Estimating effective landscape distances and movement corridors:

comparison of habitat and genetic data

MARIA C. MATEO-SÁNCHEZ,1,  NIKO BALKENHOL,2 SAM CUSHMAN,3 TRINIDAD PÉREZ,4
ANA DOMÍNGUEZ,4 AND SANTIAGO SAURA1
1
E.T.S.I Montes, Forestal y del Medio Natural, Technical University of Madrid,
Ciudad Universitaria s/n 28040, Madrid, Spain
2
Department of Wildlife Sciences, University of Goettingen, Buesgenweg 337077, Goettingen, Germany
3
United States Forest Service, Rocky Mountain Research Station, Southwest Forest Science Complex,
2500 South Pine Knoll Drive, Flagstaff, Arizona 86001 USA
4
Department of Functional Biology, University of Oviedo, Avda. Julián Claverı́a 6, 33006 Oviedo, Spain

Citation: Mateo-Sánchez, M. C., N. Balkenhol, S. Cushman, T. Pérez, A. Domı́nguez, and S. Saura. 2015. Estimating
effective landscape distances and movement corridors: comparison of habitat and genetic data. Ecosphere 6(4):59. http://
dx.doi.org/10.1890/ES14-00387.1

Abstract. Resistance models provide a key foundation for landscape connectivity analyses and are
widely used to delineate wildlife corridors. Currently, there is no general consensus regarding the most
effective empirical methods to parameterize resistance models, but habitat data (species’ presence data and
related habitat suitability models) and genetic data are the most widely used and advocated approaches.
However, the practical consequences of applying one or the other approach have not been well studied. To
address this knowledge gap, we performed a comparative study on the implications of using habitat
suitability versus genetic data for determining effective landscape distances (a proxy inversely related to
isolation among patches) based on least-cost and circuit-theoretic approaches, and for identifying potential
movement corridors. For our comparison, we used data for the Cantabrian brown bear in Spain, an
endangered population for which connectivity has been identified as a major conservation concern. Our
results show that for brown bears, habitat models tend to overestimate resistance to movement through
non-optimal areas, whereas genetic data yield higher estimates of effective distances within habitat areas.
Therefore, our results suggest that (1) dispersal might be generally less constrained by landscape
conditions than habitat utilization in home ranges, and that (2) dispersing animals might be more flexible
in their movement behavior than residents are in their habitat resource utilization behavior, with records
for residents dominating species occurrence data and subsequent habitat models. The assessed approaches
provided dissimilar connectivity models with notable differences in patterns of predicted corridors across
the study area, mainly due to differences in predicted connections between subpopulations. Our results
highlight that the functional differences in habitat vs. genetic data, as well as the assumptions and potential
limitations of different analytical approaches that use these data, need to be considered more carefully in
connectivity modeling and subsequent corridor delineation.

Key words: circuit theory; connectivity; dispersal; habitat suitability; landscape genetics; least-cost paths; resistance.

Received 16 October 2014; revised 2 December 2014; accepted 5 December 2014; final version received 4 February 2015;
published 21 April 2015. Corresponding Editor: D. P. C. Peters.
Copyright: Ó 2015 Mateo-Sánchez et al. This is an open-access article distributed under the terms of the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided
the original author and source are credited. http://creativecommons.org/licenses/by/3.0/
  E-mail: mc.mateo@upm.es

v www.esajournals.org 1 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

INTRODUCTION opinion itself or in combination with literature

review (Larkin et al. 2004, Kautz et al. 2006).
The establishment and conservation of corri- However, such approaches have frequently been
dors has been widely suggested as a key means shown to perform poorly (Beier et al. 2008, Shirk
to facilitate movement between wildlife habitat et al. 2010). To address these limitations, some
patches and to preserve and enhance landscape authors have suggested that habitat suitability
and population connectivity (Simberloff et al. models predicting species occurrence on the basis
1992). The classical concept of corridors as of empirical data may provide a better estimation
narrow strips of habitat that facilitate movements of resistance (Ferreras 2001, Chetkiewicz et al.
of organisms between habitat patches (Rosen- 2006, O’Brien et al. 2006, Beier et al. 2008). This
berg et al. 1997) has been recently broadened. approach essentially implies that animal move-
Rather than experiencing landscapes as categor- ments are influenced by the same environmental
ical mosaics (habitat vs. non habitat), it is more factors as habitat selection (e.g., Chetkiewicz et
likely that organisms experience the landscape al. 2006, Wasserman et al. 2010, Zeller et al. 2012).
matrix as a gradient of differential permeability More recently, the field of landscape genetics has
(Cushman et al. 2010). This generalization main- shown great potential to provide more rigorous
streams the concept of resistance to movement methods to parameterize resistance models by
and suggests reconceptualization of connectivity inferring the influences of landscape on realized
problems to address how species movement is population connectivity (Spear et al. 2005, Cush-
affected by landscape features across a range of man et al. 2006, Storfer et al. 2007, Balkenhol et
spatial scales. Formally, landscape resistance al. 2009). Individual-based analysis comparing
represents an integration of multiple behavioral pairwise genetic distances to pairwise effective
and physiological factors such as aversion, distances under multiple landscape resistance
energy expenditure, or mortality risk when hypothesis are a powerful tool for supporting
moving through a particular environment (Zeller conservation efforts (Cushman et al. 2006, 2013a,
et al. 2012). In a practical sense, landscape Epps et al. 2007, Segelbacher et al. 2010).
resistance is a spatial layer that reflects step-wise Understanding how different approaches can
costs of moving through each cell in a raster map affect the analyses that aim to identify conserva-
for least cost path analyses (Singleton et al. 2002) tion corridors is a critical issue given the large
or the relative probability of moving across the conservation implications and investments that
cell for circuit theory-based analyses (McRae et are potentially derived from such analyses. In
al. 2008). this paper we addressed this issue by comparing
Since modeling landscape resistance is increas- predicted movement corridors and effective
ingly used as the basis to predict population distances (derived from least cost analysis and
connectivity and therefore is a crucial step in circuit theory) across alternative resistance sur-
identifying movement corridors, resistance mod- face scenarios based on different parameteriza-
eling has received much attention from corridor tion methodologies: habitat modeling and
designers (Beier et al. 2008, Spear et al. 2010, landscape genetics. We also assessed how two
Zeller et al. 2012). However, despite the existence different variable integration approaches (addi-
of multiple and varied approaches, there is no tive and multiplicative) affected the predictions
consensus on how to best parameterize resistance and implications of landscape genetic models of
surfaces (Zeller et al. 2012). Parameterizing resistance.
resistances involves identifying and combining Our aim was to gain insights into the
the important resistant factors (e.g., land cover, differences and similarities between the results
transport infrastructure, etc.) that affect the obtained using the different approaches, their
species’ movement in an optimal way to realis- potential limitations, and consequent manage-
tically estimate the cost of movement through ment implications. We considered brown bear
any location in the landscape. Resistance values (Ursus arctos) in the Cantrabrian Range (NW
have usually been determined by assigning Spain) as the focal species. Brown bear is a long-
resistance scored on an arbitrary scale (Beier et lived omnivorous mammal with a solitary social
al. 2008, Pereira et al. 2011) on the basis of expert structure and promiscuous mating system

v www.esajournals.org 2 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

(Nores and Naves 1993, Schwartz et al. 2003). in Spain, its peripheral areas and the belt area
Males and females have intra and inter-sexually between the two subpopulations. As shown in
overlapping home ranges (Dahle and Swenson Fig. 1, both subpopulations occupy a similar area
2003) and dispersal primarily occurs by males, of about 2,500 km2 each and are separated by
while females typically are philopatric (Swenson about 50 km of unoccupied range.
et al. 1998). Brown bear is highly dependent on
large landscapes with low human-footprint and Landscape resistance parameterization
large extents of forest cover (Clevenger et al. Landscape resistance was parameterized un-
1997, Apps et al. 2004, Mateo-Sánchez et al. der three different scenarios based on different
2014a). The brown bear population in the methodological approaches (habitat scenario,
Cantabrian Range suffered a dramatic decline genetic-multiplicative scenario, and genetic-ad-
in the last several centuries as a result of human ditive scenario), as described next. All resistance
persecution and progressive loss and fragmenta- maps were produced with a spatial resolution of
tion of its habitat (Naves et al. 2003). We 100 m.
considered this population highly suited for this Habitat suitability as a proxy for resistance to
analysis for the following reasons. (1) In the movement.—A plausible way to empirically esti-
Cantabrian Range, brown bears occur in two mate relationships between connectivity and
small, apparently isolated and endangered sub- environmental conditions is to assume that
populations, with about 220 individuals in total habitat quality has a direct (inverse) relationship
(Pérez et al. 2014). Connectivity has been with resistance to movement (e.g., Pullinger and
identified as a major conservation concern for Johnson 2010, Kuemmerle et al. 2011, Mateo-
this species, with potentially large implications Sánchez et al. 2014b). We used this approach in
for actual planning and conservation measures in our first parameterization scenario (habitat sce-
the study area (Ballesteros and Palomero 2012). nario), in which we created a resistance surface
(2) Brown bears and other large mammals are of where resistance to movement was obtained
particular interest for connectivity networks through an inverse function of habitat suitability.
because these species operate at broad scales We used a multiscale suitability habitat model
and occur at low densities, which imply that their developed by Mateo-Sánchez et al. (2014a) to
populations are more likely to be affected by the predict brown bear occurrence in the study area.
loss of connectivity (Beier et al. 2008). (3) Large After transformation every pixel represented the
amounts of habitat and genetic data are available unit cost of crossing each location, so that the
for the species and recent research has focused on lowest resistance value represented the cost of
landscape resistance and connectivity for this moving through the highest quality habitat.
species (Pérez et al. 2009, 2010, Mateo-Sánchez et Landscape variables included in the model were
al. 2014a, b) which provides a unique opportunity landscape composition (percentage of landscape
for the comparative analyses needed to effective- covered by forest), forest canopy cover and
ly tackle the aforementioned objectives. density of buildings (Mateo-Sánchez et al.
2014a, b).
MATERIALS AND METHODS Landscape genetics to infer resistance to move-
ment.—We used individual-based landscape ge-
Study area netics approaches to produce resistance models
The study was carried out in the Cantabrian under two different scenarios (Mateo-Sánchez et
Range (northwestern Spain). This area is within a al., in press). Specifically, we used genetic samples
larger transnational initiative covering protected of brown bears genotyped at 17 polymorphic
areas from the Cantabrian Range to the Western microsatellite loci to quantify genetic structure
Alps (SW Europe), in which previous studies on and measured the genetic distance among
connectivity and the barrier effect of roads for samples as the proportion of shared alleles
forest mammals have focused (Gurrutxaga et al. (Bowcock et al. 1994). The relationship between
2011, Jongman et al. 2011). The region is 49,472 the genetic structure observed within the bear
km2 in extent and contains the whole known population and likely drivers of landscape
range of the native populations of the brown bear resistance was systematically evaluated through

v www.esajournals.org 3 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Fig. 1. Maps showing the location of the study area in Spain (small map) as well as the distribution of brown
bears within the study area in Spain (large map).

reciprocal causal modeling (Cushman et al. 2006, bear locations distributed across the species
2013b) and the multi-model optimization ap- range for which we also counted with genetic
proach developed by Shirk et al. (2010). The data, which allowed meaningful comparisons
resulting resistance models included variables of between results of the different approaches
landscape composition (percentage of landscape (habitat vs. genetic).
cover by mixed forest and agricultural lands), We applied two different approaches: (A) least
landscape configuration (cohesion of mixed cost path modeling, using the UNICOR software
forest and shrubland) and canopy cover. (Landguth et al. 2012), in which the movement of
One of the key questions in resistance param- individuals is assumed to follow the optimal
eterization relates to the method for combining (least costly) pathway between locations and (B)
the effects of individual landscape variables into circuit theory, using CIRCUITSCAPE (v3.5.8;
a multivariate resistance surface (Beier et al. McRae and Beier 2007, McRae et al. 2008), in
2008). Hence, we explored two different ways for which multiple available pathways (including
combining individual landscape variables: suboptimal ones) can be followed by the indi-
through multiplication (genetic-multiplicative viduals and contribute to estimated connectivity
scenario) and through addition (genetic-additive among locations. These two approaches were
scenario). used to (1) produce corridors expected to
concentrate brown bear movements in the study
Corridor identification and effective area (least cost path density and current density
distance estimation map for approaches A and B, respectively;
We used an individual-based approach to further details on our analyses can be found in
predict expected movement corridors. We con- Appendix A), and to (2) calculate the accumu-
sidered as sources and destinations for the lated cost of movement between source and
corridor mapping a set of 173 empirical brown destination areas, corresponding to the so called

v www.esajournals.org 4 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Fig. 2. Locations for effective distance and effective resistance calculations. We calculated effective distances
and resistances between (a) the cores of the ranges of the two subpopulations, (b) the periphery of the two ranges,
and (c) within the two ranges. See also Table 1.

effective distances in least cost path modeling resistance scenarios described above. Since we
and to effective resistances (or resistance dis- aimed to compare the effective distances pro-
tance) in circuit-based modeling (hereafter both duced by the three resistance scenarios, each with
referred to as effective distances). Higher effec- different ranges of variation in the resistance
tive distances among locations are assumed to values, we first normalized effective distances by
correspond to a higher degree or likelihood of dividing the effective distance in each scenario by
isolation among habitat areas or locations. the mean resistance value of all the 173 pixels
Effective distances were calculated between (a) with bear locations in the corresponding resis-
the western subpopulation core area and eastern tance surface.
subpopulation core area, (b) the western sub- We therefore produced and compared six sets
population peripheral area and eastern subpop- of corridors and effective distance/resistance
values, corresponding to the two analytical
ulation peripheral area (i.e., the edge of each core
approaches (least cost paths and circuitscape)
population area that is closest to the other
and the three resistance surface scenarios (Table
subpopulation edge) and (c) end to end within
2).
each subpopulation area (i.e., travel through the
whole occupancy area) (Fig. 2). We focused on
RESULTS
these positions due to their strategic significance
in terms of connectivity between and within the Corridor comparison
two subpopulations (see also Table 1). These The predicted movement corridor network
linkages (corridors) and effective distances were among individuals showed substantial differenc-
produced for each of the three landscape es across the three resistance scenarios and two

Table 1. Definition of analyzed effective distances.

Name Definition Meaning

a: between subpopulation Movement between two individuals located in Connectivity between subpopulations,
centers the core of the west subpopulation and the assuming most dispersers move from core
core of the east subpopulation. to core.
b: between subpopulation Movement between two individuals located in Connectivity between subpopulations,
edges the peripheral area of the west assuming that most dispersers will first try
subpopulation and the peripheral area of to stay within established ranges and
east subpopulation (border of each area eventually move outside currently occupied
closest to the other subpopulations). areas through the more hostile matrix to
Peripheral areas correspond with the limit of relocate into another range.
both subpopulations occupancy area.
c: within subpopulations Movement through the whole occupancy area Connectivity within populations.
(for each subpopulation).

v www.esajournals.org 5 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Table 2. Sets of methods being compared for delineating corridors and estimating effective distances. These
methods are based on three different approaches for parameterizing resistance models, and two different
approaches for delineating corridors and estimating effective distances based on these models.

Resistance parameterization
Corridor delineation
approach Habitat scenario Genetic-multiplicative scenario Genetic-additive scenario
Least-cost path density Corridor option 1 Corridor option 2 Corridor option 3
(Fig. 3a) (Fig. 3b) (Fig. 3c)
Circuit theory Corridor option 4 Corridor option 5 Corridor option 4
(Fig. 4a) (Fig. 4a) (Fig. 4c)

analytical approaches (Figs. 3 and 4). Important- for effective resistances based on circuit theory.
ly, major functional links did not match among However, effective distances between strategic
scenarios or methods (least cost path vs circuit locations showed considerably different values
theory). In the case of linkages defined through among scenarios (Fig. 5). Effective distances
cumulative density of least cost paths on a between two individuals located within different
resistance map derived from habitat suitability, centers of the two subpopulations were 7–9%
connections between subpopulations followed lower for genetic scenarios (multiplicative and
two main routes that converged in the peripheral additive) than for the habitat scenario. More
area of the East subpopulation (Fig. 3a). In importantly, when two individuals were located
contrast, for the resistance scenarios based on in the proximate peripheral areas (edges) of both
genetic data and least cost path analysis, connec- subpopulations, effective distance was much
tions showed more extensive networks consisting higher for the habitat scenario than for genetic
in one major route complemented by several scenarios (47% and 79% higher for the genetic-
secondary routes that converge with the princi- multiplicative and genetic-additive scenarios,
pal route when using multiplicative genetic respectively). When comparing the multiplicative
resistance (Fig. 3b). In the additive genetic and additive genetic scenarios, the normalized
resistance scenario analyzed with least cost effective distances were more similar than when
paths, three parallel and interconnected routes comparing the habitat and genetic scenarios, but
coalesce in both subpopulations (Fig. 3c). When the additive genetic scenario showed the lowest
analyzing connections designated through circuit effective distances between individuals. In con-
theory, potential connections identified through trast, effective distance between two individuals
current maps also differed across the assessed located in the same core in both subpopulations
resistance scenarios. A higher concentration of was 40–30% lower when calculated across
current in narrower and more clearly defined resistance based on habitat suitability than when
areas is found in the habitat based resistance computed for resistance scenarios based on
scenario (Fig. 4a). When resistance was based on genetic data. Effective distances within cores
genetic data, more and wider permeable areas were relatively similar in the two genetic
were detected (Fig. 4b, c), with movements less scenarios, but the multiplicative approach
concentrated in thin strips of land. Results also showed higher values.
show that within-subpopulation connectivity Effective distances (effective resistances) calcu-
network pattern was relatively similar for all lated through circuit theory followed the same
the scenarios in both approaches. general pattern as least-cost effective distances
when considering movements between popula-
Effective distances comparison tion edges and movements within populations
There were strong correlations between effec- (Fig. 5b). In these cases, effective resistances in
tive distances across the three resistance scenar- the genetic-multiplicative and genetic-additive
ios assessed when all pairs of locations where scenarios were again considerably lower than for
simultaneously considered: Pearson’s correlation the habitat-based resistance scenario (37% and
coefficient r . 0.96 in all the cases for effective 51%, respectively). Oppositely, effective resistanc-
least cost distances and r . 0.82 in all the cases es between the centers of both subpopulations

v www.esajournals.org 6 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Fig. 3. Corridors defined by using least-cost paths for the three resistance scenarios: (a) Habitat suitability-
based scenario, (b) genetic-multiplicative scenario, and (c) genetic-additive scenario.

v www.esajournals.org 7 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Fig. 4. Current maps defined by circuit theory for resistance scenarios: (a) Habitat suitability-based scenario, (b)
genetic-multiplicative scenario, and (c) additive genetic-based scenario.

v www.esajournals.org 8 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

were 18% higher in the multiplicative genetic

based model than in the habitat-based model.
However, the additive genetic model and the
habitat-based model led to very similar resis-
tances. Remarkably, effective resistance between
centers of both populations was lower than
effective resistance between population edges
for the habitat model (Fig. 5b).

DISCUSSION
Most current methods to predict population
connectivity and to identify areas important for
animal movements rely on landscape resistance
surfaces (Spear et al. 2010). However, the
methods and assumptions used to create these
surfaces are critical for their effectiveness in
guiding conservation decisions. Therefore, it is
important to explore the effects that different
analytical approaches have on predicted popula-
tion connectivity and identified corridor routes.
Here, we compared resistance surface parame-
terization and resulting connectivity models
created by two different analytical approaches
for quantifying resistances and two commonly
used methods to predict likely movement paths Fig. 5. Comparison of normalized effective distances
from these resistances. Our findings showed that among resistance scenarios. Distances are shown for
alternative approaches substantially affected the the two analytical approaches used (a) least cost paths
assessment of effective distances and largely and (b) circuit theory. Resistance scenarios are: habitat-
changed the delineation of potential corridors suitability based, genetic-multiplicative and genetic-
across the study area. additive. Between subpopulations centers corresponds
to the normalized effective distance between two
Implications of multiplicative vs. individuals located in the core of the west subpopu-
additive combination of factors lation and the core of the east subpopulation; between
The way in which factors are combined in a subpopulation corresponds to the normalized distance
resistance model has important implications for between two individuals located in the peripheral area
resulting connectivity models. Our additive of the west subpopulation and in the peripheral area of
resistance surface showed denser connectivity the east subpopulation (border of each area closest to
networks and lower effective distances/resistanc- the other subpopulation); within west and east
subpopulations corresponds to the normalized effec-
es than the multiplicative surface. This suggests
tive distance between the opposite edges of that
that multiplicative resistance models are more
subpopulation.
restrictive in identifying permeable areas for
movement and hence indicate lower connectivity.
This is because areas predicted to be highly allows compensation among factors and can lead
permeable in a multiplicative model correspond to less restrictive connectivity predictions.
to those raster cells where all the factors involved
in the models have low resistance values. Thus,
Comparison of least cost-path and
in a multiplicative model, even a single factor
with high resistance in an area can decrease circuit theory predictions
predicted landscape permeability in that area. In Least-cost path analysis and circuit theory are
contrast, an additive combination of factors based on different assumptions to model con-

v www.esajournals.org 9 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

nectivity (Spear et al. 2010). While least-cost identified through genetic-based resistances. In
analyses assume animals follow a single and addition, effective distances between subpopula-
optimal pathway, circuit theory assumes random tions were substantially higher when habitat
walks so that movement is influenced by all suitability was used as a surrogate for landscape
possible pathways. Therefore, results from both resistance. These findings suggest that resistance
methods provide different and complementary surfaces based on habitat models may tend to
insights about brown bear movement. For both overestimate landscape resistance in areas with
approaches, effective distances between subpop- low habitat suitability. In our analysis, this effect
ulation centers (the individual has to cross part of was dramatic when effective distance was calcu-
the habitat area and the matrix in-between lated between peripheral areas of both subpop-
subpopulation ranges) differed from effective ulations and individuals had to cross mostly
distances between subpopulation edges (the unsuitable areas. In contrast, effective distances
individual only crosses the matrix between between individuals located within subpopula-
habitat areas). However, and surprisingly, we tions were lowest when resistance surfaces were
found that circuit-theoretic effective resistance based on habitat suitability. Thus, congruent with
between subpopulation edges was actually high- results from Wasserman et al. (2010), we found
er than the effective resistance between popula- that habitat suitability can predict a greater
tion centers in the habitat-based resistance resistance to poor habitat than is suggested by
scenario (Fig. 5b). To have a lower effective genetic data. In other words, the fact that the
resistance between the centers of the habitat species does not occur ( permanently) in a
areas than between the less distant edges of those particular habitat does not imply that it cannot
areas is rather difficult to interpret ecologically,
move when needed (e.g., dispersal) through this
because bears moving away from the edges first
area. In this sense, landscape genetics models
have to reach the edges before crossing the
integrate the movement of many individuals
matrix to reach the other subpopulation. This
over time and thus lead to a more synoptic
result might point out to a potential limitation of
measure of landscape resistance (Zeller et al.
circuit theory (or of the implementation of it) to
2012, Cushman et al. 2013a). This suggests that
assess landscape connectivity. We ensured that
there is not necessarily a correspondence be-
our finding was not an artifact arising from some
tween habitat use patterns and dispersal move-
idiosyncratic and unnoted characteristic of our
particular brown bear spatial data by creating a ments. Conditions providing suitable habitat for
simple and purposefully controlled example permanent establishment and local resource may
depicting a resistance pattern similar to the one often be different than the conditions facilitating
of the habitat-based scenario (i.e., similar extent, dispersal movements (Cushman et al. 2013a,
resolution and arrangement). This controlled Peterman et al. 2014). Since suitability models
example confirmed our results from the real bear are based on occurrence data that usually
data set, i.e., effective resistances were smaller represent locations within home ranges, habitat
between the cores than between the edges suitability models may not adequately reflect
(further details can be found in Appendix A how environments affect animals during move-
and Appendix B: Fig. B1). This finding advocates ments outside of their usual home ranges, such
for further research to fully clarify this issue from as dispersal or mating excursions (Cushman et al.
an analytical point of view and, if possible, to 2013a). For our study species, results suggests
provide guidelines for avoiding potentially un- that dispersing bears are more flexible in their
intended results in the application of circuit movement behavior and less constrained by
theory for connectivity analyses. landscape conditions than suggested from their
occurrence in typical bear habitats. While our
Habitat models may overestimate resistance to analyses have only been conducted for a single
movement through non-optimal areas species, we believe that the results may be similar
Corridors between subpopulations predicted for many other organisms as well, particularly
from habitat suitability-based resistance did not for species that are highly mobile and not strictly
match the locations or intensity of corridors confined to a specific type of habitat.

v www.esajournals.org 10 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Genetic data estimate higher resistance to per generation can lead to a sufficient amount of
movement through suitable habitat areas gene flow between populations (e.g., Mills and
For within-population movements, the genet- Allendorf 1996, Vucetich and Waite 2000). There-
ically-derived resistance surfaces predicted larger fore, facilitating sufficient levels of inter-popula-
effective distances among individuals than the tion movements should be a priority in
habitat suitability-based resistance surface. This connectivity strategies. Importantly, it is precisely
is likely due to the fact that resistance models in this kind of movements where, according to
derived from landscape genetic analyses are our results, resistance estimates and connectivity
based on comparison of genetic distances and models varied most strongly between the ana-
effective distances. Such landscape genetic ap- lytical approaches we compared. Within the two
proaches may lead to resistance surfaces that subpopulations, effective distances and resistanc-
overestimate effective distances in areas that are es were always low, and predicted movement
highly suitable (and eventually well connected to paths were very consistent among the different
each other) when factors other than resistance to analytical approaches. However, estimated dis-
movement govern spatial-genetic structure; for tances and predicted movement paths differed
instance, even when two sampled individuals are substantially among the different approaches for
separated by a very short distance (or even connectivity between the two subpopulations.
located in the same raster cell), they will show Thus, analytical choices for parameterizing and
some genetic differentiation, which may translate utilizing resistance surfaces will be most impor-
in such genetic resistance models exacerbating tant when trying to predict inter-population
the effective distance estimates at short ranges to movements, when individuals move out of
explain such genetic structure. Spatial-genetic established ranges and cross the unsuitable
structure within populations is likely less depen- landscape matrix. Importantly, this is also the
dent on the resistance of the landscape matrix kind of analysis most relevant for corridor
(Fahrig 2007), but more strongly influenced by design, as the goal is usually the (re-) connection
many other biological and ecological factors of different subpopulations located across a
acting locally, including sex-specific space-use landscape or the support of species range shifts
behavior, local population density, survival, or in response to climate change, rather than the
reproductive success (Pflüger and Balkenhol facilitation of within-population movements.
2014). Another issue related to genetic data is
that in long-lived species there may be a Conclusions for connectivity analysis based
temporal disconnection between genetic struc- on habitat suitability
ture and the current landscape; i.e., there may be Based on our study, it seems reasonable to
legacy effects of previous landscapes (James et al. question the assumption that habitat suitability
2007, Spear et al. 2010) that could lead to models can accurately capture landscape resis-
misestimate current connectivity. However, tance to movement for corridor design. Location
Landguth et al. (2010) showed that the legacy data used to produce habitat suitability models
of past landscape features is not a particularly tend to be dominated by habitat use (i.e., shelter,
important problem in species with relatively foraging) and thus by frequent routine move-
large dispersal abilities. Hence, landscape lega- ments within established home ranges. However,
cies are unlikely to affect our conclusions about genetic structure may be strongly determined by
brown bear dispersal and gene flow. mating movements and rare dispersal events.
These types of movements are unlikely to be
The importance of inter-population movements well-captured by occurrence data and may be
for predicting connectivity therefore poorly represented in resistance models
Movements outside of typical habitat are less derived from such data. Thus, as already
frequent than within-habitat movements, but suggested by Beier et al. (2008) and Zeller et al.
they are also critical for genetic exchange and (2012), genetic-based studies are likely to be more
range expansion (Nathan et al. 2003, Chetkiewicz useful to understand connectivity of populations.
et al. 2006). Previous research has shown that a However, if genetic data are not available or
variable but generally low number of migrants when recent landscape changes are not yet

v www.esajournals.org 11 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

reflected in genetic data, parameterization based single ‘best’ management action.

on habitat models, direct movement data (e.g., For now, we advocate the use of genetically-
telemetry) or even expert opinion may be still derived resistance surfaces over the use of
required (Spear et al. 2010). Indeed, resistance occurrence data, but agree with Zeller et al.
surfaces should ideally be assessed though (2012) that more comparative research is needed
analyses of multiple data sources. For example, to fill current knowledge gaps related to land-
a resistance surface that is supported by inde- scape resistance and connectivity modeling.
pendent analysis of movement and genetic data Indeed, we argue that research identifying the
(e.g., Cushman and Lewis 2010) is much more advantages and limitations of various conceptual
likely to be robust than one developed from a and analytical approaches, such as those here
single empirical data set. For example, recent reported, is urgently needed for assessing how
research in telemetry data analysis provides meaningful and useful different resistance-based
methods that help to distinguish habitat use connectivity models actually are for practical
from dispersal locations (e.g., Dickson et al. 2005, conservation planning. We hope that our study
Squires et al. 2013, Zeller et al. 2014). Compared has highlighted some of these future research
to occurrence-based habitat models, these novel needs and that it will motivate others to further
approaches could provide resistance models that investigate this important topic.
more reliably reflect actual species movement
across complex landscapes. ACKNOWLEDGMENTS
Funding was provided by the Spanish Ministry of
Implications for conservation
Science and Innovation through research grant GE-
Connectivity models should accurately predict FOUR (AGL2012-31099), Technical University of Ma-
inter-population movement and gene flow, and drid and DAAD. We are also grateful to the Regional
enable researchers to reliably identify the most Administrations involved in the brown bear manage-
likely movement routes among subpopulations. ment: Junta de Castilla y León, Gobierno de Cantabria,
As shown in this study, connectivity models and Principado de Asturias and Xunta de Galicia for
the corridors suggested by them depend strongly providing data. Thanks also to the valuable support
on the analytical methods used for creating and provided by Fundación Oso Pardo.
utilizing underlying resistance surfaces. Several
published studies have suggested ways for LITERATURE CITED
finding optimal resistance values (e.g., Kuroe et
Apps, C. D., B. N. McLellan, J. G. Woods, and M. F.
al. 2011, Shirk et al. 2010, Graves et al. 2014), and Proctor. 2004. Estimating grizzly bear distribution
these studies have certainly improved resistance- and abundance relative to habitat and human
based connectivity modeling. However, the influence. Journal of Wildlife Management
effects of different conceptual approaches under- 68:138–152.
lying resistance models (habitat suitability vs. Balkenhol, N., F. Gugerli, S. A. Cushman, L. P. Waits,
resistance to inter-population movement) or the A. Coulon, J. Arntzen, R. Holderegger, and H. H.
analytical choices made during their creation and Wagner. 2009. Identifying future research needs in
landscape genetics: where to from here? Landscape
analysis (additive vs. multiplicative models,
Ecology 24:455–463.
least-cost vs. circuit-theory) appear to be at least Ballesteros, F., and G. Palomero. 2012. Conectividad,
as important as the numerical optimization of demografı́a y conservación del oso pardo cantáb-
resistance values. Indeed, our results suggest that rico. Pages 19–31 in A. San Miguel, F. Ballesteros,
these other aspects substantially impact the J. C. Blanco, and G. Palomero, editors. Manual de
outcome of resistance-based connectivity predic- buenas prácticas para la gestión de corredores
tions, which could be a challenge for applying oseros en la Cordillera Cantábrica. Fundación Oso
resistance-based connectivity models in practical Pardo. Ministerio de Agricultura, Alimentación y
Medio Ambiente, Madrid, Spain.
conservation. If different approaches predict
Beier, P., D. R. Majka, and W. D. Spencer. 2008. Forks in
vastly dissimilar areas to be important for the road: choices in procedures for designing
conservation efforts (e.g., optimal locations for wildland linkages. Conservation Biology 22:836–
corridors or road-crossing structures), it will be 851.
rather difficult to objectively choose and justify a Bowcock, A., A. Ruiz-Linares, J. Tomfohrde, E. Minch,

v www.esajournals.org 12 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

J. Kidd, and L. L. Cavalli-Sforza. 1994. High s10980-014-0056-5

resolution of human evolutionary trees with Gurrutxaga, M., L. Rubio, and S. Saura. 2011. Key
polymorphic microsatellites. Nature 368:455–457. connectors in protected forest area networks and
Chetkiewicz, C.-L. B., C. C. St. Clair, and M. S. Boyce. the impact of highways: A transnational case study
2006. Corridors for conservation: integrating pat- from the Cantabrian Range to the Western Alps
tern and process. Annual Review of Ecology, (SW Europe). Landscape and Urban Planning
Evolution and Systematics 37:317–342. 101:310–320.
Clevenger, A. P., F. J. Purroy, and M. A. Campos. 1997. James, P. M. A., M. J. Forteen, A. Fall, D. Kneeshaw,
Habitat assessment of a relict brown bear (Ursus and C. Messier. 2007. The effects of spatial legacies
arctos) population in northern Spain. Biological following shifting management practices and fire
Conservation 80:17–22. on boreal forest age structure. Ecosystems 10:1261–
Cushman, S. A., M. Chase, and C. Griffin. 2010. 1277.
Mapping landscape resistance to identify corridors Jongman, R. H., I. M. Bouwma, A. Griffioen, L. Jones-
and barriers for elephant movement in southern Walters, and A. M. Van Doorn. 2011. The pan
Africa. Pages 349–367 in S. A. Cushman and F. European ecological network: PEEN. Landscape
Huettmann, editors. Spatial complexity, informat- ecology 26:311–326.
ics, and wildlife conservation. Springer, New York, Kautz, R., R. Kawula, T. Hoctor, J. Comiskey, D.
New York, USA. Jansen, D. Jennings, J. Kasbohm, F. Mazzotti, R.
Cushman, S. A., and J. S. Lewis. 2010. Movement McBride, and L. Richardson. 2006. How much is
behavior explains genetic differentiation in Amer- enough? Landscape-scale conservation for the
ican black bears. Landscape Ecology 25:1613–1625. Florida panther. Biological Conservation 130:118–
Cushman, S. A., K. S. McKelvey, J. Hayden, and M. K. 133.
Schwartz. 2006. Gene flow in complex landscapes: Kuemmerle, T., K. Perzanowski, H. Resit Akcakaya, F.
testing multiple hypotheses with causal modeling. Beaudry, T. R. Van Deelen, I. Parnikoza, P.
American Naturalist 168:486–499. Khoyetskyy, D. M. Waller, and V. C. Radeloff.
Cushman, S. A., B. McRae, F. Adriansen, P. Beier, M. 2011. Cost-effectiveness of strategies to establish a
Shirley, and K. Zeller. 2013a. Biological corridors European bison metapopulation in the Carpathi-
and connectivity. Pages 284–404 in D. McDonald, ans. Journal of Applied Ecology 48:317–329.
editor. Conservation in theory and practice. Wiley, Kuroe, M., N. Yamaguchi, T. Kadoya, and T. Miyashita.
New York, New York, USA. 2011. Matrix heterogeneity affects population size
Cushman, S. A., T. N. Wasserman, E. L. Landguth, and of the harvest mice: Bayesian estimation of matrix
A. J. Shirk. 2013b. Re-evaluating causal modeling resistance and model validation. Oikos 120: 271–
with Mantel tests in landscape genetics. Diversity 279.
5(1):51–72. Landguth, E., S. Cushman, M. Schwartz, K. McKelvey,
Dickson, B. G., J. S. Jenness, and P. Beier. 2005. M. Murphy, and G. Luikart. 2010. Quantifying the
Influence of vegetation, topography, and roads on lag time to detect barriers in landscape genetics.
cougar movement in southern California. Journal Molecular Ecology 19:4179–4191.
of Wildlife Management 69:264–276. Landguth, E., B. Hand, J. Glassy, S. Cushman, and M.
Dahle, B., and J. E. Swenson. 2003. Home ranges in Sawaya. 2012. UNICOR: a species connectivity and
adult Scandinavian brown bears (Ursus arctos): corridor network simulator. Ecography 35:9–14.
effect of mass, sex, reproductive category, popula- Larkin, J. L., D. S. Maehr, T. S. Hoctor, M. A. Orlando,
tion density and habitat type. Journal of Zoology and K. Whitney. 2004. Landscape linkages and
260:329–335. conservation planning for the black bear in west-
Epps, C. W., J. D. Wehausen, V. C. Bleich, S. G. Torres, central Florida. Animal Conservation 7:23–34.
and J. S. Brashares. 2007. Optimizing dispersal and Mateo-Sánchez, M. C., N. Balkenhol, S. Cushman, T.
corridor models using landscape genetics. Journal Pérez, A. Domı́nguez, and S. Saura. In press. A
of Applied Ecology 44:714–724. comparative framework to infer landscape effects
Fahrig, L. 2007. Non-optimal animal movement in on population genetic structure: Are habitat suit-
human-altered landscapes. Functional Ecology ability models effective in explaining gene flow?
21:1003–1015. Landscape Ecology. doi: 10.1007/s10980-015-0194-4
Ferreras, P. 2001. Landscape structure and asymmet- Mateo-Sánchez, M. C., S. A. Cushman, and S. Saura.
rical inter-patch connectivity in a metapopulation 2014a. Scale dependence in habitat selection: the
of the endangered Iberian lynx. Biological Conser- case of the endangered brown bear (Ursus arctos) in
vation 100:125–136. the Cantabrian Range (NW Spain). International
Graves, T., R. B. Chandler, J. A. Royle, P. Beier, and Journal of Geographical Information Science
K. C. Kendall. 2014. Estimating landscape resis- 28:1531–1546.
tance to dispersal. Landscape Ecology. doi: 10.1007/ Mateo-Sánchez, M., S. Cushman, and S. Saura. 2014b.

v www.esajournals.org 13 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Connecting endangered brown bear subpopula- landscape impacts on effective dispersal. Molecular
tions in the Cantabrian Range (north-western Ecology 23: 2146–2156.
Spain). Animal Conservation 17:430–440. Pullinger, M. G., and C. J. Johnson. 2010. Maintaining
McRae, B. H., and P. Beier. 2007. Circuit theory predicts or restoring connectivity of modified landscapes:
gene flow in plant and animal populations. evaluating the leastcost path model with multiple
Proceedings of the National Academy of Sciences sources of ecological information. Landscape Ecol-
104:19885–19890. ogy 25:1547–1560.
McRae, B. H., B. G. Dickson, T. H. Keitt, and V. B. Shah. Rosenberg, D. K., B. R. Noon, and E. C. Meslow. 1997.
2008. Using circuit theory to model connectivity in Biological corridors: form, function, and efficacy.
ecology, evolution, and conservation. Ecology BioScience 47:677–687.
89:2712–2724. Schwartz, C. C., S. D. Miller, and M. A. Haroldson.
Mills, L. S., and F. W. Allendorf. 1996. The one- 2003. Grizzly bear. Pages 556–586 in G. A. Feld-
migrant-per-generation rule in conservation and hammer, B. C. Thompson, and J. A. Chapman,
management. Conservation Biology 10:1509–1518. editors. Wild mammals of North America: biology,
Nathan, R., G. Perry, J. T. Cronin, A. E. Strand, and management and conservation. The John Hopkins
M. L. Cain. 2003. Methods for estimating long- University Press, Baltimore, Maryland, USA.
distance dispersal. Oikos 103:261–273. Segelbacher, G., S. A. Cushman, B. K. Epperson, M.-J.
Naves, J., T. Wiegand, E. Revilla, and M. Delibes. 2003. Fortin, O. Francois, O. J. Hardy, R. Holderegger, P.
Endangered species constrained by natural and Taberlet, L. P. Waits, and S. Manel. 2010. Applica-
human factors: The case of brown bears in tions of landscape genetics in conservation biology:
Northern Spain. Conservation Biology 17:1276– concepts and challenges. Conservation Genetics
1289. 11:375–385.
Nores, C., and J. Naves. 1993. Distribución histórica del Shirk, A., D. Wallin, S. Cushman, C. Rice, and K.
oso pardo en la Penı́nsula Ibérica. Pages 13–33 in J. Warheit. 2010. Inferring landscape effects on gene
Naves and G. Palomero, editors. El oso pardo en flow: a new model selection framework. Molecular
España (Ursus arctos). ICONA, Madrid, Spain. Ecology 19:3603–3619.
O’Brien, D., M. Manseau, A. Fall, and M.-J. Fortin. Simberloff, D., J. A. Farr, J. Cox, and D. W. Mehlman.
2006. Testing the importance of spatial configura- 1992. Movement corridors: conservation bargains
tion of winter habitat for woodland caribou: an or poor investments? Conservation Biology 6:493–
application of graph theory. Biological Conserva- 504.
tion 130:70–83. Singleton, P. H., W. L. Gaines, and J. F. Lehmkuhl.
Pereira, M., P. Segurado, and N. Neves. 2011. Using 2002. Landscape permeability for large carnivores
spatial network structure in landscape manage- in Washington: a geographic information system
ment and planning: a case study with pond turtles. weighted-distance and least-cost corridor assess-
Landscape and Urban Planning 100:67–76. ment. USDA, Forest Service, Pacific Northwest
Pérez, T., J. Naves, F. Vázquez, A. Fernández-Gil, J. Research Station, Portland, Oregon, USA.
Albornoz, E. Revilla, M. Delibes, and A. Domı- Spear, S. F., N. Balkenhol, M.-J. Fortin, B. H. McRae,
nguez. 2014. Estimating the population size of the and K. Scribner. 2010. Use of resistance surfaces for
endangered brown bear through genetic sampling. landscape genetic studies: considerations for pa-
Wildlife Biology. doi: 10.2981/wlb.00069 rameterization and analysis. Molecular Ecology
Pérez, T., F. Vázquez, J. Naves, A. Fernández, A. 19:3576–3591.
Corao, J. Albornoz, and A. Domı́nguez. 2009. Non- Spear, S. F., C. R. Peterson, M. D. Matocq, and A.
invasive genetic study of the endangered Canta- Storfer. 2005. Landscape genetics of the blotched
brian brown bear (Ursus arctos). Conservation tiger salamander (Ambystoma tigrinum melanostic-
genetics 10:291–301. tum). Molecular Ecology 14:2553–2564.
Pérez, T., J. Naves, J. F. Vázquez, J. Seijas, A. Corao, J. Squires, J. R., N. J. DeCesare, L. E. Olson, J. A. Kolbe,
Albornoz, and A. Domı́nguez. 2010. Evidence for M. Hebblewhite, and S. A. Parks. 2013. Combining
improved connectivity between Cantabrian brown resource selection and movement behavior to
bear subpopulations. Ursus 21:104–108. predict corridors for Canada lynx at their southern
Peterman, W. E., G. M. Connette, R. D. Semlitsch, and range periphery. Biological Conservation 157:187–
L. S. Eggert. 2014. Ecological resistance surfaces 195.
predict fine-scale genetic differentiation in a terres- Storfer, A., M. Murphy, J. Evans, C. Goldberg, S.
trial woodland salamander. Molecular Ecology Robinson, S. Spear, R. Dezzani, E. Delmelle, L.
23: 2402–2413. Vierling, and L. Waits. 2007. Putting the ‘landscape’
Pflüger, F. J., and N. Balkenhol. 2014. A plea for in landscape genetics. Heredity 98:128–142.
simultaneously considering matrix quality and Swenson, J. E., F. Sandegren, and A. Derberg. 1998.
local environmental conditions when analysing Geographic expansion of an increasing brown bear

v www.esajournals.org 14 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

population: evidence for presaturation dispersal. Zeller, K. A., K. McGarigal, P. Beier, S. A. Cushman,
Journal of Animal Ecology 67:819–826. T. W. Vickers, and W. M. Boyce. 2014. Sensitivity of
Vucetich, J. A., and T. A. Waite. 2000. Is one migrant landscape resistance estimates based on point
per generation sufficient for the genetic manage- selection functions to scale and behavioral state:
ment of fluctuating populations? Animal Conser-
pumas as a case study. Landscape Ecology 29:541–
vation 3:261–266.
Wasserman, T. N., S. A. Cushman, M. K. Schwartz, and 557.
D. O. Wallin. 2010. Spatial scaling and multi-model Zeller, K. A., K. McGarigal, and A. R. Whiteley. 2012.
inference in landscape genetics: Martes americana in Estimating landscape resistance to movement: a
northern Idaho. Landscape ecology 25:1601–1612. review. Landscape Ecology 27:777–797.

SUPPLEMENTAL MATERIAL
APPENDIX A
DETAILS ON CORRIDOR IDENTIFICATION tween all combinations of sources and destina-
ANALYSES tions.

Least cost path density modeling Circuit theory modeling

UNICOR (UNIversal CORridor and network
Circuitscape (McRae et al. 2008) uses circuit
simulation model, Landguth et al. 2010) iden-
theory to model landscapes as conductance
tifies the shortest path between every specified
surfaces and predict important connections
species location on a landscape to every other
among locations. Based on the assumption of a
specified species location. The combination of
all the movement paths produces a least cost random walk, all plausible paths between each
path density map that represents the pattern of two locations are integrated creating a current
the most probable movement paths for brown map. Analogously to the factorial least cost path
bear in the study area (e.g., Cushman et al. density map described above, the combination of
2013a; Mateo-Sánchez et al. 2014b). To identify every pairwise current map allows the identify
putative corridors, we computed the focal corridors with higher current flow.
density of the factorial least cost path network This software also computes pairwise effective
with a moving window of 1-km radius with a resistance between all combinations of locations
GIS (Mateo-Sánchez et al. 2014b). UNICOR also (here brown bear occurrences). In this study we
calculated effective (least cost) distance be- used a four-neighbor case for the calculations.

APPENDIX B
DETAILS ON THE EXAMPLE OF CIRCUITSCAPE centre to centre effective resistance was lower
ESTIMATES OF EFFECTIVE RESISTANCES than edge to edge (see Fig. B1), confirming our
findings from the brown bear data set. This
The simulated example consists of two habitat suggests that there might be a previously
areas with low resistance to movement that are
unreported issue related to some conceptual
separated by a high resistance matrix. We
computed circuit-based effective resistances be- aspects of circuit theory or its application to
tween two points located in the center of both model landscape connectivity, or issues related to
habitat areas and two points located on their details of the implementation of the calculations
closest edges (Fig. B1). Results showed that in Circuitscape.

v www.esajournals.org 15 April 2015 v Volume 6(4) v Article 59

 MATEO-SÁNCHEZ ET AL.

Fig. B1. Simulated landscape for illustrating the results of the effective resistances through circuit theory as
implemented in Circuitscape. The landscape consists of two habitat areas with low resistance to movement
(resistance score 10) that are separated by a portion of landscape matrix with a high resistance (resistance score
100). Effective resistance was calculated (1) between two points located in the center of both habitat areas (shown
in magenta in the figure), resulting in 137.1 and 57.5 Ohms, respectively, for a four- and eight-neighbors case and
(2) between two points located in the habitat area edges (border of one habitat area closest to the other habitat
area, shown in blue in the figure), resulting in 162.3 and 69.84 Ohms, respectively, for a four- and eight-neighbors
case.

v www.esajournals.org 16 April 2015 v Volume 6(4) v Article 59