Farm characteristics in Slovene wolf habitat related to attacks on sheep

Full text

Applied
Animal
Behaviour
Science
144 (2013) 46–
56
Contents
lists
available
at
SciVerse
ScienceDirect
Applied
Animal
Behaviour
Science
journa
l
h
o
me
pag
e:
www.elsevier.com/locate/applanim
Farm
characteristics
in
Slovene
wolf
habitat
related
to
attacks
on
sheep
Diederik
van
Lierea,∗,
Cathy
Dwyerb,
Duˇ
sanka
Jordanc,
Andrea
Premik-Baniˇ
cd,
Aleˇ
s
Valenˇ
ciˇ
ce,
Drago
Kompanc,
Nataˇ
sa
Siardc
aCABWIM
consultancy,
Minervalaan
17,
9401
EB
Assen,
Netherlands
bSRUC,
Roslin
Institute
Building,
Easter
Bush,
Midlothian,
EH25
9RG,
United
Kingdom
cDepartment
of
Animal
Science,
Biotechnical
faculty,
University
of
Ljubljana,
Groblje
3,
1230
Domˇ
zale,
Slovenia
dEducational
Society
Noah
School,
Zaloˇ
ska
Cesta
155,
1000
Ljubljana,
Slovenia
eDolenje
pri
Jelˇ
sanah
55a,
6254
Jelˇ
sane,
Slovenia
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Accepted
10
December
2012
Available online 9 January 2013
Keywords:
Depredation
Wolf
attack
Night
enclosure
Electric
fence
Sheep
behaviour
Guarding
dog
a
b
s
t
r
a
c
t
We
aimed
to
characterize
differences
between
sheep
farms
in
wolf
habitat
in
Slovenia
that
either
suffered
from
wolf
attacks
(n
=
30)
or
not
(n
=
30)
during
the
pasture
seasons
2008–2010.
Main
pasture
season
was
from
April
until
November.
Median
fenced
pastures
were
2.7
ha
and
herd
size
was
93
sheep.
The
three-year
period
contained
288
attacks,
mostly
occurring
in
May
(36),
and
secondly
peaking
in
October
(23).
78%
of
all
attacks
occurred
at
night.
Significantly
fewer
non-attacked
than
attacked
farms
had
mixed
herds
(17%
versus
40%).
Wolves
killed
a
median
of
4
sheep
per
attack.
If
herds
included
goats,
2
goats
could
be
killed
in
addition.
Sheep
were
driven
to
a
night
facility
before
dusk
by
43%
of
non-attacked
farmers,
and
significantly
fewer
attacked
farms
(10%).
Significantly
fewer
attacked
than
non-attacked
farms
kept
sheep
in
closed
night
barns
or
a
separately
fenced
night-area
(20%
versus
50%).
Guarding
dogs
(usually
2
per
herd)
were
kept
by
53%
attacked
and
43%
non-
attacked
farms.
Average
fence
height
was
115
cm
and
did
not
differ
between
attacked
or
non-attacked
farms.
87%
non-attacked
farms
had
wire-mesh
fences
(either
electric
or
not)
instead
of
fences
with
horizontal
single
wires,
which
was
significantly
more
than
at
attacked
farms
(61%).
Significantly
more
attacked
(89%)
than
non-attacked
farms
(60%)
had
electric
fences
(mobile
or
fixed,
fixed
ones
could
be
combined
with
physical
fences).
In
spite
of
farmers
using
electric
fences,
annual
attack
number
was
significantly
higher
at
farms
with
a
history
of
wolf
attacks
than
at
new
farms
(4
versus
1).
Electric
fences
or
guarding
dogs
as
used
in
the
study
area
proved
ineffective:
they
did
not
prevent
wolf
attacks
or
reduce
killing
rates.
Adoption
of
mesh
instead
of
single
wires,
polarity
alternation
of
live
with
ground
wires
in
electric
fences,
and
fences
higher
than
145
cm
seem
improvements.
However,
potentially,
improved
fencing
could
also
prevent
sheep
from
breaking
out,
if
wolves
have
found
ways
to
enter
the
fenced
area,
and
might
result
in
surplus
killing.
Alternative
strategies
are:
(1)
to
keep
sheep
in
closed
night
barns
and
to
move
sheep
there
before
dusk
and
(2)
to
research
(a)
wolf
attack
rates
and
feasibility
of
separating
sheep
and
goat
herds;
(b)
sheep
and
goat
responses
to
predator
attacks
and
methods
that
assist
sheep
and
goats
to
avoid
being
attacked;
(c)
wolf
deterring
methods
focused
on
systematic
negative
reinforcement
of
chasing
and
consumption
of
livestock.
© 2012 Elsevier B.V. All rights reserved.
∗Corresponding
author.
Tel.:
+31
611391718;
fax:
+31
592406721.
E-mail
address:
dvanliere@cabwim.com
(D.
van
Liere).
1.
Introduction
In
Slovenia,
85%
of
the
whole
territory
is
considered
less
favoured
area
for
agriculture
(Rural
Development
0168-1591/$
–
see
front
matter ©
2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.applanim.2012.12.005

D.
van
Liere
et
al.
/
Applied
Animal
Behaviour
Science
144 (2013) 46–
56 47
Programme,
2007).
Sustainable
animal
production
is
mainly
limited
to
sheep
and
goat
breeding
in
mountainous
and
hilly
perennial
grasslands
with
shallow
soils
of
poor
quality,
such
as
in
the
Dinaric
karst
area.
This
Natura2000
area
is
an
EU
protected
natural
corridor
with
high
biodi-
versity
maintained
by
grazing
small
ruminants.
It
links
the
Alps
in
the
northwest
with
mountainous
Gorski
Kotar
in
the
southeast
border
with
Croatia.
It
is
also
the
main
Slovene
habitat
for
wolves
(Canis
lupus).
For
the
last
decade,
wolf
numbers
in
Slovenia
are
30–50
(Linnell
et
al.,
2002),
but
also
60–100
is
mentioned
(website
Large
Carnivore
Initiative
for
Europe).
Slovene
researchers
recently
mentioned
43
wolves,
and
10–12
packs
consist-
ing
of
4
adults
on
average
(Krofel,
2012).
The
total
area
of
Slovene
wolf
territories
is
around
4700
km2(ˇ
Cerne
et
al.,
2010),
implying
a
density
of
1
wolf/100
km2.
Esti-
mated
total
biomass
of
ungulates
in
these
wolf
territories
is
245
kg/km2(Jerina,
unpublished,
cited
by
Kavˇ
ciˇ
c
et
al.,
2011).
This
ungulate
biomass
estimate
represents
28%
of
860
kg/km2,
found
for
the
Italian
Casentinesi
forests
and
considered
very
high
(Apollonio
et
al.,
2004).
Thus,
wild
ungulate
prey
is
relatively
abundant
in
Slovenia,
as
Slovene
wolf
densities
are
5
times
lower
compared
to
the
Casenti-
nesi
forests.
Indeed,
Slovene
wolves
mainly
predate
cervides
and,
to
some
extent,
young
wild
boars
(Sus
scrofa)
(85
and
5%
of
consumed
biomass,
respectively),
whereas
domestic
ani-
mals
represent
10%
of
consumed
biomass
(Krofel
and
Kos,
2010).
However,
relations
are
ambiguous
between
prey
density
and
prey
selection
in
European
regions
with
wild
and
domestic
animals.
Some
researchers
explain
predation
on
domestic
animals
by
low
availability
of
wild
ungulates
(Capitani
et
al.,
2004;
Gula,
2008),
but
others
do
not
find
such
relations
or
point
to
contradictory
results
(Espuno
et
al.,
2004;
Kojola
et
al.,
2004;
Mattioli
et
al.,
2011;
Meriggi
and
Lovari,
1996;
Meriggi
et
al.,
1996).
In
spite
of
preferences
for
cervides,
cervid
abundance
and
low
wolf
densities,
Slovene
wolves
have
increasingly
killed
livestock.
There
were
61
cases
with
218
small
rumi-
nants
(90%
sheep,
10%
goats)
killed
in
2005
and
542
cases
with
1931
small
ruminants
(85%
sheep,
15%
goats)
killed
in
2010
(ARSO,
2005-2010).
Beside
this
livestock,
cattle,
donkeys
and
horses
were
killed
too,
but
this
was
in
5%
of
cases
on
average.
State
compensation
for
the
economic
losses
increased
almost
eleven-fold
from
D29,000
in
2005
to
D313,000
in
2010.
However,
only
a
small
number
of
farms
had
a
high
number
of
attacks
each
year:
out
of
320
known
farms
with
wolf
attacks,
25
had
over
50%
of
all
compensated
damage
(ˇ
Cerne
et
al.,
2010).
Relatively
high
number
of
attacks
with
a
few
farms
seems
common
in
Europe
(Boitani,
2000)
and
the
USA
(Breck
and
Meier,
2004;
Smallidge
et
al.,
2008).
In
Switzerland,
wolves
return
to
the
pastures
where
their
hunts
have
been
successful
(Nationale
Koordination
Herdenschutz,
2011).
Wolves’
preferences
for
specific
farms
have
also
been
found
in
Slovakia,
where
12%
of
farms
accounted
for
79–82%
of
all
losses.
Indeed,
the
same
farmers
tended
to
have
problems
each
year
(Rigg
et
al.,
2011).
Gazzola
et
al.
(2008)
also
reported
that
in
Arezzo,
a
province
in
Italy,
35
attacks
(14%
of
the
total
attacks)
involved
44%
of
the
total
number
(536)
of
sheep
and
goats
killed
in
the
whole
province
from
1998
to
2001.
General
recommendations
to
protect
livestock
from
wolves
are
to
keep
herds
within
fences,
and
electric
fences
in
particular,
to
include
one
or
more
guarding
dogs
in
the
herd,
and
to
adopt
a
night
enclosure
(Boitani,
2000;
Espuno
et
al.,
2004;
Oberle,
2010;
Plan
d’action
national
sur
le
loup,
2008;
Reinhardt
and
Kluth,
2007; ˇ
Strbenac
et
al.,
2010).
However,
it
is
not
clear
why
some
farms
are
frequently
attacked
while
others
in
the
same
wolf
region
are
not.
Mech
et
al.
(2000)
compared
cattle
farms
in
Minnesota
(USA)
and
concluded
that
the
larger
the
herd
and
the
fur-
ther
away
from
the
farm,
the
more
likely
it
was
that
the
herd
would
be
attacked.
For
sheep
flock
sizes,
it
was
the
opposite:
larger
flocks
had
fewer
attacks.
However,
farm-
ing
conditions
and
herd
sizes
differ
significantly
between
the
USA
and
Southern
Europe.
Espuno
et
al.
(2004)
did
similar
research
in
the
Mercantour
area,
but
could
not
con-
firm
Mech
et
al.’s
findings.
Espuno
et
al.
(2004)
identified
differences
between
pastures
as
by
far
the
strongest
deter-
minant
of
attack
number
and
of
number
of
sheep
killed.
They
suggested
that
differences
between
pastures
con-
cerned
habitat
attributes
and
human
attendance.
Besides
pasture
differences,
Espuno
et
al.
(2004)
suggested
factors
such
as
sheep
management
methods,
learning
differences
between
wolf
packs
and
local
densities
of
wild
prey.
How-
ever,
little
research
has
compared
farm
practises
in
wolf
habitats
with
regard
to
differences
in
cues
that
potentially
appeal
to
wolves.
Herds
in
Slovenia
commonly
number
less
than
100
sheep
(Udovˇ
c
et
al.,
2011).
Lambing
mostly
occurs
through-
out
the
year.
In
late
autumn,
a
farmer
moves
his
herd
from
the
pasture
and
keeps
it
in
a
barn
for
some
months,
mostly
near
the
farm.
The
pasture
season
is
between
March
and
December.
The
animals
are
moved
between
pastures
or
paddocks
within
a
pasture,
depending
on
when
the
grass
is
eaten.
Pastures
can
have
fixed
fences,
but
paddocks
mostly
have
mobile
fences.
Fences
can
be
with
or
without
elec-
tricity.
There
may
be
a
night
facility,
such
as
a
small
fenced
area,
with
or
without
a
night
barn.
Our
research
aimed
to
compare
practises
of
sheep
farms
with
and
without
problems
with
wolves
in
Slovene
wolf
habitat.
We
investigated
management,
pasture
character-
istics
and
potentially
appealing
cues
that
may
explain
differences
in
wolves’
detection
and
approach.
2.
Material
and
methods
2.1.
Selection
of
farmers
Using
a
farmers’
database,
provided
by
the
Department
of
Animal
Science
at
the
Biotechnical
faculty,
University
of
Ljubljana,
we
arbitrarily
selected
60
farmers
from
regions
where
wolf
attacks
on
sheep
had
regularly
occurred
for
years.
Thirty
farmers
had
wolf
attacks
and
30
had
no
wolf
attacks
in
the
period
between
2008
and
2010.
Assess-
ment
of
damage
on
livestock
caused
by
wolves
is
regulated
(Pravilnik,
2005).
Farmers
have
to
take
the
prescribed
pre-
vention
measures,
if
they
want
to
get
compensation
by
the
state
for
damage
caused
by
protected
wildlife.
Attacks
by
wolves
are
confirmed
by
an
authorized
person
from
the
state
forest
service,
including
numbers
of
sheep
and
other
animals
killed
in
attack.
Except
for
two
farmers,
all
had

48 D.
van
Liere
et
al.
/
Applied
Animal
Behaviour
Science
144 (2013) 46–
56
official
confirmation
proving
that
sheep
were
indeed
attacked
by
the
wolves.
The
two
exceptions
claimed
wolf
attacks
were
confirmed
by
hunters
and
other
experienced
farmers.
All
farmers
but
one
were
located
in
two
regions:
the
Notranjsko-
and
Obalno-kraˇ
ska
region
with
49
farmers,
and
South-East
Slovenia
around
the
town
Koˇ
cevje
with
10
farmers.
We
balanced
numbers
of
farmers
with
and
with-
out
attacks
in
both
regions
as
much
as
possible
in
order
to
balance
possible
regional
differences
between
farms.
One
farmer
with
attacked
sheep
was
from
the
Goriˇ
ska
region.
2.2.
Questionnaire
We
interviewed
the
selected
farmers
between
October
2010
and
February
2011.
We
developed
a
questionnaire
which
was
the
same
for
all
farmers
and
covered
(a)
man-
agement:
monthly
dependent
use
of
the
pastures,
breeding
regime,
frequency
of
visits
to
the
pasture,
use
of
night
enclosures,
mixing
the
flock
of
sheep
with
other
animals,
use
of
deterring
methods,
presence
of
conspicuous
objects
on
the
pasture
and
of
items
that
smell,
like
sheep
remains;
(b)
pasture
characteristics:
size,
altitude,
location,
pres-
ence
of
fences,
openness
and
main
wind
direction;
(c)
flock
characteristics:
herd
size,
sheep
breed,
behaviour
of
sheep,
response
to
an
attack
(if
applicable),
and
sounds
by
sheep
or
other
animals
present.
All
questions
were
neutrally
formu-
lated
and
of
closed
format,
because
we
needed
unbiased
answers,
and
needed
to
unambiguously
categorize
them
for
statistical
analysis.
One
trained
person,
familiar
with
sheep
breeding
practice,
interviewed
both
groups
of
farm-
ers.
Farmers
that
had
problems
with
wolves
were
also
asked
about
the
attacks:
whether
an
attack
was
witnessed,
what
number
of
sheep
and
other
animals
were
killed,
and
about
time
of
the
day
and
dates
of
the
attacks.
Most
data
came
from
files
farmers
had
been
using
over
the
years
2008–2010
to
request
compensation
money
from
the
gov-
ernment.
Not
all
farmers
could
provide
information
on
all
questions
and
if
the
farmer
was
in
doubt,
or
was
not
able
to
provide
an
answer,
the
score
for
the
particular
question
was
a
missing
value.
Consequently,
sample
size
could
be
less
than
30
per
category,
depending
on
the
variable
stud-
ied.
Farmers
that
were
interviewed
in
the
last
months
of
2010
were
contacted
again
in
2011
in
order
to
finalize
their
dataset
for
2010.
2.3.
Analysis
of
the
pasture
As
a
reference
we
used
the
pasture
that
the
farmer
used
most,
or
was
the
most
representative
according
to
the
farmer.
However,
whenever
we
interviewed
a
farmer
that
had
problems
with
wolf
attacks,
we
focussed
on
the
pasture
of
the
most
recent
attack,
or
on
the
pasture
with
attacks
the
farmer
remembered
best
in
the
period
2008–2010.
This
‘snap-shot’
approach
did
not
allow
historical
analysis
of
changes
that
were
made
at
the
farm.
Historical
anal-
ysis
was
also
hardly
possible,
because
many
farmers
did
not
have
detailed
memory
of
changes,
nor
kept
detailed
time
administrations
of
changes
to
farm
practise.
Satel-
lite
photos
based
on
GPS
coordinates
taken
with
a
GPS
device
(Garmin
eTrex
Vista
h)
at
the
spot
provided
an
overview
of
the
representative
pasture,
including
forested
areas.
These
photos
of
BirdsEye
Imagery
were
generated
by
BaseCamp
software
from
Garmin
Ltd.
The
farmer
indicated
the
boundaries
of
his
pasture,
and
main
wind
direction
in
these
photos.
We
entered
the
information
into
the
satellite
photo
program,
calculated
areas,
and
defined
whether
the
pasture
was
open
(less
than
50%
of
the
surface
was
cov-
ered
by
bushes
or
trees)
or
not.
For
24
farmers
the
BirdsEye
photos
were
not
clear
enough
and
we
used
prints
of
Geope-
dia
instead
(an
interactive
online
atlas
and
map
of
Slovenia
developed
by
Sinergise
d.o.o.)
in
combination
with
location
codes
of
farmers
provided
by
the
Department
of
Animal
Sci-
ence
at
the
Biotechnical
faculty,
University
of
Ljubljana.
We
transferred
the
information
from
the
prints
into
Geopedia.
Comparison
of
size
estimations
of
pastures
used
in
both
programs
(BaseCamp
and
Geopedia),
showed
differences
between
1%
and
4%.
2.4.
Statistical
analysis
If
variables
had
a
skewed
distribution,
we
used
modal
or
median
values
and
25th
to
75th
percentile
interquartile
range
(IQR)
as
descriptive
parameters.
Within
the
category
of
farmers
with
wolf
problems,
we
analyzed
frequencies
of
wolf
attacks,
numbers
of
animals
that
were
killed
per
attack,
as
well
as
interval
durations
between
attacks.
We
applied
related
samples
Friedman
non-parametric
analy-
sis
or
Wilcoxon
signed
rank
tests
for
comparisons
between
years
within
the
category
of
farmers
with
wolf
attacks.
We
mainly
applied
!2tests
in
case
of
comparisons
between
farms
with
and
without
wolf
attacks,
or
Fisher
exact
tests
if
expected
values
in
!2tests
were
less
than
5.
In
other
cases
we
used
Mann–Whitney
U
tests.
If
the
variable
distribution
approached
normal
distribution,
we
applied
t-tests.
Cor-
relations
between
variables
were
analyzed
with
Pearson
correlation
tests.
PASW
(SPSS)
statistical
package,
version
18.0,
was
used
for
all
analyses,
except
Fisher
exact
tests,
which
were
calculated
in
a
spreadsheet.
3.
Results
3.1.
Wolf
attacks
Direct
observations
of
wolf
attacks
were
scarce.
Out
of
the
total
of
30
farmers
that
had
wolf
attacks
at
their
farm,
4
had
witnessed
an
actual
attack:
3
by
one
wolf
and
1
by
7
wolves.
Two
farmers
had
witnessed
wolves
jumping
over
electric
fences
up
to
145
cm.
Three
farmers
described
sheep
running
away
from
the
wolf
and
1
had
seen
the
sheep
breaking
through
the
(mobile)
electric
fence.
The
fourth
farmer
had
observed
that
the
sheep
that
was
about
to
be
attacked
did
not
show
any
reaction.
Long-term
effects
that
farmers
had
mentioned
were
(a)
sheep’s
reluctance
to
return
to
the
pasture
where
the
attack
had
taken
place
(2
farmers),
(b)
a
change
from
distressed
behaviour
to
no
obvi-
ous
reactions
after
experiencing,
but
surviving
subsequent
attacks
(1
farmer)
and
(c)
ewes
that
abort
their
pregnancy,
one
or
two
weeks
after
the
attack
(2
farmers).
The
30
farmers
with
wolf
attacks
on
their
sheep
flocks
had
experienced
288
attacks
over
the
three
year
study
period:
74
attacks
in
2008,
82
in
2009,
and
132
in
2010.

D.
van
Liere
et
al.
/
Applied
Animal
Behaviour
Science
144 (2013) 46–
56 49
Fig.
1.
Frequency
distribution
of
number
of
wolf
attacks
in
Slovenia
summed
per
farm
over
the
years
2008–2010.
There
was
no
difference
between
years
in
the
number
of
attacks
(Friedman
non-parametric
analysis:
P
=
0.13).
Median
value
was
6
attacks
per
farm
(IQR
2–14;
Fig.
1),
so
2
attacks
per
year.
Within
a
year,
repeated
attacks
mostly
occurred
within
5
days
of
each
other
(median
=
11
days,
IQR
4–23).
Of
the
attacked
farms,
18
had
a
history
of
attacks
for
3
years
(or
more),
whereas
the
other
12
experienced
their
first
attacks
around
the
time
of
the
study.
The
18
farms
that
had
a
history
of
wolf
attacks,
had
used
electric
fenc-
ing
throughout
the
observation
period.
The
number
of
attacks
at
these
farms
was
5
in
2008
(median,
IQR
1–6),
4
in
2009
(IQR
1–6)
and
3
in
2010
(IQR
3–7),
but
attack
numbers
in
2008
did
not
differ
from
2009,
nor
from
2010
(related
samples
Wilcoxon
signed
rank
test:
P
=
0.86
and
0.88
respectively).
Overall,
farms
with
an
attack
history
had
a
median
of
4.33
attacks
per
year.
However,
the
12
farms
without
an
attack
history
had
1
attack
in
their
first
year.
These
attack
levels
differed
significantly
(Mann–Whitney
U
test:
P
<
0.05,
Fig.
2).
The
12
first
year
farms
did
not
use
a
fence
at
all
(2),
used
a
physical
fence
(3),
or
used
an
electric
fence
(7).
If
electric
fences
were
applied,
it
was
not
known
whether
these
were
applied
before
or
after
the
start
of
the
attacks.
Of
the
223
wolf
attacks
with
registered
dates
in
the
period
2008–2010,
a
bimodal
pattern
could
be
recognized
with
a
peak
of
36
attacks
in
May
and
a
second
peak
of
23
attacks
in
October.
Attacks
were
most
frequent
at
night
(defined
as
the
period
after
sunset
and
before
sunrise;
78%),
whereas
15%
happened
in
the
morning
(starting
at
sun-
rise),
and
7%
in
the
afternoon
(period
after
12
o’clock
until
sunset).
We
had
detailed
registrations
of
the
number
of
livestock
killed
from
18
farmers
with
flocks
of
sheep
only.
All
but
1
had
lambs
in
the
flock
during
attacks.
Wolves
mostly
killed
Fig.
2.
Median
annual
number
of
attacks
(and
IQR)
by
wolves
at
Slovene
sheep
farms
experiencing
them
in
the
first
year,
when
electric
fences
may
or
may
not
have
been
applied,
versus
farms
experiencing
wolf
attacks
for
3
years
or
more,
but
always
applying
electric
fences
(*:
P
<
0.05).
1
to
2
sheep
per
attack
(median
=
4,
IQR
1–5.5),
with
1
adult
sheep
(IQR
1–2.5)
and
2
lambs
(IQR
0–3.5).
Surplus
killing
(defined
as
a
kill
of
10
or
more
sheep)
occurred
in
10%
of
the
farms
and
in
36
attacks
(12.5%
of
all
attacks).
3.2.
Comparison
between
farms
3.2.1.
Husbandry
Table
1
provides
a
summary
of
the
most
significant
com-
parisons
in
this
study.
The
start
or
ending
month
of
the
pasture
season
did
not
differ
between
farmers
with
attacks
and
those
without
(!2(2) =
0.3;
P
=
0.86
and
!2(2) =
1.5;
P
=
0.47
respectively).
46
farmers
had
a
regular
grazing
season
and
31
of
them
started
the
season
in
April,
but
by
May
all
farmers
had
herds
in
the
pasture.
36
ended
the
grazing
season
in
November,
3
in
October
and
7
in
December.
The
other
(14)
farmers
had
variable
seasons,
depending
on
the
end
and
start
of
snow
cover
in
their
pasture.
For
each
category
of
farms
(with
and
without
wolf
attacks)
16
farms
had
sheep
breed
JSR
(Improved
Jezersko
Solˇ
cava)
only,
7
had
JS
sheep
(Jezersko
Solˇ
cava),
and
the
remaining
7
had
other
breeds
or
a
mix
of
breeds.
Lambs
were
present
during
the
pasture
season
in
almost
all
farms
(Table
1).
Median
area
of
a
fenced
pasture
was
2.7
ha
and
median
herd
size
was
93
sheep.
Median
density
was
23.3
sheep
per
ha.
There
was
no
statistical
difference
between
farmers
with
or
without
problems
with
wolves
in
the
size
of
fenced
area,
herd
size
or
density
(Mann–Whitney
U
test:
P
=
0.50,
P
=
0.10
and
P
=
0.67,
respectively).
Average
num-
ber
of
attacks
per
year
on
a
farm
did
not
correlate
with
herd
size
(rs=
0.28;
P
=
0.18;
n
=
25)
nor
with
sheep
den-
sity
(rs=
0.34;
P
=
0.11;
n
=
23).
Number
of
sheep
killed
per
attack,
however,
tended
to
correlate
positively
with
herd
size
(rs=
0.40;
P
=
0.05;
n
=
25),
and
with
density
of
sheep
(rs=
0.37;
P
=
0.08;
n
=
23).
Seventeen
out
of
60
farmers
kept
other
livestock
(goats,
horses
or
cattle)
amongst
the
sheep
(Table
1).
Five
of
them
(4
with
goats)
had
no
problems
with
wolves,
but
12
suffered

50 D.
van
Liere
et
al.
/
Applied
Animal
Behaviour
Science
144 (2013) 46–
56
Table
1
Characteristics
of
the
Slovene
sheep
farms
investigated
(n
=
60)
showing
representative
observations
where
there
were
no
differences
between
farms
(middle
column;
IQR
is
used
together
with
median
values,
sd
with
averages),
and
!2comparisons
of
numbers
of
farms
without
and
with
wolf
attacks:
ns:
not
significant;
(*):
0.1
<
P
≤
0.05;
*:
P
<
0.05;
**:
P
<
0.01;
***:
P
<
0.001.
If
!2is
significant,
it
applies
to
the
adjacent
2
×2
or
3
×2
counts.
Variable
Class
Representative
observations
for
all
60
farms
Number
of
farms
without
wolf
attacks
Number
of
farms
with
wolf
attacks
!2value
(df)
and
significance
Management pasture
season
starting
month
April
pasture
season
ending
month November
sheep
breed
JSR
and
JS
lambs
present
during
pasture
season 26 28
0.7
(1)
ns
herd
composition only
sheep 25 18 4.0
(1)
*
mixed
herds
5
12
area
of
fenced
pasture 2.7
ha
(IQR
1.5–7.3)
herd
size
93
sheep
(IQR
55–140)
density 23.3
sheep/ha
(IQR
11.0–57.5)
night
facilities open
barn
or
no
night
facilities
at
all
15 24 5.9
(1)
*
closed
barn
or
separately
fenced
area
15 6
time
of
driving
sheep
for
the
night driving
at
dusk
or
no
driving 17 27
8.5
(1)
**
driving
before
dusk 13 3
time
of
driving
sheep
from
night
enclosure
back
to
the
pasture
after
sunrise
Deterrents guarding
dogs
16
13
0.6
(1)
ns
guarding
donkeys 3
4
0.2
(1)
ns
visiting visiting
twice
a
day 16 17 0.1
(2)
ns
acoustic,
visual,
chemical
or
olfactory
deterrents
none
used
25
17
7.1
(2)
*
indicating
potential
human
presence
1
8
not
indicating
human
presence
4
5
openings
in
fence
16
14
0.3
(1)
ns
type
of
fence
wiring horizontal
single
wires 4 11 5.1
(1)
*
wire-mesh
26
17
fence
height
115
cm
(sd
17)
type
of
fence electric 10 24 16.6
(2)
***
physical 12 3
electric
combined
with
physical
8
1
Potentially
appealing
cues
predominantly
white
sheep 26 26 0
(1)
ns
no
synchronizing
lamb
delivery
23
26
1.0
(1)
ns
no
disposal
of
placenta
22
20
0.3
(1)
ns
slaughter
remains
not
to
hygiene
service 10 12 0.3
(1)
ns
sheep
at
night
reported
noisy
17
9
3.9
(1)
*
not
reported
noisy
13
21
sheep
during
shearing reported
noisy 14 4 7.1
(1)
**
not
reported
noisy 16
26
Pasture
types altitude
538
m
(sd
107)
open
pastures 12 7 1.9
(1)
ns
immediately
bordering
woods
27
26
0.2
(1)
ns
main
wind
direction
towards
woods
yes
11
14
3.2
(1)
(*)
no 16 7

D.
van
Liere
et
al.
/
Applied
Animal
Behaviour
Science
144 (2013) 46–
56 51
from
wolf
attacks
(9
with
goats).
Thus,
herds
with
mixed
species
were
significantly
associated
with
wolf
attacks.
The
9
farmers
with
mixed
flocks
of
sheep
and
goats
had
a
total
of
94
attacks
in
the
period
2008–2010.
In
28%
of
these
attacks
not
only
sheep,
but
also
goats
were
killed.
In
such
an
attack,
wolves
killed
2
goats
(median;
IQR
1–3),
2.5
adult
sheep
(IQR
2–4)
and
2
lambs
(IQR
1–3).
These
numbers
of
adult
sheep
kills
and
lamb
kills
were
similar
to
those
from
attacks
without
goat
kills
at
the
same
farms
(related
samples
Wilcoxon
signed
rank
test:
P
=
0.94
and
P
=
0.13,
respectively).
Farms
with
wolf
attacks
more
often
used
open
night
barns
or
no
night
enclosure
at
all
than
farmers
without
attacks
(Table
1).
Moreover,
the
timing
of
driving
sheep
to
night
facilities
differed
between
farmers
with
and
without
wolf
attacks:
farmers
without
attacks
were
significantly
more
likely
to
drive
their
sheep
before
dusk
(period
of
around
two
hours
after
sunset
when
objects
are
still
distin-
guishable
to
the
naked
human
eye)
to
the
night
enclosure
(Table
1).
Sheep
were
driven
back
to
the
pasture
the
fol-
lowing
day
after
sunrise.
3.2.2.
Deterrents
Farmers
with
guarding
dogs
mostly
had
2
dogs
together
with
sheep.
These
were
all
guarding
breeds
(mostly
Pyre-
nean
mountain
dog,
Hungarian
Kuvasz
or
Tornjak).
All
dogs
had
been
raised
from
2
months
of
age
together
with
sheep.
There
was
no
difference
between
farmers
with
or
with-
out
wolf
problems
in
the
use
of
dogs:
in
both
categories
about
half
of
the
farmers
used
guarding
dogs
(Table
1).
Of
the
30
farmers
with
wolf
attacks,
there
was
no
differ-
ence
between
farmers
with
dogs
and
those
without
dogs
in
the
number
of
animals
killed
per
attack
(median
number
of
animals
killed
(IQR):
2
with
dogs
(1–8),
n
=
13
farms;
4
without
dogs
(3–5.5),
n
=
17
farms;
Mann–Whitney
U
test:
P
=
0.74).
Furthermore,
4
farmers
with
wolf
attacks
and
3
farmers
without
attacks
had
guarding
donkeys.
Presence
of
donkeys
did
not
affect
the
number
of
sheep
killed
per
attack
(Mann–Whitney
U
test:
P
=
0.39).
Besides
having
dogs
or
donkeys
with
the
herd,
farmers
could
also
frequently
visit
the
pasture
or
apply
deterrents
against
wolves.
However,
visits
categorized
as
‘not
visiting
or
once
a
day’,
‘visiting
twice
a
day’
and
‘visiting
3
times
a
day
or
daily
spending
at
the
least
4
h
with
the
herd’
did
not
differ
between
farms
with
or
without
attacks
and
major-
ity
visited
twice
a
day
(Table
1).
Eighteen
farmers
used
(forms
of)
deterrents.
Deterring
against
wolves
involved
applying
radios,
flash
lights,
chemicals
against
deer
or
wild
boar,
gas
canons
and
fire
spots,
or
human
hair,
meat
in
cans,
noisy
cans
or
CD’s
attached
to
the
fence,
as
well
as
checking
the
pasture
during
the
night
by
car
or
tractor.
We
categorized
these
measures
in
those
clearly
indicat-
ing
potential
human
presence
(checks
in
the
night
by
car,
radio
playing,
human
hair
attached
to
the
fence)
and
those
that
did
not.
It
showed
that
farmers
with
wolf
problems
more
often
used
deterrents
indicating
human
presence
than
other
deterrents
(Table
1)
and
that
the
application
of
deterrents
did
not
differ
between
farmers
that
visited
the
pasture
twice
daily
or
more,
and
those
that
visited
the
pas-
ture
less
(!2(2) =
0.71;
P
=
0.70).
Five
out
of
8
farmers
with
wolf
problems
using
deterrents
indicating
human
presence
mentioned
that
attacks
occurred
at
locations
just
out
of
reach
of
sounds
or
lights
of
these
deterrents,
or
occurred
shortly
after
these
applications
had
stopped.
All
farmers
used
fences
around
the
area
where
the
sheep
grazed,
except
for
2
(who
had
wolf
attacks).
The
average
height
of
the
fence
surrounding
the
representative
pasture
or
paddock
was
115
cm
(sd
17)
and
did
not
differ
between
farms
with
or
without
wolf
attacks
(t-test,
t
=
0.05;
P
=
0.96).
However,
farmers
differed
in
the
type
of
fences
they
used
(Table
1):
most
farmers
having
wolf
problems
had
electric
fences,
whereas
most
farmers
without
wolf
problems
used
physical
fences
or
a
combination
of
electric
and
physical
fences.
Farmers
stated
that
they
applied
standard
equip-
ment,
like
from
the
Gallagher
company,
and
a
voltage
of
at
least
3500
V
to
their
electric
fence,
as
was
recommended
(Vidrih
and
Vidrih,
2009).
Five
of
10
non-attacked
farms
and
10
of
24
attacked
farms
(!2(1) =
0.2;
P
=
0.66)
using
electric
fences,
applied
mobile
electric
fences.
The
wiring
of
the
fences
(regardless
of
use
of
electricity)
differed
(Table
1):
farmers
without
wolf
attacks
strongly
preferred
wire-mesh
compared
to
farmers
with
wolf
attacks
and
hardly
used
fences
with
single
horizontal
wires.
Neverthe-
less,
at
around
half
of
the
farms
in
both
categories,
openings
were
present
in
and
under
fences
that
could
provide
wolves
opportunity
to
get
past
them
(Table
1).
3.2.3.
Presence
of
appealing
cues
We
checked
for
presence
of
fence
sounds,
sound
devices
attached
to
sheep
or
other
livestock,
such
as
bells,
and
the
farmer’s
judgement
of
whether
or
not
sheep
would
emit
sounds
when
the
farmer
arrived
or
drove
the
sheep,
whether
he
had
noisy
barking
dogs
or
braying
donkeys.
Both
categories
of
farms
did
not
differ
in
the
presence
of
noisy
objects
or
animals.
However,
farmers
without
attacks
reported
significantly
more
that
sheep
were
noisy
dur-
ing
the
night
than
farmers
with
attacks
(Table
1).
Farmers
without
attacks
also
reported
significantly
more
that
sheep
could
be
noisy
during
shearing
(Table
1).
There
were
no
differences
between
farmers
with
regard
to
presence
of
potential
visual
attractants.
This
applied
to
colour
of
the
sheep,
which
was
predominantly
white
(in
52
farms),
yellow
ear
tags
(in
55
farms),
lights
that
had
been
applied
to
the
fences
or
near
the
flock
(14
farms)
or
cisterns
in
contrasting
colours,
like
white
or
blue
(24
farms).
There
were
also
no
differences
in
olfactory
cues
related
to
farming
practises
between
farms
without
or
with
wolf
attacks.
Most
farms
did
not
synchronize
delivery
of
lambs
and
had
pregnant
ewes
during
the
pasture
season
(Table
1).
After
birth,
the
ewe’s
placenta
remained
on
the
pasture
in
most
farms,
but
18
farmers
located
the
placenta
and
dis-
posed
of
it.
There
was
also
no
difference
between
farms
with
or
without
wolf
attacks
in
the
disposal
of
slaughtered
sheep:
most
(38)
provided
the
remains
to
a
hygiene
service
(Table
1).
The
remains
could
also
be
buried,
fed
to
the
dogs,
or
left
outside
(which
could
also
be
the
pasture).
3.2.4.
Pasture
types
Pasture
altitude
averaged
538
m
above
sea
level
and
did
not
differ
between
farms
(t-test,
t
=
0.93;
P
=
0.36).
Farms
with
or
without
attacks
also
did
not
differ
in
openness:
one
third
of
all
60
farms
had
open
pastures
(Table
1).

52 D.
van
Liere
et
al.
/
Applied
Animal
Behaviour
Science
144 (2013) 46–
56
However,
farms
with
mixed
herds
with
goats
that
had
wolf
attacks,
more
often
had
non-open
pastures
than
open
pastures
(9
versus
0),
which
tended
to
differ
from
those
with
only
sheep
(11
versus
6;
Fisher
exact
probability
test:
P
=
0.05).
Distance
to
woods
did
not
differ
between
farms
(Mann–Whitney
U
test:
P
=
0.71):
woods
were
in
all
but
7
cases
immediately
bordering
pastures.
Attacked
farms
were
more
likely
to
have
wind
directions
towards
adjacent
woods
than
non-attacked
farms
(Table
1).
Of
21
farmers
with
wolf
attacks
that
had
indicated
main
wind
directions,
all
6
farms
that
had
sheep-goat
herds,
had
pastures
with
the
main
wind
towards
woods.
In
15
attacked
farms
with
herds
only
of
sheep
this
applied
to
8
farms
(Fisher
exact
probability
test:
P
=
0.06).
4.
Discussion
Farms
with
sheep
in
closed
night
barns
or
a
separately
fenced
night-area
have
less
risk
of
attacks
by
wolves,
as
shown
in
this
study.
We
have
also
shown
that
chances
of
attacks
are
reduced
if
sheep
are
moved
to
these
night
facil-
ities
before
dusk.
Protected
housing
of
sheep
during
the
night
is
important,
as
in
this
study
78%
of
wolf
attacks
were
at
night-time.
These
findings
correspond
with
the
general
view
that
wolves
are
active
during
the
night
(Kusak
et
al.,
2005;
Okarma,
1997;
Stahler
et
al.,
2006)
and
the
recommendation
to
adopt
night
enclosures
(Boitani,
2000;
Espuno
et
al.,
2004;
Oberle,
2010;
Plan
d’action
national
sur
le
loup,
2008;
Reinhardt
and
Kluth,
2007; ˇ
Strbenac
et
al.,
2010).
However,
we
could
not
find
evidence
with
Slovene
sheep
farmers
that
electric
fences
were
also
effective,
although
this
measure
is
recommended
in
general
(Boitani,
2000;
Espuno
et
al.,
2004;
Oberle,
2010;
Plan
d’action
national
sur
le
loup,
2008;
Reinhardt
and
Kluth,
2007;
ˇ
Strbenac
et
al.,
2010)
as
in
Slovenia
(Vidrih
and
Vidrih,
2009).
Slovene
authorities
recommend
farmers
apply
elec-
tric
fencing,
otherwise
they
will
not
be
compensated
for
their
losses.
Not
surprisingly
therefore,
we
found
that
most
farmers
with
wolf
attacks
applied
electric
fences.
This
also
applied
for
18
out
of
22
farmers
that
had
experience
with
wolf
attacks
for
more
than
3
years
and
had
electric
fences
during
our
observation
period.
However,
in
contrast
to
what
would
be
expected,
electric
fences
did
not
reduce
attack
rates
when
wolf
attacks
were
persistently
recurring
at
farms.
Moreover,
farms
that
were
newly
confronted
with
wolf
attacks
(whether
or
not
they
had
electric
fences)
had
low
annual
killing
rates
compared
to
rates
at
farms
that
had
electric
fences
and
wolf
attacks
for
more
than
3
years.
Electric
fences
as
used
in
our
study
area
are
therefore
inef-
fective
measures
against
repeated
wolf
attacks.
Moreover,
low
annual
killing
rates
at
new
farms
need
not
relate
to
electric
fences,
but
may
relate
to
wolves
generally
avoid-
ing
unfamiliar
sites
and
returning
to
familiar
farms
(Boitani,
2000; ˇ
Cerne
et
al.,
2010;
Gazzola
et
al.,
2008;
Nationale
Koordination
Herdenschutz,
2011;
Rigg
et
al.,
2011).
Several
reasons
may
explain
why
electric
fences
were
not
found
to
be
effective
in
the
Slovene
practise,
even
if
they
are
used
several
years.
Average
fence
height
of
115
cm,
as
we
found,
is
well
below
minimum
height
of
140
cm
of
fixed
electric
fences
that
were
tested
to
protect
against
depredation
(Acorn
and
Dorrance,
1994;
Gates
et
al.,
1978;
Linhart
et
al.,
1982;
Thompson,
1979).
Although
these
stud-
ies
concerned
coyotes
and
fixed
fences,
the
features
of
these
fences
are
adopted
and
recommended
to
protect
against
wolves
too
(Haviernick,
1998;
Paul
and
Gipson,
1994;
Shivik,
2004;
Vidrih,
2002).
Indeed,
some
farmers
in
our
research
saw
wolves
jump
over
fences
of
145
cm.
So
even
140
cm
is
not
high
enough
for
wolves
in
all
cases.
Wam
(2004a,
2004b)
found
traditional
fixed
fences
of
mesh
wire,
100
cm
high,
more
effective
against
wolves
if
improved
by
applying
electricity
and
a
height
up
to
160
cm.
Mertens
et
al.
(2002)
and
Cortés
(2007)
found
that
mobile
electric
fences
of
150
cm
were
effective
against
wolves.
Although
it
is
assumed
that
wolves
crawl
under
fences,
we
did
not
find
that
attacked
pastures
provided
more
crawling
opportunities,
as
half
of
the
farms,
both
in
attacked
and
in
non-attacked
farms,
had
openings
in
fences
that
wolves
could
use
to
crawl
through.
Coyotes
can
also
jump
between
wires
or
climb
fence
corners,
using
horizontal
corner
braces
as
toe-holds,
instead
of
crawling
under
or
jumping
over
(Acorn
and
Dorrance,
1994;
Gates
et
al.,
1978;
Linhart
et
al.,
1982;
Thompson,
1979).
There
is
no
detailed
research
into
how
wolves
pass
fences
and
therefore
it
is
not
known
to
what
extent
wolves
also
jump
between
wires
or
climb
electric
fences.
If
they
do,
wolves
in
the
Slovene
situation
are
likely
not
receiving
any
shock.
The
reason
is
that
live
wires
in
our
research
were
never
alternated
with
grounded
wires.
Live
and
ground
wire
alternation
was
applied
by
Gates
et
al.
(1978),
one
of
the
first
in
establishing
effective-
ness
of
electric
fences.
Such
alternation
is
recommended
in
Slovenia
to
guarantee
closing
of
the
electrical
circuit
when
the
animal
touches
the
fence,
and
the
dry
soil
is
not
conducting
well
(Vidrih
and
Vidrih,
2009).
Indeed,
the
karst
area
is
particularly
known
for
dry
perennial
grass-
lands
(Rural
Development
Programme,
2007).
Therefore,
fences
with
the
same
polarity
at
the
electric
wires
will
not
shock
a
wolf,
if
it
would
jump
between
these
wires.
A
wolf
may
not
even
get
a
shock
at
all
in
dry
conditions.
Similar
features
would
apply
for
wolves
climbing
a
fence.
In
particular,
mobile
fences
can
only
be
partly
stretched,
for
instance
on
uneven,
rough
terrains
as
in
the
karst,
and,
again,
all
electric
wires
have
the
same
polarity.
Wolves
may
learn
to
jump
onto
these
fences,
use
the
wires
as
toe-holds,
and
climb
over
without
being
shocked.
Furthermore,
it
is
generally
emphasized
that
electric
fences
require
regular
maintenance
and
this
requirement
may
not
have
been
met.
However,
we
have
no
data
about
fence
maintenance.
In
conclusion,
electric
fences
in
this
study
did
not
provide
a
noteworthy
physical
threshold,
whereas
polarity
of
their
live
wires
was
not
alternating,
and
electric
grounding
was
difficult
in
the
dry
karst,
which
likely
did
not
produce
a
sys-
tematic
aversive
experience
by
shocks
to
wolves.
Thereby,
we
assume
that
after
novelty
of
putting
up
a
fence
had
faded,
electric
fences
became
ineffective
tools.
In
order
to
improve
fences
and
application
of
electricity,
comparative
research
needs
to
be
done
of
what
wolves
actually
do
at
typical
entry
points
e.g.
with
stealth
and
infrared
cameras.
It
is
also
generally
assumed
that
a
guarding
dog
pro-
tects
the
herd
from
wolf
attacks
(Boitani,
2000;
Mettler,
2005;
Reinhardt
and
Kluth,
2007;
Rigg
et
al.,
2011; ˇ
Strbenac
et
al.,
2010).
However,
we
did
not
find
a
protective
effect
of