Content uploaded by Knowledge Vingi
Author content
All content in this area was uploaded by Knowledge Vingi on Mar 20, 2019
Content may be subject to copyright.
ORIGINAL ARTICLE
Ensemble modelling predicts Human Carnivore Conflict for a
community adjacent to a protected area in Zimbabwe
Kudzai Mpakairi
1
|
Henry Ndaimani
1
|
Knowledge Vingi
1
|
Tinaapi Hilary Madiri
2
|
Tendai Nekatambe
3
1
Department of Geography and
Environmental Science, University of
Zimbabwe, Harare, Zimbabwe
2
Zimbabwe Parks and Wildlife Management
Authority, Victoria Falls, Zimbabwe
3
Painted Dog Research Centre (PDRC),
Victoria Falls, Zimbabwe
Correspondence
Kudzai Mpakairi, Department of Geography
and Environmental Science, University of
Zimbabwe, Harare, Zimbabwe.
Email: kudzishaun@gmail.com
Abstract
Overlapping interests between humankind and nature have resulted in humans and
wildlife interacting. These interactions usually emanate from the absence of fences or
hard boundaries to restrict animal movement and when this happens, it results in
Human Carnivore Conflict (HCC). People residing in areas adjacent to protected areas
are the ones mostly at risk of HCC. In this study, we used ensemble modelling to
explain predation risk and indicate the key drivers of HCC in Matetsi Communal
Area, Zimbabwe. Ensemble modelling involves building a single consensus model
from several candidate models. We used seven environmental variables for the mod-
elling process, and these were distance from the park boundary, distance from rivers,
Normalised Difference Vegetation Index (NDVI), human population density and live-
stock density for cattle, goats and sheep. Livestock kill sites were used as the pres-
ence data. Our results illustrate that ensemble modelling explains predation risk with
a true skill statistic (TSS) of 0.9 for Matetsi Communal Area. This study provides the
potential application of ensemble modelling in HCC management through identifying
predation risk areas. In identifying predation risk areas, proactive and cost-efficient
management strategies for dealing with HCC in specific high-risk areas are plausible.
R
esum
e
Les int
er^
ets communs des hommes et de la nature ont abouti
a des interactions entre
les humains et la faune sauvage. Ces interactions r
esultent d’habitude de l’absence de
cl^
otures ou de limites mat
erialis
ees qui pourraient restreindre les d
eplacements des
animaux et, lorsqu’elles surviennent, elles donnent lieu
a des conflits hommes-carni-
vores (HCC). Les gens qui r
esident dans des zones voisines d’aires prot
eg
ees sont
ceux qui sont le plus menac
es d’HCC. Dans cette
etude, nous avons eu recours
a une
mod
elisation par ensemble pour expliquer le risque de pr
edation et indiquer les vec-
teurs majeurs de HCC dans l’Aire Communale de Matetsi, au Zimbabwe. La mod
elisa-
tion par ensemble implique de construire un mod
ele unique qui fait consensus
a
partir de plusieurs mod
eles candidats. Nous avons utilis
e plusieurs variables environ-
nementales pour le processus de mod
elisation,
a savoir la distance par rapport
ala
limite du parc, celle par rapport
a la rivi
ere, l’indice de v
eg
etation par diff
erence nor-
malis
ee (NDVI), la densit
e de population humaine, et la densit
edub
etail pour les
bovins, les ch
evres et les moutons. Les lieux de mise
a mort du b
etail ont servi
comme donn
ees de pr
esence. Nos r
esultats montrent que la mod
elisation par ensem-
ble explique le risque de pr
edation avec une true skill statistic de 0,9 pour l’Aire
Accepted: 27 April 2018
DOI: 10.1111/aje.12526
Afr J Ecol. 2018;1–7. wileyonlinelibrary.com/journal/aje ©2018 John Wiley & Sons Ltd
|
1
Communale de Matetsi. Cette
etude fournit une application possible de mod
elisation
par ensemble dans la gestion des HCC en identifiant les zones
a risque de pr
edation.
En identifiant des strat
egies de gestion proactives et efficaces pour ces zones
a ris-
que, il est plausible de traiter les HCC dans des zones
a risque bien d
efinies.
KEYWORDS
ensemble modelling, Human Carnivore Conflict, Matetsi communal area, Normalised Difference
Vegetation Index, predation risk
1
|
INTRODUCTION
Overlapping interests between mankind and nature have resulted in
humans and wildlife interacting (Gusset, Swarner, Mponwane, Kele-
tile, & McNutt, 2009; Zisadza Gandiwa, Gandiwa, & Muboko, 2016).
These interactions pose no threat when both parties benefit, but
when they are counterproductive on both ends, Human Wildlife
Conflict (HWC) occurs (Madden, 2004). HWC occurs in different
ways, from loss of human life, crop raiding and livestock predation
(Butler, 2000; Ficetola, Bonardi, Mairota, Leronni, & Padoa-Schioppa,
2014), to the transmission of pests and zoonotic infections (Le Bel
et al., 2011). These interactions are usually exacerbated by the
absence of fences or hard boundaries to restrict animal movement.
For instance, most carnivore species have extensive home ranges
that usually extend into adjacent communities and when carnivores
and livestock interact, it normally results in livestock predation. Live-
stock predation from wild carnivores is commonly termed Human
Carnivore Conflict (HCC) (Constant, Bell, & Hill, 2015). HCC is a sub-
set of HWC. People residing in communities adjacent to protected
areas are the ones mostly at risk of HCC. In the light of HCC, locals
tend to acclimatize a retaliatory practice of killing problem animals in
the name of problem animal control (Gandiwa, Heitk€
onig, Lokhorst,
Prins, & Leeuwis, 2013; Pachavo & Murwira, 2013).
Livestock loss from HCC is a setback for households that depend
on livestock rearing as a form of dietary and economic resource.
Unquestionably, in such a situation, it becomes impossible to achieve
conservation objectives when the local communities’perception of
wildlife is negative, yet we need them to embrace conservation. This
is because following HCC, the abundance of most carnivore species
has declined from retaliatory killings (Bauer & Van Der Merwe, 2004;
Woodroffe, Thirgood, & Rabinowitz, 2005). Such misunderstandings
between locals and resources managers are the reason why conser-
vation efforts have failed in most communal areas (Miller, 2015).
Hence, it is plausible to minimize HCC through the identification of
areas where predation is likely to occur and factors that drive HCC.
Species Distribution Models (SDM) (commonly known as Niche-
based Models) have been used to predict species range from a set
of environmental variables. This is based on the understanding that
species will occupy space where conditions are similar to where they
are presently found (Mouton, De Baets, & Goethals, 2010). The pre-
dictive power of the SDM relies on the model used and the underly-
ing assumptions. To avoid model assumptions and shortfalls
affecting the output, ensemble models have been used (e.g., in Mpa-
kairi et al. (2017)). Ensemble modelling builds a highly predictive
model from a set of other candidate models, by selecting parts of
those candidate models that are predicting well (Thuiller, Lafourcade,
Engler, & Ara
ujo, 2009). Using an ensemble of SDMs, we can obtain
a model that performs better than when candidate models are used
in isolation. Fortunately, the application of ensemble modelling in
HCC studies has allowed for better identification of predation risk
areas from a set of environmental variables (Ficetola et al., 2014;
Pandey, Shaner, & Sharma, 2015). However, the influence of each
environmental variable used affects the model performance owing
to spatial heterogeneity and species behaviour. For instance, live-
stock presence within the buffer core of Kanha Tiger Reserve, India
(Miller, Jhala, & Jena, 2016) influenced predation, whereas in the
Upper Spiti Landscape of Trans-Himalayan, India the abundance of
snow leopards and wild prey drove predation regardless of distance
from the boundary (Suryawanshi, Bhatnagar, Redpath, & Mishra,
2013).
In this study, we hypothesize that ensemble modelling can ade-
quately predict predation risk and can indicate the key drivers of
HCC in Matetsi Communal Area, Zimbabwe. We do this by relating
livestock kill site data with a set of environmental variables.
2
|
MATERIALS AND METHODS
2.1
|
Study area
The study was conducted in the Matetsi communal area (hereafter
Matetsi CA) which is adjacent to Matetsi safari area, in northwestern
Zimbabwe. We focused our study on part of Matetsi CA with human
settlements. The study area is bound between latitudes 18.38°S–
18.01°S and longitudes 26.29°E–25.79°E (Figure 1) and measures
1,638 km
2
. Vegetation within the study area is predominantly Termi-
nalia sericea,Colophospermum mopane and Setaria grasslands (Gara,
Murwira, Ndaimani, Chivhenge, & Hatendi, 2015), with altitude rang-
ing from 700 m to 900 m above sea level. Matetsi River is the major
drainage feature; it is not perennial but along the river there exist
river pools that hold water during the dry season. The area receives
mean annual precipitation of 600 mm (December-March) and experi-
ences a temperature range of 19°C (July) - 33°C (October) (Le Bel
et al., 2011). Major livestock predators within our study area are
spotted hyaena (Crocuta crocuta), lion (Panthera leo), leopard
2
|
MPAKAIRI ET AL.
(Panthera pardus), jackal (Canis mesomelas), cheetah (Acinonyx juba-
tus), civet cat (Civettictis civetta), genet cat (Genetta genetta) and wild
dog (Lycaon pictus). The Matetsi community survives on subsistence
farming, proceeds from trophy hunting and livestock rearing.
2.2
|
Presence data
We collected GPS data of all kill sites recorded (n=209) within the
study area during the dry season of 2013, 2014 and 2015 (July-Octo-
ber). These locations represented the presence data. Pseudo-absence
data (n=129) were randomly generated following suggestions by Wisz
and Guisan (2009). We could not distinguish the animal responsible for
the kill; thus, we nested all the kill sites for model building.
2.3
|
Environmental variables
We used seven environmental variables for the modelling process,
which include distance from the park boundary, distance from the
river, Normalised Difference Vegetation Index (NDVI), human popu-
lation density and livestock density data for cattle, goats and sheep.
The park boundary and rivers were digitized on the Google earth
platform (www.googleearth.com). Distance from the park and river
was computed from the Euclidean distance function in ArcMap
(ESRI, 2006). NDVI was calculated from September 2015 Landsat 8
imagery that was downloaded freely from www.earthexplorer.usgs.
gov/ (accessed 02 January 2017). NDVI was included as a proxy for
vegetation cover (Xie, Sha, & Yu, 2008). Human population density
was used as an indicator of human settlement. Areas with minimal
human presence possibly have higher predator density as distur-
bance is negligible (Thorington & Bowman, 2003). Human population
density data were obtained from the Zimbabwe Census data set
(http://www.zimstat.co.zw/cartography). Livestock density data for
cattle, goats and sheep were freely downloaded from http://livestoc
k.geo-wiki.org/download/ (accessed 01 January 2017).
To safeguard against overfitting and multicollinearity, we calcu-
lated the Variance Inflation Factor (VIF), of all the environmental
variables. All the variables were retained and later used in modelling
as they had a VIF less than the acceptable threshold of 10 (Farrar &
Glauber, 1967). Before modelling, the environmental data sets were
resampled to the same pixel size (30 930 m) for consistence.
2.4
|
Modelling
Each modelling techniques has its shortcomings, and the underlying
assumptions of each model algorithm determine the prediction out-
put (Thuiller, 2014); conversely, researchers have found a way of
building consensus models (ensemble models) from several other
models. This technique is a lot reliable as it retains areas with coher-
ent predictions from all the models used.
To model the probable HCC risk areas, two modelling algorithms
were run in R using Biomod 2 package (Thuiller et al., 2016). The
first was Gradient Boosting Model (GBM) (Ridgeway, 1999), which
yields highly predictive models (nonlinear) from combining regression
trees (Kutywayo, Chemura, Kusena, Chidoko, & Mahoya, 2013) and
Random Forest. Seventy per cent of the presence data were ran-
domly selected for model calibration and the remaining thirty per
cent were set aside for model validation. An ensemble model was
built from selecting parts of the two candidate models that had a
true skill statistic (TSS) of ≥0.6.
2.5
|
Model evaluation
The True Skill Statistic (TSS) was used in building the ensemble
model as well as in evaluating model performance. TSS is defined as
sensitivity +specificity 1 (Allouche, Tsoar, & Kadmon, 2006).
Sensitivity is the proportion of true positives and specificity is
the proportion of false positives. TSS ranges from 1to+1, with
high values (+1) indicating a perfect model and low values (<0)
FIGURE 1 Location of the Matetsi
Communal Area adjacent to Matetsi Safari
Area in northwestern Zimbabwe [Colour
figure can be viewed at
wileyonlinelibrary.com]
MPAKAIRI ET AL.
|
3
indicating a random performance (Mouton et al., 2010). The ensem-
ble model was built using a TSS values ≥0.6 following recommenda-
tions that TSS values ≥0.6 are fair for predictive objectives (Allouche
et al., 2006). Variable importance was calculated through a random-
ization method present within the Biomod 2 package (Thuiller et al.,
2009).
3
|
RESULTS
The ensemble model based on seven environmental variables (NDVI,
livestock density of cattle, goats and sheep, human population den-
sity, distance from the park boundary and river) explained well the
predation risk in our study area (TSS =0.90).
Variable influence analysis showed that NDVI contributed most
to the model performance (28%), followed by the livestock density
of goats and the least important was human population density (1%)
(Table 1). Based on the logistic threshold of equal training sensitivity
and specificity, results illustrate that predation risk was high in areas
with low NDVI (0.005-0.16), close to the park boundary (1,511-
4,373 m) and river (536-3,215 m) with a relatively low human popu-
lation (<1 human/km
2
). The livestock densities at risk of predation
were different for each species (>3.25 cattle/km
2
,<5.96 goats/km
2
and 0.08-0.11 sheep/km
2
) (Figure 2). Predation risk was high in
areas with human settlement and along Matetsi River (Figure 3).
4
|
DISCUSSION
We found out that areas with low NDVI values are potentially at risk of
livestock predation than areas with high NDVI values. Low NDVI values
usually translate to shrub lands and grasslands, commonly preferred by
livestock (Chidumayo, 2001). These areas possibly attract livestock
TABLE 1 Environmental variable contribution to predation risk
Variable Contribution (%)
Normalized Difference Vegetation Index (NDVI) 0.31
Goats density 0.24
Distance from the river 0.19
Sheep density 0.12
Distance from the park boundary 0.09
Cattle density 0.02
Human population density 0.01
FIGURE 2 Predation risk as a function
of (a) Normalised Difference Vegetation
Index (NDVI), (b) Distance from the park
boundary, (c) Livestock density of sheep,
(d) Human population density, (e) Livestock
density of goats, (f) Distance from the river
and (g) Livestock density of cattle. Dotted
lines represent the logistic threshold of
equal training specificity and sensitivity
4
|
MPAKAIRI ET AL.
because of the forage quality, which contains postingestion nutritional
benefits (Hendricks, Novellie, Bond, & Midgley, 2002). Herders also pre-
fer these areas for livestock to graze because they are open and they
can easily monitor their livestock from a distance. Additionally, these
areas are also preferential because they do not restrict livestock move-
ment during grazing, unlike high biomass areas that are highly nutritious
but not manoeuvrable (Gara et al., 2017).
Noteworthy, we also learned that of the livestock species used in
the study, goats were more likely at risk of predation than cattle and
sheep, and predation risk was elevated when goats were in small herds
(<6). Goats were at risk of predation because they can easily be
dragged away from kill sites before the landowners can retaliate as
they do not weigh (20-25 kg) as much as cattle (145-450 kg) or sheep
(30-35 kg) (Kabir, Ghoddousi, Awan, & Awan, 2013). These results are
in tandem with the literature that has reported the loss of more goats
compared to sheep or cattle in communal areas (Kissui, 2008).
Livestock predation still exists in most communal areas adjacent
to wildlife reserves (Madden, 2004) primarily because of the conflict-
ing usage of limited resource (e.g., rangelands) between wildlife and
humans (Gandiwa, Gandiwa, & Muboko, 2012; Graham, Beckerman,
& Thirgood, 2005). Most communities have acclimatized a retaliatory
approach of killing problem species, because they feel their losses
are aggravating (Childes, 1988; Suryawanshi et al., 2013). This is
despites efforts to minimize or compensate for HCC losses by sev-
eral stakeholders (Romanch, Lindsey, & Woodroffe, 2007).
For instance, following annual income losses of ~12% by each
household from HWC (Butler, 2000) in Gokwe district, Zimbabwe,
the population of predator species (e.g., wild dogs (Woodroffe & Sil-
lero-Zubiri, 2012)) has continued to decimate from landowners retali-
ating and persecution. Livestock lost to predation can be
compensated, what of wildlife loss from the retaliatory killing? Can it
be effectively restored? If not, what hope do we have in resolving
HCC? This study, like others, builds upon several efforts in existence
(e.g., the construction of bomas and night guards (Loveridge et al.,
2017)) to suppress livestock predation, through identifying risk areas,
and factors that likely drive predation. Results from the study can be
used to inform livestock owners in selecting grazing patches for live-
stock. By identifying predation risk areas, herders can utilize land-
scapes that are less at risk of predation than those that are at risk.
When areas with high predation risk areas are used, herders should
stay alert and herd their livestock in large numbers especially goats.
The strength of this research is grounded upon the usage of remote
sensing and GIS data in creating a highly predictive ensemble model for
predation risk from several species distribution models. Predation risk
models are dependent on the predictive power of the model used. By
creating an ensemble model, we try and ensure that only areas with
high predation risk are included in the final model (Carvalho, Zarco-
Gonz
alez, Monroy-Vilchis, & Morato, 2015). This study is also assenting
because unlike other studies that pooled livestock data and used it as a
single predictor variable in predation risk modelling, we managed to
include each livestock species as an independent predictor variable,
allowing an enlightened view on livestock preference by predators.
However, our results are not conclusive as we used a nested car-
nivore data set and future predation risk models should focus on
predation risk of individual carnivores rather than using nested data.
Additionally, future studies modelling predation risk from species dis-
tribution models should include preventative measures by landown-
ers to enhance model predictive ability.
5
|
CONCLUSION
An understanding of the drivers of livestock predation is important
for communities that live adjacent to wildlife reserves if they
endeavour to suppress the socioeconomic mishaps attached to HCC.
Efforts to reduce HCC are being performed at a large scale, which is
a costly exercise, yet conflicts are restricted to certain areas. This
study provides a better understanding of HCC using predictions
from ensemble modelling. Following the identification of areas with
high predation risk, proactive and cost-efficient management strate-
gies for dealing with HCC in specific high-risk areas are plausible.
ACKNOWLEDGEMENT
We would like to thank the Matetsi community for their cooperation
during data collection. We are also indebted to Edward Katiza,
Fanuel Nleya, Mathiya Zuzane and Shelter Hozhokozho who assisted
in data collection.
AUTHOR CONTRIBUTION
KM and THM conceptualized the study. KM and HN helped with
data processing and modelling. KM, KV and TN helped in the writing
of the manuscript.
FIGURE 3 Areas prone to livestock predation in Matetsi
communal land from an ensemble model. The predation risk was
threshold by equal training sensitivity and specificity to produce a
binary output [Colour figure can be viewed at
wileyonlinelibrary.com]
MPAKAIRI ET AL.
|
5
ORCID
Kudzai Mpakairi http://orcid.org/0000-0002-1929-1464
Henry Ndaimani http://orcid.org/0000-0002-8237-8140
REFERENCES
Allouche, O., Tsoar, A., & Kadmon, R. (2006). Assessing the accuracy of
species distribution models: Prevalence, kappa and the true skill
statistic (TSS). Journal of Applied Ecology,43(6), 1223–1232. https://d
oi.org/10.1111/j.1365-2664.2006.01214.x
Bauer, H., & Van Der Merwe, S. (2004). Inventory of free-ranging lions
Panthera leo in Africa. Oryx,38(1), 26–31.
Butler, J. R. A. (2000). The economic costs of wildlife predation on live-
stock in Gokwe communal land, Zimbabwe. African Journal of Ecology,
38(1), 23–30. https://doi.org/10.1046/j.1365-2028.2000.00209.x
Carvalho, E. A., Zarco-Gonz
alez, M. M., Monroy-Vilchis, O., & Morato, R.
G. (2015). Modeling the risk of livestock depredation by jaguar along
the Transamazon highway, Brazil. Basic and Applied Ecology,16(5),
413–419. https://doi.org/10.1016/j.baae.2015.03.005
Chidumayo, E. (2001). Climate and phenology of savanna vegetation in
southern Africa. Journal of Vegetation Science,12, 347–354. https://d
oi.org/10.2307/3236848
Childes, S. (1988). The past history, present status and distribution of the
hunting dog Lycaon pictus in Zimbabwe. Biological Conservation,44(4),
301–316. https://doi.org/10.1016/0006-3207(88)90022-5
Constant, N. L., Bell, S., & Hill, R. A. (2015). The impacts, characterisation
and management of human–leopard conflict in a multi-use land sys-
tem in South Africa. Biological Conservation,24, 2967–2989. https://d
oi.org/10.1007/s10531-015-0989-2
ESRI (2006). ArcMap 9.3. ESRIâArcGIS, 9, 1999-2008.
Farrar, D. E., & Glauber, R. R. (1967). Multicollinearity in regression analy-
sis: The problem revisited. The Review of Economic and Statistics,49,
92–107. https://doi.org/10.2307/1937887
Ficetola, G. F., Bonardi, A., Mairota, P., Leronni, V., & Padoa-Schioppa, E.
(2014). Predicting wild boar damages to croplands in a mosaic of
agricultural and natural areas. Current Zoology,60(2), 170–179.
https://doi.org/10.1093/czoolo/60.2.170
Gandiwa, E., Gandiwa, P., & Muboko, N. (2012). Living with wildlife and
associated conflicts in a contested area within the Northern Gonarez-
hou National Park, Zimbabwe. Journal of Sustainable Development in
Africa,14(6), 252–260.
Gandiwa, E., Heitk€
onig, I. M. A., Lokhorst, A. M., Prins, H. H. T., & Leeu-
wis, C. (2013). CAMPFIRE and human-wildlife conflicts in local com-
munities bordering Northern Gonarezhou National Park, Zimbabwe.
Ecology and Society,18(4), 7. https://doi.org/10.5751/es-05817-
180407
Gara, T. W., Murwira, A., Ndaimani, H., Chivhenge, E., & Hatendi, C. M.
(2015). Indigenous forest wood volume estimation in a dry savanna,
Zimbabwe: Exploring the performance of high-and-medium spatial
resolution multispectral sensors. Transactions of the Royal Society of
South Africa,1–9. https://doi.org/10.1080/0035919X.2015.1057269
Gara, T. W., Wang, T., Skidmore, A. K., Ngene, S. M., Dube, T., & Sibanda,
M. (2017). Elephants move faster in small fragments of low produc-
tivity in Amboseli ecosystems: Kenya. Geocarto International,32(11),
1243–1253. https://doi.org/10.1080/10106049.2016.1206625
Graham, K., Beckerman, A., & Thirgood, S. (2005). Human–predator–prey
conflicts: Ecological correlates, prey loss and patterns of manage-
ment. Biodiversity and Conservation,122, 159–171.
Gusset, M., Swarner, M., Mponwane, L., Keletile, K., & McNutt, J. (2009).
Human–wildlife conflict in northern Botswana: Livestock predation
by Endangered African wild dog Lycaon pictus and other carnivores.
Oryx,43(01), 67–72. https://doi.org/10.1017/S0030605308990475
Hendricks, H., Novellie, P., Bond, W., & Midgley, J. (2002). Diet selection
of goats in the communally grazed Richtersveld National Park. African
Journal of Range and Forage Science,19(1), 1–11. https://doi.org/10.
2989/10220110209485769
Kabir, M., Ghoddousi, A., Awan, S. M., & Awan, N. M. (2013). Assessment
of human-leopard conflict in Machiara National Park, Azad Jammu
and Kashmir, Pakistan. European Journal of Wildlife Research,60, 291–
296.
Kissui, B. (2008). Livestock predation by lions, leopards, spotted hyenas,
and their vulnerability to retaliatory killing in the Maasai steppe, Tan-
zania. Animal Conservation,11(5), 422–432. https://doi.org/10.1111/j.
1469-1795.2008.00199.x
Kutywayo, D., Chemura, A., Kusena, W., Chidoko, P., & Mahoya, C.
(2013). The impact of climate change on the potential distribution of
agricultural pests: The Case of the Coffee White Stem Borer (Mono-
chamus leuconotus P.) in Zimbabwe. PLoS ONE,8(8), e73432.
https://doi.org/10.1371/journal.pone.0073432
Le Bel, S., Murwira, A., Mukamuri, B., La Grange, M., Czudek, R., & Tay-
lor, R. (2011). Human wildlife conflicts in southern Africa: riding the
whirl wind in Mozambique and in Zimbabwe. INTECH Open Access
Publisher.
Loveridge, A. J., Kuiper, T., Parry, R. H., Sibanda, L., Hunt, J. H., Stapelk-
amp, B., & Macdonald, D. W. (2017). Bells, bomas and beefsteak:
Complex patterns of human-predator conflict at the wildlife-agropas-
toral interface in Zimbabwe. PeerJ,5, e2898. https://doi.org/10.
7717/peerj.2898
Madden, F. (2004). Creating coexistence between humans and wildlife:
Global perspectives on local efforts to address human–wildlife con-
flict. Human Dimensions of Wildlife,9(4), 247–257. https://doi.org/10.
1080/10871200490505675
Miller, J. R. (2015). Mapping attack hotspots to mitigate human–carnivore
conflict: Approaches and applications of spatial predation risk model-
ing. Biodiversity and Conservation,24(12), 2887–2911. https://doi.org/
10.1007/s10531-015-0993-6
Miller, J. R. B., Jhala, Y. V., & Jena, J. (2016). Spatial and temporal pat-
terns of livestock losses and hotspots of attack from tigers and leop-
ards in Kanha Tiger Reserve, central India. Regional Environmental
Change,16,17–29. https://doi.org/10.1007/s10113-015-0871-5
Mouton, A. M., De Baets, B., & Goethals, P. L. (2010). Ecological rele-
vance of performance criteria for species distribution models. Ecologi-
cal Modelling,221(16), 1995–2002. https://doi.org/10.1016/j.ec
olmodel.2010.04.017
Mpakairi, K. S., Ndaimani, H., Tagwireyi, P., Gara, T. W., Zvidzai, M., &
Madhlamoto, D. (2017). Missing in action: Species competition is a
neglected predictor variable in species distribution modelling. PLoS
ONE,12(7), e0181088. https://doi.org/10.1371/journal.pone.
0181088
Pachavo, G., & Murwira, A. (2013). Land-use and land tenure explain spa-
tial and temporal patterns in terrestrial net primary productivity
(NPP) in Southern Africa. Geocarto International,29(6), 671–687.
https://doi.org/10.1080/10106049.2013.837101
Pandey, P., Shaner, P.-J. L., & Sharma, H. P. (2015). The wild boar as a
driver of human-wildlife conflict in the protected park lands of Nepal.
European Journal of Wildlife Research,62, 103–108. https://doi.org/
10.1007/s10344-015-0978-5
Ridgeway, G. (1999). The state of boosting. Computing Science and Statis-
tics,31, 172–181.
Romanch, S. S., Lindsey, P. A., & Woodroffe, R. (2007). Determinants of
attitudes towards predators in central Kenya and suggestions for
increasing tolerance in livestock dominated landscapes. Oryx,41,
185–195. https://doi.org/10.1017/S0030605307001779
Suryawanshi, K. R., Bhatnagar, Y. V., Redpath, S., & Mishra, C. (2013).
People, predators and perceptions: Patterns of livestock depredation
by snow leopards and wolves. Journal of Applied Ecology,50, 550–
560. https://doi.org/10.1111/1365-2664.12061
6
|
MPAKAIRI ET AL.
Thorington, K. K., & Bowman, R. (2003). Predation rate on artificial nests
increases with human housing density in suburban habitats. Ecogra-
phy,26(2), 188–196. https://doi.org/10.1034/j.1600-0587.2003.
03351.x
Thuiller, W. (2014). Editorial commentary on ‘BIOMOD–optimizing pre-
dictions of species distributions and projecting potential future shifts
under global change. Global Change Biology,20, 3591–3592. https://d
oi.org/10.1111/gcb.12728
Thuiller, W., Georges, D., Engler, R., Breiner, F., Georges, M. D., & Thuil-
ler, C. W. (2016). Package ‘biomod2’.
Thuiller, W., Lafourcade, B., Engler, R., & Ara
ujo, M. B. (2009). BIOMOD–
a platform for ensemble forecasting of species distributions. Ecogra-
phy,32(3), 369–373. https://doi.org/10.1111/j.1600-0587.2008.
05742.x
Wisz, M. S., & Guisan, A. (2009). Do pseudo-absence selection strategies
influence species distribution models and their predictions? An infor-
mation-theoretic approach based on simulated data. BMC Ecology,9
(1), 1.
Woodroffe, R., & Sillero-Zubiri, C. (2012). Lycaon pictus. IUCN Red List of
Threatened Species v, 2.
Woodroffe, R., Thirgood, S., & Rabinowitz, A. (2005). People and wildlife,
conflict or co-existence? Cambridge, UK: Cambridge University Press.
https://doi.org/10.1017/CBO9780511614774
Xie, Y., Sha, Z., & Yu, M. (2008). Remote sensing imagery in vegetation
mapping: A review. Journal of Plant Ecology,1(1), 9–23. https://doi.
org/10.1093/jpe/rtm005
Zisadza Gandiwa, P., Gandiwa, E., & Muboko, N. (2016). Preliminary
assessment of human–wildlife conflicts in Maramani Communal Area,
Zimbabwe. African Journal of Ecology,54, 500–503. https://doi.org/
10.1111/aje.12282 https://doi.org/10.1111/aje.12282
How to cite this article: Mpakairi K, Ndaimani H, Vingi K,
Madiri TH, Nekatambe T. Ensemble modelling predicts
Human Carnivore Conflict for a community adjacent to a
protected area in Zimbabwe. Afr J Ecol. 2018;00:1–7.
https://doi.org/10.1111/aje.12526
MPAKAIRI ET AL.
|
7
