Caja PDF

Comparta fácilmente sus documentos PDF con sus contactos, la web y las redes sociales.

Compartir un archivo PDF Gestor de archivos Caja de instrumento Buscar PDF Ayuda Contáctenos



De Angelo et al 2013 Biol Cons A management landscape for jagaurs .pdf



Nombre del archivo original: De Angelo et al 2013 - Biol Cons - A management landscape for jagaurs.pdf

Este documento en formato PDF 1.4 fue generado por LaTeX with hyperref package / StampPDF Batch 3.0 Windows SPDF_1085 Oct 13 2003, y fue enviado en caja-pdf.es el 26/01/2015 a las 14:55, desde la dirección IP 200.112.x.x. La página de descarga de documentos ha sido vista 1552 veces.
Tamaño del archivo: 3 MB (37 páginas).
Privacidad: archivo público




Descargar el documento PDF









Vista previa del documento


Author's personal copy

Biological Conservation 159 (2013) 422–433

Contents lists available at SciVerse ScienceDirect

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

Understanding species persistence for defining conservation actions: A
management landscape for jaguars in the Atlantic Forest
Carlos De Angelo a,⇑, Agustín Paviolo a, Thorsten Wiegand b, Rajapandian Kanagaraj b, Mario S. Di Bitetti a
a

National Research Council (CONICET), Instituto de Biología Subtropical, Facultad de Ciencias Forestales, Universidad Nacional de Misiones, Asociación Civil Centro de
Investigaciones del Bosque Atlántico, Bertoni 85 N3370BFA, Puerto Iguazú, Misiones, Argentina
b
Department of Ecological Modelling, UFZ Helmholtz Centre for Environmental Research, P.O. Box 500136, DE-04301 Leipzig, Germany

a r t i c l e

i n f o

Article history:
Received 15 September 2012
Received in revised form 14 December 2012
Accepted 15 December 2012
Available online xxxx
Keywords:
Attractive sink
Human persecution
Land-cover conditions
Panthera onca
Two-dimensional habitat model
Upper Paraná Atlantic Forest

a b s t r a c t
Habitat models constitute useful instruments for understanding species-habitat interactions and can
constitute helpful conservation tools. The Upper Paraná Atlantic Forest (UPAF) of South America still
holds the world’s southernmost jaguar (Panthera onca) population. Our aims were: (i) to test several a
priori hypotheses on the factors affecting jaguar persistence in this region, (ii) to map habitat suitability
and identify areas with potentially conflicting habitat conditions, and (iii) to identify priority areas for
management and improve the conservation initiatives for jaguars and the UPAF. Following an information-theoretic approach, we used presence records of jaguars and pseudo-absences in generalized linear
models. We structured hypotheses into two groups which demand different management actions: land
cover and human persecution. The best model of each group was used to develop a two-dimensional habitat model. Jaguar persistence was favoured by current and historical native forest cover, and hindered by
human land uses. Protection favoured jaguar presence whereas human accessibility and high human
population density had negative effects. The two-dimensional model suggests that <8% (20,670 km2) of
the landscape represents potential core areas for jaguars (good land-cover characteristics and low human
persecution) and 11.8% (32,563 km2) stands as potentially attractive sinks where good land-cover conditions conflict with high human persecution. Reduction of human persecution is urgently needed to
increase the core areas for jaguars in this region, but improvement of land-cover conditions is important
for sustaining the connectivity among jaguar populations that seem to be isolated in different areas of the
UPAF.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction
Understanding the relationship between landscape change and
species persistence is a major issue of interest in applied ecology
because of its direct relationship with biodiversity conservation
(Tilman et al., 1994). Habitat models or species distribution models
(SDMs) constitute useful instruments for predicting species distribution and understanding the species-habitat interactions, but also
they can be used as conservation tools for delineating management
actions (Guisan and Thuiller, 2005; Guisan and Zimmermann,
2000). However, implementation of SDMs in biological conservation is not always a simple task and often demands specific approaches for transforming these models into useful management
tools (Guisan and Thuiller, 2005). Naves et al. (2003), for example,
Abbreviation: AICc, Akaike’s Information criterion corrected for small samples;
GLM, generalized linear model; UPAF, Upper Paraná Atlantic Forest eco-region.
⇑ Corresponding author. Tel.: +54 3757 423511; fax: +54 3757 422370.
E-mail address: biocda@gmail.com (C. De Angelo).
0006-3207/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.biocon.2012.12.021

proposed an approach for mapping habitat suitability for large carnivores that involves two separate models: a natural model targeting habitat suitability regarding reproduction and a human impact
model targeting habitat suitability concerning survival. This
approach allows detection of not only the conventional categories
where conditions for reproduction and survival are positively correlated (i.e., matrix, sink or poor habitat, and source or good habitat), but also otherwise undetectable areas with good conditions
for reproduction though with low survival (attractive sinks), and
areas with poor conditions for reproduction but with high survival
(refuges). These areas have important management implications,
mainly because attractive-sink areas may constitute ecological
traps with large effects on populations’ survival (Delibes et al.,
2001).
The Upper Paraná Atlantic Forest (UPAF) of Argentina, Brazil
and Paraguay, is the largest eco-region of the South American
Atlantic Forest, and it constitutes one of the world’s most endangered eco-regions (Mittermeier et al., 2005; Ribeiro et al., 2009).
The main conservation initiative developed for the UPAF is the

Author's personal copy

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

Biodiversity Vision (Di Bitetti et al., 2003), a tri-national conservation strategy designed to sustain a viable population of jaguars
(Panthera onca), considering this species as an umbrella for the vast
biodiversity that the UPAF hosts (Miller and Rabinowitz, 2002).
However, only scarce information existed about jaguars in the
UPAF when this conservation plan was developed, and one of Biodiversity Vision’s aims was the study and monitoring of jaguars
and the use of this information for validating this biodiversity conservation strategy (Di Bitetti et al., 2003). Additionally, jaguars are
among the most threatened species in the Atlantic Forest and the
UPAF hosts two Jaguar Conservation Units (JCUs) where jaguar experts encouraged research and conservation actions for this species
(Sanderson et al., 2002b). Considering these demands of knowledge about jaguars in the UPAF, different research initiatives were
developed that emphasized the urgent need of a deeper evaluation
of the remaining habitat for jaguars in this region to delineate actions at an eco-regional scale (Cullen et al., 2005; De Angelo et al.,
2011a,b; Paviolo et al., 2008).
In this study we compiled the previous information obtained
about jaguars in the UPAF and used this information to: (i) test several a priori hypotheses on the factors influencing jaguar habitat
suitability, (ii) map habitat suitability for jaguars and identify areas
with potentially conflicting habitat conditions such as attractive
sinks, and (iii) detect priority areas for implementing specific management actions and improving the conservation plans developed
for jaguars and the UPAF.
2. Material and methods
2.1. Study area
The UPAF is a subtropical and semi-deciduous forest (annual
precipitation range: 1000–2200 mm; mean temperature range:
16–22 °C), and it constitutes a highly degraded and fragmented region, where less than 8% of the forest remains (Di Bitetti et al.,
2003). The history and dynamics of human settlement, land-use
change, and fragmentation processes are heterogeneous along
the UPAF (De Angelo, 2009; Izquierdo et al., 2008; Jacobsen,
2003). In the Brazilian UPAF, most of the forest was replaced
around the middle of the last century (Ribeiro et al., 2009), while
the Paraguayan UPAF has a more recent but accelerated process
of forest destruction (Huang et al., 2007). The Argentinean UPAF
has a long history of human settlement and forest exploitation
but with much lower rates of forest replacement (Izquierdo
et al., 2008).
We selected an area of 276,843 km2 at the border shared by
Brazil, Paraguay and Argentina, which includes most of the remnants of the UPAF (Fig. 1). This area encloses all the area surveyed
by De Angelo et al. (2011b) in their monitoring of jaguar presence
and is the same area used by De Angelo et al. (2011a) in their habitat suitability analysis for pumas and jaguars using a presenceonly technique.
2.2. Species data
We utilized the presence records of jaguars collected by participatory monitoring between 2002 and 2008 (De Angelo et al.,
2011b). We obtained records from different sources (tracks, scats,
camera traps, radio-tracked animals, etc.), that were carefully selected and accurately identified for avoiding false positives. Jaguar
tracks were identified using a discriminant model developed for
recognizing jaguar tracks (De Angelo et al., 2010), and scats were
identified through specific molecular markers developed for differentiating jaguar and puma (Puma concolor) faecal samples (Haag
et al., 2009).

423

In total we obtained 974 jaguar records (De Angelo et al.,
2011b). To reduce potential pseudo-replication biases caused by
the unsystematic data collection, we superimposed a grid of 144km2 cells (the size estimated for a female jaguar home range; D.
Sana, unpublished data; A. Paviolo, unpublished data; Cullen
et al., 2005; Paviolo, 2010) and randomly selected one observation
if more than one record occurred in a cell (Kanagaraj et al., 2011).
This resulted in a total of 106 presence records to be used in our
analysis. To test if our results were influenced by this particular
selection of records, we created 10 further subsets of 106 records
following the same procedure. This allowed us to explore whether
the selected presence records were representative and whether
models constructed with alternative sets of presence records
agreed (see below Sections 2.3 and 2.4).
To obtain a binomial response variable we followed the
approach developed by Engler et al. (2004), and we generated
randomly the same number of pseudo-absences as presences (Liu
et al., 2005). To this end, we followed several rules to ensure that
pseudo-absences were located inside surveyed areas but not in
areas that were known to be suitable areas for jaguars (Appendix
A and Fig. A1). We also generated 10 further sets of pseudoabsences for model validation (see below Section 2.4).

2.3. Biological hypotheses and environmental variables
SDMs that predict average habitat suitability based on a single
function may overlook areas where habitat conditions related to
key factors with different management requirements are conflicting (Kanagaraj et al., 2011; Naves et al., 2003; Nielsen et al.,
2006). For example, it is well known that deaths of large carnivors
are mainly caused by humans, but nutritional condition determines reproductive rate (Naves et al., 2003; Woodroffe and Ginsberg, 1998). If key factors that determine reproduction and
survival are not positively correlated, a single function SDM will
overlook attractive sinks (good conditions for reproduction but
low survival) and refuges (poor conditions for reproduction but
with high survival). Thus, using a model based on two SDMs that
describe habitat suitability from the perspective of different key
factors that affect either survival or reproduction, allows for a more
subtle and management relevant assessment of habitat suitability.
Indeed, we can identify such two management–relevant key
factors for the jaguar in the UPAF. First, landscape conditions related with land cover and physical environment are important
determinants of jaguar habitat suitability at a regional scale (e.g.
forest cover, presence of water, or different human land uses; see
Table 1). The main management actions associated with these conditions are related to policies of forest restoration and territorial or
land-use planning (e.g. defining which human land uses will be
promoted in certain regions, designing corridors, protecting river
basins) (e.g. in Fernández et al., 2006; Muntifering et al., 2006;
Wikramanayake et al., 2004). Second, the presence or absence of
this species is also determined by direct human persecution of jaguars and their prey (see Table 2). The most important management
actions needed to improve habitat conditions in relation with these
threats are different from those mentioned before: the priority actions would be protection, law enforcement, and actions for reducing jaguar mortality (e.g. for reducing poaching activity and other
sources of jaguar mortality as road kills) (e.g. in Nielsen et al.,
2004; Woodroffe and Ginsberg, 1998). By analyzing habitat
suitability with respect to these two dimensions we can identify
critical areas that need to be prioritized for the different management actions. We therefore tested several a priori hypotheses on
factors that determine jaguar habitat suitability regarding the
two main key factors: land-cover and physical environment [L],
and human persecution [H] (Tables 1 and 2 respectively). We then

Author's personal copy

424

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

57°W

54°W

51°W

Jaguar presence

21°S

International border
21°S

Study area
Native forest remnants
Upper Paraná Atlantic Forest

24°S

M b a ra c a y ú
reg ion

B R A Z I L

24°S

Up p er P ar an á Pa ra n ap a n e m a
reg ion

P A R A G U AY

27°S

M i s i o n es
Gr ee n
Corr idor

A
0 25 50

57°W

R

G

E

N

T

I N

A

100
km

54°W

Fig. 1. Portion of the Upper Paraná Atlantic Forest eco-region selected for jaguar habitat suitability analysis. The right corner inset details the location of the study area in
South America. Forest remnants include native forest and marshlands and they correspond to estimates for the year 2004 done by De Angelo (2009). Dots represent the 106
presence records of jaguars used in our analysis.

used the most parsimonious SDMs of these groups as the two
dimensions for categorizing the habitat for jaguars (Fig. 2).
To describe the landscape characteristics and human pressures
that represent the different hypotheses we used a total of
9 + 4 10 = 49 variables with a spatial resolution of 330 330 m
(Appendix B; Table B1). The first nine variables described the average conditions within each cell and included topography (i.e., elevation and slope), human accessibility, distance to rivers, roads and
towns, protection category, rural population density and the mean
human population density during the last 40 years (Table B1). To
capture the jaguar perception of the different landscape elements
we also used 4 10 neighborhood variables to describe for example the frequency of cells occupied by native forest within four different neighborhood radii (1, 4, 7, and 10 km) around the focal cell
(Kanagaraj et al., 2011; Naves et al., 2003). We constructed neighborhood variables from the categories ‘current native forest’,
‘intensive agriculture’, ‘extensive pastures’, ‘pine plantations’,
‘small farms’, ‘rivers’, ‘roads’, and ‘towns’. Additionally, we added

four neighborhood variables describing the historical condition of
the forest in 1973 (forest73_r) (Table B1). The local connectivity
for radii of 1 km (connect_r1) was discarded because of its redundancy with forest_r1.
To assess whether the 106 presence records selected for the
analysis were a representative sample of the total 974 records
available or not, we compared the distribution profile of these records for each independent variable with the distribution profile
of the other 10 subsets of presence records. We observed no significant differences for any of the 48 variables (Appendix B; Table B2),
indicating that the subset selected for the analysis was representative of the variables used.
2.4. Model selection and evaluation
We combined the selected presences and pseudo-absences as a
binary response variable in generalized linear models (GLMs)
with logit-link function (McCullagh and Nelder, 1989). We also

Author's personal copy

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

425

Table 1
Description of the general and the main particular hypotheses evaluated in relation to the land-cover and physical environment as determinants of jaguar presence in the Upper
Paraná Atlantic Forest. Variables used for each hypothesis are in brackets (definitions in Table B1).
General hypotheses and justification

Particular hypotheses

Native forest (F): jaguar presence is directly conditioned by the presence and
characteristics of the native forests. Cullen (2006) described this relationship in
jaguar habitat selection in the northern UPAF. Previous studies on carnivores
also showed that the amount and connectivity of native forest have been
important predictors (Conde et al., 2010; Naves et al., 2003; Rodríguez-Soto
et al., 2011; Schadt et al., 2002). The fast and varied dynamics of UPAF
fragmentation means that the historical process of forest loss might also be
involved as predictor of jaguar persistence as occurred with other species in the
Atlantic Forest (Metzger et al., 2009) and in other areas (e.g. Brooks et al., 1999)
Physical Environment (PE): physical and geographical characteristics of the
landscape are important predictors of jaguar presence. Jaguars are positively
associated with water courses (Crawshaw and Quigley, 1991; Cullen, 2006;
Hatten et al., 2005; Somma, 2006), and they were related with higher
elevations and slopes in Mexican hills in Tamaulipas (Ortega-Huerta and
Medley, 1999) but in general with lower elevations along Mexico (RodríguezSoto et al., 2011)
Human land uses (U): human land uses have negative effects on jaguar presence at
this scaledue mainly to changes in land cover, but also lower prey availability
(e.g. intensive agriculture), and higher human presence (e.g. pastures and
farms) not only in the modified areas but also in the surroundings. Land use
characteristics have been important predictors of jaguars (Conde et al., 2010)
and many other carnivores’ presence (Kanagaraj et al., 2011; Revilla et al.,
2004).
Land-cover and physical environment combined (L): Jaguar presence is
determined by the characteristics of native forest, physical environment and/or
human land uses

(F1) The amount of forest (forest_r) favors the presence of jaguars. (F2) The local
connectivity of forest (connect_r) favors the presence of jaguars. (F3) Both, amount
and connectivity of forest are important for jaguar presence. (F4) The historical
presence of forest (forest73_r) determines the presence of jaguars. (F5) Combined
effects of current and past forest characteristics are important

(PE1) The presence of rivers (rivers_d and rivers_r) is important for jaguar
presence. (PE2) Elevation is an important predictor of jaguar presence (elevation).
(PE3) Jaguars are present mainly in higher slopes (slope). (PE4) Combined effects
of environment characteristics are important

(U1) Each land use has a particular negative impact (farms_r, int_agr_r, pastures_r
and plant_r) at this scale mostly due to differences in cover, but also in prey and
human presence. U2) Combined characteristics of different land uses are
important

L = (1) F + PE + U. (2) F + PE. (3) F + U. (4) PE + U

Table 2
Description of the general and the main particular hypotheses evaluated in relation to the human persecution of jaguars and their prey as determinants of jaguar presence in the
Upper Paraná Atlantic Forest. Variables used for each hypothesis are in brackets (definitions in Table B1).
General hypotheses and justification

Particular hypotheses

Protection and human access (PA): Jaguar presence is determined by habitat
protection and it is negatively affected by the access of humans. Both,
protection and human access, are directly related with poaching pressure on
jaguars and their prey, but also with other direct impacts on jaguars (e.g. traffic
killings) and forest (e.g. logging). Recent works demonstrated a direct
relationship between protection and jaguar density in the UPAF (Paviolo et al.,
2008), and the same pattern was observed with its main prey species where not
only protection levels but also human access were important prey abundance
predictors (Di Bitetti et al., 2008; Paviolo et al., 2009). Many authors have
described negative association of large carnivores with human access (Conde
et al., 2010; Kerley et al., 2002; Nielsen et al., 2004) and the positive effects of
protected areas (e.g. Woodroffe and Ginsberg, 1998).
Rural population (RP): Rural population density is negatively related with jaguar
presence. Human density is a good predictor of human impacts (Sanderson
et al., 2002a) and it is associated with carnivores’ extinction risk (Cardillo et al.,
2004). Altrichter et al. (2006) enhanced the importance of the history of human
settlements in the existence of jaguars in Argentinean Chaco
Human persecution combined (H): Jaguar presence is determined by the combined
effects of protection, human access and rural population density

(PA1) Jaguar presence is favoured by protection (protect_cat). (PA2) Human access
(access_cost, road_d, road_r, towns_d and towns_r) negatively affects jaguar
presence. (PA3) Protection and human access are important predictors of jaguar
presence and there is an interaction between them because jaguars often use
access ways inside protected areas (protect_cat access_cost)

compared linear, quadratic, and cubic GLM functions to test possible non-linear adjustments for each predictor variable (Guisan and
Zimmermann, 2000). Our primary objective was to compare the
support received by several a priori hypotheses on the factors that
determine jaguar presence and absence in the UPAF. We therefore
followed an information theoretic approach for model selection
(Burnham and Anderson, 2002). This method has the additional
advantage that biological knowledge is used in the process of variable selection through developing models following a priori
hypotheses, ensuring biological interpretation of the resultant
models. Also this approach allows one to assess the relative levels
of support for the competing hypotheses and to draw inferences
from the whole set of competing models (Burnham and Anderson,
2002; Johnson and Ommland, 2004).
To organize model selection, we grouped the hypotheses hierarchically, starting with the two main groups: land-cover and phys-

(RP1) Jaguars occur in areas with low rural population density (population_2000).
(RP2) Jaguars are present in areas historically low populated (population_hist)

H = PA + RP

ical environment (L) and human persecution (H) (Tables 1 and 2
respectively). Each group contained a set of general hypotheses
disaggregated into particular hypotheses. The particular hypotheses were described by GLMs through different combinations of
variables. In the case that more than one variable representing a given hypothesis were highly correlated (r > 0.7), we conducted variable reduction (details in Appendix C1).
We used the Akaike Information Criterion corrected for small
samples (AICc) for model selection, because it allows a comparison
of the models in their relative fit to the data while penalizing model complexity (Johnson and Ommland, 2004). We selected the
model with the lowest AICc for representing each particular
hypothesis (Tables 1 and 2). We then compared the models selected for the particular hypotheses within each general hypothesis using the same criterion. Finally, we compared the general
hypotheses in each main group and selected one final model for

Author's personal copy

426

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

Fig. 2. Two-dimensional habitat categorization based on Naves et al. (2003), but using a different definition of habitat dimensions according to the available information for
jaguars in the Upper Paraná Atlantic Forest (see Tables 1 and 2) and the landscape management alternatives for improving the habitat. The ‘management arrows’ indicate the
direction of habitat improvement that would occur if this actions are implemented.

the land-cover and physical environment group and one model for
the human persecution group. We also developed a global model
combining the two final models (Burnham and Anderson, 2002).
We evaluated the final and the global models by the area under
the receiver operating characteristic curve (AUC; Guisan and
Zimmermann, 2000), the percentage of correctly predicted presences and pseudo-absences, and the continuous Boyce index
(Hirzel et al., 2006). In order to evaluate potential overfitting, we
conducted a cross validation (Fernández et al., 2003; Kanagaraj
et al., 2011). To test if the particular selection of records and pseudo-absences influenced our results, we also estimated the prediction ability of the models based on the 10 alternative sets of
presence and pseudo-absences (details in Appendix C2).
We mapped the final models with a 330-m resolution to obtain
the relative probability of jaguar presence within the study area
and transformed the maps into categories of habitat quality following Naves et al. (2003) and Hirzel et al. (2006) (Appendix C2).
2.5. Two-dimensional habitat categorization
We used the selected models of the land-cover and physical
environment group and the human persecution group for categorizing the habitat for jaguars in a two dimensional way (Fig. 2). Because we modeled these dimensions only with presence data
(neither actual reproduction nor mortality data), we termed potential sources as core areas, and sinks and attractive sinks as sinklike
and attractive sinklike areas (Kanagaraj et al., 2011).
To validate our categorization particularly in relation to sinklike
and attractive sinklike areas, we used independent records of killed
or removed jaguars (n = 30) that occurred between 1998 and 2008
in the Argentinean part of the study area (De Angelo, 2009; Paviolo,
2010). Using a chi-square test, we compared the proportion of dead
and removed animals that occurred in matrix habitat, sinks (i.e., sinklike and attractive sinklike areas), and core areas, with the expected
value according to the surface available of each category in this area.
2.6. Validating and improving conservation plans
We used the two-dimensional habitat categorization to validate
the different conservation initiatives for jaguars and the study area.

We overlaid our jaguar habitat model with the Biodiversity Vision
conservation landscape (Di Bitetti et al., 2003) to observe the
agreements and disagreements between both management maps.
Based on the information available in 1999, Sanderson et al.
(2002b) defined Jaguar Conservation Units as areas that can be
considered as able to preserve a large enough (at least 50 breeding
individuals) population of resident jaguars to be potentially selfsustaining over the next 100 years. Alternatively, they included
areas containing fewer jaguars but with adequate habitat and a
stable, diverse prey base, such that jaguar populations in the area
could increase if threats were alleviated. Using these criteria and
the new information available about jaguars in the UPAF, we updated and re-defined the JCUs in this region based on our twodimensional habitat model.
To exemplify how our models can be used in adaptive management for prioritizing actions in these different conservation strategies (Sanderson et al., 2002c), we followed the least-cost corridors
approach of Rabinowitz and Zeller (2010) to identify areas for
alternative corridors connecting not only JCUs but also all core
areas where the presence of jaguars was confirmed in the UPAF.
For this analysis we used our global model as a permeability matrix
and the Corridor tool of ArcGIS to find the least cost area between
pairs of core areas with jaguars and revised JCUs (only the 0.1% of
the grid cell values were extracted, see details of the method in
Rabinowitz and Zeller, 2010).

3. Results
3.1. Determinants of jaguar presence
Land-cover conditions were important predictors of jaguar
presence in the UPAF (Tables 3 and C1). The occurrence of this species was positively related not only with the local amount of forest
and the proximity of forested areas (local connectivity), but also
with the presence of forest in the past. These three characteristics
constituted the best supported model for describing the native
forest hypothesis (F). Among the physical environment hypotheses
(PE), only the frequency of rivers was supported by the data
indicating that jaguars were found more frequently in areas

Author's personal copy

Table 3
Selection of models for predicting jaguar presence in the UPAF according to the main groups of hypotheses. Only selected models from particular hypotheses and their combinations are shown (see Tables C1 and C2). The comparison
among the final models is shown in bold type.
Variables in the final model (+ or – effect)

v2

df

p

D2adj

AICc

DAICc

wr
(%)

Null model
Land cover & physical
(L)
Forest (F)
Physical Environment
(PE)
Land uses (U)
F + PE
PE + U
F+U
F + PE + U

Intersection

0







296

119
6

0.0
3.9

forest_r1(+), connect_r7(+), forest73_r7(+), forest73_r72 ( ), forest73_r73(+)
rivers_r4(+)

121
11

5
1

<0.001
0.001

0.39
0.03

186
287

3
104

10.8
0.0

int_agr_r1( ), farms_r4 ( ), pastures_r4 ( )
forest_r1(+),connect_r7(+), forest73_r7(+), forest73_r72 ( ), forest73_r73(+), rivers_r4(+)
rivers_r4(+), agri_r1( ), farms_r4( ), pastures_r4( )
forest_r1(+), connect_r7(+), forest73_r7(+), forest73_r72( ), forest73_r73(+), int_agr_r1, ( ) farms_r4( ),pastures_r4( )
forest_r1(+), connect_r7(+),forest73_r7(+), forest73_r72 ( ), forest73_r73(+), rivers_r4(+), int_agr_r1( ), farms_r4( ), pastures_r4( )

108
125
110
130
130

3
6
4
8
9

<0.001
<0.001
<0.001
<0.001
<0.001

0.37
0.41
0.36
0.42
0.42

193
184
195
183
185

10
1
12
0
2

0.3
30.6
0.1
41.3
16.8

39

0.0

protect_cat access_cost (+), protect_cat1 access_cost (+), protect_cat2 access_cost (+)

77

3

<0.001

0.25

226

11

0.5

population_hist ( )

22

1

<0.001

0.07

276

61

0.0

87

4

<0.001

0.29

215

0

99.5

141

11

<0.001

0.45

176

0

96.1

Human persecution
(H)
Protection and human
access (PA)
Population density
(RP)
PA + RP
Global
L+H

protect_cat0 access_cost (+), protect_cat1 access_cost (+), protect_cat2 access_cost (+), population_hist ( )
2

3

forest_r1(+),forest73_r7(+), forest73_r7 ( ), forest73_r7 (+), int_agr_r1, ( ) farms_r4( ),pastures_r4( ), protect_cat0 access_cost (+),
protect_cat1 access_cost (+), protect_cat2 access_cost (+), population_hist ( )

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

Hypotheses

D2adj

Notes: Variable abbreviations are from Table B1; v2 is the Wald’s chi-square statistic;
is adjusted explained deviance; AICc is bias-corrected Akaike’s Information Criterion; DAICc is (AICc)I (AICc)min; wr is the AICc weights
expressed in percentages; (+) or ( ) indicates the direction of the effect of the variable for predicting jaguar presence; indicates interaction between variables; _r indicates the variable calculated for 1-, 4-, 7- or 10-km radius.

427

Author's personal copy

428

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

surrounded by rivers. All human land uses showed negative relationship with jaguar presence, but the combination of local frequency of intensive agriculture (1-km radius) with the frequency
of farms and pastures in the surroundings (4-km radius) yielded
the best model of the general human land use hypothesis (U).
When combining the best models of the three general hypotheses related to land-cover and physical environment, we found
that the three models that contained the native forest hypothesis
(i.e., F + PE, F + U, F + PE + U) received similar support from the data
(i.e., DAICc < 2; Tables 1 and 3). The model with the lowest AICc was
the combination of native forest and human land use (i.e., F + U,
hereafter the land-cover model), but it should be noticed that the
difference in AICc between the native forest hypothesis (F) and
the best combined model was only three (Table 3). The best model
correctly classified 83.5% of the presences and pseudo-absences
and yielded an AUC of 0.905, thus indicating good discrimination
ability (Table C4).
In the human persecution group, all particular hypotheses
yielded significant models but we did not find an overarching
hypothesis such as native forest in the land-cover conditions
group. Instead, the best model was obtained by the combination
of the two general hypotheses (Tables 2 and 3, and C2). As
expected, protected areas were positively related with jaguar
presence while the frequency and proximity to roads and towns
showed a negative effect. Human accessibility was selected for
representing the direct effect of human presence on jaguars’ occurrence. We found a higher support for the model that incorporated
the interaction between protection and human access, showing
that the influence of human accessibility changes according to
the protection level (Table C2). Jaguars occurred in areas with
low densities of rural population but both models (present and
the last 30-year average) showed similar support (Table C2).
Although the best human persecution model received less support than the best land-cover model (AICc of 215 vs. 183) it correctly classified 78% of the presences and pseudo-absences, and
yielded AUC of 0.84, indicating good discrimination ability
(Table C4).
The final models of each group were combined into one global
model (details in Tables 3 and C3). In spite of its higher complexity,
this model was selected as the most parsimonious model (lowest
AICc), indicating that both groups of hypotheses (land-cover and human persecution) were important for predicting jaguar presence in
this region. Cross validation showed that these models did not
over-fit the data, and a similar validation success was obtained with
the 10 alternative sets of presences and pseudo-absences (Table C4).

3.2. Two-dimensional habitat categorization
The final models of the land-cover and physical environment
and human persecution groups defined our two-dimensional model (Figs. 3 and D1). The different categories of suitable habitat (i.e.,
low, medium and high suitability from Fig. C1) were used for
increasing the resolution inside each habitat category of our twodimensional model. This resulted in detailed maps of priority management actions for jaguars along the UPAF (marginal areas:
Figs. D2 and D3; core areas: Figs. D4 and D5).
Core areas (suitable conditions predicted by both models) represented only 7.5% of the study area, and most of the region was
covered by matrix (Fig. 3 and Table 4). The highest surface of core
areas was located in Paraguay (42%), but the largest and more continuous core areas were located in the north part of the Argentinean region, including the Iguaçu National Park in Brazil (Fig. 3).
Sinklike and attractive sinklike areas occupied >25% of the study
area. Attractive sinklike areas were more common surrounding
the core areas in Argentina and Paraguay (Fig. 3 and Table 4).

Potential refuges were scarce along the study area and only present
in few regions of Brazil and Paraguay (Fig. 3 and Table 4).
Analyzing the location of dead and removed jaguars in the
Green Corridor showed that a higher proportion of animals were
killed in sinklike and attractive sinklike areas than expected by
the available area of these habitat categories (Fig. E1; Pearson’s
chi-squared test: v2 = 0.007; df = 2; p < 0.01). Moreover, the relatively highest mortality of jaguars occurred in the attractive sinklike areas with the best land-cover conditions, supporting our
hypothesis that these areas are ecological traps (Fig. E1).

3.3. Validating and improving conservation plans
We used the two-dimensional model for validating the conservation landscape designed by the Biodiversity Vision (Di Bitetti
et al., 2003). We found that they detected most of the core areas
that we described for jaguars (Table F1). However, 17% of Biodiversity Vision’s core areas were attractive sinklike areas for jaguars
and need protection. Additionally, our model detected two large
areas in Brazil (Ivinhema State Park and Ilha Grande National Park)
that could be incorporated as core areas (Fig. 3). More than 50% of
the areas classified by the Biodiversity Vision as ‘potential core
areas’ and as ‘forested areas that need assessment’ were classified
by our model as core areas (Table F1). However, most of the areas
classified by the Biodiversity Vision as ‘high potential of becoming
core areas’ were classified by our model as attractive sinklike areas,
which need more protection to constitute core areas. More than
30% of the ‘corridors’ were also classified as attractive sinklike
areas and >20% as matrix, showing that many of these corridors
may not be functional for jaguars (Table F1), demanding high efforts in protection and restoration to become so.
Based on our results we revised the existing Jaguar Conservation Units (JCUs; Fig. 4; Sanderson et al., 2002b) and defined JCU
as core areas with known reproductive populations (from Cullen
et al., 2005; De Angelo, 2009; De Angelo et al., 2011b; McBride,
2009; Paviolo, 2010). We also included the surrounding core areas
that were directly connected to or closer than 23 km from the
reproductive populations (23 km was the maximum distance of a
presence record to a core area). This procedure suggests a redefinition of the shape of the Misiones Green Corridor JCU, and the repositioning of the Upper Paraná – Paranapenama JCU according to the
core areas. Our models suggest the incorporation of the Mbaracayú
Biosphere Reserve and the surrounding areas as a third JCU in Paraguayan UPAF. The least-cost corridors proposed by Rabinowitz
and Zeller (2010) were observed across many of the core areas of
our model, showing the potential importance of the core areas as
stepping stones in a regional and continental conservation strategy
(Fig. 4). Their corridors also confirm the important role of the
Mbaracayú area for regional and global connectivity among jaguar
populations.
The least-cost areas that we detected across the core areas in
our models offer other alternatives for connecting the JCUs in this
region and exposed the important role of the core areas outside the
JCU for an eco-regional jaguar conservation strategy (Fig. 4).
Observing in more detail the habitat conditions in between the
core areas (in the examples shown in Fig. 4), it is possible to use
the two-dimensional model to prioritize the most urgent management actions needed for enlarging or connecting the core areas
(e.g. the different strategies needed in the potential corridors in
eastern Paraguay in Fig. 4C, and the need of increasing protection
in the Green Corridor in Fig. 4D). However, in some areas like most
of the Brazilian UPAF (Fig. 4B) and southern Paraguay (Fig. 3), the
efforts needed for implementing corridors among core areas are
higher and more challenging, demanding both protection and
land-cover improvement for transforming the matrix into suitable

Author's personal copy

429

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

Morro do
Diabo State
Park
Ivinhema State
Park

Upper Paraná
Paranapa nema
re gi on

Ilha Grande
National Park
Perobas Biological
Reserve

Mbaracayú
Reserve

Mba r a ca y ú
re gi on

B R A Z I L

PA R A G U AY
Iguaçu (Br)
and Iguazú (Ar)
National Parks

Mi si o n e s
Gre en
Corridor

San Rafael
Reserve

Matrix
Sinklike areas

A
0

25 50

R

100
km

G

E

N

T

IN

A

Attractive sinklike areas
Refugelike areas
Core areas
Dams and rivers

South of Misiones
province

International border

Fig. 3. Two-dimensional habitat model for predicting jaguar presence in the Upper Paraná Atlantic Forest and establishing priority management actions: protection/
mitigation actions in the attractive sinklike areas and restoration or land-use planning in the refuge like areas (see details in Figs. D1–D5).

Table 4
Distribution of the different suitability and management habitat categories for jaguars along the three countries that share the Upper Paraná Atlantic Forest. These categories
resulted from the two-dimensional combination of the main land-cover and human-persecution models. The percentages were calculated for each country (columns).
Habitat categories

Argentinak (m2)

Brazil (km2)

Paraguay (km2)

Total (km2)

Lakes/cities

509
(1.7%)

4598
(2.9%)

1469
(1.7%)

6576
(2.4%)

Avoided matrix

8495
(28.4%)

120,551
(75.4%)

37,047
(42.5%)

166,092
(60.0%)

Sinklike

2040
(6.8%)

19,413
(12.1%)

20,280
(23.3%)

41,733
(15.1%)

Refugelike

85
(0.3%)

5381
(3.4%)

3743
(4.3%)

9209
(3.3%)

Attractive sinklike

11,534
(38.6%)

5065
(3.2%)

15,965
(18.3%)

32,563
(11.8%)

Core areas

7253
(24.2%)

4813
(3.0%)

8604
(9.9%)

20,670
(7.5%)

Total

29,916
(100%)

159,819
(100%)

87,108
(100%)

276,843
(100%)

Author's personal copy

430

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

Upper Paraná Paranapanema
JCU

Main map legend
Core areas

Corridors
Rabinowitz & Zeller (2010)
Least cost areas (this analysis)
Dams and rivers
(B)

Jaguar Conservation Units
JCU revisited (this analysis)

Mbaracayú
JCU

JCU (Sanderson et el 2002)

BRAZIL

(C)

PA

R

U
AG

AY

Green
Corridor
JCU

(D)

AR

G

EN

TI

N

A

0

50

100

200
Km

(A)

0 20 40 km

(B)

0 10 20 km

0

(C)

10 20
km

(D)

Matrix

Corridors (Rabinowitz & Zeller 2010)

Dams and rivers

Sinklike areas

Least cost areas (this analysis)

International border

Attractive sinklike areas
Refugelike areas
Core areas
Fig. 4. Proposed redefinition of the Jaguar Conservation Units (JCUs revisited) and areas for evaluating corridors (least cost areas) for jaguar conservation in the Upper Paraná
Atlantic Forest (A). Dashed lines indicate the original JCUs described by Sanderson et al. (2002b) and in red (dark gray) the corridors proposed by Rabinowitz and Zeller (2010)
for connecting JCUs along the continent. The insets show examples of different areas of the JCUs and corridors, where the different habitat categories of the two-dimensional
model may help to prioritize management actions: in attractive sinklike areas to increase protection is a priority, in refugelike areas land-cover restoration is the most
important action; and in sinklike areas and matrix, both actions are necessary to create a corridor. (B) Possible areas for designing corridors among Morro do Diabo State Park,
Ivinhema State Park and Perobas Reserve in Brazil; (C) possible corridors between Mbaracajú JCU and the core areas of the Paraná river in Paraguay; and (D) attractive sinklike
areas separating core areas of the central and southern parts of the Green Corridor JCU in Argentina (see details in Figs. D1–D5).

Author's personal copy

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

habitat (see details in the priority actions recommended for different regions in Appendix D).

4. Discussion
4.1. Jaguar responses to habitat transformation
The Upper Paraná Atlantic Forest constituted a vast area covered by forest 200 years ago, where jaguars presumably had a continuous distribution (De Angelo et al., 2011a). Our results showed
that jaguars were seriously affected by forest loss (Fig. 3), but their
response was complex and affected not only by forest conditions
but also by other factors related to landscape transformation
and more direct human impacts (Table 3). A global model including both, land-cover conditions and human-persecution variables,
received the highest support of our data, demonstrating the importance of considering these diverse aspects for jaguar’ conservation
at a regional scale.
Not surprisingly, native forest cover and local forest connectivity are important for sustaining jaguars in the UPAF (Tables 3 and
C1). Similar results were found in other regions of jaguar distribution (Hatten et al., 2005; Ortega-Huerta and Medley, 1999; Somma,
2006). However, a significant advance in our understanding of jaguar ecology is the importance of past forest conditions for predicting current jaguar presence (Tables 3 and C1). Tilman et al. (1994)
found that habitat loss and fragmentation not only have immediate
effects on biodiversity but also produce a series of time-delayed
extinctions (i.e., the extinction debt). Clearly, such effects may be
more common in species with long generations, such as large carnivores, where a few individuals can survive in isolated fragments
for 10 years or more before the species becomes locally extinct.
However, landscape or forest history is rarely included in habitat
suitability models (Guisan and Zimmermann, 2000) and, to the
best of our knowledge, it has not been considered previously for
jaguars or other large carnivores’ habitat models.
Considering past forest conditions allowed us to understand
why jaguars were found in small and isolated fragments in eastern
Paraguay while no jaguars persist in the relatively larger forest
fragments of southern Misiones in Argentina (Fig. 3) (De Angelo
et al., 2011b). Most of the jaguars found in the small fragments
of eastern Paraguay may be survivors of recent deforestation but
probably do not constitute viable subpopulations. Southern Misiones was the most developed area of this province 30 years ago,
but many of these areas were abandoned and the forest has partially recovered (Izquierdo et al., 2008). However, the ecological
characteristics of these secondary forests are probably different
(Metzger et al., 2009) and may not sustain jaguars, or jaguars could
not recolonize these areas because high human pressures are still
persist (Figs. 3 and D3).
The physical characteristics of the landscape had a relatively
low importance in predicting jaguar presence (Tables 3 and C1),
probably due to the wide range of ecological conditions that jaguars can tolerate. In fact, jaguars had a continuous distribution
along this area in the past (Sanderson et al., 2002b). Human land
uses were more important than physical environment in predicting
jaguar presence (Table 3). The local (1-km) effect of intensive agriculture is possibly associated with the severe transformation of the
landscape (complete removal of forest) but with relatively low human presence reducing its impact in the neighborhood. Farms and
pastures are also related with a reduction of the forest-cover but
they had a larger impact in the surroundings (4-km radius;
Table C1); a fact that can be associated with higher human presence producing more disturbances in the neighboring forested
areas. Presence of cattle can also be important to explain the impact of these land uses due to the potential jaguar-cattle conflict

431

(Rosas-Rosas et al., 2010). However, our regional scale analysis
did not include detailed information about cattle abundance and
management, and therefore our conclusions regarding this issue
are limited.
Jaguars persist more frequently in inaccessible or protected
areas with historically low human density (Tables 3 and C2). The
variable accessibility (access_cost) was a useful way of representing direct human pressure, with better support than simpler measurements such as distance to roads or towns (Table C2). The
access of humans to wild areas is associated with many different
direct impacts, like poaching (Kerley et al., 2002; Nielsen et al.,
2004) and forest exploitation (Chomitz and Gray, 1996). Additionally, access ways can become an important threat to wild populations through road kills (Kerley et al., 2002; Kramer Schadt et al.,
2004). Protection reduces poaching pressure directly, but it can
also be important in reducing other direct human impacts such
as forest exploitation and transit of humans (Bruner et al., 2001).
Large carnivores often use trails and roads for their movement in
wild areas (Kerley et al., 2002; Noss et al., 1996) and this behavior
was also described for jaguars in the UPAF (Cullen, 2006; Paviolo,
2010). This may explain why the effect of access is lower inside
protected areas (Tables 3, C2 and C3), where jaguars may use the
access ways more often than in unprotected areas.
Our one-dimensional habitat models are useful for assessing the
distribution of the remaining potential habitat for jaguars and predicting unsurveyed areas where jaguars could be present (Fig. C1).
On the other hand, our two-dimensional approach allowed for a
more subtle assessment of multidimensional habitat suitability,
to detect areas where different management relevant key factors
were conflicting and to prioritize management actions (Figs. 3
and 4, D1–D5) (Naves et al., 2003). However, our models have
some limitations. First, because they were explicitly constructed
for a regional analysis, we could not include some important issues
of local jaguar habitat selection (e.g. influence of different types of
forest, a wider range in protection categories, or the relative impact
of different cattle management; Azevedo and Murray, 2007; Conde
et al., 2010; Cullen, 2006). Second, an analysis at finer scales or
with other objectives would require different hypotheses for the
two main dimensions. For example, one could include additional
human land-uses (e.g. pastures with different management of cattle or diverse human activities in the small farms) in the jaguar
persecution group to consider the potential jaguar-ranchers conflict. Finally, although our models showed a good performance
and we could validate the sinklike areas, it is important to recognize that we used only presence records for model construction, instead of reproduction (Naves et al., 2003), mortality (Nielsen et al.,
2006), or prey data (Kanagaraj et al., 2011). Collecting such data
would be important for future habitat modeling of this species.

4.2. Recommendations for jaguar conservation
The population of jaguars of the UPAF now constitutes a highly
spatially structured population (Elmhagen and Angerbjörn, 2001),
divided into several core areas, many of them completely isolated
by matrix or surrounded by attractive sinklike areas (Figs. 3 and 4;
Table 4). Less than 10% of the study area constitutes suitable habitat for jaguars and the internal structure of these areas indicates
that many of them are probably under high pressure mainly due
to human persecution (Figs. C1 and D5).
The Brazilian core areas are completely isolated by avoided matrix (Figs. 3 and 4); consequently, high efforts of active management are needed for restoring the connectivity among them and
reducing the harmful effect of their isolation (Haag et al., 2010).
Although some restoration initiatives exist along the Parana river,
higher efforts are needed to ensure structural connectivity among

Author's personal copy

432

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433

the core areas, as well as survival of migrant individuals (Cullen,
2006) (Fig. 4).
In Paraguay only a few core areas are effectively protected and
many of them are recently fragmented areas with an extinction
debt (Figs. 3 and D5) (De Angelo et al., 2011b). Although the core
areas of Paraguay showed higher connectivity through marginal
habitat than the Brazilian areas, many of these marginal areas
are sinklike or attractive sinklike areas with high potential of being
ecological traps for jaguars (Figs. 3 and 4 and D3). Even though several areas still hold jaguars and there is a potential for connectivity
among them, jaguars in Paraguay are threatened by land-cover
transformation and direct human persecution, and only few areas
are known to have potential source populations (Fig. 4).
In the Green Corridor shared by Argentina and Brazil, most of
the landscape constitutes attractive sinklike areas where jaguars
may occur because the landscape offers good structural conditions
(i.e., forest), but with a high risk of being killed by poachers or
ranchers (Figs. 3 and 4 and D3). Extremely high mortality rates
have been detected for jaguars in the surroundings of the Iguazú
and Iguaçu National Parks in Argentina and Brazil respectively
(Crawshaw, 2002; Crawshaw et al., 2004; Paviolo et al., 2008). This
high mortality, extended along the large proportion of the Green
Corridor with attractive sinklike areas, can explain the recent population crash suffered by jaguars in this region (Paviolo et al., 2008)
and the absence of jaguars in the forest fragments of the southern
part of Misiones Province (De Angelo et al., 2011b) (Figs. 3 and 4).
Reducing direct jaguar persecution and poaching of its prey are the
most urgent actions needed to regionally preserve this species in
the Green Corridor. Our model helps to identify areas where these
actions will have the highest impact in reducing jaguar mortality
and maintaining connectivity (e.g. see Fig. 4C and areas categorized
as AS3-M in Fig. D3). For this reason, this model was used to construct the Conservation Landscape in the Action Plan for Jaguar
Conservation in the Green Corridor (Schiaffino et al., 2011), approved by the National Parks Administration of Argentina in 2012.

5. Conclusions
Our modeling approach allowed us to understand that habitat
destruction for jaguars implicates not only forest loss but also
many different anthropogenic interventions on the landscape,
including those that occurred in the past. Additionally, this approach was useful for validating and improving conservation strategies for this species and the Atlantic Forest landscape, and serves
as input for an adaptive management conservation plan for both
(Sanderson et al., 2002c). To include management criteria for
selecting different dimensions for habitat modeling represents a
novel approach for modeling the distribution of endangered species, and it has the main advantage of maintaining a direct link
to landscape management and species conservation options.
Using the density estimates of jaguars along the UPAF, we calculated a mean density of 1 ind/100 km2 (Cullen, 2006; Paviolo,
2010; Paviolo et al., 2008). Extrapolating this value to the total surface covered by core areas in our study area (around 20,000 km2)
the total population of jaguars is about 200 adult individuals in
the whole eco-region (Di Bitetti et al., 2006). These individuals
are distributed along different patches, many of them isolated from
the others (Figs. 3 and 4). This reinforces the need of diminishing
direct threats and habitat loss in each of these patches, increasing
the size of the core areas through reducing human persecution
(Fig. 3), and maintaining or restoring connectivity among them
via land-use planning and land-cover restoration (Fig. 4). A
coordinated effort among the three countries would be essential
to preserve the jaguars of the UPAF (Fig. 4), by maintaining a metapopulation dynamic among the jaguar core areas not only of the

proposed JCU but including all the core areas of the eco-region
(Di Bitetti et al., 2006).
Acknowledgments
We thank Proyecto Yaguareté volunteers, and Paraguayan and
Brazilian researchers for their help in data collection, and the support provided by the Ministry of Ecology and Natural Resources of
Misiones and the National Parks Administration of Argentina. We
are grateful to C. Boiero and E. Hasson for their revision of the manuscript. Two anonymous reviewers provided useful criticism
and suggestions. Financial support was provided by CONICET, Fundación Vida Silvestre Argentina, World Wildlife Fund (WWF) – Education for Nature Program, WWF-International, WWF-Switzerland,
Conservation Leadership Programme, Society for Conservation GIS,
and Lincoln Park Zoo.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.biocon.2012.12.
021. These data include Google maps of the most important areas
described in this article.
References
Altrichter, M., Boaglio, G., Perovic, P., 2006. The decline of jaguars Panthera onca in
the Argentine Chaco. Oryx 40, 302–309.
Azevedo, F.C.C., Murray, D.L., 2007. Evaluation of potential factors predisposing
livestock to predation by jaguars. J. Wildlife Manage. 71, 2379–2386.
Brooks, T.M., Pimm, S.L., Oyugi, J.O., 1999. Time lag between deforestation and bird
extinction in tropical forest fragments. Conserv. Biol. 13, 1140–1150.
Bruner, A.G., Gullison, R.E., Rice, R.E., da Fonseca, G.A.B., 2001. Effectiveness of parks
in protecting tropical biodiversity. Science 291, 125–128.
Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference: a
Practical Information – Theoretic Approach. Springer-Verlag, New York, USA.
Cardillo, M., Purvis, A., Bielby, J., Mace, G.M., Sechrest, W., Gittleman, J.L., 2004.
Human population density and extinction risk in the world’s carnivores. PLoS
Biol. 2, 909–914.
Chomitz, K.M., Gray, D.A., 1996. Roads, land use, and deforestation: a spatial model
applied to Belize. World Bank Econ. Rev. 10, 487–512.
Conde, D.A., Colchero, F., Zarza, H., Christensen Jr., N.L., Sexton, J.O., Manterola, C.,
Chávez, C., Rivera, A., Azuara, D., Ceballos, G., 2010. Sex matters: modeling male
and female habitat differences for jaguar conservation. Biol. Conserv. 193,
1980–1988.
Crawshaw Jr., P.G., 2002. Mortalidad inducida por humanos y conservación de
jaguares: el Pantanal y el Parque Nacional Iguaçu en Brasil. In: Medellín, R.A.,
Equihua, C., Chetkiewicz, C.L., Crawshaw, P.G.J., Rabinowitz, A., Redford, K.H.,
Robinson, J.G., Sanderson, E.W., Taber, A.B. (Eds.), El jaguar en el nuevo milenio.
Editorial Ediciones Científicas Universitarias, Mexico, DF, pp. 451–464.
Crawshaw Jr., P.G., Quigley, H.B., 1991. Jaguar spacing, activity and habitat use in a
seasonally flooded environment in Brazil. J. Zool. 223, 357–370.
Crawshaw Jr., P.G., Mahler, J.K., Indrusiak, C., Cavalcanti, S.M.C., Leite, M.R.P., Silvius,
K.M., 2004. Ecology and conservation of jaguar (Panthera onca) in Iguaçu
National Park, Brazil. In: Silvius, K.M., Bodmer, R.E., Fragoso, J.M.V. (Eds.), People
in Nature: Wildlife Conservation in South and Central America. Columbia
University Press, New York, USA, pp. 271–285.
Cullen Jr., L., 2006. Jaguars as landscape detectives for the conservation of Atlantic
Forests in Brazil. In: Canterbury, University of Kent, Canterbury, USA, p. 178.
Cullen Jr., L., Abreu, K.C., Sana, D., Nava, A.F.D., 2005. Jaguars as landscape detectives
for the Upper Paraná River corridor, Brazil. Natureza e Conservação 3, 43–58.
De Angelo, C., 2009. El paisaje del Bosque Atlántico del Alto Paraná y sus efectos
sobre la distribución y estructura poblacional del jaguar (Panthera onca) y el
puma (Puma concolor), Universidad de Buenos Aires, Buenos Aires, Argentina, p.
252.
De Angelo, C., Paviolo, A., Di Bitetti, M., 2010. Traditional versus multivariate
methods for identifying jaguar, puma, and large canid tracks. J. Wildlife Manage.
74, 1141–1153.
De Angelo, C., Paviolo, A., Di Bitetti, M., 2011a. Differential impact of landscape
transformation on pumas (Puma concolor) and jaguars (Panthera onca) in the
Upper Paraná Atlantic Forest. Divers. Distrib. 17, 422–436.
De Angelo, C., Paviolo, A., Rode, D., Cullen Jr., L., Sana, D., Abreu, K.C., Xavier da Silva,
M., Bertrand, A.-S., Haag, T., Lima, F., Ricieri Rinaldi, A., Ramírez, F., Fernández, S.,
Velázquez, M., Corio, C., Hasson, E., Di Bitetti, M.S., 2011b. Participatory
networks for large-scale monitoring of large carnivores: pumas and jaguars of
the Upper Paraná Atlantic Forest. Oryx 45, 534–545.
Delibes, M., Gaona, P., Ferreras, P., 2001. Effects of an attractive sink leading into
maladaptive habitat selection. Am. Nat. 158, 277–285.

Author's personal copy

C. De Angelo et al. / Biological Conservation 159 (2013) 422–433
Di Bitetti, M.S., Placci, G., Dietz, L.A., 2003. A biodiversity vision for the Upper Paraná
Atlantic Forest Eco-region: designing a biodiversity conservation landscape and
setting priorities for conservation action. In: World Wildlife Fund, Washington,
DC, USA.
Di Bitetti, M., De Angelo, C., Paviolo, A., Schiaffino, K., Perovic, P., 2006. Monumento
Natural Nacional en peligro: el desafío de conservar al yaguareté en la
Argentina, In La Situación Ambiental Argentina 2005. In: Brown, A., Martinez
Ortiz, U., Acerbi, M., Corcuera, J. (Eds.), Fundación Vida Silvestre Argentina,
Buenos Aires, Argentina, pp. 420–431.
Di Bitetti, M.S., Paviolo, A., Ferrari, C., De Angelo, C., Di Blanco, Y., 2008. Differential
responses to hunting in two sympatric species of brocket deer (Mazama
americana and Mazama nana). Biotropica 40, 636–645.
Elmhagen, B., Angerbjörn, A., 2001. The applicability of metapopulation theory to
large mammals. Oikos 94, 89–100.
Engler, R., Guisan, A., Rechsteiner, L., 2004. An improved approach for predicting the
distribution of rare and endangered species from occurrence and pseudoabsence data. J. Appl. Ecol. 41, 263–274.
Fernández, N., Delibes, M., Palomares, F., Mladenoff, D.J., 2003. Identifying breeding
habitat for the iberian lynx: Inferences from a fine-scale spatial analysis. Ecol.
Appl. 13, 1310–1324.
Fernández, N., Delibes, M., Palomares, F., 2006. Landscape evaluation in
conservation: molecular sampling and habitat modeling for the iberian lynx.
Ecol. Appl. 16, 1037–1049.
Guisan, A., Thuiller, W., 2005. Predicting species distribution: offering more than
simple habitat models. Ecol. Lett. 8, 993–1009.
Guisan, A., Zimmermann, N.E., 2000. Predictive habitat distribution models in
ecology. Ecol. Model. 135, 147–186.
Haag, T., Santos, A.S., De Angelo, C., Srbek-Araujo, A.C., Sana, D., Morato, R.G.,
Salzano, F.M., Eizirik, E., 2009. Development and testing of an optimized method
for DNA-based identification of jaguar (Panthera onca) and puma (Puma
concolor) faecal samples for use in ecological and genetic studies. Genetica
136, 505–512.
Haag, T., Santos, A.S., Sana, D., Morato, R.G., Cullen Jr., L., Crawshaw Jr., P.G., De
Angelo, C., Di Bitetti, M.S., Salzano, F.M., Eizirik, E., 2010. The effect of habitat
fragmentation on the genetic structure of a top predator: loss of diversity and
high differentiation among remnant populations of Atlantic Forest jaguars
(Panthera onca). Mol. Ecol. 19, 4906–4921.
Hatten, J.R., Averill-Murray, A., van Pelt, W.E., 2005. A spatial model of potential
jaguar habitat in Arizona. J. Wildlife Manage. 69, 1024–1033.
Hirzel, A.H., Lay, G.L., Helfer, V., Randin, C., Guisan, A., 2006. Evaluating the ability of
habitat suitability models to predict species presences. Ecol. Model. 199, 142–
152.
Huang, C., Kim, S., Altstatt, A., Townshend, J.R.G., Davis, P., Song, K., Tucker, C.J.,
Rodas, O., Yanosky, A., Clay, R., Musinsky, J., 2007. Rapid loss of Paraguay’s
Atlantic forest between 1970s and 2000 and the status of protected areas.
Remote Sens. Environ. 106, 460–466.
Izquierdo, A., De Angelo, C., Aide, M., 2008. Thirty years of human demography and
land-use change in the Atlantic Forest of Misiones, Argentina: a test of the forest
transition model. Ecol. Soc. 13, 3.
Jacobsen, T.R., 2003. Populating the environment: human growth, density, and
migration in the Atlantic Forest. In: Galindo-Leal, C., de Gusmão Câmara, I.
(Eds.), The Atlantic Forest of South America: Biodiversity Status, Threats, and
Outlook. Island Press, Washington, pp. 426–435.
Johnson, J.B., Ommland, K.S., 2004. Model selection in ecology and evolution. Trends
Ecol. Evol. 19, 101–108.
Kanagaraj, R., Wiegand, T., Kramer-Schadt, S., Anwar, M., Goyal, S.P., 2011. Assessing
habitat suitability for tiger in the fragmented Terai Arc Landscape of India and
Nepal. Ecography 34, 970–981.
Kerley, L.L., Goodrich, J.M., Miquelle, D.G., Smirnov, E.N., Quigley, H.B., Hornocker,
M.G., 2002. Effects of roads and human disturbance on Amur tigers. Conserv.
Biol. 16, 97–108.
Kramer Schadt, S., Revilla, E., Wiegand, T., Breitenmoser, U., 2004. Fragmented
landscapes, road mortality and patch connectivity: modelling influences on the
dispersal of Eurasian lynx. J. Appl. Ecol. 41, 711–723.
Liu, C., Berry, P.M., Dawson, T.P., Pearson, R.G., 2005. Selecting thresholds of
occurrence in the prediction of species distributions. Ecography 28, 385–393.
McBride, R.T., 2009. Project Jaguar – Final Report 2002–2009, Faro Moro Eco
Research, Fundación Moises Bertoni, Secretaría del Ambiente del Paraguay
(SEAM), Asunción, Paraguay, p. 55.
McCullagh, P., Nelder, J.A., 1989. Generalized Linear Models, second ed. Chapman &
Hall, London, UK.
Metzger, J.P., Martensen, A.C., Dixo, M., Bernacci, L.C., Ribeiro, M.C., Godoy Teixeira,
A.M., Pardini, R., 2009. Time-lag in biological responses to landscape changes in
a highly dynamic Atlantic forest region. Biol. Conserv. 142, 1166–1177.
Miller, B., Rabinowitz, A., 2002. ¿Por qué conservar al jaguar? In: Medellín, R.A.,
Equihua, C., Chetkiewicz, C.L., Crawshaw, P.G.J., Rabinowitz, A., Redford, K.H.,
Robinson, J.G., Sanderson, E.W., Taber, A.B. (Eds.), El jaguar en el nuevo milenio.
Editorial Ediciones Científicas Universitarias, Mexico, DF, pp. 303–315.

433

Mittermeier, R.A., Robles Gil, P., Hoffman, M., Pilgrim, J., Brooks, T., Mittermeier,
C.G., Lamoreux, J., da Fonseca, G.A.B., 2005. Hotspots revisited: Earth’s
biologically richest and most endangered terrestrial ecoregions. In: first ed.
Conservation International, Washington, USA.
Muntifering, J.R., Dickman, A.J., Perlow, L.M., Hruska, T., Ryan, P.G., Marker, L.L., Jeo,
R.M., 2006. Managing the matrix for large carnivores: a novel approach and
perspective from cheetah (Acinonyx jubatus) habitat suitability modelling.
Anim. Conserv. 9, 103–112.
Naves, J., Revilla, E., Delibes, M., Wiegand, T., 2003. Endangered species constrained
by natural and human factors: the case of brown bears in Northern Spain.
Conserv. Biol. 17, 1276–1289.
Nielsen, S.E., Herrero, S., Boyce, M.S., Benn, B., Mace, R.D., Gibeau, M.L., Jevons, S.,
2004. Modelling the spatial distribution of human-caused grizzly bear mortalities
in the Central Rockies Ecosystem of Canada. Biol. Conserv. 120, 101–113.
Nielsen, S.E., Stenhouse, G.B., Boyce, M.S., 2006. A habitat-based framework for
grizzly bear conservation in Alberta. Biol. Conserv. 130, 217–229.
Noss, R.F., Quigley, H.B., Hornocker, M.G., Merrill, T., Paquet, P.C., 1996. Conservation
biology and carnivore conservation in the Rocky Mountains. Conserv. Biol. 10,
949–963.
Ortega-Huerta, M.A., Medley, K.E., 1999. Landscape analysis of jaguar (Panthera
onca) habitat using sighting records in the Sierra de Tamaulipas, Mexico.
Environ. Conserv. 26, 257–269.
Paviolo, A., 2010. Densidad de yaguareté (Panthera onca) en la Selva Paranaense: su
relación con la abundancia de presas, presión de caza y coexistencia con el
puma (Puma concolor), Universidad Nacional de Córdoba, Córdoba, Argentina, p.
201.
Paviolo, A., De Angelo, C., Di Blanco, Y., Di Bitetti, M., 2008. Jaguar population decline
in the Upper Paraná Atlantic Forest of Argentina and Brazil. Oryx 42, 554–561.
Paviolo, A., De Angelo, C., Di Blanco, Y., Agostini, I., Pizzio, E., Melzew, R., Ferrari, C.,
Palacio, L., Di Bitetti, M., 2009. Efecto de la caza y el nivel de protección en la
abundancia de los grandes mamíferos del Bosque Atlántico de Misiones. In:
Carpinetti, B., Garciarena, M. (Eds.), Contribuciones para la conservación y
manejo en el Parque Nacional Iguazú, Administración de Parques Nacionales,
Buenos Aires, Argentina, pp. 237–254.
Rabinowitz, A., Zeller, K.A., 2010. A range-wide model of landscape connectivity and
conservation for the jaguar, Panthera onca. Biol. Conserv. 143, 939–945.
Revilla, E., Wiegand, T., Palomares, F., Ferreras, P., Delibes, M., 2004. Effects of matrix
heterogeneity on animal dispersal: from individual behavior to
metapopulation-level parameters. Am. Nat. 164, 130–153.
Ribeiro, M.C., Metzger, J.P., Martensen, A.C., Ponzoni, F.J., Hirota, M.M., 2009. The
Brazilian Atlantic Forest: how much is left, and how is the remaining forest
distributed? Implications for conservation. Biol. Conserv. 142, 1141–1153.
Rodríguez-Soto, C., Monroy-Vilchis, O., Maiorano, L., Boitani, L., Faller, J.C., Briones,
M.Á., Núñez, R., Rosas-Rosas, O., Ceballos, G., Falcucci, A., 2011. Predicting
potential distribution of the jaguar (Panthera onca) in Mexico: identification of
priority areas for conservation. Divers. Distrib. 17, 350–361.
Rosas-Rosas, O.C., Bender, L.C., Valdez, R., 2010. Habitat correlates of jaguar kill sites of
cattle in northeastern Sonora, Mexico. Human – Wildlife Conflicts 4, 113–121.
Sanderson, E.W., Jaiteh, M., Levy, M.A., Redford, K.H., Wannebo, A.V., Woolmer, G.,
2002a. The human footprint and the last of the wild. Bioscience 52, 891–904.
Sanderson, E.W., Redford, K.H., Chetkiewicz, C.L.B., Medellin, R.A., Rabinowitz, A.R.,
Robinson, J.G., Taber, A.B., 2002b. Planning to save a species: the jaguar as a
model. Conserv. Biol. 16, 58–72.
Sanderson, E.W., Redford, K.H., Vedder, A., Coppolillo, P.B., Ward, S.E., 2002c. A
conceptual model for conservation planning based on landscape species
requirements. Landscape Urban Plan. 58, 41–56.
Schadt, S., Revilla, E., Wiegand, T., Knauer, F., Kaczensky, P., Breitenmoser, U., Bufka,
L., Cerveny, J., Koubek, P., Huber, T., Stanisa, C., Trepl, L., 2002. Assessing the
suitability of central European landscapes for the reintroduction of Eurasian
lynx. J. Appl. Ecol. 39, 189–203.
Schiaffino, K., De Angelo, C., Di Bitetti, M., Paviolo, A., Jaramillo, M., Rinas, M.,
Bertrand, A.-S., Gil y Carbó, G., Cichero, P. (Eds.), 2011. Plan de Acción para la
Conservación de la Población de Yaguareté (Panthera onca) del Corredor Verde
de Misiones. Subcomisión Selva Paranaense. Primer Borrador. Ministerio de
Ecología y Recursos Renovables de Misiones- Administración de Parques
Nacionales- Instituto de Biología Subtropical- Fundación Vida Silvestre
Argentina, Puerto Iguazú, Argentina.
Somma, D.J., 2006. Interrelated modeling of land use and habitat for the design of an
ecological corridor: a case study in the Yungas, Argentina, Wageningen
University (The Netherlands) and Universidad de Buenos Aires (Argentina),
Wageningen, The Netherlands, p. 179.
Tilman, D., May, R.M., Lehman, C.L., Nowak, M.A., 1994. Habitat destruction and the
extinction debt. Nature 371, 65–66.
Wikramanayake, E., McKnight, M., Dinerstein, E., Joshi, A., Gurung, B., Smith, D.,
2004. Designing a conservation landscape for tigers in human-dominated
environments. Conserv. Biol. 18, 839–844.
Woodroffe, R., Ginsberg, J.R., 1998. Edge effects and the extinction of populations
inside protected areas. Science 280, 2126–2128.

1

Appendix. Supplementary data

2

Supplementary online data of De Angelo et al.: Understanding species persistence

3

for defining conservation actions: a management landscape for jaguars in the Atlantic

4

Forest

5
6
7
8
9

Appendix A. Rules for pseudo-absences generation.
Fig. A1. Distribution of pseudo-absences and rules for their generation.
Appendix B. Variables
B1. Variables used for describing the hypotheses and generating the models

10

Table B1. Description of variables.

11

Table B2. Tests for the selected records.

12

Appendix C. Models

13

C.1. Variable reduction procedure.

14

C.2. Models evaluation and habitat suitability categories.

15

Tables C1 and C2. Model selection for particular hypotheses.

16

Table C3. Description of the final models.

17

Fig. C1. One-dimensional habitat suitability maps of the final models.

18

Table C4. Evaluation of the final models.

19

Appendix D. Categories and sub-categories of habitat in the two-dimensional model.

20

Fig. D1. Presences and pseudo-absences distribution along the two dimensions of

21

habitat.

22

Fig. D2. Presences and pseudo-absences distribution along the sub-categories of

23

habitat inside the marginal habitats.

24

Fig. D3. Sub-categories of habitat inside the marginal habitats.

25

Fig. D4. Presences and pseudo-absences distribution along the sub-categories of

26

habitats inside the core areas.

27

Fig. D5. Sub-categories of habitats inside the core areas.

28

Appendix E. Distribution of killed jaguars along the categories of habitat.

29

Fig. E1. Observed and expected proportion of killed jaguars in the categories of

30

habitat

31
32

Appendix F. Validation of the Biodiversity Vision of the UPAF.
Table F1. Validation of the Biodiversity Vision of the UPAF.

33
1

34
35

Appendix A. Details on the rules for pseudo-absences generation
To obtain a binomial response variable we generated pseudo-absences randomly

36

within the study area, but following several rules (Fig. A1). First, the probability of

37

occurrence of pseudo-absences was weighted by a habitat suitability index based on the

38

presence-only habitat suitability map developed for jaguars in the UPAF by De Angelo et al.

39

(2011a). This resulted in a higher proportion of pseudo-absences located in unsuitable areas

40

compared with suitable areas (Chefaoui and Lobo 2008; Engler et al. 2004; Hengl et al. 2009;

41

Titeux 2006). Second, pseudo-absences were generated only in areas identified in previous

42

analysis as unsuitable or marginal habitats for jaguars (De Angelo et al. 2011a). This rule

43

avoids location of pseudo-absences in areas expected to be suitable for jaguars (Chefaoui and

44

Lobo 2008). Third, we generated pseudo-absences only inside the area surveyed by the

45

participatory monitoring where presence data was collected (De Angelo et al. 2011b),

46

ensuring that pseudo-absences occurred only in areas that were surveyed (Mateo et al. 2010;

47

Phillips et al. 2009). This rule also prevents that pseudo-absences occur by chance only in

48

remote areas that may show, due to their large geographical distance, habitat conditions that

49

differ from what is jaguar habitat (VanDerWal et al. 2009). Finally, we generated the same

50

number of pseudo-absences as presences (n=106; Engler et al. 2004; Kanagaraj et al. 2011;

51

Liu et al. 2005) following the same rule used for stratifying presence records (no more than

52

one pseudo-absence per each 12 × 12-km grid cell; Kanagaraj et al. 2011). Additionally, we

53

generated 10 further sets of pseudo-absences for model validation. For pseudo-absence

54

generation we used the Sampling Tools of Hawth's Analysis Tools (Beyer 2004).

2

55
56

Fig. A1. Distribution of presence and pseudo-absences used for data analysis. This figure illustrates

57

the rules used for pseudo-absence random generation.

58

3

59

Appendix B. Variables

60

B.1. Variables construction

61

To describe the land-cover and physical characteristics of the landscape, and the

62

human persecution of jaguar and their prey, for representing the different hypotheses, we

63

constructed a total of 9 + 4 × 10 = 49 variables with a spatial resolution of 330 m × 330 m as

64

it is described in the main text and in Table B1. The local connectivity for radii of 1 km

65

(connect_r1) was discarded because it redundancy with forest_r1. The maps used as base

66

map for variables construction were obtained from the UPAF-GIS database compiled by Di

67

Bitetti et al. (2003) and De Angelo (2009). The land-use map for our analysis was developed

68

by De Angelo (2009) using a mosaic of Landsat-5 TM satellite images from 2004 and a

69

maximum likelihood supervised classification into seven land-uses categories (water, native

70

forest and marshlands, pine plantation, intensive agriculture, small farms with mixed land

71

uses, pastures, and urban areas). This analysis was based on the variables used by De Angelo

72

et al. (2011a). All these analyses were developed with ENVI software Version 4.2 (Research

73

Systems, Inc. 2005, USA), Spatial Analyst for ArcMap 9 (ESRI Inc. 2004) and Hawth's

74

Analysis Tools (Beyer 2004).

75

4

76

Table B1. Description of independent variables used for describing land-cover and physical

77

characteristics of the landscape, and the human persecution of jaguars and their prey in the Upper

78

Paraná Atlantic Forest (see details in De Angelo 2009; and De Angelo et al. 2011a). Nine variables

79

describe the average conditions within each cell while the other ten represent neighbourhood variables

80

(_r) that were calculated at four different neighbourhood scales of radius r = 1, 4, 7 and 10 km.
Name
access_cost
connect_r

Elevation
farms_r
forest_r

Description
Accessibility cost measured as the hours needed to access the focal cell from the
nearest town or city (De Angelo et al. 2011a; Farrow and Nelson 2001).
Frequency of cells occupied by native forest in a ring of radius r and 1-km wide (3
cells) around the focal cell. This represents an index of local connectivity of forest
around the focal cell (Naves et al. 2003; Schadt et al. 2002; Wiegand and Moloney
2004).
Elevation above sea level of the focal cell (from http://seamless.usgs.gov).
Frequency of cells occupied by small farms in a circle of radius r around the focal
cell.
Frequency of cells occupied by native forest in a circle of radius r around the focal
cell.

forest73_r

Frequency of cells occupied by native forest in 1973 in a circle of radius r around
the focal cell.
int_agr_r
Frequency of cells occupied by intensive agriculture in a circle of radius r around
the focal cell.
pastures_r
Frequency of cells occupied by extensive pastures in a circle radius r around the
focal cell.
plantat_r
Frequency of cells occupied by pine plantations in a circle of radius r around the
focal cell.
population_2000 Rural population density obtained from the most recent national census (Brazil
2000, Paraguay 2002 and Argentina 2001). This map was constructed using local
census units (municipalities in Brazil, districts in Paraguay and Departments in
Argentina), and translated to a 10-km grid of points that was used for interpolating
rural population values along the entire area with a smoothed effect (Carroll and
Miquelle 2006; De Angelo 2009)
population_hist
Mean historical rural population density. This map was constructed by the average
of rural population density maps from 1970, 1980, 1990 and 2000 built by the same
method as population_2000 (De Angelo 2009).
protect_cat
Categorical classification of relative protection levels: 0 = unprotected; 1 =
intermediate protection (e.g. private and biosphere reserves); 2 = high protection
(e.g. national and provincial parks).
rivers_d
Straight line distance to the closest river.
rivers_r
roads_d

Frequency of cells occupied by rivers in a circle of radius r around the focal cell.
Straight line distance to the closest road (including paved and dirt roads).

roads_r
slope

Frequency of cells occupied by roads (including paved and dirt roads) in a circle of
radius r around the focal cell.
Terrain slope expressed in percentage.

towns_d

Straight line distance to the closest town or city.

towns_r

Frequency of cells occupied by towns or cities in a circle of radius r around the
focal cell.

5

81

Table B2. Comparisons between the 106 presence records selected for the analysis and each of the other 10 random subsets of 106 records resampled from

82

the total 974 presence registers. For each subset, the statistic of the Mann-Whitney U test and the corresponding p value is reported for all the independent

83

variables used in the analysis. In the case of the categories of the protected areas, a Pearson’s Chi square test was used. NA indicates that this variable was

84

not used in the analysis. None of the subsets showed significant differences for any of the variables.
Subset 1

Subset 2
p

U

Subset 3
p

U

Subset 4
p

U

Subset 5
p

U

Subset 6
p

U

Subset 7
p

U

Subset 8
p

U

Subset 9
p

U

Subset 10

Variable

U

p

U

acces_cost

5593

0.955

5554

0.887

5443

0.695

5802

0.682

5621

0.996

5492

0.778

5390

0.611

5667

0.914

5529

0.842

5557

0.891

p

connect_r01

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

NA

connect_r04

5725

0.812

5700

0.856

5665

0.918

5908

0.517

5901

0.527

5535

0.853

5717

0.826

5757

0.756

5647

0.949

5705

0.847

connect_r07

5678

0.895

5605

0.977

5557

0.891

5734

0.797

5644

0.955

5629

0.982

5677

0.897

5666

0.916

5626

0.987

5607

0.981

connect_r10

5585

0.941

5563

0.903

5663

0.921

5698

0.859

5571

0.916

5673

0.904

5608

0.982

5678

0.895

5593

0.955

5613

0.991

elevation

5512

0.813

5604

0.976

5507

0.805

5614

0.994

5602

0.971

5457

0.718

5534

0.852

5585

0.942

5540

0.862

5576

0.926

farms_r01

5686

0.872

5846

0.584

5823

0.623

5680

0.884

5743

0.767

5656

0.930

5897

0.503

5480

0.743

5785

0.691

5750

0.753

farms_r04

5581

0.934

5606

0.979

5658

0.929

5404

0.632

5429

0.672

5722

0.816

5635

0.970

5413

0.646

5665

0.917

5575

0.923

farms_r07

5560

0.898

5608

0.982

5670

0.908

5405

0.633

5577

0.927

5686

0.881

5617

0.998

5538

0.859

5700

0.856

5642

0.959

farms_r10

5563

0.903

5624

0.990

5639

0.964

5399

0.624

5555

0.888

5683

0.885

5605

0.977

5562

0.901

5635

0.971

5594

0.957

forest_r01

5466

0.728

5517

0.818

5595

0.958

5644

0.954

5634

0.972

5504

0.795

5349

0.540

5673

0.901

5454

0.709

5542

0.863

forest_r04

5535

0.853

5643

0.956

5575

0.923

5783

0.713

5708

0.841

5471

0.743

5616

0.996

5716

0.828

5591

0.952

5631

0.979

forest_r07

5664

0.919

5637

0.968

5589

0.949

5822

0.649

5732

0.800

5538

0.858

5646

0.952

5733

0.799

5644

0.954

5660

0.927

forest_r10

5645

0.954

5607

0.981

5583

0.938

5770

0.734

5667

0.914

5551

0.882

5640

0.962

5698

0.859

5601

0.971

5639

0.964

forest73_r01

5684

0.883

5545

0.871

5371

0.579

5634

0.972

5629

0.982

5405

0.633

5496

0.784

5658

0.907

5585

0.941

5615

0.996

forest73_r04

5602

0.971

5604

0.976

5527

0.839

5796

0.690

5633

0.975

5552

0.882

5707

0.843

5717

0.826

5717

0.826

5590

0.951

forest73_07

5686

0.881

5656

0.933

5581

0.935

5838

0.623

5650

0.945

5618

1.000

5676

0.898

5702

0.853

5735

0.794

5619

1.000

forest73_10

5692

0.870

5655

0.935

5577

0.927

5779

0.720

5640

0.962

5569

0.913

5632

0.976

5631

0.979

5670

0.908

5605

0.977

int_agr_r01

5573

0.901

5334

0.438

5464

0.669

5495

0.732

5586

0.928

5643

0.946

5600

0.961

5488

0.718

5451

0.643

5394

0.537

int_agr_r04

5625

0.989

5386

0.601

5611

0.987

5543

0.865

5490

0.774

5734

0.795

5586

0.943

5536

0.854

5501

0.792

5474

0.746

int_agr_r07

5466

0.734

5361

0.565

5599

0.967

5503

0.798

5469

0.738

5675

0.900

5526

0.837

5584

0.940

5546

0.873

5505

0.801

int_agr_r10

5559

0.896

5459

0.722

5658

0.930

5523

0.832

5622

0.994

5607

0.981

5519

0.825

5561

0.898

5597

0.962

5591

0.952

pastures_r01

5646

0.930

5749

0.673

5760

0.647

5549

0.829

5707

0.776

5617

0.997

5773

0.617

5680

0.845

5728

0.725

5653

0.912

pastures_r04

5597

0.962

5716

0.822

5799

0.677

5513

0.810

5562

0.897

5759

0.746

5628

0.983

5571

0.915

5750

0.761

5725

0.806

pastures_r07

5561

0.898

5649

0.945

5741

0.783

5371

0.579

5546

0.872

5657

0.931

5496

0.784

5447

0.701

5637

0.967

5570

0.914

6

85
86

Table B2. It continues from the previous page.
Subset 1
Variable

U

Subset 2
p

U

Subset 3
p

U

Subset 4
p

U

Subset 5
p

U

Subset 6
p

U

Subset 7
p

U

Subset 8
p

U

Subset 9
p

U

Subset 10
p

U

p

pastures_r10

5613

0.992

5673

0.903

5712

0.835

5502

0.796

5592

0.954

5684

0.883

5562

0.901

5575

0.923

5634

0.973

5640

0.962

plantat_r01

5673

0.771

5561

0.776

5561

0.776

5460

0.453

5510

0.599

5672

0.773

5618

1.000

5622

0.988

5618

1.000

5622

0.988

plantat_r04

5674

0.852

5608

0.973

5612

0.984

5601

0.955

5563

0.855

5663

0.882

5651

0.912

5505

0.712

5647

0.923

5518

0.743

plantat_r07

5646

0.935

5592

0.940

5653

0.919

5643

0.942

5631

0.970

5658

0.907

5656

0.911

5589

0.931

5727

0.742

5610

0.982

plantat_r10

5584

0.924

5570

0.892

5535

0.816

5568

0.887

5611

0.984

5572

0.897

5531

0.806

5562

0.875

5596

0.950

5573

0.900

pop_2000

5639

0.964

5623

0.992

5614

0.993

5574

0.922

5618

1.000

5633

0.975

5620

0.997

5579

0.930

5588

0.946

5642

0.958

pop_hist

5618

1.000

5615

0.996

5644

0.955

5614

0.994

5585

0.941

5674

0.901

5665

0.918

5637

0.968

5619

1.000

5697

0.861

rivers_d

5545

0.871

5582

0.937

5555

0.888

5625

0.988

5532

0.847

5480

0.758

5572

0.919

5531

0.846

5533

0.849

5525

0.835

rivers_r01

5732

0.772

5740

0.756

5721

0.793

5810

0.624

5818

0.608

5746

0.744

5921

0.435

5821

0.603

5753

0.734

5812

0.620

rivers_r04

5672

0.904

5709

0.838

5714

0.829

5582

0.936

5640

0.961

5762

0.746

5663

0.920

5734

0.794

5680

0.889

5597

0.962

rivers_r07

5703

0.850

5785

0.709

5681

0.890

5592

0.954

5666

0.915

5749

0.770

5750

0.768

5780

0.718

5641

0.960

5720

0.820

rivers_r10

5825

0.645

5826

0.642

5704

0.849

5673

0.903

5776

0.725

5735

0.795

5803

0.679

5837

0.625

5661

0.925

5785

0.710

roads_d

5590

0.951

5662

0.922

5522

0.831

5710

0.838

5659

0.928

5410

0.642

5510

0.810

5731

0.802

5530

0.844

5665

0.917

roads_r01

5601

0.956

5682

0.833

5551

0.829

5467

0.631

5611

0.982

5710

0.759

5707

0.768

5534

0.785

5715

0.748

5652

0.913

roads_r04

5638

0.962

5495

0.764

5717

0.807

5560

0.887

5607

0.979

5695

0.851

5600

0.966

5530

0.830

5699

0.843

5560

0.886

roads_r07

5722

0.815

5526

0.836

5890

0.537

5533

0.848

5637

0.967

5752

0.763

5758

0.752

5687

0.877

5619

1.000

5565

0.905

roads_r10

5709

0.839

5574

0.921

5746

0.775

5571

0.916

5681

0.888

5755

0.760

5686

0.881

5671

0.907

5605

0.978

5626

0.987

slope

5441

0.692

5535

0.853

5618

1.000

5606

0.979

5611

0.988

5540

0.861

5491

0.776

5585

0.941

5634

0.972

5592

0.954

towns_d

5538

0.858

5646

0.951

5665

0.918

5785

0.709

5578

0.930

5645

0.953

5522

0.831

5714

0.831

5795

0.693

5574

0.922

towns_r01

5618

1.000

5618

1.000

5618

1.000

5618

1.000

5618

1.000

5618

1.000

5618

1.000

5618

1.000

5618

1.000

5618

1.000

towns_r04

5670

0.808

5619

1.000

5721

0.618

5614

0.987

5618

1.000

5619

1.000

5618

1.000

5669

0.810

5518

0.660

5669

0.810

towns_r07

5690

0.826

5786

0.601

5684

0.841

5534

0.802

5728

0.735

5535

0.803

5557

0.856

5643

0.942

5438

0.593

5696

0.810

towns_r10

5618

1.000

5563

0.888

5554

0.868

5318

0.443

5588

0.938

5425

0.620

5560

0.880

5341

0.478

5363

0.514

5629

0.978

Chi-sq
protect_cat

0.115

p
0.944

Chi-sq
0.467

p
0.792

Chi-sq
0.066

p
0.967

Chi-sq
0.312

p
0.856

Chi-sq
0.000

p
1.000

Chi-sq
0.066

p
0.967

Chi-sq
0.022

p
0.989

Chi-sq
0.062

p
0.970

Chi-sq
0.340

p
0.844

Chi-sq

p

0.095

0.954

7

87

Appendix C. Models

88

C.1. Variable reduction procedure

89

We selected the variables for representing each hypothesis using the available

90

knowledge about the biology of jaguars (see details in the Table 1 and 2) (Burnham and

91

Anderson 2002; Zuur et al. 2010; Zuur et al. 2009). To avoid problems with multicollinearity

92

in the models and model selection process (Burnham and Anderson 2002; Zuur et al. 2010),

93

we calculated Spearman’s rank correlation coefficients among the variables. When two or

94

more variables proposed for one hypothesis showed high collinearity (r > 0.7), we retained

95

the variable that better reflects the hypothesis represented by this model. For those variables

96

that we had not a biological criterion for their selection, we used a Mann-Withney U test to

97

observe the differences between presences and pseudo-absences, and we removed the

98

variable that showed the lowest univariate difference from among high-correlated variables.

99

The variable reduction procedure was applied also to combinations of models where a

100

combination was only allowed if the variables were just weakly correlated (i.e., r < 0.7).

101
102
103

C.2. Models evaluation and habitat suitability categories
We evaluated the final and the global models by the area under the receiver operating

104

characteristic curve (AUC; Guisan and Zimmermann 2000), and the percentage of correctly

105

predicted presences (sensitivity) and pseudo-absences (specificity) using a 0.5 threshold

106

based on the prevalence approach (Liu et al. 2005). Additionally, we included a presence-

107

only evaluation method, the continuous Boyce index (Hirzel et al. 2006) using Biomapper

108

software version 4.07.303 (Hirzel et al. 2008). In order to evaluate overfitting of models, we

109

conducted a cross validation (Fernández et al. 2003; Kanagaraj et al. 2011). To this end we

110

randomly partitioned the presence and pseudo-absence data into ten folds, and we used nine

111

of them for model fitting and the remaining one for model evaluation. We repeated this

8

112

procedure 10 times and we observed the average sensitivity and specificity of each of the

113

final models and the global model. Additionally, we used the 10 subsets of 106 presence

114

records and the 10 sets of 106 pseudo-absences that were not used in the analysis (see the

115

main text) for evaluating the prediction capacity of the models.

116

To transform the final and global models into habitat suitability maps, we used the

117

logistic model equation (Burnham and Anderson 2002). We combined the maps representing

118

each variable of the model with the Map Calculator of ArcGIS Spatial Analyst. The resultant

119

maps that represent the relative probability of jaguar presence were transformed into

120

categories of habitat quality following Naves et al. (2003) and Hirzel et al. (2006). Areas with

121

values where ≤ 5% of the presence records occurred were categorized as matrix (Naves et al.

122

2003). Marginal habitat was defined as the area with >5% of the presence records until the

123

value from which more presence records occurred than expected by chance (Hirzel et al.

124

2006). Areas above this value were categorized as suitable habitat which was then subdivided

125

into three suitability categories (low, medium and high suitability) using the changes in the

126

slope of the curve of the continuous Boyce index as described by Hirzel et al. (2006).

9

127

Table C1. Evaluation of hypotheses and selection of models for the particular hypotheses representing land-cover and physical conditions of the

128

landscape. Only the selected models for each particular hypothesis are shown; the best model is in bold type.
General
hypotheses
Null model
Native
forest (F)

Particular hypotheses

Variables in the final model (+ or - effect)

Null model
F1) Amount of forest
F2) Local connectivity
F3) Amount and
connectivity

intersection
forest_r7 (+)
connect_r7 (+)

F4) Forest history
F5) Combination
Physical
environment
(PE)
Human land
uses (U)

129
130
131
132
133
134

PE1) Rivers
PE2) Elevation
PE3) Slope
PE4) Combinations
U1a) Intensive agriculture
U1a) Farms
U1a) Pastures
U1a) Plantations
U2) Combinations

forest_r1 (+), connect_r7 (+)
forest73_r7(+), forest73_r72 (-),
forest73_r73(+)
forest_r1(+), connect_r7(+),
forest73_r7(+), forest73_r72 (-),
forest73_r73(+)
rivers_r4(+)
elevation (n.s.)
slope (n.s.)
elevation (n.s.), slope (n.s.), rivers_r4 (+)
int_agr_r1(-)
farms_r4 (-)
pastures_r4 (-)
plant_r10 (-), plant_r102 (+)
int_agr_r1(-), farms_r4 (-), pastures_r4 (-)

Wald’s 2

df

p

D2adj

AICc

AICc

wr
(%)

0.0
97.1
77.9

1
1

<0.001
<0.001

0.33
0.26

295.9
200.9
220.1

15.2
34.4

0.0
0.0

103.9

2

<0.001

0.35

196.2

10.5

0.5

67.3

3

<0.001

0.22

234.8

49.1

0.0

120.7

5

<0.001

0.39

185.6

0.0

99.4

11.4
0.3
0.0
11.5
19.4
71.0
30.4
18.1
108.3

1
1
1
3
1
1
1
2
3

0.001
0.583
0.838
0.009
<0.001
<0.001
<0.001
<0.001
<0.001

0.03
0.00
0.00
0.03
0.06
0.24
0.10
0.05
0.37

286.6
297.7
297.9
290.6
278.6
227.0
267.5
281.9
192.8

0.0 87.6
11.1
0.3
11.4
0.3
4.0 11.8
85.8
0.0
34.2
0.0
74.7
0.0
89.1
0.0
0.0 100.0

Notes: Variable abbreviations are from Table B1; D² adj is adjusted explained deviance; AICc is bias-corrected Akaike’s Information Criterion for fitted models; AICc is
(AICc)I − (AICc)min; wr is the AICc weights expressed in percentages; (+) or (–) indicates the direction of the effect of the variable or the variable components for predicting
jaguar presence, * indicates interaction between variables; superscripts numbers indicate quadratic or cubic adjustments; _r followed by 1,4,7 or 10 indicates the variable
calculated for 1-, 4-, 7- or 10-km radius respectively.

10

135

Table C2. Evaluation of hypotheses and selection of models for the particular hypotheses representing human persecution of jaguars and their

136

prey. Only the selected models for each particular hypothesis are shown; the best model is in bold type.
General
hypotheses
Null model
Protection and
human access
(PA)

Population
density (RP)

137
138
139
140

Particular hypotheses

Variables in the final model (+ or - effect)

Null model
PA1) Protection

intersection
protect_cat0 (+), protect_cat1 (+),
protect_cat0 (+)
access_cost (+)
protect_cat0 × access_cost (+),
protect_cat1 × access_cost (+),
protect_cat2 × access_cost (+)
population_2000 (-)
population_hist (-)

PA2) Access cost
PA3) Protection and
access cost
RP1) Present
RP2) Historical
average

Wald’s 2

df

0.0

-

D2adj

p

AIC c

AICc

wr
(%)

-

-

295.9

62.6

2 <0.001

0.21

237.4

11.9

33.2

37.9

1 <0.001

0.12

260.0

34.4

0.0

76.5

3 <0.001

0.25

225.6

0.0

66.8

21.2

1 <0.001

0.07

276.8

0.5

44.1

21.7

1 <0.001

0.07

276.3

0.0

55.9

Notes: Variable abbreviations are from Table B1; D² adj is adjusted explained deviance; AICc is bias-corrected Akaike’s Information Criterion for fitted models; AICc is
(AICc)I − (AICc)min; wr is the AICc weights expressed in percentages; (+) or (–) indicates the direction of the effect of the variable or the variable components for predicting
jaguar presence, * indicates interaction between variables; _r followed by 1,4,7 or 10 indicates the variable calculated for 1-, 4-, 7- or 10-km radius respectively.

11

141

Table C3. Variables and parameters for the final models of each main group of hypotheses (land

142

cover and human persecution) and for the global model. Variable abbreviations are from Table B1,

143

and the model selection process is detailed in Tables 3 and 4 in the manuscript. The maps representing

144

each of these models in the study area are shown in Fig. C1.
Model

Parameter

Land cover

(intersection)

β

Std.

95% Confidence interval

error

Lower

Upper

-3.311

1.898

-7.031

0.409

forest_r1

4.2E-02

2.9E-02

-1.4E-02

9.8E-02

connect_r7

3.6E-03

2.5E-03

-1.3E-03

8.6E-03

forest73_r7

0.017

0.008

0.002

0.032

forest73_r72

-1.9E-05

1.1E-05

-4.0E-05

2.1E-06

forest73_r73

6.4E-09

4.5E-09

-2.3E-09

1.5E-08

int_agr_r1

-0.062

0.041

-0.142

0.018

farms_r4

-0.009

0.003

-0.014

-0.003

past_r4

-0.005

0.004

-0.012

0.003

Human

population_hist

-0.093

0.018

-0.128

-0.058

persecution

protect_cat=0 × access_cost

3.0E-04

9.2E-05

1.2E-04

4.9E-04

protect_cat=1 × access_cost

7.9E-04

2.9E-04

2.1E-04

1.4E-03

protect_cat=2 × access_cost

0.003

0.001

0.001

0.005

protect_cat=0 × access_cost

1.8E-04

1.8E-04

-1.7E-04

5.3E-04

protect_cat=1 × access_cost

1.2E-04

3.1E-04

-4.9E-04

7.2E-04

protect_cat=2 × access_cost

0.001

0.001

0.000

0.003

-0.076

0.032

-0.139

-0.014

forest_r1

0.008

0.028

-0.047

0.063

forest73_r7

0.011

0.004

0.002

0.020

forest73_r72

-1.3E-05

7.3E-06

-2.7E-05

1.3E-06

forest73_r73

4.5E-09

3.3E-09

-1.9E-09

1.1E-08

int_agr_r1

-0.093

0.041

-0.173

-0.014

farms_r4

-0.011

0.003

-0.016

-0.005

past_r4

-0.009

0.003

-0.016

-0.002

Global

population_hist

145
146

12

b) Human persecution model

a) Land cover model

Land cover habitat suitability

147

Human impact habitat suitability

c) Global model

148
149

Fig. C1. One-dimensional habitat suitability maps of the final models: a) Land-cover model;

150

b) Human persecution model; and c) Global model.

13

151

Table C4. Evaluation of the final models developed for predicting jaguar presence along the Upper Paraná Atlantic Forest. Boyce index calculated using

152

Biomapper 4.07.303 (Hirzel et al. 2008; Hirzel et al. 2006). AUC: area under the receiver operating curve; Sensitivity: prediction of presences; Specificity:

153

prediction of pseudo-absences; General: prediction capacity including both presences and pseudo-absences.

154
Presence
Model

only

Cross validation a

Original presences and pseudo-absences

Data or pseudo-absences resampling a

Boyce

AUC

Sensitivity

Specificity

General

Sensitivity

Specificity

General

Sensitivity

Specificity

General

Land-cover

0.998

0.905

85.0 %

82.0 %

83.5%

84.0 %

82.1 %

83.1 %

84.4 %

74.4 %

79.4%

Human

0.998

0.841

74.0 %

82.0 %

78.0 %

71.7 %

77.4 %

74.6 %

60.8 %

91.3 %

76.1 %

0.977

0.915

80.0 %

79.0 %

79.5 %

84.9 %

80.2 %

82.6 %

76.5 %

88.6 %

82.5 %

persecution
Global

155

a

Mean percentages after ten evaluations with different extracted folds or resampling sets or subsets.

14

156

Appendix D. Categories and sub-categories of habitat in the two-dimensional model.

157

158
159

Fig. D1. Distribution of the presences (red large dots) and pseudo-absences (empty small dots) along

160

the habitat categories of the two final models used for the two-dimensional categorization of habitat

161

suitability for jaguars in the Upper Paraná Atlantic Forest. Most of the presence records occurred in or

162

around the core areas, and the longest distance of a presence record from a core area was 23 km.

163

However, areas classified as core areas included some small and isolated areas that may not be

164

significant for the target species (Hirzel and Le Lay 2008). The Perobas Biological Reserve (Fig. 3)

165

and surroundings in Brazil (74 km2) was the smallest isolated core area with confirmed jaguar

166

presence. We therefore re-categorized as potential refuges the core areas that were smaller than 74

167

km2 and located farther than 23 km from another core area.

168

15

169
170

Fig. D2. Sub-categories of marginal habitats according to the different levels of suitability determined

171

by the land-cover conditions and human persecution models. The graph shows the distribution of

172

presences (red dots) and pseudo absences (small empty dots) along the habitat categories. The red and

173

blue arrows indicate the priority management action needed for transforming the suboptimal areas

174

(i.e. refugelike and attractive sinklike areas), into core areas. AS: attractive sinklike areas; R:

175

refugelike areas; -M: located in marginal habitats.

16

176
177

Fig. D3. Distribution of the sub-categories of marginal habitats determined by the different levels of

178

suitability from the land-cover and human persecution models (see Fig. D2). The arrows in the legend

179

indicate the main management actions needed for increasing the suitability of each habitat category

180

(red: protection/mitigation; blue: territorial planning for land-cover recovery and restoration). Larger

181

arrows indicate priority areas (better land-cover conditions or less human persecution) where

182

management interventions will have a greater effect in transforming these areas into core areas. AS:

183

attractive sinklike areas; R: refugelike areas; -M: located in marginal habitats.

17

184
185

Fig. D4. Sub-categories of habitats inside the core areas determined by the different levels of

186

suitability from the land-cover conditions and human persecution models. The graph shows the

187

distribution of presences (red dots) and pseudo absences (small empty dots) along the habitat

188

categories. The red and blue arrows indicate the priority management action needed for transforming

189

the suboptimal areas (i.e. refugelike and attractive sinklike areas), into core areas. AS: attractive

190

sinklike areas; R: refugelike areas; C: core area sub-category; -C: located in core areas.

18

191
192

Fig. D5. Distribution of the sub-categories of habitats inside the core areas determined by the different

193

levels of suitability from the land-cover and human persecution models (see Fig. D4). The arrows in

194

the legend indicate the main management action needed for increasing the suitability of each habitat

195

category (red: protection/mitigation; blue: territorial planning for land-cover recovery or restoration).

196

Larger arrows indicate priority areas (better land-cover conditions or less human persecution) where

197

management interventions will have a greater effect in transforming these areas into potential sources.

198

AS: attractive sinklike areas; R: refugelike areas; C: core area sub-category; -C: located in core areas.

199

19

200

Appendix E. Distribution of killed jaguars along the categories of habitat.

201
a)

202
b)
Killed jaguars

14

Observed

Expected

12
10
8

6
4
2
0

203
204

Fig. E1. Observed and expected proportion of killed and removed jaguars (n=30) that

205

occurred in the different categories of habitat along the Green Corridor of Argentina plus a 23

206

km buffer area. In the upper graph (a) we joined both sinklike and attractive sinklike areas in

207

one all-sinks category for the statistical analysis (no refugelike areas occurred in this region,

208

see the results of the analysis in the main text). The bottom graph (b) shows a detailed

209

distribution of the observed and expected frequency of killed and removed jaguars, where

210

potential attractive-sinklike in marginal areas are (from worst to better land-cover

211

conditions): AS1-M, AS2-M, AS3-M. Potential attractive sinklike areas in core areas are

212

(from worst to better land-cover conditions): Sinks-C, AS1-C, AS2-C. See sub-categories in

213

Figs. D2 and D4.

214
20

215

Appendix F. Validation of the Biodiversity Vision of the UPAF.

216
217

Table F1. Validation of the Biodiversity-Vision conservation landscape (Di Bitetti et al.

218

2003) through the two-dimensional model developed for jaguars in the Upper Paraná Atlantic

219

Forest.
Biodiversity Vision categories

Jaguar model categories
Matrix

Sink

Refuge

Attractive

Core

(%)

like

like

sinklike

areas

(%)

(%)

(%)

(%)

Core areas

3

1

0

17

79

High potential for became core

3

3

0

66

28

Potential core area

6

6

9

28

51

Forested area that needs

4

5

13

12

65

Satellite areas

9

3

23

40

25

Main corridors

21

17

2

43

16

Secondary corridors

28

16

8

30

18

Lateral expansions of corridors

23

19

4

40

14

Isolated areas

67

8

1

19

5

Potential corridors

65

18

1

11

4

Areas needing a corridor

87

10

0

3

0

Priority river basin

56

20

7

12

6

area

assessment

220
221

21

222

References of the supplementary data

223

Beyer, H.L., 2004. Hawth's Analysis Tools for ArcGIS.

224

Burnham, K.P., Anderson, D.R., 2002. Model selection and multimodel inference: a practical

225
226

information-theoretic approach. Springer-Verlag, New York, USA.
Carroll, C., Miquelle, D.G., 2006. Spatial viability analysis of Amur tiger Panthera tigris

227

altaica in the Russian Far East: The role of protected areas and landscape matrix in

228

population persistence. Journal of Applied Ecology 43, 1056-1068.

229
230
231

Chefaoui, R.M., Lobo, J.M., 2008. Assessing the effects of pseudo-absences on predictive
distribution model performance. Ecological Modelling 210, 478-486.
De Angelo, C., 2009. El paisaje del Bosque Atlántico del Alto Paraná y sus efectos sobre la

232

distribución y estructura poblacional del jaguar (Panthera onca) y el puma (Puma

233

concolor), p. 252. Universidad de Buenos Aires, Buenos Aires, Argentina.

234

De Angelo, C., Paviolo, A., Di Bitetti, M., 2011a. Differential impact of landscape

235

transformation on pumas (Puma concolor) and jaguars (Panthera onca) in the Upper

236

Paraná Atlantic Forest. Diversity and Distributions 17, 422-436.

237

De Angelo, C., Paviolo, A., Rode, D., Cullen Jr, L., Sana, D., Abreu, K.C., Xavier da Silva,

238

M., Bertrand, A.-S., Haag, T., Lima, F., Ricieri Rinaldi, A., Ramírez, F., Fernández,

239

S., Velázquez, M., Corio, C., Hasson, E., Di Bitetti, M.S., 2011b. Participatory

240

networks for large-scale monitoring of large carnivores: pumas and jaguars of the

241

Upper Paraná Atlantic Forest. Oryx 45, 534-545.

242

Di Bitetti, M.S., Placci, G., Dietz, L.A., 2003. A biodiversity vision for the Upper Paraná

243

Atlantic Forest Eco-region: designing a biodiversity conservation landscape and

244

setting priorities for conservation action. World Wildlife Fund, Washington, D.C.,

245

USA.

22

246

Engler, R., Guisan, A., Rechsteiner, L., 2004. An improved approach for predicting the

247

distribution of rare and endangered species from occurrence and pseudo-absence data.

248

Journal of Applied Ecology 41, 263-274.

249

Farrow, A., Nelson, A., 2001. Accessibility Modelling in ArcView 3.1: an extension for

250

computing travel time and market catchment information. Centro Internacional de

251

Agricultura Tropical (CIAT), Cali, Colombia.

252

Fernández, N., Delibes, M., Palomares, F., Mladenoff, D.J., 2003. Identifying breeding

253

habitat for the iberian lynx: Inferences from a fine-scale spatial analysis. Ecological

254

Applications 13, 1310-1324.

255
256
257

Guisan, A., Zimmermann, N.E., 2000. Predictive habitat distribution models in ecology.
Ecological Modelling 135, 147-186.
Hengl, T., Sierdsema, H., Radovic, A., Dilo, A., 2009. Spatial prediction of species'

258

distributions from occurrence-only records: combining point pattern analysis, ENFA

259

and regression-kriging. Ecological Modelling 220, 3499-3511.

260
261
262

Hirzel, A.H., Hausser, J., Perrin, N., Le Lay, G., Braunisch, V., 2008. Biomapper 4.07.303.
Laboratory for Conservation Biology, University of Lausanne.
Hirzel, A.H., Lay, G.L., Helfer, V., Randin, C., Guisan, A., 2006. Evaluating the ability of

263

habitat suitability models to predict species presences. Ecological Modelling 199,

264

142-152.

265
266
267

Hirzel, A.H., Le Lay, G., 2008. Habitat suitability modelling and niche theory. Journal of
Applied Ecology 45, 1372-1381.
Kanagaraj, R., Wiegand, T., Kramer-Schadt, S., Anwar, M., Goyal, S.P., 2011. Assessing

268

habitat suitability for tiger in the fragmented Terai Arc Landscape of India and Nepal.

269

Ecography 34, 970-981.

23

270
271
272

Liu, C., Berry, P.M., Dawson, T.P., Pearson, R.G., 2005. Selecting thresholds of occurrence
in the prediction of species distributions. Ecography 28, 385-393.
Mateo, R.G., Croat, T.B., Felicísimo, Á.M., Muñoz, J., 2010. Profile or group discriminative

273

techniques? Generating reliable species distribution models using pseudo-absences

274

and target-group absences from natural history collections. Diversity and Distributions

275

16, 84-94.

276

Naves, J., Revilla, E., Delibes, M., Wiegand, T., 2003. Endangered species constrained by

277

natural and human factors: the case of brown bears in Northern Spain. Conservation

278

Biology 17, 1276-1289.

279

Phillips, S.J., Dudík, M., Elith, J., Graham, C.H., Lehmann, A., Leathwick, J., Ferrier, S.,

280

2009. Sample selection bias and presence-only distribution models: implications for

281

background and pseudo-absence data. Ecological Applications 19, 181-197.

282

Schadt, S., Revilla, E., Wiegand, T., Knauer, F., Kaczensky, P., Breitenmoser, U., Bufka, L.,

283

Cerveny, J., Koubek, P., Huber, T., Stanisa, C., Trepl, L., 2002. Assessing the

284

suitability of central European landscapes for the reintroduction of Eurasian lynx.

285

Journal of Applied Ecology 39, 189-203.

286

Titeux, N., 2006. Modelling species distribution when habitat occupancy departs from

287

suitability: Application to birds in a landscape context, p. 256. Université Catholique

288

de Louvain, Louvain, Belgique.

289

VanDerWal, J., Shoo, L., Graham, C., Williams, S., 2009. Selecting pseudo-absence data for

290

presence-only distribution modeling: How far should you stray from what you know?

291

Ecological Modelling 220, 589-594.

292
293

Wiegand, T., Moloney, K.A., 2004. Rings, circles and null-models for point pattern analysis
in ecology. Oikos 104, 209-229.

24

294
295
296
297

Zuur, A.F., Ieno, E.N., Elphick, C.S., 2010. A protocol for data exploration to avoid common
statistical problems. Methods in Ecology and Evolution 1, 3-14.
Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., Smith, G., 2009. Mixed effects models
and extensions in ecology with R. Springer, New York, USA.

298
299

25


Documentos relacionados


Documento PDF de angelo et al 2013 biol cons a management landscape for jagaurs
Documento PDF de angelo et al 2011 impact of landscape transformation on pumas and jaguars
Documento PDF srep37147 3
Documento PDF martinez pardo et al 2017 1
Documento PDF de angelo et al 2010 jwm track identification
Documento PDF paviolo et al 2008 jaguar pop decline oryx


Palabras claves relacionadas