The influence of prey, pastoralism and
poaching on the hierarchical use
of habitat by an apex predator
Kristoffer T. Everatt1*, Leah Andresen1& Michael J. Somers1,2
1Centre for Wildlife Management,University of Pretoria,Pretoria,South Africa
2Centre for Invasion Biology,University of Pretoria,Pretoria,South Africa
Received 27 October 2014. To authors for revision 1 December 2014. Accepted 29 January 2015
As an apex predator,habitat selection by African lions,
, is primarily determined
by bottom-up processes; however, increasing anthropogenic pressures may alter these
relationships. Using camera traps and track surveys in the Limpopo National Park,
Mozambique, we collected detection/non-detection data of lions and their prey and
combined these with occurrence data on bushmeat poaching activities and spatial data on
agro-pastoralist land use and other landscape features. We used hierarchical modelling
within an occupancy framework to determine the relative influences of ecological variables
on resource use and non-use by lions at two spatial scales. Habitat use by lions was most
strongly influenced by the occurrence of their preferred prey across both spatial scales.
However,lions were strongly negatively predicted by bushmeat poaching at the finer spatial
scale and generally negatively predicted by agro-pastoralist activities at the coarser scale.
Restricting our analysis to the home-range scale would have greatly underestimated the
impact of bushmeat poaching on the ecology of lions. The results of our study illustrate the
trophic dependency of prey resources to lions and the importance of considering scale
when investigating species habitat use. Importantly, our study also demonstrates the
limiting influence of bushmeat poaching on the use of habitat by an apex predator.
Key words: African lion, occupancy, African carnivore, bushmeat hunting, habitat ecology.
Considering the biological constraints of obligate
carnivory, the use of habitat by apex predators
should be primarily predicted by bottom-up pro-
cesses (Mitchell & Hebblewhite, 2012). However,
competition with humans may alter trophic-based
habitat relationships for predators, resulting in
reduced habitat availability (Mitchell & Hebblewhite,
., 2014). For instance, anthro-
pogenic disturbance influences home-range level
habitat selection by cougars,
(Dickson & Beier, 2002), wolves,
(Rich, Mitchell, Gude & Sime, 2012) and tigers,
Competition with humans is recognized as the
ultimate cause behind global predator declines
., 2014). However, many apex predator
populations exist in human-dominated landscapes
., 2014) or are exposed to human
pressures along reserve boundaries (Woodroffe
& Ginsberg, 1998). Understanding how humans
influence the use of habitat by apex predators
is therefore important to improve species conserva-
tion and management. For example, knowledge
of how humans alter trophic-based habitat rela-
tionships for apex predators is an important
component of determining habitat suitability and
Habitat selection can be seen as a hierarchical
process (Johnson, 1980), involving behavioural
choices that span a continuum of time, space and
ecological processes (Mitchell & Hebblewhite,
2012). Recognizing the scale dependency of
variables on species fitness is important for con-
servation planning. Since species’ fitness needs
may differ with scale, investigations limited to a
singular scale may fail to recognize the importance
of key habitat components (Nams, Mowat &
Panian, 2006).For instance, Ciarniello, Boyce, Seip
& Heard (2007) demonstrated how delineating
protected areas for grizzly bears,
based on habitat selection at the third-order would
have excluded important landscape features
whose importance only became evident at the
African Journal of Wildlife Research 45(2): 187–196 (September 2015)
ISSN 2410-7220 [Print], ISSN 2410-8200 [Online] — DOI: http://dx.doi.org/10.3957/056.045.0187
*To whom correspondence should be addressed. Present address:
Centre for African Conservation Ecology, Department of Zoology,
Nelson Mandela Metropolitan University, Port Elizabeth, 6031 South
Africa. E-mail: firstname.lastname@example.org
The contemporary distribution of the African lion,
, is largely associated with the remain-
ing extent of intact savanna (Riggio
., 2012). At
the home-range scale, lions may select for areas
with relatively higher densities of large ungulates
(Van Orsdol, Hanby & Bygott, 1985), and may
select against areas with increased threat
of human persecution (Ogutu, Bhola & Reid, 2005,
but see; Woodroffe & Frank, 2005). Lion foraging
success requires a combination of prey availability
and suitable cover from which to attack (Hopcraft,
Sinclair & Packer, 2005; Mosser, Fryxell, Eberly &
Packer, 2009). At the fourth-order scale of habitat
use, lions may select areas with preferred hunting
features over areas with higher prey densities
, 2009; Davidson
Stalking and ambush cover are less limiting to
lions in forested savannas (Funston, Mills, Biggs &
Richardson, 1998; Hopcraft
., 2005), than to
lions on open plains (Hopcraft
, 2005; Mosser
The goal of this study was to determine the
relative influence of bottom-up resource and top-
down anthropogenic factors on the hierarchical
use of habitat by lions in a system impacted by
pastoralism and poaching. We considered the
Limpopo National Park (LNP) in southwestern
Mozambique as an ideal case study location
because both lions and humans are resident in the
park, and humans freely extract resources, includ-
ing ‘bushmeat’, graze livestock, and are known to
persecute lions (Everatt, Andresen & Somers,
2014). We measured habitat use by lions within an
occupancy modelling approach that explicitly
accounts for survey and site level species
., 2002; Mitchell
& Hebblewhite, 2012). Discreet occupancy mod-
els were also developed for the variables describ-
ing resources and threats encountered by lions in
LNP. We predicted that habitat use by lions in LNP
would reflect their ecological niche as an apex
predator and be primarily predicted by bottom-up
processes, but that the threats posed to lions by
top-down anthropogenic pressures would strongly
determine the use of habitat. In addition, we
predicted that the relative importance of variables
describing lion habitat would vary with scale.
The study was conducted in a 2400 km2area
of woodland savanna plains in Mozambique’s
Limpopo National Park (LNP) (UTM X 384502
UTM Y 7432635) (Fig. 1). LNP forms a component
of the Greater Limpopo Trans-frontier Park
(GLTFP) with Kruger National Park (KNP), South
Africa and Gonarezhou National Park, Zimbabwe.
Together these parks form the core of the Greater
188 African Journal of Wildlife Research Vol. 45, No. 2, September 2015
Fig. 1. Location of study area (squares) within Limpopo National Park, which forms the Mozambican component
of the Greater Limpopo Trans-frontier Park (in green).
Limpopo Lion Conservation Unit (IUCN, 2006).
The study area is bordered to the west by KNP and
to the east by a near continuous band of agro-
pastoralist settlements along the banks of the
Limpopo River. There are additional smaller
agro-pastoralist settlements situated within the
study area. The human population is estimated
at 6500 (in 2003) in LNP and 20 000 living in the
eastern boundary villages (Huggins, Barendse,
Fischer & Sitoi, 2003), together grazing over
20 000 cattle,
Wildlife in this region of Mozambique were
largely decimated during 28 years of war
(1964–1992) (Hatton, Couto & Oglethorpe, 2001).
However, removal of portions of the South Africa–
Mozambique border fence as part of the creation
of the GLTFP in 2000 has provided opportunities
for re-colonization of wildlife into LNP (Hanks,
2000), and currently 23 spp. of ungulates occur
in the Park (Everatt, 2014).At the time of this study,
the lion population in LNP was estimated at
66 individuals or a density of 0.99 lions/100 km2
It is possible to examine habitat use at multiple
spatial scales using sampling windows of differing,
biologically relevant sizes (Baldwin & Bender,
., 2012). In this study, we exam-
ined habitat use by lions at two spatial scales
equivalent to Johnson’s (1980) second- and third-
order of habitat selection. We defined second-
order sampling sites as 100 km2grid cells, based
on average home-range sizes of lions in the adjoin-
ing and contiguous KNP (Funston, Mills, Richard-
son & Van Jaarsveld, 2003.) (Fig. 1). Included
within these grid cells we defined third-order sites
as approximately 1 km2, reasoning this size was
biologically meaningful to the scale at which lions
make short-term habitat-use decisions.
To quantify habitat use by lions we employed an
occupancy modelling approach where the estimator
(Y) was defined as the
probability of site use
., 2006). We made the following
assumptions; 1) species were not falsely identi-
fied, 2) detections were independent, 3) hetero-
geneities in occupancy or detection probabilities
were modelled using covariates. It is important to
note that the closure assumption could be relaxed
because our estimator was
probability of site use
proportion of area occupied
., 2006 p. 105).
We collected the data from temporally replicated
detection/non-detection camera trap and spoor
surveys conducted from 9 September 2011 to
26 November 2012. We deployed digital motion
cameras (15 Reconyx HC500 (Wisconsin, U.S.A.)
(trigger time of 0.97 s, detection zone approxi-
mately 24 m), 7 Spy Point Tiny-W2 (Québec,
Canada) (trigger time of 0.91 s, detection zone
approximately 17 m), 10 Bushnell Trophy Cam
(Beijing, China) (trigger time of 0.66 s, detection
zone approximately 18 m) (http://www.trailcampro.
com/trailcamerareviews.aspx) on dirt tracks,
game trails and along river edges. In addition, we
conducted track surveys on foot due to the lack
of road networks in the study area. These surveys
followed an obvious path of travel, (
trail or river edge) where substrate was adequate
Sampling occasions at the home-range scale
were represented by 189 temporally replicated
3 km transect samples (replicates separated by
> 14 days) and 326, 14-day camera-trap samples
for a combined mean of 21.6 samples per grid cell.
Of the 24 grid cells, 23 were sampled with camera-
traps (mean = 14 samples/grid cell, range = 3–30
samples/grid cell) and 23 were sampled with track
surveys (mean = 8 samples/grid cell, range = 2–16
samples/grid cell). We note here that unequal
sampling across sites is accounted for within an
occupancy model (MacKenzie
., 2002). In an
effort to meet the assumption of independence
between sampling occasions at the home-range
scale, we pooled detections (within grid cells)
when a camera-trap had sampled anytime 14 days
prior to a track transect. Sample occasions at the
short-term use scale were represented by 998
temporally replicated 1 km transects (232 sites;
638 samples) (replicates separated by 14 days)
and 14 day camera-trap samples (82 sites; 360
samples) for a combined mean of 3.6 samples per
site. Of the total 260 sites surveyed, 184 sites were
sampled only by transects, 48 were sampled
by transects and camera-traps, and 28 sampled
only by camera-traps.The detection or non-detec-
tion of lions was recorded for each (14 day)
camera trapping sample and each (1 km) track
Identification of covariates
To explain habitat use by lions in a human-
disturbed landscape, we considered five fitness-
related covariates. These included: encounter
: Factors affecting hierarchical use of habitat by an apex predator 189
probability of lion’s preferred prey, encounter
probability of alternate prey, landscape features
that facilitate prey capture (
encounter probability of bushmeat poaching and
agro-pastoralist use (Table 1).
Lions exhibit a strong preference for larger
bodied prey including African buffalo,
, (Hayward & Kerley, 2005). To quantify the
influence of preferred prey availability on habitat
use by lions, we used a probability of occurrence
model for buffalo that was developed by Everatt
., (2014) for the same survey area and time.
Other species that lions are known to preferentially
select for, including; giraffe,
, and blue wildebeest,
, (Hayward & Kerley,
2005), were excluded from the analysis because
they have a limited distribution of occurrence in the
study area (Stephensen, 2010). To quantify the
influence of alternate prey availability for lions, we
combined probability of site use for warthog,
, (Supplementary infor-
1295207) and impala,
(Andresen, Everatt & Somers, 2014) from models
developed for the same survey and time. We
assumed that the probability of prey occurrence
site use) is biologically representative of an
encounter probability for lions. To quantify the
influence of bushmeat poaching on habitat use
by lions, we used a bushmeat poaching occu-
pancy model developed by Everatt
., (2014) for
the same study area and time. Agro-pastoralist
use was measured as the mean Euclidean dis-
tance to a settlement edge per 30m×30mpixel
in a grid cell (home-range analysis) or in a buffer
(50 m diameter) placed around each camera
station or track transect (short-term site use analy-
sis) from a landscape raster (Peace Parks Foun-
dation, Stellenbosch). We considered riparian
areas as a proxy for landscape features that facili-
tate prey capture (Hopcraft
, 2005), measured
as the number of 30m×30mpixels (per grid cell
or buffer) overlapping either river (including drain-
age lines) or water (including pans) raster layers
(Peace Parks Foundation, Stellenbosch). Analy-
ses were made in the Spatial Analysis tool in Arc-
GIS 9.3.1. (ESRI, Redlands, California, U.S.A.).
We constructed a detection/non-detection matrix
for each site and spatial scale, recording a ‘1’or ‘0’
where lions were detected or not detected, respec-
tively. Following this, we constructed two survey-
specific matrices for each analysis to account for
differences in detectability between the two
sampling methods used. In the first matrix a ‘1’was
recorded where only the method ‘track’ was
employed and a ‘0’ where only cameras were
employed.In the second matrix, a ‘1’ was recorded
where each method was used and data were
pooled, and a ‘0’ where only one method was used.
The overlap of the two matrices therefore accoun-
ted for three sampling possibilities at each site;
tracks only, cameras only and pooled samples.
Additionally, we constructed season specific
dry) matrices, recording a ‘1’ for
surveys conducted during the wet season and a
‘0’ for surveys conducted during the dry season. To
account for variation in lion detection probability
) the covariates ‘track’ and ‘pooled’ (hereafter
referred to as method ‘M’) and ‘season’ were
190 African Journal of Wildlife Research Vol. 45, No. 2, September 2015
Table 1. Covariates expected to influence habitat use by lions.
Covariate Key Fitness value to lion Description Sampling range: Sampling range:
short-term habitat home-range
use habitat use
Preferred prey PP Availability of preferred prey Probability of buffalo site use 0.1–0.5 0.1–0.6
mean = 0.4 mean = 0.4
Alternate prey AP Availability of alternate prey SProbability of warthog 0.1–1.7 0.2 – 1.5
and impala site use mean = 1.0 mean = 0.9
Bushmeat poaching B Targeted or accidental snaring Probability of bushmeat 0.0–1.0 0.1 – 1.0
poaching site use mean = 0.6 mean = 0.6
Village V Persecution in defense of Proximity to agro-pastoralist 0.1–24.5 2.0 – 20.4
livestock settlements (km) mean = 10.9 mean = 11.2
Riparian R Landscape feature facilitating Amount of riparian area in site 0.0– 2.7 0.0 – 928.1
prey capture (# 30 × 30 m pixels) mean = 0.1 mean = 315.3
included in all models describing lion site use (Y).
We estimated site occupancy (Y) and detection
) using maximum likelihood functions
, 2006) and the single season
option in the program PRESENCE Version 5.5
(Hines, 2006). Continuous site covariates were
standardized on a
-scale and all covariates were
tested for collinearity using a cut-off of
Covariates found to be correlated were not included
in the same models. All possible (non-correlated)
combinations of occupancy covariates (Supple-
mentary information, http://dx.doi.org/10.6084/
m9.figshare.1295207) were considered for each
analysis (home-range scale = 11 models, short-
term site use scale = 16 models). We ranked mod-
els based on Akaike Information Criterion (AIC),
using AICc adjusted for small sample size, with the
sample size set as the number of sampling sites
(Burnham & Anderson, 2002). Models with a
DAICc <2 were considered to have strong support.
We considered a candidate set of all models
DAICc <7 whose combined weights ³0.95 (
95% confidence set). AICc weights were used to
determine the weight of evidence for each model,
and were summed for each covariate in the 95%
confidence set (Burnham & Anderson, 2002).
Variables with high summed model weights were
considered more important in explaining hetero-
geneity in occupancy. The direction of influence
of individual covariates was determined by the
sign of the b-coefficients (MacKenzie
Covariates were considered to have strong or
robust impact if b± 1.96 × S.E.from the top ranking
model were not overlapping zero. We used a
weighted model averaging technique to calculate
overall parameter estimates (Burnham & Anderson,
2002). Finally, we performed a goodness of fit test
using 10 000 bootstrap samples and a Pearson’s
chi-square statistic on the most saturated model
(MacKenzie & Bailey, 2004).
We recorded a total survey effort of 5335 camera
trap nights and 638 km of track surveys. After
pooling sampling occasions, the final data set
consisted of 251 sampling occasions at the home-
range scale and 957 sampling occasions at the
short-term site-use scale. Lions were detected
on 35 (14 day) camera samples (from 664 lion
photos) and 55 (1 km) track samples.We identified
19 individual lions from camera-trap images, with
identification based on sex, age and distinguishing
scars (Whitman & Packer, 2007). This number
is therefore the minimum sample size of individu-
als considered in this habitat analysis. The proba-
bility of site use by lion’s prey and by bushmeat
poachers is summarized in Table 2.
Habitat use at the home-range scale
The model averaged probability of detecting
lions where they occurred at the home-range scale
= 0.304 (S.E. = 0.095). The covariate
preferred prey was strongly supported and was
the principal contributing factor to habitat use
by lions at this spatial scale; the only model that
emerged with a DAICc <2 was the univariate
model Y(P)p(M+S) (Tables 3 & 4).Lions showed a
strong use of sites with a greater probability
of occurrence of their preferred prey (Tables 3 & 4).
In addition, lions generally occurred at sites with
a greater proportion of riparian areas that were
further from villages with a greater probability
of occurrence of alternate prey and lower probability
of occurrence of bushmeat poaching (Table 4).
There was no evidence lack of fit (p = 0.22) or
over-dispersion () = 1.20).
Habitat use at the short-term use scale
The model averaged probability of detecting
lions where they occurred at the short-term use
scale was $
= 0.230 (S.E. = 0.038). The greatest
contributing factors to habitat use by lions at this
: Factors affecting hierarchical use of habitat by an apex predator 191
Tab le 2. Results from discreet occupancy models describing the probability of detection ( $
) and probability of site use
Y) by bushmeat poachers, buffalo, impala and warthog from camera trapping data in Limpopo National Park, Septem-
ber 2011 to November 2012.
Bushmeat poachers* 0.165 0.027 0.799 0.050
Buffalo* 0.368 0.041 0.416 0.084
Impala‡0.285 0.038 0.482 0.090
Warthog†0.336 0.035 0.513 0.049
From * Everatt
, (2014), ‡Andresen
., (2014), †Supplementary information http://dx.doi.org/10.6084/m9.figshare.1295207
scale were the probability of occurrence of their
preferred prey and the probability of occurrence
of bushmeat poaching (Tables 3 & 4), where lions
showed a strong use of sites with a greater proba-
bility of occurrence of their preferred prey and a
strong negative use of sites with a greater proba-
bility of occurrence of bushmeat poaching (Tables
3 & 4). In addition, lions generally occurred at sites
closer to riparian areas (Table 4). There was no
evidence of a lack of fit (
= 0.41) or over-
dispersion (= 0.44).
In this study we considered use of habitat by an
apex predator that co-occurs with human activities.
Our results demonstrate that habitat use by lions
is influenced by bottom-up resources and by top-
down anthropogenic pressures (Fig. 2). In addition,
we found that the limiting influence of bushmeat
poaching was scale dependent, which has impor-
tant conservation implications.
Habitat use by an apex predator was most
predicted by bottom-up processes
Habitat use by lions in LNP was most strongly
predicted by the occurrence of buffalo. The impor-
tance of this variable was indicated by the weight
of evidence for models containing the buffalo
covariate and by the strong positive influence
of this covariate at the coarser home-range spatial
scale. That the buffalo covariate was strongly
determining across both spatial scales empha-
sizes the importance of this component of lion
192 African Journal of Wildlife Research Vol. 45, No. 2, September 2015
Tab le 3. Summary of model selection procedure for factors influencing site use (Y) by lions at the home-range scale
and at the short-term use scale. Covariates considered include; occurrence probability of preferred prey (P), occur-
rence probability of alternate prey (AP), occurrence probability of bushmeat poaching (B), distance from villages (V)
and proportion of riparian area (W).
Y(P)p(M+S) 0.00 0.372 6 151.44
Y(V)p(M+S) 2.47 0.108 6 153.91
Y(P+R)p(M+S) 2.89 0.088 7 150.27
Y(R)p(M+S) 3.45 0.066 6 154.89
Y(.)p(M+S) 3.52 0.064 5 158.57
Y(AP)p(M+S) 3.54 0.063 6 154.98
Y(P+B)p(M+S) 3.58 0.062 7 150.96
Y(AP+R)p(M+S) 4.18 0.046 7 151.56
Y(R+V)p(M+S) 4.21 0.045 7 151.59
Y(B+R)p(M+S) 5.50 0.024 7 152.88
Y(P+B+R)p(M+S) 5.72 0.021 8 148.50
Y(P+B)p(M) 0.00 0.574 6 539.79
Y(P+B+R)p(M) 1.08 0.334 7 538.76
Y(P)p(M) 4.27 0.068 5 546.15
Y(P+R)p(M) 6.33 0.024 6 546.12
Y(.)p(M) 29.18 0.000 4 573.14
Detectability (p) varies with method (M) and season (S). Y(.) assumes site use is constant, DAICc is the difference in AICc values between
each model with the low AICc model,
is the AICc model weight,
is the number of parameters in the model, and –2l is twice the negative
Tab le 4.b-coefficient estimates for covariates influenc-
ing site use (Y) by lions in order of their summed model
weights (Sw) at the home range use scale and at the
short-term use scale.
Occupancy covariate Smodel w (%) bS.E.
Preferred prey 57.6 9.82* 4.73
Riparian 20.4 0.57 0.62
Villages 16.8 –1.12 0.64
Alternate prey 10.7 2.51 1.57
Bushmeat poaching 7.0 –1.13 1.69
Preferred prey 99.9 8.62* 2.49
Bushmeat poaching 90.8 –1.50* 0.63
Riparian 35.9 0.56 0.46
*Indicates covariate has robust impact (b± 1.96 × S.E. not overlap-
habitat use (Rettie & Messier, 2000). These
results suggest that lions in LNP are making
behavioural choices to select habitat at the home-
range scale that includes the limited distribution
of buffalo herds in the park and then further select-
ing areas at a finer spatial scale that would increase
their probability of encountering individual animals.
That habitat use by lions was strongly deter-
mined by the occurrence of prey resources agrees
with trophic-based species-habitat relationships
(Krebs, 2009; Mitchell & Hebblewhite, 2012). For
instance, food resources were the primary predic-
tor of second-order habitat selection by grizzly
bears in the Canadian Arctic (McLoughlin
2002) and tigers in the Russian Far East (Miquelle
., 1999). Following this, predation risk by
wolves was the primary predictor of second-order
habitat selection by caribou,
in northern Canada (Rettie & Messier, 2000) and
predation risk by lions was the primary predictor
of second-order habitat selection by zebra, giraffe
and wildebeest on a reserve in South Africa
Habitat use by an apex predator is influenced
by top-down anthropogenic disturbance
Habitat use by lions in LNP was strongly nega-
tively predicted by bushmeat poaching at the
short-term use spatial scale. Bushmeat poaching
may limit predator habitat by depletion of prey
resources and by direct, targeted or non-targeted,
., 2013; Lindsey
, 2014). During this study we
found evidence of three lions that were killed
by bushmeat poachers, thus reducing lion occur-
rence at these sites. In addition, by modelling prey
occurrence, we were able to exclude the influence
of prey depletion by bushmeat poaching, thus
limiting our poaching covariate to describe the
direct persecution of lions in LNP. However, apply-
ing this approach could mean that the total influence
of bushmeat poaching (prey depletion and perse-
cution) is underrepresented in our hierarchical
ranking of explanatory covariates.
Finally, our results demonstrate the scale depen-
dency of lion-habitat associations. While the
importance of prey resources to the use of habitat
: Factors affecting hierarchical use of habitat by an apex predator 193
Fig. 2. Habitat use by lions in the Limpopo National Park is influenced by bottom-up resources and by top-down
anthropogenic pressures including pastoralism and bushmeat poaching.
by lions in LNP spanned the domain of both spatial
scales examined, the limiting influences of the
anthropogenic covariates varied with spatial
scale. Our results show that considering habitat
selection by lions only at the home-range scale
would have greatly underestimated the direct im-
pacts of bushmeat poaching on lion ecology. This
is concerning because failing to recognize the im-
pact of bushmeat poaching could lead to errone-
ous conclusions of lion habitat suitability and
Mechanisms responsible for the species-habitat
relationships we present may include a behavioural
or numerical response. Avoidance of pastoralism
and bushmeat poaching by lions could indicate
that lions possess a behavioural mechanism to
reduce competition with humans (Schuette, Creel
& Christianson, 2013). Alternatively, if reduced site
use indicates a numerical response by lions then
this could suggest that the human-impacted lands
of LNP are acting as sink or as attractive-sink
(Battin, 2004) habitat to the adjoining (source)
habitat in KNP. In the context of acute continental
range declines and the isolation of lion popula-
tions, sink habitats, although low in quality, may
nonetheless be important to lion conservation
by increasing lion range and maintaining genetic
connectivity (Dolrenry, Stenglein, Hazzah, Lutz
& Frank, 2014; Stoner
., 2013). Under this
scenario, LNP may offer range expansion and
connectivity and thus play an important role in the
viability of lions in the Greater Limpopo Lion Con-
servation Unit. Conversely, if lions are mistakenly
selecting for human use areas (
for cattle as
prey and/or wildlife areas used by poachers) and
suffering high levels of mortality, LNP could be
acting as an attractive sink, which could reduce the
viability of the greater lion population (Battin,
2004). Distinguishing between sink and attrac-
tive-sink habitats is therefore important for improv-
ing lion conservation prospects in the system.
We thank the Parque Nacional do Limpopo for
supporting this research and the Director of Na-
tional Conservation Areas Mozambique for grant-
ing us the research permits (005-2011/003-2012)
to conduct this study. We thank Eden Everatt for
help and enthusiasm in the field. K.T.E. and L.A.
were supported by the May and Stanley Smith
Trust, The Wipplinger KL Bursary Found, Wilder-
ness Wildlife Trust and Canada National Student
Grants. The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript. We thank Matt
Hayward and an anonymous reviewer whose
input greatly improved the manuscript.
Andreasen, A.M., Stewart, K.M., Longland, W.S., Beck-
mann, J.P. & Forister, M.L. (2012). Identification
of source-sink dynamics in mountain lions of the
, 21, 5689–5701.
Andresen, L., Everatt, K.T. & Somers, M.J. (2014). Use
of site occupancy models for targeted monitoring
of the cheetah.
Journal of Zoology
, 292, 212–220.
Baldwin, R.A. & Bender, L.C.(2008). Distribution, occu-
pancy, and habitat correlates of American Martens
) in Rocky Mountain National
Journal of Mammalogy
, 89, 419–
Barber-Meyer, S.M., Jnawali, S.R., Karki, J.B., Khanal,
P., Lohani, S., Long, B., MacKenzie, D.I., Pandav, B.,
Pradhan, N.M.B., Shrestha, B.R., Subedi, N., Thapa,
G., Thapa, K. & Wikramanayake, E.(2013). Influence
of prey depletion and human disturbance on tiger
occupancy in Nepal.
Journal of Zoology
, 289, 10–18.
Battin, J. (2004).When good animals love bad habitats:
Ecological traps and the conservation of animal
ogy, 18, 1482–1491.
Becker, M., McRobb, R., Watson, F., Droge, E.,
Kanyembo, B., Murdoch, J. & Kakumbi, C. (2013).
Evaluating wire-snare poaching trends and the im-
pacts of by-catch on elephants and large carnivores.
, 158, 26–36.
Burnham, K.P. & Anderson, D.R. (2002).
and multimodel inference: a practical information-
, 2nd edn. Springer Science, New
Yo r k .
Ciarniello, L.M., Boyce, M.S., Seip, D.R. & Heard D.C.
(2007). Grizzly bear habitat selection is scale de-
, 17, 1424–1440.
Davidson, Z., Valeix, M., Van Kesteren, F., Loveridge,
A.J., Hunt, J.E., Murindagamo, F. & MacDonald, D.W.
(2013). Seasonal diet and prey preference of the
African lion in a waterhole-driven semi-arid savanna.
8, e55182. DOI: 10.1371/journal.pone.
Dickson, B.G. & Beier, P. (2002). Home-range and habitat
selection by adult cougars in southern California.
Journal of Wildlife Management
, 66, 1235–1245.
Dolrenry, S., Stenglein, J., Hazzah, L., Lutz & Frank, L.
(2014). A metapopulation approach to African lion
Everatt, K.T.(2014). Influence of bushmeat poaching and
pastoralism on African lion,
1758), ecology in rural Mozambique. (Unpublished
M.Sc. thesis). Pretoria: University of Pretoria.
Everatt, K.T., Andresen, L. & Somers, M.J. (2014).
Trophic scaling and occupancy analysis reveals a lion
population held below carrying capacity by top-down
anthropogenic pressure in the Limpopo National
. PLOS ONE
, 9, e99389. DOI:
Funston, P.J., Mills, M.G.L., Biggs, H.C. & Richardson,
P.R.K. (1998). Hunting by male lions: ecological influ-
194 African Journal of Wildlife Research Vol. 45, No. 2, September 2015
ences and socioecological implications.
, 56, 1333–1345.
Funston, P.J., Mills, M.G.L., Richardson, P.R.K. & Van
Jaarsveld, A.S. (2003). Reduced dispersal and op-
portunistic territory acquisition in male lions
Journal of Zoology
, 259, 131–142.
Gertenbach, W.P.D. (1980). Rainfall patterns in the
Kruger National Park.
, 23, 35–43.
Hatton, J., Couto, M. & Oglethorpe, J. (2001).
Biodiversity and war: a case study of Mozambique.
Washington, D.C.: Biodiversity Support Program.
Hanks, J. (2003). Transfrontier Conservation Areas
(TFCAs) in southern Africa: their role in conserving
biodiversity, socioeconomic development and pro-
moting a culture of peace.
Journal of Sustainable For-
, 17, 1–20.
Hayward, M.H. & Kerley, G.I.H. (2005). Prey preferences
of the lion (
Journal of Zoology
Hebblewhite, M., Miquelle, D.G., Murzin, A.A., Aramilev,
V.V. & Pikunov, D.G. (2011).Predicting potential habi-
tat and population size for reintroduction of the Far
Eastern leopards in the Russian Far East.
, 144, 2403–2413.
Hines, J.E. (2006). PRESENCE-Software to estimate
patch occupancy and related parameters. USGS-
Hopcraft, G.J.C., Sinclair, A.R.E & Packer, C. (2005).
Planning for success: Serengeti lions seek prey
accessibility rather than abundance.
Journal of Ani-
, 74, 559–566.
Huggins, G., Barendse., E., Fischer, A & Sitoi, J. (2003).
Limpopo National Park: resettlement policy frame-
work. Unpublished Report.
Johnson, D.H. (1980). The comparison of usage and
availability measurements for evaluating resource
, 61, 65–71.
Krebs, C.J. (2009).
Ecology: the experimental analysis
of distribution and abundance
. San Francisco:
Lindsey, P.A., Balme, G., Becker, M., Begg, C., Bento, C.,
Bocchino, C., Dickman, A., Diggle, R.W., Eves, H.,
Henschel, P., Lewis, D., Marnewick, K., Mattheus, J.,
McNutt, W., McRobb, R., Midlane, N., Milanz, J.,
Morley, R., Murphree, M., Opyene, V., Phadima, J.,
Purchase, G., Rentsch, D., Roche, C., Shaw, J., Van
der Westhuizen, H., Van Vliet, N. & Zisadza-
Gandiwa, P. (2013). The bushmeat trade in African
savannas: Impacts, drivers, and possible solutions.
, 160, 80–96.
MacKenzie, D.I. & Bailey, L.L. (2004). Assessing the fit of
Journal of Agricultural, Bio-
logical and Environmental Statistics
, 9, 300–318.
MacKenzie, D.I., Nichols, J.D., Lachman, G.B., Droege,
S., Royle, J.A. & Langtimm, C.A. (2002). Estimating
site occupancy rates when detection probabilities are
less than one.
, 83, 2248–2255.
MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H.,
Bailey, L.L. & Hines, J.E. (2006).
tion and modeling: inferring patterns and dynamics
of species occurrence.
London, U.K.:Elsevier Press.
McLoughlin, P.D., Case, R.L., Gau, R.J., Cluff, H.D.,
Mulders, R. & Messier, F. (2002). Hierarchical habitat
selection by barren-ground grizzly bears in the
central Canadian Arctic.
, 132, 102–108.
Mitchell, M.S. & Hebblewhite, M. (2012). Carnivore habi-
tat ecology: integrating theory and application. In
L. Boitani & R.A. Powell (Eds),
Carnivore ecology and
conservation: a handbook of techniques
255). Oxford:Oxford University Press.
Miquelle, D.G., Smirnov, E.N., Merril, T.W., Myslenkov,
A.E., Quigley, H.B., Hornocker, M.G. & Scheyer, B.
(1999). Hierarchical spatial analysis of Amur tiger
relationships to habitat and prey. In P.J. Seidensticker
Riding the tiger
tiger conservation in human-
(pp. 71–99). Cambridge:
Cambridge University Press.
Mosser, A., Fryxell, J.M., Eberly, J. & Packer, C. (2009).
Serengeti real estate: density
indicators of lion habitat quality.
Nams, V.O., Mowat, G. & Panian, M.A. (2006). Deter-
mining the spatial scale for conservation purposes –
an example with grizzly bears.
, 128, 109–119.
Ogutu, J.O., Bhola, N. & Reid, R. (2005). The effects
of pastoralism and protection on the density and
distribution of carnivores and their prey in the Mara
ecosystem of Kenya.
Journal of the Zoological Soci-
ety of London
, 265, 281–293.
Rettie, W.J. & Messier, F. (2000). Hierarchical habitat
selection by woodland caribou: its relationship to
, 23, 466–478.
Rich, L.N., Mitchell, M.S., Gude, J.A.& Sime, C.A. (2012).
Anthropogenic mortality, intraspecific competition,
and prey availability influence territory sizes of
wolves in Montana.
Journal of Mammalogy
, 93, 722–
Riggio, J., Jacobson, A., Dollar, L., Bauer, H., Becker, M.,
Dickman, A., Funston, P., Groom, R., Henschel, P.,
De Iongh, H., Lichtenfeld, L. & Pimm, S. (2012). The
size of savannah Africa: a lion’s (
Biodiversity and Conservation
Ripple, W.J., Estes, J.A., Beschta, R.L. Christopher C.
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, 1241484.
Schuette, P., Creel, S. & Christianson, D. (2013).Coexis-
tence of African lions, livestock, and people in a land-
scape with variable human land use and seasonal
, 157, 148–154.
Stephensen, A. (2010). Parque Nacional do Limpopo
aerial wildlife census. Unpublished report for Parque
Nacional do Limpopo.
Stoner, D.C., Wolfe, M.L., Rieth, W.R., Bunnell, K.D.,
Durham, S.L. & Stoner, L.L.(2013). De facto refugia,
ecological traps and the biogeography of anthro-
pogenic cougar mortality in Utah.
Diversity and Distri-
Sunarto, S., Kelly, M.J., Parakkasi, K., Klenzendorf, S.,
Septayuda, E. & Kurniawan, H. (2012). Tigers need
cover: Multi-scale occupancy study of the big cat
in Sumatran forest and plantation landscapes.
Thaker, M., Vanak, A.T., Owen, C.R., Ogden, M.B.,
: Factors affecting hierarchical use of habitat by an apex predator 195
Niemann, S.M. & Slotow, R. (2011). Minimizing pre-
dation risk in a landscape of multiple predators: ef-
fects on the spatial distribution of African ungulates.
, 92, 398–407.
Van Orsdol, K.G., Hanby, J.P.& Bygott, J.D.(1985). Eco-
logical correlates of lion social organization
Journal of Zoology
, 206, 97–112
Whitman, K.L. & Packer, C. (2007).
A hunter’s guide to
aging lions in eastern and southern Africa
ton Beach, CA. Long Beach Press.
Woodroffe, R. & Ginsberg, J. (1998). Edge effects and
the extinction of populations inside protected areas.
, 280, 2126–2128.
Woodroffe, R. & Frank, L.G. (2005).Lethal control of Afri-
can lions (
): local and regional popula-
, 8, 91–98.
Responsible Editor: M.W. Hayward
196 African Journal of Wildlife Research Vol. 45, No. 2, September 2015
The influence of prey, pastoralism and poaching on the hierarchical use of habitat by an
Kristoffer T. Everatt1*, Leah Andresen1, Michael J. Somers1, 2
1 Centre for Wildlife Management, University of Pretoria, Pretoria, South Africa
2 Centre for Invasion Biology, University of Pretoria, Pretoria, South Africa
Warthog occupancy model
Twenty-three lion grid cells were surveyed for warthogs (mean = 3.6 camera sites / lion grid
cell). Active camera stations were located > 4 km apart. Sampling occasions (n = 797; mean
= 9.7 / site; range = 2 - 22) were represented by seven day intervals. We made the
assumptions of an occupancy model, but note that the closure assumption was relaxed
because the estimator was probability of site use (MacKenzie et al., 2006). Warthog spatial
use is influenced by the nutritional quality of vegetation, water availability and predation risk
(Estes, 1991). To describe heterogeneity in warthog site use, we used six landscape covariates
accounting for variation in vegetation communities, underlying geology, surface water
availability, topography and anthropogenic disturbance (Table S1).
Table S1. Covariates expected to influence occurrence of warthog.
Covariate Fitness value to
warthog Description Sampling
Nutritional variation Shrublands and thickets of
Colophospermum mopane on
Nutritional variation Woodlands and shrublands of
Combretum spp. and
Colophospermum mopane on
Lebombo hills Nutritional variation Short woodlands and shrublands
of Combretum apiculatum on
Sand plains Nutritional variation Short woodlands and thickets of
Baphia massaiensis and
Combretum apiculatum on
Water Water availability
Proximity to rivers measured in
0.0 – 9.3
mean = 3.7
Village Direct persecution Proximity to settlements
measured in ArcGIS (km)
0.5 – 22.7
mean = 11.7
A detection/non-detection matrix was constructed for each site (n = 82), recording a
‘1’ or ‘0’ where warthog were detected or not detected, respectively. Similarly, a season (wet
versus dry) specific matrix was built to account for differing detection probabilities
throughout the year (1 = November - April, 0 = May - October). First, covariates describing
heterogeneity in warthog detection probability were evaluated. The detection covariate for
season was included in all the following analysis; the model with this covariate was strongly
supported (∆AICc < 2) and ranked higher than the model that assumed detectability was
constant. Following this, we compared all possible combinations of occupancy covariates (63
models). Final covariate values were extracted as mean warthog site use from a continuous
(30 m x 30 m resolution) raster layer using the Spatial Analysis toolbox in ArcGIS 9.3.1.
Warthog site use
The model selection procedure for warthog site use is provided in Table S3. Model averaged
estimates showed that the probability of detecting warthogs at a site where they occur was
̂ = 0.336 (SE = 0.035). Site level estimates ranged from 0.008 (SE = 0.011) to 0.771 (SE =
0.004) with a weighted average of 0.513 (SE = 0.049). Site use by warthogs increased
strongly with distance from villages (Table S4). There was no evidence lack of fit (p = 1.10)
or over-dispersion ( ˆ
c = 0.20).
Table S2. Summary of model selection procedure for factors influencing warthog site use (Ψ)
across 82 sites in the Limpopo National Park, Mozambique. Covariates considered include;
distance from villages (V), combretum/mopane rugged veld (C), distance from water (W),
sand plains (SP) and mopane shrubveld (M). Detectability (p) varies with season (S).
Models ∆AICc w K -2l
Ψ(V)p(S) 0.00 0.267 4 538.42
Ψ(V+C)p(S) 0.20 0.241 5 536.35
Ψ(V+W+C)p(S) 1.12 0.152 6 534.94
Ψ(V+SP)p(S) 1.60 0.120 5 537.75
Ψ(V+W)p(S) 1.67 0.116 5 537.82
Ψ(V+M)p(S) 1.96 0.100 5 538.11
Ψ(.)p(S) 38.84 0.000 3 579.47
Ψ(.) assumes site use is constant, ∆AICc is the difference in AICc values between each model
with the low AICc model, w is the AICc model weight, K is the number of parameters in the
model, and −2l is twice the negative log-likelihood value.
Table S3. β- coefficient estimates for covariates influencing warthog site use (Ψ) in order of
their summed model weights (∑w).
Occupancy Covariate ∑ model w (%) β SE
Village 99.6 -2.95* 0.77
Combretum/Mopane rugged veld 39.4 0.61 0.52
Water 26.8 1.39 1.05
Sand plains 12.0 -0.79 1.00
Mopane shrubveld 10.0 -0.57 1.02
* Indicates covariate has robust impact (β ± 1.96 x SE not overlapping 0).
Lion occupancy model
Table S4. Results of Pearson’s r correlation test from lion occupancy models.
Home range Short term
Covariates r Covariates r
B+P 0.000 B+P 0.000
B+AP -0.386 B+AP -0.363
B+W* 0.561 B+W 0.254
B+V 0.373 B+V -0.431
P+AP* 0.702 P+AP* 0.544
P+W 0.235 P+W 0.137
P+V* -0.695 P+V* 0.526
AP+W -0.039 AP+W 0.051
AP+V* -0.943 AP+V* 0.895
W+V -0.069 W+V -0.055
* indicates covariates that are correlated using a cut-off of |r| = 0.5 and were therefore not
combined in models.
ESTES, R.D. 1991. The behavior guide to African mammals. University of California Press,
Berkley, 611 p.
MACKENZIE, D.I., NICHOLS, J.D., ROYLE, J.A., POLLOCK, K.H., BAILEY, L.L. &
HINES, J.E. 2006. Occupancy estimation and modeling: Inferring patterns and
dynamics of species occurrence. Elsevier Press, London, UK.