brown bear habitat use pattern in greece
DESCRIPTION
Επιστημονική εργασία που χρησιμοποίησε και δεδομένα που συλλέχθηκαν από την επιστημονική ομάδα του ΑΡΚΤΟΥΡΟΥ και σχετίζεται με πρότυπα κίνησης των αρκούδων στην ΕλλάδαTRANSCRIPT
PLEASE SCROLL DOWN FOR ARTICLE
This article was downloaded by: [HEAL-Link Consortium]On: 18 November 2008Access details: Access Details: [subscription number 772810500]Publisher Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Journal of Natural HistoryPublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t713192031
Brown bear (Ursus arctos L.) habitat use patterns in two regions of northernPindos, Greece - management implicationsGeorgios Mertzanis a; Athanasios S. Kallimanis b; Nikolaos Kanellopoulos c; Stefanos P. Sgardelis b;Athanasios Tragos a; Ilias Aravidis d
a NGO “Callisto”, Thessaloniki, Greece b Department of Ecology, Aristotle University of Thessaloniki,Thessaloniki, Greece c Forestry Service, Metsovo, Greece d Development Agency of Thessaloniki,Thessaloniki, Greece
Online Publication Date: 01 January 2008
To cite this Article Mertzanis, Georgios, Kallimanis, Athanasios S., Kanellopoulos, Nikolaos, Sgardelis, Stefanos P., Tragos,Athanasios and Aravidis, Ilias(2008)'Brown bear (Ursus arctos L.) habitat use patterns in two regions of northern Pindos, Greece -management implications',Journal of Natural History,42:5,301 — 315
To link to this Article: DOI: 10.1080/00222930701835175
URL: http://dx.doi.org/10.1080/00222930701835175
Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf
This article may be used for research, teaching and private study purposes. Any substantial orsystematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply ordistribution in any form to anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contentswill be complete or accurate or up to date. The accuracy of any instructions, formulae and drug dosesshould be independently verified with primary sources. The publisher shall not be liable for any loss,actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directlyor indirectly in connection with or arising out of the use of this material.
Brown bear (Ursus arctos L.) habitat use patterns in two regions ofnorthern Pindos, Greece – management implications
Georgios Mertzanisa*, Athanasios S. Kallimanisb, Nikolaos Kanellopoulosc,
Stefanos P. Sgardelisb, Athanasios Tragosa and Ilias Aravidisd
aNGO ‘‘Callisto’’, Thessaloniki, Greece; bDepartment of Ecology, Aristotle University ofThessaloniki, Thessaloniki, Greece; cForestry Service, Metsovo, Greece; dDevelopment Agencyof Thessaloniki, Thessaloniki, Greece
Conservation of brown bear (Ursus arctos L.) depends upon protecting thespecies’ core habitat and mitigating the detrimental consequences of infrastruc-ture. Therefore, a detailed knowledge of bear habitat suitability is needed. Twosites were systematically surveyed in Pindos (Gramos and Grevena) for bearsigns. Additionally, in Gramos six individuals were monitored with conventionalVHF telemetry; and in Grevena a further six individuals were monitored withGPS-Simplex collars. The aim was to analyse the habitat preference of bearswithin their home ranges, and construct habitat suitability maps, using EcologicalNiche Factor Analysis. Despite individual behavioural differences of bears andintrinsic differences between the sites, the two regional population units exhibitsimilar habitat preferences. The bears prefer agricultural land and areas near theedge between forests and open landscape formations (grasslands, cultivated fields,fallow land). Bears use alpine meadows less than randomly expected. Our resultswere used in zoning proposals for Gramos and for improvement of mitigationmeasures of Via Egnatia highway stretch in Grevena.
Keywords: brown bear; habitat use; habitat suitability; telemetry; Pindos; Greece
Introduction
The brown bear (Ursus arctos L.) range in Greece consists of two distinct nuclei
located in the Pindos mountain range (NW Greece) and the Rodopi mountain
complex (NE Greece) (Mertzanis 1992, 1994). The total area of continuous bear
range is 13,500 km2. The Pindos population encompasses two regional population
units: one close to the Albanian border (including Gramos, Voio, Mali-Madi and
Triklari Mts.) and a larger one in the southern parts of Pindos range, mainly in the
Grevena, Ioannina and Trikala prefectures (Mertzanis 2002). During recent years,
brown bear populations in Pindos have exhibited a clear trend of expansion towards
the eastern and southern parts of the species’ former range (Mertzanis et al. 2006).
The minimum brown bear regional population size in Gramos is estimated at
between 34 and 41 individuals (Defiggou 2003) and in Grevena at 44 individuals
(Drosopoulou and Scouras 2005). However, brown bear conservation status in
Greece remains critical and faces major threats from human-caused mortality,
habitat fragmentation, habitat loss and habitat degradation (Mertzanis 1999, 2002).
Management decisions with reasonable expectancy of success and appropriate
management of threatened brown bear populations require a constantly updated
*Corresponding author. Email: [email protected]; [email protected]
Journal of Natural History
Vol. 42, Nos. 5–8, February 2008, 301–315
ISSN 0022-2933 print/ISSN 1464-5262 online
# 2008 Taylor & Francis
DOI: 10.1080/00222930701835175
http://www.informaworld.com
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
knowledge and adequate information of the ecological requirements and biological
needs of the brown bear, allowing specific management recommendations (Servheen
1994). To this extent, biological information on brown bear spatial behaviour (home
range and habitat use) is of high value (Mano 1994; Wooding and Hardisky 1994).
At a coarse scale, the habitat preferences of brown bears for high productivity
forests and areas with low human pressure have already been reported in other areas
of the species range in Europe (Kobler and Adamic 2000). However, this more
general approach is not always helpful enough when it comes to environmental
planning decisions, such as zoning for conservation of core bear habitats, designing
the optimal alignment of a major highway and appropriatly locating necessary
mitigation measures. To accomplish such specific objectives, a spatially explicit,
detailed knowledge of the brown bear’s ecology is needed. The aim of this paper is to
investigate the brown bear habitat preference within a site the entire extent of which
is characterized as suitable habitat, i.e. the habitat preference within the animals’
home range. The final stage of the analysis is a map of the site, where habitat
suitability is presented as a gradient from low to high, thus enabling environmental
managers to reach practical recommendations.
Materials and methods
Study sites
Gramos (Figure 1) is located in the mountains of Gramos and NW Voio, part of
Northern Pindos range, and is delineated southwards by the Sarantaporos river,
eastwards by the Aliakmon river and northwards by the border with Albania. The
study site covers approximately 850 km2 at altitudes ranging between 600 and 2520 m
asl. The largest part of the area is located in the watershed of the Sarantaporos and
Aliakmon rivers.
Approximately 48% of the study site is covered by dense forests, 13% by partially
forested areas and 26% by grasslands. Forest vegetation is composed of: 27% black
pine (Pinus nigra J.F. Arnold), 42% oak (Quercus sp.) and 26% beech (Fagus sp.).
The remaining area is covered by agricultural land, rocky outcrops and bare ground.
Mean monthly temperatures range from a minimum of 23.1u C to a maximum of
27.4u C. Mean annual precipitation is 814 mm. Nearly all native European mammal
species are present in the area, including the wolf (Canis lupus (L., 1758)), roe deer
(Capreolus capreolus (L., 1758)), chamois (Rupicapra rupicapra (L., 1758)), wild cat
(Felis sylvestris Schreber, 1775), wild boar (Sus scrofa L., 1758) and otter (Lutra lutra
(L., 1758)). The study site is remote, characterized by low human density and
scattered human settlements. A high density of forest road network (1.5 km/km2)
related to timber activities and a relatively high level of hunting pressure are among
the human-related negative disturbance factors. The study site includes one
proposed NATURA 2000 site (Koryfes Orous Gramos; GR 1320002), one
Biogenetic reserve (Flabouro-Barouga – covering 1.3 km2) and four wildlife reserves
covering a total of 100.5 km2.
The second study site, Grevena (Figure 1), extends over approximately 800 km2
of a mixed forest and agricultural ecosystem and is located in the northeastern part
of Pindos mountain range (Lyggos and Hassia mountain massifs). Of this area 75%
is forest, 10% meadows (pasture lands), 14% agricultural land, and low population
302 G. Mertzanis et al.
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
density human settlements occupy 0.3% of the total area. Major forest vegetation
types comprise oak (Quercus sp.), black pine (Pinus nigra) and beech (Fagus sp.). The
area is characterized by a mosaic of dense forests, openings and small scale
cultivations. Altitude ranges between 500 m and 2200 m. Mean monthly tempera-
tures range from a minimum of 23.4u C to a maximum of 28.2u C. Mean annual
precipitation (1990–1994) is 589 mm. Native European mammal species present in
Gramos are also present in this area. Part of the study site is included in the North
Pindos National Park.
Study animals and data collection methods
In Gramos, six brown bears were captured (four adult males, one sub-adult male and
one adult female) during the period 1997–1999 and 2000–2002 with an Aldrich Foot
Snare trap type and were sedated with KHCl/xylazine (Rompun). For the heavier
bears the initial volume injected by a blowpipe was 750mg/3ml with a booster volume
Figure 1. Map of brown bear distribution range in Greece and location of the study areas in
Pindos Mountain range.
Journal of Natural History 303
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
of 600 mg/3 ml (administered via intra-muscular hand administered injection). For the
smaller bears the initial volume injected by a blowpipe was 250 mg/3 ml with a booster
volume of 200 mg/3 ml (administered via intra-muscular injection). Time to
anaesthesia was usually less than 15 minutes. Total immobilization time was approx.
30–45 min. The bears were fitted with conventional radiocollars (MOD-500 NH from
Telonics, USA). Telemetry data (radiolocations) were collected using a VHF TR-4
receiver (Telonics S.A., USA), a directional (‘‘three-element’’– ‘‘Yagi’’) type antenna
(Telonics, USA), a magnetic compass and a portable GPS. For each bearing, the
‘‘Raised Antenna – Null Signal Average’’ (RA – NS) technique was used to locate each
individual (Springer 1979; Kenward 2001). The exact position of each individual was
defined by the triangulation method with a minimum of three bearings (Kanellopoulos
2002; Mertzanis et al. 2005). Ground telemetry data were collected with a frequency of
1/1.1 days in order to ensure statistical independence among locations. This frequency
is considered satisfactory and has been used by many researchers in Europe for the
study of free-ranging bear populations (Roth 1983; Huber and Roth 1986, 1993;
Clevenger et al. 1990; Kaczensky et al. 1994; Mertzanis 1999, 2005). Duration of the
bear monitoring period ranged from 2.5 to 13 months (average 9.5 months).
LOCATE II and GIS (ArcInfo, ArcView) software were used for data processing
and plotting. Telemetry data were analysed by daily groups of bearings. Geographic
coordinates of the bearing points by GPS were obtained and processed through
LOCATE II (Pacer Computer Software) for triangulation. For every triangulation
the ‘‘error areas’’ were calculated with the ‘‘Maximum Likelihood’’ estimator that
determines bear position to an accuracy of 95% (Nams 1989, 1990). The average
error area for the accepted triangulations was fixed to 0.5 km2 (with a maximum of
1.9 km2), a small surface compared to the home range of each individual in the study
area, and therefore not affecting the evaluation of habitat use (Nams 1989).
In Grevena, six brown bear adults were caught (four males and two females)
during 2003–2004 using the same capture and anaesthesia protocols. These bears
were fitted with GPS (TELEVILT) radio-collars with remote download system (RX-
900 TELEVILT receiver) and 12 hours VHF beaconing. The collars were set up to
give 15–17 positions daily (the effective positions averaged 6–10 daily). Bear presence
was recorded day and night with a similar frequency. The Grevena topography is
relatively smooth without deep ravines that can dramatically affect the GPS
efficiency. In the study area there is no information about forest cover affecting the
GPS efficiency and given that the percentage of the GPS records under dense forest
is comparable with that of the ground survey records, there is no reason to assume
that forest cover affected GPS efficiency. For Grevena GPS telemetry data
coordinates (x,y) were directly mapped on geo-referenced photo-maps of the study
site. Duration of bear monitoring period ranged from one to 20 months (average
7.2 months).
Both sampling protocols were combined with systematic ground surveys for
the collection of signs of bear presence and activity. Total length of sampling
transects (600 km in Gramos site and 1,008 Km in Grevena site) followed the dense
(1.5 km/km2) forest road network which is a common feature in both study areas.
Ground surveys looked for any signs of bears, including footprints, hairs and scats
that were either on or near the road network. The ground surveys were limited to the
forest road network, and therefore this information was not used to assess the bears’
habitat preference regarding distance from human structures. However, the ground
304 G. Mertzanis et al.
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
surveys systematically covered the entire extent of the study areas and all vegetation
and land use classes in proportion to their total availability in the landscape. Finally
the ground survey samples were unrelated to the telemetry locations because they
were collected at independent time periods and relate to a different and considerably
larger number of individuals. From the hairs and scats collected in Grevena, DNAwas extracted and 44 individuals were identified (Scouras and Drosopoulou 2005).
The 12 animals tracked with radio telemetry were followed for periods of
between 1 and 20 months. If the data are separated by season the number of active
individuals or number of radiolocations will drop significantly, especially for winter
as there are only a few individuals with restricted activity and few records. Therefore,
analysing the data by season would weaken the power of the statistical analysis.
Therefore, at this stage our dataset does not allow seasonal analysis of the bears’
habitat preference, since the data were not collected for this purpose.
Landscape cover
As the source data for landscape structure quantification, raster maps of the land
cover of the study sites were used. The vegetation cover was mapped by the forestry
service (Forest Management Plans of 1994 at a resolution of 1:20,000) and updated
based upon orthorectified aerial photographs of the area. In the case of Gramos the
raster resolution was 2006200 m2, while in Grevena it was 50650 m2. The difference
of the raster resolution was mainly due to the difference in accuracy of the VHF
telemetry data collected in Gramos and the satellite GPS telemetry data from
Grevena. Land use was considered for each cell, classified as dense forest, partiallyforested area, grassland meadows, bare land, cultivated fields, infrastructure and
surface water. Woodlands were further analysed based upon the dominant species,
and forests classified as oak (Quercus sp.), beech (Fagus sp.), black pine (Pinus
nigra), white pine (Pinus leucodermis), fir (Abies borisii regis) and mixed broadleaved
species.
The analysis took into consideration not only the contents of each cell in the
raster, but also the landscape composition of its spatial neighbourhood. The
landscape composition was quantified in neighbourhoods of different size aroundeach cell and the bears’ habitat preference examined in relation to neighbourhood.
More specifically, the radius of the neighbourhoods were defined at 600 m and
1000 m at Gramos and at 250 m and 450 m for Grevena. Then, the percentage of the
neighbourhood area each land use class and each forest type covered was measured.
Third, the diversity of the land cover classes and of the vegetation types in the
neighborhood was measured using the Shannon diversity index. This process was
repeated for each cell in the raster.
The land cover maps included information regarding villages and streams. Foreach cell in our raster, the distance from the nearest village and stream was estimated
using the spatial analyst of ArcGIS 9. Topographical information included
elevation, slope and aspect, and was obtained from a 100 m Digital Elevation
Model (DEM).
Statistical analysis
For the analysis of habitat selection location data from bear sign was used, as well as
telemetry data. The analysis was performed at two levels. At the first level, for each
Journal of Natural History 305
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
background layer, the value at the location of the recorded bear presence was
estimated; the distribution of these values was compared with the distribution of
values for this layer in the entire landscape (presence versus availability) (Marcum
and Loftsgaarden 1980). The null hypothesis was that all values of the background
layer were equally suitable and that its frequency of use depended only on its
availability across the landscape. The entire extent of each study site was used as
background, because in both sites bears were observed throughout the area. Chi-
square tests were used to assess if the two distributions (background and species
presence) differed significantly statistically. When the two distributions differed
significantly, the Bonferroni 95% confidence intervals were estimated for the
frequency of bear use, for each class of the distribution, in order to identify which
classes were favoured or avoided (in accordance with Meriggi and Lovari 1996). If
the lower end of the 95% confidence interval for bear use was higher than the
availability then, this class was preferred; if the upper end of the 95% CI was lower
than the availability then this class was avoided, otherwise the class was used at
random according to its availability. The Bonferroni 95% confidence intervals were
estimated according to the formula:
pi{z0:05=2k
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pi 1{pið Þn
r
vpivpizz0:05=2k
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pi 1{pið Þn
r
where n is the total number of observations, pi is the proportion of bear presences in
the ith class of the variable, and z is the upper standard normal table value
corresponding to a probability tail area of 0.05/2k, with k the number of classes
considered.
This analysis allowed conclusions to be drawn regarding the preference or
avoidance for specific aspects of the landscape. It did not allow the building of a
habitat suitability map of the study site. To achieve that, a spatially explicit model
had to be constructed describing the relationship between the animals’ presence and
the habitat parameters. Three methods used in such modelling exercises are the
logistic regression (e.g. Mladenoff et al. 1999; Glenz et al. 2001), generalized additive
models (e.g. Guisan and Zimmermann 2000) and the classification tree analysis (e.g.
Jerina et al. 2003). The main drawback of these methods for the data is that they
require not only presence data, but also absence data. Such data (absences) may be
available for coarse scale analysis of habitat selection, e.g. bears are not observed in
the midst of towns or on the Greek islands; but they are impossible to collect for
analysis within the animals’ home range, as the one presented here. The lack of
recorded presence does not equal absence. The presence of the animal in a location
with suitable habitat may not be recorded either because of failure in detecting the
animal, or due to an animal’s effort to avoid the observers, or even due to the
idiosyncratic behaviour of each animal. Therefore an analysis method is needed that
relies only upon the recorded presences. Ecological niche factor analysis (ENFA)
was proposed to solve exactly this problem (Hirzel et al. 2002).
Ecological niche factor analysis starts off at the same point as the single variable
analysis, comparing the distribution of values where the animal is present with the
distribution of values in the background. ENFA relies on identifying differences in
the two distributions with respect to the mean (marginality) and with respect to the
standard deviation (specialization). This idea is applied to all variables in the study
306 G. Mertzanis et al.
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
related to topography, vegetation and land use, as well as the composition of the
spatial neighbourhood around each cell. The final habitat suitability is estimated
with the use of ordination techniques, such as principal component analysis. The
analysis estimates an overall marginality index, which expresses the difference
between the mean animal preference and the mean condition of the study site. Also,
the overall specialization index is estimated, which is a measure of the range of
environmental conditions the animal tolerates, compared to the range of values
recorded in the study site. For both indices, values close to 0 indicate a species which
can equally well utilize the entire area, and values close to 1 indicate a highly
specialized species that can only use a small part of the available landscape. The
analysis was performed using the Biomapper 3.0 software package (Hirzel et al.
2004).
In order to estimate the relative importance of the possible idiosyncratic
behaviour of individual bears, the analysis was repeated for each site, omitting one
individual at a time. The results were comparable, e.g. the marginality index for
Grevena varied from 0.28 to 0.32, and for Grammos from 0.51 to 0.56, and therefore
are not presented here.
Results
A total of 1299 radiolocations were obtained from sample (1) in the Gramos study
area, whereas 4564 radiolocations were obtained from sample (2) in the Grevena
study area (Table 1).
Telemetry data and total bear signs data (n5954 in Gramos and n51410 in
Grevena) suggest that bears can utilize the entire extent of both study sites, but they
appear to prefer some parts of the habitat more than others.
Even though the physiographic characteristics of the two sites appear similar
there are some significant differences. The Gramos region covers slightly higher
altitudes (600–2520 m asl) compared to the Grevena region (500–2200 m asl). In both
areas the bears utilize the entire area with the exception of the sites at high altitudes
Table 1. Bear telemetry sample main features from Gramos and Grevena study areas.
Bear Sex Age class Telemetry period Total radiolocations
Gramos sample
Nestoras male adult 10.5 months 276
Tsarnos male adult 10.5 months 266
Ondria female adult 2.5 months 68
Selinios male adult 13 months 423
Skotidas male adult 11 months 142
Zacharias male subadult 9 months 124
Grevena sample
Alexandros male adult 12 months 2060
Rohamis male adult 6 months 696
Maya female adult 2 months 164
Felicia female adult 20 months 1319
Aias male adult 1 month 175
Jeronymo male adult 2 months 150
Journal of Natural History 307
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
(.1900 m asl), which are systematically avoided. Both areas, and especially Gramos,
are characterized by rugged terrain. However, in both sites, a preference for flat
areas and an avoidance of steep slopes were observed (.70%).
In both areas there is a rich network of rivers and streams. In both cases the
bears prefer locations close to surface waters. In Gramos, where the average distance
from rivers in the site is 646 m (with a maximum distance of 6288 m), bear locations
are on average 492 m away (with a maximum of 3300 m). In Grevena, where the
average distance from rivers in the site is 800 m (with a maximum distance of
4400 m), bear locations are on average 750 m away (with a maximum of 3100 m).
In both cases, there is a constant low density human presence, with several
villages located within the study sites. In Gramos, where the average distance from
villages in the site is 1994 m (with a maximum distance of 7500 m), bear locations are
on average 1850 m away (with a minimum of 50 m and a maximum of 6000 m). In
Grevena, where the average distance from villages in the site is 3020 m (with a
maximum distance of 8800 m), bear locations are on average 2600 m away (with a
minimum of 250 m and a maximum of 8300 m). In the case of villages in both sites
bears avoid areas close to villages (,500 m) but show a strong preference for
locations at intermediate distances. Crosschecks through field observations have
identified these locations to represent key food resources for bears (e.g. orchards,
fields, cultivated fields with cereals and fallow land).
The two study sites are dominated by woodlands (Table 2). However, the
landscape composition of the open space in the two sites differs significantly. In
Gramos grasslands are the main open space formation (26% of the site), followed by
cultivated land (7%) and bare land also present (4%). In Grevena cultivated fields are
the main open space formations (14%), followed by grasslands (10%) with bare land
barely present (,1%). In both sites, bears use dense woodlands as often as randomly
expected. However the use of the other habitat types differs. In Gramos, a slight
preference for grasslands and bare land was observed, and a slight avoidance of
cultivated fields and partially forested areas. In Grevena the opposite is true, with a
Table 2. The availability of the different land use classes in the two study sites (Gramos and
Grevena) and the frequency of bear presence in each land use class.
Land use class Gramos (X2, df55, p,0.001) Grevena (X2, df55, p,0.001)
Landscape
availability
(%)
Bear
presence
(%)
95% CI for
bear use
Landscape
availability
(%)
Bear
presence
(%)
95% CI for
bear use
Dense forests 47.9 47.2 (1063) 46.7–47.7 61.6 65.0 (3881) 64.8–65.2
Partially forested 13.8 9.8 (220) 9.5–10.1 13.8 12.9 (769) 12.6–13.2
Grasslands 26.4 30.8 (693) 30.3–31.3 9.9 8.8 (525) 8.6–9.0
Agricultural
land
7.4 5.7 (127) 5.5–5.9 14.3 13.4 (799) 13.2–13.6
Bare land 3.4 6.2 (139) 6.0–6.4 0.2 0.0 (0) 0.0–0.0
Infrastructure 0.7 0.5 (11) 0.4–0.6 0.3 0.0 (0) 0.0–0.0
Note: Bear presence data include both telemetry and bear signs. For each land use class, the
95% Bonferroni confidence interval for the frequency of bear presence is also presented. The
number of brackets is the absolute number of radiolocations.
308 G. Mertzanis et al.
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
slight preference for cultivated fields and partially forested areas, and an avoidance
of grasslands.
The species composition of the woodlands was further analysed (Table 3). In
both sites oak (Quercus sp.) forests are the dominant type. In Gramos the second
commonest forest type is beech (Fagus sp.), followed by black pine (Pinus nigra) anda low presence (,1%) of other coniferous and mixed broadleaved forests. In
Grevena, black pine is the second commonest forest type, and all other forest types
(beech, fir, pine and mixed broadleaves) are present. In both sites bears show a
preference for oak forests, and an avoidance of beech forests. Black pines are slightly
preferred in Gramos and slightly avoided in Grevena; while mixed broadleaved
forests are avoided in Gramos and preferred in Grevena.
Finally, special attention was paid to the land cover composition of the
neighbourhood around each cell. The percentage of the area covered by openlandscape formations (grasslands, cultivated land and bare land) influences the
habitat preference of the bears. In both sites, animals prefer locations that include
such formations in their neighbourhood, but avoid sites that either have no open
formations or that have predominately open formations in their neighbourhoods
(.90%). In other words, the bears prefer sites near the edge of grasslands and
cultivated fields, but avoid going to the centre of large patches of grasslands and
cultivated fields. The animals do not display any preference regarding the land cover
diversity of their spatial neighbourhood.
Taking into consideration all the available information, concerning the
topography, the vegetation and the land use in the sites, as well as the composition
of the spatial neighbourhood around each grid cell, the ecological niche factor
analysis (ENFA) was performed for the presence of bears in the two sites. Figure 2
Table 3. The availability of the different forest types in the two study sites (Gramos and
Grevena) and the frequency of bear presence in each forest type.
Forest type Gramos (X2, df55, p,0.001) Grevena (X2, df55, p,0.001)
Landscape
availability
(%)
Bear
presence
(%)
95% CI
for bear
use
Landscape
availability
(%)
Bear
presence
(%)
95% CI
for bear
use
Oak (Quercus sp.) 26.4 20.3 (457) 20.0–20.7 36.3 38.5 (2302) 38.2–38.8
Black pine (Pinus
nigra)
16.7 19.2 (433) 18.8–19.6 29.7 32.3 (1931) 32.0–32.6
Beech (Fagus sp.) 16.4 17.2 (387) 16.8–17.6 3.6 2.4 (144) 2.3–2.5
White pine (Pinus
leucodermis.)
1.0 0.1 (2) 0.07–0.13 3.6 0.3 (18) 0.2–0.4
Mixed
broadleaves
0.6 0.2 (5) 0.1–0.3 1.5 3.7 (222) 3.5–3.9
Fir (Abies sp.) 0.6 0.0 (0) 0.0–0.0 0.7 0.5 (30) 0.4–0.6
Open landscape
formations
38.0 43.0 (969) 42.5–43.5 24.6 22.2 (1327) 21.9–22.5
Note: Bear presence data include both telemetry and bear signs. For open landscape
formations we include all non-forested areas, i.e. grassland, agricultural land and bare land.
For each forest type, the 95% Bonferroni confidence interval for the frequency of bear
presence is also presented. The number in brackets is the absolute number of radiolocations.
Journal of Natural History 309
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
shows the habitat suitability map produced for Gramos. In this case, the analysis
showed the species to score 0.55 in the marginality index, and 0.75 in the tolerance
index. This means that for bears a significant portion of the habitats in the site is of
high suitability but not all. As it can be seen in Figure 2, large parts of the landscape
remain underutilized. The model seems to capture the bear behaviour most
efficiently: approximately 77% of the bear presences are located in the areas
indicated as most suitable (the white areas of Figure 2) and only 13% in the least
suitable areas (black areas of Figure 2). Figure 3 shows the habitat suitability model
for Grevena, where a wider range of the available area is considered to be of high
suitability, and thus scores 0.37 in the marginality index, and 0.77 in the tolerance
index. In the case of Grevena the areas indicated by the ENFA model as most
suitable (white and light grey in Figure 3) host approximately 70% of the bear
presences, while the least suitable areas (black and dark grey in Figure 3) host only
6% of the presences. What is also interesting is that the model produced by the
analysis in the two sites is significantly different.
In Gramos the model is based on the composition of the land uses in the
neighbourhood around the point of interest, placing an emphasis on the availability
Figure 2. The habitat suitability map produced by the ecological niche factor analysis for the
Gramos study site. Habitat suitability is presented by a greyscale gradient. The lighter the
colour the more suitable habitat location for the bear.
310 G. Mertzanis et al.
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
of open landscape formations and partially forested areas and secondarily on the
species composition of the neighbourhood. In Grevena, the model is primarily based
on the distance from villages and then on the species composition of the
neighbourhood around the focal cell (mainly the availability of oak and pine
forests). However, even though the models differ, the outcomes have some common
features. In both cases, the alpine grasslands are identified as less suitable habitats,
while the forest areas adjoining open landscape formations in both models appear to
be highly suitable for the species.
Discussion
In this study, the habitat preference of bears within their home ranges was
analysed in two sites in the Pindos mountain range (north Greece). Habitat
selection is a scale dependent process and different characteristics of the landscape
Figure 3. The habitat suitability map produced by the ecological niche factor analysis for the
Grevena study site. Habitat suitability is presented by a greyscale gradient. The lighter the
colour the more suitable habitat location for the bear.
Journal of Natural History 311
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
influence habitat selection at different scales (McLoughlin et al. 2002). At the
coarse scale, the entire extent of the study sites consists of suitable habitat for the
brown bear as demonstrated by the bear signs and telemetry data found
throughout the region. Both sites are mountainous regions with rich vegetation
coverage (mainly oak forests) and moderate human presence (small scale
agriculture, human population density well below the national average and roads
of low traffic load). However, this coarse scale analysis identifying the entire area
as suitable habitat does not assist decision making for specific projects within the
study sites as it lacks appropriate refinement. In order to assist the environmental
management of the two sites, the preferences of the bears within their home ranges
had to be analysed, and areas of higher and lower suitability within a region that is
suitable had to be identified.
In the present analysis, location data from bear signs were treated in the same
way as telemetry data as in the Clevenger et al. study (1997). It was possible to treat
the bear signs data in this way since the sampling design covered the entire study area
and sampling effort was distributed in proportion to habitat type. The telemetry data
offered a detailed insight into the behaviour and habitat use of only a limited number
of individuals; and thus were spatially limited within the home ranges of these
individuals. The bear signs, on the other hand, covered the entire study site and
originated from a large number of individuals, but were biased towards the road
network. Therefore, the combination of the two data sets was the most appropriate
way to counterbalance the limitations of each data set. Radiolocation data were of
two categories, VHF collars in Gramos and satellite GPS collars in Grevena. The
two methods had significant differences. The VHF collars allowed for lower
accuracy of estimation of the exact location of the animal and considerably lower
frequency of observations, while simultaneously demanding much more manual
effort. As a result, even though the same number of individuals was tracked in both
sites for a comparable period of time, the Gramos data set consisted of significantly
fewer observations of lower accuracy, and its analysis was performed at a coarser
scale than in the case of Grevena. Furthermore, VHF telemetry demanded strenuous
field work that was possible only during day time, and night time records are absent.
However, the use of the VHF collars was obligatory, because during that period
(1997–2002), satellite collars were not available.
However, despite the limitations in the method of data acquisition, the individual
behaviour of bears (Couturier 1954) and the intrinsic differences of the two study sites,
the results for the two datasets are directly comparable. This supports the robustness
of the present model. In general, the brown bear regional population units studied in
Gramos and Grevena show habitat selection patterns appearing to be also related to
food availability. McLoughlin et al. (2002) recorded habitat selection by grizzly bears
that corresponded well with the spatial and temporal availability of food on the
landscape. In the present study, even though bears avoid the human presence, they
prefer to come close (approx. 1 km) to human settlements of the study sites in selected
locations, such as orchards and cultivated fields that are known to play an important
role as seasonal key food resources (Mertzanis 1992; Kanellopoulos et al. in press). In
such sites high trophic value plant species occur such as: wheat (Triticum sp.), corn
(Zea mays), wild apples, (Malus spp.), wild pears (Pyrus amygdaliformis and Pyrus
pyraster), berries (Rubus hirtus, Rubus canescens, Rubus idaeus), Sorbus spp. (Sorbus
torminalis, Sorbus domestica and more rarely Sorbus aucuparia), Rosa spp. (Rosa
312 G. Mertzanis et al.
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
canina, Rosa agrestis, Rosa pulverulenta and Rosa heckeliana), Prunus spp. (Prunus
domestica), chestnuts (Castanea sativa), and hazelnuts (Corylus avellana). Selective
habitat use in relation to forest associations when bears come close to human
settlements was also recorded for brown bears in Slovenia (Kobler and Adamic 2000).
In both study areas the bears were recorded to prefer open formations with rich grass
availability (grasslands in Gramos and fallow land in Grevena) and in both areas the
animals were observed to forage in these openings. Satellite telemetry data showed the
use of these forest clearings to take place mainly during the night, as it was observed
also in Alberta by Nielsen et al. (2004). Preference for woodlands with deciduous
forest species (i.e. oak forest formations), is reported for the bear populations of
central Italy as well (Possilico et al. 2004), underlining the trophic significance of
acorns and beech nuts in the bears’ diet. Preference of locations close to surface waters
such as small valleys and riparian vegetation is also a characteristic feature of bears’
habitat use in the two study areas and is also reported for other bear populations of
north America as well (Stratman et al. 2001).
The final product of the analysis was a map of habitat suitability for the brown
bears in the two sites. This spatially explicit information allowed identification of
zones of high conservation value in the Gramos site, proposed to become a protected
area. The results of this study may be used by decision makers on the need to
characterize zones of protection and human restrictions in the area. In the case of
Grevena, the output of this study found an immediate application in the design of
the Via Egnatia highway that is currently being constructed in the area. The detailed
map allowed identification of locations where the detrimental effects of the highway
are likely to be significant and mitigate them with appropriate constructions, such as
a ‘‘green bridges’’ and other specialized mitigation infrastructures, such as wildlife
underpasses and overpasses.
Acknowledgements
Telemetry research was possible in the framework of LIFE-Nature projects LIFE96NAT/GR/
03222 (1997–1999) and LIFE99NAT/GR/06498 (2000–2002) co-funded by the E.U.
(DGENV/D1/LIFE Unit) (beneficiary NGO ‘‘Arcturos’’) as well as in the framework of the
‘‘Monitoring project on impact evaluation of Egnatia highway construction (stretch 4.1) on
large mammals in the area of Grevena (2002–2005)’’. This project was co-funded by
EGNATIA ODOS SA, Hellenic Ministry of Environment, Planning & Public Works and the
EU (DG Regio). We thank the Forestry Services of Kastoria, Grevena and Kalambaka for
forestry data provision and the field team composed by D. Bousbouras, G. Giannatos,
K. Grivas, I. Isaak, G. Iliopoulos, Tr. Karahalios, Alex. Karamanlidis, M. Matthiopoulos,
H. Papaioannou, S. Pistofidis, Ath. Tragos, I. Tsaknakis, S. Riegler, A. Riegler, S. Tiliopoulos
and V. Vitaniotis for field data collection.
References
Clevenger A, Purroy F, Pelton M. 1990. Movement and activity patterns of a European brown
bear in the Cantabrian mountains, Spain. Proceeding of the International Conference of
Bear Research and Management 8:205–211.
Clevenger AP, Purroy FJ, Campos MA. 1997. Habitat assessment of a relict brown bear Ursus
arctos population in northern Spain. Biological Conservation 80:17–22.
Couturier M. 1954. L’ours brun. Grenoble, France: Artaud.
Journal of Natural History 313
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
Defiggou M. 2003. Mathematical investigation on brown bear population dynamics in Pindos
range using stage based models [Bachelor dissertation]. [Mytilene, Greece]: University of
the Aegean.
Drosopoulou E, Scouras Z. 2005. Genetic study of the brown bear sub-population in NE
Pindos-Grevena. In: Mertzanis G, editor. Monitoring and evaluation of impact of the
Egnatia highway construction (stretch 4.1 ‘‘Panagia-Grevena’’) on large mammals and
their habitats. Project final report, p. 114–197. (in Greek).
Glenz C, Massolo A, Kuonen D, Schlaepfer R. 2001. A wolf habitat suitability prediction
study in Valais (Switzerland). Landscape and Urban Planning 55:55–65.
Guisan A, Zimmermann NE. 2000. Predictive habitat distribution models in ecology.
Ecological Modelling 135:147–186.
Hirzel AH, Hausser J, Chessel D, Perrin N. 2002. Ecological-niche factor analysis: how to
compute habitat- suitability maps without absence data? Ecology 83:2027–2036.
Hirzel AH, Hausser J, Perrin N. 2004. Biomapper 3.0 [Internet]. Bern, Germany: University of
Bern; [cited 2007 November 11]. Available from: http://www.unil.ch/biomapper.
Huber D, Roth H. 1986. Home ranges and movements of brown bears in Plitvice Lakes
National Park, Yugoslavia. International Conference on Bear Research and
Management 6:93–97.
Huber D, Roth H. 1993. Movements of European brown bears in Croatia. Acta Theriologica
38:151–159.
Jerina K, Debeljak M, Dzeroski S, Kobler A, Adamic M. 2003. Modeling the brown bear
population in Slovenia – a tool in the conservation management of a threatened species.
ecological modelling 170:453–469.
Kaczensky P, Knauer F, Huber T, Jonosovic M, Adamic M. 1994. The Ljubliana-Postojna
highway – a deadly barrier for brown bears in Slovenia? Paper presented at: Symposium
1994 A Coexistance of Large Predators and Man; Poland.
Kanellopoulos N. 2002. Habitat selection of brown bear (Ursus arctos) in Gramos mountain
massif. MSc Thesis Dissertation. Aristotle University of Thessaloniki, Greece. 120 p.
Kanellopoulos N, Mertzanis G, Korakis G, Panagiotopoulou M. 2006. Selective habitat use
by brown bear (Ursus arctos L.) in northern Pindos, Greece. Journal of Biological
Research 5:23–33.
Kenward R. 2001. A manual for wildlife radio-tagging. San Diego (CA). Academic Press.
311 p.
Kobler A, Adamic M. 2000. Identifying brown bear habitat by a combined GIS and machine
learning method. Ecological Modelling 135:291–300.
Mano T. 1994. Home range and habitat use of brown bears in the southwestern Oshima
Peninsula, Hokkaido. International Conference on Bear Research and Management
9:319–325.
Marcum CL, Loftsgaarden DO. 1980. A nonmapping technique for studying habitat
preferences. Journal of Wildlife Management 44:963–968.
McLoughlin PD, Case RL, Gau RJ, Cluff HD, Mulders R, Messier F. 2002. Hierarchical
habitat selection by barren-ground grizzly bears in the central Canadian Arctic.
Oecologia 132:102–108.
Meriggi A, Lovari S. 1996. A review of wolf predation in southern Europe: does the wolf
prefer wild prey to livestock? Journal of Applied Ecology 33:1561–1571.
Mertzanis G. 1992. Aspects biogeographiques et ecologiques des populations helleniques
d’ours brun (Ursus arctos L.). Cas d’une sous-population du Pinde: application a la
conservation de l’espece et de son habitat. [dissertation]. [Montpellier, France]:
Universite de Montpellier.
Mertzanis G. 1994. Brown bear in Greece: distribution, present status – ecology of a northern
Pindus subpopulation. International Conference on Bear Research and Management
9:187–197.
314 G. Mertzanis et al.
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008
Mertzanis G. 1999. Status and management of the brown bear in Greece. In: Servheen C,
Herrero S, Peyton M, editors. Status survey and conservation action plan. Greece:
BEARS, IUCN/SSC Bear Specialist Group. p. 72–81.
Mertzanis G. 2002. Conservation of large carnivores in Greece: brown bear (Ursus arctos L.).
In: Psaroudas S, editor. Protected areas in the S. Balkans – Legislation, large carnivores,
transborder areas. Thessaloniki: Arcturos, DAC Project. p. 115–122.
Mertzanis G, Isaak I, Mavridis A, Nikolaou OI, Riegler S, Riegler A, Tragos A. 2005.
Movements activity patterns and home range of the brown bear (Ursus arctos. L.) in
Rodopi mountain range, Greece. Belgian Journal of Zoology 135(2):217–221.
Mertzanis G, Aravidis I, Tragos A, Psaroudas S. 2006. Review of brown bear distribution
status in Greece: range expansion & re-colonization of historical range. Poster
presentated at: Patras, Greece. 10th International Congress of Zoogeography and
Ecology of Greece & Adjacent Regions. 26–30 June 2006.
Mladenoff DJ, Sickley TA, Wydeven AP. 1999. Predicting gray wolf landscape recolonization:
logistic regression models vs. new field data. Ecological Applications 9:37–44.
Nams VO. 1989. Effects of radiotelemetry error on sample size and bias when testing for
habitat selection. Canadian Journal of Zoology 67:1631–1636.
Nams VO. 1990. Locate II User’s Guide. Truro (NS): Pacer Computer Software.
Nielsen SE, Boyce MS, Stenhouse GB. 2004. Grizzly bears and forestry I. Selection of
clearcuts by grizzly bears in west-central Alberta, Canada. Forest Ecology and
Management 199:51–65.
Posilico 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. Biological Conservation
118:141–150.
Roth HV. 1983. Home ranges and movement patterns of European brown bears as revealed
by radiotracking. Acta Zoologica Fenica 174:143–144.
Servheen C. 1994. Recommendations on the conservation of brown bears in Greece. LIFE-
Nature Project LIFE93NAT/GR/0180. Internal Report, 24.
Springer JT. 1979. Some sources of bias and sampling error in radio triangulation. Journal of
Wildlife Management 43(4):926–935.
Stratman M, Alden D, Pelton M, Sunquist M. 2001. Habitat use by black bears in the
sandhills of Florida. Ursus 12:109–114.
Wooding JB, Hardisky TS. 1994. Home range, habitat use and mortality of black bears in
North-central Florida. International Conference on Bear Research and Management
9:349–356.
Journal of Natural History 315
Downloaded By: [HEAL-Link Consortium] At: 07:05 18 November 2008