Content uploaded by Dawn M. Scott
Author content
All content in this area was uploaded by Dawn M. Scott on Apr 30, 2015
Content may be subject to copyright.
1 23
Biodiversity and Conservation
ISSN 0960-3115
Volume 22
Number 8
Biodivers Conserv (2013) 22:1715-1730
DOI 10.1007/s10531-013-0508-2
Characteristics and determinants of
human-carnivore conflict in South African
farmland
Michelle Thorn, Matthew Green, Dawn
Scott & Kelly Marnewick
1 23
Your article is protected by copyright and all
rights are held exclusively by Springer Science
+Business Media Dordrecht. This e-offprint
is for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.
ORIGINAL PAPER
Characteristics and determinants of human-carnivore
conflict in South African farmland
Michelle Thorn •Matthew Green •Dawn Scott •Kelly Marnewick
Received: 27 November 2012 / Accepted: 6 June 2013 / Published online: 5 July 2013
ÓSpringer Science+Business Media Dordrecht 2013
Abstract Reducing human-carnivore conflict may provide conservation, social and
economic benefits, but designing suitable mitigation activities requires information
regarding underlying anthropogenic and environmental determinants. To obtain those data,
we interviewed game and livestock farmers in Limpopo province, South Africa from
March to August 2011. We quantified (1) human-carnivore conflict characteristics; (2)
determinants of perceived carnivore predation levels and; (3) determinants of retaliatory
persecution of carnivores. Carnivores reportedly killed 1.4 % of total game and domestic
livestock holdings and the median annual rate of loss was ZAR 1.23 ha
-1
, which is
generally not sufficient to threaten farming livelihoods or the provincial economy. Farmers
reportedly killed 303 carnivores in the year prior to the interview, 44 of which were
threatened species. African wild dogs were the least tolerated species, followed by chee-
tahs, although these species were only blamed for 6 and 3 % of reported kills respectively.
The main determinants of human-carnivore conflict were high elevation, mixed purpose
farming (i.e., stocking both game and livestock), dense vegetation cover and high per-
ceived financial losses. The results suggest a number of potential conflict-mitigation
strategies which include addressing predation problems in areas with a high proportion of
vulnerable farms, improving tolerance of carnivores among low-tolerance social groups
and correcting misconceptions about the provenance and financial impact of local carni-
vore populations. This information has already lead to the implementation of several
suitably tailored mitigation activities.
M. Thorn (&)K. Marnewick
Centre for Wildlife Management, Hatfield Experimental Farm, University of Pretoria,
Private Bag 0002, Pretoria, South Africa
e-mail: Thorn_Green@Hotmail.com
M. Thorn M. Green K. Marnewick
Carnivore Conservation Programme, Endangered Wildlife Trust, Private Bag X11, Modderfontein,
Johannesburg 1645, South Africa
D. Scott
Biology Division, University of Brighton, Huxley Building, Lewes Road, Moulsecoomb,
Brighton BN2 4GJ, UK
123
Biodivers Conserv (2013) 22:1715–1730
DOI 10.1007/s10531-013-0508-2
Author's personal copy
Keywords Game Livestock Perceptions Persecution Predation Questionnaire
interview
Introduction
In the last century, many carnivore species have undergone declines in distribution and
population size (Woodroffe and Ginsberg 2000). This is often due to human-carnivore
conflict in non-protected areas (especially those bordering reserves), where the interests of
co-existing people and carnivores clash (Stein et al. 2010; Woodroffe and Ginsberg 1998). In
agricultural landscapes, human-carnivore conflict most often arises when carnivores prey
upon domestic livestock or game. For example wolves (Canis lupus) and bears (Usrus spp.)
prey upon sheep in North America and Europe (Prugh et al. 2009; Treves and Karanth 2003;
Rigg et al. 2011), tigers (Panthera tigris) and leopards (Panthera pardus) kill livestock in
Asia (Treves and Karanth 2003) and species such as lions (P. pardus), cheetahs (Acinonyx
jubatus), leopards, caracals (Caracal caracal) and black-backed jackals (Canis mesomelas)
feed on cattle, sheep, goats and game in Africa (Woodroffe and Frank 2005; Thorn et al. 2012;
Marker et al. 2003). People frequently respond by killing suspected predators and such
persecution is now among the leading world-wide threats to large carnivores (Graham et al.
2005; Inskip and Zimmermann 2009; Treves and Karanth 2003; Stein et al. 2010). At the
same time, predation of human-owned animals can cause loss of income or food, which can
have serious adverse impacts on individual farmers, agricultural production, food security
and rural development (Graham et al. 2005; Treves and Karanth 2003; Woodroffe et al.
2007). Thus, in areas where animal production is prevalent, mitigating human-carnivore
conflict may be of conservation, social and economic concern.
Here, we investigate conflict between co-existing carnivores and people in Limpopo,
South Africa. The agricultural sector is Limpopo’s second largest source of employment
(Limpopo DFED 2003) and our study area has the highest proportion of grazing land in the
province (de Klerk 2003), further accentuating the local economic importance of animal
production. We focused on non-protected areas, where intensive and extensive commercial
animal production, stud breeding, sport hunting, and wildlife-based tourism on privately-
owned farms are the predominant land uses. Anecdotal information from local researchers,
conservation workers and the Limpopo Department of Economic Development, Envi-
ronment and Tourism suggests that carnivore predation of human-owned animals and
retaliatory persecution of carnivores is common.
Several South African carnivore species, e.g. cheetahs, leopards, brown hyaenas
(Hyaena brunnea) and African wild dogs (Lycaon pictus) are protected under the National
Environmental Management: Biodiversity Act (NEMA 2004), which also prescribes per-
missible methods of lethal control for non-protected wildlife. However, the institutions
charged with managing human-wildlife conflict and enforcing biodiversity legislation in
Limpopo are hampered by corruption, overlapping responsibilities and a fundamental lack
of financial and human resource capacity (Anthony et al. 2010). Hence, compliance with
governmental policy cannot simply be compelled, especially in remote rural areas. Sus-
tainable human-wildlife conflict-mitigation strategies must therefore foster voluntary
implementation by balancing human interests with carnivore conservation objectives. This
requires a clear understanding of the characteristics and determinants of conflict.
Conflict determinants are relatively well researched in Europe and North America and
there have been a number of prior African studies (Graham et al. 2005; Inskip and
1716 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
Zimmermann 2009). Findings therein identify several potentially useful mitigation options
but also suggest that conflict arises from a range of inter-related social, economic, envi-
ronmental and ecological factors whose effects vary spatially and temporally (Graham
et al. 2005; Inskip and Zimmermann 2009; Woodroffe et al. 2007). For example, deter-
minants of conflict in Tanzania included education, predation rates and the use of effective
anti-predation measures (Holmern et al. 2007). Factors affecting attitudes to wolves in
America included social group and occupation (Naughton-Treves et al. 2003) and in
Slovakia, husbandry methods and occupation accounted for most of the variation in atti-
tudes to wolves and bears in farmland (Rigg et al. 2011). Selection and successful
application of conflict-mitigation strategies therefore depends upon accurate knowledge of
the local context and circumstances.
Few prior publications (see Anthony et al. 2010; Balme et al. 2009; Blaum et al. 2009;
Lindsey et al. 2005; St John et al. 2011) have addressed human-carnivore conflict in South
Africa and only one study in the North West Province of South Africa (Thorn et al. 2012)
has evaluated the relative contribution of social, economic and environmental drivers. By
comparison with the North West Province, our study area supports relatively high occu-
pancy of large carnivores (Lindsey et al. 2004; Daly et al. 2005; Friedmann and Daly 2004;
Thorn et al. 2011b). We expect this difference to affect carnivore dietary niches, which in
turn influence the predator–prey interactions that are at the heart of human-carnivore
conflict in farmland. Our aim was therefore to assess the characteristics and determinants
of conflict in Limpopo, thereby informing the development of suitably tailored conflict-
mitigation options. Accordingly, we interviewed game and livestock farmers throughout
the study area to obtain baseline measures of (1) human-carnivore conflict characteristics;
(2) determinants of perceived carnivore predation levels and; (3) factors determining
human persecution of carnivores.
Methods
Study area
The study was conducted in the UNESCO Waterberg Biosphere Reserve and surrounding
areas of the Waterberg District Municipality (hereafter collectively termed ‘the Water-
berg’) in Limpopo Province, South Africa (Fig. 1). The terrain within the 4,174 km
2
Biosphere Reserve consists mainly of rolling hills, but includes the Waterberg Mountain
range. Elevation is from 808 to 2,100 m and annual rainfall is between 600 and 900 mm
(de Klerk 2003). The Waterberg is within the savanna biome and sour and mixed bushveld
vegetation types predominate (de Klerk 2003).
The most common agricultural enterprise in the study area is game farming, which involves
extensive management of plains game (de Klerk 2003). Game farming income is derived
mainly from sport hunting for meat or trophies, sale of live animals, stud breeding and tourism
(Bothma 2005). Game farming is increasingly prevalent in South Africa because it is often
more profitable than livestock production (Bothma 2005). In recent years, breeding of inten-
sively managed high-value game (e.g., rare antelope colour morphs or sub-species) has also
become popular due to rapidly escalating sale prices. For example, the average 2011 auction
price for a standard impala (Aepyceros melampus) male was ZAR 1344, compared with ZAR
131 667 (9697 % higher) for a black colour morph of the same species (Vleissentraal 2011).
Domestic livestock production is also a common land use. Livestock is sometimes
managed extensively, but is usually intensively farmed in pastures. Income from livestock
Biodivers Conserv (2013) 22:1715–1730 1717
123
Author's personal copy
farming comprises mainly meat, dairy and wool sales from cattle and small-stock (pigs,
sheep and goats) (Department of Agriculture Forestry and Fisheries 2010). Stud breeding is
also common and although such high-value animals are often intensively managed, this is
not always the case. Many farmers combine game and livestock farming or stud breeding
(mixed farming).
Several carnivore species that may prey upon game and livestock are resident in
Waterberg farmland. These include caracals, black-backed jackals (hereafter referred to as
jackals), leopards, cheetahs, brown hyaenas and African wild dogs. Transient spotted hya-
enas (Crocuta crocuta) and lions (Panthera leo) occasionally occur in our study area, but are
resident only inside reserves with predator-proof fencing (Friedmann and Daly 2004).
Interview methods and questionnaire design
To obtain information on human-carnivore conflict, we visited farmers throughout the
Waterberg between March 2011 and August 2011. We interviewed 92 people who were
responsible for the management of 95 farms (Fig. 1). Initial respondents were recruited at
local farming forums and further contacts were supplied by each participant. With a refusal
rate of 8 %, non-response bias was assumed to be minimal. All but one of our respondents
were white, reflecting typical land tenure patterns in South African commercial farms
(Tladi et al. 2002). As in Marker et al. 2003, we considered white farmers to be a
homogeneous group and therefore assumed our respondents to be approximately repre-
sentative of the wider population of farmers in the study area. While not inherently
Fig. 1 Study area and interview locations (n=92). The black patch on the inset indicates the position of
the study area in South Africa
1718 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
unrealistic, this assumption could not be verified as there are no available reference sta-
tistics for land ownership by ethnic group. Our interpretation of the wider effects of conflict
therefore remains tentative.
Interviews were administered in person by M.T. Each lasted approximately 40 min and
was based on a semi-structured socio-economic questionnaire (available from M.T.)
adapted from a previous study in a neighbouring province (Thorn et al. 2012). The
questionnaire contained four sections concerning (1) characteristics of respondents and
their farms; (2) carnivore presence, predator–prey interactions and the perceived cost of
predation; (3) use of anti-predation measures and lethal control of carnivores and; (4)
degree of tolerance toward carnivores. All respondents were assured of anonymity and
confidentiality.
The characteristics we recorded about respondents included age and cultural group,
which was inferred from the respondent’s first language. Recorded characteristics of farms
included elevation, measured with a hand-held GPS at the entrance to the farm. Respon-
dents estimated the proportion of grassland and crop areas (open cover habitats) on their
land, compared with scrub and woodland areas (dense cover habitats). From this, we
determined the predominant category of vegetative cover at each farm (open, dense or
heterogeneous). This method of allocating cover categories was ground-truthed and vali-
dated in a subset of 36 farms where extensive scat searches were completed in addition to
questionnaire surveys.
Game and livestock abundance and losses were also based on land owner estimates.
Livestock abundance estimates are considered likely to be accurate as these species are
usually intensively managed and frequently counted. The accuracy of game estimates may
be more variable as these species are often extensively managed and abundance is usually
estimated annually/bi-annually. Furthermore, the survey methods employed during game
counts vary widely in their suitability and reliability. Many land owners cited lack of
population growth in game species as indirect evidence of predation, but as such losses
were unquantifiable and could be caused by factors other than carnivore predation, they
were not recorded.
To assess whether reported financial losses were congruent with market values, we
compared sale prices estimated by respondents with independent data such as sale prices at
game and livestock auctions in 2011. Although we cannot corroborate the number of
animals reportedly lost, we found that in general, respondents did not over-estimate pre-
dated game and livestock values. We were not able to verify the estimated values given for
stud animals because prices vary according to the breed, sex, size and genetic properties of
individual animals.
Analysis
We summarised human-carnivore conflict characteristics using standard descriptive sta-
tistics. We then calculated financial rates of loss (cumulative Q1, median and Q3 ZAR
losses/cumulative area surveyed) and extrapolated them over the total area of commercial
grazing land in Limpopo province to give a coarse estimate of the order of magnitude of
the wider impact of current levels of predation. Similarly, we calculated persecution rates
for each carnivore species (total number killed/total area surveyed), but limited our
extrapolation to the survey area as several of the focal species have restricted ranges, i.e. do
not occur throughout the province (Friedmann and Daly 2004).
We evaluated the effect of potential conflict determinants using two sets of generalised
linear regression models. Prior to analysis, we excluded any variables that were inter-
Biodivers Conserv (2013) 22:1715–1730 1719
123
Author's personal copy
correlated by Spearman’s rho(r
s
)[0.7 (Kolowski and Holekamp 2006). The first set of
models evaluated the effect of farm-specific and environmental predictors (Table 1) on the
number of game and livestock reportedly preyed upon at each farm, using a negative
binomial error distribution and a log link function. Secondly, we modelled the effects of
social, economic and environmental predictors (Table 1) on the probability of carnivore
persecution using the binary response variable ‘use or non-use of lethal control’, a binomial
error distribution and a logit link function.
For each response variable, we evaluated a set of candidate models containing all
additive combinations of predictors as well as all two-way interactions. We then used
standard information-theoretic multi-model inference techniques to interpret the results
(Burnham and Anderson 2002; Dochterman and Jenkins 2011) based on Akaike’s
Table 1 Predictor variables and their anticipated effect on response variables that indicate human-carni-
vore conflict
Predictor variable Predicted effect
Response variable: no. of game and livestock reportedly killed by carnivores at each farm
Anti-predation measures (lethal control,
non-lethal methods, both or neither)
Indiscriminate lethal control is unlikely to reduce
predation because it does not necessarily remove
habitual livestock predators
a
and has little long-
term effect on large carnivore density
b,c
. Non-
lethal measures such as protective enclosures have
been shown to reduce predation losses in a number
of studies world-wide
d
Land use (livestock, game or mixed
farming)
Intensive management and associated anti-predation
measures on livestock farms should reduce
predation levels compared with extensively
managed game or mixed farms
Total prey density (individuals/km
2
) High stocking densities should increase the
frequency with which carnivores encounter prey,
leading to higher levels of predation
Cover (open, dense, or equal proportions
of open and dense cover, i.e.
heterogeneous)
Dense vegetation increases the cover available to
carnivores, which has been shown to increase
predation
d,e
Response variable: use (1) or non-use (0) of lethal control
Predation losses (ZAR) High financial losses are expected to increase the
use of lethal control
b
Land use (livestock, game or mixed
farming)
Predation of intensively managed livestock is easier
to detect and quantify than loss of extensively
managed game. We therefore suspect that greater
awareness of losses coupled with lower
profitability of livestock farming
f
may provoke
increased use of lethal control among livestock
farmers
Farmer age (years) Previous studies found that younger farmers tend to
be more tolerant of carnivores
e
and should,
therefore, be less likely to use lethal control
Farmer cultural group (Afrikaans or
English)
We expect tolerance toward large carnivores and
thus, prevalence of lethal control to differ among
cultural groups
e
Sources:
a
Inskip and Zimmermann (2009),
b
Woodroffe and Frank (2005),
c
Prugh et al. (2009),
d
Woodroffe et al. (2007),
e
Thorn et al. (2012),
f
Bothma (2005)
1720 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
Information Criterion adjusted for small sample size (AIC
c
) (see Thorn et al. 2012 for a
more detailed description of the procedure).
Analysis was conducted using SPSS v17 (SPSS Inc., Chicago, USA). Financial values
throughout are given in South African Rand (ZAR 1 =GBP 0.08 =USD 0.13).
Results
Characteristics of the sample
Most respondents (73 %) were members of the Afrikaans or Afrikaans and German-speaking
cultural group. The remainder were English-speaking (23 %) and 4 % were from Bantu-
speaking or non-South African cultural groups. Most (92 %) were full time residents at their
farm with mean tenure of 16 years (±SE 1.5). The total area of the farms surveyed was
231,549 ha (median =1,300 ha, range =6.2–16,000 ha). This represents 23 % of farm-
land within the whole study area, which we calculated as the area within the minimum convex
polygon formed from the outermost interview locations. Forty three percent of farms were
used mainly for game farming, 35 % for cattle or small-stock and 22 % for mixed farming (a
roughly equal area used for game and domestic livestock). Total estimated stock holdings
(66,520 animals) consisted of 1031 sheep, 1242 goats, 1575 pigs, 11,467 cattle (including
3,861 stud animals) and 51,205 game, 3,020 of which were high-value antelopes. High-value
animals (stud livestock or rare antelope) were present in 37 % of the farms surveyed, but were
stocked in (67 %) of mixed farms. Median elevation was 1077 m (range 822–1,600 m).
Farms with predominantly dense cover accounted for 74 % of the sample, compared with
17 % farms with open cover and 9 % farms with heterogeneous cover.
Lethal control of carnivores was reportedly practiced at 45 farms (47 %), where 92 %
of the carnivores that were destroyed were shot. Non-lethal anti-predation methods were
reportedly used by 79 % of respondents, the most popular being enclosures with fencing
intended to exclude predators (n=61), moving potential prey animals to open areas with a
lower risk of predation (n=29) and natural anti-predator adaptations (stocking native,
predator-adapted breeds and not dehorning livestock) (n=17). Neither lethal nor non-
lethal methods were reportedly used by 15 % of respondents, all of whom were extensive
game farmers. We presume that they were unconcerned about predation or did not consider
anti-predation efforts practical. Thirty five percent of respondents reportedly used both
lethal and non-lethal measures.
Human-carnivore conflict characteristics
Perceived predation losses
Eighty six percent of respondents reported perceived predation losses totalling 934 animals
(1.4 % of total stock holdings used for commercial purposes) in the year preceding
interviews (Table 2). Perceived losses as a function of elevation, land use, cover and anti-
predation methods are illustrated in Fig. 2.
The median annual number of animals reportedly predated in game farms was 3
(Q1 =1, Q3 =8.5), compared with 5 (Q1 =0.5, Q3 =7.5) in livestock farms and 12
(Q1 =3.5, Q3 =25) in mixed farms. Reported median financial losses were ZAR 6000
(Q1 =ZAR 800, Q3 =ZAR 26500) in game farms, compared with ZAR 11450
(Q1 =ZAR 0, Q3 =ZAR 34100) in livestock farms and ZAR 15800 (Q1 =ZAR 8000
Biodivers Conserv (2013) 22:1715–1730 1721
123
Author's personal copy
Table 2 Kills attributed to carnivores (predation) and carnivores killed by respondents (persecution) in the
year preceding interviews (n=92)
Carnivore species Predation Persecution
Small-stock Cattle Game Total wNo. killed Persecution
rate/100 km
2
Jackal 62 24 33 119 0.98 229 9.89
Caracal 39 2 103 144 0.89 14 0.60
Leopard 18 95 155 268 0.87 16 0.69
Cheetah 0 1 26 27 0.39 0 –
African wild dog 0 7 51 58 0.42 15 0.65
Brown hyaena 0 8 13 21 0.97 27 1.17
Serval 2 0 0 2 0.46 1 0.04
Honey badger 0 0 3 3 0.59 1 0.04
Unknown 44 45 203 292
Total 165 182 587 934 303 13.09
w=proportion of farms occupied by carnivore species, persecution rate =number of carnivores killed/
total farm area surveyed
(a) (b)
(c) (d)
Fig. 2 The number of animals per farm (n=95) reportedly preyed upon by carnivores, in relation to
elevation (a), land use (b), cover (c) and anti-predation methods used (d). Box plots display median, 25 and
75 % quartiles, maximum, minimum and outliers (asterisks)
1722 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
Q3 =ZAR 66500) in mixed farms. High-value animals comprised 11 % of reportedly
predated animals in mixed farms, which means that the financial losses perceived by these
respondents arose mainly from frequent predation of relatively low-value animals. Per-
ceived losses as a function of land use and use of lethal control are illustrated in Fig. 3.
Overall, the majority of animals perceived to have been predated were game species
(63 %), followed by cattle (19 %) and small-stock (18 %) (Table 2). Animals B1 year old
accounted for 54 % of predated animals.
The estimated total value of reportedly predated animals was ZAR 2792873, with a
median annual loss per farm of ZAR 12000 (Q1 =ZAR 1000, Q3 =ZAR 30000).
Extremely high losses were reported by seven respondents, who each lost sums between
ZAR 105800 and ZAR 392500. Four of the seven were stud or high-value game breeders
who lost a relatively small number of very expensive animals.
Perceived cumulative financial losses at the median, divided by the cumulative area of the
farms surveyed, gives a median annual rate of loss of ZAR 1.23 ha
-1
(Q1 =ZAR 0.03 ha
-1
,
Q3 =ZAR 3.98 ha
-1
). Extrapolating that rate of loss across all commercial grazing land in
Limpopo produces a perceived annual predation loss of ZAR 7385509, which equates to
0.15 % of Limpopo agricultural GDP in 2010 (Statistics South Africa 2011).
Perceived culpability, persecution and tolerance of carnivores
The majority of predation losses were attributed to leopards, followed by caracals and
jackals (Table 2). Most respondents (67 %) inferred culpability from indirect evidence like
spoor, bite marks and feeding patterns, 20 % witnessed attacks or photographed carnivores
assumed to be returning to a kill, and 13 % offered no supporting evidence.
Forty two respondents reportedly killed a total of 303 carnivores in the year preceding
the interviews (Table 2). These included a serval (Leptailurus serval) and a honey badger
(Mellivora capensis) thought to have been preying on poultry. Annual species persecution
rates (Table 2) extrapolated over the whole study area equate to 1096 jackals, 129 brown
hyaenas, 77 leopards, 72 African wild dogs and 67 caracals.
Surprisingly, the species most often blamed for predation (Table 2) were not the ones
that were least tolerated. Only 23 % of respondents said that they would tolerate African
ZAR losses ('000)
MixedLivestockGame
400
300
200
100
0
Do not use lethal control
Use lethal control
Land use
Fig. 3 Reported financial losses
(median, 25 % and 75 %
quartiles, maximum, minimum
and outliers (asterisks)), in
relation to land use and lethal
control of carnivores
Biodivers Conserv (2013) 22:1715–1730 1723
123
Author's personal copy
wild dogs if they were regularly present at their farm, compared with 58 % for cheetahs,
73 % for leopards, 78 % for jackals, 79 % for caracals, and 92 % for brown hyaenas. Of 78
people who answered the question, 42 % said that they would cross the threshold from
tolerance of carnivores to intolerance if B1 % of their stock were preyed upon. Another
23 % reached that point after losing between 1 and 5 % of their stock and 17 % gave a
tolerance threshold of between 5 and 35 % of stock holdings. A further 18 % felt that
predation losses were a natural and acceptable hazard no matter how many animals were
preyed upon. The median percent loss that farmers with high-value stud or game animals
were willing to tolerate was 83 % lower than famers without high-value animals
(W=2103.5, P\0.001, n=78) and two thirds of respondents who said they were
willing to tolerate losing B1 % of their stock were farming high-value animals.
There was no correlation (r
s
=0.467, P=0.243, n=8) between the stock losses
attributed to each carnivore species and species-specific persecution levels (Table 2).
However, there was a strong positive correlation between carnivore occupancy (the pro-
portion of farms where the species was reportedly present) and species-specific persecution
levels (r
s
=0.802, P=0.017, n=8). The implication is that carnivores were persecuted
according to their availability and not in relation to their perceived culpability for pre-
dation. Indeed, when asked why they had killed carnivores, only 35 % of people who used
lethal control said they were trying to remove a specific animal suspected of attacking their
game or livestock. Another 37 % killed carnivores as a general precaution against future
predation, 21 % earned income from carnivore sport or trophy hunting, and 7 % killed
carnivores to control their population numbers.
Determinants of predation
The global additive model showed no evidence of a lack of fit (v2
80 =94.913, ^
c=1.186,
n=94). Seven of 78 candidate models were within 7 Akaike units of the top ranked model
(Table 3). Elevation (summed Akaike weight 0.86) and land use (summed Akaike weight
0.83) were the most influential variables, followed by cover (summed Akaike weight 0.79)
and anti-predation methods (summed Akaike weight 0.65). Prey density (summed Akaike
weight 0.12) and terrain (summed Akaike weight 0.01) had little effect.
Beta coefficients from regression models (Table 4) indicate that after controlling for the
effect of other variables, predicted predation frequency in farms at high elevation (1,600 m)
was almost five times that of farms at low elevation (822 m). Predicted predation frequency in
mixed farms was three times greater than in game farms. Predicted predation frequency in
farms with predominantly dense or heterogeneous cover was more than twice that of farms
with open cover. Unsurprisingly, farms that did not use any anti-predation measures were
associated with the highest predation losses. Predicted losses in those farms were over twice
as high as in farms using non-lethal measures. The same is true of farms using both lethal and
non-lethal anti-predation techniques, but in this case, cause and effect are confounded: the
increased probability of persecution may be a reaction to high predation levels or alterna-
tively, lethal control may increase predation levels (e.g. by triggering compensatory
responses such as increased natality or immigration) (Prugh et al. 2009).
Determinants of persecution
There was no evidence of a lack of fit in the global additive model (v2
88 =91.487,
^
c=1.04, n=94). Three of the 22 candidate models were within 7 Akaike units of the top
1724 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
ranked model (Table 3). Land use was by far the most influential determinant (summed
Akaike weight 0.99), followed by financial losses (summed Akaike weight 0.79). Cultural
group (summed Akaike weight 0.16) and age (summed Akaike weight 0.12) had little
effect.
Regression coefficients were taken from model 8 (Table 3), which received consider-
ably greater support than any other model (0.79 Akaike weight). After controlling for the
effect of other variables, a mixed farmer with losses at the 75th percentile (ZAR 30 000)
would be over 43 times more likely to kill carnivores than a mixed farmer with losses at the
25th percentile (ZAR 1000).
Discussion
Data from interviews comprise reported information, some of which may be subject to
bias. Declared predation rates may be negatively biased if stock abundance is poorly
monitored and if owners fail to find or report all predated carcasses. Positive bias may arise
if respondents incorrectly assume predation to be the cause of death, or deliberately inflate
losses in pursuit of ulterior motives (Rasmussen 1999). Similarly, reported persecution
rates may be negatively biased if fear of prosecution or censure inhibits respondents who
use illegal methods of lethal control or kill legally protected carnivores (St John et al.
2011). Nevertheless, capturing and incorporating information from affected communities
provides essential contextual data that are not available from any other source and this is
essential in designing effective human-wildlife conflict solutions (Anthony et al. 2010).
Furthermore, exploratory analyses such as ours are vital for the identification of key
parameters in complex systems where there is little or no pre-existing information
(Dochterman and Jenkins 2011).
Our modelling results indicate that the probability of predation was greatest in farms at
high elevation, those with predominantly dense or heterogeneous cover and those used for
mixed farming. Hence, areas with a high proportion of farms possessing these
Table 3 Generalised linear models that were within 7 AIC
c
units of the model with the lowest AIC
c
, with
associated degrees of freedom (df), number of parameters in the model (k), AIC
c
,DAIC
c
and Akaike model
weights
Model df kAIC
c
DAIC
c
Weight
Response variable: no. of game and livestock reportedly killed by carnivores at each farm
1. Land use ?cover ?elevation ?anti-predation 8 9 607.831 0.00 0.5206
2. Land use ?cover ?elevation 5 6 609.340 1.51 0.2448
3. Prey density 9anti-predation 4 5 611.365 3.53 0.0880
4. Land use 9elevation 3 4 613.136 5.31 0.0367
5. Elevation 9anti-predation 4 5 613.224 5.39 0.0351
6. Land use 2 3 613.613 5.78 0.0289
7. Prey density 9elevation 1 2 614.608 6.78 0.0176
Response variable: use (1) or non-use of lethal control (0)
8. Land use 9ZAR losses 3 4 112.521 0.00 0.7941
9. Language ?age ?land use 4 5 117.175 4.65 0.0775
10. Language ?land use 3 4 117.973 5.45 0.0520
11. Age ?land use 3 4 119.025 6.50 0.0307
Biodivers Conserv (2013) 22:1715–1730 1725
123
Author's personal copy
characteristics will be obvious focal points for future mitigation activities. The same can be
said of the large group of respondents (42 %) who considered losses in excess of 1 % of
stock holdings intolerable. That threshold seems unrealistically low, given that the per-
ceived annual predation rate for the Waterberg (1.4 % of total stock holdings) is at the
bottom of the 1.4–4.5 % range reported in other recent African studies (Holmern et al.
2007; Kolowski and Holekamp 2006; Marker et al. 2003; Schiess-Meier et al. 2007; Thorn
et al. 2012; Stein et al. 2010).
Most of those who expressed low tolerance were stud or rare game breeders and even
low predation levels in farms that stock such high-value animals can be extremely
expensive. Contextual information about the turnover and profit of affected farmers is
rarely available, making it impossible for others to assess whether losses with large
absolute values seriously affect net profit or are balanced by correspondingly high turnover
and assets. This may engender a heightened sense of financial vulnerability among high-
value stock farmers, particularly as their animals are usually intensively managed (personal
observation, M.T.), making losses relatively easy to detect (Table 1), thereby increasing
Table 4 Estimated regression coefficients ( ^
b), standard errors (±SE) and odds from generalised linear
models that were within 7 AIC
c
units of the model with the lowest AIC
c
Parameter ^
b±SE( ^
b) Exp(^
b)
Response variable: no. of game and livestock reportedly killed by carnivores at each farm Reference
categories: cover =open, land use =game, anti-predation =non lethal
Elevation (model average) 0.002 0.0007 1.002
Land use =livestock (model average) 0.290 0.3122 1.337
Land use =mixed (model average) 1.102 0.4475 3.010
Cover =heterogeneous (model average) 0.855 0.4322 2.352
Cover =dense (model average) 0.856 0.3377 2.355
Anti predation =neither (model 1) 0.931 0.4163 2.537
Anti-predation =lethal (model 1) 0.776 0.4655 2.173
Anti-predation =both (model 1) 0.848 0.3908 2.335
Prey density 9(anti-predation =non lethal) (model 3) -0.006 0.0047 0.994
Prey density 9(anti-predation =lethal) (model 3) 0.004 0.0044 1.004
Prey density 9(anti-predation =both) (model 3) 0.017 0.0049 1.017
Prey density 9(anti-predation =neither) (model 3) 0.016 0.0162 1.016
(Land use =livestock) 9elevation (model 4) 0.001 0.0009 1.001
(Land use =mixed) 9elevation (model 4) 0.002 0.001 1.002
(Land use =game) 9elevation (model 4) 0.001 0.0009 1.001
Elevation 9(anti-predation =non lethal) (model 5) 0.001 0.0007 1.001
Elevation 9(anti-predation =lethal) (model 5) 0.001 0.0008 1.001
Elevation 9(anti-predation =both) (model 5) 0.002 0.0008 1.002
Elevation 9(anti-predation =neither) (model 5) 0.001 0.0009 1.001
Prey density 9elevation (model 7) 9.13E-06 3.63E-06 1.000
Response variable: use (1) or non-use of lethal control (0)
(Land use =livestock) 9ZAR losses (model 8) 0.005 0.0038 1.005
(Land use =mixed) 9ZAR losses (model 8) 0.130 0.0538 1.139
(Land use =game) 9ZAR losses (model 8) -0.012 0.0770 0.988
Bold ^
bvalues differed significantly from zero
1726 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
awareness of predation risk. Those factors may explain why financial losses were an
important trigger for carnivore persecution in the group most likely to stock high-value
animals (mixed farmers), especially as they were also associated with the highest predicted
predation frequencies. If that interpretation is correct, recent increases in sale prices for
high-value game may have triggered an accompanying decrease in tolerance toward car-
nivores, which may be ameliorated or compounded by future movements in rare game
prices.
Despite a widespread perception that carnivores are economically damaging, financial
losses within the inter-quartile range were relatively low. For example, the reported median
rate of financial loss (ZAR 1.23 ha
-1
) represents approximately 1.5 % of net earnings in
agriculture (approximately ZAR 80 ha
-1
) (ABSA 2003) and would result in annual losses
of ZAR 1605 on an average-sized farm. Such losses seem insufficient to threaten com-
mercial farming livelihoods in general. Our extrapolation of the results to grazing areas
elsewhere in Limpopo gives only a coarse estimate of the wider impact of predation losses,
but provides an economic context for the results and identifies the order of magnitude of
the economic effect. The actual impact may differ from our estimate because elevation,
land use, vegetation cover (the three most influential variables) and consequent predation
rates in other areas of Limpopo may differ from those in the Waterberg. For this reason and
in view of the complex links between agriculture and other important economic factors
(e.g. employment), our interpretation of the wider economic impact of predation losses is
tentative. Nevertheless, at 0.15 % of Limpopo agricultural GDP, the wider economic effect
of predation losses appear negligible, particularly as agriculture, forestry and fishing
constituted only 2.3 and 2.5 % of South African and Limpopo GDP, respectively in 2010
(Statistics South Africa 2011).
The only directly comparable information about predation comes from the North West
Province, which borders Limpopo. Land use, farm management and animal husbandry
practices in the North West Province and the Waterberg are very similar (personal
observation, M.T.), yet the proportion of stock holdings predated in the North West
(2.8 %) was double that in the Waterberg. In the North West, which appears to be affected
by meso-predator release, 61 % of reported predation incidents were attributed to jackals
and caracals (Thorn et al. 2012). By comparison, jackals and caracals were blamed for just
28 % of reported predation losses in the Waterberg, where apex carnivores like leopards
and African wild dogs are still widespread (Friedmann and Daly 2004). This contrast may
exemplify the functional benefits of apex carnivores and possibly, the negative financial
consequences of removing them.
The median annual rate of loss in the North West was just (ZAR 0.22 ha
-1)
(Thorn et al.
2012), indicating that although Waterberg farmers lost a smaller proportion of their stock
holdings, prey animals were considerably more valuable. This probably reflects higher
occupancy of large carnivores (Friedmann and Daly 2004) that are able to kill large prey,
combined with greater prevalence of stud and high-value antelope breeding in the
Waterberg (personal observation). These factors may explain why financial losses were a
key determinant of conflict in the Waterberg, but not in the North West Province.
Counter intuitively, our results indicated that the species reported to be the main cause
of predation losses (leopard) is not among those that are least tolerated (African wild dogs
and cheetahs) or most persecuted (jackals). There are several possible reasons for this.
Jackals have long been considered vermin in South African farmland and receive no legal
protection. They are common and abundant despite intense long-term persecution (Sillero-
Zubiri et al. 2004), which seems to have spawned a widespread belief that lethal control of
jackals (and caracals) is necessary to prevent over-population and predation (personal
Biodivers Conserv (2013) 22:1715–1730 1727
123
Author's personal copy
observation). Leopards also have high occupancy in the Waterberg (Daly et al. 2005), but
although they pose a threat to stock holdings, hunting leopards under CITES quota permits
provides local people with a source of income, as does wildlife-based tourism. This may
explain why leopards were better tolerated than meso-carnivores that are less likely to
attract hunters or tourists or species like African wild dogs and cheetahs that cannot be
legally hunted.
However, there are alternative explanations arising from rapid growth in the area
committed to game farming over the last two decades (Bothma 2005). Historically, African
wild dogs and cheetahs were extirpated across much of their South African range (Mar-
newick et al. 2007; Sillero-Zubiri et al. 2004). Together with our occupancy results,
anecdotal information regarding the frequency of sightings suggests that both species are
re-colonising the Waterberg, probably because of the increased availability of habitat, prey
and refuge that are offered by game farms (Thorn et al. 2011a). This appears to be a
relatively recent phenomenon so low tolerance may stem from fear of unfamiliar species.
Furthermore, as they were absent from the area for a prolonged period, there is a wide-
spread belief that African wild dogs and cheetahs are not naturally-occurring in the
Waterberg (personal observation). Their reappearance has stimulated resentment of ‘for-
eign’ damage-causing animals and the organisations (conservation NGO’s and local
reserves) incorrectly blamed for introducing them into farmland (personal observation).
The estimated global population of African wild dogs (Endangered) is \5,500 with
approximately 450 mature individuals in South Africa (personal communication, Endan-
gered Wildlife Trust, South Africa), 42-106 of which are resident outside of protected areas
(Lindsey et al. 2004). The Waterberg is considered a prime target for range expansion
(Lindsey et al. 2004) and the population in northern Limpopo is thought to provide
important genetic connectivity to nearby populations in other parts of Southern Africa
(Davies-Mostert et al. 2012). Our persecution rates do not account for road kills, natural
mortality, poisoning/snare by-catch or illegal trade, which undoubtedly take a further toll
on the size and structure of the local population. It is therefore likely that persecution
constitutes a serious threat to the conservation status of African wild dogs in the
Waterberg.
The only available abundance estimates for brown hyaenas (Near Threatened) and
leopards (nationally Least Concern, globally Near Threatened) are tenuous, but suggest
that both species have much higher population numbers than African wild dogs (Mills and
Hofer 1998; Daly et al. 2005). However, as the majority of individuals occur outside
protected areas (Mills and Hofer 1998; Mackinnon and Mackinnon 1986), further research
is required to quantify local abundance and evaluate the importance of persecution-related
threats. This is true of all three species and is particularly important because changes in the
status of carnivore populations in the Waterberg may affect the stability of the wider
regional meta-population by virtue of spatial and genetic linkages with populations in
adjoining areas of southern Africa.
The results of this initial research have suggested several avenues for developing
conflict-mitigation strategies in the Waterberg. These include addressing predation prob-
lems in areas with a high proportion of vulnerable farms, improving tolerance of carnivores
among low-tolerance social groups and correcting misconceptions about the provenance
and financial impact of local carnivore populations. In each case, the literature suggests a
range of suitable initiatives. Examples include improving the physical security of enclo-
sures (Woodroffe et al. 2007), using livestock guarding dogs (Woodroffe et al. 2007; Rigg
et al. 2011) selective lethal control of habitual livestock predators (Woodroffe and Frank
2005), education and awareness raising activities (Marker et al. 2003) and constructive
1728 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
dialogue with affected communities (Anthony et al. 2010). Many such activities have
already been implemented in the study area to address the most pressing priorities revealed
by our exploratory analysis. The intention is to quantify and refine their effectiveness
before recommending them for wider use in other conflict situations with similar
characteristics.
Acknowledgments We are grateful to everyone who contributed information during interviews and to
Lapalala Wilderness and Welgevonden Private Game Reserve for logistical support. We thank anonymous
reviewers for helpful comments on the manuscript. We also thank Land Rover Centurion for sponsoring the
project and the Endangered Wildlife Trust, Knowsley Safari Park, and the Rufford Small Grants Foundation
for providing funding.
References
ABSA (2003) Game ranch profitability in Southern Africa, 3rd edn. ABSA Group: The financial sector
forum, Johannesburg
Anthony BP, Scott P, Antypas A (2010) Sitting on the fence? Policies and practices in managing human-
wildlife conflict in Limpopo Province, South Africa. Conserv Soc 8(3):225–240
Balme GA, Slotow R, Hunter LTB (2009) Impact of conservation interventions on the dynamics and
persistence of a persecuted leopard (Panthera pardus) population. Biol Conserv 142:2681–2690
Blaum N, Tietjen B, Rossmanith E (2009) Impact of livestock husbandry on small- and medium-sized
carnivores in Kalahari savannah rangelands. J Wildl Manage 73:60–67
Bothma JdP (2005) Extensive wildlife production in South Africa. Keynote address: Wildlife Seminar,
Northern Game Farmers’ Organisation, Pretoria, 11 March 2005
Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-
theoretic approach, 2nd edn. Springer-Verlag, New York
Daly B, Power J, Camancho G, Traylor-Holtzer K, Barber S, Caterrall S, Fletcher P, Martins Q, Martins N,
Owen C, Tahl T, Freidmann Y (eds) (2005) Leopard (Panthera pardus) population habitat viability
assessment workshop report. Conservation Breeding Specialist Group (SCC/IUCN)/CSBG Southern
Africa & The Endangered Wildlife Trust
Davies-Mostert HT, Kamler JF, Mills MGL, Jackson CR, Rasmussen GSA, Groom RJ, Macdonald DW
(2012) Long-distance transboundary dispersal of African wild dogs among protected areas in southern
Africa. Afr J Ecol 50(4):1365–2028
de Klerk A (2003) The Waterberg Biosphere Reserve: a land use model for ecotourism development. MSc,
University of Pretoria, Pretoria
Department of Agriculture Forestry and Fisheries (2010) Abstract of agricultural statistics. Department of
Agriculture, Forestry and Fisheries, South Africa, Pretoria
Dochterman NA, Jenkins SH (2011) Developing multiple hypotheses in behavioral ecology. Behav Ecol
Sociobiol 65:37–45
Friedmann Y, Daly D (eds) (2004) Red data book of the mammals of South Africa: a conservation
assessment. CBSG Southern Africa, Conservation Breeding Specialist Group (SSC/IUCN) & The
Endangered Wildlife Trust, Johannesburg
Graham K, Beckerman AP, Thirgood S (2005) Human–predator–prey conflicts: ecological correlates, prey
losses and patterns of management. Biol Conserv 122:159–171
Holmern T, Nyahongo J, Røskaft E (2007) Livestock loss caused by predators outside the Serengeti National
Park, Tanzania. Biol Conserv 135:518–526
Inskip C, Zimmermann A (2009) Human-felid conflict: a review of patterns and priorities worldwide. Oryx
43(01):18–34. doi:10.1017/S003060530899030X
Kolowski JM, Holekamp KE (2006) Spatial, temporal, and physical characteristics of livestock depredations
by large carnivores along a Kenyan reserve border. Biol Conserv 128(4):529–541. doi:
10.1016/j.biocon.2005.10.021
Limpopo DFED (2003) Limpopo state of the environment overview report. Limpopo Department of Finance
and Economic Development, Polokwane
Lindsey P, du Toit JT, Mills MGL (2004) The distribution and population status of African wild dogs
(Lycaon pictus) outside protected areas in South Africa. S Afr J Wildl Res 34(2):143–151
Lindsey PA, du Toit JT, Mills MGL (2005) Attitudes of ranchers towards African wild dogs Lycaon pictus:
conservation implications on private land. Biol Conserv 125(1):113–121
Biodivers Conserv (2013) 22:1715–1730 1729
123
Author's personal copy
Mackinnon J, Mackinnon K (1986) Review of the protected areas system of the Afrotropical realm. IUCN,
Gland
Marker LL, Mills MGL, Macdonald DW (2003) Factors influencing perceptions of conflict and tolerance
towards cheetahs on Namibian farmlands. Conserv Biol 17:1290–1298
Marnewick K, Beckhelling A, Cilliers D, Lane E, Mills G, Herring K, Caldwell P, Hall R, Meintjes S (2007)
The status of the Cheetah in South Africa. Cat News Special issue 3:22–31
Mills MGL, Hofer H (eds) (1998) Hyaenas. Status survey and conservation action plan. IUCN/SSC Hyaena
Specialist Group. IUCN, Gland
Naughton-Treves L, Grossberg R, Treves A (2003) Paying for tolerance: rural citizens’ attitudes toward
wolf depredation and compensation. Conserv Biol 17(6):1500–1511
NEMA (2004) National Environmental Management: National Biodiversity Act (10, 2004)
Prugh LR, Stoner CJ, Epps CW, Bean WT, Ripple WJ, Laliberte AS, Brashares JS (2009) The rise of the
mesopredator. Bioscience 59(9):779–791. doi:10.1525/bio.2009.59.9.9
Rasmussen GSA (1999) Livestock predation by the painted hunting dog Lycaon pictus in a cattle ranching
region of Zimbabwe: a case study. Biol Conserv 88:133–139
Rigg R, Findo S, Wechselberger M, Gorman ML, Sillero-Zubiri C, Macdonald D (2011) Mitigating car-
nivore–livestock conflict in Europe: lessons from Slovakia. Oryx 45(2):272–280
Schiess-Meier M, Ramsauer S, Gabanapelo T, Koenig B (2007) Livestock predation—insights from
problem animal control registers in Botswana. J Wildl Manage 71(4):1267–1274
Sillero-Zubiri C, Hoffmann M, Macdonald DW (eds) (2004) Canids: foxes, wolves, jackals and dogs. Status
survey and conservation action plan. IUCN/SCC Canid Specialist Group. IUCN, Gland
St John FAV, Keane AM, Edwards-Jones G, Jones L, Yarnell RW, Jones JPG (2011) Identifying indicators
of illegal behaviour: carnivore killing in human-managed landscapes. Proc R Soc B 279:804–812
Statistics South Africa (2011) Gross Domestic Product: annual estimates 2002–2010. Regional estimates
2002–2010. Third quarter 2011. Statistics South Africa, Pretoria
Stein AB, Fuller TK, Damery DT, Sievert LL, Marker LL (2010) Farm management and economic analyses
of leopard conservation in north-central Namibia. Anim Conserv 13:419–427
Thorn M, Green M, Bateman P, Waite S, Scott D (2011a) Brown hyaenas on roads: landscape-scale
estimation of carnivore occupancy and population size from spatially auto-correlated sign surveys
replicates. Biol Conserv 144:1799–1807
Thorn M, Green M, Keith M, Marnewick K, Bateman P, Cameron EZ, Scott D (2011b) Large-scale
distribution patterns of carnivores in northern South Africa: implications for conservation and moni-
toring. Oryx 45(4):579–586
Thorn M, Green M, Dalerum F, Bateman PW, Scott D (2012) What drives human-carnivore conflict in the
North West Province of South Africa? Biol Conserv 150:23–32
Tladi B, Baloyi T, Marfo C (2002) Chapter 6: settlement and land use patterns. In: Walmsley D, Walmsley
J, Mangold S, Kalule-Sabiti M (eds) North West Province State of the Environment Report Directorate
of Environment and Conservation Management. North West Department of Agriculture, Conservation
and Environment, Mmabatho
Treves A, Karanth KU (2003) Human-carnivore conflict and perspectives on carnivore management
worldwide. Conserv Biol 17(6):1491–1499
Vleissentraal (2011) http://www.vleissentraal.co.za/users/bosveld/web/veilingstatistiek.php. Accessed 30
March 2012
Woodroffe R, Frank LG (2005) Lethal control of African lions (Panthera leo): local and regional population
impacts. Anim Conserv 8:91–98
Woodroffe R, Ginsberg JR (1998) Edge effects and the extinction of populations inside protected areas.
Science 280(5372):2126–2128
Woodroffe R, Ginsberg JR (2000) Ranging behaviour and vulnerability to extinction in carnivores. In:
Gosling LM, Sutherland WJ (eds) Behaviour and conservation. Cambridge University Press, Cam-
bridge, pp 125–140
Woodroffe R, Frank LG, Lindsey PA, Ranah S, Romanach S (2007) Livestock husbandry as a tool for
carnivore conservation in Africa’s community rangelands: a case-control study. Biodivers Conserv
16(4):1245–1260
1730 Biodivers Conserv (2013) 22:1715–1730
123
Author's personal copy
