Biological Conservation 246 (2020) 108579

Contents lists available at ScienceDirect

Biological Conservation
journal homepage: www.elsevier.com/locate/biocon

Policy analysis

Habitat loss, extinction predictability and conservation efforts in the T

terrestrial ecoregions
⁎
Daniel Gonçalves-Souzaa, , Peter H. Verburgb, Ricardo Dobrovolskia
a
Instituto de Biologia, Universidade Federal da Bahia, Salvador, BA, Brazil
b
Institute for Environmental Studies, VU University Amsterdam, de Boelelaan 1087, 1081HV Amsterdam, the Netherlands

A R T I C LE I N FO A B S T R A C T

Keywords: The destruction of natural habitats is causing loss of biodiversity and ecosystem services. Although a “zero
Endemics-area relationship deforestation” is targeted, agriculture expansion caused by increasing human population and per capita con-
Extinction sumption might boost the destruction of natural habitats in the coming decades. Here, we estimated the current
Habitat loss and future extinction crisis in terrestrial ecoregions caused by habitat destruction and related this pattern with
Hotspots
the current conservation efforts. We applied an Endemics-Area Relationship to assess vertebrates' potential
Land cover change
extinctions in 513 ecoregions based on current land cover data and a future scenario of habitat loss. We com-
Protected areas
pared our predictions to the proportion of the ecoregions' area formally protected, testing the concordance
between threat distribution and conservation efforts. Finally, we evaluated how the distribution of threat relates
to the biodiversity hotspots delimitation. We found that 2134 endemic vertebrates are currently threatened due
to accumulated habitat loss, which is consistent with the assessment of the IUCN Red List. Further, this threat
could overtake 4209 species when considering habitat loss projections to 2040. Our findings indicate a high
concentration of threat in a few megadiverse localities, some of them outside the biodiversity hotspots. We found
little overlap between our predictions of extinction and current protected areas distribution. This study supports
current biodiversity crisis diagnoses and the expected recrudescence of Anthropocene defaunation in the future
when considering scenarios of further habitat destruction. Our analysis also contributes to the definition of
global priorities to prevent further biodiversity loss.

1. Introduction causing more impact on species populations (but see Dobrovolski et al.,
2013).
An accelerated biodiversity loss is under way on Earth (Balmford In this scenario of increasing destruction of natural vegetation and,
et al., 2003). For instance, the abundance of terrestrial vertebrates re- consequently, species' population decline, it is essential to quantify and
duced by 60% since 1970 (Grooten and Almond, 2018), and 198 ver- understand the effects of habitat loss across space. One of the main
tebrates' extinctions have been recorded since 1900, which is at least approaches to quantify the “biodiversity crisis” is the species-based
100 times more than what is naturally expected (Ceballos et al., 2015). monitoring, synthesized on the IUCN Red List (IUCN; https://www.
This crisis is the consequence of human enterprises that resulted in a iucnredlist.org/). The IUCN Red List is a valuable tool for conservation,
new age on Earth, the Anthropocene (Lewis and Maslin, 2015). Among assembling a large amount of data of species under the supervision of
the anthropic changes on Earth, habitat conversion represents the experts, yielding in a comprehensive and accessible conservation da-
leading cause of species extinctions, being agricultural activities the tabase (Rodrigues et al., 2006). However, the individual assessment of
drivers of the majority of the destruction of natural areas (Foley et al., the extinction risk of species requires considerable effort and only ad-
2005). Furthermore, the growth of the human population and the dresses the risk retroactively. Thus, finding alternative, theory-based
change in consumption patterns have been causing an increasing de- approaches that allow complementary assessments of the current and
mand for food, fuel, fiber, and livestock feed (Tilman et al., 2011). As a future state of biodiversity can help the understanding of this crisis and
result, several studies have projected an accentuated agricultural ex- guide efforts to overcome it. The species-area relationship (SAR) and
pansion in the next decades (Asselen and Verburg, 2013; Laurance endemics-area relationship (EAR) are examples of this approach. More
et al., 2014) which might require further modification on natural areas, specifically, the backward kind of SAR/EAR (i.e., an area decrease

⁎
Corresponding author.
E-mail addresses: [email protected] (D. Gonçalves-Souza), [email protected] (P.H. Verburg).

https://doi.org/10.1016/j.biocon.2020.108579
Received 16 September 2019; Received in revised form 18 February 2020; Accepted 3 April 2020
0006-3207/ © 2020 Elsevier Ltd. All rights reserved.
D. Gonçalves-Souza, et al. Biological Conservation 246 (2020) 108579

resulting in species richness decrease) has been used extensively as a EAR value of z’ is always greater than its value on SAR (Harte and
way of quantifying biodiversity loss due to habitat destruction (e.g., Kinzig, 1997), and has been reported to vary around 0.75 and 1.5 for
Thomas et al., 2004; Pereira et al., 2010; Bellard et al., 2014). To vertebrates on a continental scale (Storch et al., 2012). Here, we use the
specifically address the potential biodiversity loss at highly vulnerable intermediate value of 1 for z. (see Appendix A for further discussion).
and irreplaceable areas, Brooks et al., 2002 applied a backward EAR To assess our EAR model, we fitted a linear regression of the number
and predicted the number of vertebrates and plants at extinction risk of the predicted endemic extinctions in each ecoregion considering the
due to accrued habitat loss. However, the analysis was restricted to the habitat loss that occurred until 2015 with the number of endemics that
biodiversity hotspots (BH). Although BH represents probably the most are currently considered threatened or extinct by IUCN in the same
successful global priority scheme, they cover only 16% of the global ecoregion. The expectation is that, in each ecoregion, our predictions
terrestrial area, being most of the land bypassed in this analysis. More are approximately the number of endemic species threatened or already
recently, after the delimitation of ecoregions (Olson et al., 2001), the extinct (i.e., a regression with a slope of 1). We estimated the con-
use of ecoregions as biogeographic units has become more common in fidence intervals for the coefficient intervals using the profile likelihood
global analysis on conservation topics (e.g., Hoekstra et al., 2005; method.
Jenkins and Joppa, 2009). Therefore, using a simple tool as the EAR,
there is an opportunity to conduct a global analysis in the ecoregions of 2.2. Ecoregions
the effect of habitat loss on biodiversity and modeling scenarios for the
future considering the land-use change. Alongside, evaluating the spa- We used terrestrial ecoregions as primary biogeographic units to
tial correlation between the distribution of protected areas and the assess the risk of extinction of endemic vertebrates. The ecoregions map
threats to biodiversity is a crucial task. Such analysis can improve the used here (TNC, 2009) consists of 814 ecoregions divided into 14 dif-
knowledge of the main elements that compose the biodiversity crisis, ferent biomes. We excluded ecoregions in Antarctica and those classi-
thereby allowing more informed conservation actions. fied as Rock and Ice or Inland Water.
Here, we used EAR to quantify extinction risk in the terrestrial
ecoregions, predicting the threat of extinction of endemic vertebrates 2.3. Species
due to accumulated habitat loss. The risk assessment had two steps. In
the first step, we evaluated the risk of extinction of endemic vertebrates We quantified the original number of endemics in the ecoregions
on the ecoregions due to the historical habitat loss that occurred until using the spatial data of 27,319 species of terrestrial vertebrates,
the present time. This step also worked as a model validation since we available for amphibians, birds, mammals, and reptiles (IUCN, 2017;
compared our assessment with independent risk assessment of extinc- BirdLife International and NatureServe, 2016). This dataset represents
tion risk of the IUCN Red List. The second step was the risk assessment 80% of the amphibians, 99% of the birds, 91.8% of mammals, and
of extinction for the future, using for this the historical habitat loss 41.2% of reptiles currently described. From this data, grids of 0.1° x 0.1°
coupled to a future land use scenario. Moreover, we could assess the resolution were derived and then overlaid with the ecoregions in order
vulnerability of the protected areas network by confronting the possible to identify which species are endemics. We considered as endemic the
upcoming habitat loss and species extinction with the protected areas species that had at least 90% of their distribution area restricted to a
coverage. Finally, we explore the concordances and novelties of our single ecoregion. We found endemics in 513 ecoregions.
results with the BH to evaluate the merits and shortfalls of this ap- We obtained the assessment of extinction risk from the IUCN Red
proach. List (2018) for the endemic vertebrates we identified. We compared our
risk assessment based on current habitat loss with the IUCN assessment.
2. Methods For the risk assessments, we excluded endemics whose habitat pre-
ferences included the artificial environments category on IUCN since
2.1. EAR species that tolerate to human-modified habitats are unlikely to become
extinct after habitat loss. As our assessment with the EAR can only
The SAR is a well-documented empirical model that infers that, the identify the number of threatened species, being unable to define their
increase in available area is accompanied by an increase in the number level of threat, we reclassified the categories of IUCN Red List into a
of species (Rosenzweig, 1995). The elementary form of SAR is the binary classification. Therefore, we considered species as threatened
power function as proposed by Arrhenius (1921): with extinction, those categorized as Vulnerable, Endangered, Critically
Endangered, Extinct in the Wild, and Extinct. Species that were listed on
S = cAz
threatened categories for fitting the criteria A1 or A2 from sub-
where the S is the number of species in a given area, A, and c and z are categories d or e, or from criterion D, were not included since they are
constants. Harte and Kinzig (1997) argued that the relationship is also threatened by other factors than habitat destruction.
valid for endemic species. On the endemics-area relationship (EAR),
considering an area a that is part of a larger area A, the number of 2.4. Habitat loss
species that occurs exclusively in the area a increases alongside the area
of a. Therefore: To quantify the current habitat loss on the ecoregions, we used the
map of land cover from the Climate Change Initiative Land Cover (CCI LC)
E= c ’az ’
for the year 2015 with a resolution of 0.0028° x 0.0028° (ESA, 2017).
where the E is the number of endemic species of a given area a and c’ For the forecasting into future conditions, we used the CLUMondo
and z’ are constants, being z’ dependent of and always greater than z. model of land system change for the year 2040, with a resolution of
To predict the number of endemic species threatened with extinc- 0.083° x 0.083°. This model is based on regional economic demands,
tion in the ecoregions in the present and the future, we used a backward calculated by integrated assessment models and a dynamic social-en-
derivation of the EAR power function: vironmental context, and presents a probable forthcoming land-use
Eoriginal − Enew = Eoriginal − Eoriginal (anew /a original ) z’ scenario (Asselen and Verburg, 2013). The scenario used in this study is
based on the OECD Environmental Outlook scenario, which depicts
In this function, assuming that an initial ecoregion area aoriginal is probable developments in the economic and environmental conditions,
reduced to an area anew, the initial number of endemic species found on being regarded as a baseline scenario. The land system trends captured
that ecoregion Eoriginal declines to Enew. The value of z commonly used in this scenario resemble those of the well-known SSP2 scenario (Wolf
on SAR is 0.25. (Brooks et al., 2002; Thomas et al., 2004). However, the et al., 2018). We reclassified the classes of land use of both maps in

2
D. Gonçalves-Souza, et al. Biological Conservation 246 (2020) 108579

three categories: anthropic areas, habitat, and a category that embodies

bare areas, water bodies, and permanent ice areas (see Appendix A:
Table A1 and Table A2 in Supporting Information). To quantify the
current habitat loss, we used the CCI LC map and considered the natural
formations of vegetation categories as the available habitat for the
species. For the original area available for the species, we considered
the sum of anthropic areas and habitat areas, assuming that the an-
thropic areas were formerly habitat. We excluded water bodies, bare
areas, and permanent ice areas for the habitat analyses since these areas
are not subjected to natural vegetation clearance. For future habitat
loss, we overlaid the CCI LC map and the CLUMondo land systems map
and quantified the pixels that were converted from a natural land cover
in 2015 to an anthropic one in 2040. Data for land use for both present
and future were available for 775 ecoregions.

2.5. Protected areas

The polygon and point data from the World Database on Protected
Areas (UNEP-WCMC and IUCN, 2018; http://www.protectedplanet.
net) from May 2018 were rasterized to 0.1° x 0.1° resolution and used to
determine the coverage of protection on the ecoregions. The zone of
influence of the protection sites from the point data was estimated by
generating buffers based on the reported area of each site. We excluded Fig. 1. Comparison between the number of vertebrates expected to be extinct
due to the accumulated habitat loss until 2015 and the number of endemic
the protected areas with no reported areas from the analyses. Also, we
vertebrates listed as extinct or threatened by IUCN. The dashed line indicates
excluded protected sites that were classified as “Not Applicable” by the
the expected relationship. The solid line and the gray band indicate the ob-
IUCN criteria from the analyses. We included indigenous lands and PAs served relationship and its confidence interval, respectively (based on the
of all management categories of IUCN classification because we aimed logarithm relationship).
to measure conservation efforts, rather than the effectiveness of the
network of PAs to protect biodiversity. To evaluate the suitability of the
cover (e.g., Northwestern Hawaii Scrub, United States). An additional
current location of conservation sites in safeguarding the species on
47,232,806 km2 of natural area loss is expected as a result of the change
future agricultural expansion scenarios, we compared our predictions of
of land use projected for 2040, lowering the global average to 34%
future extinctions with the proportion of each ecoregion that is under
(sd = 30.9%) (Appendix A, Fig. A2).
protection. We also made exploratory comparisons with coverage of
protection and endemics richness and remaining habitat.
3.2. Current threat and model validation
2.6. Biodiversity hotspots
We identified 6075 endemic species distributed in 513 ecoregions,
We used the map of the BH (Brooks et al., 2006) to evaluate the with a mean of 7.6 endemics (sd = 16.8) per ecoregion. These endemics
relationship between our risk assessment of extinction to the delimi- includes 2629 amphibians (43.3% of the total of endemics), 1308 birds
tation of the BH. For this evaluation, we created our definition of a (21.5%), 874 mammals (14.4%) and 1264 reptiles (20.8%).
hotspot based on our data of endemism and habitat loss. In this defi- Considering the accumulated habitat loss until the year 2015, our EAR
nition, an ecoregion is considered a hotspot when their number of en- model predicted approximately 2134 endemic vertebrates (35.1% of
demic vertebrates is above the median of the ecoregions, and their the total of endemics) to be threatened with extinction, being ≈ 922
current accumulated habitat loss has surpassed 50% (criteria further amphibians (43.2% of the total of predictions), 444 birds (20.8%), 294
discussed in Appendix A). On the other hand, we considered ecoregions mammals (13.8%) and 473 reptiles (22.2%).
as a BH those that have half or more of their area inside the boundaries The backward EAR successfully represented the threat of extinction
of at least one BH. We then confronted the two approaches of identi- to endemics of the ecoregions, but with a tendency to overestimation,
fication of hotspots using a chi-squared test, comparing the proportion as indicated by the slope of the linear regression of our predictions
of ecoregions classified as a hotspot by our criteria with the observed against the number of threatened species according to IUCN Red List
proportions based on the delimitation of the BH (see Appendix A, Table (β = 0.7, 95% CI [0.65, 0.74]; Fig. 1). Also, by absolute numbers, we
A3). predict more risk to the endemic vertebrates from our analyses than the
IUCN Red List since 1817 (29.9%) of the endemics are considered as
2.7. Statistical analyses threatened or already extinct by the list. For the four groups of terres-
trial vertebrates separately, we found that the slope of the linear re-
For general analyses, we used Pearson's correlation coefficient as gression between our predictions against the IUCN Red List ranged from
the correlation statistics and Wilcoxon-Mann-Whitney, Kruskal–Wallis, β = 0.47 95% CI [0.4, 0.54] to 0.83 95% CI [0.75, 0.9] being the birds
and Dunn's Test for the analyses of variance. We performed all statis- the group with the weakest correlation and amphibians with the
tical analyses in the software R (R Core Team, 2017). strongest one.

3. Results 3.3. Prediction of future extinctions

3.1. Quantification of habitat loss Based on accumulated habitat loss expected until 2040, our model
predicted 4209 extinctions of endemic terrestrial vertebrates, being ≈
Presently, the terrestrial ecoregions have, on average, 63.2% 1895 amphibians (45% of the total of predictions), 855 birds (20.3%),
(sd = 29.7%) of their habitat remaining, ranging from virtually none 595 mammals (14.1%) and 865 reptiles (20.6%).
(e.g., Cape Verde Islands Dry Forests, Cape Verde) to 100% of natural The top fifteen most affected ecoregions sum 1114 species

3
 D. Gonçalves-Souza, et al.

Table 1
Ecoregions with the highest number of predictions of terrestrial vertebrates' extinction due to expected habitat loss accumulated until 2040, the countries and biomes they belong, the percentage of their areas formally
protected, and their percentage of remaining habitat in 2040.
Ecoregion Country Biome Area protected (%) Remaining habitat 2040 (%) Predictions of extinction 2040

Madagascar Lowland Forests Madagascar Tropical and Subtropical Moist Broadleaf 19.2 6.1 140.9
Forests
Northwestern Andean Montane Forests Colombia and Ecuador Tropical and Subtropical Moist Broadleaf 14.1 24.3 112.8
Forests
Solomon Islands Rain Forests Solomon Islands Tropical and Subtropical Moist Broadleaf 1 45.6 98
Forests
Eastern Cordillera Real Montane Forests Ecuador, Colombia, and Peru Tropical and Subtropical Moist Broadleaf 18.1 24.9 96.1
Forests
Peruvian Yungas Peru Tropical and Subtropical Moist Broadleaf 14.5 23 89.3
Forests
Madagascar Dry Deciduous Forests Madagascar Tropical and Subtropical Dry Broadleaf 6.4 0.8 84.3
Forests

4
Greater Negros-Panay Rain Forests Philippines Tropical and Subtropical Moist Broadleaf 11.2 2.4 65.4
Forests
Madagascar Subhumid Forests Madagascar Tropical and Subtropical Moist Broadleaf 8.3 1.6 63
Forests
Serra Do Mar Coastal Forests Brazil Tropical and Subtropical Moist Broadleaf 33.4 21.8 58.6
Forests
Magdalena Valley Montane Forests Colombia Tropical and Subtropical Moist Broadleaf 12.9 28.6 55
Forests
Mascarene Forests France and Mauritius Tropical and Subtropical Moist Broadleaf 38.9 6.3 51.5
Forests
Northern Indochina Subtropical Forests China, Laos, Myanmar, Thailand, and Vietnam Tropical and Subtropical Moist Broadleaf 4.7 3.4 51.2
Forests
Eastern Arc Forests Tanzania and Kenya Tropical and Subtropical Moist Broadleaf 19.5 2.1 50.9
Forests
Albertine Rift Montane Forests Democratic Republic of the Congo, Uganda, Rwanda, Burundi, Tropical and Subtropical Moist Broadleaf 18.2 5 50.4
and Tanzania Forests
Sierra Madre Del Sur Pine-Oak Forests Mexico Tropical and Subtropical Coniferous Forests 2.7 1.9 47.1
Biological Conservation 246 (2020) 108579
D. Gonçalves-Souza, et al. Biological Conservation 246 (2020) 108579

df = 3, p < .01; Appendix A, Table A3).

Despite the divergence between the two approaches of hotspots
classification, the concordance is apparent when we observe the
overlap between the ecoregions classified as BH with the ecoregions
with the highest number of expected future extinctions (Appendix A,
Fig. A6). When considering the expected habitat loss accumulated until
2040, ecoregions classified solely as BH have on average a higher
number of additional expected endemics extinctions when compared
with the ecoregions classified as hotspots solely by our data and criteria
(Kruskal-Wallis χ2 = 157, df = 3, p < .01; Dunn's Test p < .05;
Appendix A, Fig. A4b).

4. Discussion

This study presents an assessment of the possible effect of current

and future habitat destruction on biodiversity, by applying the EAR on
terrestrial vertebrates distributed across global terrestrial ecoregions.
Based on the premise that loss of natural areas leads to extinction of
species, we pinpointed geographical patterns of expected extinction,
presenting ecoregions that most likely have their biodiversity threa-
tened in present times and those that might have their biodiversity
threatened in the future. Using the fundamental concepts of irreplace-
Fig. 2. Scatterplot of the relationship between the predictions of extinctions of
terrestrial vertebrates based on the projected accumulated habitat loss for 2040 ability and vulnerability, we highlight priority areas for conservation,
with the percentage of the ecoregion's area formally protected. complementing the well-established biodiversity hotspots approach
(Myers et al., 2000). We also assessed how the current distribution of
protected areas relates to our predictions of threat of future extinction.
threatened with extinction due to 2040 accumulated habitat loss
We believe that the use of EAR to predict extinctions should not be
(Table 1). This amount represents more than one-quarter of the total of
considered a substitute to other methods of risk assessments. Rather, we
species predicted to go extinct and in 1.2% of the total area of all
perceive the complementary value of such approach in addressing some
ecoregions (Table 1).
of the incompleteness of more nuanced assessments. For instance, the
Considering the extinctions and extinction debt accumulated until
main limitation of species-based estimates of threat is associated with
2015, the additional extinctions after 2040 habitat loss will result in the
the lack of comprehensive knowledge on species abundance and its
increase of the debt by 2075 species (34.2% of all endemics). The top
dynamics. As a result, many species cannot be assessed regarding their
fifteen ecoregions sum 29.2% of all extra extinctions. Among these
threat of extinction by this methodological approach. In the IUCN Red
ecoregions, 13 of them are part of tropical biomes (Appendix A, Fig.
List, this gap of assessment is represented in the “Data Deficient” ca-
A3).
tegory. As the EAR is based only in species richness and habitat loss, by
including these data deficient species we were able to estimate threat of
3.4. Vulnerability of current protection network extinction in a broader, although more imprecise, manner. We found
that almost one fourth (23.5%) of the endemic vertebrates are in this
We found that 14.2% of the area of ecoregions harboring endemic Data Deficient category. This knowledge gap could partially explain the
vertebrates are formally protected. No significant correlation was found higher number of threatened species on our analysis when compared
between the percentage of protected areas on the ecoregions and our with the Red List since some species might be threatened but have not
predictions of future extinctions (r = 0.07, p = .78) (Figs. 2–3) or with been assessed yet. Data deficient species have a higher chance of being
the distribution of endemics vertebrates (r = 0.01, p > .69; Appendix threatened with extinction (Jetz and Freckleton, 2015) because they are
A, Fig. A5b). However, there is a weak correlation between the pro- generally rare, underrepresented in protected areas, and have a large
tection of ecoregions with the current remaining habitat (r = 0.26, part of their geographic area overlapping with human-modified areas
p < .05; Appendix A, Fig. A5a). (Nori and Loyola, 2015).
Another contribution to risk assessments that the use of EAR to
3.5. Comparison with biodiversity hotspots predict extinctions could bring is that it allows the conservation plan-
ning to be more up to date to the ongoing destruction of habitat since it
While 369 of the ecoregions (46.2%) analyzed have at least half of does not depend on individual assessment of species, which can be
their area inside the delimitation of the BH, we identified only 163 costly and time-consuming (Rondinini et al., 2014). The EAR potenti-
ecoregions (20.4%) as hotspots based on our criteria and data of en- ality can be improved when associated with quantitative scenarios that
demism and habitat loss (Fig. 4). We were able to highlight 44 ecor- assess the impact of alternative paths of socioeconomic development on
egions that are not regarded as a BH but have a high number of ex- biodiversity (Pereira et al., 2010). With the expected upcoming im-
pected extinctions, such as Dry Chaco, Jian Nan Subtropical Evergreen provement of high-resolution and real-time satellite data for land use
Forests, and Changjiang Plain Evergreen Forests. On the other hand, (Pettorelli et al., 2014; Hansen et al., 2013; https://www.
250 of the ecoregions considered as BH were not considered as such by globalforestwatch.org/), the EAR comes as a strong ally for a rapid
our criteria. These ecoregions have on average, less predictions of ex- assessment of current and future impacts on ecoregions and the defi-
tinction when compared with those that are solely categorized as hot- nition of priorities for conservation action. With this long-standing
spot by our classification or those that share the hotspot status by both “ecological law”, we were able to compare the level of vulnerability of
criteria (Kruskal-Wallis χ2 = 303.2, df = 3, p < .01; Dunn's Test ecoregions, as was done for “Crisis Ecoregions” (Hoekstra et al., 2005),
p < .05; Appendix A, Fig. A4a). Moreover, we found a significant but also including the irreplaceability by considering the expected ef-
difference concerning the distribution in the axis of habitat loss and fect of habitat loss on biodiversity.
endemic richness when comparing the ecoregions classified as BH with A significant result of our study is the comparison with the suc-
the ecoregions classified as hotspots by our criteria (χ2 = 1554.5, cessful approach for conservation priorities, the Biodiversity Hotspots

5
D. Gonçalves-Souza, et al. Biological Conservation 246 (2020) 108579

Fig. 3. Bivariate map displaying the relationship between the number of endemics vertebrates predicted to extinction due to the projected accumulated habitat loss
until 2040 and the percentage of the ecoregion area formally protected. Darker blue shades indicate more coverage by protected areas. Darker red shades indicate
more predictions of extinction. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

ecoregions and countries. Because of that, planning conservation po-

licies for an entire BH can be hindered because of international conflicts
and difficulties in resource allocation. As a result of that, efforts to
identify critical areas for conservation on finer scales within the hot-
spots boundaries have been arising in the conservation literature (e.g.,
Faleiro et al., 2013). By using ecoregions to identify future global
priority areas, our results collaterally highlighted smaller areas, inside
the hotspots, with different levels of threat. For instance, Madagascar
consists of eleven ecoregions and exhibited a higher number of ex-
pected extinctions on the coastal region, which comprises of Mada-
gascar Lowland Forests and Madagascar Dry Deciduous Forests ecor-
egions. Moreover, our analysis complements the hotspots delimitation,
pointing to ecoregions that are not considered hotspots but will be at a
high biodiversity risk. For example, the Dry Chaco was considered
vulnerable by our analysis, following other studies that had already
point it out as an ecoregion that is highly representative of Neotropical
biodiversity (e.g., Loyola et al., 2009; Villalobos et al., 2013).
Considering that the main effort to hold back biodiversity loss has
been the implementation of protected areas, our analyses contributed to
its evaluation. By using endemic vertebrates' distribution and focusing
on the most degraded habitats, our approach prioritized areas that are
vulnerable and harbors irreplaceable biodiversity. Our results suggest
that the current distribution of protected areas might not be consonant
with the areas that in the future might harbor most of the threatened
Fig. 4. The relationship of vulnerability (current accumulated habitat loss) and endemic vertebrates. Previous studies have already pointed to the lack
irreplaceability (predictions of extinction) in the ecoregions in the present time. of overlap between loss of habitat and the location of protection sites
The colors of the dots indicate whether the ecoregion is considered a biodi- (Hoekstra et al., 2005; Kehoe et al., 2017a). One explanation for this
versity hotspot (red) or not (black). The horizontal dashed line represents the
pattern could be that protected areas are favored to be in more intact
median of endemism, the vertical line represents 50% of accumulated habitat
areas or in areas with distinct species richness. Our data appears to
loss, and they delimit four quadrants (I-IV). By our definition, we considered an
partially support this scenario, since we found some correlation be-
ecoregion a hotspot when their habitat loss was above 50% and the endemism
above the median (II quadrant). (For interpretation of the references to colour tween protection of the ecoregions and current remaining habitat.
in this figure legend, the reader is referred to the web version of this article.) However, we found no correlation between the protection coverage and
the distribution of endemics vertebrates. It is important to note that
such orientation of the protection network is only relevant to con-
(Myers et al., 2000). The BH have received notable attention in terms of
servation when the protected areas are located in intact areas that are
the scientific debate and as a tool to attract investments to conserva-
representative of biodiversity, but are also under the pressure of human
tion, highlighting global sensitive areas due to their high plant en-
activities, otherwise the protected area presence might not represent a
demism and severe habitat loss (Brooks et al., 2006). The delimitation
change in the fate of a landscape (see Monteiro et al., 2018).
of the hotspots encompasses large areas, often harboring multiple

6
D. Gonçalves-Souza, et al. Biological Conservation 246 (2020) 108579

Considering the scenario of habitat destruction, thousands of ver- destruction. Because of its importance, the information on projected
tebrate species might face extinction risk in the future. Being aware of habitat destruction and its effect on biodiversity must be included in
the number of future extinctions and where they are more likely to conservation prioritization (Dobrovolski et al., 2013; Pouzols et al.,
occur, gives us an opportunity window to take precautionary actions 2014) if we expect to shift the ongoing trend of an even more severe
and avoid further biodiversity loss. However, to make these actions biodiversity loss. Moreover, our analysis contributes to the refinement
possible, we must mitigate the primary driver of habitat loss, the of global conservation priorities, such as the BH framework, by dis-
agriculture expansion. A “zero deforestation” policy has been proposed playing the gradient of threat inside their delimitations and by pro-
in many fora (e.g., Brown and Zarin, 2013) and should be broadened to posing the addition of new areas relevant for conservation. We also
incorporate the conservation of nonforested natural environments, suggest that the global protected area network, our fundamental con-
which are also threatened and represent habitat for many species. servation effort, might not be aligned with probable upcoming verte-
Moreover, it is crucial to recognize that agricultural activity can be brates' extinction in the terrestrial ecoregions. Mitigate the biodiversity
combined with biodiversity-friendly practices. It is possible to meet the loss figures as one of the main challenges for humanity in this century.
growing demands for agricultural products and simultaneously improve Using global data and projections of socioeconomic change allied with
the conservation of species. Foley et al. (2011) arguments that if we fundamental theoretical ecological tools can help to quantify and
direct our efforts into improve yields of low productive areas, this could overcome the upcoming challenges.
result in the actual improvement of environmental conditions without
requiring expansion of agriculture to meet the demands for agricultural CRediT authorship contribution statement
products. An instance of this is the recent history of the Brazilian
Amazon forest. From 2005 to 2013, there was a 70% decrease in the Daniel Gonçalves-Souza: Conceptualization, Methodology, Data
deforestation rates in Brazilian Amazon simultaneously with an incre- curation, Formal analysis, Validation, Writing - original draft,
ment in soy and beef production (Nepstad et al., 2014). This pattern is a Writing - review & editing, Funding acquisition. Peter H. Verburg:
result of governmental, market, and systematic societal policies of Methodology, Data curation, Formal analysis, Software, Writing -
sanctions for irregular producers and positive reinforcements to pro- review & editing. Ricardo Dobrovolski: Conceptualization, Formal
ducers that promote productivity and sustainability. This practical si- analysis, Data curation, Funding acquisition, Methodology,
tuation is a showcase of how future policies must be taken from now on Supervision, Validation, Writing - original draft, Writing - review &
to balance conservation and economic growth (but see the recent in- editing.
crease in deforestation, Dobrovolski et al., 2018).
Here, we present the possible consequences that the reduction of Acknowledgments
natural areas can have on biodiversity. However, our results should be
seen with cautious. Our model tended to overestimate the threat of DGS received fellowships of the Programa de Bolsas de Iniciação
extinction, and we acknowledge that this is probably related to the Científica (PIBIC) of UFBA, the Fundação de Amparo à Pesquisa do
uncertainty about the exact shape and slope of the EAR, that can affect Estado da Bahia (FAPESB) and the Coordenação de Aperfeiçoamento de
the accuracy of estimates based on this method (Guilhaumon et al., Pessoal de Nível Superior (CAPES). RD research is supported by CNPq
2008; Keil et al., 2015). There are also inherent problems with the more (grant 461665/2014-0) and FAPESB (grant JCB0051/2016). We would
simplistic equation of EAR we utilized here that assumes that all species like to thank Bruno Vilela, Pedro Rocha, Rilquer Mascarenhas and
respond equally to the loss of habitat, being this habitat loss permanent, Sidney Gouveia for helpful comments on the manuscript.
instantaneous, and resulting in a single contiguous fragment of habitat
surrounded by an inhospitable matrix. In the real world, habitat loss is a Declaration of competing interest
historical process, that in some situations, leaves a fragmented land-
scape with varying degrees of a tolerable matrix. The EAR is a neutral The authors declare that they have no known competing financial
model, that is, a model that assumes that the identity of the species or interests or personal relationships that could have appeared to influ-
individuals is not relevant to explain a pattern (Hubbell, 2001). Al- ence the work reported in this paper.
though neutral models have pertinent criticisms, this kind of model
remains relevant to address specific types of questions, especially for Appendix A. Supplementary data
macroecological ones. While our approach might not be the best to
predict extinction in a fragment of a landscape, a neutral model like Supplementary data to this article can be found online at https://
EAR can function as null hypothesis and as a start point for under- doi.org/10.1016/j.biocon.2020.108579.
standing the general pattern of extinction at a global scale (Maurer &
McGill, 2004), which was the main purpose of our study. In our model, References
we did not consider that some species could respond to the loss of
habitat with migration to more suitable areas or even adapt to the new Arrhenius, O., 1921. Species and area. J. Ecol. 9, 95–99. https://doi.org/10.2307/
conditions. By not considering this, we may have estimated the ex- 2255763.
Asselen, S., Verburg, P.H., 2013. Land cover change or land-use intensification: simu-
tinction due to habitat loss to a higher number than what we will ob- lating land system change with a global-scale land change model. Glob. Chang. Biol.
serve. Also, although the area is a strong predictor of species richness, 19, 3648–3667. https://doi.org/10.1111/gcb.12331.
numerous threats can have an impact on the species survivorship that Balmford, A., Green, R.E., Jenkins, M., 2003. Measuring the changing state of nature.
Trends Ecol. Evol. 18, 326–330. https://doi.org/10.1016/S0169-5347(03)00067-3.
we did not address. In this study, we presented a method to evaluate Bellard, C., Leclerc, C., Leroy, B., Bakkenes, M., Veloz, S., Thuiller, W., Courchamp, F.,
threats to biodiversity and conservation efforts using bare minimum 2014. Vulnerability of biodiversity hotspots to global change. Glob. Ecol. Biogeogr.
data. Other studies have also projected scenarios of extinction, using 23, 1376–1386. https://doi.org/10.1111/geb.12228.
BirdLife International and NatureServe, 2016. Bird Species Distribution Maps of the
multiples variables, since the stressors of biodiversity extent beyond
World. vol. Version 6.0 BirdLife International, Cambridge,UK and NatureServe,
habitat area (Kehoe et al., 2017b; Chaudhary et al., 2015; Newbold Arlington, USA.
et al., 2015). Furthermore, habitat loss combined with other threats Brooks, T.M., Mittermeier, R.A., Mittermeier, C.G., Da Fonseca, G.A., Rylands, A.B.,
Konstant, W.R., Hilton-Taylor, C., 2002. Habitat loss and extinction in the hotspots of
could lead to more extinctions than would be expected with the simple
biodiversity. Conserv. Biol. 16, 909–923. https://doi.org/10.1046/j.1523-1739.
sum of threats if we consider the synergetic effects (Romero-Muñoz 2002.00530.x.
et al., 2019). Brooks, T.M., Mittermeier, R.A., da Fonseca, G.A., Gerlach, J., Hoffmann, M., Lamoreux,
In sum, our study quantified the possible the magnitude of terres- J.F., Rodrigues, A.S., 2006. Global biodiversity conservation priorities. Science 313,
58–61. https://doi.org/10.1126/science.1127609.
trial vertebrates' extinction worldwide in relation to the habitat

7
D. Gonçalves-Souza, et al. Biological Conservation 246 (2020) 108579

Brown, S., Zarin, D., 2013. What does zero deforestation mean? Science 342, 805–807. 2017–2031. https://doi.org/10.1007/s10531-008-9570-6.
https://doi.org/10.1126/science.1241277. Monteiro, L.M., Brum, F.T., Pressey, R.L., Morellato, L.P.C., Soares-Filho, B., Lima-
Ceballos, G., Ehrlich, P.R., Barnosky, A.D., García, A., Pringle, R.M., Palmer, T.M., 2015. Ribeiro, M.S., Loyola, R., 2018. Evaluating the impact of future actions in minimizing
Accelerated modern human–induced species losses: entering the sixth mass extinc- vegetation loss from land conversion in the Brazilian Cerrado under climate change.
tion. Sci. Adv. 1, e1400253. https://doi.org/10.1126/sciadv.1400253. Biodivers. Conserv. 1–22. https://doi.org/10.1007/s10531-018-1627-6.
Chaudhary, A., Verones, F., de Baan, L., Hellweg, S., 2015. Quantifying land use impacts Maurer, B.A., McGill, B.J., 2004. Neutral and non-neutral macroecology. Basic and
on biodiversity: combining species–area models and vulnerability indicators. Applied Ecology 5, 413–422. https://doi.org/10.1016/j.baae.2004.08.006.
Environmental science & technology 49, 9987–9995. https://doi.org/10.1021/acs. Myers, N., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A., Kent, J., 2000.
est.5b02507. Biodiversity hotspots for conservation priorities. Nature 403, 853–858. https://doi.
Dobrovolski, R., Loyola, R.D., Guilhaumon, F., Gouveia, S.F., Diniz-Filho, J.A.F., 2013. org/10.1038/35002501.
Global agricultural expansion and carnivore conservation biogeography. Biol. Nepstad, D., McGrath, D., Stickler, C., Alencar, A., Azevedo, A., Swette, B., Armijo, E.,
Conserv. 165, 162–170. https://doi.org/10.1016/j.biocon.2013.06.004. 2014. Slowing Amazon deforestation through public policy and interventions in beef
Dobrovolski, R., Loyola, R., Rattis, L., Gouveia, S.F., Cardoso, D., Santos-Silva, R., Diniz- and soy supply chains. Science 344, 1118–1123. https://doi.org/10.1126/science.
Filho, J.A.F., 2018. Science and democracy must orientate Brazil's path to sustain- 1248525.
ability. Perspectives in Ecology and Conservation 16, 121–124. https://doi.org/10. Newbold, T., Hudson, L.N., Hill, S.L., Contu, S., Lysenko, I., Senior, R.A., Day, J., 2015.
1016/j.pecon.2018.06.005. Global effects of land use on local terrestrial biodiversity. Nature 520, 45–50. https://
ESA, 2017. Land cover CCI product user guide version 2.0. Available at. http://maps.elie. doi.org/10.1038/nature14324.
ucl.ac.be/CCI/viewer/download/ESACCI-LC-Ph2-PUGv2_2.0.pdf> (Access in jul 17. Nori, J., Loyola, R., 2015. On the worrying fate of data deficient amphibians. PLoS One
2018). 10, e0125055. https://doi.org/10.1371/journal.pone.0125055.
Faleiro, F.V., Machado, R.B., Loyola, R.D., 2013. Defining spatial conservation priorities Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess, N.D., Powell, G.V.,
in the face of land-use and climate change. Biol. Conserv. 158, 248–257. https://doi. Underwood, E.C., Loucks, C.J., 2001. Terrestrial ecoregions of the world: a new map
org/10.1016/j.biocon.2012.09.020. of life on earth: a new global map of terrestrial ecoregions provides an innovative tool
Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Helkowski, for conserving biodiversity. BioScience 51, 933–938.
J.H., 2005. Global consequences of land use. Science 309, 570–574. https://doi.org/ Pereira, H.M., Leadley, P.W., Proença, V., Alkemade, R., Scharlemann, J.P., Fernandez-
10.1126/science.1111772. Manjarrés, J.F., Chini, L., 2010. Scenarios for global biodiversity in the 21st century.
Foley, J.A., Ramankutty, N., Brauman, K.A., Cassidy, E.S., Gerber, J.S., Johnston, M., Science 330, 1496–1501. https://doi.org/10.1126/science.1196624.
Balzer, C., 2011. Solutions for a cultivated planet. Nature 478, 337–342. https://doi. Pettorelli, N., Laurance, W.F., O'Brien, T.G., Wegmann, M., Nagendra, H., Turner, W.,
org/10.1038/nature10452. 2014. Satellite remote sensing for applied ecologists: opportunities and challenges. J.
Grooten, M., Almond, R.E.A., 2018. Living Planet Report 2018: Aiming Higher. WWF, Appl. Ecol. 51, 839–848. https://doi.org/10.1111/1365-2664.12261.
Gland, Switzerland, pp. 22–100. Pouzols, F.M., Toivonen, T., Di Minin, E., Kukkala, A.S., Kullberg, P., Kuusterä, J.,
Guilhaumon, F., Gimenez, O., Gaston, K.J., Mouillot, D., 2008. Taxonomic and regional Moilanen, A., 2014. Global protected area expansion is compromised by projected
uncertainty in species-area relationships and the identification of richness hotspots. land-use and parochialism. Nature 516, 383–386. https://doi.org/10.1038/
Proc. Natl. Acad. Sci. 105, 15458–15463. https://doi.org/10.1073/pnas. nature14032.
0803610105. . R Core Team. (2017). R: A Language and Environment for Statistical Computing. Vienna,
Hansen, M.C., Potapov, P.V., Moore, R., Hancher, M., Turubanova, S.A.A., Tyukavina, A., Austria: R Core Team. Available at: http://www.r-project.org/.
Kommareddy, A., 2013. High-resolution global maps of 21st-century forest cover Rodrigues, A.S., Pilgrim, J.D., Lamoreux, J.F., Hoffmann, M., Brooks, T.M., 2006. The
change. Science 342, 850–853. https://doi.org/10.1126/science.1244693. value of the IUCN Red List for conservation. Trends Ecol. Evol. 21, 71–76. https://
Harte, J., Kinzig, A.P., 1997. On the implications of species-area relationships for en- doi.org/10.1016/j.tree.2005.10.010.
demism, spatial turnover, and food web patterns. Oikos 417–427. https://doi.org/10. Romero‐Muñoz, A., Torres, R., Noss, A.J., Giordano, A.J., Quiroga, V., Thompson, J.J.,
2307/3546614. Arispe, R., 2019. Habitat loss and overhunting synergistically drive the extirpation of
Hoekstra, J.M., Boucher, T.M., Ricketts, T.H., Roberts, C., 2005. Confronting a biome jaguars from the Gran Chaco. Diversity and Distributions 25, 176–190.
crisis: global disparities of habitat loss and protection. Ecol. Lett. 8, 23–29. https:// Rondinini, C., Di Marco, M., Visconti, P., Butchart, S.H., Boitani, L., 2014. Update or
doi.org/10.1111/j.1461-0248.2004.00686.x. outdate: long-term viability of the IUCN Red List. Conserv. Lett. 7, 126–130. https://
Hubbell, S.P., 2001. The unified neutral theory of biodiversity and biogeography. In: doi.org/10.1111/conl.12040.
Princeton University Press. Rosenzweig, M.L., 1995. Species Diversity in Space and Time. Cambridge University
IUCN, 2017. Red List Spatial Data. Available at. http://www.iucnredlist.org/download_ Press, Cambridge.
spatial_data> Downloaded on mar 25. 2017. Storch, D., Keil, P., Jetz, W., 2012. Universal species-area and endemics–area relation-
IUCN, 2018. The IUCN red list of threatened species. Version 2018-1. Available at. www. ships at continental scales. Nature 488, 78–81. https://doi.org/10.1038/
iucnredlist.org Downloaded on feb 02. 2018. nature11226.
Jenkins, C.N., Joppa, L., 2009. Expansion of the global terrestrial protected area system. Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C.,
Biol. Conserv. 142, 2166–2174. https://doi.org/10.1016/j.biocon.2009.04.016. Williams, S.E., 2004. Extinction risk from climate change. Nature 427, 145–148.
Jetz, W., Freckleton, R.P., 2015. Towards a general framework for predicting threat status https://doi.org/10.1038/nature02121.
of data-deficient species from phylogenetic, spatial and environmental information. Tilman, D., Balzer, C., Hill, J., Befort, B.L., 2011. Global food demand and the sustainable
Philos. Trans. R. Soc. B 370, 20140016. https://doi.org/10.1098/rstb.2014.0016. intensification of agriculture. Proc. Natl. Acad. Sci. 108, 20260–20264. https://doi.
Kehoe, L., Romero-Muñoz, A., Polaina, E., Estes, L., Kreft, H., Kuemmerle, T., 2017a. org/10.1073/pnas.1116437108.
Biodiversity at risk under future cropland expansion and intensification. Nature TNC, 2009. Terrestrial Ecoregions. Available at. http://maps.tnc.org/gis_data.html>
ecology & evolution 1, 1129–1135. https://doi.org/10.1038/s41559-017-0234-3. Downloaded on May 18. 2017.
Kehoe, L., Senf, C., Meyer, C., Gerstner, K., Kreft, H., Kuemmerle, T., 2017b. Agriculture UNEP-WCMC and IUCN, 2018. Protected Planet: [the World Database on Protected Areas
rivals biomes in predicting global species richness. Ecography 40, 1118–1128. (WDPA)/the Global Database on Protected Areas Management Effectiveness (GD-
https://doi.org/10.1111/ecog.02508. PAME)]. Online], [may/2018]. UNEP-WCMC and IUCN, Cambridge, UK Available
Keil, P., Storch, D., Jetz, W., 2015. On the decline of biodiversity due to area loss. Nat. at. www.protectedplanet.net.
Commun. 6, 8837. https://doi.org/10.1038/ncomms9837. Villalobos, F., Dobrovolski, R., Provete, D.B., Gouveia, S.F., 2013. Is rich and rare the
Laurance, W.F., Sayer, J., Cassman, K.G., 2014. Agricultural expansion and its impacts on common share? Describing biodiversity patterns to inform conservation practices for
tropical nature. Trends Ecol. Evol. 29, 107–116. https://doi.org/10.1016/j.tree. South American anurans. PLoS One 8, e56073. https://doi.org/10.1371/journal.
2013.12.001. pone.0056073.
Lewis, S.L., Maslin, M.A., 2015. Defining the Anthropocene. Nature 519, 171–180. Wolf, S., Schrammeijer, E.A., Schulp, C.J., Verburg, P.H., 2018. Meeting global land re-
https://doi.org/10.1038/nature14258. storation and protection targets: what would the world look like in 2050? Glob.
Loyola, R.D., Kubota, U., da Fonseca, G.A., Lewinsohn, T.M., 2009. Key Neotropical Environ. Chang. 52, 259–272. https://doi.org/10.1016/j.gloenvcha.2018.08.002.
ecoregions for conservation of terrestrial vertebrates. Biodivers. Conserv. 18,