The impact of habitat fragmentation by anthropogenic
infrastructures on wolves (Canis lupus)
Dissertation to obtain the degree
Doctor Rerum Naturalium (Dr. rer. nat.)
at the Faculty of Biology and Biotechnology,
Ruhr-University Bochum
International Graduate School of Biosciences
Ruhr-University Bochum
SG Behavioural Biology and Biology Education,
Professor W.H. Kirchner
submitted by
Julia Eggermann
from Witten, Germany
Bochum
April 2009
Der Einfluss der Habitatfragmentierung durch anthropogene
Strukturen auf den Wolf (Canis lupus)
Dissertation zur Erlangung des Grades
eines Doktors der Naturwissenschaften
der Fakultät für Biologie und Biotechnologie
der Ruhr-Universität Bochum
Angefertigt in der
AG Verhaltensbiologie und Didaktik der Biologie,
Prof. W.H. Kirchner
Vorgelegt von
Julia Eggermann
aus Witten
Bochum
April 2009
ERKLÄRUNG
Hiermit erkläre ich, dass ich die Arbeit selbständig verfasst und bei keiner anderen Fakultät
eingereicht und dass ich keine anderen als die angegebenen Hilfsmittel verwendet habe. Es
handelt sich bei der heute von mir eingereichten Dissertation um fünf in Wort und Bild völlig
übereinstimmende Exemplare.
Weiterhin erkläre ich, dass digitale Abbildungen nur die originalen Daten enthalten und in
keinem Fall inhaltsverändernde Bildbearbeitung vorgenommen wurde.
Bochum, den 14.04.2009
_____________________________________
(Unterschrift)
4
Contents
CONTENTS
1
GENERAL INTRODUCTION ......................................................................................................... 6
1.1
1.2
BACKGROUND..............................................................................................................................6
OBJECTIVES AND STRUCTURE OF THE THESIS..............................................................................8
2
A COMPARISON OF MONITORING METHODS FOR ESTIMATING WOLF (CANIS
LUPUS) ABUNDANCE............................................................................................................................11
2.1
ABSTRACT .................................................................................................................................11
2.2
INTRODUCTION ..........................................................................................................................11
2.3
MATERIAL AND METHODS ........................................................................................................13
Study area...........................................................................................................................................13
Data collection ...................................................................................................................................13
Data analysis......................................................................................................................................14
2.4
RESULTS ....................................................................................................................................16
2.5
DISCUSSION ...............................................................................................................................17
2.6
ACKNOWLEDGEMENTS ..............................................................................................................19
3
HABITAT PREFERENCES OF WOLVES (CANIS LUPUS) IN CENTRAL NORTHERN
PORTUGAL..............................................................................................................................................20
3.1
3.2
3.3
3.4
3.5
3.6
4
ABSTRACT .................................................................................................................................20
INTRODUCTION ..........................................................................................................................20
MATERIAL AND METHODS .........................................................................................................21
RESULTS ....................................................................................................................................23
DISCUSSION ...............................................................................................................................25
ACKNOWLEDGEMENTS ..............................................................................................................27
VALIDATION OF A WOLF (CANIS LUPUS) HABITAT MODEL.........................................28
4.1
ABSTRACT .................................................................................................................................28
4.2
INTRODUCTION ..........................................................................................................................28
4.3
MATERIAL AND METHODS ........................................................................................................31
Study area...........................................................................................................................................31
Data collection ...................................................................................................................................31
Data analysis......................................................................................................................................31
4.4
RESULTS ....................................................................................................................................36
4.5
DISCUSSION ...............................................................................................................................41
4.6
ACKNOWLEDGEMENTS ..............................................................................................................44
5
THE RELATIONSHIP BETWEEN HABITAT QUALITY AND STRESS HORMONES IN
WOLVES (CANIS LUPUS).....................................................................................................................45
5.1
ABSTRACT .................................................................................................................................45
5.2
INTRODUCTION ..........................................................................................................................45
5.3
MATERIAL AND METHODS .........................................................................................................47
Study area...........................................................................................................................................47
Data collection ...................................................................................................................................48
Fecal analysis.....................................................................................................................................49
Data analysis......................................................................................................................................51
5.4
RESULTS ....................................................................................................................................52
5.5
DISCUSSION ...............................................................................................................................53
5.6
ACKNOWLEDGEMENTS ..............................................................................................................56
Contents
5
6
GENETIC MONITORING OF A PORTUGUESE WOLF (CANIS LUPUS) POPULATION
BY MICROSATELLITE ANALYSIS....................................................................................................57
6.1
ABSTRACT .................................................................................................................................57
6.2
INTRODUCTION ..........................................................................................................................57
6.3
MATERIAL & METHODS ............................................................................................................59
Study area...........................................................................................................................................59
Sample collection & conservation......................................................................................................60
DNA extraction & amplification ........................................................................................................60
Data analysis......................................................................................................................................63
6.4
RESULTS ....................................................................................................................................65
6.5
DISCUSSION ...............................................................................................................................68
6.6
ACKNOWLEDGEMENTS ..............................................................................................................71
7
PREDICTING SUITABLE WOLF (CANIS LUPUS) HABITAT IN GERMANY ...................73
7.1
ABSTRACT .................................................................................................................................73
7.2
INTRODUCTION ..........................................................................................................................73
7.3
METHODS...................................................................................................................................75
Study area...........................................................................................................................................75
Model development ............................................................................................................................75
Model application ..............................................................................................................................76
7.4
RESULTS ....................................................................................................................................78
7.5
DISCUSSION ...............................................................................................................................81
8
GENERAL DISCUSSION...............................................................................................................85
8.1
HUMAN IMPACT ON WOLVES .....................................................................................................85
Comparability of diverse monitoring techniques ...............................................................................85
Factors influencing habitat selection.................................................................................................86
Human impact on stress hormones ....................................................................................................87
Roads as barriers to dispersal ...........................................................................................................88
8.2
GERMANY AS A WOLF COUNTRY ...............................................................................................89
Suitable habitat ..................................................................................................................................89
Connectivity of suitable habitat patches ............................................................................................90
Potential population size....................................................................................................................90
Areas with elevated risks for human-wolf conflicts............................................................................90
8.3
CONCLUSION .............................................................................................................................91
SUMMARY ...............................................................................................................................................92
ZUSAMMENFASSUNG ..........................................................................................................................95
REFERENCES..........................................................................................................................................98
ACKNOWLEDGEMENTS .................................................................................................................. 114
CURRICULUM VITAE........................................................................................................................ 115
General introduction
1
1.1
6
GENERAL INTRODUCTION
Background
Originally, wolves (Canis lupus) inhabited the whole northern hemisphere from North America
to Europe and Asia. They survived in as diverse habitats as tundra, taiga, semi-deserts, tropical
and temperate forests. However, with increasing human population numbers, human mobility
and industrialization, their formerly inaccessible habitats were transformed into anthropogenicused land, which caused men and wolf to become harsh competitors for resources. This was
followed by an era of wolf demonization and uncontrolled killing. Until the end of the 19th
century, wolves disappeared from most settled regions (Boitani 1995) and, except from small
isolated populations in the Iberian Peninsula, Italy and the Balkans (Promberger & Schröder
1993), Europe did not hold wolves anymore. This changed in the middle of the 20th century,
when people’s attitude towards wolves altered, which resulted in strictly protecting the wolf in
several European countries by the end of the 20th century. As a consequence, wolf populations
started to recover and wolves slowly reoccupy their former range.
In 1995 the first wolf immigrated to eastern Germany, from a small wolf population in western
Poland (Reinhardt & Kluth 2006). It settled in north-eastern Saxony and a colonizing population
of currently 3-4 wolf packs evolved. The people’s attitude towards the wolf’s return to Germany
is ambiguous (Kaczensky 2006). Conflicts arose with livestock farmers, hunters and persons that
fear the wolf. In other countries, where wolves were extinct and subsequently returned, similar
problems exist. In Sweden for example, hunters have the most negative attitudes towards wolves
(Ericsson & Heberlein 2003). Moreover, Ericsson and Heberlein (2003) could support the
assumption that wolf depredation negatively influences the general attitude towards the wolf. In
Finland, where wolf numbers increased from 50 individuals in 1995 to 600 in 2002 and
depredation on reindeer increased substantially, the wolf conflict between the various
stakeholders actually resulted in Finland being charged to the Court of Justice of the European
communities (Bisi et al. 2007). Many more examples of human-wolf conflicts worldwide are
available and most of them are based on competition for common prey, safety of livestock, pets
and humans (Bjerke et al. 1998, Linnell et al. 2002, Fritts et al. 2003, Kojola et al. 2004).
Therefore, the general question arose, whether wolves are able to live next to humans at all, if it
makes sense to support their recovery in semi-wild landscapes, and how they can be managed to
minimize human-wolf conflicts.
General introduction
7
The high adaptability of wolves to the coexistence with humans became apparent in regions
where wolves survived in corn fields next to human settlements (Ryabov 1987, Blanco et al.
2005), foraged on dump sites (Ciucci et al. 1997) and even crossed villages during night
(Promberger et al. 1997). Moreover, in several studies they seemed to adapt their activity pattern
to the presence of humans in their home ranges (Vilá et al. 1995, Ciucci et al. 1997, Theuerkauf
et al. 2003, Kusak et al. 2005), though this finding is still being discussed (Theuerkauf et al.
2007). Linnell et al. (2001) stressed the idea that wolves can coexist with humans, if a proper
conservation management strategy is followed. Rather, wolves in densely populated Europe will
only be able to survive if they can cope with humans (Boitani & Ciucci 1993, Chapron et al.
2003). Therefore, assessing the potential for harmonic human-wolf coexistence is the key
question for wolf conservation in Europe.
Several approaches have been used to assess human disturbance on animals, one of them being
the before mentioned analysis on activity patterns. Furthermore, analyses of stress hormones
have been conducted in several species to discover a potential increase during exposure to
human disturbance (e.g. elk, Creel et al. 2002; Pampas deer, Pereira et al.2006; black grouse,
Arlettaz et al. 2007; marten, Barja et al. 2007). A study on wolves showed, that individuals
constantly exposed to snowmobile activity, exhibited higher values of stress hormones than
individuals living in more remote areas (Creel et al. 2002). However, this study was conducted in
sparsely populated regions of two national parks in the USA, and wolves, used to the presence of
humans, might not react with an elevation in stress hormones. Still another approach to study
anthropogenic impact on animals, which currently is gaining increasing attention, is the
development of habitat utilization models. For such models, presence data of a species are set in
relation to natural and anthropogenic factors of its habitat and subsequently, they can be used to
classify areas as suitable or unsuitable habitat for this species (Anderson et al. 2003). There are
various statistic modeling approaches, such as logistic regression and Mahalanobis distance.
They differ in the input needed (presence-absence data versus presence-only data), as well as in
the output produced (actual probability of cells to present suitable habitat versus spatial patterns).
For wolves, habitat modeling was applied to regions in North America and Europe to explain
wolf presence (Mladenoff et al. 1995, Corsi et al. 1999, Grilo et al. 2002, Cayuela et al. 2004,
Jędrzejewski et al. 2004, 2005, Santos et al. 2007) and in a few studies also to predict potential
wolf habitats (Mladenoff & Sickley 1998, Glenz et al. 2001, Jędrzejewski et al. 2008) and habitat
connectivity (Rodriguez-Freira & Crecente-Maseda 2008).
General introduction
1.2
8
Objectives and structure of the thesis
The general aim of this study was to determine the impact of anthropogenic altered landscapes
on wolves, assess the ability of wolves to coexist with humans, build a model explaining the
wolves’ requirements for survival and apply the model to a currently virtually wolf-free country
to predict its further resettlement.
To gain universal knowledge in ecological and behavioral science on animals, it is essential to do
research on several populations living in different regions under diverse circumstances.
However, most studies are rather locally organized incorporating little or no collaboration with
other research units. Therefore, results of these studies are not comparable with each other,
because of variations in sampling techniques, sampling design or methods used for analysis. The
range of selectable sampling techniques is often limited, because of particular characteristics of
the study area (dry or humid, hot or cold, vegetated or not, etc.) or the studied species (available
or rare, easy to observe or cryptic, easy or difficult to catch, etc.). Consequently, it is important
to test the comparability of different sampling techniques in places where various methods can
be applied. For this reason, in chapter 2 (A comparison of monitoring methods for estimating
wolf (Canis lupus) abundance) data on wolf presence were collected in a study area in the southeast of Poland by telemetry, snow-tracking and scat surveys, and compared the results of those
three monitoring techniques to assess their comparability.
Subsequently in chapter 3 (Habitat preferences of wolves (Canis lupus) in central northern
Portugal) data on wolf presence were collected by scat surveys in a study area in central
northern Portugal and places with wolf presence and places with wolf absence in terms of land
cover, prey abundance and human presence characterized. By comparing the characteristics of
presence and absence areas, the conditions needed by wolves to survive and reproduce were
identified. Thereby, the impact of anthropogenic disturbance on wolves could be determined. To
render these findings applicable to other study regions, I developed a habitat utilization model by
logistic regression, which included those variables with the maximum importance to wolves.
For conservation management it is of particular importance to have simple models with a wide
applicability at hand (Côté & Reynolds 2002), to predict suitable habitat for a species in
currently unoccupied regions that might differ from the region for model development. To know
whether a model is transferable and can be applied to other geographical regions, it should be
validated with a second dataset, gathered in a region differing in natural (land cover, climate,
prey composition and availability) and anthropogenic parameters. Factors, influencing the
General introduction
9
studied species in both study regions, are likely to be of higher importance to it than factors
differing between presence and absence areas only in one study region. This kind of model
evaluation by independent datasets reduces the risk of false conclusions; however, such
independent datasets are seldom available (Mladenoff et al. 1999, Pearce & Ferrier 2000). In this
study, data on wolf presence and habitat variables in two study areas in Portugal and Poland
were gathered and therefore the beforehand developed habitat utilization model could be
validated on a second wolf population in chapter 4 (Validation of a wolf (Canis lupus) habitat
model).
Mostly, the negative impact of a particular parameter on a species can be detected only, when a
population decline already occurred. This, however, is often too late to rescue a population.
Therefore, it would be valuable to discover a tool for detecting such negative impact in advance.
This tool might have been found some years ago, when the stress level of an animal, determined
by the amount of stress hormones secreted by the adrenal cortex, could be measured noninvasively. Nowadays, the stress hormones (glucocorticoids) can be detected not only in the
blood, but as well in saliva, urine and feces. This makes the method more applicable, particularly
for cryptic and protected species as the wolf that cannot easily be captured for taking a blood
sample. This method was used in chapter 5 (The relationship between habitat quality and stress
hormones in wolves (Canis lupus)) to assess the stress level of wolves living in the Polish study
area. To estimate the impact of human presence on the wolves’ stress level, average amounts of
glucocorticoids from wolves living in densely populated and less forested areas were compared
to wolves living in more remote places.
Another non-invasive approach to identify negative anthropogenic influences, that decrease the
viability of wolf populations, is a molecular biological one. By means of genetically analyzing
fecal samples at four microsatellite loci, in chapter 6 (Genetic monitoring of a Portuguese wolf
(Canis lupus) population by microsatellite analysis) I tested whether roads of different size and
traffic volume act as barriers to dispersing wolves. Moreover, through determining expected and
observed heterozygosity, the dataset was explored for signs of a possible sub-structuring of the
wolf population, which might generate isolated subpopulations. This process is a major concern
in conservation biology, as isolated populations, without any genetic exchange, suffer genetic
depression due to inbreeding, which makes them more vulnerable for extinction processes. In
addition, the reliability of this monitoring method for the assessment of demographic parameters,
such as population size, pack size and dynamics, as well as pack distinction, was examined
General introduction
10
In the last chapter, chapter 7, (Predicting suitable wolf (Canis lupus) habitat in Germany) the
previously in chapter 3 developed and in chapter 4 validated habitat utilization model was
applied to Germany. Thereby, I evaluated whether there is suitable habitat for the wolf in
Germany, located these potential wolf areas and assessed the connectivity between smaller
patches of suitable habitat. Thus, it becomes possible to identify priority areas for conservation
actions and the type of management action needed can be determined (e.g. habitat protection,
habitat restoration or construction of crossing structures for facilitating dispersal and reducing
road mortality). By measuring the potential wolf areas and applying knowledge about home
range size of wolf packs and pack size of European wolves, the potential size of a German wolf
population was calculated. Moreover, areas with an elevated risk for human-wolf conflicts
within Germany, characterized by low road density and high livestock density, were identified.
This knowledge is important for the management of a conflictive species as the wolf, by showing
priority areas for livestock protection measures (e.g. livestock guarding dogs or electric fences).
A comparison of monitoring methods for estimating wolf abundance
2
2.1
11
A COMPARISON OF MONITORING METHODS FOR ESTIMATING WOLF
(Canis lupus) ABUNDANCE
Abstract
The aim of our study was to investigate whether the non-invasive scat survey is appropriate for
monitoring wolf presence and home ranges. If robust, such indirect tracking approach could
make predictive modeling and long-term monitoring of wolf distribution and habitats more
practicable than with radio-telemetry. We analyzed the marking patterns of wolves within their
home ranges. Subsequently, we compared scat surveys, radio-telemetry, and snow-tracking
regarding their accuracy and reliability in predicting the distribution of wolf pack home ranges.
We tested time-efficiency of the three methods by a sub-sampling approach. Wolves marked the
center and border of their home ranges more frequently than the rest. Consequently, the pack’s
home ranges could be distinguished by scat surveys. The distribution of home ranges estimated
by scat surveys matched the results obtained by radio-telemetry and snow-tracking. Moreover,
the sub-sampling demonstrated that scat surveys were the most time-efficient. Therefore we
suggest the use of non-invasive techniques for monitoring wolf populations on a large scale;
however for obtaining more detailed information on activity and habitat use, telemetric studies
still present the optimal approach.
2.2
Introduction
For the efficient conservation and management of wildlife it is crucial to assess the abundance,
distribution, habitat use, and population trends of the target species. This is of particular
importance for endangered species and populations. Failures in the assessment of the population
status can affect management decisions and consequently the survival of the respective species
(Reed & Blaustein 1997). For that reason efficient and reliable monitoring methods are needed.
Estimating the abundance of carnivores can be particularly difficult as they are often secretive,
occur at low densities, have large home ranges, and show nocturnal or crepuscular activity
patterns (e.g. wolves, Eggermann et al. 2009). Basically two approaches exist to monitor wildlife
populations (Wilson & Delahay 2001). Either the absolute abundance is measured by directly
counting the animals or a relative abundance is estimated by counting indices of the animals
related to their abundance. The former one is more valuable, as it reflects the actual population
size of the studied animals. However, it is applicable only to species that are easy to observe
A comparison of monitoring methods for estimating wolf abundance
12
directly. Otherwise, it requires invasive methods including capture and handling of the animals,
which is expensive and time consuming. Measures of relative abundance rely upon an index,
such as the number of droppings per km². Those methods assume that the frequency of field
signs per standardized unit of sampling effort is related to the number of animals present in the
study area. However, these indices of abundance rarely have been calibrated against true density
estimates and no studies have attempted to evaluate the precision of the calibration (Gurnell et al.
2004; Gompper et al. 2006). Therefore, there is an urgent need for assessing the accuracy,
sensitivity, and reliability of noninvasive monitoring methods in a variety of habitats and for a
range of species. Mammalian species have been surveyed by direct observations of the animal
itself or by indirect observations of the animal’s activities, such as scats, tracks, and breeding
dens. For the monitoring of carnivores classically invasive methods have been used. Radiotelemetry is one of the most common techniques, but next to the disadvantage of being invasive,
it is subject to the above mentioned restrictions of high time and financial needs (Weckerly &
Ricca 2000). Several noninvasive monitoring methods may aid in overcoming those
disadvantages. Carnivores leave characteristic signs, such as footprints, droppings or den sites
that reveal their presence, and a number of monitoring methods have been developed based on
them. Tracking animals by following their footprints in mud, sand or snow, is probably the
oldest method used for discovering mammal’s presence (Bider 1968). Snow-tracking has proven
to be adequate for the census of carnivores (Ballard et al. 1995, Oehler & Litvaitis 1996, Ciucci
et al. 2003, Patterson et al. 2004, Golden et al. 2007). The search for scats is as well an old
technique for surveying carnivores. In some species scats and urine are used as territorial marks
(Vilá et al. 1994, Hutchings et al. 2001). As the production and preservation of marks represents
a cost for the animal, it does not mark the whole territory, but it regularly deposits them at
specific, predictable sites (Zub et al. 2003). Therefore, they are easy to find for researchers.
Another advantage is that scats may persist for several months (Kohn et al. 1999). However,
even though the collection of scats for monitoring purposes has increased recently (Taberlet et al.
1999), its usefulness for surveying population size is still controversial. Various methods are
often assumed to provide the same information, but only a few studies have investigated whether
they really do so (Litvaitis et al. 1985, Alexander et al. 2005, Gompper et al. 2006). Next to the
question if different monitoring methods lead to the same results, the time- and cost-efficiency of
the various techniques has to be considered (Franco et al. 2007). Indirect methods usually are
less expensive and easier to apply than direct methods such as live trapping of animals. Thus,
A comparison of monitoring methods for estimating wolf abundance
13
they are more efficient if they provide comparable information at lower costs. Where the
calibration of an indirect monitoring technique is not feasible, the parallel use of various methods
is advisable (Mahon et al. 1998). The aim of our study was to compare the results of monitoring
wolf abundance by telemetry, snow-tracking, and scat surveys. Our focus was on assessing the
reliability of the three methods, as well as their time-efficiency.
2.3
Material and Methods
Study area
The study was conducted in an area of 1000km² in the Bieszczady Mountains region (49°19’49°50’N, 22°15’-22°45’E), south-eastern Poland. The hilly landscape forms the foothills of the
Bieszczady Mountains and is forested in 62%. The climate has a mountainous character with
continental influences. Annual precipitations average 800-1200mm and snow cover lasts for 90140 days with a depth of 10-40cm. The main prey species of wolves in this area are red deer
(Cervus elaphus), roe deer (Capreolus capreolus) and wild boar (Sus scrofa; Gula 2004, Gula
2008). Population density averages 34 inhabitants per km²; with higher densities in the northern
part of the study area and concentrated around a bigger town in the centre. Road density is
0.59km km-².
Data collection
We radio-tracked three wolves (one male and two females) of three different packs from spring
2002 until autumn 2006 for periods of 1-3 years. The wolves were caught in Belisle foot snares,
tranquilized with a mixture of ketamine and xylazine and fitted with VHF radio-collars. We
localized them by ground-triangulation during 55 continuous 24h radio-tracking sessions in 15min intervals (described in detail in Theuerkauf & Jędrzejewski 2002). The average location
accuracy was 250m (Theuerkauf et al. 2007). In each winter season from 2000/2001 until
2005/2006 we additionally followed the wolves of four packs by snow-tracking. After recent
snowfall we searched for wolf tracks in the snow along forest roads. On 146 days we found
tracks and followed them for a total of 461km. For the analysis of the snow tracking data the
wolf tracks were transformed into discrete data points in intervals of 100m. In 2006 and 2007 we
included transects in search of wolf scats into our research. We drove or walked a total of 550km
14
A comparison of monitoring methods for estimating wolf abundance
spatial marking density index
mean number of scats
250
(A)
200
150
100
50
0
50%
75%
(C)
0
-15
100%
25%
1.6
50%
75%
100%
50%
75%
100%
2.5
(B)
temporal marking density
index
marking density [scats/km²]
25%
15
1.2
0.8
0.4
0
(D)
2
1.5
1
0.5
0
25%
50%
75%
100%
home range zone [% of locations]
telemetry
snow tracking
25%
home range zone [% of locations]
telemetry
snow tracking
Fig 1: Marking behavior of wolves, assessed by scat surveys along predefined transects in 2006 and
2007. Presented are the marks (mean values of all packs, with confidence intervals) in four distinct home
range zones, each containing 25% of all locations obtained by either telemetry or snow-tracking. Marking
intensity is expressed as the mean number of scats found in each zone (A), the density of scats in each
zone (B), the ratio of expected to found scat density in each zone (C), and the quantity of scats per unit of
time, wolves spent in the respective zone (D).
of pseudorandom transects along forest paths and rarely used roads equally distributed
throughout the whole study area.
Data analysis
To analyze the marking behavior of wolves we first calculated home ranges by minimum convex
polygon method (MCP) based on data obtained by telemetry and snow-tracking, respectively.
We then divided the home ranges in four zones, with each holding 25% of data points. Within
the home range of each wolf pack we calculated the total number of scats found in the discrete
zones and calculated the mean from all packs. As the four zones were of different size and thus
had a different probability of holding wolf scats, as a next analysis we calculated an abundance
15
snowtracking [% of locations]
A comparison of monitoring methods for estimating wolf abundance
telemetry [% of locations]
6
(A)
4
2
0
0
2
4
scat surveys [% of locations]
6
6
(B)
4
2
0
0
2
4
6
scat surveys [% of locations]
Fig 2: Relationship between wolf distribution, assessed by scat surveys on transects (n = 484 1km
transects) and wolf distribution, assessed by (A) telemetry (n = 55 continuous 24h sessions) and (B)
snow-tracking (n = 149 wolf tracks), respectively. Presented are the percentages of wolf locations in the
244 sampling squares.
index given as scats per km². To clarify the relationship between zone size and quantity of scats,
we subtracted the number of scats, expected for the respective zone size, from the observed
number of scats. To test whether particular home range zones were used more often for scat
deposition, we calculated a marking intensity defined by the ratio of observed scats to expected
scats (25% of all scats in each zone). For the direct comparison of all three monitoring methods,
we divided the study area in 244 squares of 2x2km each. For each method we calculated the total
number of wolf locations in each square and consecutively the percentage of those locations
from all wolf locations in the whole study area. For the pair-wise comparison of the data
obtained by scat surveys to the data obtained by telemetry and snow-tracking, we used only
squares for which data of both methods were available. For the sub-sampling approach we first
defined basic working units representing one sampling day to standardize data obtained by the
diverse monitoring methods. One wolf track followed in the snow (independent of its length),
one 8h shift of a 24h telemetry session, or 10km of scat surveys along transects were defined as
the basic working unit. We chose a defined number of sampling days randomly from the total
dataset and counted the number of squares with wolf locations. We did this for each monitoring
method. We conducted sub-sampling in steps of five sampling days. For each step we randomly
chose five sub-samples and calculated the mean number of squares with wolf locations.
A comparison of monitoring methods for estimating wolf abundance
2.4
16
Results
Home ranges calculated by the MCP method averaged 127km² (76km² - 233km²); however,
home ranges based on snow-tracking data were smaller than the ones based on telemetry. The
size of home range zones increased from the central zone to the outermost one. Zone 1 had a
mean size of 10.5km² (9% of the total home range), zone 2 15.3km² (14% of the total home
range), zone 3 23.7km² (19% of the total home range) and zone 4 77.7km² (58% of the total
home range). We found 268 scats along the transects. They were not distributed equally
throughout the study area. The total number of scats was lowest in the central zone and increased
towards the external zones (Fig 1 (A)). However, related to the size of the respective zone, the
scat density was highest in zone 1 and decreased towards zone 4 (Fig 1 (B)). In figure 1 (C) this
result becomes most obvious. The scat density was higher than expected for the zone size in zone
1 and 2, the same as expected in zone 3 and lower than expected in zone 4. The marking
intensity, in relation to time spent in the four zones, was higher at the edges of the home range
and lower in zone 1-3 (Fig 1 (D)). This difference occurred, however, only in the analysis based
on telemetry data; based on snow-tracking data the marking intensity was equal in all four zones.
The direct comparison of the monitoring methods revealed a linear relationship among all of
them (Fig 2). The wolves were detected more often by telemetry or snow-tracking in squares,
where more scats were found during scat surveys. The concordance between scat survey data and
telemetry data was significant (Kendall’s test of concordance, W = 0.169, χ² = 27.303, p =
0.0001). The same strong concordance existed between transect data and snow-tracking data
(Kendall’s test of concordance, W = 0.103, χ² = 12.236, p = 0.0001).
For all three monitoring methods the detected wolf occurrence area increased with rising
sampling effort (Fig 3). This increase showed a logarithmic pattern, with a fast boost in the
beginning of sampling and a flattened line when including more data points into the analysis.
However, only the telemetric dataset reached an asymptote (Fig 3 (A)). The final size of the wolf
occurrence area, with all data points included, differed between the three monitoring methods.
Via telemetry we detected 92 squares with wolf presence, via snow-tracking 89 squares and via
scat surveys 78 squares. However, the sampling effort spent on scat surveys (n = 48 sampling
days) was limited to one third of the effort used for the two other sampling methods (telemetry: n
= 165 sampling days, snow-tracking: n = 149 sampling days). The increase of detected wolf area
was considerably faster by scat surveys than by the other two monitoring methods. Whilst the
A comparison of monitoring methods for estimating wolf abundance
17
maximum of 78 squares with wolf presence was reached after 48 days of scat surveys, 90
telemetry days or 95 snow-tracking days were needed to detect a wolf area of the same size.
2.5
Discussion
Our results indicate that the search for scats is a useful technique for monitoring wolf presence.
As in the study by Zub et al. (2003), wolves in our study area marked the center of their home
ranges as well as the boundaries more often than the rest. Consequently, individual packs can be
100
(A)
80
60
40
y = 24.393Ln(x) - 33.555
20
0
0
40
80
120
160
wolf area [No of squares]
telemetry days
100
(B)
80
60
40
y = 25.657Ln(x) - 42.028
20
0
0
40
80
120
160
snowtracking days
100
(C)
80
60
40
y = 29.78Ln(x) - 42.401
20
0
0
40
80
120
160
scat survey days
Fig 3: Sub-sampling in steps of five sampling days of the three datasets obtained by (A) telemetry (n =
165 sampling days), (B) snow-tracking (n = 149 sampling days), and (C) scat surveys along transects (n
= 48 sampling days). Mean values of five runs per sub-sampling step are shown with 95% confidence
intervals. Grey lines illustrate the maximum number of sampling squares with wolf presence.
A comparison of monitoring methods for estimating wolf abundance
18
recognized by this method. Although the total number of marks was higher in the periphery of
the home range we could not support the “olfactory bowl” model of Peters and Mech (1975), in
which marks accumulated along the border of the home range. When set in relation to the surface
of the different zones, wolves marked the center of their home range the same often as the
boundaries, which agrees with the results of Paquet and Fuller (1990). It points to the
explanation that wolves intensively mark those parts of their home ranges, which are of
particular value to them, as the center with the den site, or which are most vulnerable to
intruders, as the home range boundaries (Zub et al. 2003). Similar marking behavior was
observed for some male ungulates, which intensively marked the boundaries of their territories to
keep away competitors (klipspringer: Roberts & Lowen 1997, antelope: Brashares & Arcese
1999); but not for others (e.g. roe deer: Johansson & Liberg, 1996). For monitoring swift foxes
and coyotes the collection of scats has already been proposed (Sovada & Roy 1996, Olson et al.
1997, Harrison et al. 2002, Gompper et al. 2006). Detection probability was up to 70% for swift
foxes and the number of coyote feces correlated with the local coyote density. However,
Gompper et al. (2006) stressed that trail-based fecal surveys were only efficient in detecting
coyotes; for other carnivores it turned out to be inefficient. This shows that the comparability of
different monitoring methods has to be verified for each species independently. Our data showed
that wolf distribution can be estimated through scat surveys. The results reflected the ones
obtained by telemetry and snow-tracking. Though, Harrison et al. (2002) advise to verify the scat
surveys by DNA analysis to gain information on the relative abundance of swift foxes. Others
state that the relative abundance of a species might be measured by counting droppings; the
absolute abundance, however, can not (Wilson et al. 2001, Sadlier et al. 2004). These
contradictory findings might have resulted from diverse approaches how to count droppings, as
two alternatives are available. One possibility is to count the total number of scats (the standing
crop; this study); the other one is to measure the accumulation rate of scats per unit of time
(Putman 1984). In combination with knowledge about defecation rate and fecal decay rates, the
latter approach can be used to measure the absolute abundance of a species. However, one has to
consider that movements, and thus deposition of droppings, may vary between seasons and that
visibility, and thus detectability of scats, varies with habitat characteristics (Wilson & Delahay
2001, Sadlier 2004). Other studies compared habitat use of carnivores based on snow-tracking
and telemetry and the results of the two methods showed good concordance (Alexander et al.
2005, Van Etten et al. 2007), as it was the case in our study. However, as research budgets are
A comparison of monitoring methods for estimating wolf abundance
19
frequently the limiting factor, another important consideration is the time- and cost-efficiency of
the various monitoring methods. Telemetry gives precise information about home ranges and its
use, but it has higher costs and is more time consuming than the other methods (Wilson et al.
2001, Franco et al. 2007, this study). Another disadvantage is the small number of radio-tagged
animals that might not be representative of the whole population. Moreover, it is not possible to
survey on large geographic scales, though this is important for monitoring whole communities of
wide ranging species as the wolf. Snow-tracking is less expensive and has a high potential for
detecting the species when present (Gompper et al. 2006). However, it requires suitable field
conditions (e.g. fresh snow) and thus is restricted to distinctive habitats or seasons (Sadlier et al.
2004, Gompper et al. 2006). As well, trained personnel are needed for accurate identification of
the tracks (Silveira et al. 2003, Sadlier et al. 2004). All of the herein analyzed methods have their
advantages and disadvantages. The appropriateness of the selected method depends on various
factors, such as the species under study, the research question, physiographic characteristics of
the study area and funding. Our study shows that scat surveys provide reliable information about
wolf distribution and present the most time-effective monitoring method, demonstrated by the
sub-sampling approach. Thus, we propose to use this technique to gain first insights into a
population. More detailed information about activity and habitat use can subsequently be
acquired year-round by telemetry or restricted to the winter season by snow-tracking surveys.
2.6
Acknowledgements
The study was part of the “Bieszczady Wolf Project”, coordinated by Dr. Roman Gula, funded
by the Polish National Committee for Scientific Research (KBN 6P04F 006), budget of the
Museum and Institute of Zoology, PAS, and financially supported by scholarships of the German
Academic Exchange Service, “Allgemeines Promotionskolleg” and Research School of the Ruhr
University Bochum. Data on telemetry and snow-tracking were collected together with Roman
Gula, Bartosz Pirga, Jörn Theuerkauf, Barbara Brzezowska, Hiroshi Tsunoda, Sophie Rouys and
Stephan Radler. Moreover, I want to thank Stephan Radler for his help during scat surveys.
Habitat preferences of wolves in central northern Portugal
3
3.1
20
HABITAT PREFERENCES OF WOLVES (Canis lupus) IN CENTRAL NORTHERN
PORTUGAL
Abstract
From June 2005 to March 2007, we investigated wolf presence in an area of 1000km² in central
northern Portugal by line transects in search of wolf scats. We aimed at predicting wolf presence
by developing a habitat model using land cover classes, prey abundance and human influence
(e.g. population density and road density). We confirmed the presence of three wolf packs. Wolf
areas were characterized by lower human presence and higher densities of livestock. Wolves
avoided the closer surroundings of villages and roads, as well as the general proximity to major
roads. Our model, developed by binary logistic regression, included the variables livestock and
road density and correctly explained 90.7% of wolf areas. Our results show that livestock
abundance is the most important factor for wolf occurrence and that unless human impact
exceeds a certain level, wolves can coexist with humans even in areas of poor land cover.
3.2
Introduction
Wolves had been eradicated from most of their original range by the middle of the 20th century
(Boitani 2000). However, since these animals became protected in many countries, their numbers
increased and their range expanded (Boitani 2000). Most of the former wolf habitat has become
urbanized and industrialized, so the species' current expansion frequently leads to conflicts with
humans, especially in livestock farming areas (Mech 1995, Linnell et al. 1996). In Portugal, wolf
numbers and distribution decreased dramatically during the 20th century (Bessa-Gomes &
Petrucci-Fonseca 2003), until they were legally fully protected in 1988. Since then, wolf
numbers have somewhat stabilized, even though in some areas local extinctions may still occur.
Livestock farming is an important field of Portuguese economy. Livestock, often unguarded or
with just one shepherd, generally roams free in the mountains rather than in fenced pastures.
Wolf depredation on goats (Capra hircus) and sheep (Ovis aries) is therefore commonplace. As
a result, wolves are often killed illegally (by shooting, poison or snares). Other threats to
Portuguese wolves include habitat fragmentation by new roads, decrease of forest cover caused
by fires during dry summers and new settlements in formerly uncultivated areas. The humanwolf conflict therefore needs to be reduced in order to prevent a further decline of wolf numbers
21
Habitat preferences of wolves in central northern Portugal
and enable the resettlement of the species over their former range. Wolf habitat models can help
improve the carnivore’s conservation by determining priority areas, developing conservation
corridors between important wolf habitats and highlighting potential conflict zones between
wolves and humans. The aim of our study was therefore to develop a habitat model based on
data about wolf occurrence in central northern Portugal and gain insight into the important
factors for wolf distribution, particularly the impact of anthropogenic variables on the predator’s
presence.
3.3
Material and methods
The study was conducted in the Vila Real district, in central northern Portugal, and included the
Natura 2000 site “Alvão/Marão” (Fig 1). The study area covered 1000km² with mountains up to
1400 m a.s.l. (41°10’-41°51’N, 07°13’-07°59’E). The area included numerous small villages as
well as two bigger towns. Average population density was 48 humans per km² and road density
Legend
#
towns
main roads
study area
rendezvous sites
core areas
wolf area
0 2 4
10
20km
Fig 1: Location of the study area within Portugal (small map) and wolf distribution within the study area
(big map). Shown are major roads, areas of high wolf presence (green area), core areas of three wolf
packs (shaded areas), and location of rendezvous sites of three packs (dots), detected by howling
simulations.
Habitat preferences of wolves in central northern Portugal
22
around 0.83 km km-². The area was cut by three major roads and two newly constructed fenced
highways. The region was mainly shrub land (38%), agricultural land (24%) and forests (21%).
The forest was composed in 62% of coniferous, pine (Pinus pinaster) forests and secondary of
broad-leaved forests (Quercus sp. and Castanea sativa). Important economic resources included
forestry and livestock grazing.
We conducted the study from June 2005 to March 2007. We estimated wolf presence by
searching for wolf scats on 550 km of line transects. We placed a grid above the study area with
2x2km mesh size and chose transects pseudo randomly within each square. Transects were
unpaved roads accessible by four wheel drive and were distributed over the whole study area.
We inspected 200km of the 550 km every three months and the other 350km once during the
study. We assessed the position of each wolf scat by GPS. We used fixed kernel-analysis with a
band-width of 1500m to calculate probabilities of wolf occurrence from the resulting wolf
location data. Recognition of individual packs was based on a 50% probability-analysis, showing
the most intensively used areas, and confirmed by howling simulations (human imitation). In
summer, when pups still remain at rendezvous sites, wolves were stimulated to howl and, in case
of responses by adults and pups, the location was considered a rendezvous site of the pack. We
related the distance of each wolf scat from settlements and roads to the same measures obtained
for a set of random points, generated within the pack’s ranges, by Ivlev’s electivity index (Jacobs
1974):
selection index = (ps – pr)(ps + pr – 2pspr)-1
(1)
with ps being the proportion of wolf scats in a given distance to the next settlement / road and pr
the proportion of random points in the same given distance to the next settlement / road.
Selection indices vary from +1 (total selection) to –1 (total avoidance).
For the analysis of habitat preferences of wolves we used the same 2x2km squares, labeled with
high or low wolf presence. Recognition of squares with high wolf presence was based on a 95%
kernel density distribution. Squares outside this range were assumed to hold low wolf presence.
We chose variables of possible importance to wolves and assessed them for each of the 248
squares (Table 1). These variables described the land cover (forest, open areas, shrub land,
agricultural land, and urban areas), human influences (road density, number of settlements, and
human density) and prey abundance (livestock density). We focused on livestock as main prey
since Carreira and Petrucci-Fonseca (2000) found that the wolves’ diet in this area consists to
more than 80% of livestock, and wild prey, such as roe deer (Capreolus capreolus) and wild
23
Habitat preferences of wolves in central northern Portugal
Table 1: Variables used in the habitat utilization analysis, describing land cover classes, prey abundance
and human influences, together with their origin and resolution
variable
category
origin and resolution of data
road density
human influence
military maps (1996-1998), scale 1:25 000
human density
human influence
national census of the population in 2001
settlements
human influence
CORINE 2000 maps, scale 1:25 000
urban areas
land cover
CORINE 2000 maps, scale 1:25 000
agricultural land
land cover
CORINE 2000 maps, scale 1:25 000
shrub land
land cover
CORINE 2000 maps, scale 1:25 000
open areas
land cover
CORINE 2000 maps, scale 1:25 000
forest cover
land cover
CORINE 2000 maps, scale 1:25 000
livestock density
prey abundance
national census of agriculture in 2003
boar (Sus scrofa), contribute to only 18.9%. To compare wolf areas with non-wolf areas, we
calculated means of variables in squares with high wolf presence and in squares with negligible
wolf presence with 95% confidence intervals and tested differences with the Mann-Whitney Utest. For the development of the habitat model we chose logistic regression (SPSS13.0 for
Windows) as the most appropriate method, as it shows critical habitat factors for wolf presence.
We first divided the dataset into two randomly chosen sub-samples and developed the model
based on 70% of data and validated it with the remaining 30%. We used the forward stepwise
method to select variables contributing most to the model and counterchecked it by the backward
stepwise method. To estimate the fit of the produced model, we used the Hosmer-Lemeshow
goodness-of-fit test.
3.4
Results
We found a total of 1723 wolf scats between June 2005 and March 2007. We distinguished three
wolf packs and confirmed their core areas and reproduction sites by howling simulations (Fig 1).
95% kernel density distribution revealed that wolves frequently used 65% (162 squares) of the
study area, whereas they only occasionally visited the remaining 35% (86 squares). Road density
was lower in areas settled by wolves (U-test, P = 0.002; Table 2). Wolves avoided a zone of 2km
each side of major roads and usually avoided a corridor of 0.5 km each side of smaller roads (Fig
2). Wolves, however, selected a zone within 1-2km from smaller roads. Human population
24
Habitat preferences of wolves in central northern Portugal
1
selection index
(A)
0
-1
0
0.5
1
1.5
2
2.5
distance to settlement [km]
1
1
(C)
selection index
selection index
(B)
0
0
-1
-1
0
0.5
1
1.5
2
2.5
distance to major road [km]
0
0.5
1
1.5
2
2.5
distance to smaller road [km]
Fig 2: Selection and avoidance of settlements (A), major roads (B), and smaller roads (C) by wolves,
calculated by Ivlev’s electivity analysis. Selection indices can vary from +1 (total selection) to –1 (total
avoidance). The analysis is based on scat surveys along line transects from June 2005 to March 2007.
density was lower in the wolf range (U-test, P < 0.001), which included less villages and smaller
urbanized areas (U-test, P = 0.017 and P = 0.016 respectively; Table 2). Wolves avoided both,
the close proximity of settlements and areas farther than 2km from villages (Fig 2). Wolves
therefore selected areas that were within 1-1.5km from settlements. Livestock density was higher
in areas that wolves visited regularly (U-test; P < 0.001), with a stronger tendency for goats and
cows (Bos taurus) (U-test, P < 0.001) than for sheep (U-test, P = 0.008; Table 2). Areas occupied
by wolves were less forested than areas in which wolves were virtually absent (U-test, P <
0.001). Livestock density classified 77.1% of squares correctly during model development and
84.1% during model testing (Table 3). Wolf areas were correctly classified to 89.1% and 93.0%
(during model development and model testing, respectively) and non-wolf areas were correctly
classified to 53.3% and 69.2%. We subsequently added the variable road density to the model,
because the difference (calculated by the Hosmer-Lemeshow goodness-of-fit test) between
observed and predicted wolf presence was nearly significant (χ² = 13.998, P = 0.112), indicating
25
Habitat preferences of wolves in central northern Portugal
Table 2: Mean values and 95% confidence intervals (CI) for the variables “human influence”, “land cover”
and “prey abundance” in squares with and without wolf presence (based on scat surveys along line
transects conducted from June 2005 to March 2007). Results of Mann-Whitney U-test to compare means
of squares with high and low wolf presence: ** ≡ p < 0.01, * ≡ p < 0.05.
wolf presence
wolf absence
(n = 16 squares)
(n = 86 squares)
mean ± CI (95%)
mean ± CI (95%)
road density (km km-²)
0.71 ± 0.10**
1.04 ± 0.17**
human density (no km-²)
42.3 ± 6.1**
59.7 ± 10.7**
settlements (no km-²)
1.37 ± 0.20*
2.09 ± 0.43*
urban areas (km² km-²)
0.16 ± 0.04*
0.22 ± 0.05*
agricultural land (%)
24.7 ± 3.1
23.9 ± 4.3
shrub land (%)
35.8 ± 3.6
42.4 ± 6.0
open areas (%)
19.5 ± 4.1**
4.8 ± 3.1**
forest cover (%)
17.8 ± 3.0**
27.6 ± 4.8**
livestock density (no km-²)
47.0 ± 3.2**
24.6 ± 3.2**
goats
19.5 ±3.70**
9.08 ± 2.03**
sheep
11.3 ± 1.53*
9.09 ± 1.33*
cows
18.6 ± 1.94**
11.8 ± 1.75**
that the model showed a poor fit. The second model classified 78.8% of squares correctly during
model development and 82.6% during model testing, with a sensitivity (correctly classified wolf
areas) of 88.2% and 90.7% and a specificity (correctly classified non-wolf areas) of 60.0% and
69.2%. The Hosmer-Lemeshow goodness-of-fit test detected no significant difference between
observed and predicted wolf presence (χ² = 2.963, P = 0.937) and therefore provided a
considerably better fit.
3.5
Discussion
Our results imply that wolves are not habitat specific concerning land cover. Land cover
variables such as shrubs and agriculture, did not differ between wolf habitat and non-wolf
habitat. The lower amount of forest in the wolves’ ranges was probably due to the selection of
26
Habitat preferences of wolves in central northern Portugal
Table 3: Results of the logistic regression analysis and percentages of correct classifications, obtained
during model development and during model testing. Sensitivity and specificity refer to the correctly
classified wolf-areas and non-wolf areas, respectively.
variable
step1
step2
β±
Wald
S.E.
statistic
d.f. p-value
0.08
29.93
± 0.01
1
constant
-1.93 16.36
± 0.48
1
0.0001
livestock
0.08
31.19
± 0.02
1
0.0001
road
density
-0.96 12.54
± 0.27
1
0.0001
constant
-1.32 6.87
± 0.50
1
livestock
sensitivity
specificity
correct
(model/
(model/
predictions
test)
test)
(model/test)
89.1 % /
53.3% /
77.1% /
93.0%
69.2%
84.1%
88.2% /
60.0% /
78.8% /
90.7%
69.2%
82.6%
0.0001
0.0001
areas where livestock is grazing, in open habitats and shrub land. Studies from Poland showed a
considerably higher amount of forest in areas inhabited by wolves (Jędrzejewski et al. 2004,
2005). This might as well reflect the wolves’ selection of areas used predominantly by their main
prey species, which in Poland is red deer (Jędrzejewski et al. 2000, Jędrzejewski et al. 2002,
Gula 2004, and Nowak et al. 2005, Gula 2008). Our findings agree with conclusions drawn by
Mech (1995) and Fuller (1995), who state that wolves are generalists regarding their habitat
requirements. They stand at the top of the food chain and can survive wherever they have enough
to eat and are not killed by humans (Peterson 1988, Mech 1995). In our study, prey density was
the factor which best explained wolf occurrence. Fuller (1989) and Fuller et al. (1992) came to
the same conclusion. In their studies prey density explained 72% of wolf occurrences. However
they restricted their statement to unexploited wolf populations. Boitani (1992) and Carroll et al.
(1999) emphasize that studies carried out in regions with different exploitation histories, e.g. in
North America and Europe, have to be regarded in their own context. Boitani and Ciucci (1993)
state, that the human attitude towards wolves is the determining factor of wolf occurrence.
However, the relationship between wolves and humans is very complex (Linnell et al. 2001), as
for instance wolves are nocturnal in southern Europe (Italy and Spain) with human densities of
20-30 people km-², but are rather diurnal in southeastern Poland with a human density of 44
Habitat preferences of wolves in central northern Portugal
27
people km-² (Theuerkauf et al. 2007). In our study, areas with higher human presence (measured
by human population density, road density, and urban areas) were avoided by wolves, with roads
having the highest impact on them. However, the negative impact of roads might just reflect the
negative impact of humans (Thiel 1985 and Mech et al. 1988). It seems that wolves select roads
seldom used by humans for ease of travel, but avoid bigger roads (Thurber et al. 1994,
Theuerkauf et al. 2003, Theuerkauf et al. 2007, this study). Other studies of habitat use also
pointed to the negative impact of roads on wolf occurrence (Mladenoff et al. 1995, Cayuela
2004, Jędrzejewski et al. 2005). In north-central Spain, however, Blanco et al. (2005) did not
find a major impact of highways on wolves, which crossed them frequently via bridges. In other
studies, wolf avoidance of people was indirect through selection of high altitudes, where people
seldom appear (Glenz et al. 2001, Grilo et al. 2002). In this part of Portugal, higher altitudes
have been a major refuge for wolves until recently (Carreira & Fonseca 2000). With the
increasing construction of wind farms and road networks to enable their access, human
disturbance in highland regions is becoming an issue for wolf conservation. In spite of the
present human pressure we conclude that the most important factor for wolf occurrence in this
region is the abundance of prey. Human presence has some negative impact on wolves and might
even prevent the settlement of wolves in a given area. However, this is likely to occur only in
areas where uncontrolled killing of wolves is frequent, human activities reach a threshold or
hiding conditions (forest and shrub cover) are insufficient. The behavioral plasticity of wolves is
the main reason for their survival, despite persecution throughout the centuries in Europe, as well
as for their recent range expansion.
3.6
Acknowledgements
The study was funded by Norscut, S.A. and Aenor, S.A. and financially supported by a
scholarship of the “Allgemeines Promotionskolleg” of the Ruhr University Bochum. Sample
collection was part of an environmental impact study coordinated by “Grupo Lobo” and
supervised by Prof. Francisco Fonseca. I want to thank Ana Guerra, Gonçalo Costa and Sílvia
Silva for their help during fieldwork and the Alvão Natural Park services for their cooperation.
Validation of a wolf habitat model
4
4.1
28
VALIDATION OF A WOLF (Canis lupus) HABITAT MODEL
Abstract
Habitat modeling has become a central tool in conservation biology, though modeling techniques
are diverse and so are the resulting models. Our aim was to design a model for explaining wolf
occurrence in south-eastern Poland, compare it to other wolf habitat models and test the
transferability of such models. For this purpose we conducted scat surveys to determine wolf
occurrence and characterized the study area in terms of land cover, anthropogenic influences and
prey density. By means of a Principle Component Analysis we identified four ecological factors
as significant for explaining wolf presence. All of these factors belonged to the categories of
anthropogenic influences and prey density. Through logistic regression we designed a wolf
habitat model for predicting wolf occurrence within the study area, which included livestock
density and forest cover. Via meta-analysis we detected prey availability and anthropogenic
influences as the common basis of all habitat models, while both factors were frequently masked
by land cover variables. Moreover prey availability presents a fairly complex parameter and is
difficult to measure. We suggest the design of a standardized procedure to characterize wolf
habitats, in order to gain standardized habitat models, which can be applied worldwide as
predictive tools in conservation biology.
4.2
Introduction
As wolves and humans have always been competitors for food, hundreds of years ago wolves
have been demonized and people tried to exterminate them. In many European countries, as well
as in the majority of the USA, people were successful in doing so. Nowadays the opinion about
wolves is slowly changing and people start to accept the wolf’s right to live. Since the wolf was
protected by law in most western European countries (Boitani 2000), its numbers rapidly
increased and it started reoccupying areas, which are cultivated by humans. As remote areas
have become scarce, the wolf nowadays has to coexist with people in close proximity. This leads
to human-wolf conflicts, especially in areas, where people live on livestock and wild prey has
become limited. Forests were clear cut to produce grazing ground and agricultural land and
consequently retreat areas for wolves and their natural prey decreased. As a result wolves learned
to exploit other food resources and started feeding on livestock (Vilá et al. 1995, Cayuela et al.
Validation of a wolf habitat model
29
2004, Treves et al. 2004, Kusak et al. 2005, Bisi et al. 2007, chapter 3 this thesis). This, in turn,
increased poaching as a major cause of wolf mortality. Another threat wolves had to face was an
increasing fragmentation of their habitat by anthropogenic barriers such as roads and settlements
(Thurber et al. 1994, Wydeven et al. 2001, Theuerkauf et al. 2003b, Whittington et al. 2004,
2005, Jaeger et al. 2005, Kaartinen et al. 2005, Theuerkauf et al. 2007). Nevertheless wolves live
in close neighborhood with humans in various countries (Ciucci et al 1997, Jhala & Sharma
1997, Theuerkauf et al. 2007, Gula et al. in press, chapter 3 this thesis), showing that coexistence
is possible. Even in agricultural habitats in Spain wolves survive, breeding in cereal and corn
fields and feeding on wild boar, rabbits and hare (Blanco et al. 2005). There is no simple
relationship between human density and wolf persistence in a given area (Linnell et al. 2001). In
Poland for example wolves survive alongside 44 inhabitants km-² (Theuerkauf et al. 2007),
whereas they have been exterminated amongst 1.3 people km-² in Sweden (Zimen & Boitani
1979). Thus a variety of environmental factors influence the survival of wolves and possible
cumulative effects of those factors have to be considered. Knowing under which conditions
wolves are able to live and what kind of environmental characteristics make a habitat unsuitable
for wolves, can help in supporting their resettlement. This can be realized either directly by
protecting crucial habitat for wolves and mitigating the negative impact of anthropogenic
infrastructure such as roads and settlements (Cain et al. 2003, Ng et al. 2004, Clevenger &
Waltho 2005) or indirectly by reducing human-wolf conflicts through supporting farmers in
protecting their livestock (Musiani et al. 2003, Gula 2008) and keeping wild prey densities high
enough for a predator-prey equilibrium (Gula 2008). Another promising move is to visualize
connectivity between suitable wolf areas and allow wolf management and conservation efforts to
be focused on critical areas. Such efforts might include measures to encourage wolf dispersal
through suitable areas and consequently minimize conflicts with humans, thus contributing
positively to the management of a socially conflictive species (Rodriguez-Freira & CrecenteMaseda 2008). One central possibility to gain information on suitable wolf habitats is the
application of ecological modeling. During ecological modeling wolf occurrence data are
combined with environmental variables (as well biotic as abiotic factors) and a model of the
species’ requirements for the examined variables is created (Anderson et al. 2003). Several
habitat models have been build in different countries so far (Mladenoff et al. 1995, Massolo &
Meriggi 1998, Glenz et al. 2001, Grilo et al. 2002, Cayuela et al. 2004, Jedrzejewski et al. 2004,
Validation of a wolf habitat model
30
Fig 1: Map of the study area with bigger cities in the closer surrounding (dots) and state borders to
Slovakia and the Ukraine. Presented are major roads, areas with wolf presence (gray area), as well as
the four core areas of wolf packs in that region (circles).
2005, Rodriguez-Freira & Crecente-Maseda 2008), but environmental variables identified as
most important for wolves seem to differ a lot. By comparing different habitat models and
transferring them to other regions, it might be possible to extract those factors, which are crucial
for wolves in all regions, and separate them from factors, which are important merely in specific
habitats. Based on these variables it will be possible to design a universal model, apply it to
presently wolf free areas, and predict their future resettlement. For this reason the initial aim of
our study was to build a habitat model for a wolf population in south-eastern Poland. The
proximate aim was to compare it to other models within Europe and worldwide, in order to find
similarities and differences. As a final step we applied a model, designed for a Portuguese wolf
Validation of a wolf habitat model
31
population, to the herein studied wolf population and tested the transferability by revealing the
model fit.
4.3
Material and Methods
Study area
The study was conducted in the foothills of the Bieszczady Mountains (49°19’-49°50’N, 22°15’22°45’E), which are situated in south-eastern Poland and flanked by the state borders to Slovakia
and the Ukraine (Fig 1). The study area comprised approximately 1000km². Human population
density averages 34 inhabitants km-²; with considerably lower densities throughout most of the
study area (29 inhabitants km-²) and one bigger town of about 10,000 inhabitants. Road density
is 0.59km km-², with one main road axis from the south to the north and two main road axes in
east-western direction. The area is forested to 66% and presents an optimal habitat for red deer
(Cervus elaphus), roe deer (Capreolus capreolus) and wild boar (Sus scrofa), the main prey
species of the wolf. The climate is mountainous with continental influences. Annual precipitation
averages 800-1200mm, with a snow depth of 10-40cm for a period of 90-140 days during winter
season.
Data collection
From winter 2005/2006 until autumn 2007 we conducted scat surveys along 550km of
predefined transects. Transects were chosen pseudo-randomly, with the requirement of an equal
distribution throughout the whole study area. We predominantly walked or drove, with an upper
speed limit of 10km/h, along dirt roads and an additionally few paved forest roads with restricted
use by cars. For each scat that was found we acquired the location by GPS.
Data analysis
For the calculation of wolf distribution within the study area we conducted a kernel probability
analysis, including 90% of all scat samples. For the distinction of separate wolf packs we
calculated core areas of higher usage rates by kernel analysis with 55% probability of
occurrence. For further analyses we divided the entire study area into squares of 4km² by
mounting a grid of 2x2km mesh size. The resulting 244 squares were split into squares with
regular wolf presence (core areas), irregular wolf presence, and wolf absence, based on the 90%
32
Validation of a wolf habitat model
Table 1: Variables used for habitat modeling, describing land cover classes, prey abundance and human
influences, together with their origin and resolution
variable
category
origin and resolution of data
road density
human influence
digitalized with GPS by the authors
human density
human influence
CORINE 2000 and LUCAS 2001, resolution 1ha
settlements
human influence
CORINE 2000 maps, scale 1:25 000
urban areas
land cover
CORINE 2000 maps, scale 1:25 000
agricultural land
land cover
CORINE 2000 maps, scale 1:25 000
forest cover
land cover
CORINE 2000 maps, scale 1:25 000
wild ungulate density
prey abundance
census in 2007, resolution of hunting districts
livestock density
prey abundance
agricultural census in 2004
and 55% kernel probability analyses. For some additional analyses, regular and irregular wolf
presence was joined as wolf areas (WA) and wolf absence defined as non-wolf areas (NWA).
Further, squares were characterized by variables of land cover (forest cover, agriculture), prey
density (wild ungulate density and livestock density) and anthropogenic factors (road density,
human population density, number of settlements, and extent of urban areas, Table 1). All
geographic analyses were performed in ArcView 3.2 (ESRI 1992) and ArcInfo 8.1 (ESRI 1999).
To compare the habitat characteristics of areas with dissimilar wolf presence, we calculated
means of the before mentioned variables in squares with regular wolf presence, irregular wolf
presence and wolf absence with 95% confidence intervals and tested for possible differences
with the Kruskal-Wallis test. To cluster the variables into ecologically meaningful groups and
identify their importance for explaining the variance between the studied squares, we conducted
a Principle Component Analysis (PCA). Only those variables with significant differences
between wolf abundance classes, discovered by the Kruskal-Wallis test, were entered into the
PCA. To illustrate differences between the three wolf abundance classes we computed mean
scores of the ecological factors identified by PCA and tested them by the Kruskal-Wallis test (3
wolf abundance classes) and the Mann-Whitney U-test (WA vs. NWA). We then entered the
factors identified by PCA into a Discriminant Function Analysis (DFA) to rank their importance
for wolf abundance discrimination. The contribution of each factor was evaluated by correlation
analysis. The factors were entered into the model simultaneously to study their combined effects.
Table 2: Mean values and 95% confidence intervals (CI) for variables concerning “human influence”, “land cover” and “prey abundance” for three wolf abundance
classes. Results of the Kruskal-Wallis test are indicated by ** ≡ p < 0.01 and * ≡ p < 0.05.
CLASSES OF WOLF ABUNDANCE
absence
irregular presence
regular presence
total
mean ± CI (95%)
mean ± CI (95%)
mean ± CI (95%)
mean ± CI (95%)
0.69 ± 0.09**
0.54 ± 0.08**
0.50 ± 0.13**
0.59 ± 0.05
major roads 0.11 ± 0.05**
0.14 ± 0.04**
0.01 ± 0.01**
0.11 ± 0.03
secondary roads 0.58 ± 0.09**
0.41 ± 0.07**
0.49 ± 0.13**
0.49 ± 0.05
-
road density [km km ²]
human density [no km-²]
34.3 ± 4.4**
37.8 ± 9.5**
23.1 ± 3.8**
34.1 ± 4.8
settlements [no km-²]
0.95 ± 0.16**
0.74 ± 0.14**
0.48 ± 0.21**
0.77 ± 0.10
urban areas [km² km-²]
agricultural land [%]
0.34 ± 0.08**
0.38 ± 0.09**
0.14 ± 0.09**
0.33 ± 0.05
38.0 ± 5.2**
32.5 ± 4.5**
21.0 ± 5.3**
32.7 ± 3.1
forest cover [%]
61.0 ± 5.5**
65.0 ± 4.8**
79.8 ± 5.3**
65.9 ± 3.2
broad-leaved forest 11.0 ± 3.1**
20.9 ± 4.7**
25.1 ± 7.8**
17.9 ± 2.9
coniferous forest 19.6 ± 4.0
18.1 ± 4.5
22.2 ± 7.2
17.9 ± 2.8
mixed forest 29.9 ± 4.6
26.1 ± 3.9
31.3 ± 7.1
28.4 ± 2.7
3.37 ± 0.12**
2.40 ± 0.13**
2.27 ± 0.14**
2.74 ± 0.10
roe deer 2.20 ± 0.10**
1.43 ± 0.10**
1.23 ± 0.09**
1.68 ± 0.08
red deer 0.85 ± 0.07
0.78 ± 0.04
0.83 ± 0.05
0.82 ± 0.03
0.18 ± 0.02**
0.21 ± 0.03**
0.25 ± 0.02
wild ungulate density [no km-²]
wild boar 0.34 ± 0.03**
-
livestock density [no km ²]
0.27 ± 0.17**
2.62 ± 0.53**
2.21 ± 0.54**
1.68 ± 0.30
-2
3.66 ± 0.17**
5.02 ± 0.48**
4.48 ± 0.63**
4.42 ± 0.26
total prey density [no km ]
34
Validation of a wolf habitat model
Table 3: Results of the Principle Component Analysis. Shown are correlation coefficients between
ecological factors and environmental variables.
factor
variable
I
II
forest
-0.852
-0.128
-0.186
-0.150
agriculture
0.793
0.173
0.227
0.194
major roads
0.724
0.010
-0.199
-0.161
settlements
0.542
0.124
0.331
0.295
urban areas
0.783
-0.114
0.192
0.143
humans
0.727
-0.094
-0.008
-0.093
-0.468
-0.497
-0.109
-0.090
roe deer density
0.054
0.874
0.130
-0.153
wild boar density
-0.030
0.827
-0.087
-0.100
wild ungulate density
-0.021
0.957
0.085
-0.154
secondary roads
-0.019
0.058
0.987
-0.038
roads
0.357
0.061
0.857
-0.120
livestock density
0.079
-0.379
-0.081
0.902
total prey density
0.083
-0.067
-0.057
0.961
Eigenvalue
3.677
2.849
2.034
2.021
explained variance (%)
26.267
20.350
14.532
14.439
cum % of variance
26.267
46.617
61.149
75.588
broad-leaved forest
III
IV
We used 70% of all study squares for building the discriminant function and the other 30% for
validation of the model. Furthermore, we conducted binary logistic regression analyses for
designing models to explain wolf presence. The equation parameters were estimated by the
Validation of a wolf habitat model
35
stepwise forward method. For model development we used habitat variables that differed
significantly between wolf abundance classes on the one hand and ecological factors identified
by PCA on the other hand. Again we used 70% of the data for model development and 30% for
model testing. By means of the Wald statistic we tested our model against the null model and by
the Hosmer-Lemeshow goodness-of-fit test we assessed the model fit. Further we compared
factors from our study to factors identified as crucial for wolf abundance in various studies
worldwide. To test the transferability of habitat models from one wolf population to others, we
applied a wolf habitat model, constructed on data collected in central-northern Portugal, on our
dataset. As habitat variables differed between the two study areas to a great extend, it was
impossible to use the Portuguese model directly. To circumvent this problem, we entered the
variables detected as crucial for wolf presence in the Portuguese study area into our new model,
but recalculated the variable coefficients.
Fig 2: Mean scores of factor I (anthropic pressure), factor III (road density), factor II (wild prey density),
and factor IV (domestic prey density) discovered by the Principle Component Analysis for three wolf
abundance classes.
Validation of a wolf habitat model
4.4
36
Results
During scat surveys we found a total of 268 wolf scats. Wolves were present in 61% of the study
area, calculated by 90% kernel probability analysis. The southern part of the study area was
entirely used by wolves, whereas the north-western most part was hardly ever used (Fig 1). By
55% kernel probability analysis we detected four distinct areas of intensified use, belonging to
four wolf packs, which were separated from each other by main roads. By these analyses WA
(150 squares) were distinguished from NWA (94 squares) and areas with regular wolf presence
(40 squares) from areas with irregular wolf presence (110 squares). The total density of roads, as
well as secondary and main road density regarded separately, differed significantly between the
three wolf abundance classes (Kruskal-Wallis test, p = 0.01, p = 0.007 and p = 0.001,
respectively; Table 2). There were virtually no main roads passing through areas with regular
wolf presence (0.01km km-²); however the density of secondary roads was only slightly lower
than in wolf free areas and even higher than in areas with irregular wolf presence. Within areas
of irregular wolf presence main road density was higher than in areas with regular wolf presence
as well as in wolf free areas. Moreover, variables concerning human impact, such as human
population density, number of settlements, extend of urban areas, and agricultural land differed
between wolf abundance classes (Kruskal-Wallis test, p = 0.004, p = 0.002, p = 0.002 and p =
0.002 respectively; Table 2). All anthropogenic influences were lowest in areas with regular wolf
presence. Forest cover increased from wolf free areas through areas with irregular wolf presence
to areas with regular wolf presence (Kruskal-Wallis test, p = 0.001). Furthermore, forest
composition changed between wolf abundance classes. Mixed forest cover was highest in all
three classes; however in NWA coniferous forest dominated, whereas in WA broad-leaved forest
was wider spread. Concerning food resources, livestock density was higher in areas with wolf
presence than in wolf-free areas (Kruskal-Wallis test, p < 0.001), whereas wild ungulate density
was significantly higher in wolf-free areas (Kruskal-Wallis test, p < 0.001). This difference was
caused by roe deer and wild boar densities (Kruskal-Wallis test, p < 0.001), the difference
between red deer densities was not significant (Kruskal-Wallis test, p > 0.05). Except mixed
forest, coniferous forest and red deer density, all variables were entered into the PCA. The PCA
discriminated between four ecologically meaningful factors (Table 3). Factor I represented
anthropogenic influences, with forest cover, agricultural land, main roads, settlements, urban
areas and human population density. Factor II corresponded to wild ungulate density, with the
37
Validation of a wolf habitat model
Table 4a: Correlation coefficients of the two models, designed by Discriminant Function Analysis.
ecological factors
correlation coefficients
DF 1
DF 2
Factor II
0.84*
0.06
Factor I
0.10
0.92*
Factor III
0.16
-0.31*
Factor IV
-0.25
0.29*
Eigenvalue
0.91
0.07
% of variance
92.6
7.4
Χ²
116.2
11.4
significance level
0.001
0.010
* largest absolute correlation between each variable and any discriminant function
variables roe deer density, wild boar density and wild ungulate density. Moreover the variable
broad-leaved forest was also included into this factor, though an ecologically meaningful link
with wild ungulates is not apparent. Factor III included secondary road and total road density,
and factor IV livestock density and total prey density. All four factors together explained 76% of
the variance, with 47% being explained by factor I and II. The mean scores of factors I and III
underlined the fact that anthropogenic influences in areas with regular wolf presence are lower
compared to the other two abundance classes (Fig 2). Furthermore, in Fig 2 the above mentioned
differences between densities of wild prey (factor II) and domestic prey (factor IV) become
clear. Wild prey densities were lower in areas with wolf presence, whereas domestic prey
densities were higher in these areas. The DFA produced two models (DF1 and DF2), which
significantly discriminated between the three wolf abundance classes (Table 4). DF1 identified
factor II as most important and explained 92.6% of the variance. Altogether the DFA correctly
classified 67.1% of grouped cases during model development and 59.7% during model testing.
By binary logistic regression we developed a first model, which included roe deer density as the
only variable (Table 5). However, the Hosmer-Lemeshow goodness-of-fit test revealed a poor fit
(p = 0.016). As a second variable wild boar density was added to the model, but it still didn’t fit
to the data (p = 0.039). The final model (Hosmer-Lemeshow goodness-of-fit test, p =0.632)
contained the variables roe deer density, wild boar density and secondary roads. It correctly
classified 85.0% of squares during model development and 80.5% during model testing.
38
Validation of a wolf habitat model
Table 4b: Model coefficients of the four ecological factors and percentages of correct classifications of
the three wolf abundance classes by the discriminant functions.
classes of wolf abundance
ecological factors
absence
irregular presence
regular presence
factor I
0.27
0.07
-0.65
factor II
1.68
-0.86
-0.77
factor III
0.58
-0.21
0.06
factor IV
-0.41
0.39
0.14
constant
-1.95
-1.35
-1.46
% correct classification (model
development) a
84.1
53.8
66.7
% correct classification (crossvalidated) b
84.1
52.5
58.3
% correct classification (model
testing)c
75.0
51.5
50.0
a: 67.1% correct classifications; b: 65.3% correct classifications; c: 59.7% correct classifications
However, a lower density of wild ungulates within wolf areas does not present an ecologically
meaningful criterion to explain wolf presence, but rather a secondary adaptation of the wild
ungulate species to the presence of wolves. Moreover the reliability of the wild prey data is not
confirmed. Therefore, in a next step we excluded all variables concerning wild ungulates from
model development. This second model included forest cover and livestock density as variables
(Table 5). The Hosmer-Lemeshow goodness-of-fit test showed a good fit (p = 0.112), though the
model performed somewhat worse than the previous, with 73.1% correct classifications during
model development and 71.4% during model testing. However, wolf areas were correctly
identified to 80.8% and 77.6%; consequently the model was almost as efficient as the first one.
When using the before defined ecological factors I – IV, we reached similar results when factor
II (wild prey factor) was included, but considerably worse results when factor II was excluded
during model development (Table 6). In this case wolf-free areas were falsely classified to more
than 70% as suitable habitat for wolves.
The meta-analysis revealed that in nearly all studies factors related to anthropogenic influences
were identified as most important for explaining the presence of wolves (Table 7). The factors
themselves however differed between studies. In several studies density or length of roads were
Table 5: Results of the logistic regression analysis and percentages of correct classifications, obtained during model development and testing. Sensitivity and
specificity refer to the correctly classified wolf areas and non-wolf areas, respectively.
variable
all
variables
model 1
model 2
model 3
wild
ungulates
excluded
model 1
model 2
β ± S.E.
Wald
statistic
d.f. p-value
roe deer
-3.046 ± 0.423
51.940
1
0.0001
constant
6.072 ± 0.833
53.148
1
0.0001
roe deer
-2.476 ± 0.473
27.443
1
0.0001
wild boar
-7.812 ± 3.008
6.746
1
0.009
constant
7.265 ± 1.064
46.648
1
0.0001
secondary
roads
-1.436 ± 0.586
6.011
1
0.014
roe deer
-2.495 ± 0.487
26.238
1
0.0001
wild boar
-8.787 ± 3.092
8.075
1
0.004
constant
8.366 ± 1.242
45.360
1
0.0001
livestock
1.012 ± 0.220
21.058
1
0.0001
constant
-0.478 ± 0.220
4.730
1
0.030
forest
0.028 ± 0.008
12.109
1
0.001
livestock
1.088 ± 0.230
22.444
1
0.0001
constant
-2.380 ± 0.613
15.075
1
0.0001
sensitivity
(model / test)
specificity
correct predictions goodness(model / test) (model / test)
of-fit
86.5 / 85.7
79.4 / 64.3
83.8 / 77.9
0.016
87.5 / 87.8
82.5 / 71.4
85.6 / 81.8
0.039
86.5 / 87.8
82.5 / 67.9
85.0 / 80.5
0.632
72.1 / 65.3
69.8 / 82.1
71.3 / 71.4
0.001
80.8 / 77.6
60.3 / 60.7
73.1 / 71.4
0.112
Table 6: Results of the logistic regression analysis and percentages of correct classifications obtained, during model development and testing. Ecological factors
discovered by the PCA were entered into the model. Sensitivity and specificity refer to the correctly classified wolf areas and non-wolf areas, respectively.
variable
all factors
factor II
excluded
model 1
model 1
model 2
model 3
β ± S.E.
Wald
statistic
d.f. p-value
factor II
-2.114 ± 0.319
43.940
1
0.0001
constant
1.072 ± 0.255
17.747
1
0.0001
factor IV
0.538 ± 0.191
7.983
1
0.005
constant
0.513 ± 0.166
9.616
1
0.002
factor III
-0.307 ± 0.164
3.486
1
0.062
factor IV
0.543 ± 0.193
7.879
1
0.005
constant
0.562 ± 0.171
10.859
1
0.001
factor I
-0.230 ± 0.162
2.004
1
0.157
factor III
-0.324 ± 0.166
3.808
1
0.051
factor IV
0.547 ± 0.193
8.039
1
0.005
constant
0.576 ± 0.172
11.181
1
0.001
sensitivity
(model / test)
specificity
correct predictions goodness(model / test) (model / test)
of-fit
87.5 / 85.7
81.0 / 71.4
85.0 / 80.5
0.236
86.5 / 79.6
3.2 / 7.1
55.1 / 53.2
0.001
80.8 / 87.8
17.5 / 14.3
56.9 / 61.0
0.019
84.6 / 83.7
28.6 / 17.9
63.5 / 59.7
0.053
Validation of a wolf habitat model
41
included into the model, in others human population density or number of villages. Prey
availability was the category selected second most frequently during modeling, even though in
three studies it was excluded from analysis because of unavailability of reliable data. Land cover
was included in all analyses, but in 5 out of 11 studies it was not considered important for wolf
occurrence.
The model with variables identified as most important for explaining wolf presence in chapter 3
(livestock density and road density) classified 72.5% of the studied squares correctly during
model development and 74.0% during model testing (Table 8). It explained wolf areas and nonwolf areas to a similar degree and was considered as suitable by the Hosmer-Lemeshow
goodness-of-fit test (p = 0.506).
4.5
Discussion
This study supports the findings of chapter 3 that wolves choose areas with lower anthropogenic
influences for living. Especially areas with regular wolf presence, presenting core areas with
reproduction sites, are less fragmented by roads and settlements and hold lower human
population densities. Forest cover in these areas is particularly high. Already Ciucci et al. (1997)
and Theuerkauf et al. (2003a) stated that wolves choose places away from villages and roads and
with dense forest cover as breeding sites. After the wolf female gives birth, for some weeks she
stays within or close to the den together with her pups and is thus more vulnerable to human
disturbance (Tsunoda et al. 2008). Outside the core areas, in regions with irregular wolf
presence, human presence does not seem to have a major influence on wolves. In our study,
human population density and size of urban areas in these locations was even higher than in
wolf-free areas. This becomes particularly obvious in regions, where wolves forage within or
close to towns during night (Ciucci et al. 1997). During their movements they often use paved
roads for ease of travel (Ciucci et al. 1997, Whittington et al. 2005), however main roads are
usually avoided (Kaartinen et al. 2005, Theuerkauf et al. 2007, chapter 3 of this thesis) and in our
study area they represent barriers between neighboring wolf packs. The avoidance of settlements
is not that obvious. In some studies the proximity is avoided (chapter 3 of this thesis), but in
others it is not avoided (Theuerkauf et al. 2007) or even visited on a regular basis (Ciucci et al.
1997). Our data suggest that a high number of small settlements seem to have a stronger negative
effect on wolves than a few bigger settlements, as the former cause a higher degree of
fragmentation. Next to anthropogenic influences our Principle Component Analysis selected
Table 7: Meta-analysis of wolf habitat models created worldwide, showing variables identified as good predictors for the presence of wolves.
country method
critical variables
USA
logistic regression road density, fractal dimension
Swiss
logistic regression
variable category
anthropogenic influence
deer density
pasture, coniferous forest, crops,
wetland, open water
(road density)
human population, distance to
roads
livestock
land cover
linear regression
Spain
Mahalanobis
distance
Poland
multiple regression distance to border, forest cover
land cover
Poland
logistic regression forest cover
land cover
distance to major road, distance
to population nuclei
land cover
livestock density
Portugal logistic regression
road density
Poland
logistic regression
livestock density
forest cover
Glenz et al. 2001
86%
Grilo et al. 2002
Treves et al. 2004
anthropogenic influence
prey density
anthropogenic influence
land cover
Mladenoff et al. 1995
83%
prey density
land cover
anthropogenic influence
USA
cost surface
analysis
performance
authors of the study
[%]
93% / 86%
population density, arable lands anthropogenic influence
ungulate diversity index
prey density
Portugal logistic regression altitude, mixed forest
Spain
variables
excluded
prey density
excluded
prey density
excluded
82%
Cayuela 2004
59%
Jedrzejewski et al. 2004
73% / 88%
Jedrzejewski et al. 2005
prey density
excluded
Rodriguez-Freira &
Crecente-Maseda 2008
prey density
anthropogenic influence
79% / 83%
chapter 3 of this thesis
prey density
land cover
73% / 71%
this study
Validation of a wolf habitat model
43
prey density as most important for explaining wolf presence. In most other studies of wolf
habitat modeling worldwide, prey density was as well identified as the most important factor for
wolves (Glenz et al. 2001, Treves et al. 2004, Cayuela et al. 2004, chapter 3 of this thesis). In the
majority of studies, where prey density was not selected as an explaining variable during
modeling, it was not included at all, as no data existed (Jędrzejewski et al. 2004, 2005,
Rodriguez-Freira & Crecente-Maseda 2008). In those studies however factors were selected
during modeling, which are directly correlated to prey density. For example Grilo et al. (2002)
identified altitude as a good predictor for wolf occurrence though altitude itself most probably
has no importance for wolves. Rather it stands for a good equilibrium between avoiding human
presence in valley bottoms and the lack of prey in too high altitudes. In another habitat model,
generated on wolf occurrence data from Spain, land cover was classified as particularly
important for wolves, however they argued that it is associated with refuge from people and
availability of prey (Rodriguez-Freira & Crecente-Maseda 2008). Similarly, in two studies
conducted by Jędrzejewski et al. (2004, 2005), forest cover was a good predictor for wolf
presence, as forested areas usually hold lower human population densities and higher densities of
wild ungulate species. In our study as well forest was included into the model, however during
the PCA it was assigned to the ecological factor of anthropogenic influences and negatively
correlated with it. Blanco et al. (2005) could show that forest itself is not essential for wolf
occurrence, as wolves survived in agricultural landscapes and bred in cereal and maize fields.
Another issue to be considered is the substantial difference in behavior between wolf populations
that depend on dissimilar food resources as wild ungulates, livestock or garbage. As food
acquisition is the central aspect for the survival of a wolf, it strongly influences its behavior and
habitat selection. This explains why the availability of food is the most important factor for
wolves. However it is not easy to measure, as prey density is not equal to prey availability. For
example wild prey and domestic prey are not the same available for wolves. This became
obvious during our study, as wild ungulate densities were considerably lower in wolf areas than
in non-wolf areas, while for livestock densities it was the other way round. This difference can
be explained by the avoidance behavior of the wild ungulate species towards areas inhabited by
wolves (Stephens & Peterson 1984). On the contrary livestock is bound to its grazing grounds
and consequently can’t escape from wolves. Thus, wolves can choose habitats with high
livestock densities. This makes it difficult to apply habitat models based on wolf populations that
44
Validation of a wolf habitat model
Table 8: Results of the logistic regression analysis with variables selected in chapter 3 of this thesis and
percentages of correct classifications, obtained during model development and testing. Sensitivity and
specificity refer to the correctly classified wolf areas and non-wolf areas, respectively.
variable
livestock
road
density
constant
β±
S.E.
Wald
statistic
1.05 ±
0.23
-0.87
± 0.45
0.08 ±
0.36
d.f. p-value sensitivity specificity correct
goodness
(model /
(model /
predictions -of-fit
test)
test)
(model /
test)
20.99
1
0.001
3.73
1
0.054
0.05
1
0.822
73.1 / 73.5 71.4 / 75.0
72.5 / 74.0
0.506
prey mainly on livestock to wolf populations that depend more on wild ungulates, as it was the
case in this study. The meta-analysis revealed that wolf habitat models created worldwide do not
differ much, with all of them including the categories of anthropogenic influences and prey
density, even though sometimes indirectly through land cover variables. If the analysis and
variables entered into the modeling process had been standardized for all studies, we expect the
results to be identical.
4.6
Acknowledgements
The study was part of the “Bieszczady Wolf Project”, coordinated by Dr. Roman Gula, funded
by the Polish National Committee for Scientific Research (KBN 6P04F 006), budget of the
Museum and Institute of Zoology, Polish Academy of Sciences, and financially supported by a
scholarship of the “Allgemeines Promotionskolleg” and the Research School of the RuhrUniversity Bochum. I thank Stephan Radler for his help during field work.
45
The relationship between habitat quality and stress hormones in wolves
5
5.1
THE RELATIONSHIP BETWEEN HABITAT
HORMONES IN WOLVES (Canis lupus)
QUALITY
AND
STRESS
Abstract
The aim of our study was to evaluate possible influences of human presence on the stress level of
wolves, while controlling for confounding variables such as storage conditions of fecal samples,
seasonal effects and rank. The negative impact of anthropogenic pressure has been shown for a
variety of species, however frequently it becomes obvious only when the population is already
declining. Therefore, more sensitive methods for detecting such harmful influences in an early
stage are needed. We tested a non-invasive endocrine approach and assessed the stress level of
wolves by measuring the stress hormone concentration in 62 fecal samples with a cortisol
enzyme immunoassay. Stress hormone concentrations were not influenced by weather conditions
during sampling (except from feces with a high water content sampled during snow melt) and
did not show seasonal variation. Possibly high ranking wolves, identified by their marking
behavior at crossroads, exhibited higher levels of stress hormones. Furthermore, after controlling
for the before mentioned parameters, the stress level of the studied wolves was significantly
higher in habitats with a higher anthropogenic pressure. Therefore, we conclude that wolves are
negatively affected by the presence of humans and that non-invasive stress hormone analysis
represents an effective means for assessing such negative anthropogenic impacts prior to causing
severe harm on the population level.
5.2
Introduction
For the conservation and management of endangered species it is of particular importance to
assess their sensibility towards humans and anthropogenic structures. Human-induced stress
responses of several wildlife species by outdoor recreational activities and snow sports has
already been debated (Creel et al. 2002, Taylor & Knight 2003 and Arlettaz et al. 2007).
Moreover, next to human-induced stress there exists a variety of other environmental and
behavioral stressors that affect an animal. Possible stressors to an organism can be based on
changes in environment or food resources, seasonal parameters as weather and reproduction
(Kotrschal et al. 1998, Pereira et al. 2006, Petrauskas & Atkinson 2006, Arlettaz et al. 2007;
reviewed in Millspaugh & Washburn 2004), behavior and social status (Sands & Creel 2004,
Creel 2005, Mooring et al. 2006, Barja et al. 2008). Stress has a considerable impact on virtually
The relationship between habitat quality and stress hormones in wolves
46
all bodily functions; it can disrupt reproduction and cognition, alter the animal’s behavior and
degrade the performance of the animal’s immune system, resulting in reduced resistance to
disease (McEwan & Sapolsky 1995, Wingfield & Sapolsky 2003, Millspaugh & Washburn 2004,
Romero 2004, Korte et al. 2005, Touma & Palme 2005). That is why it affects not only the
fitness of the individual, but as well the probability of survival of the whole population. Several
approaches exist how to measure an animal’s stress level, such as behavioral observations or
measurement of physiological parameters, for instance heart rates (MacArthur et al. 1979, 1982,
Moen et al. 1982) or stress hormones. The latter has high potential, as stress provokes several
endocrine processes, which finally result in measurable alterations of hormone levels in various
body fluids or excreta (Möstl & Palme 2002). The primary response to a stressor is the activation
of the sympathetic nervous system (SNS), which induces the secretion of epinephrine
(adrenaline) and norepinephrine (Boonstra 2005, Reeder & Kramer 2005). To deal with the
stressor, these hormones provide extra energy through glucogenolysis and lipolysis and
accelerate its transport through the body by increasing heart rate, blood pressure and breathing.
Within minutes, this first cascade is followed by the stimulation of the hypothalamic-pituitaryadrenocortical (HPA) axis. The hypothalamus releases corticotrophin-releasing hormone, which
stimulates the anterior pituitary to secrete adrenocorticotropic hormone (ACTH). ACTH, in turn,
causes glucocorticoid (GC) secretion into the circulation by the adrenal cortex. In birds and small
mammals corticosterone is the primary glucocorticoid, whereas in medium to large mammals, as
well as in fish, cortisol or a mixture of both dominates (Millspaugh & Washburn 2004). For this
reason, GCs have been used as physiological indicators of stress in a variety of species (Touma
& Palme 2005). In the short-term, the stress response is an evolutionary adaptive process
(Arlettaz et al. 2007), which diverts energy from physiological actions not immediately required
for survival, toward actions needed to escape from the stressor. However, if it is prolonged or
repeated, it gets harmful by producing an array of pathologies and reducing the individual’s
fitness (Creel et al. 2002, Sands & Creel 2004, Arlettaz et al. 2007). The endocrine assessment of
stress used to be performed through measuring GCs in blood samples. This, however, brings
some disadvantages about, especially when endangered species are to be studied. As the animals
have to be captured to collect the blood samples, they get disturbed, eventually even brought into
danger during capture or handling, and accurate assessment of stress might be compromised by
the additional stress of capture (Millspaugh et al. 2002). However, in 1977 first noninvasive
endocrine analyses were carried out in birds by Czekala and Lasley and short time later in
The relationship between habitat quality and stress hormones in wolves
47
mammals by Möstl et al. (1983). Since then, this method was used gradually more often and in a
wider range of species, as it represents a remarkable improvement for the monitoring of
reproduction and stress associated with environmental disturbances in free-ranging wildlife
species (Wasser et al. 2000, Khan et al. 2002, Palme 2005). As capture and handling of the
animals is omitted, stress measurements are unaffected by the researcher and thus reflect the
actual stress level of the animal more accurately (Kotrschal et al. 1998). Another advantage of
noninvasive hormone measurements by fecal analyses is the pooling effect of glucocorticoid
metabolites over a time period, with gut passage time and dynamics of excretion determining the
temporal resolution of the method (Scheiber et al. 2005). Thus, the hormone concentration
represents a more accurate assessment of chronic stress, instead of showing short-term
fluctuations (Harper & Austad 2000). For these reasons, we chose noninvasive fecal GC
metabolite measurements to assess the stress level of wolves in the present study. The aim of the
study was to evaluate the impact of habitat quality on stress levels in wolves, controlling for
confounding factors as sample preservation, season and social rank.
5.3
Material and methods
Study area
The study was conducted in two distinct areas. One of the study regions is situated in the
southeast of Poland, in the foothills of the Bieszczady Mountains. The region (described in detail
in chapters 2 and 4), inhabited by five wolf packs, extends over an area of about 1000km². It is
directly connected to the eastern wolf range in Slovakia and the Ukraine. The second region is
situated in the Holy Cross Primeval Forest in central Poland, halfway between Krakow and
Warsaw. The wolf pack living in this forest was only discovered in 2006, when one of the
authors started monitoring it (Gula 2008). The last wolf record from this place was from 1953;
however, through state-organized killing campaigns wolves were subsequently extirpated in that
region and only returned decades later (Gula 2008). There has been occasional evidence of wolf
presence in the farther surrounding of the forest since the middle 80’s, but no documentations of
wolves living inside the forest exist. Connectivity to the eastern wolf range is questionable.
48
The relationship between habitat quality and stress hormones in wolves
(B)
6
(A)
8
frequency
frequency of occurrence
8
6
4
2
0
0
4
20
40
60
80
100
120
140
s t e ro id c o nc e nt ra t io n [ ng/ g]
2
0
0
10
20
30
40
50
60
steroid concentration [ng/g]
Fig 1: Frequencies of fecal glucocorticoid metabolite concentrations [ng / g wet feces] of wolves in all
fecal samples (A) and fecal samples selected for further analyses (B).
Data collection
We collected scats occasionally in 2004 (n = 2) and 2005 (n = 2). However, intensive scat
surveys we carried out from March 2006 until March 2007 (n = 58). We organized the search
along predefined, randomly chosen transects throughout the whole study area. We found most
fecal samples during winter season (December to February, n = 16) and early spring (March and
April, n = 30), when they were either frozen or at least remained cold since defecation. During
summer (May to August, n = 11) and autumn (October and November, n = 5) we acted with
particular caution to sample only those scats, classified as certainly fresh (based on texture and
smell). This is of particular importance, as fecal concentrations of cortisol metabolites are
dependent on microbial fermentation; consequently they continue changing after defecation and
thus inflate actual hormone concentrations (Khan et al. 2002, Washburn & Millspaugh 2002,
Sauerwein et al. 2004). We sampled roughly half of the fecal mass for stress hormone analysis,
whereas the other half was used for genetic and diet analyses. Samples were frozen unpreserved
within a few hours after collection and stored at -20°C until analysis, as recommended by Hunt
& Wasser 2003 and Millspaugh & Washburn 2004. Shipping was conducted within 24 hours on
dry ice. Additional information, if the sample was found at a crossroad or along the trail, was
noted next to date, coordinates and assumed age of the scat.
The relationship between habitat quality and stress hormones in wolves
(A)
49
(B)
Fig 2: Mean cortisol metabolite concentrations [ng/g wet feces] in (A) fecal samples of wolves collected
during winter (n = 23) or the rest of the year (n = 36) and (B) fecal samples deposited by wolves during
mating season (n = 15) or non-mating season (n = 44). Presented are box plots with medians, interquartile range, outliers and extreme cases.
Fecal analysis
Parent glucocorticoids circulating in the blood are metabolized in the gut and liver to a variety of
species-specific metabolites (Wasser et al. 2000), which are subsequently excreted via feces and
urine. Thus, parent GCs themselves are rarely detected in fecal samples and as well extraction
methods as immunoassays, developed for measuring GC concentrations in the blood, have to be
adapted accordingly (Hunt & Wasser 2003, Young et al. 2004). In our study, GC metabolites
were extracted from the feces by adding 5ml of 80% methanol to 0.5g of the fecal mass, shaking
(30 min) and centrifuging the suspension (15 min, 2500g), and decanting the supernatants. As
fecal steroid metabolites are a mixture of metabolites with different polarities, it is of importance
to test the efficiency of the extraction method to be used, beforehand. The above mentioned
protocol was developed by Schatz & Palme (2001) and tested on dogs and cats. It proved best for
all mammalian species studied so far (Palme 2005). For the measurement of GC metabolite
concentrations, either radioimmunoassays (RIAs) or enzyme immunoassays (EIAs) can be
conducted. During these assays labeled steroids compete with steroids from the fecal sample for
binding at the steroid-specific or group-specific antibody (Möstl et al. 2005). The difference
between the two immunoassays is the label of the competitive steroids, with radioactive labeled
The relationship between habitat quality and stress hormones in wolves
50
steroids used in a RIA and enzyme labeled steroids in an EIA. The latter is often preferred, as no
special laboratory for the work with radioactive substances is needed. One important point is the
validation of the selected immunoassay for the species of interest (Touma & Palme 2005), which
can be performed analytically, physiologically and behaviorally. During analytical validation,
radioactive labeled hormone is injected into the bloodstream of an animal and labeled hormone
metabolites measured in the feces. Hereby, the sensitivity, accuracy, precision, and crossreactivity of the respective immunoassay can be measured (Möstl et al. 2005). For physiological
validation, ACTH is injected, which mimics a natural adrenal stress response by causing a rapid
rise in circulating native GCs, and the increase of GC metabolite concentrations in the fecal
samples is measured (Wasser et al. 2000). A second possibility is the administration of
dexamethasone, which suppresses the adrenocortical activity, causing a decline in GCs, and the
subsequent measurement of decreased GC metabolite concentrations in the feces (Touma &
Palme 2005). Similarly, during biological validation the adrenocortical activity is stimulated by
bringing the animal into a stressful situation (e.g. capture, immobilization or transport) and
measuring the increase in GC metabolite concentrations afterwards (Touma & Palme). In this
study, a cortisol EIA was applied. This assay was validated analytically, physiologically, and
behaviorally in the red wolf (Canis rufus; Young et al. 2004), as well as physiologically (Schatz
& Palme 2001) and behaviorally (Palme et al. 2001) in the dog (Canis familiaris). For labeling,
DADOO-labeled cortisol-3-carboxymethyloxime (cortisol-3-CMO) was used (Palme & Möstl
1997). The steroid-specific antibody (against cortisol-3-CMO) was linked to bovine serum
albumin (BSA). The antibody cross-reacts to 100% with cortisol, to 6.2% with corticosterone, to
4.6% with 5α-pregnane-11β,17α,21-triol-3,20-dione and to less than 1% with 5β-pregnanes, 5αandrostanes, and 5β-androstanes (Palme & Möstl 1997). The enzyme immunoassay procedure
consists of 10 steps and was developed by Palme & Möstl (1997). As a first step, microtiter
plates were coated with a coating antibody and BSA. After incubating and washing the plates,
standards, pools, and samples were added. Afterwards biotin-labeled steroids were added,
followed by the steroid specific antibody. The microtiter plates were then incubated at 4°C-8°C
overnight, during which time the sample steroids and the biotin-labeled steroids compete for
binding at the steroid-specific antibody. Subsequently, the plates were washed, enzyme-solution
(streptavidin-peroxidase) added, incubated, and once again washed. As a last step, substratesolution (TMB) was added for starting the color reaction and H2SO4 for stopping it again. By
The relationship between habitat quality and stress hormones in wolves
51
measuring the optical density (O.D.) at 450nm and comparing the O.D. of the samples to the
standard curve, the steroid concentration could be calculated.
Data analysis
For comparing samples collected during the cold season (December to March, N = 23) to the
ones collected during the rest of the year (April to November, N = 36), we calculated means and
tested for significant differences with the Mann-Whitney U-test. We performed the same
analysis for comparisons between mating season (January to March, N = 15), during which
wolves show courtship behavior, and the rest of the year (April to December, N = 44).
Furthermore, we related the marking behavior to the concentration of stress hormones, by
dividing samples into feces deposited at crossroads and feces deposited along the trail. We
suppose that feces deposited at crossroads have a more pronounced marking function than those
placed along the trail, as proposed by Barja et al. (2008). Moreover we classified the home range
quality of the six studied wolf packs by estimating percentage of cover (forest and shrubs) and
percentage of artificial surfaces (industries, settlements and agricultural land). Subsequently, we
related these two parameters to mean stress hormone concentrations of the six wolf packs. We
calculated significant correlations by the Spearman rank correlation test.
Fig 3: Mean cortisol metabolite concentrations [ng/g wet feces] in fecal samples deposited by wolves at
crossroads (n = 9) or along the trail (n = 30). Presented are box plots with medians, inter-quartile range,
outliers and extreme cases.
52
The relationship between habitat quality and stress hormones in wolves
5.4
Results
As a total, 62 fecal samples were collected and analyzed. Cortisol metabolite concentrations
ranged from 0.6ng/g wet fecal mass to 139.2ng/g wet fecal mass (Fig 1(A)). Three samples
contained considerably higher concentrations (with 91.8ng/g, 123.4ng/g and 139.2ng/g) than all
other samples (maximum 53.9ng/g; Fig 1(B)). This might be due to high amounts of water in the
samples, as they were collected during snow melt. As Washburn & Millspaugh (2002) showed in
experiments with simulated rainfall, high water contents increase fecal GC metabolite
concentrations dramatically (nearly doubling within 7 days). They argue, that additional moisture
provides a perfect growth environment for microbes and detritivores, which heavily metabolize
steroid hormones and thus increase concentrations of immunoreactive steroids (e.g. by
deconjugation). Consequently, we excluded these fecal samples from further analyses. The mean
steroid concentration of the remaining 59 fecal samples was 11.4 ± 2.8ng/g wet fecal mass.
Mean steroid metabolite concentrations differed between months (minimum 3.1ng/g, maximum
18.4ng/g); however, the differences were not statistically significant (Kruskal-Wallis test, N =
10, P = 0.335). Moreover, no differences in immunoreactive steroid concentrations between cold
season and rest of the year existed (Mann-Whitney U-test, N = 59, P = 0.774; Fig 2 (A)), with
mean hormone levels of 11.6 ± 4.3ng/g in winter and 11.3 ± 3.7ng/g during the remaining year.
Likewise, GC metabolite concentrations did not change significantly during mating season
(Mann-Whitney U-test, N = 59, P = 0.086; Fig 2 (B)), though in this time, when wolves display
Table 1: Sample size, mean fecal cortisol metabolite concentrations [ng/g wet feces] and 95% confidence
intervals for the six studied wolf packs.
wolf pack
study area
n
mean hormone
concentration
[ng/g]
95%
confidence
interval
Andro
Bieszczady
6
11.0
10.2
Nika
Bieszczady
11
13.5
9.6
Łodyna
Bieszczady
5
16.3
9.4
Stebnik
Bieszczady
10
8.5
4.0
Paniszczew
Bieszczady
7
13.7
6.6
Holy Cross
Central Poland
20
9.8
4.3
total
total
59
11.4
2.8
The relationship between habitat quality and stress hormones in wolves
53
courtship behavior, hormone concentrations were somewhat higher (14.7 ± 5.9ng/g) than during
the rest of the year (10.3 ± 3.1ng/g). Marked differences in cortisol metabolite concentrations,
however, existed between fecal samples found at crossroads and those found along the trail (Fig
3). Hormone concentrations were significantly higher in samples from crossroads (24.3 ±
10.0ng/g wet feces) than in the other samples (8.6 ± 2.5ng/g wet feces; Mann Whitney U-test, N
= 39, P = 0.001). However, as sample size at crossroads was fairly small (N = 9), consequently
the confidence interval was larger than in the second group of samples (N = 30). Mean hormone
concentrations differed between wolf packs (Table 1), however, this difference was not
statistically significant (Kruskal-Wallis test, N = 59, P = 0.543). On the level of wolf packs,
cortisol metabolite concentrations were elevated in home ranges with a higher percentage of
urbanization (Fig 4). This correlation was significant (Spearman’s rho, N = 6, P = 0.019).
Respectively, cortisol metabolite concentrations were significantly lower in home ranges with a
higher percentage of by forest or shrub cover (Spearman’s rho, N = 6, P = 0.019). Other
anthropogenic influences, such as human population density and density of main roads, showed a
slightly positive correlation with stress hormone concentrations (R² = 0.384 and R² = 0.336,
respectively); however this correlation was not significant (Spearman’s rho, N = 5, P = 0.624 and
P = 0.624, respectively). Neither wild prey density nor domestic prey density correlated with
stress hormone concentrations (Spearman’s rho correlation coefficient, N = 5, R² = 0.048, P =
747 and R² = 6.8*10-4, P = 0.874, respectively).
5.5
Discussion
With our study on stress hormone levels in wolves we could support the findings of Barja et al.
(2008) that scats used for territory marking held elevated GC metabolite concentrations. Those
scats were supposed to function as territory marks as they were deposited at crossroads and thus
at locations frequently passed by other wolves. The importance of crossroads for fecal marking
was already described for various carnivores, e.g. Iberian lynx (Lynx pardinus, Robinson &
Delibes 1988), red fox (Vulpes vulpes, Macdonald 1980) and Iberian wolf (Barja et al. 2004). In
the wolf population studied by Barja et al. (2004), scats were preferentially placed at
intersections with several, easily accessible roads and on conspicuous substrates. This is thought
to enhance visibility of marks and consequently the possibility of being detected by other passing
The relationship between habitat quality and stress hormones in wolves
54
Fig 4: Correlation between mean cortisol metabolite concentrations [ng/g wet feces] in fecal samples of 6
wolf packs and percentages of the respective home ranges occupied by artificial surfaces.
wolves. In particular, crossroads leading to areas of exceptional importance, such as den sites or
rendezvous sites, were more intensively marked. Moreover, marking behavior in carnivores is
usually exhibited by dominant animals (Peters & Mech 1975, Zub et al. 2003), though mainly
urine and secretions from diverse glands (e.g. anal gland or inter-digital gland) were regarded to
have an olfactory marking function. However, some studies imply that as well feces are used for
territory marking, if they are deposited in small amounts or in conspicuous places (Asa et al.
1985, Barja et al. 2004, 2005). Barja et al. (2008) showed endocrinologically that the breeding
pair, identified by higher sex hormone levels during breeding season, marked crossroads by feces
and these ‘marking feces’ held higher GC metabolite concentrations than other ‘non-marking
feces’. These findings, as well as our herein presented results, support the hypothesis that
dominant animals suffer higher levels of stress than subordinates, as proposed for a variety of
species, especially for those living in permanent social groups with stable membership that were
studied in the wild (Sands & Creel 2004; reviewed by Creel 2001, 2005). However, several
authors advise to assess the relationship between fecal GC metabolite concentrations and
seasonal patterns before interpreting differences in stress hormone levels (Millspaugh et al. 2001,
Pereira et al. 2006). Without the knowledge of normal seasonal stress hormone levels, it becomes
difficult to assess the reason for observed variations (Millspaugh & Washburn 2004). Increased
The relationship between habitat quality and stress hormones in wolves
55
stress hormone levels during breeding season, as described for many species (reviewed in
Millspaugh & Washburn 2004), could not be detected in our study. The GC metabolite level was
slightly elevated during time of courtship; however, this difference was not statistically
significant. For wolves, higher stress hormone concentrations during breeding season were
observed by McLeod et al. (1996). However, this study was carried out in captivity and thus
can’t be directly compared to studies on free-ranging wolves. The reason for elevated GC levels
in captive wolf packs is based on two main differences to wild living wolves. First of all, in the
wild, wolf packs represent family units, consisting of the breeding pair and its offspring of the
previous two years (Mech 1999). This is frequently not the case for captive wolf packs, which
include several adult, unrelated wolves that have to determine their hierarchy in aggressive
encounters (e.g. Schenkel 1947, Rabb et al. 1967). The second difference is, that the grown-up
offspring as well as subordinate wolves in captivity can’t migrate to start breeding in their own
territory or to escape from the stress of subordination (Creel 2001). These tensions escalade
during breeding season and can result in elevated stress hormone concentrations. However, also
for wild wolves, an increase in GC metabolite concentration during breeding season has been
reported (Sands & Creel 2004). The difference to our study might be the size of wolf packs
under observation. With 7 to 22 individuals per pack in the Yellowstone National Park, the
tension during breeding season might have been considerably higher than in our studied wolf
packs, consisting of no more than 7 individuals. Influences of weather on the stress level of
wolves in our study area, as well as influences of weather on the conservation of GC metabolites
in the fecal samples can be ruled out for our study, as no differences of fecal GC metabolite
concentrations existed between cold season and rest of the year. As Washburn & Millspaugh
(2002) could show by their study on effects of simulated environmental conditions on fecal GC
metabolites, differences in ambient temperature do not inflate the measurements. As well, Khan
et al. (2002) emphasis that storage effects might mask subtle physiological differences on the
individual level, but not substantial differences between study groups. Nevertheless, Washburn
and Millspaugh (2004) stress the importance of understanding and acknowledging effects of
sample age, if samples can not be collected fresh. The main finding of our study, though, is the
correlation of the stress hormone level with habitat quality; with higher fecal GC metabolite
concentrations in samples of wolves living under higher human pressure. Similarly, Creel et al.
(2002) found significantly higher stress hormone levels in fecal samples of wolves from areas
intensively traversed by snowmobiles, than in samples of wolves living in strict reserves.
The relationship between habitat quality and stress hormones in wolves
56
Moreover, in those areas fecal GC metabolite concentrations were only elevated during times of
intensive snowmobile traffic. In the same study they could show that elk (Cervus elaphus)
showed a stronger physiological stress response toward snowmobiles than toward wheeled
vehicles, and that the stress response increased with increasing numbers of national park visitors.
This might be explained by the fact that snowmobiles and visitors can unexpectedly appear
anywhere in the animals’ home range, whereas wheeled vehicles are more predictable by only
driving along the roads. Such negative effects of unpredictable events and uncontrollable
stressors have already been shown for other species (reviewed by Muller & Wrangham 2004 and
Petrauskas & Atkinson 2006). Likewise, correlations between stress hormone level and human
disturbance could be documented for a variety of other mammalian and avian species (e.g.
Arlettaz et al. 2007 [black grouse Tetrao tetrix], Thiel et al. 2008 [capercaillie Tetrao urogallus],
Pereira et al. 2006 [Pampas deer Ozotoceros bezoarticus bezoarticus], Barja et al. 2007
[European pine marten Martes martes] and Martinez-Mota et al. 2007 [black howler monkey
Alouatta pigra]). These findings show that increasing home range fragmentation by
anthropogenic infrastructure and high levels of human disturbance, for instance by outdoor
recreational activities and snow sports, represent a serious threat for wildlife (Arlettaz et al.
2007). Though, non-invasive endocrine analyses provide a sensitive method to assess the stress
level of wildlife species prior to a demographic response or decline in population size (Creel et
al. 2002) and thus make it a powerful tool for wildlife managers and conservationists.
5.6
Acknowledgements
The study was part of the “Bieszczady Wolf Project”, coordinated by Dr. Roman Gula, funded
by the Polish National Committee for Scientific Research (KBN 6P04F 006), budget of the
Museum and Institute of Zoology, PAS, and financially supported by scholarships of the
“Allgemeines Promotionskolleg” and Research School of the Ruhr University Bochum. I thank
Stephan Radler, Artur Milanowski and Krzysztof Król for their help during field work. Hormone
analyses by EIA were conducted at the University of Veterinary Medicine in Vienna at the
Institute of Biochemistry by the working group of Prof. Rupert Palme.
57
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
6
6.1
GENETIC MONITORING OF A PORTUGUESE
POPULATION BY MICROSATELLITE ANALYSIS
WOLF
(Canis
lupus)
Abstract
Non-invasive genetic monitoring of wildlife species has become increasingly popular in the last
decade, as it allows the assessment of a broad range of important population parameters, while
eliminating the need to capture and handle the target animals. I applied this technique to answer
questions on the genetic diversity of the studied wolf population, as a tool to identify a possible
barrier effect of major roads crossing the study area, as well as for pack recognition and
assessment of pack dynamics. In addition, I estimated the total population size in the study area.
By means of four highly polymorphic microsatellite systems I genotyped 69 fecal samples,
which revealed a total of 23 distinct genotypes. The population showed heterozygosity excess,
with an expected heterozygosity of HE = 0.77 and an observed heterozygosity of HO = 0.85. The
F-statistics revealed no noteworthy barrier effect of any of the major roads. Through
microsatellite analysis all four wolf packs in the study area could be detected and each individual
wolf unambiguously assigned to one of them. The number of wolves in each pack was between
four and eight, which represents a reasonable pack size. The overall population size was
extrapolated based on an exponential model and estimated as 25-26 wolves. With 23 genotyped
wolves, consequently, about 90% of the total wolf population within the study area was sampled.
The method employed proved useful for the assessment of population parameters, such as
genetic diversity and population size. Moreover, it could be applied successfully for detecting
packs and pack dynamics, as well as for an environmental impact study of anthropogenic barriers
such as major roads.
6.2
Introduction
To formulate predictions about the viability of populations in the long-term, reliable estimates of
their population size, as well as sex ratio, age structure and genetic variation are needed (Eggert
et al. 2003). The more, for an effective conservation management the effect of environmental
changes and anthropogenic disturbances on population size have to be known. However,
determining accurate estimates of population size for rare or elusive species, such as the wolf,
can be difficult (Creel et al. 2003, Piggott & Taylor 2003). Traditional field census techniques,
such as direct observations, capture-recapture methods or radio-tracking, often lack the precision
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
58
needed, as animals are difficult to detect or capture (Kohn et al. 1999), or suffer from small
sample sizes, especially in case of endangered species and carnivores (Mills et al. 2000). For a
social species as the wolf, other interesting aspects, such as population dynamics, inter-pack
connectivity and dispersal rates, can’t be estimated exclusively by field observations (Lucchini et
al. 2002). Therefore, non-invasive methods are needed, which do not require capture and
handling of the target animal (Hausknecht et al. 2005), but nevertheless give precise
demographic information about the target species. The appropriate tool was found some years
ago, when the use of molecular techniques in ecological research gained more importance (Haig
1998). Meanwhile, it was used to monitor dangerous or difficult-to-observe species, such as
bears (Kohn et al. 1995, Taberlet et al. 1997, Paetkau et al. 1998, Woods et al. 1999), marine
mammals (Palsbøll et al. 1997, Reed et al. 1997), mountain lions (Ernest et al. 2000), elephants
(Eggert et al. 2003), and coyotes (Kohn et al. 1999). As a source of DNA mainly feces were
used; however, hairs and feathers are the same effective. These remains can be collected without
disturbing the target animal, thus making the technique feasible not only for ecological, but as
well for behavioral studies (Taberlet & Luikart 1999). It can be used to gain information on the
minimum number of animals in an area, the sex ratio and genetic relatedness, movement
patterns, scent marking and home range size (Smith et al. 2006). Individuals can be distinguished
through microsatellite analysis. Microsatellites are repetitive sequences consistent of iterations of
1-5 base pairs, scattered throughout the genomes of higher organisms (Fredholm & Winterø
1995). These microsatellites are often highly polymorphic due to an elevated mutation rate, as
they vary in the number of repeats (Fredholm & Winterø 1995, Slatkin et al. 1995), and thus
provide information on relationships and divergences of geographically proximate populations
(Gerlach & Musolf 2000). They can be amplified by PCR, which is an enzymatic process during
which a specific region of DNA is repeatedly replicated to gain millions of copies of this
particular sequence (Taberlet et al. 1996). For a PCR, nanogram quantities of genomic DNA are
sufficient (Hausknecht et al. 2005). However, one has to keep in mind that non-invasive samples
usually provide degraded and low-quantity DNA. Therefore, PCR success rates are lower and
genotyping errors higher than in standard population genetic surveys (Santini et al. 2007) that
use blood or tissue samples. As microsatellites are comparatively short fragments, they can
nevertheless be amplified with sufficient success rates even from fairly degraded DNA.
Based on previous information on the studied subpopulation of wolves gained by non-invasive
monitoring techniques (chapter 3), I performed this study to achieve additional information on
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
59
population and pack size, pack dynamics, relatedness within and between packs, as well as
genetic diversity of this subpopulation. As a supplementary aim, I attempted to assess the impact
of anthropogenic barriers on the genetics of the studied wolves, by analyzing the effect of three
major roads cutting the study area into four parts.
6.3
Material & Methods
Study area
The study was conducted in central northern Portugal, north of the river Douro, in an area of
about 2000km² (Fig 1). The study area extends 95km from southwest (2km north of the river
Douro) to northeast (up to the border with Spain), with a width varying between 8km and 32km.
The studied wolves belong to the Iberian wolf population, which was, next to Italy, one of the
isolated populations surviving the wolf eradication programs throughout Western Europe in the
18th and 19th century (Breitenmoser 1998, Boitani 2003). The studied subpopulation was always
NW
E
W
Fig 1: Location of the study area within Portugal (small map) and detailed information on the study area
(major map). Indicated are major roads, state border and sampling locations of the genotyped scats (each
symbol represents one genotype). NW, E and W mark the three possible subpopulations separated by
major roads.
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
60
connected to the Iberian wolf range in Portugal and Spain, though still being continuously
reduced in numbers and driven back to the North of the country. As their wild prey species (i.e.
roe deer Capreolus capreolus and wild boar Sus scrofa) are rare in the study area, and wolves
consequently prey on livestock (i.e. goats Capra hircus and sheep Ovus aries, Carreira &
Petrucci-Fonseca 2000), conflicts with local farmers are frequent. As a consequence, numbers of
illegal killing through shooting or poisoning are high, though wolves are protected by law yearround in Portugal since 1988. Another threat to wolves in this area is a continuously progressing
fragmentation by anthropogenic infrastructure, such as three major roads with up to 700 cars per
hour that cut the study area into 4 parts.
Sample collection & conservation
From June 2005 until March 2007 we conducted a total of 718km of transects in search of wolf
scats. 368km were checked for scats regularly every three months. These transects were pseudorandomly distributed over the whole study area. We collected fresh scats for DNA analysis and
all other scats for diet analyses. Fecal DNA quality is reduced by ageing of the samples,
particularly under disadvantageous environmental conditions (e.g. high temperature, high
humidity or direct exposure to sun or rain, Nsubuga et al. 2004), through degradation by
decomposer organisms, such as bacteria, fungi, worms or insects (Santini et al. 2007). Such poor
quality DNA from non-invasive samples can result in contamination, allelic dropout or false
alleles (Taberlet et al. 1996). Allelic dropout can occur when template DNA is pipetted from a
very dilute extract into the PCR reaction and only one of two alleles is pipetted and amplified
(Taberlet & Luikart 1999). As a consequence, a heterozygote is becoming a false homozygote.
False alleles are artifacts (so called shadow bands) below the main band, produced by slippage
during the amplification process (Taberlet et al. 1996). Dinucleotide microsatellites are more
likely to have shadow bands (Taberlet & Luikart 1999). To avoid such genotyping errors, we
immediately placed the fresh scats in plastic bags and vacuum closed them. Within a few hours
we froze them at -4°C and after a few weeks at -26°C. We did not dry them or treat them with
any chemicals.
DNA extraction & amplification
For DNA extraction, 50-300mg of each frozen scat was scratched from the surface. The rest of
the sample was directly frozen again, as for some samples the extraction was not successful and
61
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
Table 1: Microsatellite systems used for genotyping of wolf scat samples
core
sequence
allele
size
CPH2
F: TTCTGTTGTTATCGGCACCA
R: TTCTTGAGAACAGTGTCCTTCG
(AC)15
96108
6
CPH4
F:ACTGGAGATGAAAACTGAAG
ATTATA
R: TTACAGGGGAAAGCCTCATT
(TG)17
141149
5
FH2088
F: CCCTCTGCCTACATCTCTGC
R: TAGGGCATGCATATAACCAGC
(TTTA)10
(TTCA)4
114132
7
FH2096
F: CCGTCTAAGAGCCTCCCAG
R: GACAAGGTTTCCTGGTTCA
(GAAT)9
96108
5
loci
primer sequence
no. of
alleles
reference
Fredholm
&
Winterø
1995
Fredholm
&
Winterø
1995
Francisco
et al.
1996
Francisco
et al.
1996
had to be repeated. The samples were incubated in CTAB buffer with the addition of proteinase
K over night at 37°C. On the following day, phenol-chloroform purification was conducted to
get rid of PCR inhibitors, followed by alcohol precipitation and storage of the extracted DNA in
TE buffer at -26°C. For removing RNA, which can act as a PCR inhibitor, RNAse A was added
to the extracts and another phenol-chloroform purification and alcohol precipitation step
conducted. The purified DNA was stored in TE buffer at -26°C until further treatment. To detect
the presence of DNA in the extract, an agarose gel electrophoresis was run and the DNA
visualized by ethidiumbromide under UV-light. I used four microsatellite loci (Table 1) to
genotype the fecal samples by PCR. The number of loci used for genotyping should be as small
as possible required to reach a low probability of identity (PI), as more loci increase the
likelihood of errors producing false genotypes and thus more individuals than actually present
(Creel et al. 2003). The PI describes the probability that two individuals within a population have
the same multilocus genotype. I selected the 2 dinucleotide microsatellite systems (CPH2 &
CPH4; Fredholm & Winterø 1995) and 2 tetranucleotide microsatellite systems (FH2088 &
FH2096; Francisco et al. 1996) due to their polymorphism, shown for wolves in several other
genetic studies within Europe (Lucchini et al. 2002, Hausknecht et al. 2005, Jędrzejewski et al.
2005, Aspi et al. 2006, Fabbri et al. 2007). Furthermore, I chose these microsatellites, as
Lucchini et al. (2002) could show that they do not amplify DNA from ungulates, and thus prey
DNA does not affect genotype determination. As well they, demonstrated that there exists a large
62
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
Table 2: Allele frequencies, expected (HE) and observed (HO) heterozygosity and F-statistics of the four
microsatellite systems used to genotype wolf fecal samples
microsatellite
loci
CPH2
CPH4
FH2088
FH2096
allele size [bp]
96
98
100
104
106
108
141
143
145
147
149
114
116
118
122
125
128
132
96
98
102
106
108
allele
frequency
0.25
0.20
0.05
0.23
0.23
0.05
0.30
0.16
0.23
0.23
0.07
0.07
0.10
0.05
0.21
0.21
0.29
0.07
0.14
0.16
0.43
0.14
0.14
HE
HO
F
0.79
0.83
-0.05
0.77
0.87
-0.13
0.81
0.91
-0.13
0.73
0.78
-0.07
variance in allelic dropout rates in between microsatellite systems, with three of the herein used
four microsatellites (CPH2, FH2088 and FH2096) being more efficiently amplified than others.
Microsatellites were amplified by PCR in a volume of 10µl, containing 1-2µl template DNA
solution (adjusted to the concentration of the respective DNA extract), 1µl PCR buffer, 3mM
MgCl2, 2mM of each dNTP, 1µl forward and reverse primer, and 0.25µl Taq polymerase
(5prime). To detect possible contaminations, negative controls without template DNA solution
were run with each PCR. PCR conditions included an initial denaturation step (4min at 94°C),
followed by 40 cycles of denaturation for 30s at 94°C, annealing for 30s at temperatures
optimized for each primer (the annealing temperature was decreased by 2°C after 10 cycles and
by another 2°C after the next 10 cycles) and elongation for 30s at 72°C, and finished by a final
elongation step of 4min at 72°C. In the first amplification cycles higher annealing temperatures
were used to increase the specificity of amplification, i.e. the annealing accuracy of the primers
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
63
to the target sequence. Later during PCR, annealing temperatures were decreased to increase the
amount of target DNA yielded. The analysis of the microsatellites was performed on a
polyacrylamide gel electrophoresis (PAGE) under UV-light with the program BioDocAnalyze
(Biometra).
Data analysis
I assessed the impact of the scats’ age at sampling, as well as the impact of the sampling season
on the PCR amplification success by comparing DNA extracts with PCR product to those ones
without PCR product. Based on the four microsatellites, I assessed the number of individual
genotypes, as well as the number of alleles and the allele frequency at each of the microsatellite
loci. As a means to detect inbreeding through bottleneck effects, I calculated the expected
heterozygosity HE and observed heterozygosity HO and tested for heterozygosity excess with the
inbreeding coefficient F:
F = (HE – HO) / HE
(1)
Since the average heterozygosity at several polymorphic loci is a relative measure of
heterogeneity within the population, analysis of heterozygosity within microsatellite loci is
suggested as a measure for the diversity of populations (Fredholm & Winterø 1995).The
expected heterozygosity is defined as:
HE = 1- ∑pi2
(2)
Further, I calculated the probability of identity (PI) as recommended by Paetkau and Strobeck
(1994) according to the following equation:
PI = ∑ pi4 + ∑ (2 pi pj) ²,
(3)
with pi and pj being the frequencies of the ith and jth alleles of the respective microsatellite locus.
I computed the multi-locus PI by multiplying single locus probabilities, starting with the highest
PI, assuming that loci are independent, as suggested by the microsatellite linkage map of the
domestic dog (Neff et al. 1999). In natural populations containing many siblings, a high
probability of the PI being biased and underestimated exists (Donnelly 1995). To assess the
upper limit of the population’ proportion having genotypes that cannot be distinguished from at
least one other individual, I calculated a PI for sibs as proposed by Taberlet and Luikart (1999):
PIsibs = 0.25 + (0.5 ∑ pi2) + [0.5 (∑ pi2) 2] – (0.25 ∑ pi4)
(4)
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
64
0.45
probability of identity
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
1
PI
PI sibs
2
3
4
number of microsatellite loci
Fig 2: Probability of identity (PI) and PI for sibs calculated for 1-4 microsatellites used in this study
Moreover, I assessed the distribution of distances between collection locations of (i) randomly
chosen pairs of scats belonging to the same genotype and (ii) randomly chosen pairs of scats
belonging to any genotype. To detect a potential substructure in the studied population, I
compared the geographic distance between two scat samples with the genetic distance by the
Spearman’s rho correlation coefficient. The genetic distance between two samples was
determined by the number of common alleles over all microsatellite loci, divided by the possible
number of common alleles, i.e. eight alleles at four microsatellite loci. Furthermore, I tested if
the three major roads, segregating the study area into four parts, act as barriers to a genetic
exchange between wolves. As a first measurement, I explored the data set for genotypes found
on both sides of the roads. As a second measurement, I calculated allele frequencies on both
sides of the roads and examined them for significant differences. Based on information about
wolf packs in the study area (chapter 3), I searched for private alleles, i.e. alleles that occur only
in particular packs, determined the number of individual genotypes per wolf pack, as well as the
degree of relatedness among the four wolf packs. To learn about pack dynamics, I compared the
genotypes found in two consecutive sampling years. To test if the number of unique genotypes,
identified in this study, is a reasonable number of wolves in the study area, I estimated the total
size of the studied wolf population by extrapolating from my sub sample. Therefore, I first used
an accumulation curve and plotted the number of scat samples analyzed against the number of
unique genotypes identified in these samples. I randomly selected a given number of samples in
100 sub-sampling runs and calculated the mean number of genotypes found in these 100
65
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
simulations. From the resulting curve I extrapolated by means of nonlinear regression an
exponential population size model described in Eggert et al. (2003) by the following formula:
E(x) = a (1 – e (bx))
(5)
This model is commonly used for estimating species richness (reviewed in Gotelli & Colwell
2001).
50
50
40
(B)
frequency of occurence
frequency of occurence
(A)
30
20
10
0
40
30
20
10
0
5
15
25 35
45
55 65
75
geographic distance [km]
85
5
15
25
35
45
55
65
75
85
geographic distance [km]
Fig 3: Distribution of geographic distances between randomly chosen pairs of fecal samples with the
same genotype (A) and randomly chosen pairs of all fecal samples (B)
6.4
Results
Of 84 samples used for DNA extraction, 69 could be successfully amplified and genotyped. The
amplification success depended on the season during which scat samples were collected, with
significantly lower failure rates for samples collected during winter than during the rest of the
year (U-test, p = 0.0001). Older scat samples resulted in lower amplification success; however,
this difference was not statistically significant (Kruskal-Wallis test, p = 0.089). In the 69 scat
samples that could be successfully genotyped, the four microsatellite systems revealed 23
genotypes. The number of recaptures of the unique genotypes ranged from 1 to 4, with an
average of 2 recaptures. All microsatellites were polymorphic, with 5 to 7 alleles (Table 1). The
allele frequencies were not distributed equally, but some alleles occurred frequently, whereas
others could be detected only sporadically (Table 2). The expected heterozygosity HE ranged
from 0.73 to 0.81, with an average heterozygosity over all four loci of 0.77. Thus it resembles
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
66
Table 3: Genetic structure calculated by F statistics (according to Weir & Cockerham 1994) for three
subpopulations of wolves separated by major roads. NW, E and W indicate the three possible
subpopulations separated by major roads.
Sample regions
FIS
FST
NW / W
-0.15
0.056
W/E
-0.13
0.033
NW / E
-0.16
0.026
the observed heterozygosity HO with 0.78 – 0.91 and an average of 0.85 (Table 2). The
inbreeding coefficient was close to zero; therefore, a bottleneck of this population can be
excluded. A slight heterozygosity excess, with more heterozygotes observed than expected,
supported the absence of inbreeding in this population. The probability of identity (PI) was
above 0.1 for one microsatellite locus, but already with two microsatellite loci the PI was close
to zero (Fig 2). However, the upper limit calculated by the PI for sibs was much higher and still
exceeded 0.01 with 4 loci. The distance between samples ranged from 0km to 76km. However,
there was an apparent difference in the distribution of distances when (i) randomly chosen pairs
of samples with the same genotype (Fig 3 (A)) were built, compared to (ii) randomly chosen
pairs of all samples (Fig 3 (B)). In the former analysis most sample pairs were found 0-15km
apart from each other. Only very few samples were found more than 25km away from each
other. In the latter analysis, however, the distances between sampling locations were almost
equally distributed throughout the whole distance range. The comparison of the geographic
distance between two samples to their genetic distance showed, that individuals closer related to
each other were also found in a smaller distance from each other. This correlation was highly
significant (Spearman’s rho correlation; p = 0.0001). One of the major roads, cutting the study
area into two parts from north to south, was crossed by five individuals during the study period.
Two of them crossed it in the north of the study area and three others in the center. This could be
detected, as samples of the respective genotypes were found on both sides of the road. Another
major road in the south of the study area was crossed by one individual. I compared allele
frequencies on both sides of the roads and calculated corresponding FIS and FST values (Table 3).
The FIS parameter showed negative values close to zero for all pair-wise comparisons of the three
possible subpopulations. These negative values indicate heterozygosity excess and thus are
another prove against inbreeding within the hypothesized subpopulations. The average FST values
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
67
CHAVES
r = 0.42
SAMARDÃ
PADRELA
r = 0.37
r = 0.41
VAQUEIRO
Fig 4: Schematic diagram presenting the mean relatedness among the four studied wolf packs (arrows).
Indicated are major roads (dashed grey lines) cutting the population, consisting of four wolf packs, into
four subpopulations.
over all four microsatellite loci ranged from 0.026 to 0.056 for the comparison of the three
subpopulations. The highest value occurred when I compared the allele frequencies on both sides
of the major road cutting the western half of the study area into a northern and a southern part.
The three wolf packs, located in chapter 3 (“Vaqueiro”, “Samardã” and “Padrela”) consisted of
6, 4 and 5 individuals, detected in 18, 15 and 16 fecal samples, respectively. Another pack in the
north of the study area (“Chaves”) comprised 8 individuals, detected from 20 fecal samples.
From all 23 alleles that occurred over the 4 microsatellite loci only one was a private allele,
which occurred exclusively in the central-eastern pack (“Padrela”) of the study area. In this pack,
however, the allele was as well detected only in two of the 5 individuals. The overall relatedness
of all individuals over all packs was r = 0.363 (N = 253). There was no significant difference
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
68
between the degree of relatedness within packs, compared to the one between packs (U-test, N =
253, p = 0.323). However, wolves within two packs (“Padrela”: r = 0.44, N = 10 and “Samardã”:
r = 0.42, N = 6) were closer related than the ones within the two other packs (“Chaves”: r = 0.36,
N = 28 and “Vaqueiro”: r = 0.36, N = 15). As well, there were variations in relatedness among
the different packs (Fig 4). The “Padrela” pack was significantly closer related to the “Chaves”
pack (r = 0.42, N = 40) than to the other two packs (U-test, p < 0.001 and p < 0.01, respectively).
The two other packs (“Vaqueiro” and “Samardã”) were closer related to each other (r = 0.41, N =
24), than both of them to the “Padrela” pack (U-test, p < 0.001 and p < 0.05, respectively). In the
first sampling year 21 of the 23 genotypes were detected. In the second sampling year only 15
genotypes were sampled, with 13 recaptures and two new genotypes. Eight individuals of the
previous year were not resampled. Based on an extrapolation of the accumulation curve (Fig 5)
the calculated population size model E(x) = 25.2 (1 – e (-0.042 x)) predicted a total population size
of 25 to 26 individuals in the study area, with a probability of 99.8%. Based on this model 92%
of the total population were sampled.
6.5
Discussion
With 69 out of 84 scat samples that could be successfully amplified and genotyped, the PCR
success rate was around 80%. In the remaining 20% of the samples the DNA might have been
too dilute, and low amounts of template DNA are known to severely limit the PCR success
(Santini et al. 2007). One possibility is to increase the amount of extract added to the PCR
reaction and thus increase the concentration of template DNA for amplification; however, this
raises the problem of exhausting the sample material before all microsatellite loci have been
successfully amplified (Morin et al. 2001). Another measure that can be taken is to increase the
number of PCR cycles, in order to get a detectable PCR product (Taberlet & Luikart 1999).
Additionally I decreased the annealing temperature during PCR to yield higher amounts of
amplificate. Theoretically, only a single molecule can be amplified, as was shown by the
amplification of DNA extracted from a single spermatozoid (Arnheim et al. 1990). Another
reason for the high failure rate might have been a progressed degradation of the DNA or the
presence of PCR inhibitors, which Monteiro et al. (1997) identified as complex polysaccharides.
Moreover, they stress that fecal samples are still the most difficult specimens for DNA extraction
and amplification. In concordance with Piggott & Taylor (2003) season- and age-related
differences in PCR success were found in this study. Fresh samples collected during winter
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
69
number of genotypes
30
20
10
0
0
50
100
150
number of samples
observed
predicted
Fig 5: Accumulation curve of the average number of genotypes, detected from an increasing number of
samples analyzed (black dots). The grey line represents the extrapolation of the accumulation curve,
calculated by an exponential model (Eggert et al. 2003).
contained the highest amplifiable DNA contents. Already Lucchini et al. (2002) recommended
collecting fresh scats in winter along wolf snow tracks, if possible, to improve the quality of the
laboratory data. To save time and money during laboratory work, it is valuable to know the
lowest number of microsatellite loci needed to reliably distinguish between individuals. The
number of microsatellites needed depends on their polymorphism. Selecting highly polymorphic
loci can drastically reduce the number of loci that have to be scored to achieve a given level of
accuracy (Blouin et al. 1996). This accuracy is particularly important in studies requiring
individual identification, as for example the herein presented genetic census study, because a
single error in a multilocus genotype already creates a false individual (Creel et al. 2003). This
can have disastrous consequences, if the aim is to assess the population size of endangered
species. Unfortunately, exactly such ecological or conservation biological studies have to rely on
non-invasively collected samples, as it was the case in this study. Moreover, these studies often
deal with highly inbred populations with a low level of heterozygosity, which in turn increases
the PI and possible misidentifications of individuals (Taberlet et al. 1999). This was fortunately
not the case for the herein studied wolf population, as an inbreeding coefficient close to zero and
heterozygosity excess indicated that no bottleneck occurred in the closer past. The average value
of expected heterozygosity was fairly high, compared to values of 0.41 to 0.70 calculated for the
bottlenecked Italian wolf population (Lucchini et al. 2002, Randi & Lucchini 2002, Fabbri et al.
2007). This supports the postulation that the largest wolf population in Western Europe survived
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
70
in the Iberian Peninsula, with a population size of 2000-2500 wolves in the early 1990s (Blanco
et al. 1992). The calculated heterozygosity value was similar to a Polish wolf population, with an
expected heterozygosity of 0.73 (Jędrzejewski et al. 2005). For that population as well
heterozygosity excess was documented. They argued that this might be due to a high human
caused mortality of wolves in that area, followed by a frequent immigration of wolves from the
east. The herein studied wolf population as well underlies a high anthropogenic pressure,
resulting in high mortality rates due to conflicts with livestock farmers (Carreira & PetrucciFonseca 2000). Consequently, frequent immigration of wolves from the stable population in
Spain is probable. This high genetic diversity of the studied wolves rendered the use of a large
number of microsatellite loci unnecessary, as it is needed for highly inbred populations (Kohn &
Wayne 1997). Moreover, no signs of a barrier effect of major roads in the study area were
detected. This could be confirmed, as samples of five wolves were found on both sides of the
roads, indicating repeated road crossings of these individuals. As well the FIS statistics, with
values close to zero, supported the genetic exchange between wolves from different sides of the
roads. However, the heterozygosity excess, resulting in slightly negative FIS values, hind to some
segregation. Also the FST values indicated that subdivision of wolf populations by roads, possibly
due to genetic drift, accounts for 2.6-5.6% of the total genetic variation. This segregation,
though, might have been an artifact, as in subpopulations, consisting of a family unit with
offspring of a single pair, significantly more heterozygotes are present than expected under
Hardy-Weinberg equilibrium (Gerlach & Musolf 2000). This is the case for wolf packs, which
are family units consisting of related individuals (Lehman et al. 1992). Such high genetic
similarity within packs was supported by the herein presented study. Fabbri et al. (2007) found
that the genetic similarity within groups is higher at the border of the wolf distribution range.
This fits to data of this study, with a high overall relatedness, as the study area is the southern
border of the continuous wolf range in Portugal. However, as well the degree of relatedness
between wolf packs varied, and packs living next to each other were closer related. This is in
concordance with further findings of Fabbri et al. (2007), who showed that genotypes in closer
proximity are more similar than genotypes with greater spatial separation, and hypothesized that
during recolonization processes local inbreeding hotspots are generated with the before
mentioned result. Furthermore, this study detected a migration of one individual from one wolf
pack – probably its natal pack – to a second wolf pack. During migration, this wolf crossed two
of the major roads in the study area, supporting the idea that they merely act as semi-permeable
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
71
barriers for wolves, though they do not disturb gene flow in a critical extent. The before
mentioned process of accepting strange wolves to already existing packs was also observed in a
Polish wolf population, which suffered high mortality due to legal and illegal hunting
(Jędrzejewski et al. 2005). They speculated that packs reduced to a few members might be more
tolerant to adopt strange wolves, in order to survive and rear young. Moreover, Meier et al.
(1995) deduced that even in the absence of a high anthropogenic pressure wolf packs are not as
stable and rigid as previously assumed. Through fecal DNA analyses such pack dynamics can be
traced back and long distance excursions, as well as core areas of wolf packs, detected.
Moreover, the distribution of distances among scats with the same genotype might present a new
approach to estimating home range size non-invasively (Smith et al. 2006). However, this might
require a substantial collection effort (Taberlet et al. 1997). In the herein presented study, scats
of the same genotype were predominantly found within the distance of 15km, which matches the
diameter of a typical wolf home range. Compared to previously collected field data (chapter 3),
samples with identical genotypes were almost exclusively collected within the borders of the
before calculated home ranges of the wolf packs. This shows the high potential of combining
field data with genetic data. Further, non-invasive genetics can also be used in ecological studies
to obtain an independent estimate of population size, in addition to field estimates (Smith et al.
2006). However, as Creel et al. (2003) could show, the method tends to overestimate the
population, as one genotyping error already creates a new individual. By using the minimum
number of microsatellites needed to obtain a low PI and by allowing genotypes with one
mismatch to be scored as identical, though, the overestimation can be reduced. A central
advantage is the potential to considerably increase the number of sampled animals in secretive
species, increasing the accuracy of estimates. Moreover, one should keep in mind that such
estimates are impressively accurate, considering the logistical hurdles required to estimate
population sizes for rare organisms such as carnivores (Mills et al. 2000).
6.6
Acknowledgements
The study was conducted in the laboratory of the SG Behavioural Biology and Biological
Education and financially supported by a scholarship of the Allgemeines Promotionskolleg and
the Research School of the Ruhr-University Bochum. Sample collection was part of an
environmental impact study coordinated by “Grupo Lobo” and supervised by Prof. Francisco
Fonseca. I want to thank Ana Guerra, Gonçalo Costa and Sílvia Silva for their help during
Genetic monitoring of a Portuguese wolf population by microsatellite analysis
72
fieldwork and the Alvão Natural Park services for their cooperation. Moreover, I want to thank
Markus Hold for introducing me into the laboratory work.
Predicting suitable wolf habitat in Germany
7
7.1
73
PREDICTING SUITABLE WOLF (Canis lupus) HABITAT IN GERMANY
Abstract
Wolves started resettling areas that used to be wolf-free since several decades and were
anthropogenically altered in the meantime. This will possibly result in elevated human-wolf
conflicts due to a competition for resources between wolves and humans. To reduce potential
conflicts and support the recovery of a German wolf population, it would be advantageous to
know in advance where dispersing individuals are likely to pass through and settle. Therefore,
the aim of this study was to identify potential wolf habitat in Germany, assess the potential wolf
population size, evaluate the connectivity between potential wolf habitats and discover areas of
elevated risks for human-wolf conflicts. By means of a habitat utilization study, developed by
logistic regression based on data of a Portuguese wolf population and evaluated on data of a
Polish wolf population, patches of optimal wolf habitat could be identified in the north, northeast and west of Germany. Approximately 4% of Germany presents optimal habitat and another
32% good habitat. These areas provide space for roughly 500-1500 wolves. For the resettlement
of north-eastern Germany structures facilitating dispersal (e.g. green bridges) are most important.
However, for the resettlement of western Germany more extensive conservation measures are
needed. Particularly the high risk areas in the centre and south of Germany require proper
conservation management plans to minimize human-wolf conflicts. The herein developed model
has the potential to increase the efficiency of conservation management actions for the
reestablishment of a viable German wolf population by setting priorities, thereby reducing costs
and increasing the acceptance of wolves in an as highly industrialized state as Germany.
7.2
Introduction
After a long time of wolf persecution resulting in a progressive decline of wolves over most of
Europe, slowly their numbers are increasing and their range is expanding again. As a
consequence, wolves return back to previously wolf-free areas, which is likely to produce an
increase in human-wolf conflicts. For this reason, it is of major importance for wildlife
conservation management not only to know where a species currently occurs, but also where it
could occur and which areas it most probably will resettle in the near future (Peterson &
Dunham 2003). This information can be helpful for supporting a further expansion of the wolf
Predicting suitable wolf habitat in Germany
74
through several conservation actions. First of all, the knowledge about potentially suitable wolf
habitats can be integrated into landscape planning (Rodes et al. 2006) and a further urbanization
and fragmentation of the identified areas kept as low as possible from a socio-economic point of
view. Moreover, isolated patches of suitable habitat can be connected by restoring natural
corridors, as well as constructing green bridges for facilitating the crossing of barriers (e.g.
motorways) during dispersal. To gain the information about where wolves are likely to appear in
the near future, habitat utilization models, the tool for such predictions, become of primary
importance. These models combine occurrence data with natural (biotic and abiotic) and
anthropogenic factors to create a model of the species’ requirements, which subsequently can be
mapped over the whole area of interest to predict the probability of occupancy (Anderson et al.
2003, Naves et al. 2003). Based on such predictions, moreover, a risk map can be constructed,
identifying areas with high potential of human-wolf conflicts, caused by livestock depredation
(Ciucci & Boitani 1998, Mech et al. 2000). The avoidance or minimization of such conflicts
would lead to a better acceptance of wolves by humans, which is basically the most important
aspect for wolf conservation in Europe (Boitani & Ciucci 1993, Chapron et al. 2003). Therefore,
habitat modeling is the key to a good conservation management, through the identification of
conservation priorities and thus an increase of its efficiency (Rhodes et al. 2006).
Germany is one of the western European countries for which such habitat suitability predictions
for wolf conservation are currently of particular importance, as wolves started immigrating from
a Polish subpopulation, settled in Saxony and currently expand their range westwards. Wolves
had been virtually extinct in Germany since the middle of the 19th century; however, only one
century later single wolves occasionally emigrated from Poland to the east of Germany, where a
total of 21 wolves were legally shot from 1948 until 1989 (Ansorge et al. 2006, Reinhardt &
Kluth 2007). After being fully protected throughout all Germany in 1990, another thirteen
wolves, mostly solitary dispersing animals, were killed through illegal shooting and car accidents
(Reinhardt & Kluth 2007). In 2000 the first reproduction of wolves in Germany could be
confirmed and wolf numbers increased during the following years to currently about 35-40
individuals. The general attitude of the German public towards wolves is positive (Kaczensky
2006), however illegal shootings and several losses due to car accidents demonstrate the
importance of further conservation measures.
The principle aims of this study were (1) the identification of suitable wolf habitats within
Germany, (2) the assessment of the capacity of identified suitable wolf habitats in terms of
Predicting suitable wolf habitat in Germany
75
potential wolf numbers, (3) the evaluation of connectivity between suitable wolf habitats and (4)
the identification of areas with an elevated risk for human-wolf conflicts.
7.3
Methods
Study area
The study was conducted in three separate study areas in Portugal, Poland and Germany. The
two small scale study areas in Portugal and Poland were used to develop and evaluate the wolf
habitat selection model. They comprised about 1000km² each. The areas were inhabited
continuously by wolves in the last decades, indicating that they represent good wolf habitat (for
detailed descriptions of the areas see chapter 3 and 4). The large scale study area incorporated all
Germany, which did not hold resident wolves since the end of the 19th century, until recently
when they started resettling eastern Saxony. Germany is a highly industrialized state, with an
average population density of 228 inhabitants per km². Nearly 8% of Germany is urbanized and
the traffic network is dense, with a total of 12,500km of motorways and 39,000km of highways.
Forest cover contributes to 28% of the country. Wild ungulates are represented by roe deer
(Capreolus capreolus), red deer (Cervus elaphus), fallow deer (Dama dama), sika deer (Cervus
nippon), wild boar (Sus scrofa), moufflon (Ovis musimon) and in a few places chamois
(Rupicapra rupicapra).
Model development
The wolf habitat utilization model was developed based on data collected in the Portuguese
study area (chapter 3). The model build in chapter 3 was, however, transformed to be applicable
to the data on livestock and wild ungulates from the large scale study area. Instead of using
absolute livestock densities in numbers per km², a livestock biomass index was calculated. The
respective sheep (Ovis aries) and goat (Capra capra) densities were therefore multiplied with a
coefficient of 0.5, to account for their lower weight in comparison to bigger ungulates (e.g. wild
boar or red deer). For validating the computed wolf habitat utilization model, data from the
second small scale study area in Poland were used (chapter 4). As already discussed in chapter 4,
densities of wild ungulates were not a good predictor for wolf abundance, as both variables were
negatively correlated, probably due to an adaptation of the prey to the presence of predators
resulting in avoidance of areas inhabited by wolves. Therefore, only data on sheep were
included, transformed into a livestock biomass index as described above. To assess the suitability
Predicting suitable wolf habitat in Germany
76
Table 1: Origin of the variables used for identifying and characterizing possible wolf habitats in Germany
variable
origin of data
road density
digitized by the author from a 1:600.000 scale road map
human density
European Environment Agency
urban areas
CORINE 2000 maps, scale 1:25,000
forest cover
CORINE 2000 maps, scale 1:25,000
livestock density
Landesämter für Statistik (2007)
wild ungulate density
Obere Jagdbehörden (2003-2008)
of the model, the calculated probabilities for wolf occurrence were compared by Spearman rank
correlation to the wolf abundance determined through Kernel probability analysis.
Model application
The large scale study area was divided into 92,738 squares of 2km x 2km and habitat variables
calculated for each of the squares (Table 1). For the application of the habitat model, livestock
density represented by number of sheep per km² was calculated and transformed into a livestock
biomass index. Data on cattle density were not included, as in Germany cows mostly remain
indoors and thus are not available for wolves. Furthermore, wild ungulate densities were
calculated based on yearly hunting bags for each county (except for North-Rhine Westphalia,
where data were only available on federal state level). Therefore, first the average yearly hunting
bag of the last 2-4 years was build for each ungulate species and multiplied by two to obtain an
approximate estimation of true ungulate densities (Reinhardt & Kluth 2007). The resulting
numbers, however, must be treated with caution, as they only represent a rough approximation
and not true densities. For the calculation of the biomass index, then, a coefficient of 1 was used
for red deer, fallow deer, sika deer and wild boar, a coefficient of 0.5 for moufflon and chamois
and a coefficient of 0.3 for roe deer, accounting for the difference in biomass between species.
The roads included into the habitat models, constructed based on data of the small scale study
areas, were considerably smaller with a lower traffic volume than the roads assessed for the large
scale study area. During model application this discrepancy had to be considered, as bigger roads
with a higher traffic volume are likely to have a larger negative impact on wolves and might
therefore decrease habitat quality more. To mitigate this discrepancy of the variable “road
density” in the different study areas, motorways and highways were buffered with 2km and a
77
Predicting suitable wolf habitat in Germany
combination of road density and buffered area around roads used as the explaining variable in
the habitat model. The distance of 2km was chosen, as this was the average area around bigger
roads found to be avoided by wolves in various studies (Theuerkauf et al. 2003, Kaartinen et al.
2005, Theuerkauf et al. 2007, chapter 3 of this study). The before calculated model was then
applied to the large scale study area. In a first model application step the suitability for wolves
was calculated for each of the 2km x 2km squares based on the variables “road density” and
“wild ungulate biomass index”. The following four suitability classes were distinguished: (1) 030% “unsuitable habitat”, (2) 31-50% “poor habitat”, (3) 51-70% “good habitat” and (4) 71100% “optimal habitat”. To further characterize the different suitability classes by land cover
and anthropogenic influences, for each of them average human population density, percentage of
urbanized areas and forest cover were calculated. Furthermore, patches of optimal habitat were
identified and measured. Based on information about the mean home range size and pack size of
(A)
(B)
Fig 1: Habitat suitability map of Germany, with “unsuitable habitat” in white, “poor habitat” in bright grey,
“good habitat” in grey and “optimal habitat” in black (A) and risk map of Germany, with “no risk” in white,
“low risk” in bright grey, “elevated risk” in grey and “high risk” in black (B). Red lines represent motorways
and the red polygon the present wolf area in Germany.
78
Predicting suitable wolf habitat in Germany
wolves in Europe, the capacity of these suitable patches in terms of potential population size of
wolves (number of wolf packs and wolves) was calculated. Moreover, the distance to the present
wolf area in Saxony was estimated, as well as barriers to wolf dispersal (e.g. motorways)
identified. The performance of the habitat utilization model was assessed by computing the
predicted suitability of the present wolf area in Saxony. In a second model application step, a
risk map was assessed, based on the variables “road density” and “livestock biomass index”,
identifying areas with low road density and high livestock density as suitable wolf areas with a
high risk of human-wolf conflicts, caused by depredation on livestock. Analogous to the
suitability model the following four risk classes were distinguished: (1) 0-30% “no risk”, (2) 3150% “low risk”, (3) 51-70% “elevated risk” and (4) 71-100% “high risk”.
7.4
Results
The adapted Portuguese wolf habitat utilization model included the two variables “livestock
biomass index” and “road density”. The coefficients of the new model are shown in Table 2,
together with the model performance in terms of sensitivity and specificity. When applying the
model to the second small scale study area in Poland, the overall predicted suitability was very
low. Such low estimates evolved from low livestock numbers in that area. However, as livestock
for food supply in that area is only of secondary importance to wolves and wild ungulates as the
main prey can’t be used for predicting wolf presence (for detailed explanations see chapter 4),
these low estimates do not represent the absolute suitability of that area correctly. Though, a
Table 2: Parameters of the habitat utilization model determined by logistic regression, based on data of
the small scale study area in Portugal and percentages of correct classifications. Sensitivity and
specificity refer to the correctly classified wolf areas and non-wolf areas, respectively.
variable
β ± S.E.
livestock
0.189 ±
biomass index 0.031
-0.738 ±
road density
0.224
-0.645 ±
constant
0.350
Wald
d.f. p-value sensitivity specificity
statistic
36.16
1
0.001
10.81
1
0.001
3.40
1
0.065
84.6%
54.7%
correct
predictions
74.2%
79
Predicting suitable wolf habitat in Germany
40
(B)
(A)
30
600
land cover [%]
population density [inh/km²]
800
400
20
10
200
0
0
unsuitable
poor
good
habitat suitability class
optimal
unsuitable
forest
urban areas
poor
good
optimal
habitat suitability class
Fig 2: Characterization of the four habitat suitability classes. Presented are mean population densities
(A), mean percentages of urbanization and forest cover (B) with 95% confidence intervals.
Spearman rank correlation test showed good concordance with wolf presence-absence data (p <
0.001), indicating that the model classified wolf areas as significantly more suitable for wolves
than non-wolf areas, although suitability values only ranged from 13% to 43%. This shows that
the transferability of the model is limited and rather presents relative, instead of absolute, habitat
suitability, if the explaining parameters and their biological relevance differ between study areas.
The first application of the model to the large scale study area classified 23% as “unsuitable
habitat”, 45% as “poor habitat”, 32% as “good habitat” and 4% as “optimal habitat” (Fig 1(A)).
The unsuitable habitat, with an average human population density of 540 inhabitants per km²,
was significantly denser populated than the other three habitat suitability classes with 180
inhabitants/km², 96 inhabitants/km² and 77 inhabitants/km², respectively (Mann-Whitney U-test,
P < 0.001; Fig 2 (A)). Forest cover was significantly elevated in optimal habitats, with 30%
compared to 16-20% (Mann-Whitney U-test, p < 0.001). There was, however, no obvious
difference between suitability classes concerning percentages of urbanized areas, with values
between 16.6% and 19.6% (Fig 2 (B)). The model identified 11 patches of optimal habitat with
the precondition that (1) a patch is not crossed by a motorway and (2) a patch covers more than
200km² (Table 3). Without precondition (1), five patches in the west of Germany (“EifelWesterwald-Hunsrück”) build a cluster of 7,760km², as well as two patches in the “OdenwaldSpessart” region with 1,860km² and two patches in the north-east of Germany (“HavellandUckermark”) with 4,650km². The other two patches (“Plön” and “Wittenberg”) are isolated and
Predicting suitable wolf habitat in Germany
80
cover 1,110km² and 1,580km², respectively. If an average home range size for wolf packs in
central Europe of 150-200km² is assumed (Ciucci et al. 1997, Findo et al. 2004, Kusak et al.
2005), a total of 85 to 113 wolf packs could settle in the before mentioned patches of optimal
habitat. The two bigger patch connections in the west and north-east of Germany could hold 3952 and 23-31 wolf packs, respectively. The other three smaller patches could hold between 6 and
12 wolf packs each. With an average pack size of 4-7 wolves (Apollonio et al. 2004, Findo et al.
2004, Jędrzejewski et al. 2004), consequently, a minimum of 340 to 791 wolves could inhabit
Germany. The suitable areas in the north-east and east of Germany (in the federal states of
Mecklenburg-Western Pomerania and Brandenburg) are 80-150km apart from the present wolf
area in Saxony (Table 3). However, they are separated by several highways and a few
motorways. The distance between the wolf area in Saxony and the potential wolf habitats in
western and northern Germany ranges between 350km and 450km. When additionally the areas
classified as good habitat are considered as possible wolf habitats, a vast area in the north-east of
Germany and another large area in the south-west might be settled by wolves. These areas
comprise approximately one third of Germany and could theoretically hold more than 800 wolf
packs with over 3,000 wolves. However, about half of the patches are either too small to hold
more than one pack or too isolated to hold a viable subpopulation of wolves. Therefore, to gain a
more reasonable estimate, pack numbers should be divided by two, resulting in an approximate
estimate of 1,500 wolves. For validating the applicability of the Portuguese model to
environmental conditions of Germany, the predicted suitability was calculated for the present
wolf area in Saxony. It contained no optimal habitat, but as well virtually no unsuitable habitat
(3.4%). Two-thirds of the area (67.8%) were classified as good habitat and the remaining 28.8%
as poor habitat. Consequently, the model can be accepted as applicable to the large scale study
area. In the second model application step, with “livestock biomass index” as the second
explaining variable next to “road density”, only 3% of all Germany had a high risk of humanwolf conflicts due to possibly high depredation on livestock and 15% had an elevated risk. The
areas identified as optimal wolf habitat only overlapped with high risk areas in less than 0.1%
and with elevated risk areas in 1%. Good wolf habitat coincided with high risk areas as well in
1% and with elevated risk areas in 6.5%. The main conflict areas were situated in the very north
of Germany (North Frisia, East Frisia and Pinneberg), in the centre of Germany (around the
Thuringian Forest) and in the south of Germany (Swabian Mountains, Fig 1 (B)).
81
Predicting suitable wolf habitat in Germany
Table 3: Characterization of areas identified as “optimal habitat” for wolves. Only areas bigger than
200km² were included. Minimum wolf pack numbers assume a home range size of 200km² per pack and
maximum wolf pack number a home range size of 150km² per pack. Minimum wolf numbers are based on
a pack size of 4 wolves and maximum numbers on a size of 7 wolves.
„optimal habitat“
patch
size
[km²]
Eifel –
Westerwald –
Hunsrück
EWH
Odenwald –
Spessart
OS
Havelland –
Uckermark
HU
Plön
P
395
1085
1280
1395
3605
970
890
1440
3210
1110
Wittenberg
W
1580
7.5
potential
no. of
packs
39 - 52
potential
no. of
wolves
156 – 364
distance to
present wolf area
[km]
420
9 – 12
36 – 84
350
23 – 31
92 – 217
140
6–7
24 – 49
380
8 - 11
32 – 77
80
Discussion
Suitable wolf habitats were identified based on road density as the anthropogenic variable
negatively influencing habitat suitability and on wild ungulate biomass as the natural variable
positively influencing habitat suitability. These two variables were selected previously, based on
data collected on a Portuguese wolf population (chapter 3). In several other studies on habitat
modeling the negative impact of roads on wolves has already been discussed (Thurber et al.
1994, Mladenoff et al. 1995, Mladenoff & Sickley 1998, Corsi et al. 1999, Cayuela et al. 2004,
Jędrzejewski et al. 2008, Rodriguez-Freira & Crecente-Maseda 2008). Newly colonizing wolves
were found to choose areas with road densities below 0.6km km-² (Thiel 1985); however, when
they expanded their range they started using also less favorable habitats with higher road
densities (Fuller et al. 1992). Roads decrease habitat quality as they are positively correlated with
the frequency of road deaths, ease of access for hunters, and the probability of meeting humans
(Rodriguez-Freira & Crecente-Maseda 2008). This correlation became obvious in the present
study as well, since human population density and percentage of urbanized areas decreased with
Predicting suitable wolf habitat in Germany
82
increasing habitat suitability for wolves, predicted by the model containing “road density” as one
of the explaining variables. Roads, however, might be the better predictor for unsuitable habitat
than population density or urbanized areas, as they combine local boosts of anthropogenic
pressure around cities with linear features connecting these boosts and fragmenting the
landscape. These linear elements might be even the more dangerous features for wolves,
compared to cities with high road densities, as collisions with vehicles are a major factor for the
mortality of wide-roaming carnivores (Lovari et al. 2007, Klar et al. 2008). Already Linnell et al.
(2001) found that human density alone is not a good predictor for the persistence of wolves,
since they may become extirpated in sparsely populated areas while surviving in regions with
high population densities. The natural predictive variable is not that straightforward and differs
in the varying studies on habitat suitability modeling for the wolf. In a Polish study for example,
forest and other natural land cover types were selected during a logistic regression analysis
(Jędrzejewski et al. 2008). They assumed that their model might be applicable to the Belarus
Republic, the Baltic States and Germany, because of broadly similar geographic conditions. In
the herein presented study, forest was not included into the model, though a retrospective
analysis showed that the optimal and good habitat patches contained significantly more forest
than less suitable patches. As wolves mainly feed on wild ungulates in both countries, this
concordance is reasonable, with wild ungulates staying mostly within or close to forests.
However, the herein produced model was based on data from a region less similar to Germany in
terms of land cover and prey composition, and this modeling approach discovered that instead of
forest cover the abundance of prey seems to be the explaining variable. Forest cover is rather an
indicator of low population density and high wild ungulate density. Already Mech (1995) and
Fuller (1995) stated that the wolf is a habitat generalist and does not require specific land cover
types, but exclusively depends on food resources and low killing by humans. Therefore, a model
as the one used in this study, combining anthropogenic influences and food abundance, is likely
to be the model with the widest applicability. One disadvantage of this model is that the
availability of food is not a straightforward variable and difficult to measure. However, if models
are to be used for supporting conservation actions, one important condition is that they should
stay as simple as possible and be based on empirical parameters that can be measured easily
(Côté & Reynolds 2002). Another issue that needs to be focused on is to standardize model
parameters. This kind of possible shortcoming could be observed in the herein presented study,
since the variable “road density” was not directly comparable between the three datasets for
Predicting suitable wolf habitat in Germany
83
model development, model evaluation and model application. The small scale study areas
contained only national and regional roads with a moderate traffic volume. The large scale study
area, on the contrary, was crossed by several thousands of kilometers of motorways and
highways with high to very high traffic volume, and regional roads were not included during
modeling. This fact turned the transferability of the produced habitat suitability model
problematic. To overcome this deficiency and to take into account the knowledge that wolves
avoid the surrounding of roads, additionally to the length of the roads a buffer zone was created
around each road and integrated into the model. Nevertheless, it is possible that the model is
positively biased and consequently the suitability as well as the calculated potential wolf
numbers overestimated. The final estimate of a potential population size of 1,500 wolves is in the
same dimension as predicted wolf numbers for Poland (Jędrzejewski et al. 2008). As the two
countries are of the same size, but Poland holds only half of Germany’s population density
(128inh/km² versus 228inh/km²) and much lower road densities (0.14km km-² in Poland with and
in Germany without regional roads), it was to be expected that Poland provides space for more
wolves than Germany. The similar predictions for potential wolf numbers in both countries
might also be explained by using no standardized variables. As additionally smaller road
categories were included into the Polish model, the overall suitability was similar as the one
assessed in this study. Including regional highways into the habitat suitability analysis for
Germany would decrease the overall suitability estimates and less patches would belong to the
good and optimal habitat classes. Nevertheless, the general distribution of suitable and unsuitable
habitats would stay the same as predicted in the herein presented study, even though potential
wolf numbers might turn out to be lower than predicted. As it could be shown that Germany
contains suitable habitat for wolves, the next important concern was the possibility of wolves to
resettle those selected areas. The north-eastern patches of good and optimal habitat are within the
range of a typical wolf dispersing from its natal pack (Kojola et al. 2006, Rodriguez-Freira &
Crecente-Maseda 2008) and thus close enough to the present wolf area to be resettled. A total of
11 wolves that were found dead in the selected areas in the last two decades prove this
(Reinhardt & Kluth 2007). However, it is not just the distribution of dispersal habitat that limits
patch connectivity, but also factors contributing to dispersal mortality, such as the dense road
network (Kramer-Schadt et al. 2004). The influence of this road mortality on wolf dispersal
became already obvious in Germany, as so far a minimum of three wolves that dispersed from
the present wolf area were killed on roads (Reinhardt & Kluth 2007). Therefore, conservation
Predicting suitable wolf habitat in Germany
84
actions should focus on facilitating the crossing of motorways and highways during dispersal, as
otherwise even expensive restoration and maintenance actions of good wolf habitat will be
useless, if it is not accessible for dispersing animals (Malo et al. 2004). More problematic,
however, is a natural resettlement of the potential wolf habitats in western Germany, as they are
separated from the eastern wolf range by numerous barriers and a large zone of unsuitable
habitat. The only corridor between these two potential wolf areas is a zone in central Germany
around the Thuringian Forest, which however was assigned to the high risk class. This area,
therefore, has the highest potential for human-wolf conflicts in Germany and consequently
should be in the focus of management actions. The reliability of the risk map developed in this
study could be validated for the present wolf area, which was classified as an area with a low risk
for human-wolf conflicts. Accordingly, a study on the wolves’ diet in that region discovered that
wolves feed to 97% on wild ungulates and remains of livestock were not found in the analyzed
scats (Ansorge et al. 2006). As high risk areas and suitable habitats for wolves overlap only in a
very few places, it is advisable to conduct conservation management actions to keep wolves out
of these high risk areas to avoid human-wolf conflicts.
This study could show that Germany contains regions suitable for wolves and that these regions
are large enough to hold some hundreds of wolves. However, it is important to constantly
evaluate and correct the model during the further expansion of wolves. Moreover, a more
detailed and extensive connectivity study should be performed, integrating factors such as
dispersal behavior of wolves and barriers to dispersal, as done in the study of Kramer-Schadt et
al. (2004) for lynx. As well, a viability study including data on the life history of wolves, as well
as on future demographic and land use changes, comparable to the study of Kramer-Schadt et al.
(2005) for the reintroduction of lynx in Germany, is recommended to legitimate costly
conservation management actions.
General discussion
8
8.1
85
GENERAL DISCUSSION
Human impact on wolves
Comparability of diverse monitoring techniques
To improve the power and transferability of study results, it is advisable to test them on
independent datasets (Pearce & Ferrier 2000). However, the appropriateness of monitoring
methods depends on the geographic location, physiography of the study area, funding and
logistics (Alexander et al. 2005), which are not always the same in different study areas. I
decided to conduct my research in three distinct study regions in Portugal, Poland and Germany
that differ with regards to climate, land cover, prey abundance and human presence. Due to
differences in climate between the Portuguese and Polish study area, not the same methods could
be applied to monitor wolf presence. In climates with cold and snowy winters, as in Poland,
snow tracking is a widely used tool to detect the presence of wolves and learn about their
movement patterns (Ciucci et al. 2003, VanEtten et al. 2007). It is a favorable technique, since it
is easily applicable and produces a high amount of information at low costs (Alexander et al.
2005, chapter 2). Not only the presence of wolves can be proved, but moreover the number of
wolves and packs estimated, as well as their movement patterns and marking behavior studied.
Further, during snow tracking, urine and fecal samples for additional analyses (such as diet,
hormone and genetic analyses) can be collected (VanEtten et al. 2007). Moreover, in the Polish
study area radio-telemetry was conducted, as three wolves were fitted with radio-collars. In the
Portuguese study area, due to the lack of snow in winter and funding for radio-tracking wolves,
the only systematic possibility to detect the presence of wolves was to conduct scat surveys.
Advantages of this method are the applicability in all seasons and over large areas, low costs,
and its non-invasive character. This difference in methods used, however, raises the question
whether the results of the different studies are comparable. This is a common problem and
several studies have focused on solving this trouble (Mayle 2000, Harrison et al. 2002, Silveira
et al. 2003, Gurnell et al. 2004, Sadlier et al. 2204, VanEtten et al. 2007). However, the detection
probability of the diverse monitoring methods differs between species (Gompper et al. 2006) and
therefore comparative studies should be conducted for each species separately. For wolves, such
studies are scarce (Alexander et al. 2005), making it necessary to first test the comparability of
the monitoring techniques used. This study showed that the detection probability for wolves by
General discussion
86
the three methods was similar; indicating that the results, obtained in the diverse study regions,
are comparable. This coincides with findings of Alexander et al. (2005), who proved that snowtracking data and data obtained by telemetry could both be used to study wolf-environment
relationships with identical outcomes. The methods used in this study, however, differed in their
time- and cost-efficiency, as well as in the information content they revealed. Scat surveys were
the least time- and cost-consuming; on the other hand they provide the smallest amount of
additional information.
Factors influencing habitat selection
The formerly wide distribution of the wolf throughout the whole northern hemisphere already
showed the high adaptability of wolves. They seem to be able to live in all kinds of geographic
and physiographic regions, indicating that they are not habitat specific. Nevertheless, they got
extinct in most parts of their former range, showing that humans have a strongly negative impact
on them. This impact is, however, not straightforward and easy to explain, as wolves survived in
some regions with a high human population density (Theuerkauf et al. 2007), but got extinct in
more remote areas (Zimen & Boitani 1979). Therefore, a combination of various factors is likely
to contribute either to the extinction or the survival of wolves. One method to identify these
factors, as well as the interaction of these factors, is ecological modeling. During the process of
ecological modeling, ecological and environmental variables (biotic and abiotic) are combined
with occurrence data of the respective species (Anderson et al. 2003). The outcome is an
equation that predicts the probability of the species occurring under given conditions, in other
words that measures the suitability of a given habitat for that respective species. With the herein
presented study on two wolf populations in Portugal and Poland, the assumption that wolves are
habitat generalists (Fuller 1995, Mech 1995), could be supported. In the Portuguese study area
wolves preferred open habitats, whereas in the Polish study area they selected forested locations.
This difference can be easily explained by corresponding differences in the wolves’ diet. In
Portugal they feed mainly on livestock (Carreira & Petrucci-Fonseca 2000), which stays in open
places for grazing, whereas in Poland they predominantly feed on wild ungulates (Gula 2008),
which reside within or close to forests. On the contrary, variables concerning prey abundance
and anthropogenic impacts were classified as important to wolves in both study areas. This
shows the more general significance of these factors for the survival of wolves, regardless of the
geographic region they were studied in. The habitat utilization model build with data of the
General discussion
87
Portuguese wolf population identified the variables “road density” and “livestock density” as
most important for explaining wolf presence. A combination of these two variables was as well
detected as vital for the survival of wolves in a few other studies (Mladenoff & Sickley 1998,
Cayuela et al. 2004). In the majority of other studies, however, either only one of the two
variables in combination with others was selected (Mladenoff et al. 1995, Treves et al. 2004,
Jędrzejewski et al. 2005, Rodriguez-Freira & Crecente-Maseda 2007) or, indeed variables from
the categories “prey abundance” and “anthropogenic impact” were selected, but not the same as
the ones selected in the herein presented model (Glenz et al. 2001). In still other studies, land
cover types were identified as important to wolves; however, in those studies no data on “prey
abundance” were included (Jędrzejewski et al. 2004, 2005, Rodriguez-Freira & Crecente-Maseda
2007). For this reason, an indicator of “prey abundance” (e.g. forest) was selected during
modeling. During validation of the study on an independent dataset from a Polish wolf
population, it became apparent that “prey abundance” is a quite complicated variable and
difficult to measure. Wolves occurred in areas with significantly lower livestock densities than in
the Portuguese study area. On the other hand, the human dimension was as well much smaller in
the Polish study region, with an average road density in wolf absence areas similar to the average
road density in wolf presence areas in Portugal. The relative distribution of suitable and
unsuitable habitat was, however, correctly predicted by the model, though absolute probabilities
for the presence of wolves were underestimated. Consequently, the findings suggest that a
combination of human impact and food abundance determines the suitability of a habitat for
wolves.
Human impact on stress hormones
A new method to non-invasively measure human influences on animals prior to visible impacts,
such as population decline, is a stress hormone analysis. An animal reacts to a stressor with the
secretion of glucocorticoids via the hypothalamic-pituitary-adrenocortical axis (Möstl & Palme
2002, Boonstra 2005). These glucocorticoids were formerly measured in the blood of the
respective individual. Since a few years, though, the employment of non-invasive methods to
measure stress hormones is increasing and gaining importance. Instead of a blood sample, saliva,
urine or feces are used. The advantage is that animals no longer need to be captured for taking a
sample and, consequently, the method can be applied to cryptic and endangered species as well
(Wasser et al. 2000, Khan et al. 2002, Palme 2005). Moreover, the stress during handling of the
General discussion
88
individual no longer distorts the results (Kotrschal et al. 1998) and, instead of showing shortterm fluctuations, more accurate assessments of chronic stress are obtained (Harper & Austad
2000). Using this tool, in several studies the negative impact of human disturbance on animals
could be shown (Pereira et al. 2006, Arlettaz et al. 2007, Barja et al. 2007, Martinez-Mota et al.
2007, Thiel et al. 2008). Several authors, however, as well reported seasonal differences in stress
hormone concentrations, with elevated stress levels during mating season (reviewed in
Millspaugh & Washburn 2004). Therefore, it is advisable to first test for these confounding
variations before interpreting the influence of human presence on the animal’s stress hormone
production (Millspaugh et al. 2001, Pereira et al. 2006). In this study, though, no significant
seasonal differences in the stress hormone level of the studied wolves were found that might
distort other correlations. Moreover, no influence of weather conditions, reported by other
researchers (Washburn & Millspaugh 2002, Pereira et al. 2006), could be detected. Therefore, I
assume that possible differences in the amount of glucocorticoid metabolites measured in scats,
found in regions with dissimilar human presence, are probably due to anthropogenic influences.
Indeed, an increase in stress hormones could be shown for regions with a higher percentage of
urbanized areas, in contrast to areas with more forest cover. These findings are in concordance
with findings of Creel et al. (2002), who measured increased stress hormone levels in samples of
wolves living in a region intensively used by snowmobile drivers. Consequently, a negative
impact of human disturbance to wolves could be detected by stress hormone analyses in this
study, where other methods were not able to reveal such a negative influence (Theuerkauf et al.
2007).
Roads as barriers to dispersal
Still another method, which only recently was started to be used for monitoring wildlife
populations, is a molecular biological one. Individuals of a species can be identified noninvasively by microsatellite analysis of DNA, extracted from hairs, feathers, urine or feces. The
potential of this new tool is immense, as such important aspects to wildlife conservation as
effective population size, population dynamics and genetic depression can be estimated.
Moreover, aspects interesting for the study of social species as the wolf, like pack dynamics,
inter-pack connectivity and dispersal rates, can be studied (Lucchini et al. 2002). The advantage
of a genetic analysis, compared to traditional field census techniques, is the high precision of
identifying individuals and therefore getting a reliable estimate of population size, pack size and
General discussion
89
numbers of packs in a given area. Further, because of analyzing the remains of an animal, which
are easy to collect without disturbing the animal itself, sample size can be significantly
increased. By means of genetic analyses four wolf packs were distinguish in the study area and
the level of relatedness among them calculated. They were rather closely related with each other,
which might point to a low exchange with other populations, resulting in a high degree of
inbreeding. However, a fairly high expected heterozygosity compared to bottlenecked
populations in Italy (Lucchini et al. 2002, Randi & Lucchini 2002, Fabbri et al. 2007) disproved
this assumption. A comparison of expected and observed heterozygosity further revealed a slight
heterozygosity excess, which already was found in a wolf population in north-eastern Poland,
with a high human caused mortality and subsequent high immigration rates of wolves from the
east (Jędrzejewski et. al 2005). As human-caused mortality in this study area is elevated as well,
immigrations from the stable Spanish wolf population are probable. These results demonstrate
the negative impact of humans on the studied wolf population, which however currently can be
compensated by immigrations from a source population in Spain. A barrier effect of roads,
though, could be disproved by means of the F-statistics, as well as by detecting identical
genotypes on both sides of major roads. This finding agrees with findings of Blanco et al. (2005)
that highways did not act as barriers to wolf dispersal, but only delayed wolf expansion for some
15 years.
8.2
Germany as a wolf country
Suitable habitat
Habitat utilization models, developed based on presence-absence data of a species that were
collected in a region with a stable population of this species, can be used for predicting suitable
habitat in other regions where the species presently is absent (Anderson et al. 2003, Naves et al.
2003). Based on this knowledge, in suitable habitats either a natural resettlement can be
supported or an artificial resettlement carried out. The habitat prediction study of Germany
identified around one-third of the country as suitable habitat for wolves, though only 4% were
identified as optimal habitat. Five patches of optimal habitat exist in the north, west and northeast of Germany. These patches could hold between 340 and 790 wolves. The good habitat was
not spread over the whole country in small patches, but rather it was located in bigger zones in
the north-east and south-west. Thus, the application of the habitat utilization model, developed in
Portugal and validated in Poland, rejected the common assumption that Germany is not a proper
General discussion
90
country for wolves and showed that an industrialized state as Germany indeed can hold a
population of wolves.
Connectivity of suitable habitat patches
Important for the conservation of wolves is, however, not only the question if suitable habitat is
existent, but as well if it is accessible for dispersing wolves for initial settlement (Kramer-Schadt
et al. 2004, Malo et al. 2004) and if it is interconnected with other populated habitats to assure
the genetic exchange between the diverse wolf populations. A preliminary coarse analysis
showed that the north-eastern suitable habitat patches are within the range of a dispersing wolf
(Kojola et al. 2006, Rodriguez-Freira & Crecente-Maseda 2008), and thus theoretically
accessible, yet they are separated from the present wolf region in Saxony by a few motorways
and several regional highways. This decreases the probability that dispersing wolves can reach
them without road casualties. On the contrary, the suitable habitats in western Germany are
rather unlikely to be settled by wolves in the near future without extensive conservation
measures.
Potential population size
A first rough estimation of a potential wolf population size in Germany resulted in unlikely high
estimates of 3000 wolves. When habitat patches, too small to hold at least 2-3 wolf packs, the
minimum number of packs necessary to support a viable population as suggested by Fuller
(1995), or too isolated to provide the required genetic exchange between subpopulations, were
subtracted, a more reliable estimate of 1500 wolves resulted. This number agreed with the
estimated potential wolf population size for Poland, calculated by Jędrzejewski et al. (2008).
Still, this value is fairly high and only achievable with conservation measures for protecting,
restoring and connecting suitable habitat.
Areas with elevated risks for human-wolf conflicts
Of great importance to wildlife conservation management, particularly for a conflictive species
as the wolf, is the minimization of human-animal conflicts (Boitani & Ciucci 1993, Chapron et
al. 2003). These conflicts arise due to a competition between humans and animals for resources.
Wolves and humans compete for space and food, and in regions where people diminished wild
ungulate species to low numbers, wolves adapt their behavior and focus on livestock as their
General discussion
91
major food resource (Bjerke et al. 1998, Linnell et al. 2002, Fritts et al. 2003, Kojola et al. 2004).
For conservation management it is therefore vital to know areas with a high risk of human-wolf
conflicts in advance. By applying the previously developed habitat utilization model, such areas,
characterized by low road densities and high livestock densities, could be identified.
8.3
Conclusion
The human dimension played a critical role in the habitat selection of wolves in both study areas
in Portugal and Poland (chapter 3 and chapter 4). However, still more important was the
availability of food and wolves could survive, by adapting to the human presence through
avoiding roads and settlements, if prey density was high. Fragmentation of their habitat by roads
did not generate isolated subpopulations and no genetic depression due to inbreeding could be
detected (chapter 6). Nevertheless, even in the study region with lower human disturbance in
Poland, wolves experienced elevated stress hormone levels in areas with lower forest cover and
higher percentages of urbanization (chapter 5). Therefore, I assume that wolves can coexist with
humans, if other characteristics of the habitat are favorable (e.g. high prey density and retreat
areas). Nevertheless, human presence does stress the animals, which in the long term might have
negative impacts on their physical condition and reproduction and thus affect population
viability. The findings of this study stress the importance of collecting data for answering a
research question on several independent animal populations and by means of diverse methods,
as this additional information might influence the interpretation of results initially obtained by
studying only one population with one method.
Summary
92
SUMMARY
For the conservation of wolves it is essential to know how they can cope with anthropogenic
changes of their habitat. As the impact of humans on wolves is not straightforward and easy to
assess, I used a wide array of methods to approach the topic from various different perspectives.
Moreover, I studied two distinct wolf populations living in diverse environments to detect
transferability and limitations of my conclusions.
In chapter 2 I approached the question of comparability of diverse monitoring methods to
assess the presence of wolves. Therefore, telemetry on three radio-collared wolves, snowtracking during winter and scat surveys during the rest of the year were conducted.
Subsequently, the distribution of areas that varied in wolf abundance, detected by the three
monitoring methods, was compared. All three methods revealed similar results on wolf presence.
Scat surveys were the most time-efficient, though also provided the lowest information content.
Snow-tracking was similar time-efficient, however only restricted to cold climates and the winter
season. Telemetry can’t be recommended for monitoring wolf populations on a large scale due to
its high costs in terms of funding and time; however, to gain information on activity and habitat
use on a small scale it is the preferred technique.
In chapter 3, then, I addressed the topic of anthropogenic influences on habitat selection by
wolves. After characterizing wolf habitats in terms of land cover, prey density and anthropogenic
impacts, they were compared to areas with wolf absence. Wolf habitats in Portugal were
characterized by significantly higher livestock densities, lower human presence and more open
areas. A habitat utilization model developed by logistic regression identified livestock density
and road density as the major factors influencing wolf presence. The negative influence of
anthropogenic infrastructure on wolves could be, moreover, demonstrated by an avoidance
behavior of wolves towards bigger roads and settlements.
As wolves seem to react dissimilar to the presence of humans in different environments,
indicated by diverse studies on habitat selection by wolves, it is advisable to test the
transferability of initial conclusions to distinct wolf populations. Therefore, in chapter 4 I
repeated the analyses of chapter 3 on data collected in a second study area, in Poland. Areas with
wolf presence were characterized by higher livestock and lower wild ungulate densities, lower
human presence and more forested areas. These findings show that the most constant factor for
indicating wolf presence is the anthropogenic influence. Distribution of land cover types in wolf
Summary
93
habitats of the two study areas was reverse; however, it mirrored differences in the prey
composition of the two wolf populations. The previously developed habitat utilization model
predicted the distribution of suitable and unsuitable habitat correctly, though it underestimated
the overall suitability of the study area.
To test for more subtle negative impacts of humans on wolves that did not yet result in a
population decline, but might do so in the near future, in chapter 5 I conducted stress hormone
analyses. The concentration of stress hormones was measured in fecal samples of wolves living
in unequally humanized areas and compared to each other. This analysis revealed that a higher
degree of urbanization within a wolf pack home range leads to an increase in stress hormone
concentrations. Influences of other factors on stress hormone levels, such as mating season or
weather conditions, could be ruled out. However, feces deposited on crossroads, thus showing a
potential marking function, exhibited higher concentrations of stress hormones.
Another potential reason for the decline of wildlife populations is a genetic depression in
isolated subpopulations due to anthropogenic barriers such as roads. Therefore, in chapter 6 I
used genetic analyses as another method to detect subtle negative impacts of humans on wolves
prior to severe consequences. By microsatellite analysis 23 distinct genotypes were identified.
Six of them were detected on two sides of major roads crossing the study area. This first hind for
a semi-permeability of roads to wolves could be confirmed by an inbreeding coefficient close to
zero and heterozygosity excess. FST values indicated that subdivision of populations by major
roads accounted for 2.6-5.6% of the total genetic variation. Moreover, values of expected
heterozygosity were high compared to bottlenecked populations.
In chapter 7, finally, I applied the habitat utilization model, developed in chapter 3 and
validated in chapter 4, to Germany for detecting potential suitable wolf habitats. The model
predicted two bigger areas of optimal wolf habitat in the north-east and west of Germany, plus
some smaller patches in the north, east and west. Good habitat stretched over the whole northeast and south-west of Germany, connected only by a narrow corridor. Areas of elevated risk for
human-wolf conflicts, characterized by low road density and high livestock density, are situated
in the north, center and south of Germany. Overall, one third of the country represents good
habitat for wolves, but only 4% were classified as optimal habitat. A potential wolf population
size for Germany was computed, with a maximum of 1500 wolves.
The findings of this study suggest that anthropogenic factors influence the habitat selection
behavior of wolves. Nevertheless, with adequate prey densities wolves can settle as well in areas
Summary
94
highly fragmented by human infrastructure. However, for the conservation of wolves it is
important to minimize the barrier effect of roads and render them permeable for wolves, to avoid
the formation of genetically isolated subpopulations. Moreover, more subtle impacts of human
presence on wolves need to be considered, as for example chronically elevated stress levels can
decrease the viability of a population in the long term. The habitat prediction study showed that
Germany possesses suitable habitats for wolves and can be naturally resettled, if proper
management and conservation measures are conducted.
Zusammenfassung
95
ZUSAMMENFASSUNG
Für den Schutz von Wölfen ist es notwendig zu wissen wie sie mit anthropogenen
Veränderungen ihrer Umwelt zu Recht kommen. Da die Beeinflussung des Wolfes durch den
Menschen nicht eindeutig und leicht zu messen ist, benutzte ich vielfältige Methoden, um das
Thema von verschiedensten Seiten zu bearbeiten. Zudem untersuchte ich zwei unterschiedliche
Wolfspopulationen, welche unter sehr verschiedenen Umweltbedingungen leben, um die
Übertragbarkeit und Limitationen meiner Schlussfolgerungen zu testen.
In Kapitel 2 bearbeitete ich die Fragestellung, ob diverse Methoden zur Erhebung von
Wolfsvorkommen gleiche Ergebnisse liefern. Hierzu wurden drei mit Sendehalsbändern
ausgestattete Wölfe telemetrisch überwacht, während des Winters wurde Schneekartieren
durchgeführt, sowie während des restlichen Jahres systematisch nach Wolfskot gesucht.
Anschließend wurde die Verteilung von Gebieten mit unterschiedlichem Wolfsvorkommen,
basierend auf Daten, welche an Hand der zuvor genannten Erhebungsmethoden gesammelt
wurden, verglichen. Alle drei Methoden lieferten ähnliche Ergebnisse bezüglich des
Wolfsvorkommens. Die systematische Suche nach Wolfskot war zeitlich gesehen die effektivste
Methode, jedoch lieferte sie auch die geringsten Informationen. Das Schneekartieren war ähnlich
zeiteffektiv, kann jedoch nur in kalten Gegenden und im Winter eingesetzt werden. Die
Telemetrie ist nicht empfehlenswert zur großflächigen Erhebung von Wolfsvorkommen, da sie
sehr kostspielig und zeitaufwendig ist. Um jedoch Informationen über Aktivität und
Habitatnutzung in kleineren Untersuchungsgebieten zu erhalten, ist sie die bevorzugte Methode.
Im 3. Kapitel befasste ich mich mit dem Aspekt des anthropogenen Einflusses auf die
Habitatwahl von Wölfen. Nach einer Charakterisierung der Wolfshabitate bezüglich
Landnutzung, Beutedichte und anthropogenen Einflüssen, wurden sie mit wolfsfreien Gebieten
verglichen. Wolfshabitate in Portugal wiesen stark erhöhte Dichten an Vieh, geringere Präsenz
von Menschen, sowie weitläufige offene Landschaften auf. Ein Habitatnutzungsmodell, welches
mit Hilfe logistischer Regression erstellt wurde, identifizierte Viehdichte und Straßendichte als
die Hauptfaktoren bei der Beeinflussung von Wolfsvorkommen. Der negative Einfluss
anthropogener Infrastrukturen auf Wölfe wurde des Weiteren deutlich durch von Wölfen
gezeigtes Meidungsverhalten gegenüber großen Straßen und Siedlungen.
Da Wölfe unter verschiedenen Umweltbedingungen unterschiedlich auf die Anwesenheit
von Menschen zu reagieren scheinen, was durch mannigfaltige Studien zur Habitatwahl von
96
Zusammenfassung
Wölfen festgestellt wurde, ist es ratsam, die Übertragbarkeit der zuvor gezogenen
Schlussfolgerungen auf andere Wolfspopulationen zu testen. Aus diesem Grund wiederholte ich
in Kapitel 4 die zuvor in Kapitel 3 durchgeführten Analysen mit Daten, welche in einem zweiten
Untersuchungsgebiet in Polen erhoben wurden. Wolfsgebiete wiesen höhere Viehdichten,
niedrigere Schalenwilddichten, geringere anthropogene Einflüsse und eine stärkere Bewaldung
auf. Diese Ergebnisse zeigen, dass der beständigste Faktor für die Vorhersage von
Wolfsvorkommen
die
anthropogenen
Einflüsse
sind.
Die
Verteilung
verschiedener
Landnutzungstypen innerhalb der Wolfshabitate war in den zwei Untersuchungsgebieten
gegenläufig. Sie spiegelte jedoch die Unterschiede in der Beutezusammensetzung der zwei
Wolfspopulationen wider. Das zuvor erstellte Habitatnutzungsmodell sagte die Verteilung
geeigneter und ungeeigneter Habitate voraus, insgesamt unterschätzte es jedoch die Eignung des
Untersuchungsgebietes für Wölfe.
Um auch weniger offensichtliche, negative Einflüsse des Menschen auf den Wolf zu
erkennen, welche zwar bisher noch keinen Populationsrückgang hervorgerufen haben, dies aber
in naher Zukunft tun könnten, führte ich in Kapitel 5 Stresshormonanalysen durch. Die
Konzentration an Stresshormonen wurde in Kotproben von Wölfen, welche in unterschiedlich
stark anthropogen genutzten Gebieten leben, gemessen und miteinander verglichen. Die Analyse
zeigte, dass ein höherer Anteil urbaner Gebiete innerhalb des Streifgebietes eines Wolfsrudels zu
einer erhöhten Stresshormonkonzentration führt. Einflüsse anderer Faktoren auf die
Stresshormonkonzentration, wie zum Beispiel Paarungszeit oder Wetterbedingungen, konnten
ausgeschlossen werden. Kot, welcher an Wegkreuzungen abgesetzt wurde und somit eine
Markierungsfunktion vermuten lässt, wies jedoch höhere Konzentrationen an Stresshormonen
auf.
Genetische Verarmung bei isolierten Teilpopulationen, herbeigeführt durch anthropogene
Barrieren, wie zum Beispiel Straßen, ist ein weiterer potentieller Grund für einen
Populationsrückgang. Folglich nutzte ich in Kapitel 6 genetische Analysen als eine weitere
Methode zum Erkennen unterschwelliger, negativer, menschlicher Einflüsse auf Wölfe, bereits
bevor
diese
weitreichende
Auswirkungen
hervorrufen
können.
Durch
eine
Mikrosatellitenanalyse konnten 23 individuelle Genotypen bestimmt werden. Von diesen wurden
sechs Genotypen beidseitig großer Strassen, welche das Untersuchungsgebiet kreuzen, entdeckt.
Dieses erste Indiz, welches für die Semipermeabilität der Strassen für Wölfe spricht, konnte
bestätigt werden durch einen Inzuchtkoeffizienten nahe null und einen Überschuss an
Zusammenfassung
97
Heterozygoten in der Population. Die FST -Werte zeigten an, dass die Trennung der Population
durch Hauptstraßen 2.6-5.6% der gesamten genetischen Variation ausmacht. Des Weiteren lagen
die erwarteten Heterozygotiewerte höher als in Populationen, welche in naher Vergangenheit
einem genetischen Flaschenhals unterlagen.
Im 7. Kapitel wendete ich schließlich das zuvor in Kapitel 3 erstellte und in Kapitel 4
validierte Habitatnutzungsmodell auf den Raum Deutschland an, um potentiell geeignete
Wolfshabitate zu identifizieren. Das Modell wies zwei großräumige Gebiete mit optimaler
Eignung für Wölfe im Nordosten und Westen Deutschlands aus, sowie einige kleinere Gebiete
im Norden, Osten und Westen des Landes. Potentiell geeignetes Habitat erstreckte sich über den
gesamten Nordosten und Südwesten Deutschlands, die beiden Gebiete wurden jedoch nur durch
einen schmalen Korridor geeigneten Habitates miteinander verbunden. Gebiete mit einem
erhöhten Risiko für Konflikte zwischen Menschen und Wölfen, welche durch eine geringe
Straßendichte und hohe Viehdichte gekennzeichnet sind, lagen im Norden, Zentrum und Süden
des Landes. Insgesamt ein Drittel des Landes zeichnet sich durch eine mögliche Eignung für
Wölfe aus, wobei jedoch nur 4% als optimales Wolfshabitat eingestuft wurden. Für ganz
Deutschland wurde eine potentielle Wolfspopulationsgröße von maximal 1500 Tieren berechnet.
Die Ergebnisse dieser Studie legen die Vermutung nahe, dass anthropogene Einflüsse die
Habitatwahl von Wölfen beeinflussen. Dennoch können Wölfe sich auch in Gebieten
niederlassen, welche stark durch anthropogene Infrastrukturen fragmentiert sind, wenn
angemessene Beutedichten vorliegen. Jedoch ist es für den Schutz der Wölfe wichtig, den
Barriereeffekt von Straßen zu minimieren und diese permeabel für Wölfe zu gestalten, um die
Bildung von genetisch isolierten Teilpopulationen zu vermeiden. Des Weiteren müssen weniger
offensichtliche, negative Einflüsse des Menschen auf Wölfe bedacht werden, da zum Beispiel
ein chronisch erhöhter Stresslevel die Überlebensfähigkeit von Populationen langfristig
verringern kann. Die Habitatstudie zeigte, dass Deutschland über geeignete Wolfshabitate
verfügt und diese, im Falle entsprechender Management- und Schutzmaßnahmen, auch auf
natürlichem Wege von Wölfen wiederbesiedelt werden können.
References
98
REFERENCES
ALEXANDER, S.M., PAQUET, P.C., LOGAN T.B. AND SAHER, D.J. 2005. Snow-tracking versus
radiotelemetry for predicting wolf-environment relationships in the Rocky Mountains of Canada. Wildlife
Society Bulletin 33: 1216-1224.
ANDERSON, R.P., LEW, D. AND PETERSON, A.T. 2003. Evaluating predictive models of species’
distributions: criteria for selecting optimal models. – Ecological Modelling 162: 211-232.
ANSORGE, H., KLUTH, G. AND HAHNE, S. 2006. Feeding ecology of wolves Canis lupus returning to
Germany. – Acta Theriologica 51: 99-106.
APOLLONIO, M., MATTIOLI, L., SCANDURA, M., MAURI, L., GAZZOLA, A. AND AVANZINELLI, E. 2004.
Wolves in the Casentinesi forests: insights for wolf conservation in Italy from a protected area with a rich
wild prey community. – Biological Conservation 120: 249-260.
ARNHEIM, N., LI, H.H. AND CUI, X.F. 1990. PCR analysis of DNA sequences in single cells: single sperm
gene mapping and genetic disease diagnosis. – Genomics 8: 415-419.
ARLETTAZ, R., PATTHEY, P., BALTIC, M., LEU, T., SCHAUB, M., PALME, R. AND JENNI-EIERMANN, S.
2007. Spreading free-riding snow sports represent a novel serious threat for wildlife. – Proceedings of the
Royal Society Britain 274: 1219-1224.
ASA,C.S., MECH, L.D. AND SEAL, U.S. 1985. The use of urine, faeces and anal secretions in scentmarking by a captive wolf (Canis lupus) pack. – Animal Behaviour 33: 1034-1036.
ASPI, J., ROININEN, E., RUOKONEN, M., KOJOLA, I. AND VILÁ, C. 2006. Genetic diversity, population
structure, effective population size and demographic history of the Finnish wolf population. – Molecular
Ecology 15: 1561-1576.
BALLARD, W.B., MCNAY, M.E., GARDNER, C.L. AND REED, D.J. 1995. Use of line-intercept track
sampling for estimating wolf density. In: Carbyn, L.N., Fritts, S.H., Seip, D.R. (eds). Ecology and
conservation of wolves in a changing world. Canadian Circumpolar Institute, Edmonton, Alberta,
Canada, pp. 469-480.
BARJA, I., DE MIGUEL, F.J. AND BARCENA, F. 2004. The importance of crossroads in faecal marking
behaviour of the wolves (Canis lupus). – Naturwissenschaften 91: 489-492.
BARJA, I., SILVÁN, G. AND ILLERA, J.C. 2008. Relationships between sex and stress hormone levels in
feces and marking behavior in a wild population of Iberian wolves (Canis lupus signatus). – Journal of
Chemical Ecology 34: 697-701.
BARJA, I., SILVAN, G., ROSELLINI, S., PINEIRO, A., GONZALEZ-GIL, A., CAMACHO, L. AND ILLERA, J.C.
2007. Stress physiological responses to tourist pressure in a wild population of European pine marten. –
Journal of Steroid Biochemistry and Molecular Biology 104: 136-142.
BESSA-GOMES, C. AND PETRUCCI-FONSECA, F. 2003. Using artificial neural networks to assess wolf
distribution patterns in Portugal. – Animal Conservation 6: 221-229.
References
99
BIDER, J.R., 1968. Animal activity in uncontrolled terrestrial communities as determined by sand transect
technique. Ecological Monographs 38: 269-308.
BISI, J., KURKI, S., SVENSBERG, M. AND LIUKKONEN, T. 2007. Human dimensions of wolf (Canis lupus)
conflicts in Finland. – European Journal of Wildlife Research 53: 304-314.
BJERKE, T., REITAN, O. AND KELLERT, R. 1998. Attitudes towards wolves in southeastern Norway.
Society & Natural Resources 11: 169-178.
BLANCO, J.C., CUESTA, L. AND REIG, S. 1992. Distribution, status, and conservation problems of the wolf
Canis lupus in Spain. – Biological Conservation 60: 73-80.
BLANCO, J.C., CORTES, Y. AND VIRGÓS, E. 2005. Wolf response to two kinds of barriers in an
agricultural habitat in Spain. – Canadian Journal of Zoology 83: 312-323.
BLOUIN, M.S., PARSONS, M., LACAILLE, V. AND LOTZ, S. 1996. Use of microsatellite loci to classify
individuals by relatedness. – Molecular Ecology 5: 393-401.
BOITANI, L. 1992. Wolf research and conservation in Italy. – Biological Conservation 61: 125-132.
BOITANI, L. 1995. Ecological and cultural diversities in the evolution of wolf-human relationships. – In:
Carbyn, L.N., Fritts, S.H., Seip, D.R. (eds.). Ecology and Conservation of Wolves in a Changing World.
Canadian Circumpolar Institute, Alberta, Occasional Publication No. 35, pp. 3-11.
BOITANI, L. 2000. Action plan for the conservation of the wolves (Canis lupus) in Europe. – Nature and
Environment, Council of Europe Publishing 113.
BOITANI, L. 2003. Wolf conservation and recovery. In: Mech, L.D., Boitani, L. (eds.). Wolves, Behavior,
Ecology, and Conservation. University of Chicago Press, Chicago, pp. 317-340.
BOITANI, L. AND CIUCCI, P. 1993. Wolves in Italy: critical issues for their conservation. - In: Promberger,
C., Schröder, W. (eds). Wolves in Europe: status and perspectives. Munich Wildlife Society, Ettal,
Germany, pp. 74-90.
BOONSTRA, R. 2005. Equipped for life: the adaptive role of the stress axis in male mammals. – Journal of
Mammalogy 86: 236-247.
BRASHARES, J.S. AND ARCESE, P. 1999. Scent marking in a territorial African antelope: I. The
maintenance of borders between male oribi. Animal Behaviour 57: 1-10.
BREITENMOSER, U. 1998. Large predators in the Alps: the fall and rise of man’s competitors. – Biological
Conservation 83: 279-289.
CARREIRA, R.S. AND PETRUCCI-FONSECA, F. 2000. Lobo na região oeste de Trás-os-Montes (Portugal). –
Galemys 12: 123-134.
CARROLL, C., PAQUET, P.C. AND NOSS, R.F. 1999. Modeling carnivore habitat in the Rocky Mountain
Region: a literature review and suggested strategy. - Toronto, WWF Canada.
CAYUELA, L. 2004. Habitat evaluation of the Iberian wolf Canis lupus in Picos de Europa National Park,
Spain. – Applied Geography 24: 199-215.
References
100
CHAPRON, G., LEGENDRE, S., FERRIÈRE, R., CLOBERT, J. AND HAIGHT, R.G. 2003. Conservation and
control strategies for the wolf (Canis lupus) in western Europe based on demographic models. – Comptes
Rendus Biologies 326: 575-587.
CIUCCI, P. AND BOITANI, L. 1998. Wolf and dog depredation on livestock in central Italy. – Wildlife
Society Bulletin 26: 504-514.
CIUCCI, P., MASI, M. AND BOITANI, L. 2003. Winter habitat and travel route selection by wolves in the
northern Apennines, Italy. Ecography 26: 223-235.
CIUCCI, P., BOITANI, L., FRANCISCI, F. AND ANDREOLI, G. 1997. Home range, activity and movements of
a wolf pack in central Italy. – Journal of Zoology 243: 803-819.
CLEVENGER, A.P. AND WALTHO, N. 2005. Performance indices to identify attributes of highway crossing
structures facilitating movement of large mammals. – Biological Conservation 125: 453-464.
CORSI, F., DUPRÈ, E. AND BOITANI, L. 1999. A large-scale model of wolf distribution in Italy for
conservation planning. – Conservation Biology 13: 150-159.
CÔTÉ, I.M. AND REYNOLDS, J.D. 2002. Predictive ecology to the rescue? – Science 298: 1181-1182.
CREEL, S. 2001. Social dominance and stress hormones. – Trends in Ecology and Evolution 16: 491-497.
CREEL, S. 2005. Dominance, aggression, and glucocorticoid levels in social carnivores. – Journal of
Mammalogy 86: 255-264.
CREEL, S., FOX, J.E., HARDY, A., SANDS, J., GARROTT, B. AND PETERSON, R.O. 2002. Snowmobile
activity and glucocorticoid stress responses in wolves and elk. – Conservation Biology 16: 809-814.
CREEL, S., SPONG, G., SANDS, J.L., ROTELLA, J., ZEIGLE, J., JOE, L., MURPHY, K.M. AND SMITH, D. 2003.
Population size estimation in Yellowstone wolves with error-prone non-invasive microsatellite genotypes.
– Molecular Ecology 12: 2003-2009.
CZEKALA, N.M. AND LASLEY, B.L. 1977. A technical note on sex determination in monomorphic birds
using fecal steroid analysis. – International Zoo Yearbook 17: 209-211.
DONNELLY, P. 1995. Nonindependence of matches at different loci in DNA profiles: quantifying the
effect of close relatives on the match probability. – Heredity 75: 26-34.
EGGERMANN, J., GULA, R., PIRGA, B., THEUERKAUF, J., TSUNODA, H., BRZEZOWSKA, B., ROUYS, S. AND
RADLER, S. 2009. Daily and seasonal variation in wolf activity in the Bieszczady Mountains, SE Poland.
Mammalian Biology 74: 159-163.
EGGERT, L.S., EGGERT, J.A. AND WOODRUFF, D.S. 2003. Estimating population sizes for elusive animals:
the forest elephants of Kakum National Park, Ghana. – Molecular Ecology 12: 1389-1402.
ERICSSON, G. AND HEBERLEIN, T.A. 2003. Attitudes of hunters, locals, and the general public in Sweden
now that the wolves are back. – Biological Conservation 111: 149-159.
References
101
ERNEST, H.B., PENEDO, M.C.T., MAY, B.P., SYVANEN, S.M. AND BOYCE, W.M. 2000. Molecular tracking
of mountain lions in the Yosemite Valley region in California: genetic analysis using microsatellites and
faecal DNA. – Molecular Ecology 9: 433-441.
FABBRI, E., MIQUEL, C., LUCCHINI, V., SANTINI, A., CANIGLIA, R., DUCHAMP, C., WEBER, J.-M.,
LEQUETTE, B., MARUCCO, F., BOITANI, L., FUMAGALLI, L., TABERLET, P. AND RANDI, E. 2007. From the
Apennines to the Alps: colonization genetics of the naturally expanding Italian wolf (Canis lupus)
population. – Molecular Ecology 16: 1661-1671.
FINDO, S. AND CHOVANCOVA, B. 2004. Home ranges of two wolf packs in the Slovak Carpathians. –
Folia Zoologica 53: 17-26.
FRANCISCO, L.V., LANGSTON, A.A., MELLERSH, C.S., NEAL, C.L. AND OSTRANDER, E.A. 1996. A class
of highly polymorphic tetranucleotide repeats for canine genetic mapping. – Mammalian Genome 7: 359362.
FRANCO, A.M.A., PALMEIRIM, J.M. AND SUTHERLAND, W.J. 2007. A method for comparing effectiveness
of research techniques in conservation and applied ecology. Biological Conservation 134: 96-105.
FREDHOLM, M. AND WINTERØ, A.K. 1995. Variation of short tandem repeats within and between species
belonging to the Canidae family. – Mammalian Genome 6: 11-18.
FRITTS, S.H. AND CARBYN, L.N. 1995. Population viability, nature reserves, and the outlook for gray wolf
conservation in North America. – Restoration Ecology 3: 26-38.
FRITTS, S., STEPHENSON, R., HAYES, R. AND BOITANI, L. 2003. Wolves and humans. – In: Mech, D.,
Boitani, L. (eds.). Wolves: behaviour, ecology, and conservation. University of Chicago Press, Chicago.
FULLER, T.K. 1989. Population dynamics of wolves in north-central Minnesota. - Wildlife Monographs
105: 41pp.
FULLER, T.K. 1995. Guidelines for gray wolf management in the Northern Great Lakes Region. –
International Wolf Center Technical Publication 271, Ely, Minnesota.
FULLER, T.K., BERG, W.E., RADDE, G.L., LENARZ, M.S. AND JOESLYN, G.B. 1992. A history and current
estimate of wolf distribution and numbers in Minnesota. – Wildlife Society Bulletin 20: 42-55.
GERLACH, G. AND MUSOLF, K. 2000. Fragmentation of landscape as a cause for genetic subdivision in
bank voles. – Conservation Biology 14: 1066-1074.
GLENZ, C., MASSOLO, A., KUONEN, D. AND SCHLAEPFER, R. 2001. A wolf habitat suitability prediction
study in Valais (Switzerland). – Landscape and Urban Planning 55: 55-65.
GOLDEN, H.N., HENRY, J.D., BECKER, E.F., GOLDSTEIN, M.I., MORTON, J.M., FROST, D., SR. AND POE,
A.J. 2007. Estimating wolverine Gulo gulo population size using quadrate sampling of tracks in snow.
Wildlife Biology 13: 52-61.
References
102
GOMPPER, M.E., KAYS, R.L., RAY, J.C., LAPOINT, S.D., BOGAN, D.A. AND CRYAN, J.R. 2006. A
comparison of non-invasive techniques to survey carnivore communities in north-eastern North America.
Wildlife Society Bulletin 34: 1142-1151.
GOTELLI, N.J. AND COLWELL, R.K. 2001. Quantifying biodiversity: procedures and pitfalls in the
measurement and comparison of species richness. – Ecology Letters 4: 379-391.
GRILO, C., MOCO, G., CANDIDO, A.T., ALEXANDRE, A.S. AND PETRUCCI-FONSECA, F. 2002. Challenges
for the recovery of the Iberian wolf in the Duoro river south region. – Revista de Biologia 20: 121-133.
GULA, R. 2004. Influence of snow cover on wolf Canis lupus predation patterns in Bieszczady
Mountains, Poland. Wildlife Biology 10: 17-23.
GULA, R. 2008. Wolves return to Poland’s Holy Cross Primeval Forest. – International Wolf Magazine
18: 17-21.
GULA, R. 2008. Wolf depredation on domestic animals in the Polish Carpathian Mountains. Journal of
Wildlife Management 72: 283-289.
GULA, R., HAUSKNECHT, R. AND KUEHN, R. 2009. Evidence of wolf dispersal in anthropogenic habitats
of the Polish Carpathian Mountains. Biodiversity and Conservation, in press.
GURNELL, J., LURZ, P.W.W., SHIRLEY, M.D.F., CARTMEL, S., GARSON, P.J., MAGRIS, L. AND STEELE, J.
2004. Monitoring red squirrels Sciurus vulgaris and grey squirrels Sciurus carolinensis in Britain.
Mammalian Review 34: 51-74.
HAIG, S. 1998. Molecular contributions to conservation. – Ecology 79: 413-425.
HARPER, J.M. AND AUSTAD, S.N. 2000. Fecal glucocorticoids: a non-invasive method of measuring
adrenal activity in wild and captive rodents. – Physiological and Biochemical Zoology 73: 12-22.
HARRISON, R.L., BARR, D.J. AND DRAGOO, J.W. 2002. A comparison of population survey techniques for
swift foxes (Vulpes velox) in New Mexico. American Midland Naturalist 148: 320-337.
HAUSKNECHT, R., KÜHN, R., PIRGA, B. AND GULA, R. 2005. Molekulargenetisches Monitoring von
Wildtierpopulationen am Beispiel von Wolfsrudeln in Polen. – Beiträge zur Jagd- und Wildforschung 30:
203-211.
HUNT, K.E. AND WASSER, S.K. 2003. Effect of long-term preservation methods on fecal glucocorticoid
concentrations of grizzly bear and African elephant. – Physiological and Biochemical Zoology 76: 918928.
HUTCHINGS, M.R., SERVICE, K.M. AND HARRIS, S. 2001. Defecation and urination patterns of badgers
Meles meles at low density in south west England. Acta Teriologica 46: 87-96.
JACOBS, J. 1974. Quantitative measurements of food selection: a modification of the forage ratio and
Ivlev’s electivity index. – Oecologia 14: 413-417.
JAEGER, J.A.G., BOWMAN, J., BRENNAN, J., FAHRIG, L., BERT, D., BOUCHARD, J., CHARBONNEAU, N.,
FRANK, K., GRUBER, B. AND TLUK VON TOSCHANOWITZ, K. 2005. Predicting when animal populations
References
103
are at risk from roads: an interactive model of roads avoidance behavior. – Ecological Modeling 185:
329-348.
JĘDRZEJEWSKI, W., JĘDRZEJEWSKA, B., OKARMA, H., SCHMIDT, K., ZUB, K. AND MUSIANI M. 2000. Prey
selection and predation by wolves in BiałowieŜa Primeval Forest, Poland. – Journal of Mammalogy 81:
197-212.
JĘDRZEJEWSKI, W., SCHMIDT, K., THEUERKAUF, J., JĘDRZEJEWSKA, B., SELVA, N., ZUB, K. AND
SZYMURA, L. 2002. Kill rates and predation by wolves on ungulate populations in BiałowieŜa Primeval
Forest (Poland). – Ecology 83: 1341-1356.
JĘDRZEJEWSKI, W., NIEDZIAŁKOWSKA, M., NOWAK, S. AND JĘDRZEJEWSKA, B. 2004. Habitat variables
associated with wolf (Canis lupus) distribution and abundance in northern Poland. – Diversity and
Distributions 10: 225-233.
JĘDRZEJEWSKI, W., SCJMIDT, K., JĘDRZEJEWSKA, B., THEUERKAUF, J., KOWALCZYK, R. AND ZUB, K.
2004. The process of a wolf pack splitting in BiałowieŜa Primeval Forest, Poland. – Acta Theriologica 49:
275-280.
JĘDRZEJEWSKI, W., NIEDZIAŁKOWSKA, M., MYSLAJEK, R.W., NOWAK, S. AND JĘDRZEJEWSKA, B. 2005.
Habitat selection by wolves Canis lupus in the uplands and mountains of southern Poland. – Acta
Theriologica 50: 417-428.
JĘDRZEJEWSKI, W., JĘDRZEJEWSKA, B., ZAWADZKA, B., BOROWIK, T., NOWAK, S. AND MYSŁAJEK, R.W.
2008. Habitat suitability model for Polish wolves based on long-term national census. – Animal
Conservation 11: 377-390.
JĘDRZEJEWSKI, W., BRANICKI, W., VEIT, C., MEDUGORAC, I., PILOT, M., BUNEVICH, A.N.,
JĘDRZEJEWSKA, B., SCHMIDT, K., THEUERKAUF, J., OKARMA, H., GULA, R., SZYMURA, L. AND FÖRSTER,
M. 2005. Genetic diversity and relatedness within packs in an intensely hunted population of wolves
Canis lupus. – Acta Theriologica 50: 3-22.
JHALA, Y.V. AND SHARMA, D.K. 1997. Child-lifting by wolves in eastern Uttar Pradesh, India. – Journal
of Wildlife Research 2: 94-101.
JOHANSSON, A. AND LIBERG, O. 1996. Functional aspects of marking behavior by male roe deer
(Capreolus capreolus). Journal of Mammalogy 77: 558-567.
KAARTINEN, S., KOJOLA, I. AND COLPAERT, A. 2005. Finnish wolves avoid roads and settlements. –
Annales Zoologici Fennici 42: 523-532.
KACZENSKY, P. 2006. Akzeptanzstudie für Wölfe in Deutschland. In: Kaczensky, P. (ed.).
Medienpräsenz- und Akzeptanzstudie “Wölfe in Deutschland”. Universität Freiburg, Germany, pp. 9-88
KHAN, M.Z., ALTMANN, J., ISANI, S.S. AND YU, J. 2002. A matter of time: evaluating the storage of fecal
samples for steroid analysis. – General and Comparative Endocrinology 128: 57-64.
References
104
KLAR, N., FERNÁMDEZ, N., KRAMER-SCHADT, S., HERRMANN, M., TRINZEN, M., BÜTTNER, I. AND
NIEMITZ, C. 2008. Habitat selection models for European wildcat conservation. – Biological Conservation
141: 308-319.
KOHN, M.H. AND WAYNE, R.K. 1997. Facts from feces revisited. – Trends in Ecology and Evolution 12:
223-227.
KOHN, M.H., KNAUER, F., STOFFELA, A., SCHRÖDER, W. AND PÄÄBO, S. 1995. Conservation genetics of
the European brown bear – a study using excremental PCR of nuclear and mitochondrial markers. –
Molecular Ecology 4: 95-103.
KOHN, M.H., YORK, E.C., KAMRADT, D.A., HAUGHT, G., SAUVAJOT, R.M. AND WAYNE, R.K. 1999.
Estimating population size by genotyping faeces. Proceedings of the Royal Society of London B 266:
657-663.
KOJOLA, I., HUITU, O., TOPPINEN, K., HEIKURA, K., HEIKKINEN, S. AND RONKAINEN, S. 2004. Predation
on European wild forest reindeer (Rangifer tarandus fennicus) by wolves (Canis lupus) in Finland. –
Journal of Zoology 263: 229-235.
KOJOLA, I., ASPI, J., HAKALA, A., HEIKKINEN, S., ILMONI, C. AND RONKAINEN, S. 2006. Dispersal in an
expanding wolf population in Finland. – Journal of Mammalogy 87: 281-286.
KORTE, S.M., KOOLHAAS, J.M., WINGFIELD, J.C. AND MCEWAN, B.S. 2005. The Darwinian concept of
stress: benefits of allostasis and costs of allostatic load and the trade-off in health and disease. –
Neuroscience and Behavioral Physiology 29: 3-38.
KOTRSCHAL, K., HIRSCHENHAUSER, K. AND MÖSTL, E. 1998. The relationship between social stress and
dominance is seasonal in greylag geese. – Animal Behaviour 55: 171-176.
KRAMER-SCHADT, S., REVILLA, E., WIEGAND, T. AND BREITENMOSER, U. 2004. Fragmented landscapes,
road mortality and patch connectivity: modelling influences on the dispersal of Eurasian lynx. – Journal
of Applied Ecology 41: 711-723.
KRAMER-SCHADT, S., REVILLA, E. AND WIEGAND, T. 2005. Lynx reintroductions in fragmented
landscapes of Germany: projects with a future or misunderstood wildlife conservation? – Biological
Conservation 125: 169-182.
KUSAK, J., SKRBINŠEK, A.M. AND HUBER, D. 2005. Home ranges, movements, and activity of wolves
(Canis lupus) in the Dalmatian part of Dinarids, Croatia. – European Journal of Wildlife Research 51:
254-262.
LEHMAN, N., CLARKSON, P., MECH, L.D., MEIER, T.J. AND WAYNE, R.K. 1992. A study of the genetic
relationships within and among wolf packs using DNA fingerprinting and mitochondrial DNA. –
Behavioural Ecology and Sociobiology 30: 83-94.
References
105
LINNELL, J., SWENSON, J. AND ANDERSEN, R. 2001. Predators and people: conservation of large
carnivores is possible at high densities if management policy is favourable. – Animal Conservation 4:
345-349.
LINNELL, J.D.C., SMITH, M.E., ODDEN, J., KACZENSKY, P. AND SWENSON, J.E. 1996. Strategies for the
reduction of carnivore-livestock conflicts: a review. – NINA Oppdragsmelding 443: 1-118.
LINNELL, J.D.C., ANDERSEN, R., ANDERSONE, Z., BALCIAUSKAS, L., BLANCO, J.C., BOITANI, L.,
BRAINERD, S., BREITENMOSER, U., KOJOLA, I., LIBERG, O., LÖE, J., OKARMA, H., PEDERSEN, H.C.,
PROMBERGER, C., SAND, H., SOLBERG, E.J., VALDMANN, H. AND WABAKKEN, P. 2002. The fear of
wolves: A review of wolf attacks on humans. - NINA Oppdragsmelding 731: 1-65.
LITVAITIS, J.A., SHERBURNE, J.A. AND BISSONETTE, J.A. 1985. A comparison of methods used to
examine snowshoe hare habitat use. Journal of Wildlife Management 49: 693-695.
LOVARI, S., SFORZI, A., SCALA, C. AND FICO, R. 2007. Mortality parameters of the wolf in Italy: does the
wolf keep himself from the door? – Journal of Zoology 272: 117-124.
LUCCHINI, V., FABBRI, E., MARUCCO, F., RICCI, S., BOITANI, L. AND RANDI, E. 2002. Noninvasive
molecular tracking of colonizing wolf (Canis lupus) packs in the western Italian Alps. – Molecular
Ecology 11: 857-868.
MACARTHUR, R.A., JOHNSTON, R.H. AND GEIST, V. 1979. Factors influencing heart rate in free-ranging
bighorn sheep: a physiological approach to the study of wildlife harassment. – Canadian Journal of
Zoology 57: 2010-2021.
MACARTHUR, R.A., GEIST, V. AND JOHNSTON, R.H. 1982. Cardiac and behavioral responses of mountain
sheep to human disturbance. – Journal of Wildlife Management 46: 351-358.
MACDONALD, D.W. 1980. Patterns of scent marking with urine and faeces amongst carnivore
communities. – Symposia of the Zoological Society of London 45: 107-139.
MAHON, P.S., BANKS, P.B. AND DICKMAN, C.R. 1998. Population indices for wild carnivores: a critical
study in sand-dune habitat, south-western Queensland. Wildlife Research 25: 11-22.
MALO, J.E., SUÁREZ, F. AND DÍEZ, A. 2004. Can we mitigate animal-vehicle accidents using predictive
models? – Journal of Applied Ecology 41: 701-710.
MARTINEZ-MOTA, R., VALDESPINO, C., SANCHEZ-RAMOS, M.A. AND SERIO-SILVA, J.C. 2007. Effects of
forest fragmentation on the physiological stress response of black howler monkeys. – Animal
Conservation 10: 374-379.
MASSOLO, A. AND MERIGGI, A. 1998. Factors affecting habitat occupancy by wolves in northern
Apennines (northern Italy): a model of habitat suitability. – Ecography 21: 97-107.
MAYLE, B.A., PUTMAN, R.J. AND WYLLIE, I. 2000. The use of trackway counts to establish an index of
deer presence. – Mammal Review 30: 233-237.
References
106
MCEWAN, B.S. AND SAPOLSKY, R.M. 1995. Stress and cognitive function. – Current Opinion in
Neurobiology 5: 205-216.
MCLEOD, P.J., MOGER, W.H., RYON, J., GADBOIS, S. AND FENTRESS, J.C. 1996. The relation between
urinary cortisol levels and social behavior in captive timber wolves. – Canadian Journal of Zoology 74:
209-216.
MECH, L.D. 1995. The challenge and opportunity of recovering wolf populations. - Conservation Biology
9: 270-278.
MECH, L.D. 1999. Alpha status, dominance, and division of labor in wolf packs. – Canadian Journal of
Zoology 77: 1196-1203.
MECH, L.D., FRITTS, S.H., RADDE, G.L. AND PAUL, W.J. 1988. Wolf distribution and road density in
Minnesota. – Wildlife Society Bulletin 16: 85-87.
MECH, L.D., HARPER, E.K., MEIER, T.J. AND PAUL, W.J. 2000. Assessing factors that may predispose
Minnesota farms to wolf depredation on cattle. – Wildlife Society Bulletin 28: 623-629.
MEIER, T.J., BURCH, J.W., MECH, L.D. AND ADAMS, L.G. 1995. Pack structure and genetic relatedness
among wolf packs in a naturally-regulated population. In: Carbyn, L.N., Fritts, S.H., Seip, D.R. (eds.).
Ecology and Conservation of wolves in a changing world. Canadian Circumpolar Institute, University of
Alberta, Edmonton, pp. 293-302.
MILLS, L.S., CITTA, J.J., LAIR, K.P., SCHWARTZ, M.K. AND TALLMON, D.A. 2000. Estimating animal
abundance using non-invasive DNA sampling: promise and pitfalls. – Ecological Applications 10: 283294.
MILLSPAUGH, J.J. AND WASHBURN, B.E. 2004. Use of fecal glucocorticoid metabolite measures in
conservation biology research: considerations for application and interpretation. – General and
Comparative Endocrinology 138: 189-199.
MILLSPAUGH, J.J., WOODS, R.J., HUNT, K.E., RAEDEKE, K.J., BRUNDIGE, G.C., WASHBURN, B.E. AND
WASSER, S.K. 2001. Fecal glucocorticoid assays and the physiological stress response of elk. – Wildlife
Society Bulletin 29: 899-907.
MILLSPAUGH, J.J., WASHBURN, B.E., MILANICK, M.A., BERINGER, J., HANSEN, L.P. AND MEYER, T.M.
2002. Non-invasive techniques for stress assessment in white-tailed deer. – Wildlife Society Bulletin 30:
899-907.
MLADENOFF, D.J. AND SICKLEY, T.A. 1998. Assessing potential gray wolf restoration in the northeastern
United States: a spatial prediction of favorable habitat and potential population levels. – Journal of
Wildlife Management 62: 1-10.
MLADENOFF, D.J., SICKLEY, T.A. AND WYDEVEN, A.P. 1999. Predicting gray wolf landscape
recolonization: logistic regression models vs. new field data. - Ecological Applications 9: 37-44.
References
107
MLADENOFF, D.J., SICKLEY, T.A., HAIGHT, R.G. AND WYDEVEN, A.P. 1995. A regional landscape
analysis and prediction of favorable Gray wolf habitat in the northern great lakes region. – Conservation
Biology 9: 279-294.
MOEN, A.N., WHITTEMORE, S. AND BUXTON, B. 1982. Effects of disturbance by snowmobiles on heart
rate of captive white-tailed deer. New York Fish and Game Journal 29: 176-183.
MONTEIRO, L., BONNEMAISO, D., VEKRIS, A., PETRY, K.G., BONNET, J., VIDAL, R., CABRITA, J. AND
MÉGRAUD, F. 1997. Complex polysaccharides as PCR inhibitors in faeces: Helicobacter pylori Model. –
Journal of Clinical Microbiology 35: 995-998.
MOORING, M.S., PATTON, M.L., LANCE, V.A., HALL, B.M., SCHAAD, E.W., FETTER, G.A., FORTIN, S.S.
AND
MCPEAK, K.M. 2006. Glucocorticoids of bison bulls in relation to social status. – Hormones and
Behavior 49: 369-375.
MORIN, P.A., CHAMBERS, K.E., BOESCH, C. AND VIGILANT, L. 2001. Quantitative polymerase chain
reaction analysis of DNA from non-invasive samples for accurate microsatellite genotyping of wild
chimpanzees (Pan troglodytes verus). – Molecular Ecology 10: 1835-1844.
MÖSTL, E. AND PALME, R. 2002. Hormones as indicators of stress. – Domestic Animal Endocrinology 23:
67-74.
MÖSTL, E., NÖBAUER, H., CHOI, H.S., WURM, W. AND BAMBERG, E. 1983. Trächtigkeitsdiagnose bei der
Stute mittels Östrogenbestimmung im Kot. – Prakt. Tierarzt 64: 491-492.
MÖSTL, E. RETTENBACHER, S. AND PALME, R. 2005. Measurement of corticosterone metabolites in birds’
droppings: an analytical approach. – Annals of the New York Academy of Sciences 1046: 17-34.
MULLER, M.N. AND WRANGHAM, R.W. 2004. Dominance, cortisol and stress in wild chimpanzees (Pan
troglodytes schweinfurthii). – Behavioral Ecology and Sociobiology 55: 332-340.
MUSIANI, M., MAMO, CH., BOITANI, L., CALLAGHAN, C., GATES, C.C., MATTEI, L., VISALBERGHI, E.,
BRECK, S. AND VOLPI, G. 2003. Wolf depredation trends and the use of fladry barriers to protect livestock
in Western North America. – Conservation Biology 17: 1538-1547.
NAVES, J., WIEGAND, T., REVILLA, E. AND DELIBES, M. 2003. Endangered species constrained by natural
and human factors: the case of brown bears in northern Spain. – Conservation Biology 17: 1276-1289.
NEFF, M.W., BROMAN, K.W., MELLERSH, C.S., RAY, K., ACLAND, G.M., AGUIRRE, G.D., ZIEGLE, J.S.,
OSTRANDER, E.A. AND RINE, J. 1999. A second-generation linkage map of the domestic dog, Canis
familiaris. – Genetics 151: 803-820.
NG, S.J., DOLE, J.W., SAUVAJOT, R.M., RILEY, S.P. AND VALONE, T.J. 2004. Use of highway
undercrossings by wildlife in southern California. – Biological Conservation 115: 499-507.
NOWAK, S., MYSLAJEK, R.W. AND JĘDRZEJEWSKA, B. 2005. Patterns of wolf Canis lupus predation on
wild and domestic ungulates in the Western Carpathian Mountains (S Poland). – Acta Theriologica 50:
263-276.
References
108
NSUBUGA, A.M., ROBBINS, M.M., ROEDER, A.D., MORIN, P.A., BOESCH, C. AND VIGILANT, L. 2004.
Factors affecting the amount of genomic DNA extracted from ape faeces and the identification of an
improved sample storage method. – Molecular Ecology 13: 2089-2094.
OEHLER, J.D. AND LITVAITIS, J.A. 1996. The role of spatial scale in understanding responses of mediumsized carnivores to forest fragmentation. Canadian Journal of Zoology 74: 2070-2079.
OLSON, T.L., DIENI, J.S. AND LINDZEY, F.G. 1997. Swift fox survey evaluation, productivity, and
survivorship in southeast Wyoming. In: Giddings, B. (ed.). Swift fox conservation team 1997 annual
report. Montana Department of Fish, Wildlife and Parks, Helena, pp. 57-76.
PAETKAU, D. AND STROBECK, C. 1994. Microsatellite analysis of genetic variation in black bear
populations. – Molecular Ecology 3: 489-495.
PAETKAU, D., WAITS, L.P., CLARKSON, P.L., CRAIGHEAD, L., VYSE, E., WARD, R. AND STROBECK, C.
1998. Variation in genetic diversity across the range of North American brown bears. Conservation
Biology 12: 418-429.
PALME, R. 2005. Measuring fecal steroids – guidelines for practical application. – Annals of the New
York Academy of Sciences 1046: 75-80.
PALME, R. AND MÖSTL, E. 1997. Measurement of cortisol metabolites in faeces of sheep as a parameter
of cortisol concentration in blood. – International Journal of Mammalian Biology 62 (supplement 2): 192197.
PALME, R., SCHATZ, S. AND MÖSTL, E. 2001. Influence of a vaccination on faecal cortisol metabolite
concentrations in cats and dogs. – Deutsche Tierärztliche Wochenschau 108: 23-25 [in German with
English summary].
PALSBØLL, P.J., ALLEN, J., BÉRUBÉ, M., CLAPHAM, P.J., FEDDERSEN, T.P., HAMMOND, P.S., HUDSON,
R.R., JØRGENSEN, H., KATONA, S., LARSON, A.H., LARSEN, F., LIEN, J., MATTILA, D.K., SIGURJÓNSSON,
J., SEARS, R., SMITH, T., SPONER, R., STEVICK, P. AND ØIEN, N. 1997. Genetic tagging of humpback
whales. Nature 388: 767-769.
PAQUET, P.C. AND FULLER, W. 1990. Scent marking and territoriality in wolves of Riding Mountain
National Park. In: Macdonald, D.W., Muller-Schwarze, D., Natynczuk, S.E. (eds.). Chemical signals in
vertebrates 5. Oxford University Press, New York, pp. 394-400.
PATTERSON, B.R., QUINN, N.W.S., BECKER, E.F. AND MEIER, D.B. 2004. Estimating wolf densities in
forested areas using network sampling of tracks in snow. Wildlife Society Bulletin 32: 938-947.
PEARCE, J. AND FERRIER, S. 2000. Evaluating the predictive performance of habitat models developed
using logistic regression. – Ecological Modelling 133: 225-245.
PEREIRA, R.J.G., DUARTE, J.M.B. AND NEGRÃO, J.A. 2006. Effects of environmental conditions, human
activity, reproduction, antler cycle and grouping on fecal glucocorticoids of free-ranging Pampas deer
stags (Ozotoceros bezoarticus bezoarticus). – Hormones and Behavior 49: 114-122.
References
109
PETERS, R.P. AND MECH, L.D. 1975. Scent-marking in wolves. American Scientist 63: 628-637.
PETERSON, R.O. 1998. The pit or the pendulum: issues in large carnivore management in natural
ecosystems. – In: Agee, J.K., Johnson, D.R. (eds). Ecosystem management for parks and wilderness.
University of Washington Press, Seattle, Washington, USA, pp. 105-117.
PETERSON, J.T. AND DUNHAM, J.A. 2003. Combining inferences from models of capture efficiency,
detectability, and suitable habitat to classify landscapes for conservation of threatened bull trout. –
Conservation Biology 17: 1070-1077.
PETRAUSKAS, L.R. AND ATKINSON, S.K. 2006. Variation of fecal corticosterone concentrations in captive
Steller sea lions (Eumetopias jubatus) in relation to season and behavior. – Aquatic Mammals 32: 168174.
PIGGOTT, M.P. AND TAYLOR, A.C. 2003. Remote collection of animal DNA and its applications in
conservation management and understanding the population biology or rare and cryptic species. –
Wildlife Research 30: 1-13.
PROMBERGER, C. AND SCHRÖDER, W. 1993. Proceedings of the workshop “Wolves in Europe – current
status and prospects”. – Munich Wildlife Society, Oberammergau, Germany.
PROMBERGER, C., IONESCU, O., PETRE, L., ROSCHAK, C., SÜRTH, P., FÜRPAß, B., TODICESCU, L.,
SANDOR, A., MINCA, M., STAN, T., HOMM, H., PREDOIU, G. AND SCURTU, M. 1997. Carpathian Large
Carnivore Project – annual report 1996/97. - Munich Wildlife Society and Romanian Wildlife Research
Department.
PUTMAN, R.J. 1984. Facts from faeces. Mammalian Review 14: 79-97.
RABB, G.B., WOOLPY, J.H. AND GINSBURG, B.E. 1967. Social relationships in a group of captive wolves.
– American Zoologist 7: 305-311.
RANDI, E. AND LUCCHINI, V. 2002. Detecting rare introgression of domestic dog genes into wild wolf
(Canis lupus) populations by Bayesian admixture analyses of microsatellite variation. – Conservation
Genetics 3: 31-45.
REED, J.M. AND BLAUSTEIN, A.R. 1997. Biologically significant population declines and statistical
power. Conservation Biology 11: 281-282
REED, J.Z., TOLLIT, D.J., THOMPSON, P.M. AND AMOS W. 1997. Molecular scatology: the use of
molecular genetic analysis to assign species, sex and individual identity to seal faeces. – Molecular
Ecology 6: 225-234.
REEDER, D.M. AND KRAMER, K.M. 2005. Stress in free-ranging mammals: integrating physiology,
ecology, and natural history. – Journal of Mammalogy 86: 225-235.
REINHARDT, I. AND KLUTH, G. 2007. Wölfe in Deutschland - Status. – In: Reinhardt, I., Kluth, G. (eds.).
Leben mit Wölfen – Leitfaden für den Umgang mit einer konfliktträchtigen Tierart in Deutschland. BfNSkripten 201, Bonn-Bad Godesberg, Germany, pp. 34-69.
References
110
REVILLA, E., PALOMARES, F. AND DELIBES M. 2001. Edge-core effects and the effectiveness of
traditional reserves for conservation: Eurasian badgers in Doñana National Park. – Conservation Biology
15: 148-158.
ROBERTS, S.C. AND LOWEN, C. 1997. Optimal patterns of scent marks in klipspringer (Oreotragus
oreotragus) territories. Journal Zool. Lond. 43: 565-578.
ROBINSON, I.H. AND DELIBES, M. 1988. The distribution of faeces by the Spanish lynx (Felis pardina).
Journal of Zoology 216: 577-582.
RODRIGUEZ-FREIRA, M. AND CRECENTE-MASEDA, R. 2008. Directional connectivity of wolf (Canis
lupus) populations in northwest Spain and anthropogenic effects on dispersal patterns. – Environmental
Modeling & Assessment 13: 35-51.
RHODES, J.R., WIEGAND, T., MCALPINE, C.A., CALLAGHAN, J., LUNNEY, D., BOWEN, M. AND
POSSINGHAM, H.P. 2006. Modeling species’ distributions to improve conservation in semiurban
landscapes: Koala case study. – Conservation Biology 20: 449-459.
ROMERO, L.M. 2004. Physiological stress in ecology: lessons from biomedical research. – Trends in
Ecology and Evolution 19: 249-255.
RYABOV, L.S. 1987. On the wolf synanthrophy in the Central Black Earth Belt. – Byulleten
Moskovskogo Obshchestva Ispytatelei Prirody, Otdelene Biologii 92: 3-12 [In Russian with English
summary].
SADLIER, L.M.J., WEBBON, C.C., BAKER, P.J. AND HARRIS, S. 2004. Methods of monitoring red foxes
Vulpes vulpes and badgers Meles meles: are field signs the answer? Mammalian Review 34: 75-98.
SANDS, J. AND CREEL, S. 2004. Social dominance, aggression and fecal glucocorticoid levels in a wild
population of wolves, Canis lupus. – Animal Behaviour 67: 387-396.
SANTINI, A., LUCCHINI, V., FABBRI, E. AND RANDI, E. 2007. Ageing and environmental factors affect
PCR success in wolf (Canis lupus) excremental DNA samples. – Molecular Ecology 7: 955-961.
SANTOS, M., VAZ, C., TRAVASSOS, P. AND CABRAL, J.A. 2007. Simulating the impact of socio-economic
trends on threatened Iberian wolf populations Canis lupus signatus in north-eastern Portugal. – Ecological
Indicators 7: 649-664.
SAUERWEIN, H., MÜLLER, U., BRÜSSEL, H., LUTZ, W. AND MÖSTL, E. 2004. Establishing baseline values
of parameters potentially indicative of chronic stress in red deer (Cervus elaphus) from different habitats
in western Germany. – European Journal of Wildlife Research 50: 168-172.
SCHATZ, S. AND PALME, R. 2001. Measurement of fecal cortisol metabolites in cats and dogs: a noninvasive method for evaluating adrenocortical function. – Veterinary Research Communications 25: 271287.
References
111
SCHEIBER, I.B.R., KRALJ, S. AND KOTRSCHAL, K. 2005. Sampling effort / frequency necessary to infer
individual acute stress responses from fecal analysis in greylag geese (Anser anser). – Annals of the New
York Academy of Sciences 1046: 154-167.
SCHENKEL, R. 1947. Expression studies of wolves. – Behaviour 1: 81-129.
SILVEIRA, L., JÁCOMO, A.T.A. AND DINIZ-FILHO, J.A.F. 2003. Camera trap, line transect census and track
surveys. Biological Conservation 114: 351-355.
SLATKIN, M. 1995. A measure of population subdivision based on microsatellite allele frequencies. –
Genetics 139: 457-462.
SMITH, D.A., RALLS, K., HURT, A., ADAMS, B., PARKER, M. AND MALDONADO, J.E. 2006. Assessing
reliability of microsatellite genotypes from kit fox faecal samples using genetic and GIS analyses. –
Molecular Ecology 15: 387-406.
SOVADA, M.A. AND ROY, C.C. 1996. Summary of swift fox research activities conducted in western
Kansas – Annual report. In: Luce, B., Lindzey, F. (eds.). Annual report of the Swift fox conservation
team. Wyoming Game and Fish Department, Lander, pp. 64-68.
STEPHENS, P.W. AND PETERSON, R.O. 1984. Wolf-avoidance strategies of moose. – Ecography 7: 239244.
TABERLET, P. AND LUIKART, G. 1999. Non-invasive genetic sampling and individual identification. –
Biological Journal of the Linnean Society 68: 41-55.
TABERLET, P., WAITS, L..P. AND LUIKART, G. 1999. Noninvasive genetic sampling: look before you leap.
Trends in Ecology and Evolution 14: 323-327.
TABERLET, P., GRIFFIN, S., GOOSSENS, B., QUESTIAU, S., MANCEAU, V., ESCARAVAGE, N., WAITS, L.P.
AND
BOUVET, J. 1996. Reliable genotyping of samples with very low DNA quantities using PCR. –
Nucleic Acids Research 24: 3189-3194.
TABERLET, P., CAMARRA, J.-J., GRIFFIN, S., UHRES, E., HANOTTE, O., WAITS, L.P., DUBOIS-PAGANON,
C., BURKE, T. AND BOUVET, J. 1997. Non-invasive genetic tracking of the endangered Pyrenean brown
bear population. – Molecular Ecology 6: 869-876.
TAYLOR, A.R. AND KNIGHT, R.L. 2003. Wildlife responses to recreation and associated visitor
perceptions. – Ecological Applications 13: 951-963.
THEUERKAUF, J. AND JĘDZEJEWSKI, W. 2002. Accuracy of radiotelemetry to estimate wolf activity and
locations. Journal of Wildlife Management 66: 859-864.
THEUERKAUF, J., ROUYS, S. AND JĘDZEJEWSKI, W. 2003a. Selection of den, rendezvous, and resting sites
by wolves in the Bialowieza Forest, Poland. – Canadian Journal of Zoology 81: 163-167.
THEUERKAUF, J., JĘDZEJEWSKI, W., SCHMIDT, K. AND GULA, R. 2003b. Spatiotemporal segregation of
wolves from humans in the Bialowieza Forest (Poland). – Journal of Wildlife Management 67: 706-716.
References
112
THEUERKAUF, J., GULA, R., PIRGA, B., TSUNODA, H., EGGERMANN, J. BRZEZOWSKA, B., ROUYS, S. AND
RADLER, S. 2007. Human impact on wolf activity in the Bieszczady Mountains, SE Poland. – Annales
Zoologici Fennici 44: 225-231.
THIEL, R.P. 1985. The relationship between road densities and wolf habitat suitability in Wisconsin. American Midland Naturalist 113: 404-407.
THIEL, D., JENNI-EIERMANN, S., BRAUNISCH, V., PALME, R. AND JENNI, L. 2008. Ski tourism affects
habitat use and evokes a physiological stress response in capercaillie Tetrao urogallus: a new
methodological approach. – Journal of Applied Ecology 45: 845-853.
THURBER, J.M., PETERSON, R.O., DRUMMER, T.D. AND THOMASMA, S.A. 1994. Gray wolf response to
refuge boundaries and roads in Alaska. - Wildlife Society Bulletin 22: 61-68.
TOUMA, C. AND PALME, R. 2005. Measuring fecal glucocorticoid metabolites in mammals and birds: the
importance of validation. – Annals of the New York Academy of Sciences 1046: 54-74.
TREVES, A., NAUGHTON-TREVES, L., HARPER, E.K., MLADENOFF, D.J., ROSE, R.A., SICKLEY, T.A. AND
WYDEVEN, A.P. 2004. Predicting human-carnivore conflict: a spatial model derived from 25 years of data
on wolf predation on livestock. – Conservation Biology 18: 114-125.
TSUNODA, H., GULA, R., THEUERKAUF, J., ROUYS, S., RADLER, S., PIRGA, B., EGGERMANN, J. AND
BRZEZOWSKA, B. 2009. How does parental role influence the activity and movements of breeding
wolves? – Journal of Ethology 27: 185-189.
VAN ETTEN, K.W., WILSON, K.R. AND CRABTREE, R.L. 2007. Habitat use of red foxes in Yellowstone
National Park based on snow tracking and telemetry. Journal of Mammalogy 88: 1498-1507.
VILÁ, C., URIOS, V. AND CASTROVIEJO, J. 1994. Use of faeces for scent marking in Iberian wolves (Canis
lupus). Canadian Journal of Zoology 72: 374-377.
VILÀ, C., URIOS, V. AND CASTROVIEJO, J. 1995. Observations on the daily activity patterns in the Iberian
wolf. In: Carbyn, L.N., Fritts, S.H., Seip, D.R. (eds.), Ecology and Conservation of Wolves in a Changing
World. Canadian Circumpolar Institute, Alberta, Occasional Publication No. 35, pp. 335-340.
WASHBURN, B.E. AND MILLSPAUGH, J.J. 2002. Effects of simulated environmental conditions on
glucocorticoid metabolite measurements in white-tailed deer feces. – General and Comparative
Endocrinology 127: 217-222.
WASSER, S.K., HUNT, K.E., BROWN, J.L., COOPER, K., CROCKETT, C.M., BECHERT, U., MILLSPAUGH,
J.J., LARSON, S. AND MONFORT, S.L. 2000. A generalized fecal glucocorticoid assay for use in a divers
array of nondomestic mammalian and avian species. – General and Comparative Endocrinology 120:
260-275.
WECKERLY, F.W. AND RICCA, M.A. 2000. Using presence of sign to measure habitats used by Roosevelt
elk. Wildlife Society Bulletin 28: 146-153.
References
113
WEIR, B.S. AND COCKERHAM, C.C. 1984. Estimating F-statistics for the analysis of population structure.
– Evolution 38: 1358-1370.
WHITTINGTON, J., ST. CLAIR, C.C. AND MERCER, G. 2004. Path tortuosity and the permeability of roads
and trails to wolf movement. – Ecology and Society 9: 4.
WHITTINGTON, J., ST. CLAIR, C.C. AND MERCER, G. 2005. Spatial responses of wolves to roads and trails
in mountain valleys. – Ecological Applications 15: 543-553.
WILSON, G.J. AND DELAHAY, R.J. 2001. A review of methods to estimate the abundance of terrestrial
carnivores using field signs and observation. Wildlife Research 28: 151-164.
WINGFIELD, J.C. AND SAPOLSKY, R.M. 2003. Reproduction and resistance to stress:when and how. –
Journal of Neuroendocrinology 15: 711-724.
WOODS, J.G., PAETKAU, D., LEWIS, D., MCLELLAN, B.N., PROCTOR, M. AND STROBECK, C. 1999.
Genetic tagging of free-ranging black and brown bears. – Wildlife Society Bulletin 27: 616-627.
WYDEVEN, A.P., MLADENOFF, D.J., SICKLEY, T.A., KOHN, B.E., THIEL, R.P. AND HANSEN, J.L. 2001.
Road density as a factor in habitat selection by wolves and other carnivores in the Great Lakes Region. –
Endangered Species UPDATE 18: 110-114.
YOUNG, K.M., WALKER, S.L., LANTHIER, C., WADDELL, W.T., MONFORT, S.L. AND BROWN, J.L. 2004.
Noninvasive monitoring of adrenocortical activity in carnivores by fecal glucocorticoid analysis. –
General and Comparative Endocrinology 137: 148-165.
ZIMEN, E. AND BOITANI, L. 1979. Status of the wolf in Europe and the possibility for conservation and
reintroduction. – In: E. Klinghammer (ed.), The behaviour and ecology of wolves. STPM Press, New
York, pp. 43-83.
ZUB, K., THEUERKAUF, J., JĘDRZEJEWSKI, W., JĘDRZEJEWSKA, B., SCHMIDT, K. AND KOWALCZYK, R.
2003. Wolf pack territory marking in the BiałowieŜa Primeval Forest (Poland). Behaviour 140: 635-648.
Acknowledgements
114
ACKNOWLEDGEMENTS
Ich danke Prof. Dr. Wolfgang H. Kirchner dafür, dass er mich als Doktorandin in seiner
Arbeitsgruppe aufgenommen hat und dafür, dass er mich immer meinen eigenen Weg gehen ließ.
Prof. Dr. K.-P. Hoffmann danke ich für die Übernahme des Koreferats.
Ich danke allen Mitarbeitern der AG Verhaltensbiologie, im Besonderen Dr. Pia Aumeier, dem
Herz der AG, und Regina Lehnart, die immer den Überblick behielt, für die nette Atmosphäre
und den guten Zuspruch in Zeiten der Krise.
A special thank goes to Dr. Roman Gula, for all he taught me about good wildlife science, for his
support during fieldwork and getting along in the day-by-day live in Ustrzyki Dolne, for always
having an open door and an open ear for me, as well as for detailed revisions on the manuscript.
Bei Dr. Jörn Theuerkauf möchte ich mich dafür bedanken, dass er jederzeit und rekordverdächtig
schnell auf meine Fragen vom anderen Ende der Welt geantwortet hat, wiederholt behilflich war
bei der Literaturbeschaffung, für Korrekturen an einem Teil der Arbeit und dafür, dass er mich
mit seiner unendlichen Power immer wieder angesteckt hat – schade, dass wir uns nur so selten
persönlich getroffen haben.
I thank Prof. Dr. Francisco Fonseca for welcoming me in his research team.
A special thank goes also to Bartosz Pirga and Barbara Brzezowska for making me feel like
home in the far-away south-east of Poland.
Bei meinem Lebenspartner, Martin Singer, bedanke ich mich dafür, dass er, vor allem in der
Endphase der Promotion, meine Gemütsschwankungen geduldig ertragen und mir das Motto
„positiv denken“ näher gebracht hat – es war nicht immer leicht mit mir.
Last but not least gilt mein unermesslicher Dank meinen Eltern, Friedhelm und Christa
Eggermann, sowie meiner Schwester, Karen Eggermann, die immer für mich da waren, mich in
allen Lebenssituationen und bei allen Unternehmungen unterstützt haben und mir auf alle
erdenklichen Weisen behilflich waren – ohne Euch hätte ich es nicht geschafft.
115
Curriculum Vitae
CURRICULUM VITAE
PERSÖNLICHE DATEN Name
SCHULBILDUNG
STUDIUM
Julia Eggermann
Geburtsdatum
18.05.1979
Geburtsort
Witten
Staatsangehörigkeit
deutsch
Familienstand
ledig
1986 – 1990
Harkortschule, Witten
1990 – 1999
Albert-Martmöller Gymnasium, Witten
09 / 2001
Vordiplom, Ruhr-Universität Bochum
11 / 2003 – 08 / 2004 Diplomarbeit
„Activity pattern of wolves (Canis lupus) in the
Bieszczady Mountains, Poland”
AG Verhaltensbiologie & Didaktik der Biologie
Prof. W.H. Kirchner
08 / 2004
Diplom, Ruhr-Universität Bochum
ab 11 / 2004
Promotion, Ruhr-Universität Bochum
WISSENSCHAFTLICHE 11 / 2003 – 04 / 2004 Stipendium des DAAD
TÄTIGKEITEN
06 / 2005 – 05 / 2007 Stipendium des Allgemeinen Promotionskollegs
der Ruhr-Universität Bochum
02 / 2008 – 05 / 2009 Wissenschaftliche Hilfskraft in der AG
Verhaltensbiologie & Didaktik der Biologie,
Prof. W.H. Kirchner
Ruhr-Universität Bochum