Content uploaded by Christopher R. Dickman
Author content
All content in this area was uploaded by Christopher R. Dickman on Nov 29, 2020
Content may be subject to copyright.
Effects of low-level culling of feral cats in open populations:
a case study from the forests of southern Tasmania
Billie T. Lazenby
A,B,D
, Nicholas J. Mooney
C
and Christopher R. Dickman
A
A
School of Biological Sciences, A08, University of Sydney, NSW 2006, Australia.
B
Department of Primary Industries, Parks, Water and Environment, 134 Macquarie Street, Hobart,
Tas. 7000, Australia.
C
PO Box 120, Richmond, Tas. 7025, Australia.
D
Corresponding author. Email: Billie.Lazenby@dpipwe.tas.gov.au
Abstract
Context. Feral cats (Felis catus) threaten biodiversity in many parts of the world, including Australia. Low-level culling is
often used to reduce their impact, but in open cat populations the effectiveness of culling is uncertain. This is partly because
options for assessing this management action have been restricted to estimating cat activity rather than abundance.
Aims. We measured the response, including relative abundance, of feral cats to a 13-month pulse of low-level culling in
two open sites in southern Tasmania.
Methods. To do this we used remote cameras and our analysis included identification of individual feral cats. We
compared estimates of relative abundance obtained via capture–mark–recapture and minimum numbers known to be alive,
and estimates of activity obtained using probability of detection and general index methods, pre- and post-culling. We also
compared trends in cat activity and abundance over the same time period at two further sites where culling was not conducted.
Key results. Contrary to expectation, the relative abundance and activity of feral cats increased in the cull-sites, even
though the numbers of cats captured per unit effort during the culling period declined. Increases in minimum numbers of cats
known to be alive ranged from 75% to 211% during the culling period, compared with pre- and post-cull estimates, and
probably occurred due to influxes of new individuals after dominant resident cats were removed.
Conclusions. Our results showed that low-level ad hoc culling of feral cats can have unwanted and unexpected outcomes,
and confirmed the importance of monitoring if such management actions are implemented.
Implications. If culling is used to reduce cat impacts in open populations, it should be as part of a multi-faceted approach
and may need to be strategic, systematic and ongoing if it is to be effective.
Received 12 February 2014, accepted 18 October 2014, published online 20 February 2015
Introduction
The introduction and establishment of feral cats (Felis catus)on
islands has caused the extinction of several animal taxa (Nogales
et al.2004,2013) and, although their impact in open systems is
less clear (Dickman 1996), feral cats are nevertheless recognised
as an important threat to biodiversity in many parts of the world.
In Australia, they are formally recognised as a key threat to
the country’s biodiversity under the Environment Protection
and Biodiversity Conservation Act 1999. The documentation
associated with this recognition nominates 35 birds, 36
mammals, seven reptiles, and three amphibians that are under
threat of extinction and that are thought to be adversely affected by
feral cats (DEWHA 2008). Cats have been estimated to kill many
billions of small mammals and birds each year in the United States
(Loss et al.2013), and are recognised as an extinction driver for
at least 33 species of endemic insular vertebrates in other parts of
the world (Nogales et al.2013).
Feral cats have been successfully eradicated from some islands
and virtual islands (e.g. within fenced exclosures), resulting in
positive population responses for many species of small and
medium-sized mammals, birds and reptiles (Campbell et al.
2011). However, eradication of feral cats from open systems
has seldom been achieved. Open systems have more potential
for re-invasion by cats, usually contain a greater diversity of
non-target species that may complicate control actions, and
pose practical difficulties owing to the large areas that must be
covered (Denny and Dickman 2010; Dickman et al.2010). As a
result, control of feral cats in such systems is often undertaken
by the continuous removal of individuals via culling. This places
downward pressure on the cat populations and may alleviate,
but does not remove, the predation threat and impact on prey
species.
Strategic culling (Braysher 1993; Olsen 1998) is often
recognised as a sound conceptual approach to mitigate the
impacts of target species (Hone 2007), but may require long-
term effort using multi-faceted control methods (Dickman et al.
2010). Such effort is likely to be particularly intensive for feral
cats, owing to their cryptic and wary behaviour. In addition, there
CSIRO PUBLISHING
Wildlife Research,2014, 41, 407–420
http://dx.doi.org/10.1071/WR14030
Journal compilation CSIRO 2015 www.publish.csiro.au/journals/wr
are limited options for effectively monitoring either the relative
or absolute numbers of feral cats, the relationship between
abundance and impact, and therefore the level of control
necessary to successfully reduce their impact (Fisher et al.
2001; Reddiex and Forsyth 2004; Reddiex et al.2004; Denny
and Dickman 2010). These difficulties might partly explain
why feral cats are the least-frequently controlled vertebrate
pest species in Australia (Reddiex et al.2006), despite their
recognition as a key threat to the country’s biodiversity.
Past studies have measured trends in feral cat populations by
using activity or presumed relative abundance estimates that do
not require identification of individuals derived from methods
such as sand-padding and spotlight counts. One of the limitations
of such methods is that it is often unknown how they relate
to abundance or density or, in other words, to the number of
individuals within a defined area. Interpretation of activity
and index estimates often requires the assumption that the
relationship between them and abundance is positive and
linear, which may not always be the case (Forsyth et al.2005).
The apparent fast-learning ability and neophobic tendencies of
feral cats mean that traditional live-trap, tag and recapture
techniques are generally unsuitable, especially when they are
used by inexperienced operators, and that less invasive and
more humane techniques for monitoring abundance are
required. This consideration applies to many carnivores (Long
et al.2008). Two non-invasive methods are emerging as potential
tools for effectively monitoring feral cat abundance, namely,
identification of individuals on the basis of collection of hair
samples and analysis of genetic markers (Hanke and Dickman
2013), and cameras that are sensitive to infrared heat and motion
(referred to henceforth as remote cameras) (Forsyth et al.2005;
Robley et al.2010). Both methods have existed for some time;
however, recent advances in technology have increased their
potential effectiveness and cost accessibility, allowing for more
widespread deployment.
Remote cameras have been used to successfully estimate the
abundance of a range of wild felines including tigers, Panthera
tigris (Karanth and Nichols 1998), pumas, Puma concolor (Kelly
et al.2008), and snow leopards, Uncia uncia (Jackson et al.
2006). Few trials have assessed the suitability of remote cameras
for monitoring the abundance and population trends of feral cats
in Australia; however, those that have been conducted, have
yielded promising results. Robley et al.(2010) were able to detect
a difference in occupancy rates using remote cameras between
two different habitats, and Bengsen et al.(2011) were able to
measure a reduction in feral cat numbers following a culling
operation in a closed population.
In the present study, we focus on one aspect of the
conceptual framework underpinning effective feral-cat control,
namely, the population response of feral cats to culling. The
culling intensity we employed was low level to simulate the
resource-effort that typically might be available to and expended
by natural resource managers. We systematically deployed
remote cameras at four sites, twice a year, for 3 years. Two
sites were experimental sites where we conducted a 13-month
pulse of low-level culling, with one site being a control, and
the other an observational site. We developed a system for
identifying individual feral cats on the basis of coat colours
and patterns to produce estimates of relative abundance
using capture–mark–recapture and minimum-known-to-be-
alive estimators. We compared these estimates to indexes
of activity derived from the same dataset, using probability of
detection and general index estimators. Our specific aims were to
measure the effect of a 13-month pulse of low-level culling on the
relative abundance of feral cats, and to compare this to indexes of
activity. We hypothesised that our culling efforts would result in a
temporary reduction in feral cat numbers at the two experimental
sites. In addition, we hypothesised that index of activity estimates
would reasonably reflect our relative abundance estimates at the
low densities of feral cats that we were dealing with.
Materials and methods
Study sites and remote camera surveys
Four study sites, referred to henceforth as South-west, Mount
Field, Tasman Peninsula and Wellington Ranges, were selected in
southern Tasmania between the 51 and 55 northing grid on the
GDA 94 map datum based on flora, fauna and land use (Fig. 1).
The central coordinate for each site is South-west –4248028.800,
14622026.400; Mount Field –4239046.800 , 14641016.800; Tasman
Peninsula –433050.400, 1475403600 ; and Wellington Ranges
–4252044.400, 14714016.800 . The habitat at the first three of
these sites, namely, cool, dense, temperate forest and heath, is
described in detail elsewhere (Lazenby and Dickman 2013);
habitat in the Wellington Ranges was very similar to that in
the other sites. Cat culling was conducted at Mount Field and
Tasman Peninsula, commencing in July 2009 and running for
13 months. The South-west was a control (i.e. non-culling) site.
The Wellington Ranges site was set up initially to be a further
control, but became an observational site when we discovered that
Tasman Peninsula
Wellington Ranges
Mt Field
South-west
Fig. 1. Map of Tasmania showing study sites. Pulsed culling of feral cats
was conducted over 13 months between July 2009 and August 2010 at the
Mount Field and Tasman Peninsula sites; the South-west was a control
(non-culling site) and the Wellington Ranges an observational site. All
sites were monitored with remote cameras twice a year from 2009 to 2011
(excluding the Wellington Ranges which was monitored only once in 2009).
408 Wildlife Research B. T. Lazenby et al.
feral cat culling operations, that could not be accurately
quantified, were being conducted at two waste-management
centres that were within 2 km of the outer margins of this
study site.
Standardised remote-camera surveys were carried out at all
sites from 2009 to 2011. The protocol has been described in detail
by Lazenby and Dickman (2013) but, in brief, we deployed 15–18
DigitalEye7.2 trail cameras (Pixcontroller, Pittsburgh, PA) in
a systematic grid pattern 1–1.5 km apart at the four study sites,
allowing them to operate for 1–2 weeks per survey, commencing
in April and June in each of the three years of the study. The
distance between remote cameras was such that the same
individual cat could be recorded at multiple survey devices.
This had implications for closed capture–mark–recapture
analyses where multiple survey devices within a home range
are likely to increase individual capture probability and hence the
robustness of abundance estimates (Otis et al.1978). However,
lack of independence of survey devices violates one of the major
assumptions of site occupancy analyses (MacKenzie et al.2006).
The degree to which this assumption was violated was quantified
for feral cats. Given that all study sites were sampled with the
same remote-camera survey methodology, and we were using
multi-season models, relative estimates of site occupancy within
and among study sites were interpreted with due caution, but were
still considered relevant.
The cameras featured an infrared flash that is less likely to
disturb animals compared with visible white light flashes (Meek
et al.2012). They also had a passive infrared triggering system
that detects body heat and motion before triggering a photo.
We set the passive infrared sensitivity (PIR) switch at medium
(standard factory setting), and the switch control board to record
in ‘trail mode’so that photographs were taken at least once every
second after the PIR sensor had been activated by an animal. The
switch control board was also set such that the camera would
record photographs 24 h a day. A scent lure and food reward
consisting of Jurotuna emulsion (Juro Oz Pro tackle, Australia)
and fish-based tinned cat food in jelly was placed 1.5–2.0 m from
each camera unit for each survey. Two dessert spoons of tinned cat
food and 50–75 mL of tuna emulsion were spread in a 0.25-m
2
area that was the focal point for the camera. Tuna emulsion was
also squirted on one or two branches up to 2 m off the ground,
above the focal point of the camera, to maximise the chances of the
lure scent entering air streams. Additional tuna emulsion was
placed in a perforated film canister that was staked into the ground
after the first two standard camera surveys. Mid-trip checks were
conducted during surveys lasting more than a week (i.e. all the
2010 and 2011 surveys) to re-bait, and to replace camera batteries
and memory cards.
We did not deploy cameras in the Wellington Ranges in April
2009, resulting in 23 site surveys in total.
Data storage
We collected >58 500 photographs over the course of our surveys.
These were stored in an Access database, referred to hence as
the remote-camera surveys database. The database consisted of
six relational tables, and included information pertaining to the
species of animal photographed, surety of species identification,
24-h survey period, survey and site identification.
We classified an occurrence of a feral cat as a photograph or
set of consecutive photographs separated by no longer than
5 min. Each 5-min break in photographs was classed as a new
occurrence, unless a different individual could be clearly
discerned for breaks less than 5 min. Larger animals, including
feral cats, tended to have occurrence breaks in the order of hours
or days. On rare occasions where a ‘blank’photo was recorded
within an occurrence, it was not counted as a blank.
Sampling effort within each camera survey was broken down
into 24-h sampling periods, with sampling periods defined
as the 24 h between 1700 hours and 1700 hours the next day
(i.e. 5 p.m. –5 p.m.). Twenty-four-hour periods defined by
calendar dates were not used because calendar days change at
midnight, whereas a continuous night or day is likely to have most
biological significance to a wild animal. Although still arbitrary
and variable depending on the time of year and concomitant
length of day, 24-h periods defined by 1700 hours are towards the
end of one day and the beginning of a night.
Identification of individual feral cats
To identify individual cats, we recorded coat colours and
patterns by using the formally recognised descriptions used by
cat breeders and fanciers (Turner and Bateson 2000).
Classifications for coat colour were black, white, orange, grey
or cream; for pattern, they were solid, bicolour, tricolour and
tabby. Tabby variations included classic, mackerel, spotted,
ticked and tortoiseshell.
Inspection of individual markings, particularly in the foreleg
region, revealed consistent and recognisable differences among
different individuals with similar coat colours and patterns,
provided there were enough good-quality photographs for
comparison. In light of this, we designed a scoring system
similar to that used for dolphins and whales (Slooten et al.
1992; Friday et al.2000; Markowitz et al.2003), in which
each occurrence of a feral cat at a camera site was given a
within-session confidence score for accuracy of identification.
Within-session was defined as any occurrence within a survey
period (either 1–7or1–14 days) and was rated as low, medium or
high, using the following criteria:
Low: no distinct marks and/or other identifying features and/or
poor-quality photos and/or few photos and/or poor camera angle.
Medium: distinct marks or other identifying features despite
poor-quality photos and/or few photos and/or poor camera angle.
High: distinct marks and/or other identifying features, good-
quality photos, many photos from several angles.
Identifying features that were unlikely to change over the
course of 1–2 weeks were included, such as missing patches
of fur, body condition, and kitten-like features, including size.
We constructed a file for each individual cat that contained
photographs from different angles; these were used for
reference and added to each additional occurrence.
Our system for identifying individuals with low, medium or
high confidence could not discern the difference between two
individuals with exactly the same coat colours, patterns and other
identifying features such as body shape. We reasoned that the
chances of observing two such identical cats would be greatest
between siblings. To check similarities in sibling appearance, we
photographed the right- and left-hand sides of kittens from several
Feral-cat population response to low-level culling Wildlife Research 409
litters at the Hobart Cat Centre and then calculated the number
of siblings that we could discern in each litter by using our
identification system.
Feral-cat trapping
Feral cats were live-trapped and then euthanased at the Mount
Field and Tasman Peninsula study sites. Culling operations
commenced in July 2009, following the completion of camera
surveys in April and June at each of the 4 study sites. Feral-cat
trapping was conducted every 1–4 months until August 2010 at
the two culling sites.
The total area covered by live traps was 40 km
2
at Mount
Field and 45 km
2
on the Tasman Peninsula, but individual
trapping trips within the 13-month culling period covered
slightly smaller areas. Up to 46 traps were set at each site for
6–16 nights. Five different styles of trap were deployed, each with
unique characteristics. Trap styles were small and large Mascots
(wire-mesh cage with treadle trigger and down-swinging door),
tent and stubby (soft-shell, treadle trigger with snapping door)
(DPIPWE 2009), and Mersey box (wire cage, treadle trigger with
sliding door). The same proportion of each style of trap was
deployed at each site; although monthly trapping effort varied,
the overall trapping effort was similar between study sites. We
did not use padded leg-hold traps, given the high density of
non-target species, including Tasmanian devils (Sarcophilus
harrisii), spotted-tailed (Dasyurus maculatus) and eastern
quolls (D. viverrinus) within our study sites.
Trapping at Mount Field was a collaborative effort with
Tasmanian Parks and Wildlife Service staff based at the
Mount Field National Park. Traps were set in sheltered
positions, out of public sight, within easy access to gravel
roads and 4 4 tracks. Trap sites were selected on the basis of
availability of shelter, proximity to landscape junctions such
as forks in tracks, track–creek junctions, or vegetation–creek
junctions, and signs of feral cat activity such as scats, footprints or
live sightings. Where adequate natural shelter was not available
for the wire-cage traps, a hessian sack was placed over the back of
the cage and secured.
Traps were baited with different baits during each trapping
session, but the numbers of trapping sessions conducted with a
particular bait or bait combination were standardised between
both culling sites. Baits were fish-based tinned cat food and fish
oil, dried cat food and fish oil, lamb kidneys and lightly fried
chicken necks. Cat urine (collected from the opposite site, i.e.
Mount Field trap sites deployed urine collected from feral cats on
the Tasman Peninsula and vice versa) was sprinkled on the ground
at the trap sites near one of the above food lures for one of the
cat-trapping trips at each site. We minimised by-catch of devils
in wire-cage traps because of the potential for diseased devils
with late-stage tumours to suffer further damage to their faces
from abrasions. To do this, we set soft-shell traps in areas where
devils were known to frequent, and temporarily avoided setting
traps where there were fresh devil scats, tracks or photographs.
Traps were checked in the morning and afternoon each day and
baits were replaced every third day. Non-target animals were
released and dirty traps replaced with clean traps at each capture
site. Trapped cats were scanned for possible micro-chips and
investigated for signs of lactation. Micro-chipped cats were to be
returned to their owners, in line with Tasmania’s proposed cat-
management strategy (DPIPWE 2008). However, none was
trapped, perhaps owing to the remoteness of the study sites.
Similarly, no lactating females were trapped. When no micro-chip
or signs of lactation could be detected, feral cats were euthanased
by a single shot to the head from a 0.22 rifle using hollow point
ammunition. Euthanasia followed standard operating procedure
number CAT002-6 (Sharp and Saunders 2005). Rapid death
was confirmed by absence of rhythmic breathing and loss of
an eye protection reflex (or blink). Following euthanasia, cats
were weighed, body condition assessed, head width measured,
coat colour recorded and photographed, sex determined, age
estimated, digestive tracts removed, and tissue was collected
from the ear and stored in 70% ethanol.
Analyses
Capture per unit effort in traps
The capture per unit effort of feral cats in traps was calculated
by dividing the number of individuals captured per 100 trap-
nights for each trapping trip. We did not adjust our calculation of
trapping effort for sprung but empty traps because we checked
and reset, where necessary, our traps twice within each 24-h cycle.
Site occupancy
Site-occupancy matrices were generated using the data
collected from the remote-camera surveys. Matrices were
study-site specific and consisted of a column of camera
numbers and a series of columns representing each 24-h
survey period. The same classification of a 24-h sampling
period was used for both site occupancy and capture–mark–
recapture analyses, with a sampling period defined as the 24 h
between 1700 hours and 1700 hours the next day. Data were
ordered chronologically for each sampling period and each study
site survey. Twenty-four-hour sampling periods constituted
secondary sampling sessions, and different trips to the same
study site, i.e. study site surveys, were primary sampling
sessions (Donovan and Hines 2007). There were six primary
periods for each study site (i.e. two site surveys a year for each
of the Years 2009, 2010 and 2011), excluding the Wellington
Ranges where there was only one survey in 2009 and two for
each of 2010 and 2011. There were either 7 (2009) or 14 (2010
and 2011) secondary sampling sessions for each primary session.
Each site-occupancy matrix was constructed from an array
of 1s, 0s and ‘.’s, which denoted a species detection, non-
detection or camera failure, respectively, for each 24-h
sampling period at each camera. The data were copied from an
excel file into PRESENCE 4.0 (Hines 2006) and analysed
using single-species multi-season occupancy models. The
framework for investigating single-species multi-season data
in PRESENCE is a nested model set, ranging from a global
model of psi(season)p(day)gamma(season)epsilon(season) (i.e.
site occupancy varies with season or study-site survey,
probability of detection varies with each 24-h survey period,
probability of colonisation of a site varies with season, and
extinction probability varies with season) to the most
constrained model of psi(.)p(.)gamma(.)epsilon(.) (i.e. constant
site occupancy, probability of detection, colonisation and
extinction). Site-occupancy models can also incorporate
410 Wildlife Research B. T. Lazenby et al.
covariates such as habitat. However, we kept our site-occupancy
models as simple as possible because we were interested primarily
in recording any relative changes in probability of site occupancy,
detection, colonisation or extinction within sites between
successive study-site surveys.
We conducted two types of site-occupancy modelling. In the
first, we modelled the potential for seasonal changes in probability
of site occupancy, detection, colonisation or extinction as a direct
function of the study-site survey and then used the most supported
models to test for correlations with our other estimators (e.g.
general index, MKTBA and relative population abundance
derived from capture–mark–recapture). To do this, we limited
our analyses to parameterisation number two in PRESENCE,
because we wanted to directly estimate the probability of site
occupancy for each season, rather than deriving it from the initial
season (Donovan and Hines 2007). We then used PRESENCE
to estimate the probability of site occupancy, colonisation and
detection, under the most universally supported of these very
simple models.
Second, we tested specifically for the effect of culling
by incorporating culling as a covariate in our models. We
reasoned that other potentially important covariates such as
vegetation types associated with the cameras would remain
constant throughout the course of the study and therefore we
did not include them. Importantly, we looked at the probability of
detecting devils after each survey at each site in relation to feral
cats, given that devils can reduce the probability of detecting
feral cats at remote cameras (Lazenby and Dickman 2013).
Devils were observed too infrequently to estimate their
probability of detection in the Wellington Ranges; however,
there was no indication of marked changes in their estimated
probability of detection or site occupancy for each site survey for
the remaining three sites and, therefore, we did not include this as
a covariate in our models (Lazenby 2012). We modelled each site
separately so that we could maintain our focus on within-study
site changes rather than potential differences in the estimated
parameters between sites. Our site-occupancy analyses that
included culling as a covariate were limited to the South-west,
Mount Field and Tasman Peninsula because we could not
accurately quantify culling in the Wellington Ranges.
To include culling as a seasonal covariate, we specified a
design matrix where Seasons 3 and 4 (which corresponded to
the period when we were culling cats) were different from each
other, and different from Seasons 1, 2, 5 and 6 (which were set to
be equal and which corresponded to the period when we were not
culling cats). We used information-theoretic methods to compare
the relative fit of our models, using Akaike’s information criterion
scores corrected for small sample sizes (AIC
c
; Burnham and
Anderson 2002), which are one of the outputs from program
PRESENCE. Models with the lowest value were considered to be
the best fit for the data, and models within two AIC
c
points of the
lowest value were considered to be a reasonable fit for the data.
MKTBA
The minimum number of feral cats known to be alive
(MKTBA) was calculated by summing the total number of
individuals identified with medium to high confidence within
each study-site survey. The density of feral cats was calculated by
dividing the MKTBA estimate of cats by the effective area; that is,
the area sampled by allowing for cats with a home range on the
outer margins of each study site (Hayne 1949). Effective areas
were calculated by placing a minimum convex polygon around
the camera locations and buffering them by 1 km. The effective
areas for each study site were as follows: South-west 117 km
2
,
Mount Field 68 km
2
, Tasman Peninsula 69 km
2
and Wellington
Ranges 57 km
2
.
Capture–mark–recapture estimates
Closed capture–mark–recapture analyses were used to estimate
the abundance of feral cats at each site for each survey. We did not
attempt to model individual covariates for the probability of first
capture, or the probability of recapture given our small dataset,
and therefore we used full likelihood models where abundance
is included in the likelihood of capture, as implemented in
program CAPTURE (Otis et al.1978; Rexstad and Burnham
1991), as opposed to a Huggins-analysis approach, which can be
implemented in MARK (Huggins 1989; White and Burnham
1999). Program CAPTURE was used to assist in the selection of
an appropriate model, to conduct statistical tests for population
closure based on the experimental data, and to generate
population estimates and average capture probabilities for feral
cats for each study-site survey (Otis et al.1978; Rexstad and
Burnham 1991). The recapture matrices for individual feral cats
were constructed using individual occurrence data for 24-h
survey periods extracted from the remote-camera survey
database. Only feral-cat occurrences that could be attributed to
an individual with medium to high confidence were included in
the analyses. There were too few movements of individual feral
cats among different cameras within the same session, to conduct
spatially explicit capture–recapture analyses.
Model selection in CAPTURE indicated that the null model,
M(0), was the single-best fitting model for 13 of the 20 study-site
surveys where there were sufficient data. The second-most
supported model was M(h) jackknife (hereafter M(h)), which
was selected with the M(0) model for 6 of 20 surveys. In all, 3 of
the total 23 study-site surveys that were conducted did not yield
sufficient data for model selection testing or any other capture–
mark–recapture analyses. We subsequently used the jackknife
version of model M(h), which accounts for individual
heterogeneity in capture probability (a feature common to
carnivores, particularly felids; e.g. Karanth and Nichols 1998),
and produces estimates that are a linear function of recorded
capture frequencies (Otis et al.1978) to estimate population
abundance. This model performs well for small sample sizes
(Otis et al.1978; Pollock and Otto 1983), has been used in past
studies of carnivores, and was supported by our observations of
individual heterogeneity in capture probability during the culling
phase of the project.
General index
General index (GI) estimates of activity were calculated using a
modified form of the Allen index for remote-camera data. The
mean numbers of hourly occurrences of feral cats per day for each
camera site were calculated, and the GI was calculated by taking
the mean of the daily means, using the following equation from
Engeman (2005):
Feral-cat population response to low-level culling Wildlife Research 411
GI ¼1
dX
d
j¼1
1
sj X
sj
i¼1
xij;
where d= days, s= stations, and x
ij
is the measurement from the
ith station on the jth day.
The associated variance was estimated using a linear
mixed-effects model in SPSS 20 (SPSS, Chicago, IL, USA)
(Engeman 2005). Three components of variability (camera-to-
camera variability, daily variability and random observational
variability) associated with each station each day were calculated
using the VARCOMP procedure and the following syntax:
VARCOMP Activity BY Station Day
/RANDOM = Station Day
/METHOD = REML
/CRITERIA = ITERATE(50)
/CRITERIA = CONVERGE(1.0E-8)
/DESIGn = Day Station
/INTERCEPT = INCLUDE.
Comparisons between population abundance
and activity estimators
Associations between MKTBA, capture–mark–recapture, GI and
probability of detection across all study-site surveys and sites
were explored using Pearson’s correlations. Observations and
tests for linearity, normality and outliers were conducted before
correlation analyses, by constructing scatter plots of the
estimates generated using each estimator against the other,
Q–Q plots and Shapiro–Wilk tests for normality. Capture–
mark–recapture estimates generated from study-site surveys
where the estimated individual capture probability was <0.1
were excluded from the analyses because they were likely to
be unreliable (Otis et al.1978).
Results
There were 353 feral-cat occurrences (i.e. records of a cat visiting
a camera, regardless of individual identity) at cameras from
2009–2011 across the four study sites from over 4600 camera
trap-nights; 335 (95%) of these could be attributed with medium
to high within-session confidence to 86 individuals. Overall, 14%
of the occurrences that were attributable with medium to high
confidence to an individual feral cat were recorded at more than
one camera site within a session. Twenty-six feral cats were
trapped and euthanased; 10 from Mount Field after 1319 trap-
nights, and 16 from the Tasman Peninsula after 1445 trap-nights.
These data form the basis for the subsequent results.
Identification of individuals
The minimum number of individual feral cats identified within
each session ranged from 1 to 12. Most feral cats could be
identified from photos with medium to high within-session
confidence, with just 4–6% (n= 18 of 353) falling into the
low-confidence category across all study sites and surveys. Of
the 18 occurrences with low within-session confidence, six
were classified as solid black cats and one as a black and grey
mackerel; the colour or pattern on the remaining 11 cats could not
be ascertained. Nine of these latter occurrences consisted of only
one photograph taken at a poor angle, whereas the remaining were
either poor-quality photos resulting from condensation on the
lens, limited number of photos and/or poor angle, or the distance
from camera to the cat was too great to ascertain detail. The 18
low-confidence occurrences were not used in any analyses
requiring identification of individuals.
Individuals from six of seven litters that we photographed at
the Hobart Cat Centre displayed distinct coat-colour and -pattern
differences compared with their siblings. There were no clear
differences between the siblings of one litter that consisted of
two kittens; however, this litter was young (4–6 weeks of age),
and the pelage had substantial amounts of longer grey hairs that
obscured the pattern underneath. It is common for younger kittens
to have this fluffy appearance, which disappears as they become
older.
Feral-cat culling
The number of feral cats physically captured per 100 trap-nights
within each trapping session ranged from 0 to 2.6, and decreased
over the 13-month course of trapping at each site (Fig. 2). The
percentages of culled feral cats that were recognisable as
individuals photographed in previous survey or pilot sessions
were 44% and 56% for the Mount Field and Tasman Peninsula
sites, respectively.
0
0.5
1.0
1.5
2.0
2.5
3.0 (a)
(b)
Jul 09 Aug 09 Oct 09 Dec 09 Mar 10 May 10 Jul 10
0
0.5
1.0
1.5
2.0
2.5
3.0
Aug 09 Sep 09 Jan 10 Aug 10
No. cats per 100 trap nights
Month and year of trapping expedition
Fig. 2. The number of feral cats captured in cage and soft-shell traps per 100
trap-nights during trapping sessions ranging from 6–16 nights with up to 46
traps at (a) Mount Field (total of seven expeditions) and (b) Tasman Peninsula
(total of four expeditions) study sites.
412 Wildlife Research B. T. Lazenby et al.
Culling operations were conducted near the Wellington
Ranges study site by the Hobart and Glenorchy City Councils
as part of ongoing feral-cat control at the McCrobies Gully and
Jackson Street Landfill Centres. Over 100 cats were removed
from the McCrobies Gully site each year including 2008, 2009,
2010 and 2011. Removal figures for the Jackson Street Landfill
Centre have not been kept, but culling operations were conducted
in 2010 and ceased in 2011. The relatively high numbers of
cats removed from the McCrobies Gully site compared with the
number that we live-trapped are not surprising, given that landfill
centres or ‘tip’sites are highly modified landscapes with an
abundance of food resources that can support remarkably high
densities of cats (Denny 2005).
Estimation of relative abundance and activity
Site occupancy –potential for seasonal changes
Site-occupancy analyses indicated that the spatial distribution
of feral cats was generally stable during the 3-year survey period,
although there was some support for the detection probability
varying between site surveys. Cameras were not always spatially
independent, with an average of 14% of individual cats being
recorded at more than one camera site within sessions; thus,
changes in site occupancy may have been underestimated.
Models with constant site occupancy, colonisation and
probability of detection were the most highly supported, with
models with constant site occupancy, colonisation and
probability of detection that varied with study site survey
having next-most support (Table 1). Most models converged;
however, three models across two sites where the probability of
site occupancy or colonisation was set to vary with study-site
survey, did not.
Estimates of the probabilities of site occupancy, detection and
colonisation under the most universally supported model, psi(.)
gamma(.)p(.), revealed remarkably similar estimates of the
probability of site occupancy among all four sites. Probabilities
of detection indicated some site variation, with higher values for
the Tasman Peninsula and the Wellington Ranges than the South-
west and Mount Field. The South-west had the highest gamma
estimate; that is, the highest probability of an unoccupied site at
Time tbecoming occupied at time t+1 (Table2).
There was support for a model with survey-specific variation
in detection probability. Detection estimates generated under the
psi(.)gam(.)p(survey) model were used as a measure of feral-cat
activity at cameras, and were used for comparison with other
relative abundance and activity estimators.
Table 1. Feral cat site-occupancy model results for four sites in southern Tasmania
Models are based on Akaike’s Information Criterion corrected for small sample sizes (AICc); those shown are for delta AICc =0.00, models within 2 points,
the next closest model based on model selection, and models that include a variable of interest estimating occupancy (psi), probability of detection (p), and
probability of colonisation (gam, abbreviated from gamma) for feral cats at four sites in Tasmania from 2009 to 2011: (a) South-west, (b) Mt Field, (c) Tasman
Peninsula, and (d) Wellington Ranges. Site occupancy modelling was performed using PRESENCE 4.0.
Model AIC
c
Delta
AIC
c
AICc wt Model
likelihood
No.
parameters
–2LogLik Converged?
South-west
psi(.)gam(.)p(.) 451.27 0.00 0.5529 1.0000 3 445.25 Yes
psi(.)gam(survey)p(.) 451.95 0.68 0.3935 0.7118 7 437.86 +4.67 sig digits
psi(survey)gam(.)p(.) 457.52 6.25 0.0243 0.0439 8 441.40 –15.47 sig digits
psi(.)gam(.)p(survey) 459.68 8.41 0.0084 0.0149 8 443.56 Yes
Mount Field
psi(.)gam(.)p(survey) 348.56 0.00 0.5599 1.0000 8 332.44 Yes
psi(.)gam(.)p(.) 350.98 2.42 0.1670 0.2982 3 344.96 Yes
Tasman Peninsula
psi(.)gam(.)p(.) 558.64 0.00 0.3042 1.0000 3 552.62 Yes
psi(.)gam(.)p(survey) 561.13 2.49 0.0876 0.2879 8 545.01 Yes
Wellington Ranges
psi(.)gam(.)p(.) 527.72 0.00 0.5138 1.0000 3 521.72 Yes
psi(.)gam(.)p(survey) 528.42 0.68 0.3657 0.7118 7 514.33 Yes
psi(.)gam(survey)p(.) 531.47 3.73 0.0796 0.1549 6 519.40 +4.87 sig digits
Table 2. Parameter estimates from the most strongly supported feral-cat site-occupancy model at four sites in southern Tasmania
Estimates of parameters and associated 95% confidence intervals were generated for the model psi(.)gamma(.)p(.) in a multi-season analysis of
feral-cat occupancy at four study sites in Tasmania from 2009 to 2011 using PRESENCE 4.0. Occupancy (psi), probability of detection (p), and
95% confidence intervals (95% conf int) are shown
Site Psi 95% conf int p 95% conf int gamma 95% conf int
South-west 0.6352 0.4063–0.8159 0.0842 0.0582–0.1205 0.4308 0.1949–0.7030
Mount Field 0.5029 0.2582–0.7463 0.0881 0.0566–0.1346 0.1388 0.0328–0.4337
Tasman Peninsula 0.6260 0.4432–0.7787 0.1372 0.1072–0.1739 0.3061 0.1478–0.5286
Wellington Ranges 0.6344 0.4223–0.8047 0.1154 0.0865–0.1524 0.2891 0.0998–0.5988
Feral-cat population response to low-level culling Wildlife Research 413
Population trends and comparison between estimators
Compared with the average MKTBA for feral cats during the
pre- and post-culling surveys, cat numbers increased by 211%
and 75%, respectively, during the 13-month period of
culling at the Mount Field and Tasman Peninsula study sites.
Conversely, estimates of MKTBA remained relatively stable
in the South-west where culling was not conducted, and
fluctuated at the Wellington Ranges observation site (Fig. 3).
These patterns accorded generally with relative abundance
estimates from capture–mark–recapture analyses and with
estimates of activity from the GI and probability of detection
modelling (Fig. 3, Appendix 1).
Scatterplots, Q–Q plots and Shapiro–Wilk tests confirmed
linearity and normality in distributions of the population
estimators, and pairwise comparisons revealed strongly
positive correlations between estimators across all study site
surveys (Table 3). Eight of the population estimates generated
by capture–mark–recapture were excluded from analyses
because they were based on average individual capture
probabilities of <0.1 from study-site surveys, and a further two
study-site surveys yielded data that were too sparse to produce
capture–mark–recapture estimates.
Site occupancy –the effect of culling
There was strong support for a model that included culling as a
seasonal covariate for detection probability at the two sites where
culling was conducted (Table 4). In contrast, there was no support
for this model in the South-west control site. However, the degree
of support for a model where probability of colonisation varied
with culling in the South-west could not be adequately assessed
because this model did not converge (Table 4).
Discussion
We observed marked changes in the relative abundance and
activity of feral cats in response to culling. Contrary to our
prior expectations, however, these changes were positive
rather than negative, with average increases in MKTBA
ranging from 75% to 211% over the period of culling at the
Tasman Peninsula and Mount Field sites, respectively. Cat
numbers fell, and were comparable with those in the pre-
culling period, when culling ceased. Culling operations near
the Wellington Ranges observation site probably differed in
magnitude and intensity from those at the Tasman Peninsula
and Mount Field, so that results between these sites are not
10
0.05
0.10
0.15
0.20
0.25
0
2
4
6
8
10
12
14 (a)(b)
(c)(d)
GI and p-detection estimate
MKTBA and M(h) estimate
MKTBA M(h) GI p-detection
0
0.05
0.10
0.15
0.20
0.25
0
2
4
6
8
10
12
14
GI and p-detection estimate
MKTBA and M(h) estimate
0
0.05
0.10
0.15
0.20
0.25
0
Apr 09 Jun 09 Mar 10 Sep 09 Jun 10 Au
g
10 Ma
y
11 Au
g
11Jun 10 Apr 11 Jun 11
Apr 09 Jun 09 Mar 10 Jun 10 Apr 11 Jun 11 Apr 09 Jun 09 Mar 10 Jun 10 Apr 11 Jun 11
2
4
6
8
10
12
14
GI and p-detection estimate
MKTBA and M(h) estimate
0
0.05
0.10
0.15
0.20
0.25
0
2
4
6
8
10
12
14
GI and p-detection estimate
MKTBA and M(h) estimate
Fig. 3. Comparisons between minimum number of feral cats known to be alive (MKTBA), capture–mark–recapture estimates generated using a model
accounting for individual heterogeneity in capture probability (M(h)), general index (GI), which is a modified version of the Allen index, and p-detection
(probability of detection estimates generated in PRESENCE 4.0 under a multi-season site-occupancy model, which accounts for survey-specific heterogeneity
in detection probability (psi(.)gamma(.)p(survey)) for feral cats at four study sites in Tasmania. (a) South-west, (b) Mount Field, (c) Tasman Peninsula and
(d) Wellington Ranges. Note that some population estimates generated by capture–mark–recapture are not shown if they were based on average individual
capture probabilities of <0.1 during study-site surveys or on data that were too sparse to yield capture–mark–recapture estimates.
414 Wildlife Research B. T. Lazenby et al.
directly comparable. There was, nonetheless, some similarity,
with a decreasing population trend associated with the cessation
of culling at one waste-management site near the Wellington
Ranges. In addition to these results, we found strong and positive
correlations between estimates of relative abundance and activity
across the four estimators that we used at the four study sites.
These results contradict our first hypothesis that culling would
reduce cat numbers, but support the second hypothesis that our
estimates of abundance and activity would be concordant.
Culling has led to reductions in the numbers or activity of
feral cats in other studies, but these results generally have been
obtained after periods of sustained and intensive effort (Algar and
Burrows 2004) or in situations, such as on peninsulas or in fenced
compounds, where the cat population is relatively closed (Short
et al.1997; Read and Bowen 2001). The greatest reductions in cat
populations, including complete eradication, have been achieved
on small islands where refuges are limited and immigration is
non-existent (Domm and Messersmith 1990; Algar et al.2002).
However, our study was intended to simulate low-level culling
effort. This is often applied to open cat populations to reduce
depredation on small game, free-ranging poultry, threatened
native species and other biodiversity values (Coman 1991;
Booth 2010), and perhaps represents the level of control that
might occur where sport-shooting is permitted (Harding et al.
2001). In the discussion below, we first assess whether our
methods were sufficient to allow us to detect actual changes in
cat populations, and then consider biological mechanisms
that might explain our results. We conclude by discussing the
implications of our results for managers.
Detecting changes in cat populations
In the first instance, the strength of inference from our results
rests on the accuracy with which we could identify individual feral
cats at cameras. Although we used a systematic and practical
method for identification that was based on the coat and general
Table 3. Comparisons of abundance and activity estimates for feral cats
Pearson correlations among study-site surveys at four sites in Tasmania (South-west, Mount Field, Tasman Peninsula and Wellington Ranges) for
estimates of relative abundance and activity of feral cats, represented by minimum known to be alive (MKTBA), general index (GI –a modified
form of the Allen index), p-detection (probability of detection generated under a model with constant site occupancy, colonisation, and probability
of detection that varied with study site survey in PRESENCE 4.0), and abundance (Mh jackknife) generated in CAPTURE under the jackknife
model where individual capture probability varies. Correlations significant at 0.001 (2-tailed) are denoted by **
MKTBA GI p-detection Mh jackknife
MKTBA Pearson correlation 1 0.799** 0.641** 0.836**
P(2-tailed) <0.001 0.001 <0.001
N23 23 23 13
GI Pearson correlation 0.799** 1 0.921** 0.902**
P(2-tailed) <0.001 <0.001 <0.001
N23 23 23 13
p-detection Pearson correlation 0.641** 0.921** 1 0.847**
P(2-tailed) 0.001 <0.001 <0.001
N23 23 23 13
Mh jackknife Pearson correlation 0.836** 0.902** 0.847** 1
P(2-tailed) <0.001 <0.001 <0.001
N13 13 13 13
Table 4. Model selection results showing the effect of culling on feral cats
Models are based on Akaike’s information criterion corrected for small sample sizes (AIC
c
); those shown are for delta AIC
c
= 0.00, models within two points,
and the next-closest model based on model selection. Data were obtained from remote cameras operated over 3 years (2009–2011) at three sites in Tasmania,
namely, South-west (where no culling was conducted), Mount Field and Tasman Peninsula, where low-level culling of feral cats was undertaken over a period
of 13 months. Site-occupancy modelling was performed using PRESENCE 4.0. Delta AIC
c
is the difference in AIC values between each model and the model
with the lowest AIC. AIC
c
wt is the model weight. –2LogLik is twice the negative log-likelihood
Model AIC
c
delta AIC
c
AIC
c
wt Model
likelihood
No. par –2LogLik Converged?
South-west
psi(.)gam(cull)p(.) 448.58 0.00 0.6919 1.0000 5 438.53 +5.00 sig digits
psi(.)gam(.)p(.) 451.27 2.69 0.1803 0.2605 3 445.25 Yes
Mt Field
psi(.)gam(.)p(cull) 345.35 0.00 0.8184 1.0000 5 335.30 Yes
psi(.)gam(.)p(survey) 350.36 5.01 0.0668 0.0817 8 334.24 Yes
Tasman Peninsula
psi(.)gam(.)p(cull) 561.35 0.00 0.4365 1.0000 5 551.30 Yes
psi(.)gam(.)p(.) 561.84 0.49 0.3416 0.7827 3 555.82 Yes
psi(.)gam(.)p(survey) 564.91 3.56 0.0736 0.1686 8 548.79 Yes
Feral-cat population response to low-level culling Wildlife Research 415
appearance of cats, the accuracy of the method remains unknown.
Ideally it would be tested against an independent dataset that
included concurrent identification of individuals using remote
cameras and genetic fingerprinting, and/or a blind trial of
known marked individuals at remote cameras. Despite this, we
have no reason to expect that our camera-based identification of
individuals was inaccurate, particularly at the low cat densities
and with the great variation in coat colours and patterns that
prevailed at our study sites. Moreover, we could recognise and
distinguish sibling kittens at the Hobart Cat Centre using this
method. The reliability of individual identification could be
affected by large changes in either the density of feral cats or
variation in coat colour or pattern, but this was unlikely to be a
problem here.
Second, we identified relatively few cats (86 individuals);
these showed a low recapture or within-session photograph rate
(0.03–0.24), and little movement between cameras (14%). Otis
et al.(1978) recommended aiming for sample sizes of 200
individuals or more within a session, with capture probabilities
~0.2, for robust estimation of abundance. Clearly, our data were
too sparse to achieve this recommendation, as evidenced by
the low individual capture probabilities for our capture–mark–
recapture analyses and lack of convergence of some of our site-
occupancy models. Data-paucity is an issue common to many
studies of carnivores (Long et al.2008), particularly feral cats
(Reddiex et al.2004; Denny and Dickman 2010), and in this case,
could have been improved by greatly increasing the density
and/or area covered by cameras. This may also have facilitated
spatially explicit capture–mark–recapture analyses, which have
been used for other rare and cryptic carnivores (Efford and
Fewster 2013), but that require greater recapture rates of the
same individuals between different survey devices than we were
able to achieve. However, an increase in the density of cameras
would likely also have increased the spatial dependence between
sites, making the data unsuitable for site-occupancy analyses
unless the dataset was subsampled (MacKenzie et al.2006). An
increase in the number and recapture rate of individuals would
improve the reliability of our estimates of relative abundance
(Williams et al.2002).
We describe our estimates of abundance as relative abundance
because our remote-camera methodology was biased towards
tracks and trails within study sites and, although feral cats have
often been recorded preferentially using tracks and trails (Mahon
et al.1998; Denny 2005), some individuals may still have avoided
them. Because our remote cameras were set at the same sites for
the duration of all surveys, and the four study sites were selected
on the basis of similar vegetation, topography, and land use, we
expect that our results reasonably reflected relative changes in
abundance through time within and between sites.
Because of the challenges associated with accurately
estimating the abundance of feral cats, there is some appeal
in using activity estimators that do not rely on individual
identification and generally cope well with sparse data. Indeed,
the correlations that we observed between our estimates of
activity and relative abundance indicated that activity
reasonably reflected cat numbers during the surveys at our
study sites. The congruence between estimators also adds
weight to their accuracy in depicting general population trends
at the sites. Although we would caution against extending this
observation beyond the present study, because changes in activity
could result from shifts in behaviour rather than abundance,
our combined results suggested that the increases in
abundance of feral cats at the culling sites were real and not
artefacts of methodology.
The effects of low-level culling on feral cats
Two main biological mechanisms could explain the increase in
abundance of feral cats at the culling sites. In the first instance,
culling operations could have removed dominant individuals and
allowed greater access to resources by remaining cats, thus
promoting an increase in juvenile survival. Such compensatory
responses have been documented in a wide range of species
following low-level culling (Sinclair et al.2006), and can return
a population swiftly to levels equivalent to, or greater than, its
numbers pre-culling. We were able to identify kitten-like features
such as fluffy pelage and small stature from our photographs;
however, we were not confident that we could reliably identify
subadult from adult cats. Regardless, increased survival of
juvenile cats cannot explain the full extent of the population
increases that we observed; at most, it could have provided only a
marginal boost. This is because the reproductive potential of
female feral cats within and around the study sites is unlikely to
have been large enough over a 13-month period to produce the
rapid changes in numbers that we observed.
Alternatively, and probably more plausibly, the culling
sites experienced influxes of new individuals after dominant
resident cats were removed. Several observations support this
interpretation. First, 22 of the 26 cats that were culled were adults,
and these were sufficiently large to indicate that they could have
been residents. Thus, males averaged 3.44 kg (0.50 s.d., n= 12)
and non-pregnant females 2.47 kg (0.70 s.d., n= 10; Lazenby
2012), well within the range of sizes documented for adult
feral cats elsewhere in Australia (Denny et al.2002; Denny
and Dickman 2010).
Second, several studies have suggested that the spatial
organisation of feral cats is socially structured (Dards 1983;
Kerby and Macdonald 1994). For example, Liberg (1984)
found that subordinate male feral cats in Sweden had larger
ranges than dominant males, and suggested that this was
because they were temporarily excluded from certain areas by
dominant individuals, or they failed to settle permanently. Liberg
(1984) also showed that males had larger home ranges than did
females and that the ranges of socially dominant males were
determined by the density and distribution of female cats. If our
culling did indeed remove dominant cats, the observations
from previous studies suggest that subordinate individuals, or
‘floaters’(individuals with no fixed range that traverse the fixed
territories of other individuals), may have been responsible for
the temporary increase in feral-cat numbers at the culling sites.
Bruinzeel and van de Pol (2004) found similarly that floaters
constituted the majority of recolonisers in a manipulative study
where territorial oyster catchers (Haematopus ostralegus) were
removed. Bruinzeel and van de Pol (2004) suggested that these
floating individuals were able to recolonise rapidly because they
had made regular intrusions into the formerly occupied territories
and were able to rapidly recognise a vacancy. Culling has been
shown to disrupt the social organisation of many other species,
416 Wildlife Research B. T. Lazenby et al.
including rodents such as black rats (Rattus rattus) (King et al.
2011) and carnivores including badgers (Meles meles) (Carter
et al.2007) and foxes (Vulpes vulpes) (e.g. Doncaster and
Macdonald 1991; Cavallini 1996).
Third, site occupancy modelling indicated that more feral
cats were detected at the same sites when abundance estimates
increased in 2010, thus supporting the idea that the displacement
of resident animals resulted in an influx of new individuals from
surrounding areas. Detection probability generally increased
during the 2010 surveys at Mount Field and the Tasman
Peninsula, when cat abundances were higher, with the May
2010 survey on the Tasman Peninsula being one exception,
and there was very little support for a model that incorporated
a colonisation function (the probability that an unoccupied site in
Season tis occupied by the species in Season t+ 1). The increased
number of individual cats photographed during the 2010 surveys
also had a lower probability of capture in live traps, which could
have resulted from higher levels of neophobia, or wariness, in
subordinate or floater individuals (Denny and Dickman 2010).
Subordinate animals that move into newly vacated areas are
often male (e.g. King et al.2011). Moseby et al.(2009a) found
that male feral cats fitted with GPS collars moved into new areas
less than 2 days after poison baits had been laid (for feral cats and
foxes). It would be enlightening to know whether there was a
gender bias in the sex ratio of feral cats at Mount Field and the
Tasman Peninsula during the period of higher abundance and
activity in 2010, but unfortunately we could not reliably discern
cat sex on photographs. There was no gender bias in the sex ratio
of culled feral cats, although most of these were trapped during
the early stages of the culling program. It has also been suggested,
but not demonstrated, that dominant feral cats may exclude
subordinates living in the same areas near camera stations
(Lane et al.2013); if correct, subordinate individuals may not
be detected until the dominants have been removed. This is clearly
an area that would benefit from further investigation.
Implications
The potential for feral cats to increase in abundance, and possibly
impact, following low-level culling has important ramifications
for understanding the magnitude of effort necessary to effectively
control feral cats using this method in open populations. Clearly,
the low-level culling effort we used did not constitute a sustained,
multi-faceted, long-term downward pressure on our study
populations, which may be required if culling is to be used
in programs of feral-cat control (Braysher 1993; Olsen 1998;
Hone 2007; Dickman et al.2010). Rather, this study provides
evidence that ad hoc culling of feral cats may be not only
ineffective, but has the potential to increase the impact of feral
cats in open populations. Even worse, the decline in numbers of
cats trapped over the 13-month cull could have led to erroneous
conclusions that the culling was effective if we had not been able
to monitor the populations independently. We cannot say whether
culling responses similar to those we observed might apply in
other areas, but our results do highlight the need for monitoring
the effectiveness of any operations that seek to reduce the impacts
of feral cats. Even very well planned management programs
can have unexpected outcomes, such as those observed on
Macquarie Island when rabbits (Oryctolagus cuniculus)
erupted and removed vegetation following the eradication of
feral cats (Bergstrom et al.2009). Our study also suggests
that the risks of invasion or re-invasion should factor into the
decision-making process for prioritising feral-cat control sites
(Dickman et al.2010).
It is important to note that there are cases where culling-
introduced vertebrate predators in open populations can be used
as a successful strategic short-term management tool. For
example, the stoat (Mustela erminea) is culled in New Zealand
during irruptions of its prey that are driven by heavy falls of seeds
of southern beech (Nothofagus spp.). This strategic culling
reduces damaging predation on nesting endemic birds (King
and Powell 2011). If the intensity of effort needed for either
strategic or long-term culling is not available, the impacts of feral
cats may be reduced in some situations by using exclusion fencing
(Moseby et al.2009b; Denny and Dickman 2010), poison baiting
(Algar et al.2007), increasing the structural complexity of the
environment to provide refuges for prey (Stokes et al.2004), or
managing for healthy populations of dominant native carnivores
that may reduce the impacts of feral cats on prey populations (e.g.
Soulé et al.1988). In all situations, as we have shown, monitoring
is essential to ensure that the control methods are effective.
Acknowledgements
We thank anonymous reviewers for constructive comments on a draft of this
manuscript. This study was conducted under approvals from the University
of Sydney Animal Ethics Committee (numbers L04/8/2008/3/4878 and
L04/7-2009/2/5091) and scientific permits were issued by the Tasmanian
Department of Primary Industries, Parks, Water and Environment for research
on native wildlife (TFA108121, TFA10046, and TFA11137). This work
would not have been possible without logistical support from the Tasmanian
Department of Primary Industries, Parks, Water and Environment, including
the Threatened Species Section, Mount Field Parks and Wildlife Service
base, and Wildlife Management Branch. The Parks and Wildlife Service team
based at Mount Field conducted several culling trips at that site, in addition to
providing invaluable advice and logistical support.
References
Algar, D., and Burrows, N. D. (2004). Feral cat control research: Western
Shield review –February 2003. Conservation Science Western Australia
5, 131–163.
Algar, D. A., Burbidge, A. A., and Angus, G. J. (2002). Cat eradication on
Hermite Island, Montebello Islands, Western Australia. In ‘Turning
the Tide: the Eradication of Invasive Species’. (Eds C. R Veitch and
M. N. Clout.) pp. 14–18. (IUCN: Gland, Switzerland.)
Algar, D., Angus, G. J., Williams, M. R., and Mellican, A. E. (2007). Influence
of bait type, weather and prey abundance on bait uptake by feral cats
(Felis catus) on Peron Peninsula, Western Australia. Conservation
Science Western Australia 6, 109–149.
Bengsen, A., Butler, J., and Masters, P. (2011). Estimating and indexing feral
cat population abundances using camera traps. Wildlife Research 38,
732–739. doi:10.1071/WR11134
Bergstrom, D. M., Lucieer, A., Kiefer, K., Wasley, J., Belbin, L., Pedersen,
T. K., and Chown, S. L. (2009). Indirect effects of invasive species
removal devastate World Heritage island. Journal of Applied Ecology 46,
73–81. doi:10.1111/j.1365-2664.2008.01601.x
Booth, C. (2010). Hunting and feral animal control: conservation or con?
In ‘Convergence or Conflict: Animal Welfare in Wildlife Management
and Conservation’. (Eds M. Tensen and B. Jones.) pp. 25–31. (RSPCA
Australia: Canberra.)
Feral-cat population response to low-level culling Wildlife Research 417
Braysher, M. (1993). ‘Managing Vertebrate Pests: Principles and Strategies.’
(Bureau of Resource Sciences: Canberra.)
Bruinzeel, L. W., and van de Pol, M. (2004). Site attachment of floaters
predicts success in territory acquisition. Behavioral Ecology 15, 290–296.
doi:10.1093/beheco/arh019
Burnham, K. P., and Anderson, D. R. (2002). ‘Model Selection and
Multimodel Inference: a Practical Information-theoretic Approach’.
2nd edn. (Springer-Verlag: Berlin.)
Campbell, K. J., Harper, G., Algar, D., Hanson, C. C., Keitt, B. S., and
Robinson, S. (2011). Review of feral cat eradication on islands. In ‘Island
Invasives: Eradication and Management’. (Eds C. R. Veitch, M. N. Clout
and D. R. Towns.) pp. 37–46. (IUCN: Gland, Switzerland.)
Carter, S. P., Delahay, R. J., Smith, G. C., Macdonald, D. W., Riordan, P.,
Etherington, T. R., Pimley, E. R., Walker, N. J., and Cheeseman, C. L.
(2007). Culling-induced social perturbation in Eurasian badgers Meles
meles and the management of TB in cattle: an analysis of a critical
problem in applied ecology. Proceedings. Biological Sciences 274,
2769–2777. doi:10.1098/rspb.2007.0998
Cavallini, P. (1996). Variation in the social system of the red fox. Ethology
Ecology and Evolution 8, 323–342. doi:10.1080/08927014.1996.
9522906
Coman, B. (1991). The ecology and control of feral cats in Australia. In ‘The
Impact of Cats on Native Wildlife’. (Ed. C. Potter.) pp. 70–74. (Australian
National Parks and Wildlife Service: Canberra.)
Dards, J. L. (1983). The behaviour of dockyard cats: interaction of adult males.
Applied Animal Ethology 10, 133–153. doi:10.1016/0304-3762(83)
90117-7
Denny, E. A. (2005). Ecology of free-living cats exploiting waste disposal
sites: diet, morphometrics, population dynamics and population genetics.
Ph.D. Thesis, University of Sydney, Sydney.
Denny, E. A., and Dickman, C. R. (2010). ‘Review of Cat Ecology and
Management Strategies in Australia.’(Invasive Animals Cooperative
Research Centre: Canberra.)
Denny, E., Yakovlevich, P., Eldridge, M. D. B., and Dickman, C. R. (2002).
Social and genetic analysis of a population of free-living cats (Felis
catus L.) exploiting a resource-rich habitat. Wildlife Research 29,
405–413. doi:10.1071/WR02092
DEWHA (2008). ‘Threat Abatement Plan for Predation by Feral Cats.’
(Department of Environment, Water, Heritage and the Arts: Canberra.)
Dickman, C. R. (1996). ‘Overview of the Impacts of Feral Cats on Australian
Native Fauna.’(Australian Nature Conservation Agency: Canberra.)
Dickman, C. R., Denny, E., and Buckmaster, T. (2010). Identification of
sites of high conservation priority impacted by feral cats. Report for
the Department of the Environment, Water, Heritage and the Arts,
Canberra.
Domm, S., and Messersmith, J. (1990). Feral cat eradication on a Barrier
Reef island, Australia. Atoll Research Bulletin 338,1–4. doi:10.5479/
si.00775630.338.1
Doncaster, C. P., and Macdonald, D. W. (1991). Drifting territoriality in the
red fox Vulpes vulpes. Journal of Applied Ecology 60, 423–439.
doi:10.2307/5288
Donovan, T. M., and Hines, J. (2007). Exercises in occupancy modeling
and estimation. Available at http://www.uvm.edu/envnr/vtcfwru/spread
sheets/occupancy.htm. [Verified October 2013]
DPIPWE (2008). ‘Cat Management in Tasmania: Taking the Initiative.’
(Department of Primary Industries, Parks, Water and Environment:
Hobart.)
DPIPWE (2009). ‘Tasmanian Community Forest Agreement Research into
Alternatives to 1080, Newsletter 17.’(Department of Primary Industries,
Parks, Water and Environment: Hobart.)
Efford, M. G., and Fewster, R. M. (2013). Estimating population size by
spatially explicit capture–recapture. Oikos 122, 918–928. doi:10.1111/
j.1600-0706.2012.20440.x
Engeman, R. M. (2005). Indexing principles and a widely applicable
paradigm for indexing animal populations. Wildlife Research 32,
203–210. doi:10.1071/WR03120
Fisher, P., Algar, D., and Johnston, M. (2001). Current and future feral cat
control management for conservation outcomes. In ‘Veterinary
Conservation Biology; Wildlife Health and Management in
Australasia: Proceedings of an International Joint Conference, Taronga
Zoo, Sydney, Australia 1–6 July 2001’. (Eds A. Martin and L. Vogelnest.)
pp. 1–8. (Australian Veterinary Association: Canberra.)
Forsyth, D. M., Robley, A. J., and Reddiex, B. (2005). ‘Review of
Methods used to Estimate the Abundance of Feral Cats.’(Arthur
Rylah Institute for Environmental Research, Department of
Sustainability and Environment: Melbourne.)
Friday, N., Smith, T. D., Stevick, P. T., and Allen, J. (2000). Measurement
of photographic quality and individual distinctiveness for the
photographic identification of humpback whales, Megaptera
novaeanglia. Marine Mammal Science 16, 355–374. doi:10.1111/
j.1748-7692.2000.tb00930.x
Hanke, P. U., and Dickman, C. R. (2013). Sniffing out the stakes: hair-
snares for wild cats in arid environments. Wildlife Research 40,45–51.
doi:10.1071/WR12210
Harding, E. K., Doak, D. F., and Albertson, J. D. (2001). Evaluating the
effectiveness of predator control: the non-native red fox as a case study.
Conservation Biology 15, 1114–1122. doi:10.1046/j.1523-1739.2001.
0150041114.x
Hayne, D. W. (1949). An examination of the strip census method for
estimating animal populations. The Journal of Wildlife Management
13, 145–157. doi:10.2307/3796084
Hines, J. E. (2006). ‘PRESENCE2: Software to Estimate Patch Occupancy
and Related Parameters.’(USGS–PWRC.) Available at http://www.
mbr_pwrc.usgs.gov/software/presence.html. [Verified month year]
Hone, J. (2007). ‘Wildlife Damage Control.’(CSIRO Publishing:
Melbourne.)
Huggins, R. M. (1989). On the statistical analysis of capture experiments.
Biometrika 76, 133–140. doi:10.1093/biomet/76.1.133
Jackson, R. M., Roe, J. D., Wangchuck, R., and Hunter, D. O. (2006).
Estimating snow leopard population abundance using photography and
capture–recapture techniques. Wildlife Society Bulletin 34, 772–781.
doi:10.2193/0091-7648(2006)34[772:ESLPAU]2.0.CO;2
Karanth, K. U., and Nichols, J. D. (1998). Estimation of tiger densities in India
using photographic captures and recaptures. Ecology 79, 2852–2862.
doi:10.1890/0012-9658(1998)079[2852:EOTDII]2.0.CO;2
Kelly, M. J., Noss, A. J., Di Bitetti, M. S., Maffei, L., Arispe, R. L., Paviolo, A.,
De Angelo, C. D., and Di Blanco, Y. E. (2008). Estimating puma densities
from camera trapping across three study sites: Bolivia, Argentina and
Belize. Journal of Mammalogy 89, 408–418. doi:10.1644/06-MAMM-A-
424R.1
Kerby, G., and Macdonald, D. W. (1994). Cat society and the consequences
of colony size. In ‘The Domestic Cat: the Biology of its Behaviour’.
(Eds D. C. Turner and P. Bateson.) pp. 67–81. (Cambridge University
Press: Cambridge, UK.)
King, C., and Powell, R. (2011). Managing an invasive predator pre-adapted
to a pulsed resource: a model of stoat (Mustela erminea) irruptions in
New Zealand beech forests. Biological Invasions 13, 3039–3055.
doi:10.1007/s10530-011-9993-y
King, C. M., Innes, J. G., Fitzgerald, N., Winstanley, T., O’Brien, B.,
Bridgman, L., and Cox, N. (2011). Reinvasion by ship rats (Rattus
rattus) of forest fragments after eradication. Biological Invasions 13,
2391–2408. doi:10.1007/s10530-011-0051-6
Lane, C., Bengsen, A., and Murphy, E., eds. (2013). ‘Proceedings of the
National Feral Cat Management Workshop, Mantra Southbank,
Melbourne, Australia, 30 November, 2010.’(Invasive Animals
Cooperative Research Centre: Canberra.)
418 Wildlife Research B. T. Lazenby et al.
Lazenby, B. T. (2012). Do feral cats affect small mammals? A case studyfrom
the forests of southern Tasmania. Ph.D. Thesis, University of Sydney.
Lazenby, B. T., and Dickman, C. R. (2013). Patterns of detection and capture
are associated with cohabiting predators and prey. PLoS ONE 8, e59846.
doi:10.1371/journal.pone.0059846
Liberg, O. (1984). Home range and territoriality in free ranging house cats.
Acta Zoologica Fennica 171, 283–285.
Long, R. A., Mackay, P., Ray, J., and Zielinski, W. (2008). ‘Non-Invasive
Survey Methods for Carnivores.’(Island Press, University of Michigan:
Ann Arbor, MI.)
Loss, S. R., Will, T., and Marra, P. P. (2013). The impact of free-ranging
domestic cats on wildlife of the United States. Nature Communications
4, 1396. doi:10.1038/ncomms2380
MacKenzie, D. I., Nichols, J. D., Royle, J. A., Pollock, K. H., Bailey,L. L., and
Hines, J. E. (2006). ‘Occupancy Estimation and Modeling: Inferring
Patterns and Dynamics of Species Occurrence.’(Elsevier: London.)
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. doi:10.1071/WR97007
Markowitz, T. M., Harlin, A. D., and Wursig, B. (2003). Digital photography
improves efficiency of individual dolphin identification. Marine Mammal
Science 19, 217–223. doi:10.1111/j.1748-7692.2003.tb01103.x
Meek, P. D., Ballard, G., and Fleming, P. (2012). ‘An Introduction to
Camera Trapping for Wildlife Surveys in Australia. PestSmart Toolkit
Publication.’(Invasive Animals Cooperative Research Centre: Canberra.)
Moseby, K. E., Stott, J., and Crisp, H. (2009a). Movement patterns of feral
predators in an arid environment –implications for control through poison
baiting. Wildlife Research 36, 422–435.
Moseby, K. E., Hill, B. M., and Read, J. L. (2009b). Arid Recovery –a
comparison of reptile and small mammal populations inside and
outside a large rabbit, cat and fox-proof exclosure in arid South
Australia. Austral Ecology 34, 156–169. doi:10.1111/j.1442-9993.2008.
01916.x
Nogales, M., Martín, A., Tershy, B. R., Donlan, C. J., Veitch, D., Puerta, N.,
Wood, B., and Alonso, J. (2004). A review of feral cat eradication on
islands. Conservation Biology 18, 310–319. doi:10.1111/j.1523-1739.
2004.00442.x
Nogales, M., Vidal, E., Medina, F. M., Bonnaud, E., Tershy, B. R., Campbell,
K. J., and Zavaleta, E. S. (2013). Feral cats and biodiversity conservation:
the urgent prioritization of island management. Bioscience 63, 804–810.
doi:10.1525/bio.2013.63.10.7
Olsen, P. (1998). ‘Australia’s Pest Animals: New Solutions to Old Problems.’
(Bureau of Resource Sciences: Canberra; and Kangaroo Press: Sydney.)
Otis, D. L., Burnham, K. P., White, G. C., and Anderson, D. R. (1978).
Statistical inference from capture data on closed animal populations.
Wildlife Monographs 62,1–135.
Pollock, K. H., and Otto, M. C. (1983). Robust estimation of population size
in closed animal populations from capture–recapture experiments.
Biometrics 39, 1035–1049. doi:10.2307/2531337
Read, J., and Bowen, Z. (2001). Population dynamics, diet and aspects of the
biology of feral cats and foxes in arid South Australia. Wildlife Research
28, 195–203. doi:10.1071/WR99065
Reddiex, B., and Forsyth, D. (2004). Review of existing red fox, wild dog,
feral cat, rabbit, feral pig and feral goat control in Australia. II. Information
gaps. Unpublished final report for the Department of the Environment and
Heritage, Canberra.
Reddiex, B., Forsyth, D., McDonald-Madden, E., Einoder, L. D., Griffion,
P. A., Chick, R. R., and Robley, A. J. (2004). Review of existing red fox,
wild dog, feral cat, feral rabbit, feral pig, and feral goat control in Australia.
I. Audit. Report for the Department of the Environment and Heritage,
Canberra.
Reddiex, B., Forsyth, D. M., McDonald-Madden, E., Einoder, L. D.,
Griffioen, P. A., Chick, R. R., and Robley, A. J. (2006). Control of
pest mammals for biodiversity protection in Australia. I. Patterns of
control and monitoring. Wildlife Research 33, 691–709. doi:10.1071/
WR05102
Rexstad, E., and Burnham, K. P. (1991). ‘User’s Guide for Interactive
Program CAPTURE. Abundance Estimation of Closed Animal
Populations.’(Colorado State University: Fort Collins, CO.)
Robley, A., Gormley, A., Woodford. L., Lindeman, M., Whitehead, B.,
Albert, R., Bowd, M., and Smith, A. (2010). Evaluation of camera
trap sampling designs used to determine change in occupancy rate and
abundance of feral cats. Technical Report No. 201.Arthur Rylah Institute
for Environmental Research, Melbourne.
Sharp, T., and Saunders, G. (2005). ‘Humane Pest Animal Control: Codes
of Practice and Standard Operating Procedures.’(New South Wales
Department of Primary Industries: Orange, NSW.)
Short, J., Turner, B., Risbey, D. A., and Carnamah, R. (1997). Control of feral
cats for nature conservation. II. Population reduction by poisoning.
Wildlife Research 24, 703–714. doi:10.1071/WR96071
Sinclair, A. R. E., Fryxell, J. M., and Caughley, G. (2006). ‘Wildlife Ecology,
Conservation, and Management.’2nd edn. (Blackwell Publishing:
Oxford, UK.)
Slooten, E., Dawson, S. M., and Lad, F. (1992). Survival rates of
photographically identified Hector’s dolphins from 1984 to 1998.
Marine Mammal Science 8, 327–343. doi:10.1111/j.1748-7692.1992.
tb00049.x
Soulé, M. E., Bolger, D. T., Alberts, A. C., Wright, J., Sorice, M., and Hill, S.
(1988). Reconstructed dynamics of rapid extinctions of chaparral-
requiring birds in urban habitat islands. Conservation Biology 2,
75–92. doi:10.1111/j.1523-1739.1988.tb00337.x
Stokes, V. L., Pech, R. P., Banks, P. B., and Arthur, A. D. (2004). Foraging
behaviour and habitat use by Antechinus flavipes and Sminthopsis murina
(Marsupialia: Dasyuridae) in response to predation risk in eucalypt
woodland. Biological Conservation 117, 331–342. doi:10.1016/j.biocon.
2003.12.012
Turner, D., and Bateson, C. (2000). ‘The Domestic Cat: the Biology of its
Behaviour.’(Cambridge University Press: Cambridge, UK.)
White, G., and Burnham, K. P. (1999). ‘Program MARK.’(Colorado State
University: Fort Collins, CO.) Available at http://www.phidot.org/
software/mark. [Verified month year]
Williams, B. K., Nichols, J. D., and Conroy, M. J. (2002). ‘Analysis and
Management of Animal Populations: Modelling, Estimation and Decision
Making.’(Academic Press: San Diego, CA.)
Feral-cat population response to low-level culling Wildlife Research 419
Appendix 1. Population estimates for feral cats at four study sites in Tasmania
Density was calculated by dividing the MKTBA estimate of cats by the effective area of each study site. Effective areas were calculated by placing a minimum
convex polygon around the camera locations and buffering by 1 km. The effective areas for each study site were as follows: South-west, 117 km
2
; Mount Field,
68 km
2
; Tasman Peninsula, 69 km
2
; and Wellington Ranges, 57 km
2
Study
area
Survey MKTBA M(h)
Jackknife
SE
M(h)
Average
capture
probability (p)
M(h)
Density
(individual
cats km
–2
)
General
index
Lower and
upper 95% CI
for general
index
Probability of
detection-site
occupancy
modelling
Lower and
upper 95% CI
for site-occupancy
prob of det
SW April 2009 6 6 2.278 0.238 0.051 0.079 0.072–0.086 0.113 0.051–0.235
SW June 2009 4 0.034 0.045 0.041–0.049 0.070 0.027–0.172
SW April 2010 6 7 1.467 0.143 0.051 0.063 0.060–0.066 0.088 0.049–0.154
SW June 2010 3 4 1.547 0.161 0.026 0.042 0.040–0.044 0.072 0.035–0.146
SW March 2011 6 8 1.996 0.125 0.051 0.094 0.088–0.100 0.092 0.049–0.166
SW June 2011 4 6
B
1.993 0.083 0.034 0.040 0.038–0.042 0.063 0.027–0.138
MtF April 2009 1 0.015 0.016 0.009–0.023 0.034 0.008–0.134
MtF June 2009 2 0.029 0.029 0.023–0.035 0.074 0.022–0.223
MtF April 2010 9 11 2.001 0.136 0.132 0.125 0.120–0.130 0.147 0.083–0.248
MtF July 2010 5 6 1.414 0.143 0.088 0.051 0.049–0.053 0.092 0.040–0.199
MtF April 2011 3 17
B
9.075 0.025 0.044 0.021 0.020–0.022 0.037 0.014–0.094
MtF July 2011 3 4
B
1.419 0.089 0.044 0.027 0.025–0.029 0.053 0.023–0.118
TP April 2009 6 8 3.185 0.179 0.087 0.098 0.091–0.105 0.165 0.084–0.300
TP June 2009 4 7 1.851 0.204 0.058 0.101 0.093–0.109 0.176 0.089–0.319
TP May 2010 12 24
B
7.383 0.051 0.174 0.088 0.084–0.091 0.108 0.065–0.176
TP August 2010 9 13 2.487 0.132 0.130 0.138 0.133–0.143 0.221 0.142–0.327
TP April 2011 7 9
B
3.618 0.095 0.101 0.059 0.057–0.061 0.105 0.054–0.192
TP July 2011 7 7 2.438 0.133 0.101 0.071 0.068–0.074 0.104 0.058–0.178
WR September 2009 5 4 0.936 0.214 0.088 0.065 0.061–0.069 0.080 0.045–0.139
WR May 2010 12 38
B
11.614 0.06 0.211 0.112 0.108–0.116 0.139 0.089–0.210
WR August 2010 10 10 2.041 0.136 0.175 0.121 0.117–0.125 0.160 0.099–0.247
WR
A
May 2011 5 7
B
1.982 0.092 0.088 0.045 0.043–0.047 0.076 0.034–0.158
WR
A
August 2011 5 7
B
1.982 0.092 0.088 0.057 0.054–0.060 0.075 0.035–0.152
A
The population estimates for May and August 2011 camera surveys in the Wellington Ranges are the same; however, they are based on different data.
B
These population estimates were generated from a survey, with an average capture probability of <0.1 and are likely to be unreliable (Otis et al.1978). Blanks
for M(h) population estimates represent surveys where there were insufficient data to produce an estimate. Probability of detection figures were generated
under a multi-season site-occupancy model, with constant site occupancy, constant colonisation and probability of detection that varied with study-site survey
(i.e. psi(.)Gamma(.)P(study site survey)).
420 Wildlife Research B. T. Lazenby et al.
www.publish.csiro.au/journals/wr
