such as community structure
and species’ responses to landscape changes (Banks-­Leite
Manuscript received 1 October 2015; revised 29 December
2015; accepted 1 January 2016; final version received 19
February 2016. Corresponding Editor: J. Goheen.
3
E-mail: [email protected]
et al. 2014) can be obtained through examination of
response curves to landscape characteristics, which are
not always linear (Fahrig 2003). Based on such information, it may be possible to develop more efficient
­conservation strategies. Knowledge of effects of anthropogenic impacts on the contemporary biota has direct
implications to defining strategies of landscape management for conservation and habitat restoration
(Rodrigues et al. 2009, Tambosi et al. 2013).
Variation in habitat diversity, connectivity, and total
amount of habitat have direct effects on species
­distribution and diversity within heterogeneous landscapes (Estavillo et al. 2013). However, among these
different landscape properties, quantity of remaining
habitat is claimed to be the characteristic that best
explains ­ecological responses related to ecosystem functions for many organisms (Fahrig 2013). Often there
are threshold responses describing the relationship
between amount of habitat and other biological
1854
September 2016
BAT RICHNESS DECLINES WITH FRAGMENTATION
variables, such as number of species or community composition (Fahrig 2002, Banks-­Leite et al. 2014), whereby
precipitous changes occur beyond a certain level of
reduction of habitat.
Different types of habitat are used by Neotropical bats,
and several highly mobile species may persist in fragmented landscapes because of such heterogeneity
(Bernard and Fenton 2003, Montiel et al. 2006, Meyer
et al. 2008). Some species can traverse such heterogeneity
and benefit from anthropogenically modified habitats
around fragments of native vegetation (Fahrig and
Merriam 1985, Cisneros et al. 2014). However, there is
no consensus about the influence of landscape structure
and habitat fragmentation on assemblage characteristics
of Neotropical bats, such as species richness or composition. In fact, studies have recorded positive effects
(Medellín et al. 2000, García-­Morales et al. 2013), negative effects (Fenton et al. 1992, Brosset et al. 1996), or
even an absence of significant effects (Bernard and
Fenton 2007) of landscape structure and fragmentation
on characteristics of biodiversity. Such inconsistency
among studies (Cunto and Bernard 2012, García-­García
et al. 2014) precludes the generalization of a common
pattern of bat responses to habitat loss. Moreover,
studies in different regions have demonstrated population decline, species loss, and turnover in impacted
areas, although the magnitude of these effects on community structure is highly variable (Brosset et al. 1996,
Jones et al. 2003, Safi and Kerth 2004, García-­Morales
et al. 2013). Despite this, bats represent a promising
group for study of responses to fragmentation, due to
their diversity, high abundance, and relative sampling
ease (Medellín et al. 2000).
Among bats that exhibit the greatest response to landscape change, the guild of gleaning animalivores (e.g.,
Phylostominae) seems to be highly edge-­sensitive, and
responds negatively to disturbance and forest fragmentation (Fenton et al. 1992, Medellín et al. 2000, Meyer
et al. 2008, Farneda et al. 2015). Such a pattern may not
be broadly pervasive, since a positive and significant
association with edge density was found for bats in
Peruvian Amazonia (Klingbeil and Willig 2009).
However, those results also deserve a cautious interpretation since the broad context of the studied area (Iquitos,
Peru) was that of recent fragmentation embedded in a
large block of continuous forest. Larger regional landscapes that are highly fragmented and that have been
affected for a long period may not exhibit the same kinds
of patterns as systems that have been recently fragmented
or are embedded in a larger region of continuous forest.
For this reason it is necessary to better understand
responses of phyllostomine bats to a variety of landscape
attributes.
Considering that fragmentation and loss of natural
habitats can lead to irreversible changes in biodiversity
and species interactions (Chapin et al. 2000), the concept
of fragmentation thresholds may be helpful to inform
conservation initiatives. Thresholds arise when “small
1855
changes in an environmental driver produce large
responses in the ecosystem” (Groffman et al. 2006). Thus,
a fragmentation threshold is a breaking point below
which effects of reductions in amount of habitat become
more exaggerated than above the threshold (Huggett
2005). For example, the abundance of birds and non-­
flying mammals is more affected by changes in landscapes with less than 30% of the amount of habitat
(Andrén 1994, Banks-­
Leite et al. 2014) than those
occurring in landscapes above this threshold, probably
due to effects of isolation or lower resource availability
in areas of high fragmentation (Andrén 1994, Pardini
et al. 2010). This breaking point is referred to as Andrén’s
fragmentation threshold (Pardini et al. 2010). Identifying
whether there are fragmentation thresholds for other
animal groups, such as bats, is fundamental to better
understand ecosystems in a fragmented world, where
conservation decisions often are influenced by existing
habitat characteristics (Huggett 2005).
Substantial differences in bat species composition
between heterogeneous forested habitats and adjacent
non-­forested areas such as pastures and forest edges have
been observed in many systems (Estrada et al. 1993,
Swihart et al. 2006, Medina et al. 2007). Furthermore,
bats use different landscape elements to facilitate
movement while they forage (e.g., Fahrig and Merriam
1985), but they may actually live in forests more frequently than in other landscape elements (Estrada et al.
1993, Bernard and Fenton 2003). This reinforces the
potential for a fragmentation threshold, with higher bat
species richness in areas of higher amounts of habitat,
with an expected abrupt decrease in richness below a
certain amount of habitat.
Here, we evaluated whether there was a threshold of
bat species richness as a function of habitat loss. Then,
we investigated the relative contributions of quantity of
different habitat types in the landscape to bat abundance
(i.e., total number of individuals) within feeding guilds.
In particular, we were interested in whether accounting
for individual habitat types was as effective as using just
total amount of habitat when trying to explain abundance within guilds of frugivores, nectarivores, animalivores, and sanguivores. We predicted that: (1) there is a
threshold effect on richness as a function of amount of
habitat, and (2) different bat guilds will respond differently to variation in amounts of different habitat types
within fragmented landscapes.
We expected a threshold model would best explain the
relationship between percentage of forest and richness of
bats than linear or additive models (Fig. 1). We also
expected that frugivore abundance would be best
explained by multiple habitat types, and to exhibit a
hump-­
shaped pattern, with increasing abundances at
intermediate levels of total amount of habitat, probably
related to higher amounts of edge in the landscape
(Fahrig 2003, Bolívar-­Cimé et al. 2013). We expected that
abundances of nectarivores and animalivores would be
better explained by amount of native habitat, since these
1856
RENATA L. MUYLAERT ET AL.
Ecological Applications
Vol. 26, No. 6
Fig. 1. Working hypotheses for bat diversity in response to the amount of forest within fragmented landscapes of southeastern
Brazil.
bats roost in these areas and seem to be more sensitive
to habitat loss (Meyer et al. 2008). We expected that
sanguivore (Desmodus rotundus) abundance would not
be affected by amount of native habitat since this species
is common in both disturbed and forested areas (Fenton
et al. 1992).
Methods
Study area
We adopted the term “cerrado” as the set of forestl­ ike savannah formations together with riparian forests
and other forest physiognomies, such as semideciduous
interior Atlantic Forest (Coutinho 2006). The Cerrado,
the second largest Brazilian ecoregion and one of
world’s conservation hotspots (Ratter et al. 1997), is a
highly human-­modified ecoregion, being restricted to
less than 1% of its original area in São Paulo State
(Kronka et al. 2005). As a consequence of anthropogenic activity, biodiversity of cerrado is highly
threatened; only 20% of its natural areas in Brazil
remain undisturbed, and only 1.2% is within protected
areas (Myers et al. 2000).
Landscape selection and calculation of amount of habitat
The study region is composed of a distributed network
of very small patches (Durigan et al. 2004). More than
80% of patches are less than 50 ha (Ribeiro et al. 2009).
Nonetheless, the region also harbors the largest protected area of cerrado in the state of São Paulo, located
at the Jataí Ecological Station (21°33′ S, 47°45′ W), with
~9000 ha. Therefore, we had a wide range of habitat
patch sizes available for our study. We selected 15
­circular 2.5 km radius landscapes along a gradient from
9% to 88% remaining amount of forest (Table S1,
Fig. 2) centered at the sampling sites’ centroid. We
selected four control landscapes (areas 8, 10, 13, and
14), that were located in two protected areas: the Jataí
Ecological Station and Porto-­Ferreira State Park. These
control areas were selected because they were protected
areas and because they were composed of predominantly remnant forest cerrado (cerradão). The surrounding areas around all sampled landscapes included
mostly sugarcane plantations, but also ­
forestry
(Eucalyptus spp. and Pinus sp.), and less f­requently
fruit crops (Citrus sp.), cattle pastures, dams, and
urban areas. The region harbors different vegetation
September 2016
BAT RICHNESS DECLINES WITH FRAGMENTATION
1857
Fig. 2. Spatial distribution of 15 landscapes with the scales represented as buffers with dimensions of 2.5, 2, and 1 km in
southeastern Brazil. Green areas are vegetation patches, and the percent of forest coverage is at the upper left of each landscape.
The geographic coordinates of the centroids are presented in S1. Coordinates are in UTM.
formations, mostly cerradão forests (a woodland
cerrado, Coutinho 1978), semideciduous forests, and
riparian forests. The study landscapes were centered on
cerradão fragments.
The climate is tropical of central Brazil (Cwa according
to Köppen classification) and the studied locations are
in the subtropical moist forest life zone according to the
Holdridge classification (Holdridge 1947). Mean temperatures of sampled sites are reported in Table S1. The
region has two defined seasons, one hot and rainy from
October to mid-­
April, and the other cold and dry,
between April and mid-­September.
We used forest cover as a surrogate for amount of
habitat, hereafter amount of forest. Amount of forest is
a measure of landscape composition, where the total
quantity of remaining forested vegetation is estimated,
irrespective of other landscape characteristics (quantity
of different habitats, size and distribution of fragments,
amount or shape of edge). Here, we used amount of
forest to inversely represent amount of landscape degradation and habitat loss by anthropogenic impacts: the
higher the amount of forest in a landscape, the lower the
landscape degradation (Fahrig 2003). Using a vegetation
cover map, amount of forest was estimated for each landscape and divided by the area within the buffer. Metrics
were calculated using Quantum GIS 2.0.1 (QGIS
Development Team 2014).
We also characterized amount of different vegetation
classes in each buffer. A land cover map was derived from
a georeferenced Landsat satellite image (acquired in
2010, spatial resolution of 30 m, cloud free). Our landscape classification procedure followed three steps:
(1) identification of vegetation types according to the
map of remaining vegetation of São Paulo State (resolution of 30 m, Kronka et al. 2005); (2) map editing; and
(3) field validation. Map editing used high-­resolution
(1 m) images available from Google Earth via the Open
Layer plug-­in in QGIS 2.0.1 (cartographic data from
2013; QGIS Development Team 2014), where different
vegetation types were classified. The working scale using
Open Layer was set to 1:8,000. Field validation consisted
of visits to all sampling sites during the day, marking
points with the real land cover type in situ, and then
confirming the mapped areas classification. Vegetation
classes used for analysis were: native vegetation (sum of
cerrado, secondary, semideciduous, and riparian forest),
cerrado, semideciduous forest, riparian forest, tree plantations (mainly Eucalyptus spp.), and matrix (open
anthropogenic areas, such as sugarcane, pasture, and
agriculture).
1858
RENATA L. MUYLAERT ET AL.
Sampling of bat diversity
We conducted the fieldwork under research permits
granted by the Chico Mendes Institute for Conservation
and Biodiversity (SISBIO 31163-­1, 35901-­1), São Paulo
Forest Institute (IF/COTEC 260108-­007.043 for sampling
in protected areas), São Paulo State University (Ethics
Commitee) and Campus Administration of the Federal
University of São Carlos (022/07 DISG/PU). For the
capture and handling of bats we followed the Guidelines
of the American Society of Mammalogists (Sikes et al.
2011). Bats were identified to the species level in the field
using a combination of taxonomic keys (Vizotto and
Taddei 1973, Gardner 2008). Vouchers were deposited in
the mammal collection coordinated by Adriano Peracchi
(Federal University of Rio de Janeiro UFRRJ).
In each landscape we sampled four points in similar
types of habitat patches (mosaic-­level sampling sensu
Bennett et al. 2006). Two points were sampled in winter
(2012) and two in summer (2013). Within each landscape,
sampling points were separated by at least 50 m
(mean = 287 ± 107 m). In each season one sampling point
was set inside the cerrado patch and the other was set at
the nearest edge in contact with a sugarcane matrix (Plate
S1). We chose sampling points based on vegetation
structure, always preferring to set nets on trails close to
denser cerrado formations (with canopy height higher
than 7 m; sensu Ribeiro and Walter 2008). We estimated
canopy height for each sampling point with telescopic
sticks (8 m). Edge nets were also set based on vegetation
structure and deployed 1 m from the boundary of native
vegetation. We conducted five capture nights per landscape (three nights in summer, two in winter). In each
sampling night we deployed 12 mist nets (model 716/7P,
12 × 3 m; denier 70/2, mesh 16 × 16 mm; Ecotone,
Gdynia, Poland) that remained open for 6 h from sunset,
the period when Neotropical bats are more active (Aguiar
and Marinho-­Filho 2004). We did not sample on nights
with full moons or heavy rains because Neotropical bats
can exhibit lunar-­phobia (Mello et al. 2013) or cease
activity during rain (Thies et al. 2006). Sampling effort
was calculated by multiplying the area and total time
mist nets were open (Straube and Bianconi 2002).
We did not consider differences in species-­
capture
probabilities. Although mist nets are selective for phyllostomids (Kunz and Parsons 2009), the capture of insectivorous bats from other families is facilitated in Cerrado
due to its simplified vertical structure (Aguirre 2002,
Zortéa and Alho 2008). We included all bats captured
when determining species richness. To assess differences
between captures in rainy and dry season, we performed
a paired t test (Zar 1999).
Data analyses
All sites were sampled with the same effort. Sample
completeness and nestedness allowed us to estimate sampling bias of particular landscapes in terms of estimates
of species richness. Differences between rare and common
Ecological Applications
Vol. 26, No. 6
taxa in terms of abundance can translate into methodological difficulties in determining species richness when
samples are not represented by the same number of individuals, even when equal effort is employed for each
sample (Gotelli and Colwell 2001). Samples from areas
of high species richness may have low sample completeness but not necessarily compromise assessments of
patterns of species richness. As long as rank order among
sites is maintained, such a sampling effect will only make
comparisons conservative. In nested assemblages, those
sites whose estimate of species richness is compromised
by a sampling bias will fall out toward the bottom of the
ranked nested-­matrix (Blüthgen 2010). To better identify
landscapes in which the estimate of species richness may
have been affected by sampling bias we used a combination of completeness and nestedness. Thus, for the
analysis involving Andrén’s fragmentation threshold, we
excluded landscapes that had a sample completeness of
less than 75% and fell in the lower half of the ranked
nested matrix. We determined sample completeness
(Magurran 2004) in the studied landscapes (Table S1,
Fig. S1) by dividing observed species richness from sampling by the mean of the first and second order jackknife
species richness estimates. We used the NODF index
(Fig. S2, Almeida-­Neto et al. 2008) to examine assemblage nestedness using the software ANINHADO
(Guimarães and Guimarães 2006) with nestedness
significance calculated by Monte Carlo simulations
­
(999). Based on these criteria, we excluded sites 7, 9, and
15 in the test of Andrén’s fragmentation threshold.
Scale selection
Scale dependence can affect the response of animal
abundance to landscape characteristics (Fortin and Dale
2005). Because of differential mobility, different guilds can
respond to landscape characteristics measured at different
scales (Gorresen et al. 2005, Avila-­Cabadilla et al. 2012).
We used a scale-­selection procedure to choose the best
scales of response for each dependent variable as a function
of the total amount of forest and the other vegetation
classes. Here, we define scale as the spatial extent of a
measured landscape (buffer size), and use this as a surrogate for extent of realized dispersal by bats within that
area (Wu and Hobbs 2007). For scale selection, a total of
seven models were analyzed per spatial scale and per
response variable (null model, 0.5, 1, 1.5, 2, 2.5, 5 km).
Before model fitting and scale selection we conducted a
correlation analyses between all landscape variables to
check for between-­scale collinearity. Landscape variables
included in scale selection were not strongly correlated for
a particular scale (with Pearson’s r < 0.7; Zuur et al. 2009).
This was not the case for the same variable at different
scales. For example total amount of cerrado measured at
0.5 km is correlated with total amount of cerrado at 1 km.
Thus, we corrected for this effect by reducing collinearity
of the same metric at different scales. To reduce collinearity, we recasted each variable as a linear combination of
September 2016
BAT RICHNESS DECLINES WITH FRAGMENTATION
the other scales for the same variable (Zuur et al. 2009).
We did not apply any transformation to the data set for
predictor variables or for abundance. We used multivariate model selection with generalized additive models
(GAM) with smoothness selection (Hastie 2013). As
response variables were counts, we modeled errors as
Poisson distributed with a log-­link function.
Estimating richness thresholds related to amount of forest
To examine a response to fragmentation, we used
species richness as the dependent variable and total
amount of forest as the independent variable. We used
data on amount of forest at a 2 km spatial scale to
conduct model selection. The model selection procedure
contrasted four types of models: linear, piecewise,
forest + abundance, and null (Fig. 1). We also constructed generalized additive models whereby different
classes of vegetation were independent variables and
species richness was the dependent variable.
We used piecewise regression to evaluate Andrén’s
(1994) fragmentation threshold. Piecewise regression
models use more than one line to fit data, and use a break
point to unite regression lines (Toms and Lesperance 2003).
In our case, the break point represents a fragmentation
threshold whereby further reductions in amount of forest
translate into precipitous declines in bat species richness.
We conducted a piecewise regression based on a generalized linear model using an iterative fitting procedure to
estimate model parameters (Muggeo 2003). Values of
parameters are fitted repeatedly until estimates lead to
the maximum likelihood. We used the R language
(R Development Core Team 2014) and the packages bbmle
(Bolker and R Development Core Team 2014) and gam
(Hastie 2013). Standard errors (SEs) and confidence
intervals (CIs) of break points were obtained by the
package segmented (Muggeo 2014). We identified the most
plausible models primarily considering model weights
(wi; but report ΔAICc), selection frequencies (πi), and R2.
Contribution of different habitat types to bat abundance
within feeding guilds
In the Neotropics, mist nets primarily capture phy­
llostomid bats. Therefore, we analyzed only the following
feeding guilds (adapted from Gardner 1977): frugivores,
animalivores (subfamily Phyllostominae), nectarivores,
and sanguivores. We used total number of captures as a
surrogate for bat abundance within feeding guilds.
For frugivorous, nectarivorous, and animalivorous
bats, we used a 1-­km scale and for sanguivores we used
the 2.5-­km scale when analyzing effects of amounts of
different vegetation classes. We chose those scales
because they were the most plausible scales for explaining
abundance of those feeding guilds (Table S2).
Eight models were constructed to examine if amount
of individual vegetation classes (their proportion), or
their summed values (amount of forest) better explained
1859
abundance within each guild (i.e., the total number of
individuals of that guild irrespective of species affiliation). Explanatory variables included in each model
were amount of each vegetation class that potentially
affected abundance within guilds: cerrado, riparian
forest, semideciduous forest, tree plantations, secondary
forest, native vegetation, amount of forest, and the null
model. The null model represented a model of no effect,
consisting of a random number sequence as the explanatory variable. The total amount of forest model represents a scenario where abundance is influenced primarily
by total forest amount, irrespective of the composition
of different habitat types giving rise to that total. The
best single-­habitat predictor of bat abundance within
guilds was the amount of cerrado (Fig. S3; see Pearson’s
correlation among habitat types and amount of forest).
We used those types of habitat that did not exhibit high
collinearity in the analysis, with the exception of cerrado
and native forest. Thus, we constructed 32 models (eight
models for each of four guilds).
We selected models based on the Akaike’s information
criterion corrected for small sample sizes (AICc; Burnham
and Anderson 2002). As abundance within guilds may
exhibit nonlinear responses to landscape attributes
(Suarez-­
Rubio et al. 2013), we applied generalized
additive models assuming a Poisson distributed error
term. We consider as best models those with the best
predictive accuracy according to AICc weights (evidence
strength) of the plausible models selected (ΔAICc = 0),
which were based on restricted maximum likelihood estimates. To verify ­prediction strength, we used selection
frequencies (πi, Burnham and Anderson 2002) assessed
with data sets bootstrapped 10 ,000 times. The πi values
are sums of the number of times each model was selected
as the most plausible, divided by 10, 000.
Model validation and spatial autocorrelation
Plausible models were validated with residual analysis,
verifying homogeneity and normality of regression
residuals. We selected landscapes distant from each other
to minimize spatial autocorrelation. To verify if there was
significant spatial autocorrelation we followed Fortin and
Dale (2005) using the Pearson’s residuals of the plausible
models to build a Moran’s I correlogram and test significance of spatial autocorrelation using 10, 000 permutations and a 5% threshold for significance. These analyses
were conducted using SAM 4.0 (Rangel et al. 2010).
Significant autocorrelation was not detected in model
residuals; thus we did not include additional model terms
to account for it.
Results
Andrén’s fragmentation threshold
We employed a total effort of 162 ,000 m2/h resulting
in 1,482 captures from 31 species. Ninety-­two percent
1860
RENATA L. MUYLAERT ET AL.
(n = 1,362) of captures were phyllostomids (19 species,
Table S3). Most sites were adequately sampled (i.e.,
observed richness/estimated richness > 0.75). There were
no differences in captures within sites between summer
and winter (paired t test: t = −1.87, P = 0.07). Assemblages
were significantly nested (NODF = 68.76, P < 0.0001,
Table S1, Figure S2).
Bats in the Brazilian Cerrado indeed exhibit Andrén’s
threshold. Species richness was best explained by total
amount of forest (wi = 0.32), with minor positive contributions of amount of cerrado, riparian, and semideciduous
forest (Table 1). Of all species recorded, 44% (n = 12) were
found only in landscapes above the theoretical threshold
(the threshold value, ψ = 30% of habitat, Andrén 1994),
whereas four species were found only in areas below the
lower level of Andrén’s threshold (10%, Andrén 1994). Of
the models presented in Fig. 1 hypothesizing the relationship between species richness and amount of forest,
the most plausible was the piecewise model (wi = 0.86, πi
= 0.99, Table 2). The estimated threshold was 47.81% of
amount of forest (Fig. 3, Table S4). The piecewise model
had ΔAICc equal to zero and weight of 0.92, being a better
predictor of the relationship between richness and
amount of forest than the linear model (ΔAICc = 6.7, wi
= 0.03). Null and total forest + abundance models were
weak. The piecewise model did not show significant
Table 1. Generalized additive models explaining bat richness
as a function of amount of different vegetation classes and
their respective evidence weight.
Model
Total forest
Cerrado
Riparian
Semideciduous
Secondary
Forestry
Native
Null
ΔAICc
K
wi
πi
0
0.7
2
2.1
2.8
3.6
3.6
11.2
2
3
3
3
3
3
3
2
0.32
0.22
0.11
0.11
0.07
0.05
0.05
0.001
0.9987
0
0
0
0
0
0
0.0013
Ecological Applications
Vol. 26, No. 6
spatial autocorrelation of residuals (Table S5, Fig. S4).
Landscapes above the threshold had a median of the total
number of species per landscape that was 50% higher
than that of landscapes below the threshold (above
threshold landscapes = 15 ± 1.32 species, below threshold
= 10 ± 1.08 species, means and SE, respectively, Fig. 4).
Fig. 3. Piecewise regression model demonstrating that there is
a threshold in bat richness as a function of amount of forest. The
estimated break-point, is 47% (Pψ = 0.03, SEψ = 7.51) of amount
of forest remaining in landscape and is significant at 5% threshold
for significance. There is a rapid decline of bat richness after the
breaking point. The study was conducted among 12 landscapes in
southeastern Brazil. Dashed lines indicate 95% CIs of fitted values.
The line at the bottom shows the 95% interval around the
estimated breaking point.
Notes: The best models (AICc = 0) are in bold and the plausible models (AICc < 2) are in bold and italics. K represents number of estimated parameters, AICc is the corrected Akaike
Information Criterion, ΔAICc is the Akaike difference, wi is the
Akaike weight, and πi is selection frequency (10 000
bootstraps).
Table 2. Models describing the relationship between bat richness in response to amount of forest sampled within 12 landscapes of forest–savanna formations in southeastern Brazil.
Model
ΔAICc
K
wi
R2
πi
Piecewise
Linear
Habitat + abundance
Null
0
6.7
9
17.7
5
4
3
2
0.86
0.08
0.05
<0.001
0.34
0.21
0.22
0.016
0.99
0.01
0
0
Notes: The best model (ΔAICc = 0) is in bold. Model descriptors are the same as in Table 1.
Fig. 4. Boxplot with median values and standard errors
showing total richness per landscape, divided in above-­threshold
landscapes and below-­predicted threshold landscapes.
BAT RICHNESS DECLINES WITH FRAGMENTATION
September 2016
Contribution of different habitats types to bat abundance
The contribution of different habitat types to variation
in bat abundance varied among guilds (Table 3, Fig. 5).
For both nectarivorous and animalivorous bats, the best
habitat model included amount of cerrado (wi = 0.72 and
wi = 0.98, respectively). The response of sanguivores
(only represented by Desmodus rotundus) to all habitat
types was weak, and the only plausible model included
amount of semideciduous forest (wi = 0.45). Frugivorous
1861
bat abundance responded to amount of tree plantations,
and the relationship was hump-­shaped (wi = 1.00, Fig.
S5).
Discussion
Our results suggest that bats respond precipitously to
forest loss. Moreover, different habitat types influenced
abundance of bats within guilds differently. This indicates that among bat guilds there are different sensitivities to habitat loss. Andrén’s threshold (1994) was
corroborated in the present study, and the amount of
Table 3. Generalized additive models describing the relation- available habitat needed to sustain bat diversity is subship between abundance within bat guilds in response to the
amount of forest sampled within 15 landscapes of forest-­ stantially more than for other vertebrates. Different
savanna formations in southeastern Brazil.
habitat types did not seem to affect richness when considered separately. Instead, what matters for increased
wi
Model, by response
πi
ΔAICc
bat richness is greater amount of forest within the landvariable (abundance)
scape, considering all kinds of habitat represented.
and spatial scale
1 km
Frugivores
Forestry
Secondary
Cerrado
Riparian
Total forest
Native
Semideciduous
Null
Nectarivores
Cerrado
Native
Total habitat
Riparian
Semideciduous
Forestry
Secondary
Null
Animalivores
Cerrado
Secondary
Semideciduous
Native
Total forest
Riparian
Forestry
Null
2.5 km
Sanguivores
Semideciduous
Forestry
Total forest
Cerrado
Riparian
Native
Secondary
Null
0
166.2
327.4
342.8
481.6
548.8
692.2
817
1
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.93
0.05
0.01
0.01
0
0
0
0
0
1.9
22.1
227.3
230.9
244.4
258.7
291.4
0.72
0.28
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.36
0.2
0.02
0.33
0.04
0.03
0
0
0
8.3
11.5
12.5
29.3
43.1
51.3
51.6
0.98
0.02
0
0
<0.001
<0.001
<0.001
<0.001
0.6
0.27
0.01
0
0.04
0.06
0
0
0
3.2
3.9
4.5
6.7
8.2
10.9
13.7
0.66
0.13
0.1
0.07
0.02
0.01
0
<0.001
0.45
0.44
0.05
0.04
0
0.03
0
0
Notes: The best model (ΔAICc = 0) is in bold. See model coefficients in Table S4. Model descriptors are as in Table 1.
Richness, amount of forest, and the fragmentation
threshold
Our calculated threshold of 47% amount of forest
needed to maintain richness is above the threshold proposed by Andrén (1994). It is also above the threshold
found for other groups. For example, for small mammals,
the value was 30% of remaining forest (Estavillo et al.
2013), and was between 30% and 50% for understory birds
in Atlantic Forest (Martensen et al. 2012). Other studies
have failed to corroborate the threshold hypothesis at all
(grassland invertebrates in Parker and MacNally 2002,
bird and lizard richness in Lindenmayer et al. 2005).
From a community perspective, fragmentation can
have a strong influence on persistence of biodiversity and
can be a major cause of extinction (Fahrig 2002). Most
studies on thresholds in response to habitat loss have
been performed on particular taxonomic groups
examined in isolation (Andrén 1994), i.e., woodland
birds (Radford et al. 2005, Betts et al. 2007). Since the
loss of species can be preceded by loss of interactions
(Estes et al. 2011), it would be interesting to investigate
fragmentation thresholds from the perspective of key
ecological processes, by focusing on more inclusive
groups such as all pollinators (Aguilar et al. 2006),
instead of focusing on specific, narrow taxonomic groups.
Given such an approach, consistency of the threshold
concept, and mechanisms driving responses to habitat
loss could be better understood.
The threshold above that proposed by Andrén (1994),
is an alarming result since even for bats, a group with
high mobility, there is a precipitous decline in species
number and abundance in more degraded landscapes.
We believe bat assemblages in these landscapes, fragmented for more than 40 years (Durigan et al. 2007),
might be impoverished in the number of sensitive species,
compared to more-­preserved areas. For example, we can
contrast our study with that of Zortéa and Alho (2008),
conducted in a well preserved cerrado area in the
1862
RENATA L. MUYLAERT ET AL.
Ecological Applications
Vol. 26, No. 6
Fig. 5. Plausible nonlinear models (GAM) explaining bat abundance within 15 fragmented landscapes in Brazil in a gradient of
amount of forest (forestry refers to plantations, often of Eucalyptus spp.). Landscape metrics were measured at a 1-­km scale for
frugivorous, nectarivorous, and animalivorous bats, and at a 2-­km scale for sanguivores bats. The values of significance (threshold
for significance at 5%), with P for smooth’s parametric effects, F for smooth’s parametric F test, and R2 are: (A) frugivores P < 0.001,
F = 65.73, R2 = 0.42; (B) nectarivores P < 0.001, F = 34.67, R2 = 0.26; (C) Phyllostominae bats P = 0.04, F = 4.07, R2 = 0.33; (D)
sanguivore P < 0.001, F = 10.88, R2 = 0.02.
center-­west of Brazil. In a total of 60 nights of capture,
25 bat species were registered, and eight species were
phyllostomines (Zortéa and Alho 2008). In our study we
registered only five species of phyllostomine bats, and we
sampled more areas (n = 15), with greater effort. We
performed 75 capture nights vs. 60 nights and sampled
for more hours per night than did Zortéa and Alho
(2008). Moreover, among the 16 species of Phyllostominae
that could occur in the cerrado domain (Gardner 2008),
we only captured five species with our considerable sampling effort. Bat communities in to interior of São Paulo
State are impoverished in comparison to well-­preserved
cerrados. Restoration strategies should seek to increase
the amount of habitat in fragmented landscapes, since
much of this region presents low or moderate landscape
resilience (Tambosi et al. 2013).
Meyer et al. (2008) found that habitat loss (reduction
in amount of forest) and not fragmentation (increase in
the number and change in the distribution of fragments)
is the main process after isolation (between fragments)
that influences phyllostomid responses on islands. Here,
we did not account for other fragmentation effects, since
fragmentation per se (i.e., number of patches) was correlated with amount of forest. Our results do highlight
September 2016
BAT RICHNESS DECLINES WITH FRAGMENTATION
the importance of amount of habitat in explaining bat
biodiversity in fragmented landscapes. This agrees with
a recent review by Fahrig (2013), who argued that in most
studies, the effects of habitat fragmentation per se (i.e.,
independent of habitat loss) are absent, too weak to be
detected, or may only become apparent at low amounts
of habitat. This suggests that conservation efforts that
solely focus to minimize fragmentation effects for a given
amount of habitat may often be inadequate. Moreover,
the process of anthropogenic degradation may only
proceed in a fashion whereby fragmentation and habitat
loss are correlated. It may be unrealistic to model or
experiment as if these are two independent effects. For
bat conservation efforts, focusing on amount of habitat
and the concept of thresholds is a simple and effective
means of detecting landscapes that still have considerable
conservation value. In addition, landscape composition
rather than configuration may be driving biodiversity in
these landscapes (sensu Fahrig et al. 2003). Configuration
may be more important in low-­
resilience landscapes,
where restoration initiatives would not be worth the
effort when compared to moderate-­resilience landscapes
(Tambosi et al. 2013). Furthermore, high vulnerability
to fragmentation in bats is associated primarily with high
sensitivity to edge habitats (Meyer et al. 2008). This is a
menacing observation since moderately fragmented
landscapes present the highest amounts of edge (Fahrig
2003).
Relative contributions of different habitat types to
abundance within feeding guilds
The present study suggests that variation in amount of
different habitat types is perceived similarly by animalivores and nectarivores. In contrast, frugivores responded
differently to habitat loss in terms of abundance. This
emphasizes the role of other aspects of landscape structure
such as landscape heterogeneity that may determine bat
abundance within particular guilds. Animalivore and nectarivore guilds tend to respond more positively to higher
amounts of native vegetation, and not as strongly to forestry and secondary forests (wi values in Table 3). Since
heterogeneity seems to be important for the persistence
of frugivores, the quality of habitats should be taken into
account when protected areas are planned, modified, or
created in sugarcane-­dominated landscapes.
Frugivorous bat abundance was best explained by
amount of tree plantations, with higher abundances at
intermediate amounts; other models were less plausible.
Thus, frugivores may be benefiting from exotic forest
plantations such as Eucalyptus spp., which frequently
have a native understory with many chiropterochoric
plants, such as Cecropia pachystachya, Piper aduncum,
and Solanum variabile (Silva et al. 1995, Muylaert et al.
2014, personal observation). This pattern probably is
related to higher amounts of edge in landscapes at intermediate levels of forestry or amount of forest (see plot in
Figure S4; Fahrig 2003). The persistence of frugivorous
1863
bat species may not be as influenced by amount of native
vegetation and this suggests some adaptability to environmental changes, at least over the short term.
Nectarivores, being present at all sites, were most
strongly affected by cerrado loss (Fig. 5). Despite a tendency of increase of small phytophagous species in more
fragmented areas reported by Farneda et al. (2015) in the
Amazon, this might not be the case for nectarivorous
bats. But it is important to point that Farneda et al.
(2015) did not analyze responses of the most common
nectarivore species we found in the present study;
Glossophaga soricina. Thus, more studies should address
responses of nectarivores on different fragmentation contexts. Models were very weak for sanguivores (R2 = 0.01
for the best model), since we mostly captured them in
more degraded landscapes, with <10% of semidecidous
forest. We suggest that these weak patterns are due to
our sample design that prioritized areas mostly surrounded by sugarcane, and not by pasture, where vampire
bats tend to be more abundant (del Pietro et al. 1992).
We expected that animalivorous bats (subfamily
Phyllostominae) would respond positively to large
amounts of forest, due to the small home ranges for some
species and their characteristic foraging strategies
(Cosson et al. 1999). We found these bats in landscapes
with low amounts of forest, but even in high-­habitat
landscapes they exhibited low abundance. Due to the
extensive fragmentation of São Paulo state (Durigan
et al. 2007), it is possible that rare species have become
locally extinct in landscapes with lower levels of forest
cover. As phyllostomines were not abundant in this
study, and under-­sampling could mask effects of habitat
loss, we recommend that a broader analysis of response
of this group be performed for cerrado and other biomes
where information from long-­term sampling is available.
Whereas many studies have related bat diversity to
patch size (Estrada et al.1993, Cosson et al. 1999, Schulze
et al. 2000; Faria 2006), ours is novel in that we have
confirmed the existence of a fragmentation threshold
within a landscape perspective. Studies addressing spatial
and temporal processes that affect distribution and abundance of species provide advances in our understanding
of ecosystems. Furthermore, observed patterns contribute to amplify our predictive knowledge of consequences of landscape alterations for animal populations
(Fortin and Dale 2005). This study corroborates that
animalivorous and nectarivorous bats are rarer in disturbed landscapes (see Fenton et al. 1992, Wilson et al.
1996, Medellín et al. 2000), and that these guilds could
be good indicators of ecosystem integrity (Rapport
1995). We believe that, together with studies that have
been published over the last 20 years (Fenton et al. 1992,
Meyer and Kalko 2008, Meyer et al. 2008, García-­García
et al. 2014, Farneda et al. 2015), we are forming a robust
body of evidence about the sensitivity of certain bats to
anthropogenic habitat modification in the Neotropics.
The presence or persistence of certain bat subfamilies
in particular the Phyllostominae likely indicates
1864
RENATA L. MUYLAERT ET AL.
ecosystem integrity (Fenton et al. 1992, Brosset et al.
1996 in the Neotropics; and Duchamp and Swihart 2008
in temperate forests). The importance of these animal
groups to ecological functions is still unclear, which
should further encourage studies on their role in nutrient
cycling and energy flow, predation (Boyles et al. 2011),
mutualisms (Saldaña-­
Vázquez 2014), and coexistence
(Weber et al. 2011), as well as about their sensitivity and
persistence in different habitats (Martensen et al. 2012).
Identifying reduced species richness in landscapes with
low amounts of habitat can focus conservation strategies
by identifying priority areas with major restoration
potential. Bats play an interesting role in this sense. They
can be sensitive bioindicators (phyllostomine bats and
the small nectarivores) as well as effective tools for restoration (frugivores and nectarivores) since they are persistent and very abundant in fragmented landscapes and
important seed dispersers and flower pollinators.
Our research highlights the effects of habitat loss and
landscape composition on a highly abundant and mobile
animal group. Native vegetation cover in the landscape,
particularly the amount of cerrado was important in
explaining response of abundance of the two most sensitive groups of bats to fragmentation: animalivores and
nectarivores. We suggest that a threshold relationship
of species richness with amount of forest should be more
broadly investigated in order to better understand effects
of habitat loss on distribution of organisms and also on
ecological processes. Moreover, species loss is more
­dramatic below the threshold level. This does not mean
that species loss does not occur in areas with higher
amounts of habitat (Mönkkönen and Reunanen 1999).
Moreover, even a high threshold point might be underestimating the magnitude of species loss in highly fragmented landscapes.
Acknowledgments
We thank the colleagues, landowners, reserve guards, and
keepers who helped us. M. Nogueira, A. Peracchi, C. Aires,
and G. Sabino-­Santos Jr helped us with species identification.
M. Mello and M. Cortês gave us suggestions for the manuscript. C. Mendes and P. Dodonov helped us with programming.
C. Sanches, A. and G. C.de Morais, M. Schaefer, C. Mendes,
G. Sabino-­Santos Jr, M. Eigenher, V. Kavagutti, P. Rogeri, and
many others assisted us in the field. The Unesp, UFSCar,
COTEC/IF (S. Aparecida, E. Montilla, and L. Tadeo), and
IBAMA provided us with fieldwork infrastructure. The Brazilian
Research Council (CNPq RLM 131169/2012-­
2, MCR
312045/2013-­1) and the São Paulo Research Foundation (Fapesp
2012/04096-­0, 2013/18465-­0, 2013/50421-­2), IDEA WILD, and
Ecotone Inc. (“Do Science and Get Support”) funded our study.
This is m
­ anuscript number T-­9-­1275, College of Agricultural
Sciences and Natural Resources, Texas Tech University.
Literature Cited
Aguiar, L. M. S., and J. S. Marinho-Filho. 2004. Activity
patterns of nine phyllostomid bat species in a fragment
of the Atlantic Forest in southeastern Brazil. Revista
Brasileira de Zoologia 21:385–390.
Ecological Applications
Vol. 26, No. 6
Aguilar, R., L. Ashworth, L. Galetto, and M. Aizen. 2006.
Plant reproductive susceptibility to habitat fragmentation:
review and synthesis through a meta-­
analysis. Ecology
Letters 9:968–980.
Aguirre, L. F. 2002. Structure of a Neotropical savanna bat
community. Journal of Mammalogy 83:775–784.
Almeida-Neto, M., P. Guimarães, P. R. Guimarães, R. D.
Loyola, and W. Ulrich. 2008. A consistent metric for
nestedness analysis in ecological systems: reconciling concept
and measurement. Oikos 117:1227–1239.
Andrén, H. 1994. Effects of habitat fragmentation on birds
and mammals in landscapes with different proportions of
suitable habitat: a review. Oikos 71:355–366.
Avila-Cabadilla, L. D., G. A. Sanchez-Azofeifa, K. E. Stoner,
M. Y. Alvarez-Añorve, M. Quesada, and C. A. PortilloQuintero. 2012. Local and landscape factors determining
occurrence of phyllostomid bats in tropical secondary forests.
PLoS ONE 7:e35228.
Banks-Leite, C., R. Pardini, L. R. Tambosi, W. D. Pearse,
A. A. Bueno, R. Bruscagin, and J. P. Metzger. 2014.
Using ecological thresholds to evaluate the costs and benefits
of set-­
asides in a biodiversity hotspot. Science 345:
1041–1045.
Bennett, A. F., J. Q. Radford, and A. Haslem. 2006. Properties
of land mosaics: implications for nature conservation in
agricultural environments. Biological Conservation 133:
250–264.
Bernard, E., and M. B. Fenton. 2003. Bat mobility and
roosts in a fragmented landscape in Central Amazonia,
Brazil. Biotropica 35:262–277.
Bernard, E., and M. Fenton. 2007. Bats in a fragmented
landscape: species composition, diversity and habitat
interactions in savannas of Santarém, Central Amazonia,
Brazil. Biological Conservation 134:332–343.
Betts, M. G., G. J. Forbes, and A. W. Diamond. 2007.
Thresholds in songbird occurrence in relation to landscape
structure. Conservation Biology 21:1046–1058.
Blüthgen, N. 2010. Why network analysis is often disconnected
from community ecology: a critique and an ecologist’s
guide. Basic and Applied Ecology 11:185–195.
Bolívar-Cimé, B., J. Laborde, G. M. MacSwiney, C. MuñozRobles, and J. Tun-Garrido. 2013. Response of
phytophagous bats to patch quality and landscape attributes
in fragmented tropical semi-­
deciduous forest. Acta
Chiropterologica 15:399–409.
Bolker, B., and R Development Core Team. 2014. Bbmle:
tools for general maximum likelihood estimation. R package
version 1.0.16. http://CRAN.R-project.org/package=bbmle
Boyles, J. G., P. M. Cryan, G. F. Mccracken, and T. H.
Kunz Th. 2011. Economic importance of bats in agriculture.
Science 332:41–42.
Brosset, A., P. Charles-Dominique, A. Cockle, J. F. Cosson,
and D. Masson. 1996. Bat communities and deforestation
in French Guiana. Canadian Journal of Zoology
74:1974–1982.
Burnham, K. P., and D. R. Anderson. 2002. Model selection
and multimodel inference. Second edition. Springer-Verlag,
New York, New York, USA
Chapin, F. S. III, E. S. Zavaleta, V. T. Eviner, R. L. Naylor,
P. M. Vitousek, H. L. Reynolds, and S. Díaz. 2000. Consequences
of changing biodiversity. Nature 405:234–242.
Cisneros, L. M., M. E. Fagan, and M. R. Willig. 2014.
Effects of human-­
modified landscapes on taxonomic,
functional and phylogenetic dimensions of bat biodiversity.
Diversity and Distributions 21:523–533.
Cosson, J. F., J. M. Pons, and D. Masson. 1999. Effects of
forest fragmentation on frugivorous and nectarivorous bats
September 2016
BAT RICHNESS DECLINES WITH FRAGMENTATION
in French Guiana. Journal of Tropical Ecology
15:515–534.
Coutinho, L. M. 1978. O conceito de cerrado. Revista Brasileira
de Botânica 1:17–23.
Coutinho, L. M. 2006. O conceito de bioma. Acta Botanica
Brasilica 20:13–23.
Cunto, G. C., and E. Bernard. 2012. Neotropical bats as
indicators of environmental disturbance: what is the
emerging message? Acta Chiropterologica 14:143–151.
Dirzo, R., H. S. Young, M. Galetti, G. Ceballos, N. J.
Isaac, and B. Collen. 2014. Defaunation in the
Anthropocene. Science 345:401–406.
Duchamp, J. E., and R. K. Swihart. 2008. Shifts in bat
community structure related to evolved traits and features
of human altered landscapes. Landscape Ecology
23:849–860.
Durigan, G., G. A. D. C. Franco, and M. F. Siqueira. 2004.
A vegetação dos remanescentes de cerrado no estado de
São Paulo. Pages 29–56 in M. D. Bitencourt, and R. R.
Mendonça, editors. Viabilidade de Conservação dos
remanescentes de cerrado no Estado de São Paulo.
Annablume/FAPESP, São Paulo, São Paulo, Brazil.
Durigan, G., M. F. Siqueira, and G. A. D. C. Franco. 2007.
Threats to the cerrado remnants of the state of São Paulo,
Brazil. Scientia Agricola 64:355–363.
Estavillo, C., R. Pardini, and P. L. B. da Rocha. 2013.
Forest loss and the biodiversity threshold: an evaluation
considering species habitat requirements and the use of
matrix habitats. PLoS ONE 8:e82369.
Estes, J. A., et al. 2011. Trophic downgrading of planet
Earth. Science 333:301–306.
Estrada, A., R. Coates-Estrada, and D. Merritt. 1993. Bat
species richness and abundance in tropical rain forest
fragments and in agricultural habitats at Los Tuxtlas,
Mexico. Ecography 16:309–318.
Fahrig, L. 2002. Effect of habitat fragmentation on the
extinction threshold: a synthesis. Ecological Applications
12:346–353.
Fahrig, L. 2003. Effects of habitat fragmentation on
biodiversity. Annual Review of Ecology Evolution and
Systematics 34:487–515.
Fahrig, L. 2013. Rethinking patch size and isolation effects:
the habitat amount hypothesis. Journal of Biogeography
40:1649–1663.
Fahrig, L., and G. Merriam. 1985. Habitat patch connectivity
and population survival. Ecology 66:1762–1768.
Faria, D. 2006. Phyllostomid bats of a fragmented landscape
in the north-­
eastern Atlantic forest, Brazil. Journal of
Tropical Ecology 22:531–542.
Farneda, F. Z., R. Rocha, A. López-Baucells, M. Groenenberg,
I. Silva, J. M. Palmeirim, P. E. D. Bobrowiec, and C. F.
J. Meyer. 2015. Trait-­
related responses to habitat
fragmentation in Amazonian bats. Journal of Applied
Ecology 52:1381–1391.
Fenton, M. B., L. Acharya, D. Audet, M. B. Hickey, C.
Merriman, M. K. Obrist, D. M. Syme, and B. Adkins.
1992. Phyllostomid bats (Chiroptera, Phyllostomidae) as
indicators of habitat disruption in the Neotropics. Biotropica
24:440–446.
Fortin, M. J., and M. T. T. Dale. 2005. Spatial analysis: a
guide for ecologists. Cambridge University Press, Cambridge,
UK.
García-García, J. L., A. Santos-Moreno, and C. KrakerCastañeda. 2014. Ecological traits of phyllostomid bats
associated with sensitivity to tropical forest fragmentation
in Los Chimalapas, Mexico. Tropical Conservation Science
7:457–474.
1865
García-Morales, R., E. I. Badano, and C. E. Moreno. 2013.
Response of Neotropical bat assemblages to human land
use. Conservation Biology 27:1096–1106.
Gardner, A. L. 1977. Feeding habits. Pages 293–350 in
R. J. Baker, J. K. Jones Jr, and D. C. Carter,
editors. Biology of bats of the new world family Phyl­
lostomidae. Part II. Texas Tech University, Lubbock,
Texas, USA.
Gardner, A. L. 2008. Mammals of South America, Volume
1. Marsupials, xenarthrans, shrews, and bats. University
of Chicago Press, Chicago, Illinois, USA.
Gorresen, P. M., M. R. Willig, and R. E. Strauss. 2005.
Multivariate analysis of scale-­
dependent associations
between bats and landscape structure. Ecological
Applications 15:2126–2136.
Gotelli, N. J., and R. K. Colwell. 2001. Quantifying
biodiversity: procedures and pitfalls in the measurement
and comparison of species richness. Ecology Letters
4:379–391.
Groffman, P. M., et al. 2006. Ecological thresholds: The key
to successful environmental management or an important
concept with no practical application? Ecosystems 9:
1–13.
Guimarães, P. R., and P. Guimarães. 2006. Improving the
analyses of nestedness for large sets of matrices.
Environmental Modelling and Software 21:1512–1513.
Hastie, T. 2013. gam: Generalized Additive Models. R package
version 1.09.1. http://CRAN.R-project.org/package=gam
Holdridge, L. R. 1947. Determination of world plant
formations
from
simple
climatic
data.
Science
105:367–368.
Huggett, A. J. 2005. The concept and utility of ‘ecological
thresholds’ in biodiversity conservation. Biological
Conservation 124:301–310.
Jones, K., A. Purvins, and J. L. Gittleman. 2003. Biological
correlates of extinction risk in bats. American Naturalist
161:601–614.
Klingbeil, B. T., and M. R. Willig. 2009. Guild-­
specific
responses of bats to landscape composition and configuration
in fragmented Amazonian rainforest. Journal of Applied
Ecology 46:203–213.
Kronka, F. J. N., C. K. Matsukuma, M. A. Nalon, I. H.
D. Cali, M. Rossi, J. F. A. Mattos, M. S. Shin-Ike, and
A. A. S. Pontinha. 2005. Inventário florestal do estado
de São Paulo. Instituto Florestal, São Paulo, São Paulo,
Brazil.
Kunz, T. H., and S. Parsons. 2009. Ecological and behavioral
methods for the study of bats. Johns Hopkins University
Press, Baltimore, Maryland, USA.
Lindenmayer, D. B., J. Fischer, and R. B. Cunningham.
2005. Native vegetation cover thresholds associated
with species responses. Biological Conservation 124:
311–316.
Magurran, A. E. 2004. Ecological diversity and its
measurement. Princeton University Press, Princeton, New
Jersey, USA.
Martensen, A. C., M. C. Ribeiro, C. Banks-Leite, P. I. Prado,
and J. P. Metzger. 2012. Associations of forest cover,
fragment area, and connectivity with Neotropical understory
bird species richness and abundance. Conservation Biology
26:1100–1111.
Medellín, R. A., M. Equihua, and M. A. Amin. 2000.
Bat diversity and abundance as indicators of disturbance
in Neotropical rainforests. Conservation Biology 14:
1666–1675.
Medina, A., C. A. Harvey, D. S. Merlo, S. V. Vílchez, and
B. Hernández. 2007. Bat diversity and movement in an
1866
RENATA L. MUYLAERT ET AL.
agricultural landscape in Mantiguás, Nicarágua. Biotropica
39:120–128.
Mello, M. A., E. K. Kalko, and W. R. Silva. 2013. Effects
of moonlight on the capturability of frugivorous
phyllostomid bats (Chiroptera: Phyllostomidae) at different
time scales. Zoologia 30:397–402.
Meyer, C. F., and E. K. Kalko. 2008. Assemblage-­
level
responses of phyllostomid bats to tropical forest
fragmentation: land-­
bridge islands as a model system.
Journal of Biogeography 35:1711–1726.
Meyer, C. F., J. Fründ, W. P. Lizano, and E. K. V. Kalko.
2008. Ecological correlates of vulnerability to fragmentation
in Neotropical bats. Journal of Applied Ecology
45:381–391.
Mönkkönen, M., and P. Reunanen. 1999. On critical thresholds
in landscape connectivity: a management perspective. Oikos
88:302–305.
Montiel, S., A. Estrada, and P. León. 2006. Bat assemblages
in a naturally fragmented ecosystem in the Yucatan Peninsula,
Mexico: species richness, diversity and spatio-­
temporal
dynamics. Journal of Tropical Ecology 22:267–276.
Muggeo, V. M. R. 2003. Estimating regression models with
unknown break-­
points. Statistics in Medicine 22:3055–
3071.
Muggeo, V. M. R. 2014. Package ‘segmented’. Biometrika
58:525–534.
Muylaert, R. L., D. Matos, and M. A. R. Mello. 2014.
Interindividual variations in fruit preferences of the yellow-­
shouldered. Mammalia 78:93–101.
Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B.
Da Fonseca, and J. Kent. 2000. Biodiversity hotspots for
conservation priorities. Nature 403:853–858.
Pardini, R., A. A. Bueno, T. A. Gardner, P. I. Prado, and
J. P. Metzger. 2010. Beyond the fragmentation threshold
hypothesis: regime shifts in biodiversity across fragmented
landscapes. PLoS ONE 5:e13666.
Parker, M., and R. MacNally. 2002. Habitat loss and the
habitat fragmentation threshold: an experimental evaluation
of impacts on richness and total abundances using grassland
invertebrates. Biological Conservation 105:217–229.
del Pietro, H. A., N. Marchevsky, and E. Simonetti. 1992.
Relative population densities and predation of the common
vampire bat Desmodus rotundus in natural and cattle-­raising
areas in north-­
east Argentina. Preventive Veterinary
Medicine 14:13–20.
QGIS Development Team. 2014. QGIS Geographic
Information System. Open Source Geospatial Foundation
Project. http://qgis.osgeo.org
R Development Core Team. 2014. R: a language and
environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. www.R-project.org
Radford, J. Q., A. F. Bennett, and G. J. Cheers. 2005.
Landscape level thresholds of habitat cover for woodland-­
dependent birds. Biological Conservation 124:317–337.
Rangel, T. F., J. A. F. Diniz-Filho, and L. M. Bini. 2010.
SAM: a comprehensive application for spatial analysis in
macroecology. Ecography 33:46–50.
Rapport, D. J. 1995. Ecosystem health: an emerging integrative
science. Springer Berlin Heidelberg, Germany.
Ratter, J. A., J. F. Ribeiro, and S. Bridgewater. 1997. The
Brazilian cerrado vegetation and threats to its biodiversity.
Annals of Botany 80:223–230.
Ribeiro, J. F., and B. M. Walter. 2008. As principais
fitofisionomias do Bioma Cerrado. Pages 152–212 in
S. M. Sano, S. P. Almeida, and J. F. Ribeiro, editors.
Cerrado ecologia e flora. Embrapa, Brasília, Distrito Federal,
Brazil.
Ecological Applications
Vol. 26, No. 6
Ribeiro, M. C., J. P. Metzger, A. C. Martensen, F. G.
Ponzoni, and M. M. Hirota. 2009. The Brazilian Atlantic
Forest: how much is left, and how is the remaining forest
distributed? Implications for conservation. Biological
Conservation 142:1141–1153.
Rodrigues, R. R., R. A. F. Lima, S. Gandolfi, and A. G.
Nave. 2009. On the restoration of high diversity forests:
30 years of experience in the Brazilian Atlantic Forest.
Biological Conservation 142:1242–1251.
Safi, K., and G. Kerth. 2004. A comparative analysis of
specialization and extinction risk in temperate-­
zone bats.
Conservation Biology 18:1293–1303.
Saldaña-Vázquez, R. A. 2014. Intrinsic and extrinsic factors
affecting dietary specialization in Neotropical frugivorous
bats. Mammal Review 44:215–224.
Schulze, M. D., N. E. Seavy, and D. F. Whitacre. 2000. A
comparison of the phyllostomid bat assemblages in
undisturbed neotropical forest and in forest fragments of
a slash-­
and-­
burn farming mosaic in Petén, Guatemala.
Biotropica 32:174–184.
Sikes, R. S., Gannon, W. L., andThe Animal Care and Use
Committee of the American Society of Mammalogists. 2011.
Guidelines of the American Society of Mammalogists for
the use of wild mammals in research. Journal of Mammalogy
92:231–253.
Silva, C. M., F. R. Scarano, and F. Souza Cardel. 1995.
Regeneration of an Atlantic forest formation in the
understorey of a Eucalyptus grandis plantation in south-­
eastern Brazil. Journal of Tropical Ecology 11:
147–152.
Straube, F. C., and G. V. Bianconi. 2002. Sobre a grandeza
e a unidade utilizada para estimar esforço de captura com
utilização de redes-­
de-­
neblina. Chiroptera Neotropical
8:150–152.
Suarez-Rubio, M., S. Wilson, P. Leimgruber, and T.
Lookingbill. 2013. Threshold responses of forest birds to
landscape changes around exurban development. PLoS ONE
8:1–11.
Sutherland, W. J., et al. 2013. Identification of 100 fundamental
ecological questions. Journal of Ecology 101:58–67.
Swihart, L. R. K., J. J. Lusk, J. E. Duchamp, C. M. Rizkalla,
and J. E. Moore. 2006. The roles of landscape context,
niche breadth, and range boundaries in predicting species
responses to habitat alteration. Diversity and Distributions
12:277–287.
Tambosi, L., A. C. Martensen, M. C. Ribeiro, and J. P.
Metzger. 2013. A framework to optimize restoration efforts
based on habitat amount and landscape connectivity.
Restoration Ecology 22:169–177.
Thies, W., E. K. V. Kalko, and H. A. Schnitzler. 2006.
Influence of environment and resource availability on activity
patterns of Carollia castanea (Phyllostomidae) in Panama.
Journal of Mammalogy 87:331–338.
Toms, J. D., and M. L. Lesperance. 2003. Piecewise regression:
a tool for identifying ecological thresholds. Ecology
84:2034–2041.
Vizotto, L. D., and V. A. Taddei. 1973. Chave para
determinação de quirópteros brasileiros. Editora da
Universidade Estadual de São Paulo, São José do Rio
Preto, São Paulo, Brazil 72 pp.
Weber, M. M., J. L. De Arruda, B. O. Azambuja, V. L.
Camilotti, and N. C. Cáceres. 2011. Resources partitioning
in a fruit bat community of the southern Atlantic Forest,
Brazil. Mammalia 75:217–225.
Wilson, D. E., C. F. Ascorra, and S. Solari. 1996. Bats as
indicators of habitat disturbance. Pages 613–625 in D. E.
Wilson, and A. Sandoval, editors. Manu: the biodiversity
September 2016
BAT RICHNESS DECLINES WITH FRAGMENTATION
of southeastern Peru. Smithsonian Institution Press,
Washington, D.C., USA.
Wu, J., and J. R. Hobbs, editors. 2007. Key topics in landscape
ecology. Cambridge University Press, Cambridge, UK.
Zar, J. H. 1999. Biostatistical analysis. Prentice Hall, Upper
Saddle River, New Jersey, USA 662 p.
1867
Zortéa, M., and C. J. R. Alho. 2008. Bat diversity of a
Cerrado habitat in central Brazil. Biodiversity and
Conservation 17:791–805.
Zuur, A., E. N. Ieno, N. Walker, A. A. Saveliev, and G. M.
Smith. 2009. Mixed effects models and extensions in ecology
with R. Springer, New York, New York, USA. 574 pp.
Supporting Information
Additional supporting information may be found in the online version of this article at http://onlinelibrary.wiley.com/
doi/10.1890/15-1757.1/suppinfo