ArticlePDF Available
Effectiveness of thermal cameras compared to spotlights for
counts of arid zone mammals across a range of ambient
temperatures
Hugh McGregor
A
,
B
,
C
,
F
,Katherine Moseby
C
,
D
,Christopher N. Johnson
A
,
B
and Sarah Legge
A
,
E
A
National Environmental Science Program Threatened Species Recovery Hub, Centre for Biodiversity and
Conservation Science, University of Queensland, St Lucia, Qld 4075, Australia.
B
School of Natural Sciences, Private Bag 55, University of Tasmania, Hobart, Tas. 7001, Australia.
C
Arid Recovery, PO Box 147, Roxby Downs, SA 5725, Australia.
D
University of New South Wales, Sydney, NSW 2052, Australia.
E
Fenner School of Environment and Society, The Australian National University, Canberra, ACT 2601,
Australia.
F
Corresponding author. Email: hugh.mcgregor@utas.edu.au
Abstract. Effective monitoring of mammal species is critical to their management. Thermal cameras may enable more
accurate detection of nocturnal mammals than visual observation with the aid of spotlights. We aimed to measure
improvements in detection provided by thermal cameras, and to determine how these improvements depended on ambient
temperatures and mammal species. We monitored small to medium sized mammals in central Australia, including small
rodents, bettongs, bilbies, European rabbits, and feral cats. We conducted 20 vehicle-based camera transects using both a
spotlight and thermal camera under ambient temperatures ranging from 108Cto358C. Thermal cameras resulted in more
detections of small rodents and medium sized mammals. There was no increased benefit for feral cats, likely due to their
prominent eyeshine. We found a strong relationship between increased detections using thermal cameras and
environmental temperature: thermal cameras detected 30% more animals than conventional spotlighting at approximately
158C, but produced few additional detections above 308C. Spotlighting may be more versatile as it can be used in a greater
range of ambient temperatures, but thermal cameras are more accurate than visual surveys at low temperatures, and can be
used to benchmark spotlight surveys.
Keywords: arid zone mammals, conservation, detectability, distance sampling, monitoring, spotlighting, thermal
imagery.
Received 4 June 2020, accepted 6 February 2021, published online 12 March 2021
Introduction
Conservation actions are best evaluated with effective moni-
toring (Legge et al. 2018). Monitoring of many mammal species
across Australia is limited by difficulties in detecting animals,
and by challenges in obtaining estimates of actual abundance
rather than indices of activity. This is especially relevant in cases
where it is difficult to conduct a trapping program, such as where
animals occupy habitats that are hard to sample (e.g. arboreal
animals living high up in tree canopies), or where some animals
are hyperabundant and cause trap saturation (Wayne et al. 2008;
Page et al. 2013). As an alternative to trapping, travelling at
night with a spotlight or torch and counting all animals seen
(spotlighting), while recording distance of animals from the
transect line to allow distance-sampling analysis, remains a
common technique to monitor animals when trapping is
impractical and abundance estimates are required.
Although spotlighting with distance sampling is an important
tool for monitoring wildlife in many situations, it has limitations.
One drawback is that some sources of detection error are not
measurable or able to be estimated. Although some sources of
error such as distance from transect (where detections close to
the observer are more likely than those further away), observer
experience, or wind can be measured and factored into analysis,
other major causes of error cannot be measured. For example,
many animals are principally detectable from eyeshine, but
some animals in a population might not have their face directed
towards the observer or their head may be obscured by vegeta-
tion (Lindenmayer et al. 2001;Goldingay and Sharpe 2004).
This might not change with distance to transect, and hence may
not be considered part of the estimated error. The head angles of
undetected animals cannot be measured, and so animals facing
away from the observer or with their heads hidden from view
CSIRO PUBLISHING
Australian Mammalogy
https://doi.org/10.1071/AM20040
Journal compilation Australian Mammal Society 2021 www.publish.csiro.au/journals/am
would constitute a ‘ghost’ subpopulation that could not be
integrated into any form of abundance analysis, resulting in
underestimates of actual density.
Thermal camera technology might be able to overcome errors
with detectability associated with spotlight surveys, creating a
more robust estimate of wildlife density (Gill et al. 1997;
Cilulko et al. 2013). These cameras see only heat, so detectability
is not dependent on whether an animal faces the observer, diplays
eyeshine, has visibly noticeable features (e.g. black fluffy tail), or
is moving. Instead, detection relies on whether the animal’s body
heat contrasts with the background. In addition to this, many
thermal camera modules have the ability to record video of the
feed, allowing independent verification.
There are limits to the effectiveness of thermal cameras. If
animals are to be detected their body temperature must contrast
with the environment (Kays et al. 2019). This condition would
not hold when ambient temperatures are similar to or greater
than average mammalian body temperatures (,358C: Morrison
1962). As most mammals have insulating fur and their exterior
temperature is lower than body temperature (Gerken 1996;
Simpson et al. 2016), even lower ambient temperatures might
be needed for thermal cameras to be effective. For example,
Vinson et al. (2020) found detectability of furred arboreal
mammals decreased when ambient temperatures were above
248C. Added to this, many microhabitats in the landscape retain
heat (e.g. under logs), making it difficult to discern the heat
generated by animals from background variability. As well as
environmental constraints, many animal species have beha-
vioural or physical traits that make them better suited to detec-
tion by spotlighting. For example, Focardi et al. (2001) did not
find thermal cameras to be better than spotlights to survey foxes,
due to their bright eyeshine and propensity to look at the
observer.
Our study aimed to assess the relative effectiveness of
adding thermal cameras to traditional visual spotlight surveys
in different survey contexts, and to understand in which
species and climates thermal cameras would provide the most
benefit. We monitored small rodents, medium sized native
mammals like bettongs and bilbies, European rabbits, feral cats
and kangaroos. We asked the following questions. First, are
detection rates from surveys using a thermal camera and
spotlight together different from those derived from just a
spotlight, and how do these differences vary among mammal
species? Second, does including thermal camera detections
alter detection functions and estimated strip widths in distance
sampling? Third, what temperatures are required for thermal
imaging to work optimally? Finally, can insights learnt from
using thermal cameras enable more accurate density estimates?
We had an opportunity to measure the accuracy of density
estimates derived from thermal imagery and spotlighting in
a situation where we had a relatively reliable benchmark
of the actual population of rabbits in a fenced paddock
(see McGregor et al. 2019).
Methods
The comparison of thermal cameras to spotlight surveys was
conducted at Arid Recovery, a conservation reserve in central
South Australia (308290S, 1368530E), in an arid landscape with
160 mm average rainfall per year (records from Olympic Dam
Aerodrome, www.bom.gov.au/climate/data, accessed 30 October
2019). The area is dominated by swales with scattered saltbush
,50 cm high and longitudinal dunes dominated by shrubs typi-
cally 1–5 m high. Arid Recovery is a fenced conservation reserve.
In over half of the fenced reserve (60 km
2
), all European rabbits
(Oryctolagus cuniculus), feral cats (Felis catus) and red foxes
(Vulpes vulpes) have been removed (Moseby and Read 2006),
whereas the other half (62 km
2
) is where populations of native and
invasive mammals coexist in manipulative experiments (Moseby
et al. 2012;West et al. 2018;McGregor et al. 2019;Ross et al.
2019). Numerous threatened native mammals have been rein-
troduced, including bilbies (Macrotis lagotis), burrowing bet-
tongs (Bettongia lesueur), and stick-nest rats (Leporillus
conditor). Small mammals including the spinifex hopping mouse
(Notomys alexis) and plains mouse (Pseudomys australis)are
abundant inside the reserve (Moseby et al. 2009). Outside the
fenced reserve, European rabbits and feral cats are abundant, and
there are occasional records of dingoes (Canis lupus)and
red foxes.
The comparison between detections using spotlighting and
thermal cameras was conducted on five different transects
through four areas of the Arid Recovery reserve and adjacent
land (Fig. 1), sampled a total of 20 times. Each transect took
between 2 and 5 h. Four samples each were taken from the
northern natives and feral coexistence paddocks (see Fig. 1),
four from outside the reserve on the boundary, and eight from
inside the natives-only feral-free paddocks. All transects were
between 12 km and 18 km long.
All vehicle-based spotlight/thermal camera surveys were
conducted between June 2016 and December 2018. Surveys
began at least 1 h after sunset. Before beginning a transect, we
would record time, wind (none, light, medium, strong), moon
phase, and whether each observer would be considered expe-
rienced (more than 50 h spotlighting for animals) or inexperi-
enced. The front passenger seat was occupied by someone
using a 100 W spotlight scanning to the left of the vehicle (not
the right-hand side), and a second surveyor used a thermal
camera from the rear passenger seat. The thermal camera
(FLIR scout III 640, FLIR_ Systems, Wilsonville, USA) had
a resolution of 640 512 pixels. This was the highest known
resolution of thermal cameras commercially available in
Australia for under AU$10 000 in 2016. The vehicle was
driven at a speed between 5 and 20 km h
1
. We used the
mode InstAlert
TM
, where objects hotter than the background
would appear red.
Records were kept of all mammals seen, and the perpendicu-
lar distance from the road of each sighted animal was recorded.
Most distances were estimated, but at least once every five
detections the distance was measured with a rangefinder (800 m
Laser Rangefinder, Kogan, Melbourne, Australia) to avoid drift
of estimated distance. Animals observed on the road by the
driver were recorded as a distance of zero. As well as recording
animals seen, the user of the thermal camera also recorded
whether an animal seen by the spotlighter was not seen by
thermal camera and vice versa. In some cases, it was ambiguous
as to whether one would have seen it had the other not (e.g. the
spotlighter might look more intently for an animal if the
thermal camera user announced that they had spotted an animal).
BAustralian Mammalogy H. McGregor et al.
For the few ambiguous cases, we discussed the sighting as a
group and made a collective decision on how the detection
should be recorded. However, the vast majority of instances
(.98% of sightings) were unambiguous for numerous reasons.
First, neither the spotlighter or thermal camera user can see
where the other was looking (a spotlight beam is invisible
through a thermal camera). Second, transects were short enough
that fatigue was rarely an issue (i.e. people were looking intently
most of the time regardless of whether the other person detected
an animal). As the spotlighter was in the front seat typically
scanning forward-left, while the thermal camera user was in the
back scanning left to back-left, most error would be from the
thermal camera user alerted to animal presence by the spot-
lighter, not the other way around as the former is scanning
ground already scanned by the latter. Overall error associated
with this issue was deemed to be minimal in regard to our core
aims, given that we were comparing spotlighting versus a
spotlight/thermal camera combination. Therefore, our tally of
how many animals would have been seen were there no thermal
camera should be largely accurate.
Animals seen were identified where possible; however, all
small-mammal detections (i.e. animals less than 100 g) were
grouped together. In some cases where only the thermal camera
detected a medium-sized mammal (1–3 kg), we would assume the
identity based on which paddock we were in. For example, all
medium mammals in the dingo paddock were classed as rabbit,
and those in the core reserve counted as bettongs or bilbies.
To estimate the detection functions of each species group
(small mammals, bettongs/bilbies, rabbits, cats), we used distance
sampling in R ver. 3.5.1 (www.r-project.org) using library
‘Rdistance’ ver. 2.1.3 (McDonald et al. 2019). For each group,
we tested whether thermal cameras altered the detection function,
Legend
Spotlighting transects
Arid Recovery fences
Fig. 1. Map of the Arid Recovery reserve in central South Australia, including the boundary fence (black line with
marks) and spotlighting transects where thermal cameras were also used (white dashed lines).
Thermal cameras and arid zone mammals Australian Mammalogy C
and the best formof assuming the detection function. For this, we
tested four models that included whether an animal was detected
using a thermal camera, a model without this variable, then a
version of both with either a half-normal pattern (representing a
gradual decline in detections and a slight ‘tail’ at greater
distances) or negative-exponential (represents a rapid decline
but low probabilities of detecting distant animals). The top model
was selected using Information Theory by calculating Akaike
weights corrected for small sample sizes (AICc) and selection of
the model with the lowest score (Burnham and Anderson 1998).
To understand how extra thermal camera sightings affect the
estimated strip width (the most critical parameter for estimating
density), we compared this from the top models above with the
same model run with only spotlight records.
Once we had estimated a density for each spotlight/thermal
camera survey, we then measuredthe effect of the thermal camera
in adding detections under different conditions. We calculated the
percentagedifference between animal detections for each transect
if thermal camera detections were added, where 0 represents no
extra benefit and 2 would represent a 200% increase in density
estimates derived from using thermal cameras. We did not use the
density estimates derived from distance sampling, as these reflect
both the count and estimated strip widths. For example, in a
transecta thermal cameramight add many sightings but these alter
the estimated strip width, and result in a lower density estimate.
However, we would have no way of knowing whether this lower
estimate was more accurate or not. Therefore, the portion of extra
sightings was a better and more tangible refection of the value of
thermal cameras. To ensure adequate sample size, we combined
the counts for all medium-sized mammals (bettongs, bilbies, and
rabbits). Then, we compared this ratio as our response variable
against nightly temperature and observer experience as predictor
variables. Temperature was recorded from the nearest Bureau of
Meteorology weather station at the start of each transect (Olympic
Dam Aerodrome, www.bom.gov.au/products/IDS60801/). We
used generalised linear models in R and compared four models
(one with each variable, one with both, and null).
Finally, we tested the accuracy of density estimates for rabbits
in an area with a known inferred density. This was a fenced
paddock of 37 km
2
, where a rabbit reduction experiment was
conducted (see McGregor et al. 2019). We removed 2215
individuals within one month.Two lines of evidence (track counts
and spotlight counts) agreed in suggesting that rabbit activity
declined by 80% as a result of thisremoval. We therefore inferred
that the population consisted of approximately 2769 rabbits
before the removal. Two transect counts were conducted in the
three months preceding this removal, and we compared density
estimated from spotlighting counts alone, and with a thermal-
camera derived correction factor, with the known density.
Results
From 2016 to 2018, we conducted 20 spotlight/thermal camera
surveys in which mammals were recorded both by an observer
using a spotlight and by an observer using a thermal camera. On
average, transects were 14.5 km (minimum 8.7 km to maximum
18.6 km), covering a total 289.2 km. In total, 853 animals were
sighted: 258 rabbits, 185 burrowing bettongs, 66 cats, 269 small
mammals, 41 red kangaroos, 28 greater bilbies, 3 dingoes, 2
western barred bandicoots, and 1 fox. Thermal cameras resulted
in many additional sightings of animals compared with
spotlighting (Table 1).
The addition of thermal camera detections altered the
distance-detection curves for many species. Detections of Euro-
pean rabbits were best described by a half-normal function
(Table 2). Thermal cameras altered the form of the relationship
by showing relatively fewer detections at greater distance
(Fig. 2). Detections of feral cats were best described by a
Table 1. Detections of different species during spotlighting, along with the number of extra detections contributed by the thermal camera
Spotlight detections Additional thermal camera detections % increase in detections
European rabbits 221 31 14
Feral cats 66 0 0
Burrowing bettong 142 43 23
Small mammals 228 41 15
Greater bilby 25 3 11
Red kangaroo 6 3 33
Table 2. Details of the top models of distance functions for animal groups with enough detections
Top models are those with the lowest AICc scores for each group. ‘Funct’ refers to the form of the detection function (HN, half-normal; NE, negative-
exponential), ‘Thermal’ refers to whether the model included a variable relating to whether the animal was detected by thermal cameras, and ‘Prob. det’ is the
probability of detection
Species Funct Thermal Prob. det y PDAICc
European rabbits HN Yes 0.49 18.5 ,0.001 0.62
Feral cats HN No 0.42 10.6 ,0.001 0.97
Burrowing bettongs NE Yes 0.16 –17.3 ,0.001 0.89
Small mammals NE Yes 0.10 –15.6 ,0.001 1.00
DAustralian Mammalogy H. McGregor et al.
half-normal function (Table 2), and there were no extra thermal
camera detections. Detections of both burrowing bettongs and
small mammals were best described by a negative-exponential
function (Table 2), and thermal cameras increased detections at
greater distances. When the top models were then rerun with
only spotlight data, the effective strip widths were similar for
rabbits, bettongs and small mammals (Table 3).
We calculated the proportion of sightings that thermal
cameras added to each transect of medium-sized mammals
and small mammals, and found top models for both groups
included ambient temperature (i.e. had the lowest AICc values:
Table 4). User experience was present in the top model for small
mammals, but a model that did not include this variable also had
strong support, suggesting it is uninformative. Both top models
revealed a strong significant relationship between the proportion
of extra detections added by the thermal camera and ambient
temperature, for both medium-sized mammals (coef ¼–0.02,
t¼–4.89, d.f. ¼0.16,18, P¼0.001) (Fig. 3) and small mammals
Distance (m)
Probability of detection
0
0
0.2
0.4
0.6
0.8
1.0
20 40 60
Spotlight
Thermal
Spotlight
Thermal
Spotlight
Thermal
80
Distance (m)
Feral cat European rabbit
Burrowing bettong Small mammal
Probability of detection
0
0
0.2
0.4
0.6
0.8
1.0
50 100 150 200 250 300
Distance (m)
Probability of detection
0
0
0.2
0.4
0.6
0.8
1.0
20 40 60 80 100 120
Distance (m)
Probability of detection
0
0
0.2
0.4
0.6
0.8
1.0
10 20 30 40
Fig. 2. Detection functions of the top models for each animal type, including the different forms from thermal (solid)
and spotlighted (dashed) detections. No extra feral cats were detected by thermal cameras, so there was no difference in
thermal prediction for this species.
Table 3. Differences in the estimated strip width and detection probability of the top models of detection functions (Table 2) when both thermal
camera and spotlight detections are included (Both), or when applied to only the spotlight detections (Spotlight)
Species Estimated strip width Detection probability
Both Spotlight Both Spotlight
European rabbits 61 61.8 0.49 0.5
Burrowing bettongs 11 8.9 0.13 0.09
Small mammals 4.4 3.7 0.1 0.08
Thermal cameras and arid zone mammals Australian Mammalogy E
(coef ¼–0.03, t¼–2.34, d.f. ¼1.18,14, P¼0.03) (Fig. 3). This
model suggested that when temperature was 108C that the
thermal camera detected 31% more medium-sized mammals
(95% CI ¼26–34%) and 48% more small mammals (95%
CI ¼34–59%) than standard spotlight counts, but this benefit
decreased by half by 208C, and there were no additional detec-
tions when temperatures were above 308C.
The accuracy of density estimates for rabbits in a fenced
paddock was tested. As transects were conducted during
temperatures above 258C, we were not able to incorporate
thermal camera detections. Using the top detection functions
for rabbits (Table 2,Fig. 2) based on spotlighting only,
the two spotlighting transects conducted before the rabbit
removal (November 2016 and January 2017) estimated 2118
Table 4. Information Theory selection table for comparing models of the portion of extra sightings that thermal cameras added, against ambient
temperature (Temp) and whether the thermal camera user was experienced (Exp)
Top models were considered as those with the lowest D; logLik is the log-likelihood of model
Model description d.f. logLik AICc DWeight
Small mammals Temp and Exp 4 0.44 10.8 0 0.54
Temp 3 –1.87 11.7 0.98 0.33
Null 2 –4.63 14.2 3.42 0.1
Exp 3 –3.99 16 5.23 0.04
Medium mammals Temp 3 19.43 –31.4 0 0.73
Temp and Exp 4 20.02 –29.4 1.98 0.27
Null 2 10.95 –17.2 14.15 0
Exp 3 12.23 –17 14.39 0
Medium mamals
Small mamals
Maximum daily temperature (°C)
Maximum daily temperature (°C)
Portion extra detections
10
0
0.1
0.2
0.3
0.4
0.5
Portion extra detections
0
0.25
0.50
0.75
1.00
15 20 3025 35
10 15 20 3025 35
Fig. 3. Relationships between the proportion of extra detections provided by thermal cameras to spotlight surveys against maximum daily temperature for
medium sized mammals (0.5–3 kg) and small mammals (,0.5 kg). Grey shading represents 95% confidence intervals.
FAustralian Mammalogy H. McGregor et al.
(95% CI ¼2005–2232) and 2061 rabbits (95% CI ¼1951–
2172) inside the paddock. If we assume, based on Fig. 3,that
31% were missed in spotlighting, the population estimates
become 2775 and 2700 rabbits (95% CI ¼2430–3082) from
the two surveys, much closer to the independent estimate of
population size and approximately 2769.
Discussion
We found that thermal cameras increased detection rates of
small mammals by approximately 50% and medium-sized
mammals by approximately 30%, when ambient temperatures
were 208C or lower. This result agrees with other studies that
have found a similar improvement in detection (Gill et al. 1997).
Although there were differences in detection functions between
thermal cameras and spotlighting, these would not alone account
for approximately 30–50% additional detections. We propose
that most additional detections were due to thermal cameras
accounting for animals that by positional circumstances would
not normally be detected (e.g. not facing the observer and cre-
ating no eyeshine). The approximately 30% improvement
reported here was almost identical to the increase in the detec-
tion rates of kangaroos reported when combining a thermal
camera with human observers in aerial censuses. However, the
improvement in rabbit detections found with thermal cameras
was lower than the 60% increase of Focardi et al. (2001). This
could be due to differences in habitat. Because we conducted
surveys in an area with fewer shrubs and less-dense ground
vegetation than Focardi et al. (2001), detection rates of rabbits
by spotlighting would have been higher.
We found that thermal cameras resulted in different detection
functions compared with spotlights, shifting the average detec-
tion distance for most species. Broadly, we found that thermal
cameras increased sightings of animals in the 5–50 m range,
which we suspect reflects the limit of thermal camera use for
animals of such sizes. With more detections added within this
distance band, this altered the detection functions and estimated
strip widths of species relative to where most of their spotlight
detections were located. Most spotlight detections of small
mammals and bettongs were close to the road (,10 m), so
thermal cameras increased detections at intermediate distances,
expanding the detection functions/estimated strip widths. Yet
most spotlight detections of rabbits were over a wider area with
many beyond 50 m, so the addition of thermal camera detections
inversely decreased the estimated strip widths. We recorded no
benefit of using thermal cameras to detect cats, most likely for
the same reason. Their eyeshine is so bright and can be seen from
distances of up to 310 m (in this study), so there was little
potential benefit of having extra sightings in the 5–50 m band.
This suggests that while thermal cameras can add extra sightings
for certain species, these are focused at intermediate distances
from the road (5–50 m) and thermal cameras provide little
benefit either at very close or long range or for species with
bright eyeshine.
We were able to achieve our aims and provide an under-
standing of where and when thermal cameras might be of use;
however, we were not able to investigate the statistical implica-
tions within distance sampling. By necessity, we pooled detec-
tions from multiple transects to obtain detection functions.
These transects spanned areas and times with vastly different
animal assemblages, so we were not able to assume density as
stable and investigate the full implications of model parameters
on density estimates. However, our research does suggest
areas for consideration. Of particular concern would be the
interplay between increasing detection rates, temperature and
detection functions and estimated strip widths. We demonstrate
a substantial linear relationship between the former two, but had
insufficient data to examine the latter. As estimated strip width is
the most important variable affecting density estimates
(Buckland et al. 2001), further research for each species would
be required before accurate inferences of density can be made
(e.g. Augusteyn et al. 2020).
Our results might be biased by our study circumstances, and
not completely relevant elsewhere. The vast majority of the cat
detections reported here were from a fenced paddock that had a
high density of feral cats, with no major mammalian predators
(e.g. dingos, foxes) nor any active conservation hunting for at
least two years prior to our surveys. This period of safety could
have meant that the cats were less wary, less likely to avoid
vehicles and lights, and therefore more detectable by spotlight
survey. In situations where cats are more cryptic, thermal
cameras may have a larger effect on detectability of cats but
only at short to medium distances.
As expected, ambient temperature had a strong effect on the
performance of thermal cameras relative to standard spotlight
counts. At night-time temperatures above 308C, thermal cam-
eras provided almost no additional benefit, and would often
detect virtually no animals at all. Even at temperatures above
208C, there were substantially fewer detections. This agrees with
Vinson et al. (2020), who recorded similar results for Australian
arboreal mammals. This means that thermal cameras should not
be used exclusively for long-term monitoring programs that
include warmer months, or in areas where minimum nightly
temperatures regularly exceed 308C. In such instances, variabil-
ity in detection rates would more likely relate to temperature
than differences in mammal abundance or activity. Using
thermal cameras in dense forests should not have such limita-
tions as background temperatures are typically below 308C
(Burke et al. 2019;Kays et al. 2019), yet for many open habitats
there may be seasons through the year when thermal cameras are
not useful. Spotlighting has no such limitations and thus may be
a more temporally versatile method of monitoring.
Although thermal cameras were not ideal in warm tempera-
tures for small to medium sized mammals (20–5000 g) at our
study site, they were a valuable addition for vehicle-based
spotlighting counts in cooler temperatures. Not only did our
use of thermal cameras result in higher total detections, they also
provided a more accurate estimation of the absolute size of an
independently censused rabbit population. We suggest that other
studies could also use thermal cameras to provide a greater
understanding of whether spotlighting might be underestimating
detections of monitored mammal species. Thermal cameras
would be especially useful for studies where detecting a large
proportion of the population of a small to medium sized species
is critical, for example in surveys to locate populations of
threatened species where every detection is important or for
monitoring populations at low density where population esti-
mates have large confidence intervals.
Thermal cameras and arid zone mammals Australian Mammalogy G
Conflict of interest
The authors declare no conflicts of interest.
Acknowledgements
Arid Recovery hosted and supported this project. Arid Recovery is an
independent conservation and research initiative supported by BHP, The
University of Adelaide, Bush Heritage Australia and The South Australian
Department for Environment and Water. Samantha Bryson-Kirby, Guy
Nelson, Monica Griffiths, Nathan Beerkens, Emily Gregg, and Melissa
Jensen assisted with field data collection. Katherine Tuft and Jose´ Joaquı´n
Lahoz-Monfort assisted with project development. HMcG and SL are sup-
ported by the National Environmental Program’s Threatened Species
Recovery Hub. This research did not receive any specific funding.
References
Augusteyn, J., Pople, T., and Rich, M. (2020). Evaluating the use of thermal
imaging cameras to monitor the endangered greater bilby at Astrebla
Downs National Park. Australian Mammalogy 42, 329–340.
doi:10.1071/AM19040
Buckland, S. T., Anderson, D. R., Burnham, K. P., Laake, J. L., Borchers,
D. L., and Thomas, L. (2001). ‘Introduction to Distance Sampling:
Estimating Abundance of Biological Populations.’ (Oxford University
Press: Oxford, UK.)
Burke, C., Rashman, M., Wich, S., Symons, A., Theron, C., and Longmore,
S. (2019). Optimizing observing strategies for monitoring animals using
drone-mounted thermal infrared cameras. International Journal of
Remote Sensing 40, 439–467. doi:10.1080/01431161.2018.1558372
Burnham, K. P. and Anderson, D. R. (1998). ‘Model Selection and Multi-
model Inference: A Practical Information-theoretic Approach.’ 2nd edn.
(Springer Science & Business Media, Inc.)
Cilulko, J., Janiszewski, P., Bogdaszewski, M., and Szczygielska, E. (2013).
Infrared thermal imaging in studies of wild animals. European Journal of
Wildlife Research 59, 17–23. doi:10.1007/S10344-012-0688-1
Focardi, S., De Marinis, A. M., Rizzotto, M., and Pucci, A. (2001).
Comparative evaluation of thermal infrared imaging and spotlighting
to survey wildlife. Wildlife Society Bulletin 133–139.
Gerken, M. (1996). Application of infrared thermography to evaluate
the influence of the fibre on body surface temperature in llamas.
In ‘Proceedings of the 2nd European Symposium on South American
Camelids’. Vol. 30.
Gill, R. M. A., Thomas, M. L., and Stocker, D. (1997). The use of portable
thermal imaging for estimating deer population density in forest habitats.
Journal of Applied Ecology 34, 1273–1286. doi:10.2307/2405237
Goldingay, R. L., and Sharpe, D. J. (2004). How effective is spotlighting for
detecting the squirrel glider? Wildlife Research 31, 443–449.
doi:10.1071/WR03010
Kays, R., Sheppard, J., McLean, K., Welch, C., Paunescu, C., Wang, V.,
Kravit, G., and Crofoot, M. (2019). Hot monkey, cold reality: surveying
rainforest canopy mammals using drone-mounted thermal infrared
sensors. International Journal of Remote Sensing 40, 407–419.
doi:10.1080/01431161.2018.1523580
Legge, S., Robinson, N., Lindenmayer, D., Scheele, B., Southwell, D., and
Wintle, B. (2018). ‘Monitoring Threatened Species and Ecological
Communities.’ (CSIRO Publishing: Melbourne.)
Lindenmayer, D., Cunningham, R., Donnelly, C., Incoll, R., Pope, M.,
Tribolet, C., Viggers, K., and Welsh, A. (2001). How effective is
spotlighting for detecting the greater glider (Petauroides volans)?
Wildlife Research 28, 105–109. doi:10.1071/WR00002
McDonald, T. L., Carlisle, J., and McDonald, A. (2019). Rdistance:
Distance-Sampling Analyses for Density and Abundance Estimation.
R package version 2.1.3. Available at https://CRAN.R-project.org/
package¼Rdistance
McGregor, H., Moseby, K. E., Johnson, C. N., and Legge, S. M. (2019). The
short-term response of feral cats to rabbit population decline: are
alternative native prey more at risk? Biological Invasions 22, 799–811.
doi:10.1007/S10530-019-02131-5
Morrison, P. (1962). Body temperatures in some Australian mammals. II.
Peramelidae. Australian Journal of Biological Sciences 15, 386–394.
doi:10.1071/BI9620386
Moseby, K. E., and Read, J. L. (2006). The efficacy of feral cat, fox
and rabbit exclusion fence designs for threatened species protection.
Biological Conservation 127, 429–437. doi:10.1016/J.BIOCON.2005.
09.002
Moseby, K. E., Hill, B. M., and Read, J. L. (2009). 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
Moseby, K. E., Neilly, H., Read, J. L., and Crisp, H. A. (2012). Interactions
between a top order predator and exotic mesopredators in the Australian
rangelands. International Journal of Ecology 2012, 1–15. doi:10.1155/
2012/250352
Page, M., Kuiper, J., Kabat, A., and Legge, S. (2013). A way to reduce
interference with Elliott traps. Australian Mammalogy 35, 128–130.
doi:10.1071/AM11045
Ross, A. K., Letnic, M., Blumstein, D. T., and Moseby, K. E. (2019).
Reversing the effects of evolutionary prey naivete´ through controlled
predator exposure. Journal of Applied Ecology 56, 1761–1769.
doi:10.1111/1365-2664.13406
Simpson, K., Johnson, C. N., and Carver, S. (2016). Sarcoptes scabiei: the
mange mite with mighty effects on the common wombat (Vombatus
ursinus). PloS One 11, e0149749. doi:10.1371/JOURNAL.PONE.
0149749
Vinson, S. G., Johnson, A. P., and Mikac, K. M. (2020). Thermal cameras as
a survey method for Australian arboreal mammals: a focus on the
greater glider. Australian Mammalogy 42, 367–374. doi:10.1071/
AM19051
Wayne, A., Rooney, J., Morris, K., and Johnson, B. (2008). Improved bait
and trapping techniques for chuditch (Dasyurus geoffroii): overcoming
reduced trap availability due to increased densities of other native fauna.
Conservation Science Western Australia 7, 49–56.
West, R., Letnic, M., Blumstein, D. T., and Moseby, K. E. (2018). Predator
exposure improves anti-predator responses in a threatened mammal.
Journal of Applied Ecology 55, 147–156. doi:10.1111/1365-2664.12947
www.publish.csiro.au/journals/am
HAustralian Mammalogy H. McGregor et al.
... An increasing number of studies have highlighted the potential of thermal imaging as an effective, non-invasive method of mammal wildlife census (Vinson et al. 2020;McGregor et al. 2021;Pocknee et al. 2021). This technology enables the operator to locate an individual or object with a temperature of above absolute zero by detecting the heat signature it emits through infrared radiation (for a full explanation see (McCafferty 2013;Sharp and Havens 2015), with operators using a hand-held or vehicle/dronemounted thermal imaging device depending on survey location and/or the target species (Havens and Sharp 1998;Kays et al. 2019;McGregor et al. 2021). ...
... An increasing number of studies have highlighted the potential of thermal imaging as an effective, non-invasive method of mammal wildlife census (Vinson et al. 2020;McGregor et al. 2021;Pocknee et al. 2021). This technology enables the operator to locate an individual or object with a temperature of above absolute zero by detecting the heat signature it emits through infrared radiation (for a full explanation see (McCafferty 2013;Sharp and Havens 2015), with operators using a hand-held or vehicle/dronemounted thermal imaging device depending on survey location and/or the target species (Havens and Sharp 1998;Kays et al. 2019;McGregor et al. 2021). ...
... This thermal technology has successfully out-performed other detection methods for a range of mammal species globally, including mature and neo-natal deer in the temperate regions of the US and the UK (Gill et al. 1997;Ditchkoff et al. 2005;Hodnett 2005;Collier et al. 2007), rabbits, foxes and polar bear dens in the Arctic (Boonstra et al. 1994;Amstrup et al. 2004), lagomorphs and wild boar (Focardi et al. 2001), as well as native and introduced terrestrial mammals in arid Australia (McGregor et al. 2021). ...
Article
Full-text available
This study developed and tested the efficacy of a real-time thermography technique to improve survey methods for Australian arboreal mammal species, with a focus on the greater glider. Development involved the use of thermal imaging cameras combined with spotlight transect surveys of an endangered greater glider population at Seven Mile Beach National Park. Over 30h of nocturnal spotlight transect surveys were undertaken over 14 (1km) transects within 70ha of dry sclerophyll forest. A protocol for the use of thermography to survey greater gliders was developed. The efficacy of the thermography protocol was then experimentally tested in comparison to traditional spotlighting. Overall, thermography was better at detecting arboreal mammals than spotlighting (P<0.05). However, the effect was not significant for greater gliders (P=0.79), even though there was a trend towards improved detection of the species using thermal cameras. Thermography is a novel approach to undertaking arboreal mammal surveys and future studies should consider its relevance, effectiveness and associated costs to improve survey designs, especially for threatened species.
Article
Full-text available
The impacts of invasive predators can be amplified by high densities of invasive prey species. In Australia, hyper abundant rabbit populations lead to high densities of feral cats and correspondingly high impact of cats on native species, especially small mammals. Therefore, it would be expected reducing rabbits could also reduce abundance of cats, and thereby alleviate predation on native small mammals. However, cats might respond to the loss of rabbits by prey-switching to native small mammals, resulting in increased predation on those species. Our aim was to understand the short-term effects of an experimental reduction of rabbit abundance on feral cats and their small-mammal prey in arid South Australia. We reduced the rabbit population in a 37 km² experimental enclosure by ~ 80%, while monitoring an adjacent unmanipulated area as a control. Cat activity and survival of VHF-collared cats in the enclosure decreased by 40% following the rabbit reduction. Surviving cats increased their consumption of reptiles, birds and invertebrates, but they nevertheless evinced hunger by increased intake of experimentally-supplied sausages. There was no change in either the proportion of cat scats that contained remains of small mammals, or the rate at which video-collared cats were recorded killing small mammals, even though the activity of small mammals declined. Our results demonstrate that individual feral cats prey-switch in response to removal of their primary prey. However, we also show that survival and overall activity of cats decreased, which could result in net, long-term benefits for native prey threatened by cats. Management of feral cats using food lures or baits would also be more effective when introduced prey are scarce, as cats are more likely to eat novel food.
Article
Full-text available
Inappropriate anti‐predator responses (naiveté) towards introduced predators is a key factor contributing to the extinction and endangerment of prey species world‐wide and the failure of wildlife reintroductions. Here, we test the idea that success of reintroduction can be improved by exposing a predator naïve prey species to introduced predators under controlled conditions (in situ predation) prior to reintroduction, such that prey adopt increased wary behaviours to aid in survival. We exposed a population of a naïve marsupial, the greater bilby (Macrotis lagotis), to a controlled number of introduced predators (feral cats, Felis catus) for 2 years within a large fenced paddock and then compared the pre‐release behaviour and post‐translocation survival of predator‐exposed and predator‐naïve bilbies over 40 days. Behavioural assays conducted in a small pen prior to reintroduction suggested that predator‐exposed bilbies were warier as they spent less time moving and more time in cover than predator naïve bilbies. After translocation, predator‐exposed bilbies were more likely to survive to 40 days and were less likely to be preyed upon by cats than predator‐free bilbies. Synthesis and applications. Naiveté towards predators is a major problem thwarting successful reintroductions world‐wide. Our study demonstrates that exposure to predators under controlled conditions can increase survival of reintroduced prey and is a promising approach to overcome the problem of naiveté towards introduced predators and the global problem of prey naiveté. Future conservation of naïve prey species may depend on such training methods prior to releasing into areas where predators are present.
Article
Full-text available
Animals of the rainforest canopies are often endangered by deforestation or hunting but are difficult to survey and study because of the inaccessibility of the treetops, combined with the visual camouflage of many species. Drone-based thermal sensors have the potential to overcome these hurdles by rapidly scanning large forested areas from above, detecting and mapping wildlife based on the contrast between their warm body temperatures and the cool tree canopies. We tested this concept by flying a drone-mounted thermal infrared radiometric sensor over the wildlife-rich rainforests of Barro Colorado Island, Panama. Arboreal mammals had body temperatures around 27°C and were conspicuous in the thermal infrared imagery at night and early morning when the forest canopy was cool (23–25°C), but were difficult to detect by mid-morning, by which time the direct sunshine had heated up canopy vegetation to over 30°C. Species were difficult to identify from thermal infrared imagery alone, but could be recognized from synchronized visual images taken during the daytime. Simultaneous drone and ground-based surveys of the same area proved that the aerial thermal camera did detect high-canopy species missed by the terrestrial observer, but that substantially more animals were detected by the human than by the drone. Because animal detection was so much better at night, when species ID was difficult, we suggest that future work could combine automated detection of animals from thermal infrared imagery with flash photography or IR illumination to enable species ID during nocturnal surveys. We conclude by discussing various logistical challenges that limit the utility of drone-based thermal infrared today, but that could be overcome by continued improvement of technology and collaboration with permitting agencies.
Article
Full-text available
Incorporating an understanding of animal behaviour into conservation programmes can influence conservation outcomes. Exotic predators can have devastating impacts on native prey species and thwart reintroduction efforts, in part due to prey naïveté caused by an absence of co-evolution between predators and prey. Attempts have been made to improve the anti-predator behaviours of reintroduced native prey by conducting laboratory-based predator recognition training but results have been varied and have rarely led to improved survival in reintroduction programmes. We investigated whether in situ predator exposure could improve anti-predator responses of a predator-naïve mammal by exposing prey populations to low densities of introduced predators under controlled conditions. We reintroduced 352 burrowing bettongs to a 26-km² fenced exclosure at the Arid Recovery Reserve in South Australia and exposed them to feral cats (density 0.03-0.15 cats/km²) over an 18-month period. At the same time, we translocated a different group of bettongs into an exclosure free of introduced predators, as a control. We compared three behaviours (flight initiation distances, trap docility and behaviour at feeding trays) of cat-exposed and control bettongs before the translocations, then at 6, 12 and 18 months post-translocation. Cat-exposed bettongs displayed changes in behaviour that suggested increased wariness, relative to control bettongs. At 18 months post-reintroduction, cat-exposed bettongs had greater flight initiation distances and approached feed trays more slowly than control bettongs. Cat-exposed bettongs also increased their trap docility over time. Synthesis and applications. Translocation is recommended as a conservation tool for many threatened species yet success rates are generally low. We demonstrate that controlled levels of in situ predator exposure can increase wariness in the behaviour of naïve prey. Our findings provide support for the hypothesis that in situ predator exposure could be used as a method to improve the anti-predator responses of predator-naïve threatened species populations.
Article
Full-text available
Parasitism has both direct and indirect effects on hosts. Indirect effects (such as behavioural changes) may be common, although are often poorly described. This study examined sarcoptic mange (caused by the mite Sarcoptes scabiei) in the common wombat (Vombatus ursinus), a species that shows severe symptoms of infection and often causes mortality. Wombats showed alterations to above ground behaviours associated with mange. Infected wombats were shown to be active outside of the burrow for longer than healthy individuals. Additionally, they spent more time scratching and drinking, and less time walking as a proportion of time spent above ground when compared with healthy individuals. They did not spend a higher proportion of time feeding, but did have a slower feeding rate and were in poorer body condition. Thermal images showed that wombats with mange lost considerably more heat to the environment due to a diminished insulation layer. Infection status did not have an effect on burrow emergence time, although this was strongly dependent on maximum daily temperature. This study, through the most detailed behavioural observations of wombats to date, contributes to a broader understanding of how mange affects wombat health and abundance, and also to our understanding of the evolution of host responses to this parasite. Despite being globally dispersed and impacting over 100 species with diverse intrinsic host traits, the effects of mange on hosts are relatively poorly understood, and it is possible that similar effects of this disease are conserved in other host species. The indirect effects that we observed may extend to other pathogen types.
Article
Spotlight surveys are widely used to monitor arid-zone-dwelling species such as the greater bilby (Macrotis lagotis). These surveys require a sufficient sample size to adequately model detection probability. Adequate sample sizes can be difficult to obtain for low-density populations and for species that avoid light and or have poor eyeshine like the bilby. Abundance estimates based on burrow counts can be problematic because of the variable relationship between the number of burrows used and bilby abundance. In 2013, feral predators devastated a Queensland bilby population and a method was required that could locate and monitor the remaining bilbies. We report on a study that compared density estimates derived from spotlighting and thermal cameras. Bilbies were surveyed annually over three years, using spotlights and thermal cameras on different nights but using the same transects to compare the methods. On average, thermal cameras detected twice the number of bilbies per kilometre surveyed than spotlighting. Despite this difference in the number of bilbies detected, density estimates (bilbies km−2) were similar (thermal camera versus spotlight: 0.6 versus 0.2 (2014), 3.4 versus 3.4 (2015) and 4.8 versus 3.3 (2016)). Nevertheless, the larger sample size obtained using thermal cameras gave greater confidence in modelling detection probability.
Article
The proliferation of relatively affordable off-the-shelf drones offers great opportunities for wildlife monitoring and conservation. Similarly the recent reduction in the cost of thermal infrared cameras also offers new promise in this field, as they have the advantage over conventional RGB cameras of being able to distinguish animals based on their body heat and being able to detect animals at night. However, the use of drone-mounted thermal infrared cameras comes with several technical challenges. In this article, we address some of these issues, namely thermal contrast problems due to heat from the ground, absorption and emission of thermal infrared radiation by the atmosphere, obscuration by vegetation, and optimizing the flying height of drones for a best balance between covering a large area and being able to accurately image and identify animals of interest. We demonstrate the application of these methods with a case study using field data and make the first ever detection of the critically endangered riverine rabbit (Bunolagus monticularis) in thermal infrared data. We provide a web-tool so that the community can easily apply these techniques to other studies (http://www.astro.ljmu.ac.uk/aricburk/uav_calc/).
Article
Body temperature measurements on the short-nosed bandicoot (Thylacis obeaulus) have shown a nocturnal cycle with a range of 1� 2�C and a short active phase at 2200-0400 hr. The bilby or rabbit bandicoot (Macrotis lagoti8) had a sharply defined temperature cycle, with a range of almost 3�C after several months of captivity, during which the day-time resting temperature was progressively lowered from 36� 4 to 34� 2�C. Forced activity raised the diurnal temperature substantially but not to the nocturnal level. Forced activity did not raise the nocturnal level which was similar in the two species (37' O�C). Both species could regulate effectively at an ambient temperature of 5�C, but only Thylaci8 showed regulation at ambient temperatures of between 30 and 40�C.
Article
Elliott traps are widely used in Australia to capture small terrestrial mammals. However, in some situations non-target species disturb the traps, resulting in fewer traps being available for the target species. This situation occurred on Faure Island Wildlife Sanctuary (Western Australia) where boodies (burrowing bettongs) were disturbing the Elliott traps deployed to monitor populations of shark bay mice and western barred bandicoots. This note presents the method used to alleviate this problem on Faure Island but would be more widely applicable in other situations where Elliott traps are disturbed by animals other than the target species.