Biodiversity and Conservation 11: 10631079, 2002.

2002 Kluwer Academic Publishers. Printed in the Netherlands.

A comparison on the response to forest fragmentation

by medium-sized Iberian carnivores in central Spain

1,2, *, JOSE LUIS TELLERIA

EMILIO VIRGOS 1 and TOMAS
SANTOS 1
1
Department of Animal Biology I ( Vertebrate Zoology), Faculty of Biological Sciences, Universidad
2
Complutense, E-28040 Madrid, Spain; Current address: Instituto de Investigacion en Recursos

Cinegeticos ( IREC), CSIC-UCLM, Ronda de Toledo S /n, E-13005 Ciudad Real, Spain; * Author for
correspondence (e-mail: evirgos@ irec.uclm.es; fax: 1(1) 926 -225184)

Received 23 November 2000; accepted in revised form 25 June 2001

Key words: Forest fragmentation, Geographic effects, Iberian Peninsula, Medium-sized carnivores,
Spatial scale

Abstract. We studied the use of forest fragments by five medium-sized carnivore species in 280 forest
fragments on the two Iberian plateaus. We looked for indirect evidence (faeces, tracks, dens) of fragment
use (occurrence) by the species and analysed whether occurrence could be related to four groups of
variables: local (vegetation structure and patch size), landscape (distance to possible colonisation
sources), regional (fragment location on the northern or southern plateau) and the vegetation type of the
fragments. We analysed the differential response of species according to their life-history and behavioural
traits. The relationship between use by each species and the factors studied was analysed using stepwise
logistic regressions. Results indicate that three factors are crucial to explain fragment use: fragment size,
geographic location and vegetation type. Large fragments are used more than smaller ones, the northern
plateau is more suitable than the southern plateau, and holm oak and broad-leaved oak forests are more
used than pine woods. The effects and magnitude varied slightly among species, depending on habitat
requirements or life-history traits. Data indicate that conservation strategies in fragmented environments
must take into account elements functioning at different spatial scales, and that it is essential to consider
each case within a characteristic regional context.

Introduction

Forest fragmentation adversely affects many forest vertebrates (Opdam 1991;

Andren 1994; Bender et al. 1998). Mammalian studies on this subject have mainly
focused on small mammals (Verboom and van Apeldoorn 1990; Tellera et al. 1991;
Van Apeldoorn et al. 1992; Tew and Macdonald 1993; Oehler and Litvaitis 1996),
and there is a lack of studies on carnivore species. However, due to their relatively
large size, these species may be more vulnerable to processes of habitat deteriora-
tion (Laurance 1991). This lack of information is particularly important since many
of these animals are endangered or affecting other key species such as game or
threatened ones.
This study explores the patterns of fragment use by medium-sized Iberian
carnivores of central Spain. In particular, it analyses the role of factors that may
affect fragment occupation by these species at different spatial scales (Kotliar and
Wiens 1990; Brown and Litvaitis 1995). The use of each fragment by a species may
1064

depend on factors that act on a local scale (e.g. forest area and physiognomic
structure), and also on factors that act at landscape scale (e.g. distances to possible
sources of animals for colonisation), or on its regional abundance because popula-
tions in fragments may interact with larger populations (Gilpin and Hanski 1991). It
is interesting to know, therefore, the relative importance of the different factors
given that conservation measures for these animals may be ineffective if not applied
at the appropriate scale (Hansen et al. 1993). Conversely, we suggest that different
carnivore species may show a contrasted response to fragmentation according to
their different life-history traits. Red foxes are able to survive in very deforested
landscapes, so they could be less susceptible to forest lost than other species. On the
other hand, badgers are very constrained by the presence of good soils to dig their
setts (Roper 1993), and this probability is enhanced in large forests with high
environmental heterogeneity (Freemark and Merriam 1986). So, we expect that
badgers may be more constrained by large forests than other species due to this
particular behavioural trait. Wildcats, genets and stone martens need a high availa-
bility of shelter and food (especially woodmouse and microtines in the case of
genets and wildcats). As abundance of rodents is very high in small forests (Tellera
et al. 1991), we cannot expect these species to be limited in fragment size by this
factor. Moreover, it is probable that these species may be recorded both in large and
small fragments, although they may prefer large forest because of their necessity of
shelter. Finally, both genets and stone martens show a rather termophilic affinity
(Delibes 1983), whereas badger is more abundant in wet areas (Virgos and
Casanovas 1999). So, we expect genets and stone martens to show higher occur-
rence in the south than in the north, and the opposite in the case of the badger.
The Iberian plateaus are a good system for studying medium-sized carnivore
distribution in fragmented forests at different spatial scales. This group of mammals
is still relatively rich in Spain, although it is usually concentrated in the best
conserved mountain areas where forests are common (e.g. Delibes 1983). In
contrast, north and south plateaus are intensively farmed with patchily distributed
forests across the landscape (Santos and Tellera 1998). The plateaus differ in
climatic conditions and predominant land uses. The northern plateau is rainier and
cooler than the southern plateau, two traits which in Mediterranean climates
positively affect primary productivity and the abundance of primary and secondary
consumers (Oksanen et al. 1981). Moreover, the northern plateau is devoted to cattle
raising, whereas hunting on small game species is common in the south fragments,
which is associated with intensive predator control activities. It is, therefore,
possible that these uses give rise to differences in regional abundance of mammals
and that these differences affect their ability to colonise or survive in the fragments.
The specific aim of this study is to analyse whether the use of fragments by
medium-sized carnivore species is linked to (a) their location in the northern or
southern plateau (regional effect), (b) landscape structure at local scale, especially
fragment distance to other forests (landscape effect), (c) the size and physiognomic
and botanical features of each fragment (fragment effect), (d) the vegetation type
(tree dominant cover) of each fragment and (e) the particular life-history traits of
each species studied. Supposing that several of these factors are linked to fragment
 1065

use, we were interested in finding out the relative importance of each one in order to
draw conclusions that can be applied to the conservation of medium-sized carni-
vores.

Material and methods

Study area

The Iberian plateaus are a characteristic plain area of Spain divided by the
mountains of the Central System. The northern plateau is higher (800 m asl.), wetter
and cooler than the southern plateau (above 600 m asl.) (Font 1983). Both areas are
dominated by extensive cereal farming, with isolated remains of the old Mediterra-
nean forest (holm oak forests, mainly Quercus ilex) and conifer plantations (mainly
Pinus pinea, P. pinaster). These forests present an intensely fragmented distribution
pattern in which small fragments are common in the landscape (Santos and Tellera
1998).
A total of 280 forest fragments distributed among eight different archipelagos (a
group of neighbour fragments in a particular area) in the two plateaus were sampled:
169 and 111 on the northern and southern plateaus, respectively (see archipelago
location in Figure 1). Selection of the forests was made using recent aerial
photographs (1:20 000). Two independent criteria were taken into account for this
selection: (a) its usefulness in the gradient of patch sizes (0.1.100 ha); (b) each
forest patch should be separated from other patches (at least 25 m) by an agricultural
matrix (devoted to intensive cereal crops). The distribution of sizes, isolation and
vegetation type in each archipelago is given in Table 1.

Species selection and sampling

We chose the five commonest medium-sized carnivore species of the Iberian

Peninsula and we assume that they need woodland or scrubland areas to survive,
even though they use the ecotones and peripheral deforested areas to feed. These
species use rather large home ranges, so that they may regularly visit several
fragments, and so locating them in a particular forest indicates the current use of this
fragment but does not necessarily imply the stable presence of an isolated popula-
tion. The selected species were: red fox (Vulpes vulpes), badger (Meles meles), stone
marten (Martes foina), genet (Genetta genetta) and wild cat (Felis silvestris). We
also included rabbit (Oryctolagus cuniculus) for its potential role in carnivore
distribution because it is an important prey for many carnivore species in Spain
(Jaksic and Soriguer 1981).
We searched for indirect evidences of use in each forest studied (tracks, faeces
and dens / burrows). This method has been proposed by several researchers (e.g.
Clevenger 1993; Staender 1998) as a good alternative to trapping or scent-stations
indexes in carnivore-habitat studies conducted at landscape or regional scales. The
latter methods are too expensive and time-consuming to be an adequate option in
1066

Figure 1. Location of eight forest archipelagos in the two Iberian plateaus containing the study forest
fragments.

large-scale studies. In addition, other authors have indicated the good fit between
faeces or dens recording and population density or frequency of occurrence (e.g.
Cresswell et al. 1989; Cavallini 1994; Staender 1998). The fragments were visited
from May to August of 1994. One visit per patch was made during this study period,
providing a snapshot of the current use of these fragments for each species. This
procedure allowed us to describe the pattern of fragment use for the individuals of
different species over a range of sizes, degrees of isolation and habitat features.
The fragments were sampled using survey routes along roads and paths of similar
width (to locate scats of stone marten, red fox, wildcats and rabbits) and cross-
country (out of the paths) to locate badger setts and genet latrines. In fragments
under 2 ha, the whole area was exhaustively surveyed for sign search (both in paths
and cross-country). In fragments over 2 ha, path and cross-country routes were
surveyed alternatively in each forest. We sampled survey routes of 500 m long. The
routes were 1 m wide in paths and roads and 10 m wide cross-country. In particular,
the genet latrines and dens (for badgers) were searched in the same type of habitat
features along the survey routes: fences, edges, rocks, slopes, riverbanks. We
attenuated the problems associated with the use of signs as a measure of current use
by sampling under similar climatic conditions (late spring and summer) and
searching for signs with the same protocol across habitats. The first attenuates the
problem with differences in decay rates over seasons. The second alleviates the
problem of differences in detectability in different habitats.
Table 1. Forest archipelagos studied in the Iberian plateaus.
Archipelago N Location Habitat Inter-fragment Distance to nearest ,1 ha 110 ha 1050 ha 50100 ha .100 ha
type distance continuous forests
Jamuz 41 North Holm oak 116.34 (25520) 14.90 (6.021.4) 19 15 5 1 1
Lerma 29 North Holm oak 97.24 (25495) 10.20 (1.121.8) 13 4 6 2 4

Sahagun 38 North Broad-leaved 48.95 (25220) 4.09 (0.229.0) 13 15 5 2 3
Vitoria 22 North Broad-leaved 202.73 (25460) 1.48 (0.63.3) 6 9 7 0 0
Medina 39 North Pine 72.56 (25280) 23.09 (18.329.8) 10 21 4 3 1
Villatobas 39 South Holm oak 101.28 (25760) 41.3 (28.445.3) 15 16 4 3 1
Mota 31 South Holm oak 88.06 (25440) 10.8 (6.617.4) 11 8 7 3 2
La Roda 41 South Pine 174.88 (25460) 28.05 (22.330.9) 14 16 7 1 3
For each archipelago are given: Number of forest patches (N), geographic location (plateau), habitat type (dominant tree species), mean and range of inter-fragment
distances (in m), mean and range of distances (in km) from forest higher than 1000 ha, and number of forest fragments in each size category.
1067
1068

The number of these survey routes in each forest fragment follows an approxi-
mately logarithmic relationship with fragment size, in order to homogenously cover
the whole patch. Thus, the number of routes varied from 2 in the 23 ha fragments
to a minimum of 10 in fragments over 100 ha.

Fragment features

The fragments were characterised at the three spatial scales of the study by using a
series of factors that could potentially affect mammal use. Firstly, regional effect
involved classification into two groups according to location on the northern or
southern plateau. Secondly, for landscape effect isolation was quantified in relation
to possible colonisation sources (distance to nearest fragment, distance to forests
greater than 100 ha, and distance to forests greater than 1000 ha in the mountains or
riparian woods). The measurements were made on aerial photos (1:18 000) in the
case of the first two estimates and on LANDSAT images (1:100 000) for the third
variable. For fragment effect, the size of each fragment was determined in hectares
on photos 1:18 000, and vegetation type was determined in the field according to the
dominant type of tree in each fragment (holm oak wood Q. ilex; broad-leaved
forests: Pyrenean oak woodland Q. pyrenaica and / or Portuguese oak Q. faginea;
and pinewood P. pinaster and P. pinea).Vegetation structure was quantified in 25 m
radius circles by visually estimating the cover of variables that could be of potential
importance to the different species for food, shelter, or both: tree cover, shrub .50
cm cover, shrub ,50 cm cover and herb cover (see Morrison et al. 1992; Santos and
1998).
Tellera

Analyses

The occurrence of species in each fragment was used as a dependent variable in a

stepwise logistic regression analysis (forward Wald method; Hosmer and
Lemeshow 1989) in which location (in north or south plateau), distances, size,
vegetation structure variables, and vegetation type (dominant tree species, see
above) for each fragment were used as predictors. The vegetation type and the
plateau were included as dummy variables (Neter et al. 1985). In order to reduce
collinearity in multiple logistic regression analysis, we analysed the potential
association between continuous predictor variables through Pearson correlations,
then we used the different subsets of independent variables to run stepwise multiple
logistic regression and to evaluate the adequacy of the different models from a
deviance analysis among G-statistics values from the final models (Manly et al.
1993). In these analyses, all possible two-way interactions between variables were
considered.
All the variables were transformed before the analyses by logarithmic (fragment
size and isolation) and angular (vegetation covers) transformation. The logistic
regression analyses were carried out with the statistical package SPSS 6.0 and the
remaining analyses were done with Statistica 4.5.
 Table 2. Pearson correlation matrix among the different variables used in this study.
Variable Distance Distance Distance Size of the Fragment Herb Shrub Shrub Tree
nearest fragment forest nearest size cover ,50 cm .50 cm cover
fragment .100 ha .1000 ha fragment cover cover

Distance nearest fragment 0.038 P 5 0.532 0.050 P 5 0.406 0.047 P 5 0.436 20.098 P 5 0.103 20.085 P 5 0.155 0.073 P 5 0.226 0.090 P 5 0.132 0.072 P 5 0.231
Distance fragment . 100 ha 0.102 P 5 0.088 20.108 P 5 0.071 20.476 P , 0.001 0.156 P 5 0.009 20.257 P , 0.001 20.269 P , 0.001 0.147 P 5 0.014
Distance forest . 1000 ha 0.145 P 5 0.015 0.119 P 5 0.048 0.008 P 5 0.892 20.148 P 5 0.013 20.247 P , 0.001 20.210 P , 0.001
Size of the nearest fragment 0.190 P 5 0.001 20.007 P 5 0.902 0.059 P 5 0.329 0.024 P 5 0.686 0.031 P 5 0.605
Fragment size 20.180 P 5 0.002 0.164 P 5 0.006 0.175 P 5 0.003 20.236 P , 0.001
Herb cover 20.365 P , 0.001 20.405 P , 0.001 0.105 P 5 0.080
Shrub ,50 cm cover 0.294 P , 0.001 20.277 P , 0.001
Shrub .50 cm cover 20.255 P , 0.001
1069
1070

Results

Correlations among variables

Only three variables did not show significant correlations between them: fragment
size, distance to nearest forest greater than 1000 ha and distance to nearest fragment.
All vegetation structure variables were correlated with some or all of the above-
mentioned variables, so they were not included in the following analyses (see Table
2). We examined relationships of these three variables with geographic location and
vegetation type (fixed factors), which should be also included in the regression
analysis, with protected ANOVA after MANOVA analysis (Scheiner 1993). MAN-
OVA results indicated that there were significant differences between plateaus (x 2 5
0.77; 2 276 d.f.; P , 0.001) and vegetation types (x 2 5 0.33; 6 550 d.f.; P , 0.001).
The protected ANOVA for each dependent variable indicated that only the distance
to nearest forest greater than 1000 ha showed statistical differences both for
geographic location and vegetation type (Table 3). Southern forest fragments were
further from large forests than northern ones (26.19 6 8.84 vs. 13.05 6 12.50 km),
and broad-leaved forests were less isolated than holm oak and pine ones (3.13 6 4.6
vs. 18.64 6 12.93 and 25.57 6 3.41 km). Therefore, the distance to forests greater
than 1000 ha was not independent of vegetation type and geographic location, so
both in least squares and logistic stepwise multiple regressions we checked two
independent models including variables non-intercorrelated: (1) fragment size (local
effects), distance to nearest forest (local effects), geographic location (regional
effects) and vegetation type; (2) fragment size (local effect), distance to nearest
fragment (local effect) and distance to nearest forest greater than 1000 ha (landscape
effects).

Species occurrence in forest fragments

The five species studied presented an unequal pattern of use among fragment sizes
(Table 4). The model that included fragment size, distance to nearest fragment,
vegetation type and geographic location (model 1) performed better than the model
with fragment size, distance to nearest fragment and distance to nearest forest
greater than 1000 ha for all species (model 2; see deviance analyses in Table 5).

Table 3. Results of the protected ANOVA performed with geographic location (plateau) and habitat types
(holm oak, broad-leaved oak and pine forests) as factors, and fragment size, distance to nearest fragment
and distance to nearest forest greater than 1000 ha.
Variable F value d.f. Significance P
Distance to nearest fragment 2.173 (Plateau) 1 278 0.14
0.098 (Habitat type) 2 277 0.91
Distance to nearest forest .1000 ha 81.103 (Plateau) 1 278 ,0.001
279.224 (Habitat type) 2 277 ,0.001
Fragment size 1.258 (Plateau) 1 278 0.26
1.029 (Habitat type) 2 277 0.36
Table 4. Number of fragments used by the study species according to fragment size (in brackets, number of surveyed fragments in each size category).
Species ,0.5 ha 0.52 ha 25 ha 510 ha 1050 ha .50 ha North South Total
(n: 60) (n: 85) (n: 43) (n: 17) (n: 45) (n: 30) Plateau Plateau (n: 280)
(n: 169) (n: 111)
Vulpes vulpes 15 28 14 9 30 29 84 41 125
Meles meles 0 2 0 1 3 9 14 1 15
Martes foina 1 5 2 1 5 9 18 5 23
Genetta genetta 0 4 1 1 5 5 14 2 16
Felis silvestris 2 10 2 0 3 7 22 2 24
1071
1072

Table 5. Diagnostic results from logistic regressions for the two different models considered and for each
species considered: concordance between observed and predicted values (%) ((a) overall, (b) presence
correctly predicted, (c) absence correctly predicted), G-model, P-values and deviance analysis (obtained
from the expression Gmodel 1 2Gmodel 2 with d.f. model 1 2 d.f. model 2 degrees of freedom).
Vulpes vulpes Meles meles Martes foina Genetta genetta Felis silvestris
Model 1 69.29(a) 97.14(a) 92.86(a) 94.64(a) 91.79(a)
59.20(b) 53.33(b) 13.04(b) 18.75(b) 4.17(b)
77.42(c) 99.62(c) 100(c) 99.24(c) 100(c)
G4 5 74.48 G3 5 55.07 G4 5 38.23 G4 5 39.64 G3 5 22.51
P , 0.001 P , 0.001 P , 0.001 P , 0.001 P , 0.001
Model 2 71.79(a) 96.43(a) 92.14(a) 94.29(a) 91.43(a)
58.40(b) 33.33(b) 8.7(b) 0(b) 0(b)
82.58(c) 100(c) 99.61(c) 100(c) 100(c)
G2 5 66.31 G2 5 45.09 G2 5 31.78 G1 5 14.44 G2 5 12.03
P , 0.001 P , 0.001 P , 0.001 P , 0.001 P 5 0.002
Deviance G2 5 8.17 G1 5 9.98 G2 5 6.45 G3 5 25.2 G1 5 10.48
P 5 0.017 P 5 0.0015 P 5 0.04 P , 0.001 P 5 0.0012

Particularly, model 1 was largely more suitable than model 2 for genets and
wildcats, species with strong differences between models both in G values and
percent of concordance (Table 5).
In addition, both models showed low percentages of concordance for the species
presence (high omission rate), with exceptions. Intermediate levels of omission
errors were found in badgers and red foxes from model 1. The 0% correct
assignment for presence data in genets and wildcats from model 2 is noticeable.
Because of their low predictive power and low G values, we concentrate on the
results from model 1 (Table 5).
A regional effect was found for red foxes, badgers and wildcats, which showed
higher occurrences on the north than on the south plateau. Fragment size was an
important factor linked to occurrence in red foxes and genets. In all these cases,
species occurrence was higher in large than in small fragments.Vegetation type also
explained the occurrence of red foxes and genets, finding higher values in holm oak
and broad-leaved forests than in pine fragments for red foxes and the opposite
pattern for the genet. A relatively important part of the occurrence pattern of
badgers, stone martens, genets and wildcats was explained by two-way interactions
between some predictors. The occurrence of badgers was higher in large oak forests
than in small or large pine ones (habitat 3 size interaction). The occurrence of stone
martens was explained by two interactions: habitat 3 size and habitat 3 distance to
nearest fragment. Stone martens showed higher occurrences in large oak fragments
than in large or small pine fragments. The second interaction indicated that stone
martens were more frequent in fragments near other patches in holm oak woodlands
than in both near and isolated broad-leaved and pine forests. However, this pattern
could be spuriously affected by the low power of the model (Table 5). Genets were
more frequent in large pine fragments compared to both large and small oak
fragments. For wildcats, the occurrence was higher in large holm oak forests than in
both large and small broad-leaved or pine fragments. In summary, fragment size,
 1073

Table 6. Summary of the selected variables in the final models from the stepwise logistic regressions
(forward Wald) for each species considered.
Species Variable B P R
Vulpes vulpes Size 1.221 ,0.001 35.02
Plateau 0.710 ,0.05 8.87
Habitat ,0.05 7.5
Holm oak 0.755 ,0.05
Broad-leved 0.750 0.07
Meles meles Size 3 Habitat ,0.001 42.85
Size 3 Holm oak 2.498 ,0.001
Size 3 Broad-leaved 1.852 ,0.001
Plateau 0.854 ,0.01 20.07
Martes foina Habitat 3 distance to ,0.01 17.15
nearest forest .1000
Holm oak 3 distance 20.668 0.07
Broad-leaved 3 distance 0.857 ,0.01
Habitat 3 size ,0.001 33.7
Holm oak 3 size 1.814 ,0.001
Broad-leaved 3 size 0.498 0.15
Genetta genetta Size 1.368 ,0.001 31.97
Habitat ,0.001 29.81
Holm oak 23.198 ,0.001
Broad leaved 22.193 ,0.01
Plateau 3 distance to 21.263 ,0.01 22.04
nearest forest .1000
Felis silvestris Plateau 2.370 ,0.01 21.23
Habitat 3 size ,0.01 19.10
Holm oak 3 size 0.846 ,0.01
Broad-leaved 3 size 20.111 0.78

regional location and vegetation type effects were important variables for all the
species, while landscape effects only affected stone marten occurrence. Variables
included in the final model for the different species are given in Table 6.

Discussion

Regional and landscape effects

Forest location on the northern or southern plateau is an important factor in

determining the likelihood of the different species to use the fragments. This may be
related to a hypothetical higher regional abundance of carnivore species on the
northern plateau. The greater suitability of northern forests has been demonstrated in
and Santos 1993; roe deer: Virgos
other vertebrates (birds: Tellera and Tellera

1998), and it has been shown that badger abundance is lower in southern than in
northern areas of the mountains of central Spain (Virgos and Casanovas 1999).
Although genets and stone martens are termophilic species and they should be more
abundant in the south than in the north plateau, the data indicated the opposite
1074

pattern. This presumed southwards decrease in abundance may be related both with
the growing deterioration of the forests (perhaps linked to rarity of stone martens,
genets and wildcats) and increased xericity in the south of the Iberian Peninsula
(probably linked to badger rarity), where the formerly intense pressure (felling,
charcoal making, cattle ranching) has brought about an increasing loss in the mesic
tree cover in favour of xeric and rather shrubby species (Costa et al. 1995; Santos
1998). This man-made change in the dominant vegetation has also
and Tellera
occurred in the north, but in a less marked way. In any case, given that the
populations of species established in the fragments are not independent entities, it
seems logical to assume that differences in the regional abundance of the study
species may determine differences in fragment colonisation rates and, therefore, in
the richness of species using them. Hence, the lower use in the south compared to

Figure 2. Graphs showing the percentage of fragment presence of the different carnivore species over a
range of fragment sizes. Range interval for fragment size is large (200 ha) in order to indicate the
different sensitivity of the species to increases in forest size above 200 ha (asssumed as enough to
maintain populations of the different species).
 1075

the north should be a result of the presumed low species density on regional forests
(including the mountain forests) of the southern plateau.
However, the role of isolation from large forests in mountains or riparian woods
(landscape effects) probably is not negligible to explain differences in occurrence
between plateaus. Both predictors are confounded because northern fragments are
less isolated than southern ones. The distance to nearest fragments greater than 1000
ha was an important variable in the second model and allowed to explain an
important part of the occurrence variation for the different species. This is not
surprising because carnivores are less mobile and less able to overcome habitat
barriers than other species such as birds. From this point of view, the most isolated
forests might be expected to be less used than the less isolated ones; a similar fact
has been shown by a medium-sized ungulate species, the roe deer (Capreolus
and Virgos
capreolus), in central Spain (Tellera 1997). So, colonisation and rescue
effects are more difficult in fragments in the southern parts than in their northern
counterparts and they may partially explain the low occurrence of carnivore species
in the south plateau. It could be suggested that both factors, regional habitat quality
(xericity and shrubby structure) and isolation, may explain together the recorded
pattern but the data and spatial contexts do not allow us to test their relative role.
In addition, most parts of southern forests are used as hunting grounds for small
game and predators have been persecuted and killed for decades. Therefore, the
regional effect is probably a combination of several underlying processes that
interact to decrease the suitability of forest fragments in the south of the Iberian
Peninsula. More experimental studies may help us to understand the proximate
causes of this pattern and their relative importance.

Local effects

Fragment size was an excellent predictor of carnivore occurrence in fragments. This

association between fragment size and occurrence has been largely shown for
1994; Bender et al. 1998). This fact may be
different vertebrate species (Andren
explained because large fragments are able to maintain high populations of the
species (Fahrig and Merriam 1994) that should be less prone to extinction (Pimm et
al. 1988). This might be linked to greater use of these fragments and therefore to
higher probability of locating signs of species presence. Moreover, the large
fragments are more heterogeneous than small ones (Freemark and Merriam 1986),
which favours the presence of different scarce resources for the species.
However, there is a surprisingly low use of the smaller fragments by true edge
species, which could make appropriate use of the patch edges and the peripheral
cultivated areas. In any case, sensitivity to size varied considerably according to
species, although the incidence was clearly higher in fragments larger than 100200
ha than in small fragments (see Figure 2 for incidence functions over large
fragments). In accordance with our initial predictions, the species least affected by
fragment size is the red fox, which is able to live in deforested agricultural
environments provided that there are small remnants of bushes or trees where it can
establish itself (Macdonald et al. 1981). Badgers and wildcats are also habitat
1076

generalist species, although they respond less well to area reduction; both tend to
disappear in fragments of similar size (Figure 2), with small differences in their
rarefaction in fragments under 100 ha, but with a higher occurrence of wildcats in
the small fragments. Badgers are omnivores, able to exploit a wide range of food
resources (Roper 1994), but they need places to locate its setts (dens) from which to
make routine movements (Roper 1993). Wildcats, however, appear to depend more
on prey abundance than on availability of den sites (Stahl and Leger 1992). In this
sense, the great abundance of small mammals and the presence of rabbits (the basic
prey of wildcat; Stahl and Leger 1992) in small fragments and in ecotonical areas
et al. 1991; Tew and Macdonald 1993) might facilitate its more frequent
(Tellera
appearance in small fragments, which has been suggested in our initial predictions.
Finally, stone marten and genet do not appear to be uniquely restricted to the
larger fragments, in general being notably scarce in the forest fragments of the study
area. This result is paradoxical given that the two species are recognised trophic and
habitat generalists, well able to move over relatively broad territories (see Livet and
Roeder 1987 for the genet; Libois and Waechter 1991 for the stone marten). It is
possible that the thermal preferences of genets may restrict it to certain specific sites
in central Spain (Delibes 1983; Virgos and Casanovas 1997), reducing its regional
abundance and its probability of fragment occupancy. However, it is difficult to
explain the rarefaction of the stone marten on the Iberian plateaus and their
concomitant withdrawal to larger fragments. It could be suggested that the frag-
ments show a lack of suitability due to their low rock cover and hilliness, two
important habitat features for stone marten presence in central Spain (Virgos and
Casanovas 1998; Virgos et al. 2000).

Vegetation type

The floristic composition of the tree cover does appear to determine certain
differences among species, given that most of them were more frequent in oak than
in pine forests, and so the species richness in pine forests was lower than in oak
ones. An exception to the general low occurrence of the species in pine forests was
the genet, a species more restricted by thermal conditions than by habitat or
physiognomy (own unpublished data). The greater importance of both holm oak and
broad-leaved forests compared to pine forests may be explained because conifer
fragments are mostly plantations on agricultural lands. The relative youth of many
of them (2540 years) could negatively affect their use by the species due to the fact
that they have not had time to be colonised by the study species. On the other hand,
pine forest structure is less diverse (e.g. low shrub cover), providing probably less
refuge and food for the involved species.

Conclusions

In summary, although our results should be interpreted with caution for wildcats,
genets and stone martens due to the relatively low predictive power of the models,
 1077

the data support previous suggestions about the need to consider the conservation
problems of species in fragmented habitats at different spatial scales (see also
Kotliar and Wiens 1990; Hansen et al. 1993; Oehler and Litvaitis 1996). This
involves limitations when planning any strategy to conserve these animals in the
study area. For example, the two plateaus represent different scenarios from the
perspective of carnivore use of fragmented forests. Their conditions depend on
climatic factors and probably on historical and land-use factors that are difficult to
manipulate in the short term or at a local level. However, fragment size seems to be
a more easily manageable factor from the perspective of species conservation. Any
forest alteration that tends to reduce overall size, or any reforestation plan on the
plateaus that includes recovery of these mammals should take into consideration the
basic need to maintain large forests (always over 100 ha) in order to ensure that
these species are conserved. Conversely, reforestation with pine plantations seems
to be a wrong decision if the interest is to maintain medium-sized carnivores in the
region.

Acknowledgements

We are very grateful for the help that the following people provided during the
fieldwork: Marcos Llorente, Francisco J. Garca, Roberto Carbonell, Marino Lopez

Yolanda Cortes
de Carrion, and Roco A. Baquero. Sara Cabezas helped us with the
English. This study is a contribution and was funded by the projects Determination
of Minimal Areas for Forest Vertebrates on the Iberian Plateaux (ICONA-UCM)
and Effects of Forest Fragmentation on the Biology of Forest Vertebrates (PB92 /
0238 DGICYT). E. Virgos was under a post-doc grant in the Instituto de Inves-
en Recursos Cinegeticos
tigacion during the last phases of manuscript submission.

References
H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes with different
Andren
proportions of suitable habitat: a review. Oikos 71: 355366.
Bender D.J., Contreras T.A. and Fahrig L. 1998. Habitat loss and population decline: a meta-analysis of
the patch size effect. Ecology 79: 517533.
Brown A.L. and Litvaitis J.A. 1995. Habitat features associated with predation of New England
cottontails: what scale is appropiate? Canadian Journal of Zoology 73: 10051011.
Cavallini P. 1994. Faces count as an index of fox abundance. Acta Theriologica 39: 417424.
Clevenger A.P. 1993. Sign surveys as an important tool in carnivore conservation research and
management programmes. Seminar on the Management of Small Populations of Threatened Mam-
mals. Council of Europe, Sofia, pp. 4454.
Costa M., Garca M., Morla C. and Sanz H. 1995. La evolucion de los bosques de la Pennsula

Iberica:
basada en datos paleobiogeograficos.
una interpretacion 1: 3158.
Ecologa
Cresswell P., Harris S., Bunce R.G.H. and Jefferies D. 1989. The badger, Meles meles in Britain: present
status and future population changes. Biological Journal of the Linnean Society 38: 91101.
Delibes M. 1983. Distribution and ecology of the Iberian carnivores: a short review. XV Congreso

Internacional de Fauna Cinegetica pp. 359378.
y Silvestre, Trujillo, Espana,
1078

Fahrig L. and Merriam G. 1994. Conservation of fragmented populations. Conservation Biology 8:

5059.
Font I. 1983. Climatologa de Espana y Portugal. Instituto Nacional de Meteorologa, Madrid.
Freemark K.E. and Merriam G. 1986. Importance of area and habitat heterogeneity to bird assemblages in
temperate forest fragments. Biological Conservation 36: 115141.
Gilpin M.E. and Hanski I. 1991. Metapopulation Dynamics: Empirical and Theoretical Investigations.
Academic Press, London.
Hansen A.J., Garman S.L. and Marks B. 1993. An approach for managing vertebrate diversity across
multiple-use landscapes. Ecological Applications 3: 481496.
Hosmer D.W. and Lemeshow S. 1989. Applied Logistic Regression. Wiley Interscience, Chichester, UK.
Jaksic F.M. and Soriguer R.C. 1981. Predation upon the European rabbit Oryctolagus cuniculus in
Mediterranean habitats of Chile and Spain: a comparative analysis. Journal of Animal Ecology 50:
269281.
Kotliar N.B. and Wiens J.A. 1990. Multiple scales of patchiness and patch structure: a hierarchical
framework for the study of heterogeneity. Oikos 59: 253260.
Laurance W.F. 1991. Ecological correlates of extinction proneness in Australian tropical rain forest
mammals. Conservation Biology 5: 7988.
Libois R. and Waechter A. 1991. La Fouine (Martes foina). In: Artois M. and Delattre P. (eds),

Encyclopedie des carnivores de France BY. Societe Francaise pour lEtude et Protection des
`
Mammiferes, Bohallard, Puceul.
Livet F. and Roeder J.-J. 1987. La genette (Genetta genetta). In: Artois M. and Delattre P. (eds),
Encyclopedie des carnivores de France BY. Societe Francaise pour lEtude et Protection des
`
Mammiferes, Bohallard, Puceul.
Macdonald D.W., Bunce P.H. and Bacon P.J. 1981. Fox populations, habitat characterization and rabies
control. Journal of Biogeography 8: 145151.
Manly B., McDonald L. and Thomas D. 1993. Resource Selection by Animals. Chapman & Hall,
London.
Morrison M.L., Marcot B.G. and Mannan R.W. 1992. Wildlife Habitat Relationships. Concepts and
Applications. University of Wisconsin Press, Madison, Wisconsin.
Neter J., Wasserman W. and Kutner M.H. 1985. Applied Linear Statistical Models. Richard D. Irwin,
Homewood, Illinois.
Oehler J.D. and Litvaitis J.A. 1996. The role of spatial scale in understanding responses of medium-sized
carnivores to forest fragmentation. Canadian Journal of Zoology 74: 20702079.
Oksanen L., Fretwell S.D., Arruda J. and Niemela P. 1981. Exploitation ecosystems in gradients of
primary productivity. American Naturalist 118: 240261.
Opdam P. 1991. Metapopulation theory and habitat fragmentation: a review of Holarctic breeding bird
studies. Landscape Ecology 5: 93106.
Pimm S.L., Jones H.L. and Diamond J. 1988. On the risk of extinction. American Naturalist 122:
757785.
Roper T.J. 1993. Badger setts as a limiting resource. In: Hayden T.J. (ed.), The Badger BY. Royal Irish
Academy, Dublin, pp. 2634.
Roper T.J. 1994. The European badger Meles meles: food specialist or generalist? Journal of Zoology,
London 234: 437452.
J.L. 1998. Efectos de la fragmentacion
Santos T. and Tellera de los bosques sobre los vertebrados de las

mesetas Ibericas.
Organismo Autonomo de Parques Nacionales. Coleccion Tecnica,
Madrid.
Scheiner S.M. 1993. MANOVA: multiple response variables and multispecies interactions. In: Scheiner
S.M. and Gurevitch J. (eds), Design and Analysis of Ecological Field Experiments. Chapman & Hall,
London, pp. 94112.
Staender P.E. 1998. Spoor counts as indices of large carnivore populations: the relationship between
spoor frequency, sampling effort and true density. Journal of Applied Ecology 35: 378385.
Stahl P. and Leger F. 1992. Le chat sauvage (Felis silvestris). In: Artois M. and Maurin H. (eds),

Encyclopedie des carnivores de France BY. Societe Francaise pour lEtude et la Protection des
`
Mammiferes, Bohallard, Puceul.
 1079

J.L. and Santos T. 1993. Distributional pattern of insectivorous passerines in the Iberian forests:
Tellera
does abundance decrease near the border? Journal of Biogeography 20: 235240.
J.L. and Virgos
Tellera E. 1997. Distribution of an increasing roe deer population in a fragmented
Mediterranean landscape. Ecography 20: 247252.
J.L., Santos T. and Alcantara
Tellera M. 1991. Abundance and food searching intensity of wood mice
(Apodemus sylvaticus) in fragmented forests. Journal of Mammalogy 72: 183187.
Tew T.E. and Macdonald D.W. 1993. The effects of harvest on arable wood mice, Apodemus sylvaticus.
Biological Conservation 65: 279283.
Van Apeldoorn R.C., van Oostenbrink W.T., Winden A. and van der Zee F.F. 1992. Effects of habitat
fragmentation on the bank vole Clethrionomys glareolus in an agricultural landscape. Oikos 65:
265274.
Verboom J. and van Apeldoorn R. 1990. Effects of habitat fragmentation on the red squirrel Sciurus
vulgaris. Landscape Ecology 4: 171176.
E. and Casanovas J.G. 1997. Habitat selection of genet Genetta genetta in the mountains of
Virgos
central Spain. Acta Theriologica 42: 169177.
E. and Casanovas J.G. 1998. Distribution patterns of the stone marten (Martes foina) in
Virgos
Mediterranean mountains of central Spain. Zeitschrift fur Saugetierkunde
63: 193199.
E. and Casanovas J.G. 1999. Environmental constraints at the edge of a species distribution, the
Virgos
Eurasian badger (Meles meles): a biogeographic approach. Journal of Biogeography 26: 559564.
E. and Tellera
Virgos J.L. 1998. Habitat selection of the roe deer in Spain: constraints on a species
distribution. Canadian Journal of Zoology 76: 12941299.
E., Recio M.R. and Cortes
Virgos Y. 2000. Stone marten (Martes foina) use of different landscape types
Saugetierkunde
in the mountains of central Spain. Zeitschrift fur 65: 375379.