Estimating occurrence and detectability
of a carnivore community in eastern
Botswana using baited camera traps
Lauren C. Satterfield1,2*, Jeffrey J. Thompson3,4, Andrei Snyman5,7,
Luis Candelario1, Brian Rode6& John P. Carroll7
1University of Georgia, Warnell School of Forestry and Natural Resources, 180 E. Green Street, Athens, GA 30602, U.S.A.
2University of Washington, School of Environmental and Forest Sciences, Box 352100, Seattle, WA 98195-2100, U.S.A.
3Guyra Paraguay, Parque Ecológico Asunción Verde, Avda, Carlos Bóveda, Asunción, Paraguay
4National Science and Technology Council (CONACYT) of Paraguay, Justo Prieto 223 esq. Teófilo del Puerto,
Barrio Villa Aurelia, Asunción, Paraguay
5Northern Tuli Predator Project, Mashatu Private Game Reserve, Northern Tuli, P.O. Box 26, Lentswe Le Moriti, Botswana
6BC Field Guide Services (Pty) Ltd., Private Bag X3008, Suite 75, Hoedspruit, 1380 South Africa
7University of Nebraska-Lincoln, School of Natural Resources, 3310 Holdrege St., P.O. Box 830989, Lincoln, NE 68583, U.S.A.
Received 21 June 2016. To authors for revision 2 August 2016. Accepted 3 February 2017
Human–wildlife conflict and habitat loss are threatening carnivore populations in southern
Africa, where the bulk of research focuses on large predators. However, scant research
exists on medium and small carnivore (mesocarnivore) ecology. We employed hierarchical
community modelling to estimate the effect of habitat on species occurrence and the effect
of bait on detection probabilities for the carnivore community in the Mashatu Game Reserve,
Botswana. Wetested sites baited with either impala (
) meat (meat sites)
or cheesecloth soaked in used cooking fat (fat rag sites) against unbaited sites (control
sites). Within each bait classification, we divided our sampling effort between two habitat
classifications, riverine and non-riverine sites. Thirteen of 16 carnivore species inhabiting
the area (81%), including 10 of 12 species of mesocarnivore (83%), were recorded. Occu-
pancy rates were higher in riverine habitat for several species, in particular African civet
), brown hyaena (
), and large-spotted genet (
), demonstrating the importance of riverine habitat, which is declining in the study
region. Our results suggest that the use of bait improves detectability. Several large carni-
vores, including spotted hyaena (
), brown hyaena, and leopard (
), were detected at highest rates at meat sites. Many mesocarnivores, including
black-backed jackal (
) and African civet responded equally to meat and fat
rag sites, with detections greater than at control sites.Notably, large-spotted genet showed
highest detection rates at fat rag sites, and brown hyaena showed higher rates at fat rag sites
than control sites. Our detection results indicate that spent cooking fat may be used as an
effective bait alternative to meat when studying mesocarnivore communities in southern
Africa, whereas meat may still be the most effective for studying large carnivore communi-
ties. Identifying effective methods of increasing detection rates will benefit carnivore survey
and monitoring initiatives, especially for cryptic species.
Key words: Africa, bait, Botswana, camera trap, Carnivora, carnivore, cooking fat, detection,
habitat, Mashatu, meat, occupancy.
With large carnivore populations in decline glob-
ally, continued research on effective conservation
and management strategies is vital for stabilizing
and protecting populations (Ray, Hunter &
Zigouris, 2005; Ripple
, 2014). These declines
are of particular concern due to the potential role
carnivores play in altering ecosystem dynamics
through trophic interactions (Estes
, 2014). Consequently, a thorough
understanding of carnivore community ecology is
essential for the conservation of not only carnivore
species, but also the ecosystems that they inhabit.
In southern Africa, extensive research of large
carnivores has a long and enduring history (Caro &
Stoner, 2003; Ogada, Woodroffe, Oguge & Frank,
, 2005; Dalerum, Somers, Kunkel
& Cameron, 2008; Snyman, Jackson & Funston,
*To whom correspondence should be addressed.
African Journal of Wildlife Research 47(1): 32–46 (April 2017)
ISSN 2410-7220 [Print], ISSN 2410-8200 [Online] — DOI: http://dx.doi.org/10.3957/056.047.0032
2015). However, studies on medium and small
carnivores (hereafter mesocarnivores) in the region
are conspicuously lacking given their ecological
importance and the precarious conservation
status of some species (Martinoli
There is an important need for the development
of efficient, cost-effective, non-invasive sampling
methodologies which adequately survey the major-
ity of species within the carnivore communities of
southern Africa. This need is due to the relative
rarity and cryptic nature of most carnivores, scant
research resources, and often restrictive regula-
tions on data collection in southern Africa’s wildlife
parks and game reserves. Advancement in digital
photography and remote camera (camera trap)
technology meets these needs as it allows for
sampling methods that are minimally invasive, and
low in cost and manpower, while generating high-
quality data (Rowcliffe & Carbone, 2008).
To increase the effectiveness of data collection
using camera trapping, species- or guild-specific
baits are often used to selectively draw carnivore
species to camera trap stations.A literature review
of baits used to lure carnivores yielded the follow-
ing options: peanut butter, eggs, fruit, fish, meat,
offal, blood, live animals, pet food, Calvin Klein
Obsession aftershave, synthetic fatty acid lures,
and carnivore urine (Wilson & Delahay, 2001;
Barea-Azcón, Virgós, Ballesteros-Duperón,
Moleón & Chirosa, 2006; Thorn, Scott, Green,
Bateman & Cameron, 2009; Gil-Sánchez
2011). Many commercial baits, such as carnivore
urine, are difficult to obtain in remote areas of
southern Africa. Others, such as meat, can be
impractical or unsafe to keep in a field camp setting
where refrigeration is scarce. Furthermore, stored
baits have the potential to draw large predators to
field camps and thus into proximity with people.
It is important to consider how baits might influ-
ence behaviour, and thus whether or not bait is
appropriate for the goal of a particular study.
Sampling designs have the potential to signifi-
cantly impact species detectability (Mackenzie &
Royle, 2005; Hamel
, 2013). Baits have the
potential to alter behaviour through age- or sex-
based divisions in attraction, avoidance, or habitu-
ation, which can further vary by individual and
thus influence heterogeneity of results (Royle,
Magoun, Gardner, Valkenburg & Lowell, 2011;
Foster & Harmsen, 2012). If not accounted for in
study design or analysis, these impacts can skew
interpretation of results. In particular, bait use has
notable potential to influence results of studies
focused demographics due to impacts on species
behaviour (Foster & Harmsen, 2012). However,
studies that rely more heavily on high detection
rates, such as live trapping, population estimation,
or monitoring of cryptic species, can benefit from
bait use (Foster & Harmsen, 2012; du Preez,
Loveridge & Macdonald, 2014). In these cases,
bait can be a valuable tool to improve study design
and save time and resources.
We investigated the importance of riverine habi-
tat to carnivores on the Mashatu Game Reserve,
Botswana. At the time of our study, the remaining
sections of large fever berry (
forest were in narrow strips of 15–100 m along
river banks. Historical photographs of the reserve
show a drastic loss of this riparian habitat over the
last 50 years. While many factors could influence
this habitat change, the leading theory is that this
loss is the direct result of an acute flood amid
sustained drought, combined with increased
) populations on the
site (O’Connor, 2010).
These habitat changes are particularly important
in light of increasing human impact on landscapes
throughout southern Africa. These influences
have significant effects on predator communities,
both direct and indirect, which can cascade
throughout trophic levels (Woodroffe, 2000;
Woodroffe, Lindsey, Romañach, Stein & ole
Ranah, 2005; Gusset, Swarner, Mponwane,
Keletile & McNutt, 2009). Therefore it is increas-
ingly important to understand the role of meso-
carnivores in these systems. Reducing stress on
populations, especially through identifying and
protecting vital habitat, can help to mitigate
anthropogenic pressures on car nivores. Regretta-
bly, research investigating the effect of human
activities on the mesocarnivore community is
scarce (Thorn, Green, Dalerum, Bateman & Scott,
, 2014), perhaps due to limited
funds for these less charismatic species and
limited research on effective sampling techniques.
Fortunately, smaller members of the order
Carnivora are gaining some research attention. A
study in the Okavango Delta of Botswana recently
showed that small mammalian species diversity,
and separately carnivore species diversity,
increases in more remote parts of conservation
areas. Results were not reported for small carni-
vores or medium carnivores specifically (Rich,
Miller, Robinson, McNutt & Kelly, 2016). However,
a study on carnivores in Kenya did investigate
species-specific impacts of environmental and
: Estimating occurrence and detectability of a carnivore community in Botswana 33
anthropogenic variables. That study found that
while environmental variables best explain carni-
vore species occupancy, impacts were demon-
strated related to human activities (Schuette,
Wagner, Wagner & Creel, 2013). These general
trends underscore the need for more effective
sampling techniques to support more detailed
In addition to assessing the importance of habi-
tats, we evaluated the efficacy of cheesecloth
soaked in used cooking fat against locally harvested
) meat in attracting
carnivores on the Mashatu Game Reserve of
eastern Botswana. Meat is a common carnivore
bait but increasingly difficult to obtain in the region
due to hunting restrictions and conflicts with
tourism activities. By contrast, used cooking fat is a
plentiful waste product found in most bush camp
kitchens, making it more accessible and less logis-
tically problematic than meat bait.
To this end, we employed hierarchical community
modelling within a Bayesian framework to estimate
the effect of habitat on occupancy probabilities
and the effect of bait type on detection probabilities
for the carnivore community in the Mashatu Game
Reserve, Botswana. We predicted that habitat
would influence species-specific occupancy rates,
and that riverine habitat would yield higher occu-
pancy rates than non-riverine sites for some
species. This hypothesis is based on years of
presence-only data taken from opportunistic wild-
life surveys conducted by local guides driving and
walking on the property, who observed several
species almost exclusively in riverine areas. We
expected that detection rates for carnivores would
likely be species-specific among baits, and that
detection rates at baited sites generally would be
higher than at unbaited sites. This hypothesis is
due to a presence of both obligate carnivores and
omnivores in the suite of predators studied, and
the predisposition of some species (e.g. brown
)) to use scent to find
food, which could increase attraction to bait (Mills,
1987; IUCN, 2013). To our knowledge this was the
first time that used cooking fat was tested as bait
on carnivores in southern Africa, and the first time
that a community occupancy analysis was con-
ducted on the study site.
MATERIALS AND METHODS
The 3300 ha study area lies in a western land
holding of the 29 000 ha Mashatu Game Reserve
(henceforth ‘Mashatu’) (22°13’S, 28°58’E), an
unfenced patchwork of private land holdings in
eastern Botswana (Fig. 1a), making it unique in
southern Africa. Mashatu lies within the Northern
Tuli Game Reserve, a 72 000 ha property bordered
on three sides by the Motloutse, Limpopo, and
Shashe rivers. Thus, while connected to vast
expanses of unfenced conservation land, the site
is in close proximity to farms, villages, and other
The region consists of rocky outcrops dispersed
on ancient floodplains along the Motloutse and
Limpopo rivers. Dominant flora on the plains
include mopane (
umbrella thorn (
), and mustard bush
). Thickets along riverbanks
contain large fever berry trees. According to rainfall
data collected by researchers on the study site
since 1996, average annual total rainfall is approxi-
mately 360 mm, with most precipitation (310 mm
on average) falling during the wet summer months
from November to April and little to no rain (50mm
on average) falling during the region’s dry winter
season from May to October. The study area
supports a suite of 16 carnivore species (Table 2).
We deployed 20 remote infra-red-triggered cam-
eras (Moultrie M80 Game Spy, EBSCO Industries
Inc., Birmingham, AL) housed in locked metal
boxes to prevent theft and damage by animals,
especially elephants, from June to August, 2013
34 African Journal of Wildlife Research Vol. 47, No. 1, April 2017
Tab le 1. Characteristics of study design and camera sampling effort for a study of carnivore occupancy on Mashatu
Game Reserve in eastern Botswana during June–August 2013.
Dates Bait type
Sites 24 h periods Effort
Riverine Non-riverine (days)
12 Jun – 02 Jul Control 415 20 180
03 Jul – 22 Jul Meat 5 5 19 190
12 Jun – 22 Jul Fat Rag 5 5 40 400
1One of the initial 10 cameras failed during the control study and had to be removed from that analysis.It was replaced with a Bushnell
camera (Bushnell Trophy Cam HD, Bushnell Outdoor Products, Overland Park, KS) for the subsequent meat study.
: Estimating occurrence and detectability of a carnivore community in Botswana 35
Fig. 1. Map of southern Africa (inset) and location of the Mashatu Game Reserve study area within the Northern Tuli
Game Reserve, Botswana (a). Location of the Mashatu Game Reserve within the Greater Mapungubwe Transfrontier
Conservation Area (inset) and the location of camera trap sites by bait and habitat classification (b). Research was
conducted during June, July, and August of 2013.
36 African Journal of Wildlife Research Vol. 47, No. 1, April 2017
Tab le 2. Species list within order Carnivora found on Mashatu Game Reserve, Botswana during June–August of 2013. Count of observations at each bait type (CT =
control, MT = meat, FR = fat rag) and, separately, within each habitat classification (RI = riverine, NR = non-riverine) are listed, along with total counts by species. Effort is
given as the cumulative number of sites-days during the study period. International Union for Conservation of Nature and Natural Resources Red List status (LC = Least
Concern, VN = Vulnerable, NT = Near Threatened, EN = Endangered) and IUCN population trend are also listed (IUCN, 2013).
CT MT FR RI NR TOTAL
(site-days) 180 190 400 375 395 770
Family Species Common name IUCN IUCN Trend
Black-backed jackal LC Stable 32 85 199 113 203 316
Bat-eared fox LC Unknown 2 03145
Felis silvestris lybica
African wildcat LC Decreasing 2 2 15 17 2 19
Leopard NT Decreasing 8 17 7 22 10 32
Slender mongoose LC Stable 1 11213
White-tailed mongoose LC Stable 0 11202
Banded mongoose LC Stable 2 34729
Spotted hyaena LC Decreasing 9 27 17 18 35 53
Brown hyaena NT Decreasing 3 39 37 63 16 79
Honey badger LC Decreasing 0 5 8 11 2 13
African civet LC Unknown 3 20 44 67 0 67
Small-spotted genet/ LC Stable 0 371910
South African large- LC Unknown 1 6 98 101 4 105
1One photograph believed to be
was recorded. However, it could not be verified due to poor resolution, and thus it was excluded from analysis.
Sample sites were unbaited (control sites), or
baited with either impala meat (meat sites) or with
rags soaked with used cooking oil (fat rag sites).
These bait classifications were evaluated by
sampling 10 sites each. Sampling occurred for 40,
20, and 19 days at fat rag, control, and meat sites,
respectively (Table 1). We were most interested in
the effectiveness of fat rags, and chose our design
to allow periods of simultaneous sampling of fat
rags with each of the other two bait types.Thus, fat
rag sites were tested at the same time as control
sites for the first half of the study, and concurrent
with meat sites for the second half of the study,
which lasted 40 days in total.
Sampling sites were placed every ~0.5 km.
Systematic random site placement was used such
that fat rag and control sites were placed alter-
nately (Fig. 1b). Fat rag sites remained in the same
locations throughout the study. After 20 days, con-
trol sites were replaced with meat sites (hung inac-
cessibility within a tree 100 m or less of previous
control sites) and sampling continued for 19 days.
This preserved ~1 km distance between sites of
the same bait classification and at least 0.5 km
between all sites throughout the study.
We were not able to sample all three bait types
simultaneously due to limited camera availability
and logistical constraints in checking camera sites.
Furthermore, our design reduced bias. For exam-
ple, if a meat site was changed to a fat rag site, it
would be difficult to determine if initial visits to the
fat rag bait were due to attraction to the fat rag bait
or memory of the meat bait. Thus, our sampling
design prevented any confounding effects result-
ing from changing a site from one bait type to an-
Placement of sampling sites was stratified by
habitat so that five of the 10 sites within each bait
classification each year were placed in riparian
forest along the Motloutse River (riverine sites),
while the other five were placed away from the r iver
in mopane, mustard bush, and umbrella thorn flats
(non-riverine sites). Thus, an equal number of
sites were sampled for each of two habitat classifi-
cations within each of three bait classifications for
all sampling periods. There was one exception for
control sites, where one camera in a riverine area
was censored due to camera malfunction (Table 1).
From researcher observation, habitat changes
abruptly from riverine to non-riverine habitat. All
riverine sites were at least 1 km away from non-
riverine sites to prevent sampling in any transi-
Because we were investigating a group of carni-
vores with widely varying home range sizes, rather
than an individual species, we do not present an
effective trapping area. Instead, we acknowledge
the confounding effect of indeterminate plot
sizes, as discussed by Efford and Dawson (2012).
Further, we acknowledge the undefined distance
from which an individual could be drawn in to bait
from outside the study area due to acute sense of
smell, making an estimate of maximum site area
particularly problematic for sites at the edge of the
study area. For example, brown hyaenas can smell
scents of even old carcasses from at least 2 km
away (Mills & Hofer 1998). After considering these
factors, combined with the fact that this study was
more concerned with the effects of covariates on
occupancy and detection rather than with spatial
use, we did not attempt to interpret species spatial
use resulting from the occupancy portion of this
analysis. Rather, we focus on covariate effects of
habitat on occupancy probabilities and bait on
Fat rags consisted of 30 cm × 30 cm pieces of
cheese cloth soaked in liquid cooking fat (sun-
flower oil, olive oil, butter) left over after cooking
vegetables, eggs, chicken, beef, and pork in the
field camp kitchen. Also, fat was collected and
stored each week after cleaning the kitchen’s fat
trap. The camp has not reported any issues with
carnivores attempting to enter the kitchen to
access food or stored fat.Fat rag sites required two
trees or shrubs located approximately 3–6 m
apart. One plant served as the camera post, while
the other was cut at 0.5 m as a bait post. A single
fat rag was strapped to the top of each bait post
using cable ties. For the bait post, the use of a live
plant was preferred to the use of a stake to hinder
displacement or extraction of the bait post by
Meat was obtained from two male impalas
harvested off the property on the first day of the
study. Each impala was field dressed and divided
into five parts consisting of the combined head and
ribcage plus four body quarters. Thus, two impala
yielded all 10 baits needed for each meat site in the
study. Meat was hung on a low, overhanging tree
branch. Branches were selected to make access
by tree-climbing carnivores difficult or impossible.
A 50/50 mixture of water and impala rumen was
poured over the bait. A second tree or shrub,
located 8–10 m away, was used to attach the
camera. By hanging the meat bait in this manner,
we provided a carrion scent lure without risking
: Estimating occurrence and detectability of a carnivore community in Botswana 37
loss of the meat itself. Loss of the bait would
necessitate the harvest of additional impala, an
undesirable scenario given the extreme limits to
acquiring research harvest permits within the
study site. These restrictions were further com-
pounded by the nationwide halt of safari hunting
permits in 2013 as preparation for a blanket ban on
sports hunting across Botswana in 2014 (National
Geographic Society, 2012; Africa Geographic,
For control sites, a camera was placed on a tree
or shrub and pointed toward an unobstructed
viewing space of at least 8–10 m in diameter.
Cameras were placed along roads or game trails
and baits were refreshed weekly with new fat rags
or by pouring a 50/50 mixture of water and impala
rumen over meat baits. No meat was replaced
during the study period, again due to the difficulty
in securing harvest permits. Cameras were also
checked for functionality and battery life at all sites
during bait refreshment and at least once between
bait refreshments. Cameras operated 24 h/day.
Community occupancy modelling
To estimate site occupancy while accounting for
imperfect detection, we employed a single season
occupancy model within a multi-species hierarchi-
cal Bayesian modelling framework (Dorazio &
Royle, 2005; Dorazio, Royle, Söderström &
Glimskär, 2006). Detection probabilities allow the
model to improve occupancy estimates by predict-
ing how many sites were occupied by the species
in question, even when the species was not
detected. Our modelling approach allowed us to
estimate the effects of habitat classification on
occupancy and baiting method on detectability,
while also permitting estimations for some of the
rarer species through data borrowing that would
not have been possible employing a single species
framework (Dorazio & Royle, 2005). Since the
placement of our sample sites did not ensure that
detections among sites were independent, we
interpret occupancy as site use.
In our model, modified from Dorazio & Royle
(Dorazio & Royle, 2005), we assume that occu-
pancy and detection (R
each influenced by a species-level random effect
independent of site,
, respectively, and by a
site-level random effect independent of species, "
respectively (Dorazio & Royle, 2005). We
with linear combination of
parameters representing site-level categorical
covariates for habitat and bait type.
This resulted in the following models:
are binary indictor
vectors for riverine sites, meat bait sites, and fat
rag bait sites, respectively. Thus, intercept values
represent non-riverine sites for the occupancy
equation and control sites for the detection equa-
tion. We defined a sampling occasion as a 24-hour
period beginning and ending at sunrise each day.
Days that cameras did not collect data (e.g.battery
died, elephants redirected the camera field of
view) were censored in the analysis.
We plotted mean occupancy and detection point
estimates and their 95% Bayesian credibility
intervals that resulted from the model (Figs 2 & 3).
Separately, we reported Bayesian p-values and
95% Bayesian credibility intervals (p-value/BCI)
for the occupancy (Table 3) and detection parameter
estimates (Table 4). Bayesian p-values (hereafter
‘p-values’) serve as an indicator of model fit, with
values near 1 showing strong evidence for the
model covariate and values near 0 showing strong
evidence for the intercept model.For bait covariates,
p-values represented the probability that a bait
had a positive effect on detections as compared to
control sites. For the habitat covariate, p-values
represented the probability that riverine sites
increased occupancy as compared to non-riverine
sites. The significance of this covariate effect was
determined from the mean parameter estimates
and their associated BCIs, where an effect was
considered significant when the credibility interval
did not include zero.
We ran the analysis using WinBUGS (Lunn,
Thomas, Best & Spiegelhalter, 2000) and the
R2WinBugs package (Sturtz, Ligges & Gelman,
2005) for program R v. 3.1.2 (R Core Team, 2014).
We assigned non-informative priors with normal
distributions to parameters and intercepts (Dorazio
& Royle, 2005) using three Markov Chain Monte
Carlo simulation chains with 30 000 iterations, dis-
carding the first 5000 iterations as a burn-in and
thinning chains by 5:1.
All research was conducted under Institutional
Animal Care and Use Committee approval under
Animal Use Protocol number A2013 04-002-Y1-A0
and Botswana Ministry of Environment, Wildlife,
and Tourism Research Permit EWT 8/36/4 XXV
38 African Journal of Wildlife Research Vol. 47, No. 1, April 2017
Thirteen of the 16 carnivore species known to
inhabit the study area (81%) were detected (Table 2).
African lion (
), dwarf mongoose
), and Selous’ mongoose (
) occur on the site but were not
observed during the study. Cheetahs (
) occur at very low densities in the area,
and were thus not included in our list of resident
species (Brassine & Parker, 2015). Wild dogs
) have been observed on the site,
but a poaching event in 2012 killed the remaining
mature adults in the area (Andrei Snyman, pers.
obs., June 2013). Thus, we did not consider this
species to be resident as their continued presence
on the site has not been confirmed.
Occupancy results of plotted point estimates
showed that occupancy rates of African civets
) and brown hyaenas were higher
at riverine than non-riverine sites (Fig. 2). Parame-
ter estimates (p-value/BCI) gave further informa-
tion about covariate effects.Riverine habitat had a
significant and positive effect on occupancy for
African civet, African wildcat (
), brown hyaena, honey badger (
), large-spotted genet (
and leopard (
) (Table 3).
Detection results of point estimates showed
differences in detection rates among bait types for
five species (Fig. 3).Detectability of brown hyaena
was significantly different among all bait types,
with meat sites yielding a higher detection rate
than fat rag sites, and fat rag sites yielding a higher
detection rate than controls. A significant differ-
ence was found for spotted hyaenas (
), with meat yielding a higher detection rate
than either fat rag or control sites.Meat and fat rag
sites had similar detection estimates for both
black-backed jackals (
) and Afri-
can civets, which were both significantly higher
than control sites. Also, fat rag sites had signifi-
cantly greater detection point estimates for large-
spotted genets than either unbaited control or
meat baited sites (Fig. 3). Parameter estimates
(p-value/BCI) showed through covariate effects
that meat had a significant and positive effect on
detection for African civet, black-backed jackal,
brown hyaena, honey badger, leopard, and spot-
ted hyaena (Table 4). Fat rags had a significant and
positive effect on detection for African civet, Afri-
can wildcat, black-backed jackal, brown hyaena,
and large-spotted genet (Table 4).
Our study represents the first comprehensive
occupancy survey of carnivores in eastern Bot-
swana. These are among the first reported occu-
pancy estimates for many of the species detected
in the Northern Tuli Game Reserve and among the
first to take a quantitative community perspective
on a carnivore complex in southern Africa (e.g.
, 2013; Ramesh, Kalle, Rosenlund &
Downs, 2016; Rich
, 2016). Our detection of
most of the species found on the study site
suggests that our methods are comprehensive for
surveying and monitoring the carnivore commu-
nity. Our methodology employing baited camera-
trapping methods proved successful in addressing
the issue of the general low detectability of carni-
vores and also demonstrated species-specific
relationships in bait types that will be relevant for
future research. The efficacy of our methodology
in detecting rare and cryptic species is notable in
that we obtained the first recorded evidence of
white-tailed mongoose for our study site, a species
whose occurrence had been suspected, but not
In addition, we found differences in habitat use
among congeneric species. For example, large-
spotted genets had higher occupancy rates at
riverine sites, which are dominated by large
fever berry forest, whereas small-spotted genets
) exhibited higher use of non-
riverine regions (Table 3). Large-spotted genets
are known to range widely, but prefer wooded
habitats (Fuller, Biknevicius & Kat, 1990), whereas
small-spotted genets will use less wooded habitat if
necessary (Herrero & Cavallini, 2007). To our
knowledge, this is the first study to quantify differ-
ences in habitat use by these sympatric
species. Similarly brown hyaena primarily occupied
areas along rivers whereas spotted hyaenas were
recorded mostly at non-riverine sites (Table 3).
and hyaena species on
the reserve is possibly a function of niche separa-
tion associated with use of riverine
Our lack of detections of African lion, dwarf
mongoose, Selous’ mongoose, cheetah, and wild
dog are likely due to disparate reasons.Wild dogs
were likely extirpated due to a poisoning event just
prior to the start of our research (Andrei Snyman,
pers. obs., June 2013). African lions and cheetahs
are present, but our study area is generally not
within core areas on the reserve for those species.
Additionally, our short sampling period might not
: Estimating occurrence and detectability of a carnivore community in Botswana 39
40 African Journal of Wildlife Research Vol. 47, No. 1, April 2017
Fig. 2. Mean occupancy of carnivore species in the Mashatu Game Reserve, Botswana, during June, July and August of 2013. Mean occupancy estimates and 95%
Bayesian credibility intervals are shown for riverine and non-riverine habitat by species.
: Estimating occurrence and detectability of a carnivore community in Botswana 41
Tab le 3. Estimates for covariate effects of habitat classification on occupancy during 2013 for carnivores on the Mashatu Game Reserve in eastern Botswana. Positive
mean values suggest that the species yielded higher occupancy values in the associated habitat type, while negative values suggest lower occupancy values in the associ-
ated habitat (mean ± 95% Bayesian Credibility Interval). Significant effects (where the BCI does not include 0) are in boldface and italicized. Bayesian p-values assess the
proportion of times out ofn=60000iterations (3 chains × 20 000 iterations) that the parameter on the covariate was greater than that of the control or reference (here
non-riverine habitat is the reference), indicating strength of the habitat effect, with values over 0.5 indicating preference of riverine habitat over non-riverine habitat, and
Covariate Riverine Non-riverine (intercept values)
Species p-value Mean L_BCI U_BCI p-value Mean L_BCI U_BCI
1.00 4.97 2.26 8.91
0.00 –3.16 –6.77 –1.02
1.00 6.24 0.86 20.98
0.15 –0.87 –2.70 0.87
Banded mongoose 0.97 6.46 –0.19 21.18 0.50 0.39 –2.18 5.29
Bat-eared fox 0.39 0.13 –3.72 9.77 0.16 –1.04 –3.19 2.08
Black-backed jackal 0.89 4.62 –1.34 19.96
1.00 2.62 0.87 5.89
1.00 7.25 1.46 22.01
0.46 –0.05 –1.19 1.14
0.99 6.85 0.98 21.93 0.31 –0.34 –2.41 2.98
1.00 6.69 0.79 21.78 0.17 –0.70 –2.18 0.73
0.99 6.73 0.44 22.24 0.79 1.15 –0.81 5.75
Slender mongoose 0.90 5.21 –1.36 20.75 0.48 0.21 –2.93 5.00
Small-spotted genet 0.32 0.04 –6.06 14.34 0.92 1.79 –0.47 6.02
Spotted hyaena 0.12 –1.44 –4.65 0.99
1.00 2.11 0.42 5.05
White-tailed mongoose 0.96 6.09 –0.30 20.76 0.22 –1.43 –6.09 3.15
42 African Journal of Wildlife Research Vol. 47, No. 1, April 2017
Fig. 3. Mean detection of carnivore species in the Mashatu Game Reserve, Botswana, during June, July and August of 2013.Mean detection estimates and 95% Bayesian
credibility intervals are shown for meat and fat rag baits and control sites by species. Note that while detection estimates span from 0 to 1, the
-axis is truncated at the high-
est observed detection value.
: Estimating occurrence and detectability of a carnivore community in Botswana 43
Tab le 4. Estimates for covariate effects of bait type on detection based on 20 000 iterations comparing each bait type (fat rag, meat) against unbaited (control) sites for
carnivore baits evaluated on Mashatu Game Reserve, Botswana from June–August 2013. Positive mean values suggest that the bait site yielded greater detection values
than control sites (Mean ± 95% Bayesian Credibility Interval). Significant effects (where the BCI does not include 0) are in boldface and italicized. Bayesian p-values
assess the proportion of times out of n = 60 000 iterations (3 chains × 20 000 iterations) that the parameter on the covariate was greater than that of the control or reference,
indicating strength of the bait effect, with values over 0.5 indicating preference of the bait over the control, and
Covariate Fat Rag Meat Control (Intercept values)
Species p-value Mean L_BCI U_BCI p-value Mean L_BCI U_BCI p-value Mean L_BCI U_BCI
1.00 1.85 1.01 2.77 1.00 1.70 0.86 2.66
0.00 –2.84 –3.73 –2.07
1.00 1.58 0.42 2.76
0.85 0.67 –0.75 1.73 0.00 –4.31 –5.42 –3.22
Banded mongoose 0.69 0.33 –1.03 1.65 0.93 0.89 –0.36 1.98 0.00 –4.59 –5.82 –3.46
Bat-eared fox 0.86 0.91 –0.74 2.60 0.85 0.81 –1.14 2.33 0.00 –4.03 –5.87 –2.46
1.00 1.38 0.92 1.84 1.00 1.23 0.74 1.72
0.00 –1.39 –1.81 –0.97
1.00 1.24 0.35 2.23 1.00 2.03 1.18 3.05
0.00 –3.14 –4.08 –2.33
Honey badger 0.97 1.17 –0.08 2.45
0.99 1.36 0.27 2.54
0.00 –4.72 –5.93 –3.57
1.00 3.52 2.35 4.62
0.93 0.84 –0.34 1.87 0.00 –3.93 –5.01 –2.79
Leopard 0.23 –0.41 –1.48 0.62
0.98 0.91 0.03 1.73
0.00 –3.17 –3.93 –2.39
Slender mongoose 0.55 0.08 –1.89 1.85 0.90 0.91 –0.64 2.13 0.00 –5.36 –7.03 –3.70
Small-spotted genet 0.96 1.20 –0.16 2.58 0.97 1.22 –0.02 2.40 0.00 –4.93 –6.35 –3.61
Spotted hyaena 0.32 –0.21 –1.06 0.63
1.00 1.36 0.60 2.13
0.00 –2.72 –3.44 –2.02
White-tailed mongoose 0.68 0.44 –1.53 2.29 0.94 1.07 –0.40 2.36 0.00 –5.64 –7.54 –3.80
encompass an entire territorial patrol event for
larger carnivores. These factors, coupled with a
relatively tight camera placement of ~1 km
between similar baits, could render the study area
absent of these species during our sampling
period. Dwarf and Selous’ mongoose are small
and common species that have been observed on
the reserve; however, they are known to have very
small home ranges and are generally considered
to be insectivorous (Creel, 2013). Furthermore,
sites were selected to increase the likelihood to
capturing species along travel routes while consid-
ering the safety of a researcher walking on foot.
Thus, site placement along roads and game trails
may have influenced detection of some species.
Consequently, the spatial scale, site placement,
and baits used in our study might have fundamen-
tally reduced the likelihood of detecting these
species. Finally, limitations in possible survey
days along with small study area size resulted in
low sampling effort for this investigation, as repre-
sented in Table 1, which likely influenced low
detection rates for some species.
Combined information from the detection analy-
sis and parameter estimates give strong evidence
that meat baits are best for members of the large
carnivore guild, in particular leopard and spotted
hyaena, and possibly brown hyaena. While meat
was most effective for brown hyaenas, results
showed that fat rags could also be used effectively
to survey this species.A recent study of baited
unbaited camera traps in Zimbabwe, where
) meat was used as bait,
supported the need for bait in leopard surveys (du
, 2014). Conversely, studies on the
mesocarnivore complex in southern Africa using
fat rags as a lure will likely benefit from similar or
increased species detections with lower costs and
far fewer logistical and permit requirements as
compared to those that employ meat as bait. This
is especially true in the case of large-spotted
genets which showed significantly higher detec-
tion rates at fat rags sites as compared to any other
bait type. Used cooking fat is generated on-site in
camp kitchens and does not present the storage
issues that exist with meat bait.The fat rag method
eliminates the need for harvesting meat on site,
storing meat in camp, or shipping meat from an
external source. Further, it provides a use for a
waste product found in most field camps in south-
Finally, and perhaps most importantly, fat rags
may also help in studying mesocarnivores in areas
where large predators are common, by specifically
targeting and attracting these subordinate carni-
vores. Use of a universal bait preferred by large
and small carnivores alike, such as impala meat,
when attempting to trap a specific and perhaps
uncommon carnivore species, can logically result
in a situation whereby the target species avoids
the bait site due to the presence of dominant,
Through employing a systematic sampling
design we were able to quantify the composition
and habitat use of the carnivore community in the
Mashatu Game Reserve in eastern Botswana
while also demonstrating an effective methodology
using baits to address issues of low detectability
typical in carnivore research. Given the inter-
specific difference in habitat use highlighting the
importance of riverine habitats, and given that this
is the first community level study of the carnivore
community in the region, our results serve as an
important reference point for future research,
monitoring, and management of the carnivore
community in eastern Botswana.
To determine if our findings are consistent
across different regions, we suggest that future
research should repeat this study within carnivore
communities similar to the one tested here, and
within communities different to that described
here, in a larger study area employing a higher
sampling effort than our study allowed. Further-
more, future studies should continue to investigate
the effect of fat rag bait on detection by comparing
fat rags to other commonly used bait types such as
eggs or offal in addition to meat. We found that
some species occur at higher occupancy rates in
riverine habitat than in non-riverine habitat. To
determine if finer-scale aspects of habitat (e.g.
vegetation density) also affect occupancy, a
larger-scale study should look at factors influenc-
ing heterogeneity within riverine and non-riverine
habitat areas. Finally, we recommend that future
investigations specifically address the role of habi-
tat in niche separation of sympatric congeneric
species within the order Carnivora to determine if
habitat diversity is integral to the coexistence of
We thank the Ministry of Environment, Wildlife
and Tourism and the Department of Wildlife and
National Parks, Botswana, for their cooperation
and permission to conduct our research as well as
Mashatu Game Reserve, in particular Pete le
44 African Journal of Wildlife Research Vol. 47, No. 1, April 2017
: Estimating occurrence and detectability of a carnivore community in Botswana 45
Roux and David Evans. Additionally, we would like
to thank Anton Lategan and Stuart Quinn for their
continued support of our research and collabora-
tion. We greatly appreciate the advice of Drew
Tyre in finalizing quantitative aspects of the manu-
script. Finally, we thank the University of Georgia
Botswana Field Course students for field assis-
tance. This work was supported by the University
of Georgia Warnell School of Forestry and Natural
Resources, and EcoTraining Inc. We thank these
organizations and individuals for their extensive
financial and logistical support.
Africa Geographic. (2014). Botswana hunting ban takes
effect. Retrieved from: http://africageographic.com/
blog/botswana-hunting-ban-takes-effect/ on 8 Octo-
Barea-Azcón, J.M., Virgós, E., Ballesteros-Duperón, E.,
Moleón, M. & Chirosa, M. (2006). Surveying carni-
vores at large spatial scales: a comparison of four
Biodiversity and Conserva-
, 16(4), 1213–1230.
Brassine, E.I. & Parker, D.M. (2015). Trapping elusive
cats: using intensive camera trapping to estimate the
density of a rare African felid.
Caro, T.M. & Stoner, C.J. (2003). The potential for
interspecific competition among African carnivores.
, 110(1), 67–75.
Creel, S. (2013).
dwarf mongoose. In
J. Kingdon & M. Hoffmann (Eds),
ins, equids and rhinoceroses
(Vol. V, pp. 368–373).
London, U.K.: Bloomsbury.
Dalerum, F., Somers, M.J., Kunkel, K.E.& Cameron, E. Z.
(2008). The potential for large carnivores to act as
biodiversity surrogates in southern Africa.
diversity and Conservation
, 17(12), 2939–2949.
Dorazio, R.M. & Royle, A.J. (2005). Estimating size and
composition of biological communities by modeling
the occurrence of species.
Journal of the American
, 100(470), 389–398.
Dorazio, R.M., Royle, J.A., Söderström, B. & Glimskär, A.
(2006). Estimating species richness and accumula-
tion by modeling species occurrence and detect-
, 87(4), 842–854.
Du Preez, B.D., Loveridge, A.J. & Macdonald, D.W.
(2014). To bait or not to bait: a comparison of camera-
trapping methods for estimating leopard
, 176, 153–
Efford, M.G. & Dawson, D.K. (2012). Occupancy in con-
, 3(4), 1–15.
Estes, J.A., Terborgh, J., Brashares, J.S., Power, M.E.,
Berger, J., Bond, W.J., Carpenter, S.R., Essington,
T.E., Holt, R.D., Jackson, J.B.C., Marquis, R.J.,
Oksanen, L., Oksanen, T., Paine, R.T., Pikitch, E.K.,
Ripple, W.J., Sandin, S.A., Scheffer, M., Schoener,
T.W., Shurin, J.B., Sinclair, A.R.E., Soulé, M.E.,
Virtanen, R., Wardle, D.A. (2011). Trophic downgrad-
ing of planet earth.
, 333(6040), 301–306.
Foster, R.J. & Harmsen, B.J.(2012). A critique of density
estimation from camera-trap data.
Journal of Wildlife
, 76(2), 224–236.
Fuller, T.K., Biknevicius, A.R. & Kat, P.W. (1990). Move-
ments and behavior of large spotted genets (
Gray 1830) near Elmenteita Kenya
Gil-Sánchez, J.M., Moral, M., Bueno, J., Rodríguez-
Siles, J., Lillo, S., Pérez, J., Martín, J.M., Valenzuela,
G., Garrote, G., Torralba, B., Simón-Mata, M.Á.
(2011). The use of camera trapping for estimating
Iberian lynx (
) home ranges.
Journal of Wildlife Research
, 57(6), 1203–1211.
Gusset, M., Swarner, M.J., Mponwane, L., Keletile, K. &
McNutt, J. W. (2009). Human–wildlife conflict in
northern Botswana: livestock predation by Endan-
gered African wild dog
and other carni-
, 43(01), 67–72.
Hamel, S., Killengreen, S.T., Henden, J-A., Eide, N.E.,
Roed-Eriksen, L., Ims, R. A. & Yoccoz, N. G. (2013).
Towards good practice guidance in using camera-
traps in ecology: influence of sampling design on
validity of ecological inferences.
Methods in Ecology
, 4(2), 105–113.
Herrero, J. & Cavallini, P. (2007). Genetta genetta (The
IUCN Red List of Threatened Species 2007) (p.
IUCN. (2013).The IUCN Red List of Threatened Species
[Version 2013.2]. Retrieved from: http://www.
iucnredlist.org/ on 12 February 2014.
Lunn, D.J., Thomas, A., Best, N. & Spiegelhalter, D.
(2000). WinBUGS–aBayesian modelling frame-
work: concepts, structure, and extensibility.
, 10, 325–337.
Mackenzie, D.I. & Royle, J.A. (2005). Designing occu-
pancy studies: general advice and allocating survey
Journal of Applied Ecology
, 42(6), 1105–1114.
Martinoli, A., Preatoni, D., Galanti, V., Codipietro, P.,
Kilewo, M., Fernandes, C.A.R., Wauters, L.A., Tosi,
G. (2006).Species richness and habitat use of small
carnivores in the Arusha National Park (Tanzania).
Biodiversity and Conservation
, 15(5), 1729–1744.
Mills, M.G.L. (1987). Behavioural adaptations of brown
and spotted hyaenas in the southern Kalahari.
Afrikaanse Tydskrif vir Wetenskap
, 83, 595–598.
Mills, G. & Hofer, H.(1998). Status Survey and Conserva-
tion Action Plan: Hyaenas. IUCN/SSC Hyaena
Specialist Group. Retrieved from: http://www.
s.pdf on 12 February 2014.
National Geographic Society. (2012). No more hunting of
any kind in Botswana. Retrieved from: http:/
-more-hunting-of-any-kind-in-botswana/ on 9 Octo-
O’Connor, T.G. (2010). Transformation of riparian forest
to woodland in Mapungubwe National Park, South
Africa, between 1990 and 2007.
Ogada, M.O., Woodroffe, R., Oguge, N.O.& Frank, L.G.
(2003). Limiting depredation by African carnivores:
the role of livestock husbandry.
Ramesh, T., Kalle, R., Rosenlund, H. & Downs, C.T.
(2016). Native habitat and protected area size
matters: preserving mammalian assemblages in
the Maputaland Conservation Unit of South Africa.
Forest Ecology and Management
, 360, 20–29.
Ray, J.C., Hunter, L.& Zigour is, J. (2005).Setting conser-
vation and research priorities for larger African carni-
vores (Vol. 24). Wildlife Conservation Society New
York. Retrieved from http://programs.wcs.org/Por tals/
R Core Team. (2014). R: A language and environment
for statistical computing. (Version 3.1.2). Vienna,
Austria: R Foundation for Statistical Computing.
Retrieved from: http://www.R-project.org/ on 27 Feb-
Rich, L.N., Miller, D.A.W., Robinson, H.S., McNutt, J.W. &
Kelly, M.J. (2016). Using camera trapping and hierar-
chical occupancy modelling to evaluate the spatial
ecology of an African mammal community.
, 53(4), 1225–1235.
Ripple, W.J., Estes, J.A., Beschta, R.L., Wilmers, C.C.,
Ritchie, E.G., Hebblewhite, M., Berger, J., Elmhagen,
B., Letnic, M., Nelson, M.P., Schmitz, O.J., Smith,
D.W., Wallach, A.D., Wirsing, A.J.(2014). Status and
ecological effects of the world’s largest carnivores.
, 343(6167), 1241484.
Rowcliffe, J.M. & Carbone, C. (2008). Surveys using
camera traps: are we looking to a brighter future?
, 11(3), 185–186.
Royle, A.J., Magoun, A.J., Gardner, B., Valkenburg, P. &
Lowell, R.E. (2011).Density estimation in a wolverine
population using spatial capture-recapture models.
Journal of Wildlife Management
, 75(3), 604–611.
Schuette, P., Wagner, A.P., Wagner, M.E. & Creel, S.
(2013). Occupancy patterns and niche partitioning
within a diverse carnivore community exposed to
Snyman, A., Jackson, C.R. & Funston, P.J. (2015).
The effect of alternative forms of hunting on the
social organization of two small populations of
in southern Africa.
Sturtz, S., Ligges, U. & Gelman, A.E. (2005). R2Win-
BUGS: a package for running WinBUGS from R.
Journal of Statistical Software
, 12(3), 1–16.
Thorn, M., Green, M., Dalerum, F., Bateman, P.W. &
Scott, D.M. (2012). What drives human–carnivore
conflict in the North West Province of South Africa?
, 150(1), 23–32.
Thorn, M., Scott, D.M., Green, M., Bateman, P.W. &
Cameron, E.Z. (2009). Estimating brown hyaena
occupancy using baited camera traps.
Journal of Wildlife Research
, 39(1), 1–10.
Wilson, G.J. & Delahay, R.J.(2001). A review of methods
to estimate the abundance of terrestrial carnivores
using field signs and observation.
Woodroffe, R. (2000). Predators and people: using
human densities to interpret declines of large carni-
, 3(02), 165–173.
Woodroffe, R., Lindsey, P., Romañach, S., Stein, A. & ole
Ranah, S. M. K. (2005). Livestock predation by
endangered African wild dogs (
Responsible Editor: F. Dalerum
46 African Journal of Wildlife Research Vol. 47, No. 1, April 2017