Content uploaded by Santiago Saura
Author content
All content in this area was uploaded by Santiago Saura on Dec 23, 2015
Content may be subject to copyright.
RESEARCH ARTICLE
Seasonal and temporal changes in species use
of the landscape: how do they impact the inferences
from multi-scale habitat modeling?
Marı
´a C. Mateo-Sa
´nchez .Aitor Gasto
´n.Carlos Ciudad .
Juan I. Garcı
´a-Vin
˜as .Jorge Cuevas .Ce
´sar Lo
´pez-Leiva .
Alfredo Ferna
´ndez-Landa .Nur Algeet-Abarquero .Miguel Marchamalo .
Marie-Jose
´e Fortin .Santiago Saura
Received: 14 May 2015 / Accepted: 8 December 2015
ÓSpringer Science+Business Media Dordrecht 2015
Abstract
Context Multi-scale approaches to habitat modeling
have been shown to provide more accurate under-
standing and predictions of species-habitat associa-
tions. It remains however unexplored how spatial and
temporal variations in habitat use may affect multi-
scale habitat modeling.
Objectives We aimed at assessing how seasonal and
temporal differences in species habitat use and distri-
bution impact operational scales, variable influence,
habitat suitability spatial patterns, and performance of
multi-scale models.
Methods We evaluated the environmental factors
driving brown bear habitat relationships in the Cantab-
rian Range (Spain) based on species presence records
(ground observations) for the period 2000–2010,
LiDAR data on forest structure, and seasonal estimates
of foraging resources. We separately developed multi-
scale habitatmodels for (i) each season (spring,summer,
fall and winter) (ii) two sub-periods with different
population status: 2000–2004 (with brown bear distri-
bution restricted to the main population nuclei) and
2005–2010 (with expanding bear population and range);
and (iii) the entire 2000–2010 period.
Results Scales of effect remained considerably
stable across seasonal and temporal variations, but
not the influence of certain environmental variables.
The predictive ability of multi-scale models was lower
in the seasons or periods in which populations used
larger areas and a broader variety of environmental
conditions. Seasonal estimates of foraging resources,
together with LiDAR data, appeared to improve the
performance of multi-scale habitat models.
Conclusions We highlight that the understanding of
multi-scale behavioral responses of species to spatial
Special issue: Multi-scale habitat modeling.
Guest Editors: K. McGarigal and S. A. Cushman.
M. C. Mateo-Sa
´nchez (&)A. Gasto
´n
C. Ciudad J. I. Garcı
´a-Vin
˜as J. Cuevas
C. Lo
´pez-Leiva S. Saura
ECOGESFOR Research Group, E.T.S.I Montes, Forestal
y del Medio Natural, Technical University of Madrid,
Ciudad Universitaria s/n, 28040 Madrid, Spain
e-mail: mc.mateo@upm.es
A. Ferna
´ndez-Landa
Agresta Cooperative Society, C/Duque Ferna
´nNu
´n
˜ez 2,
28012 Madrid, Spain
N. Algeet-Abarquero M. Marchamalo
Department of Land Morphology and Engineering,
Technical University of Madrid, ETSI Caminos, Canales
y Puertos, c/Profesor Aranguren s/n, 28040 Madrid, Spain
M.-J. Fortin
Department of Ecology and Evolutionary Biology,
University of Toronto, 25 Harbord Street, Toronto,
ON M5S 3G5, Canada
123
Landscape Ecol
DOI 10.1007/s10980-015-0324-z
patterns that continually shift over time may be
essential to unravel habitat relationships and produce
reliable estimates of species distributions.
Keywords Multi-scale habitat modeling
Operational scale Habitat seasonality Foraging
resource Population size Brown bear
Introduction
The sensitive dependence of ecological processes to
drivers acting across a range of scales has been
constantly revisited from the most theoretical per-
spectives of landscape ecology (Urban 1987; Levin
1992) to the most practical and recent ecological
modeling (Vicente et al. 2014). There is a consensus in
literature that species respond to habitat at different
scales (Johnson 1980; Schaefer and Messier 1995;
Rettie and Messier 2000; Grand et al. 2004; Fortin
et al. 2012), yet many studies have usually considered
this response only at a single spatial scale (e.g. Holland
and Fahrig 2000; Triantis et al. 2003; Posillico et al.
2004). Spatial scale may refer to the grain size at
which the environmental variables are mapped, to the
spatial extent at which the effect of the environmental
variable is quantified around a given location (typi-
cally through a mean value of the variable over the
considered spatial extent), or to a combination of both.
When characterizing species-habitat relationships,
it is critical not only to identify the correct drivers but
also the scale (spatial extent) at which they have their
strongest effect (hereafter referred as operational
scale). Recent studies (e.g. Grand et al. 2004; Holland
et al. 2004; Shirk et al. 2012; Wasserman et al. 2012;
Weaver et al. 2012; Mateo-Sa
´nchez et al. 2014) have
supported this point and demonstrated that multi-scale
approaches to habitat modeling yield improved pre-
dictions of species occurrence. Multi-scale habitat
modeling may capture dimensions that have been
ignored by usual approaches conducted at single scale
(Pearson et al. 2004). Failure to address the variability
of species responses with scale can strongly affect
inferred habitat relationships and lead to a misinter-
pretation of interactions between species and the
environment with potentially large conservation
implications (Bradley et al. 2010; Shirk et al. 2012).
Several authors have emphasized the importance of
analytical processes aimed to select the operational
scale for each predictor when modeling habitat
selection (Kotliar and Wiens 1990; Holland et al.
2004; Wasserman et al. 2012). One advocated method
to estimate the spatial scale at which a species
perceives a particular landscape variable is to model
the species-habitat relationship at a number of scales
and determine which scale fits the model best (Sav-
ignac et al. 2000; Wasserman et al. 2012; Mateo-
Sa
´nchez et al. 2014). However, multi-scale habitat
models so far have mainly focused on a single period
of time with a particular distribution of individuals
given by the presence data that usually feed these
models. Some authors have suggested that operational
scales may be influenced by species traits such as
space use (i.e., home ranges in mammals; see Jackson
and Fahrig 2012) or perceptual abilities (Zollner
2000). There are evidences that the species space use
can vary in relation to seasonal variation in food
availability (Anderson et al. 2005), breeding (Pope
et al. 2000), population size and status (i.e., growing or
decreasing population; McFarland et al. 2014) or other
factors (Kie et al. 2002; Anderson et al. 2005).
However, it remains yet unexplored how seasonal or
temporal changes in species distributions may affect
the operational scales, inferences and habitat suitabil-
ity patterns of multi-scale habitat modeling.
Our aim in this study was to address this gap in
knowledge by assessing the potential impact of
seasonal and temporal changes on multi-scale model
performance and on the ecological understanding of
multi-scale species responses and distribution drivers.
We focus on the effect of scale defined as the spatial
extent at which the measured environmental variables
are perceived by species, while keeping the grain size
constant. We investigate brown bear (Ursus arctos)
habitat relationships in the Cantabrian Range (NW
Spain) as the focal species for this study for different
reasons. First, brown bear is one of the most endan-
gered mammals in Spain. Even when recent studies
(Ballesteros and Palomero 2012) reveal that the
population is growing, limited habitat availability
has been identified as one of the major conservation
concerns and potential constraints for the long-term
persistence of this species. Second, brown bear has, as
other large mammals, broad spatial requirements with
variable pattern of space use due to seasonality
(Bo
¨rger et al. 2006), which, together with the recent
population expansion in the Cantabrian Range, makes
this species a good model organism for the objectives
Landscape Ecol
123
of this study. Third, brown bear in the Cantabrian
Range counts with a large long-term occurrence data
set and its habitat requirements have been broadly
studied (e.g. Clevenger et al. 1992; Naves and
Palomero 1993; Naves et al. 2003; Mateo-Sa
´nchez
et al. 2014), providing valuable support on under-
standing species-habitat relationship. Furthermore, we
were able to estimate, based on LiDAR data, forest
shelter as well as seasonal foraging resources for
brown bears over the &35,700 km
2
study area. All
these factors offer a unique opportunity to assess the
variability of the scale (spatial extent) of habitat
relationships along temporal and spatial processes,
and to test how multi-scale habitat models predictions
change over the seasons and different time periods.
We addressed our general objective through two
specific analyses. First, we separately considered the
seasonal (spring, summer, fall and winter) brown bear
presence records and foraging resources for the period
2000–2010, in order to gain insights on how seasonal
habitat use may affect the operational scales that are
identified through multi-scale habitat modeling as well
as the predictions of species occurrence. Then, we
investigated how two different multi-annual periods
(2000–2004 vs. 2005–2010), with distinctive popula-
tion size and status, may affect the extents at which
individuals perceive and use the landscape and the
ability of multi-scale models to predict species
occurrence. In both cases, the results of these analyses
were compared to those of a conventional (year-
round) multi-scale habitat model for the entire period
2000–2010.
Methods
Study area
The study was carried out in a &35,700 km
2
area
located in the Cantabrian Range (NW Spain, 42°510N,
5°340W) (Fig. 1). This area comprises the whole
known range of the brown bear native populations in
the Iberian Peninsula, and its peripheral zones where
future population expansion may be likely (buffer of
*25 km around current species range). The study
region has a complex topography (elevations ranging
from sea level to 2648 m) and a humid Atlantic
climate with mild temperatures and short summers.
According to the forest map of Spain (FMS) the
landscape is composed by a mosaic of forests (39 %),
shrublands (24 %), natural grasslands (3 %), rocky
areas (2 %), croplands (22 %) and other artificial or
highly modified areas (10 %). Forests are dominated
by deciduous oaks (Quercus robur and Q. petraea),
beeches (Fagus sylvatica) and chestnuts (Castanea
sativa) mainly in northward slopes of the range.
Extensive forests of semi-deciduous oaks (Q. pyre-
naica and Q. faginea) and also evergreen oaks (Q. ilex)
are found on southern slopes. Pinus sp. and Eucalyptus
globulus are also relatively abundant in the area. A
large variety of shrubland species is found in the
Cantabrian Range (Erica sp., Calluna vulgaris,Vac-
cinium myrtillus,Crataegus sp., Rubus sp., Prunus
spinosa,Rosa sp., Cytisus sp., Genista sp., Ulex sp.,
Rhamnus alpina, etc.). Some parts of the study area
have low human population density, whereas others
have experienced extensive urban and agriculture
development connected by a network of transport
infrastructures. As a result of the extensive and long-
lasting anthropogenic activities in this region, the
original forest landscape has suffered significant
modifications.
Species presence data
We used available brown bear presence records for the
period 2000–2010 collected in the Cantabrian Range
by trained observers and rangers. Presence records
were obtained through direct observation and indirect
evidence (feces, footprints, hair, beehives attacks); the
diversity of presence data sources might allow a
comprehensive representation of brown bear occur-
rence patterns, particularly given that the same type of
records is used for comparing the different seasons and
time periods. We reformatted all these occurrence data
and integrated them to create a complete presence-
only dataset. Locations were resampled with 1 ha cell
size to summarize potential locations of the same
individual in consecutive days and to allow compu-
tational efficiency. This is the same cell size used in
previous studies for this species (Mateo-Sa
´nchez et al.
2014). We obtained a total number of 6207 locations
for the entire period 2000–2010 that were used in
subsequent analyses. The same procedure was fol-
lowed to obtain bear locations separately for each
season: spring (2034 locations), summer (2226 loca-
tions), fall (1316 locations) and winter (844 locations).
We considered a non-hibernation hypothesis for
Landscape Ecol
123
winter since there are evidences that in the Cantabrian
Range bears reduce only partially, and not completely,
their activity in winter (Palomero 1995; Naves et al.
2001; Palomero et al. 2011). To assess the effects of
temporal changes in population size on habitat use, we
subdivided and summarized, following the same
procedure, the bear locations in two multi-annual
periods: a ‘‘constrained’’ period (2000–2004, 1746
locations) with a smaller population size of about 80
individuals (Wiegand et al. 1998; Naves et al. 1999)
and a ‘‘growth’’ period (2005–2010, 4744 locations)
with a noticeable population increase up to about 200
individuals in 2011 (Ballesteros and Palomero 2012),
i.e. more than doubling the previously estimated
population size.
Overview of environmental data and assessed
scales
We used environmental variables related to foraging
resources, shelter and human pressure that have shown
to be suitable predictors of brown bear occurrence
Fig. 1 Location of the study area and brown bear range (gray squares) in the Cantabrian Range. Dashed rectangle indicates a focused
area shown in Figs. 2and 3. The names of the Spanish provinces are indicated
Landscape Ecol
123
(Clevenger et al. 1997; Naves et al. 2003; Garcı
´a et al.
2007; Koren
ˇet al. 2011; Mateo-Sa
´nchez et al. 2014).
A set of 12 variables (Table 1) was referenced to a
UTM projection (ETRS89, zone 30) and resampled
with 1 ha cell size. To generate raster layers for
multiple spatial extents, we used circular moving
windows around each location with radii 0.25, 0.5, 1,
2, 4, 8, 16, 32 and 64 km as the scales for each variable
(see Mateo-Sa
´nchez et al. 2014 for details). This range
of radii covers the species
´responses to environmental
gradients across all the scales (i.e. extents) relevant for
meeting its ecological requirements, from resources
within habitat patches to the extent of home ranges and
broader landscape-scale impacts. We used ArcGIS
10.1 (ESRI) for calculations.
LiDAR data
LiDAR information was provided by the Spanish
National Plan for Aerial Orthophotography (PNOA;
Ministerio de Fomento 2015) with a mean density of
0.5 points per m
2
and vertical root mean square error
(RMSE) B0.15 m. A total of 10,544 tiles (2 92 km),
weighting 1.3 Tb, were processed with FUSION
software (McGaughey and Carson 2003). A prede-
fined height of 3.5 m was used as the threshold to
separate trees from understory vegetation. Forest
height was estimated as the 95th percentile of
vegetation height from the aboveground vegetation
returns ([3.5 m from the ground), i.e. of laser pulses
intercepting vegetation and returning to the sensor.
Forest canopy cover was computed as the ratio
between the number of first returns above 3.5 m and
the total number of first returns (a laser pulse may
intercept several vegetation targets from the top of the
tree to branches or understory vegetation; only the first
return was considered).
Foraging resources
Foraging resources for brown bears in each 1 ha cell
were estimated for each particular season from LiDAR
data, the FMS at a 1:50,000 scale (developed in
coordination with the Third Spanish National Forest
Inventory), and specific plant species abundance
models (Table 1). We estimated foraging resources
from the combination of those plant species (trees,
shrubs and herbs) which sequentially provide food
supply for brown bears along the different seasons
(Marquı
´nez 2002; Palomero et al. 2011). Main forage
resources comprise herbaceous plants in spring
(mostly hydrophylous tall herbs). Fleshy fruits and
berries largely provided by shrublands (species such as
V. myrtillus,Crataegus sp., Rubus sp., P. spinosa and
R. alpine) seem to be especially important for bears. In
fall and winter acorns, beech nuts and chestnuts are the
fundamental component of brown bear diet. The
abundance and importance of the resources provided
by each plant species were both accounted for, as in
Nielsen et al. (2010). The importance of each foraging
resource was determined by previous seasonal scat-
analysis for brown bears in the study area (Delibes
1999; Fernandez-Calvo et al. 2001; Marquı
´nez 2002;
Ballesteros et al. 2012). The abundance of tree species
was obtained from FMS (species identity and occupied
area in a given map polygon) and LiDAR data (canopy
cover within the occupied area). The abundance of
non-tree species was estimated through the informa-
tion in available floristic inventories, ecological niche
modeling (penalized logistic regression) based on
previously available species distribution models that
predict occurrence probability in each cell of the study
area for each shrub species using climate and lithology
(Gasto
´n and Garcı
´a-Vin
˜as 2011), and expert knowl-
edge on the compatibility of the presence of particular
plant species within plant communities mapped all
throughout the study area by FMS. Only plant
resources were considered, due to the lack of infor-
mation for other foraging resources (e.g. wild ungu-
lates, insects). In any case, we believe this variable is a
good estimator of foraging resources given that plant
matter represents the most important food for brown
bears (Clevenger et al. 1992; Naves et al. 2006; Ciucci
et al. 2014). Year-round map of foraging resources
was estimated as the sum of the seasonal foraging
resources. We calculated the focal mean (i.e. mean
value for each cell location within a specific neigh-
borhood depending on the extent of measurement) of
seasonal and total year-round foraging resources at the
nine scales mentioned above.
Shelter
Brown bears have high dependence on forest cover
(Clevenger et al. 1992). Beyond food availability,
previous studies have highlighted the importance of
forest and shrublands in term of shelters and habitat
continuity (Go
´mez-Manzanedo et al. 2012; Mateo-
Landscape Ecol
123
Sa
´nchez et al. 2014) as they provide cover for roosting,
escaping dangers, rearing young, and loafing. In
addition, in these forest areas human pressures are
fewer Palomero et al. 2011). Therefore, we used
variables of forest and shrubland structure as potential
surrogates for brown bears shelter (Table 1). LiDAR
data provided estimates of forest canopy cover and
forest height. We calculated the focal mean of the
previous variables across the nine assessed scales. In
order to evaluate the effect of connectedness of forest
and shrubland cover, we also calculated the cohesion
index for these two cover types in FRAGSTATS 4.2
(McGarigal et al. 2012). Forest area was obtained from
LiDAR data (forest canopy cover C30 %). Cohesion
of forest and cohesion of shrubland were evaluated
only for seven extents (excluding 32 and 64 km) to
reduce ‘‘boundary effects’’; the cohesion index cannot
be calculated for cells located at a distance from the
edge of the study area smaller than the considered
extent, which would have excluded a large proportion
of the study area (including locations with significant
densities of brown bear presence records) for these
large extents of 32 and 64 km; see McGarigal et al.
(2012).
Human pressure
We included density of buildings and transport
infrastructures as predictors of human pressure
(Table 1). Buildings were extracted from a topo-
graphic map developed by Spanish Geographic
National Institute (CNIG) and transport infrastructures
from open street map (OSM, www.openstreemap.
org). We divided transport infrastructures in highways
and conventional roads to try to assess potential dif-
ferent effects depending on traffic volume and physi-
cal restrictions on bear distribution. The density of
these elements was evaluated at the proposed nine
scales.
Modelling seasonal changes in habitat use
First, univariate models were performed to identify in
every season (astronomical seasons were used) the
operational scale, i.e. the spatial extent at which each
environmental variable was most strongly related to
brown bear occurrence (e.g., Grand et al. 2004; Mateo-
Sa
´nchez et al. 2014). We used lrm and pentrace
functions from the rms package (Harrell 2014) in the R
Table 1 Environmental variables used for analyzing brown bear habitat suitability
Metric Description Data source Unit
Foraging resources
SpFR Focal mean of spring foraging resources LiDAR ?FMS ?specific models %
SuFR Focal mean of summer foraging resources LiDAR ?FMS ?specific models %
AFR Focal mean of autumn foraging resources LiDAR ?FMS ?specific models %
WFR Focal mean of winter foraging resources LiDAR ?FMS ?specific models %
TotFR Focal mean of total (year-round) foraging
resources (sum of foraging resources in
each season)
LiDAR ?FMS ?specific models %
Shelter
FCC Focal mean of forest canopy cover LiDAR %
FHei Focal mean of forest height LiDAR m
ForCI Cohesion index of forests LiDAR Dimensionless
ShrubCI Cohesion index of shrubland FMS Dimensionless
Human pressure
BuildDens Building density CNIG Buildings/km
2
HwDens Highways density OSM km/km
2
RoadDens Conventional roads density OSM km/km
2
Focal mean is the mean value of each variable within a specific neighborhood (depending on the scale of measurement) around each
location. Foraging resources were expressed as percentage of the maximum possible resources in a location
FMS forest map of Spain, CNIG Spanish Geographic National Institute, OSM open street map
Landscape Ecol
123
environment for statistical computing (R Core Team
2014) to fit penalized logistic regression models
(Harrell 2001) with each predictor as a single linear
term. For every season we chose the operational scale
(i.e., scale with best performance) of each predictor
using Akaike’s Information Criterion (AIC; Johnson
and Omland 2004). Twenty thousand random pseudo-
absence points were drawn inside limits of the study
area (cell size =1 ha). The pseudo-absences were the
same for every single model.
Secondly, multi-scale models (e.g. Grand et al.
2004; Shirk et al. 2012; Wasserman et al. 2012) were
developed to assess the effect of each predictor in
brown bear distribution. According to the results of the
univariate models, we developed for each season a
model combining environmental variables at their
specific operational scales. For a meaningful compar-
ison and understanding of changes in habitat use
between seasons, we used in all the cases the full set of
explanatory variables. Except for seasonal changes in
foraging resources, explanatory variables were treated
as static across seasons. We developed penalized
logistic regression models with linear terms only and
without interactions among predictors. We used lrm
and pentrace functions to fit the logistic regressions
models. We assessed the predictive performance of
multi-scale models using the area under the receiver
operating characteristic curve (AUC; Fielding and
Bell 1997) estimated with ten-fold cross-validation.
As the main objective of these models is to rank sites
according to the suitability for the species, AUC is a
good option because it measures discrimination, i.e.,
the probability that the predicted value is higher in a
presence point than in a pseudo-absence point, both
picked at random.
Modelling temporal changes in habitat use
Data analysis followed the same procedures and set of
explanatory variables as for seasonal models (i.e.,
univariate models to select the operational scale of
each predictor and multi-scale models to assess
predictor influence). We separately used locations of
the two periods 2000–2004 and 2005–2010 (instead of
seasonal locations), as well as locations for the entire
period 2000–2010, and used in all cases the same set of
explanatory variables. We used year-round foraging
resources (foraging resources were not considered
separately for each season in this multi-annual
analysis). Similarly, we considered that variables did
not change over the studied period, since important
land-use changes have not happened in the study area
in the assessed ten-year period, particularly in com-
parison with the sharp increase experienced by the
bear population, which is mainly the result of its
protection status for the last three decades and related
changes in human attitudes towards bears.
Results
Seasonal changes in habitat use
The results of the univariate models showed a similar
pattern of operational scales across seasons (Table 2).
Environmental variables related to foraging resources
affected bear habitat suitability at broad scales (16 km
in most cases), especially during fall. Variables related
to human pressure also affected habitat suitability at
broad scales. However, potential predictors of bear
shelter showed more discrepancy: shrubland cohesion
(ShrubCI) presented the highest AIC values at a large
extent (16 km), while the highest AIC for forest
predictors was mainly found at a fine or medium scale
(1 km in most cases). All the set of evaluated
multivariate models showed good performance, with
AUC ranging from 0.897 to 0.938. The AUC values
were higher (p \0.001) in fall and winter than in
spring and summer (Table 3). As we expected,
independent variables related to human pressure had
a negative effect on bear habitat suitability (Table 3).
Building density had an important negative effect in
all seasons, especially in spring. Road density did not
have a significant effect in spring and fall, but it
affected negatively in summer and winter. Highway
density had a more negative effect on habitat use than
road density (higher coefficients in absolute value),
but it did not influence habitat suitability in winter.
Habitat suitability was significantly related to foraging
resources in all seasons (Table 3). Variables related to
bear shelter had a general positive effect on bear
habitat suitability, particularly shrubland cohesion
(high positive coefficients), but forest height did not
influence significantly habitat use in summer and
winter. In like manner, seasonal changes were also
noticeable in the spatial pattern of habitat suitability
(Fig. 2). Habitat suitability showed the most extensive
Landscape Ecol
123
pattern in spring and summer, and the most local
pattern in winter.
Temporal changes in habitat use
Selected operational scales followed the same trend
for both partial periods (2000–2004 and 2005–2010)
and for the entire 2000–2010 period and broadly
coincided with seasonal patterns (Table 2). Foraging
resources affected bear habitat suitability at broad
scales (16 km). Human pressure variables also
affected habitat suitability at large extents (16 km).
Shelter variables influenced at broad scales in some
cases (ShrubCI, ForCI) and at fine scales (0.5–1 km)
in others (FCC, FHei).
All the evaluated multivariate models showed good
model performance (AUC ranging from 0.900 to
0.947) (Table 4). AUC values were remarkably higher
(p \0.001) in the ‘‘constrained’’ period (2000–2004)
than in the ‘‘growth’’ (population expansion) period
(2005–2010). The synoptic pattern of habitat suitabil-
ity was also wider in the second period of population
Table 2 Operational scale (km) for each environmental variable and season/period used for analyzing brown bear habitat suitability
Predictor Spring Summer Fall Winter Year-round
(2000–2010)
2000–2004 2005–2010
SpFR 16
SuFR 16
AFR 32
WFR 16
TotFR 16 16 16
FCC 1 0.5 1 1 111
FHei 0.5 0.5 0.5 0.5 0.5 0.5 0.5
ForCI 16 16 4 1 16 16 16
ShrubCI 16 16 16 16 16 16 16
BuildDens 16 16 16 16 16 16 16
HwDens 16 16 16 16 16 16 16
RoadDens 16 16 16 16 16 16 16
Seasonal results correspond to the entire period 2000–2010. See Table 1for variable description
Table 3 Seasonal changes in habitat use as inferred from seasonal presence records for the entire period 2000–2010
Predictor Spring Summer Fall Winter Total (year round)
Scale Coef Scale Coef Scale Coef Scale Coef Scale Coef
SFR 16 0.337*** 16 0.744*** 32 0.697*** 16 1.108*** 16 1.165***
FCC 1 0.305*** 0.5 0.332*** 1 0.312*** 1 0.357*** 1 0.163***
FHei 0.5 0.129* 0.5 -0.017 0.5 0.286*** 0.5 0.107 0.5 0.291***
ForCI 16 0.768*** 16 0.494*** 4 0.829*** 1 0.559*** 16 0.179***
ShrubCI 16 1.341*** 16 1.759*** 16 2.729*** 16 2.833*** 16 2.156***
BuildDens 16 -2.269*** 16 -1.717*** 16 -1.694*** 16 -1.48*** 16 -1.171***
HwDens 16 -0.616*** 16 -0.51*** 16 -1.429*** 16 0.079 16 -0.616***
RoadDens 16 -0.045 16 -0.196* 16 -0.205 16 -0.883*** 16 0.051
AUC 0.897 0.901 0.928 0.938 0.910
See Table 1for variable description. SFR corresponds to seasonal foraging resource and varies depending of the season/period
analyzed
Significance levels: *** \0.001, ** \0.01, * \0.05
Landscape Ecol
123
expansion and more concentrated in core areas during
the first period with smaller population size (Fig. 3).
Building density had an important negative effect
during all seasons, especially in the ‘‘constrained’’
period (Table 4). Highway density had a significant
negative effect only in the ‘‘growth’’ period. Habitat
suitability was significantly related to foraging
resources during all the studied periods (Table 4).
Potential predictors of bear shelter also had a positive
effect on habitat use, especially shrubland cohesion.
Discussion
Robustness in operational scales across seasons
and time periods
Distribution of species is driven by processes linked to
several levels of ecological complexity, and therefore
expressed at different scales (Vicente et al. 2011).
Previous research (Mateo-Sa
´nchez et al. 2014)
showed that explicitly optimizing the scale of habitat
suitability models considerably improved single-scale
modeling for brown bears in terms of (i) model
performance and spatial prediction and (ii) avoiding
an oversimplification and lack of appreciation of
details that were blurred or missed in habitat models
conducted at a single scale. Our results support the
conclusion by Mateo-Sa
´nchez et al. (2014) that brown
bears respond to environmental factors at different
scales, but provide further new insights into the
influence of processes underlying temporal and spatial
variations in habitat use. We found that operational
scales were scarcely affected by the different ways in
which bears use habitat due to seasonal changes or
population status. This suggests that bears may have a
scale of perception of the environment not mediated
by changes in species distribution or trends, despite the
fact that several authors have related operative scales
to movement ranges of the organism (Wiens 1989;
Fig. 2 Habitat suitability maps provided by the optimized multi-scale model for every season: aspring, bsummer, cfall, dwinter and
eyear-round. Seasonal and year-round results correspond to the entire period 2000–2010. See Fig. 1for the location of the area here
shown
Landscape Ecol
123
Vos et al. 2001; Dungan et al. 2002). We did not find
evidence that different activity patterns potentially due
to environmental conditions, population trends, nutri-
tional or reproductive status impacted on operative
scales that rule habitat relationships. A plausible
explanation might be that in brown bears the extensive
seasonal movements (such as mating or dispersal
movements beyond currently occupied areas) are
mostly undertaken by males while females are typi-
cally philopatric (Swenson et al. 1998), while in our
study the available presence records did not allow for a
differentiation of the genders. More detailed informa-
tion (e.g., age and gender of the individuals in the
species occurrence records, tracking telemetry) and
additional research may be necessary to shed further
light on the influence of the habitat use during these
movements on operational scales. It has to be noted,
however, that so far there has been very little empirical
evidence of the effects of home range size or dispersal
distance on the operational scales (Desrochers et al.
2010; Fisher et al. 2011). In any case we recognize that
these aspects might be crucial when selecting the
range of scales at which landscape variables should be
measured to identify accurate and biologically-justi-
fied operational scales.
Additionally, for periods when habitat use is more
limited (i.e., winter or period with smaller population
size) and thus local selection could be expected to be
stronger (more restrictive), our results showed than
brown bears also have a large-scale perception of
Table 4 Temporal changes
in habitat use
See Table 1for variable
description
Significance levels:
*** \0.001, ** \0.01,
*\0.05
Predictor 2000–2004 2005–2010 Total (2000–2010)
Scale Coef Scale Coef Scale Coef
TotFR 16 1.607*** 16 0.978*** 16 1.165***
FCC 1 0.177* 1 0.139** 1 0.163***
FHei 0.5 0.243*** 0.5 0.314*** 0.5 0.291***
ForCI 16 0.208* 16 0.182*** 16 0.179***
ShrubCI 16 3.457*** 16 2.003*** 16 2.156***
BuildDens 16 -2.996*** 16 -1.083*** 16 -1.171***
HwDens 16 0.200 16 -0.704*** 16 -0.616***
RoadDens 16 0.489*** 16 0.749 16 0.051
AUC 0.947 0.900 0.910
Fig. 3 Habitat suitability maps provided by the optimized multi-scale model for every period: a2000–2004, b2005–2010 and
c2000–2010. Map (c) in this figure is the same as map (e) in Fig. 2. See Fig. 1for the location of the area here shown
Landscape Ecol
123
factors related to habitat configuration, food resource
and human disturbance, similar to other periods when
habitat use is more extensive. Whereas other factors
related to forest characteristics (such as forest cover
and height) were always supported at the finest scales
regardless the season or population status. This result
suggests that both large-scale and fine-scale factors
contribute to determine, although to different degrees,
brown bear perception and use of habitat.
Some similar dominant drivers of brown bear
habitat suitability across seasons and time periods
Landscape configuration together with human pres-
sure and food availability appeared to be dominant
drivers for brown bear habitat selection whatever the
period of the year and the status/size of the population,
suggesting the importance of large extents of forest
cover with low human footprint. Particularly, our
results highlighted the importance of landscape with
extent of connected shrubland at broad scales without
seasonal or temporal differences, even showing a
stronger relationship with bear occurrence than for-
aging resources and cohesion of forest areas. This may
suggest the importance of shrubland as linkage areas
through which bears can move between other habitat
patches and their potential contribution to habitat
suitability in terms of shelter, matrix permeability and
habitat continuity.
Human pressure: building density more influential
than roads
Medium and large mammals generally suffer strong
human pressures (Cardillo et al. 2005) and therefore
they are more abundant in areas with low human
footprint (Woodroffe 2000). In our case study, brown
bears showed a marked avoidance of predictors of
human pressure at broad scales. Interestingly, these
predictors did not contribute to the multi-scale models
prediction in the same way. The density of buildings
was the most influential variable in the group and
showed a strong negative relationship with bear
habitat. This was also consistent with previous habitat
studies (Naves et al. 2003; Apps et al. 2004; Mateo-
Sa
´nchez et al. 2014), but our results furtherly showed
that the avoidance is similar for all the seasons and
periods with different population size. This suggests
that brown bear strongly avoids areas in the landscape
to a considerable distance (16 km) away from dense
human settlements.
However, linear infrastructures provided less pre-
dictive ability and different contribution according to
the season and size of the population. Highways
showed much higher effect than roads; the latter one
only had significant effect in summer and winter. This
fact may be related to the species space requirements.
Summer and winter are two seasons with a lower
activity and smaller home ranges (Huber and Roth
1993), thus individuals may select more localized
areas. This aspect is also illustrated by the spatial
pattern of habitat suitability, with optimal areas
constrained to a comparatively lower number of high
quality patches in the core areas in winter in compar-
ison to a wider range of suitable areas in seasons with a
broader habitat use (i.e., spring) and to the year-round
habitat use (Fig. 2). Interestingly, highways did not
influence multi-scale model prediction in winter
whereas conventional roads did. This result may be
due to the fact that the more extensive network of
conventional roads may affect bear habitat require-
ments, whereas highways are scarcer and their spatial
distribution in the landscape may not influence the
more remote and localized used by bears in that season.
The same trend was observed in the first analyzed
period with smaller population size in which individ-
uals were constrained to the areas with best habitat
quality and low human influence, as is also reflected in
the habitat suitability spatial pattern (Fig. 3a).
Food resource as a key factor in habitat use
Landscape composition was mostly introduced in the
models in terms of food availability since it has been
advocated as pervasive predictors of animal move-
ment, space use and habitat selection (Fretwell and
Lucas 1970; Fryxell et al. 2004; Ciudad et al. 2009;
Van Beest et al. 2011). Food resource added predictive
power to all the multi-scale models and showed a
higher contribution in year-round models, no matter
the status and distribution of the population, suggesting
that food resources are critical regulating factors
affecting individual growth and population density
(Mattson et al. 2004). This finding shows that bears are
well suited to forage-based definitions of habitat
quality (Nielsen et al. 2010). Seasonal food availability
is particularly important for bears due to high nutri-
tional demands in periods such as hyperphagia (i.e.,
Landscape Ecol
123
last part of summer and fall when individuals accu-
mulate fat to survive winter; Berland et al. 2008) and
during spring, particularly for females with cubs (Rode
and Robbins 2000). However, the contribution of food
resource to seasonal multi-scale models was variable
along the seasons and did not show this pattern. Food
resources had also a stronger contribution to the multi-
scale seasonal models in summer and winter (see
Table 3). This furtherly supports our conclusion that
due to the more reduced (localized) use of landscape in
these seasons, brown bear occurrence would be
associated with more specialized habitat requirements
and therefore distributed into more localized highest
quality spots (Fig. 2d). Suitable habitat is restricted to
places with forage supplies and shelter, more scarce in
winter (Fig. 2d) and more spread in summer (Fig. 2b).
However, the extent of suitable habitat increased in the
other seasons (e.g. spring) in which individuals
required to use a wider variety of habitat resources as
shown in Fig. 2a. In other words, the species may
exhibit more generalist traits, as shown for other
species, under environmental or density changes
(Colles et al. 2009; Barnagaud et al. 2011) and seasonal
food availability might lose predictive ability. Even in
periods in which foraging is crucial, such as hyper-
phagia, and in spring, the broader use of resources
might reduce the inferred strength of the foraging
resource variable in the habitat models.
Shelter and forest structure are important at fine
spatial scales
Habitat quality attributes related to forest structure
were significant at fine scales in most of the multi-
scale models. Mature stands may be important not
only for their higher productivity of food resources
(Apps et al. 2004), but also for their suitability to be
used as shelter areas (Clevenger et al. 1997). Forest
height was significant for seasonal habitat use in fall
and summer. During fall the effect may be correlated
to foraging use of these habitats due to their potential
higher fruit production and during summer to refuge
and thermal regulation (Apps et al. 2004). Forest
canopy cover was consistently a good predictor of bear
occurrence along the seasons and periods, suggesting
that forests with higher canopy cover provide more
protection to bears. These findings are consistent with
previous ecological knowledge on habitat selection of
this species in the study area. For example, Clevenger
(1991) found that brown bear in the Cantabrian Range
was largely dependent on forest cover. Naves et al.
(2003) and Mateo-Sa
´nchez et al. (2014) also found
that forest cover was positively related to habitat use.
However, this conclusion differs from other studies for
brown bear in other study areas (e.g. Canada), where
canopy cover seemed to be inversely related to use,
likely favoring other food resources in more open
areas. A plausible explanation for this difference is the
nature and size of the study area. The Cantabrian
Range has, as most of the Iberian Peninsula, an old
history of human uses and bear persecution. In
addition, in this area humans and bears cohabit in a
relatively small extent where closed forest cover may
provide crucial protection from humans. On the
contrary, other landscapes in North America or North
Scandinavia with very low human pressure over large
areas might offer remote and inaccessible lands
regardless of canopy closure.
Influence of population size and seasonal resources
for multi-scale models predictive ability: the more
spread the population is, the more difficult
prediction becomes
We found that multi-scale model
´s ability to predict
brown bear occupancy was influenced by the status and
size of this population as well as by seasonal changes.
In the early years with small population size, presum-
ably quite below the carrying capacity of the available
habitat in the core population nuclei, bear occurrence
was associated with a narrow range of habitat charac-
teristics (hence more predictable) as apparent from the
spatial distribution of bear habitat (Fig. 3a). However,
the population increase and potential saturation in parts
of the core nuclei might have resulted in the newly
expanding populations using an increased variety of
land covers and other environmental characteristics in
the landscape. This broader range and diversity of
environmental characteristics may force individuals to
be more plastic (more generalist in their habitat
preferences) when selecting new territories outside
the original (and saturated) core populations, including
settlement into more heterogeneous and lower-quality
areas (Fig. 3b), agreeing with recent results for other
mammal species in Spain such as the Iberian lynx
(Gasto
´n et al. 2015). Such trend would translate into
reduced model prediction ability for the period
2005-2010 (that with an expanding species range).
Landscape Ecol
123
This finding is congruent with McFarland et al. (2014),
who found that multi-scale species distribution models
produced poor estimations for a bird species with a
growing population.
Similar considerations apply to seasonal models,
which showed higher predictive ability in winter, when
the home ranges are smaller and use of resources is
confined to more localized areas. In contrast, predic-
tions were weaker in spring, when longer distance
dispersal and mating trigger the use of suboptimal
habitat. Good predictions for fall might be given by a
broader use of resources than in winter but narrower
than in spring, considering that in fall availability and
reachability of high quality habitat is needed due to
nutritional requirements (i.e., hyperphagia).
In any case, all the multi-scale models in this study
outperformed previous multi-scale models for the
species in the study area (Mateo-Sa
´nchez et al. 2014).
Optimal multi-scale models in the previous study by
Mateo-Sa
´nchez et al. (2014) included variables related
to landscape composition (percentage of the landscape
occupied by forest), human disturbance and canopy
cover with AUC =0.862, compared to 0.910 of the
year-round model in this study. This remarkable
increase (p \0.001) in model performance may be
due to (i) the direct quantification of foraging
resources as a predictive variable (including seasonal
and year-round estimates) and (ii) the use of a finer
spatial resolution of the environmental data, particu-
larly LiDAR data on vegetation structure, concurring
with the results from previous single-scale habitat
studies (Tattoni et al. 2012; Zellweger et al. 2014).
Additional comparative research should however be
tackled to specifically clarify how different data on
forest cover may influence inferences and predictive
power of multi-scale habitat suitability models.
Conclusions
Multi-scale habitat models have shown to accurately
predict areas where species can meet their ecological
requirements, and enable researchers to reliably under-
stand the species-habitat relationships. However, very
little was known about how the spatiotemporal changes
in species distributions may affect the operational scales
and the inferred habitat relationships. As shown in this
study, operative scales appeared to be robust across
variations in bear habitat use and distribution due to
seasonal changes and population increase, showing a
likely stable perception of environment. However, the
hierarchy and strength of factors influencing habitat
relationships is different depending on the distinctive
use of seasonal and temporal resources. Similarly, key
differential factors among seasons, such as seasonal
food resource, are crucial to provide more accurate and
informative habitat suitability models. Here we intended
to highlight that the understanding of multi-scale
behavioral responses by species to spatial patterns that
continually shift over time may be essential to produce
reliable species distribution modelling that feed conser-
vation measures. We admit that our habitat models
could be further strengthened in twodirections. First, by
explicitly accounting for temporal changes in landscape
structure (e.g. changes in forest cover or structure
through time), besides the considered changes in
population status. Second, through more detailed infor-
mation about brown bear use of the landscape, partic-
ularly through movement data gathered through a
sufficiently large set of GPS collars for individuals of
different age, gender and condition. Regardless these
potential directions for future improvements, we hope
that our study has contributed to the understanding of
multi-scale species-habitat relationships in dynamic
landscapes and populations, and hope to motivate
further research on this topic with important implica-
tions in landscape ecology. Finally, our study provides
the most detailed (spatially and temporally) habitat
models available so far for the endangered brown bear
populations in the Cantabrian Range, therefore being of
considerable value to support conservation strategies
aiming to ensure the persistence of this emblematic
species in the Iberian Peninsula.
Acknowledgments Funding was provided by the Spanish
Ministry of Science and Innovation research grant GEFOUR
(AGL2012-31099) and Technical University of Madrid. We are
also grateful to the Regional Administration involved in the
brown bear management: Junta de Castilla y Leo
´n, Gobierno de
Cantabria, Principado de Asturias and Xunta de Galicia for
providing data. Thanks also to the valuable support provided by
Fundacio
´n Oso Pardo.
References
Anderson DP, Forester JD, Turner MG, Frair JL, Merrill EH,
Fortin D, Mao JS, Boyce MS (2005) Factors influencing
female home range sizes in elk (Cervus elaphus) in North
American landscapes. Landscape Ecol 20(3):257–271
Landscape Ecol
123
Apps CD, McLellan BN, Woods JG, Proctor MF (2004) Esti-
mating grizzly bear distribution and abundance relative to
habitat and human influence. J Wildl Manag 68(1):
138–152
Ballesteros F, Palomero G (2012) Conectividad, demografı
´ay
conservacio
´n del oso pardo Canta
´brico. In: San Miguel A,
Ballesteros F, Blanco JC, Palomero G (eds) Manual de
buenas pra
´cticas para la gestio
´n de corredores oseros en la
Cordillera Canta
´brica. Fundacio
´n Oso Pardo. Ministerio de
Agricultura, Alimentacio
´n y Medio Ambiente. Serie
Especies Amenazadas, Madrid, pp 21–33
Ballesteros F, Martı
´n B, Blanco JC (2012) Calidad de ha
´bitat y
presencia de osos en el corredor interpoblacional. In: San
Miguel A, Ballesteros F, Blanco JC, Palomero G (eds)
Manual de buenas pra
´cticas para la gestio
´n de corredores
oseros en la Cordillera Canta
´brica. Fundacio
´n Oso Pardo.
Ministerio de Agricultura, Alimentacio
´n y Medio Ambi-
ente. Serie Especies Amenazadas, Madrid, pp 33–56
Barnagaud JY, Devictor V, Jiguet F, Archaux F (2011) When
species become generalists: on-going large-scale changes
in bird habitat specialization. Global Ecol Biogeogr
20(4):630–640
Berland A, Nelson T, Stenhouse G, Graham K, Cranston J
(2008) The impact of landscape disturbance on grizzly bear
habitat use in the Foothills Model Forest, Alberta, Canada.
For Ecol Manag 256(11):1875–1883
Bo
¨rger L, Franconi N, Ferretti F, Meschi F, De Michele G,
Gantz A, Coulson T (2006) An integrated approach to
identify spatiotemporal and individual-level determinants
of animal home range size. Am Nat 168(4):471–485
Bradley BA, Wilcove DS, Oppenheimer M (2010) Climate
change increases risk of plant invasion in the Eastern
United States. Biol Invasions 12(6):1855–1872
Cardillo M, Mace GM, Jones KE, Bielby J, Bininda-Emonds
ORP, Sechrest W, Orme CDL, Purvis A (2005) Multiple
causes of high extinction risk in large mammal species.
Science 309(5738):1239–1241
Ciucci P, Tosoni E, Di Domenico G, Quattrociocchi F, Boitani L
(2014) Seasonal and annual variation in the food habits of
Apennine brown bears, central Italy. J Mammal
95(3):572–586
Ciudad C, Robles H, Matthysen E (2009) Postfledging habitat
selection of juvenile middle spotted woodpeckers: a multi-
scale approach. Ecography 32:676–682
Clevenger AP (1991) Seleccio
´ndeha
´bitat. In: Clevenger AP,
Purroy FJ (eds) Ecologı
´a del oso pardo en Espan
˜a. Madrid.
Museo Nacional de Ciencias Naturales, CSIC (Mono-
grafı
´as 4), pp 73–84
Clevenger A, Purroy F, Pelton M (1992) Brown bear (Ursus
arctos L.) habitat use in the Cantabrian Mountains, Spain.
Mammalia 56(2):203–214
Clevenger AP, Purroy FJ, Campos MA (1997) Habitat assess-
ment of a relict brown bear Ursus arctos population in
northern Spain. Biol Conserv 80(1):17–22
Colles A, Liow LH, Prinzing A (2009) Are specialists at risk
under environmental change? Neoecological, paleoeco-
logical and phylogenetic approaches. Ecol Lett 12(8):
849–863
Delibes M (1999) Osos y madron
˜os. Biolo
´gica 39:62–63
Desrochers A, Renaud C, Hochachka WM, Cadman M (2010)
Area-sensitivity by forest songbirds: theoretical and
practical implications of scale-dependency. Ecography
33:921–931
Dungan JL, Perry JN, Dale MRT, Legendre P, Citron-Pousty
Fortin M-J, Jakomulska A, Miriti M, Rosenberg MS (2002)
A balanced view of scale in spatial statistical analysis.
Ecography 25(5):626–640
Fernandez-Calvo I, Naves J, Fernandez-Gil A, Iglesias JM
(2001). Variacio
´n interanual en la dieta del oso pardo
canta
´brico (Ursus arctos) y su relacio
´n con la disponibili-
dad. In: V National Meeting of the Spanish Society for
study and conservation of mammals. SECEM, Vitoria,
Spain, pp 64–65
Fielding AH, Bell JF (1997) A review of methods for the
assessment of prediction errors in conservation presence/
absence models. Environ Conserv 24(01):38–49
Fisher JT, Anholt B, Volpe JP (2011) Body mass explains
characteristic scales of habitat selection in terrestrial
mammals. Ecol Evol 1:517–528
Fortin M-J, James PMA, MacKenzie A, Melles SJ, Rayfield B
(2012) Spatial statistics, spatial regression, and graph the-
ory in ecology. Spat Stat 1:100–109
Fretwell SD, Lucas HL (1970) On territorial behavior and other
factors influencing habitat distribution in birds. Acta Bio-
theor 19(1):16–36
Fryxell JM, Wilmshurst JF, Sinclair AR (2004) Predictive
models of movement by Serengeti grazers. Ecology
85(9):2429–2435
Garcı
´a P, Lastra J, Marquı
´nez J, Nores C (2007) Detailed model
of shelter areas for the Cantabrian brown bear. Ecol Inform
2(4):297–307
Gasto
´n A, Garcı
´a-Vin
˜as JA (2011) Modelling species distribu-
tions with penalised logistic regressions: a comparison with
maximum entropy models. Ecol Model 222(13):
2037–2041
Gasto
´n A, Bla
´zquez-Cabrera S, Garrote G, Mateo-Sa
´nchez MC,
Beier P, Simo
´n MA, Saura S (2015). Contrasting responses
to different agricultural covers and dispersal plasticity of
woodland species: the case of the endangered Iberian lynx
Go
´mez-Manzanedo M, Urchaga A, Roig S, San Miguel A
(2012) Gestio
´n de la vegetacio
´n arbustiva y herba
´cea. In:
San Miguel A, Ballesteros F, Blanco JC, Palomero G (eds)
Manual de buenas pra
´cticas para la gestio
´n de corredores
oseros en la Cordillera Canta´brica. Fundacio
´n Oso Pardo
Ministerio de Agricultura, Alimentacio
´n y Medio Ambi-
ente. Serie Especies Amenazadas, Madrid, pp 101–140
Grand J, Buonaccorsi J, Cushman SA, Griffin CR, Neel MC
(2004) A multiscale landscape approach to predicting bird
and moth rarity hotspots in a threatened pitch pine–scrub
oak community. Conserv Biol 18(4):1063–1077
Harrell FE (2001) Regression modeling strategies: with appli-
cations to linear models, logistic regression and survival
analysis. Springer, New York
Harrell FE (2014) rms: regression modeling strategies. R
package version 4.2-1. http://cran.r-project.org/packages=
rms
Holland J, Fahrig L (2000) Effect of woody borders on insect
density and diversity in crop fields: a landscape-scale
analysis. Agric Ecosyst Environ 78(2):115–122
Holland JD, Bert DG, Fahrig L (2004) Determining the spatial
scale of species’ response to habitat. Bioscience
54(3):227–233
Landscape Ecol
123
Huber D, Roth HU (1993) Movements of European brown bears
in Croatia. Acta Theriol 38:151–159
Jackson HB, Fahrig L (2012) What size is a biologically relevant
landscape? Landscape Ecol 27(7):929–941
Johnson DH (1980) The comparison of usage and availability
measurements for evaluating resource preference. Ecology
61(1):65–71
Johnson JB, Omland KS (2004) Model selection in ecology and
evolution. Trends Ecol Evol 19:101–108
Kie JG, Bowyer RT, Nicholson MC, Boroski BB, Loft ER
(2002) Landscape heterogeneity at differing scales: effects
on spatial distribution of mule deer. Ecology 83(2):
530–544
Koren
ˇM, Find’o S, Skuban M, Kajba M (2011) Habitat suit-
ability modelling from non-point data: the case study of
brown bear habitat in Slovakia. Ecol Inform 6(5):296–302
Kotliar NB, Wiens JA (1990) Multiple scales of patchiness and
patch structure: a hierarchical framework for the study of
heterogeneity. Oikos 59(2):253–260
Levin SA (1992) The problem of pattern and scale in ecology:
the Robert H. MacArthur award lecture. Ecology
73(6):1943–1967
Marquı
´nez J (2002) Delimitacio
´ndea
´reas crı
´ticas para el oso
pardo y cartografı
´a de calidad de ha
´bitat. Instituto de
Recursos Naturales y Ordenacio
´n del Territorio, Univer-
sidad de Oviedo. Consejerı
´a de Medio Ambiente. Gobierno
del Principado de Asturias
Mateo-Sa
´nchez MC, Cushman SA, Saura S (2014) Scale
dependence in habitat selection: the case of the endangered
brown bear (Ursus arctos) in the Cantabrian Range (NW
Spain). Int J Geogr Inf Sci 28(8):1531–1546
Mattson DJ, Barber K, Maw R, Renkin R (2004) Coefficients of
productivity for Yellowstone’s grizzly bear habitat. US
Department of the Interior, US Geological Survey
McFarland TM, Grzybowski JA, Mathewson HA, Morrison ML
(2014) Presence-only species distribution models to predict
suitability over a long-term study for a species with a
growing population. Wildl Soc Bull 39:218–224
McGarigal K, Cushman SA, Ene E (2012) FRAGSTATS v4:
spatial pattern analysis program for categorical and con-
tinuous maps. Computer software program produced by the
authors at the University of Massachusetts, Amherst. http://
www.umass.edu/landeco/research/fragstats/fragstats.html
McGaughey RJ, Carson WW (2003) Fusing LIDAR data, pho-
tographs, and other data using 2D and 3D visualization
techniques. In: Proceedings of terrain data: applications
and visualization–making the connection, October 28–30,
pp 16–24
Ministerio de Fomento (2015) Plan Nacional de Ortofotografı
´a
Ae
´rea. Instituto Geogra
´fico Nacional. www.pnoa.ign.es
Naves J, Palomero G (1993) El oso pardo en Espan
˜a(Ursus
arctos). ICONA, Madrid
Naves J, Wiegand T, Ferna
´ndez AF, Stephan T (1999) Riesgo de
extincio
´n del oso pardo canta
´brico. La poblacio
´n occi-
dental. Fundacio
´n Oso de Asturias, Oviedo
Naves J, Ferna
´ndez-Gil A, Delibes M (2001) Effects of recre-
ation activities on a brown bear family group in Spain.
Ursus 12:135–140
Naves J, Wiegand T, Revilla E, Delibes M (2003) Endangered
species constrained by natural and human factors: the case
of brown bears in northern Spain. Conserv Biol
17(5):1276–1289
Naves J, Ferna
´ndez-Gil A, Rodrı
´guez C, Delibes M (2006)
Brown bear food habits at the border of its range: a long-
term study. J Mammal 87(5):899–908
Nielsen SE, McDermid G, Stenhouse GB, Boyce MS (2010)
Dynamic wildlife habitat models: seasonal foods and
mortality risk predict occupancy-abundance and habitat
selection in grizzly bears. Biol Conserv 143(7):1623–1634
Palomero G (1995) Winter activity of the brown bear in the
Cantabrian Mountains (Spain). In: Abstracts of the 10th
tenth international conference on bear research and man-
agement, Mora, Sweden, 11–14th September
Palomero G, Ballesteros F, Blanco JC, Garcı
´a-Serrano A, Her-
rero J, Nores C (2011) Osas, el comportamiento de las osas
y sus crı
´as en la Cordillera Canta
´brica, 2nd edn. Fundacio
´n
Oso Pardo, Fundacio
´n biodiversidad, Madrid
Pearson RG, Dawson TP, Liu C (2004) Modelling species dis-
tributions in Britain: a hierarchical integration of climate
and land-cover data. Ecography 27(3):285–298
Pope SE, Fahrig L, Merriam HG (2000) Landscape comple-
mentation and metapopulation effects on leopard frog
populations. Ecology 81(9):2498–2508
Posillico M, Meriggi A, Pagnin E, Lovari S, Russo L (2004) A
habitat model for brown bear conservation and land use
planning in the central Apennines. Biol Conserv
118(2):141–150
R Core Team (2014) R: a language and environment for sta-
tistical computing. R Foundation for Statistical Comput-
ing, Vienna, Austria. http://www.R-project.org/. Version
3.1.2
Rettie WJ, Messier F (2000) Hierarchical habitat selection by
woodland caribou: its relationship to limiting factors.
Ecography 23(4):466–478
Rode K, Robbins C (2000) Why bears consume mixed diets
during fruit abundance. Can J Zool 78(9):1640–1645
Savignac C, Desrochers A, Huot J (2000) Habitat use by pileated
woodpeckers at two spatial scales in eastern Canada. Can J
Zool 78(2):219–225
Schaefer JA, Messier F (1995) Habitat selection as a hierarchy:
the spatial scales of winter foraging by muskoxen. Ecog-
raphy 18(4):333–344
Shirk AJ, Wasserman TN, Cushman SA, Raphael MG (2012)
Scale dependency of American marten (Martes americana)
habitat relations. In: Aubry KB, Zielinski WJ, Raphael
MG, Proulx G, Buskirk W (eds) Biology and conservation
of martens, sables, and fishers: a new synthesis. Cornell
University Press, Ithaca, pp 269–283
Swenson JE, Sandegren F, So-Derberg A (1998) Geographic
expansion of an increasing brown bear population: evi-
dence for presaturation dispersal. J Anim Ecol
67(5):819–826
Tattoni C, Rizzolli F, Pedrini P (2012) Can LiDAR data improve
bird habitat suitability models? Ecol Model 245:103–110
Triantis KA, Mylonas M, Lika K, Vardinoyannis K (2003) A
model for the species–area–habitat relationship. J Biogeogr
30(1):19–27
Urban DL (1987) Landscape ecology. Bioscience 37:119–127
Van Beest FM, Rivrud IM, Loe LE, Milner JM, Mysterud A
(2011) What determines variation in home range size
Landscape Ecol
123
across spatiotemporal scales in a large browsing herbivore?
J Anim Ecol 80(4):771–785
Vicente J, Randin CF, Gonc¸alves J, Metzger MJ, Lomba A,
Honrado J, Guisan A (2011) Where will conflicts between
alien and rare species occur after climate and land-use
change? A test with a novel combined modelling approach.
Biol Invasions 13(5):1209–1227
Vicente JR, Gonc¸alves J, Honrado JP, Randin CF, Pottier J,
Broennimann O, Lomba A, Guisan A (2014) A framework
for assessing the scale of influence of environmental factor
on ecological patterns. Ecol Complex 20:151–156
Vos CC, Verboom J, Opdam PF, Ter Braak CJ (2001) Toward
ecologically scaled landscape indices. Am Nat 157(1):
24–41
Wasserman TN, Cushman SA, Wallin DO, Hayden J (2012)
Multi scale habitat. Relationships of Martes americana in
northern Idaho, USA. USDA forest service RMRS research
paper RMRS-RP-94
Weaver JE, Conway TM, Fortin M-J (2012) An invasive spe-
cies’ relationship with environmental variables changes
across multiple spatial scales. Landscape Ecol 27:
1351–1362
Wiegand T, Naves J, Stephan T, Ferna
´ndez A (1998) Assessing
the risk of extinction for the brown bear (Ursus arctos)in
the Cordillera Cantabrica, Spain. Ecol Monogr 68(4):
539–570
Wiens JA (1989) Spatial scaling in ecology. Funct Ecol
3:385–397
Woodroffe R (2000) Predators and people: using human den-
sities to interpret declines of large carnivores. Anim Con-
serv 3(2):165–173
Zellweger F, Morsdorf F, Purves RS, Braunisch V, Bollmann K
(2014) Improved methods for measuring forest landscape
structure: LiDAR complements field-based habitat
assessment. Biodivers Conserv 23(2):289–307
Zollner PA (2000) Comparing the landscape level perceptual
abilities of forest sciurids in fragmented agricultural
landscapes. Landscape Ecol 15(6):523–533
Landscape Ecol
123
A preview of this full-text is provided by Springer Nature.
Content available from Landscape Ecology
This content is subject to copyright. Terms and conditions apply.
