Movement patterns of cheetahs (Acinonyx jubatus) in farmlands in Botswana

29 Pages·2016·0.95 MB·English

©2017.PublishedbyTheCompanyofBiologistsLtd|BiologyOpen(2017)6,118-124doi:10.1242/bio.021055
RESEARCHARTICLE
Movement patterns of cheetahs (Acinonyx jubatus) in farmlands in
Botswana
L.K.VanderWeyde1,2,*,T.Y.Hubel3,J.Horgan1,J.Shotton3,R.McKenna1andA.M.Wilson3
ABSTRACT potentialadaptionstotheseenvironmentsmayrevealhowcheetah
behaviouronfarmlandsmaydifferfromthatobservedinthemore
Botswanahasthesecondhighestpopulationofcheetah(Acinonyx
commonlystudied protected areas (Durant et al., 2004; Kellyand
jubatus)withmostlivingoutsideprotectedareas.Asaresult,many
Durant,2007).Insightoncheetahspatialecologyinfarmlandswill
cheetahs are found in farming areas which occasionally results in
enable improved methods in human-wildlife conflict mitigation, a
human-wildlife conflict. This study aimed to look at movement
vitalcomponentofcheetahsurvivalintheseareas.
patternsofcheetahsinfarmingenvironmentstodeterminewhether
Botswana has the second largest population of cheetah and
cheetahshaveadaptedtheirmovementsinthesehuman-dominated
retains large connecting habitats across the country (Klein, 2007;
landscapes.Wefittedhigh-timeresolutionGPScollarstocheetahsin
Purchase et al., 2007) with 38% of the land set aside for wildlife
the Ghanzi farmlands of Botswana. GPS locations were used to
utilisation (Winterbach et al., 2015). Cheetahs, like other large
calculatehomerangesizesaswellasnumberanddurationofvisitsto
carnivores, have large home ranges, but despite several protected
landscape features using a time-based local convex hull method.
areas being large enough to accommodate these ranges, cheetahs
Cheetahshadmedium-sized homerangescomparedtopreviously
exist at low densities in these areas (Klein, 2007) and are more
studiedcheetahinsimilarfarmingenvironments.Resultsshowedthat
commonlyfoundinnon-protectedareas.Studiesofcheetahhome
cheetahs actively visited scent marking trees and avoided visiting
rangeshaveshownlargevariabilityinsize,primarilyasafunction
homesteads.Aslightpreferenceforvisitinggamefarmsovercattle
ofpreyavailability,habitat,sexandinterspecificcompetition(Caro,
farms was found, but there was no difference in duration of visits
1994;Durant,2000).Earlycheetahstudiesinlargeopenprotected
betweenfarmtypes.Weconcludethatcheetahsselectedforareas
habitats,suchastheSerengetiNationalPark,revealedaveragehome
thatareimportantfortheirdietaryandsocialneedsandprefertoavoid
range sizes around 700km2 for non-resident males and around
human-occupied areas. Improved knowledge of how cheetahs use
800km2forfemales(Caro,1994).Conversely,insimilarprotected
farmlands can allow farmers to make informed decisions when
habitatsintwofencedSouthAfricannationalparks,homerangesof
developingmanagementpracticesandcanbeanimportanttoolfor
lessthan 200km2for bothsexeswererecorded (Broomhall et al.,
reducinghuman-wildlifeconflict.
2003;Welchetal.,2015).Inthesecases,cheetahresidencystatusor
KEYWORDS:T-LoCoH,Homerange,Resourceselection, preyavailabilityandtheirmigratorypatternsmayberesponsiblefor
Non-protected,Human-wildlifeconflict thedifferinghomerangesizes.Thenon-protectedareasfavouredby
cheetahsinsomeregionsmaybeduetotheloworlackofpresence
INTRODUCTION oflargercompetitorssuchaslion(Pantheraleo)andspottedhyaena
Cheetah populations continue to decline worldwide, despite (Crocuta crocuta). The presence of these competitors has been
conservation efforts and increasing knowledge of cheetah showntoaffecttemporalandspatialmovementofcheetahs(Bissett
ecology. Key drivers of decline include habitat loss, disease and et al., 2015; Cristescu et al., 2013; Durant, 2000), with cheetahs
persecution, which is widespread in much of southern Africa activelyavoidingareasandmovingawayfromlargercompetitors.
(Purchaseetal.,2007).Infarmingareas,wherecheetahshavebeen Whilstcheetahsoccupyingareaswithlowinterspecificcompetition
shown to occur in relatively high densities (Klein, 2007; Marker mightbeexpectedtohavesmallerhomeranges(astheneedtoavoid
et al., 2003), they must co-exist with humansto survive, but how more competitive species is reduced), this is not always the case
theymayadapttheirmovementsandbehavioursintheselandscapes (Purchase and du Toit, 2000). Similarly, in Namibian farmlands
is not well understood. Knowledge of both broad-scale and with low densities of large carnivores (Marker et al., 2008a), the
fine-scale movement patterns that show selection or avoidance of home range sizes of cheetahs are exceptionally large, averaging
both natural and man-made features is important for cheetah 1600km2. Furthermore, the authors found that these large home
conservation in farming areas (Kent, 2011). Understanding rangeswerenotrelatedtosex,socialstatusorseason.Infarmlands
ofBotswana,malecheetahhomerangesaveraged668km2(Houser
et al., 2009a), smaller than those found in Namibia, but still
1CheetahConservationBotswana,B5,KgaleSidingOfficePark,Plot1069-KO,
Gaborone,Botswana.2SanDiegoZoo,InstituteforConservationResearch,15600 considerably larger than the male home ranges of 190-310km2
SanPasqualValleyRoad,Escondido,CA92027-7000,USA.3StructureandMotion observedinSouthAfricanranches(MarnewickandCilliers,2006).
Laboratory,RoyalVeterinaryCollege,UniversityofLondon,HatfieldAL97TA,UK. These findings suggest that whilst the low density of large
*Authorforcorrespondence([email protected]) competitive carnivores in farmlands may have some effect on en
ranging patterns of cheetahs, prey availability may be more p
L.K.V.,0000-0003-0981-8856 influential(Houseretal.,2009a). O
ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommonsAttribution Farming environments, such as those in the Ghanzi district of gy
License(http://creativecommons.org/licenses/by/3.0),whichpermitsunrestricteduse, Botswana, generally contain resources such as permanent water o
distributionandreproductioninanymediumprovidedthattheoriginalworkisproperlyattributed. l
sourcesnecessaryforlivestock,andnaturallyoccurringpreyspecies o
i
Received2August2016;Accepted30November2016 which are attractive to predators (Marker-Kraus and Kraus, 1994; B
118
RESEARCHARTICLE BiologyOpen(2017)6,118-124doi:10.1242/bio.021055
Winterbach et al., 2015). However, farmlands are also highly (0.75±0.05 s.e.m.). Home range sizes varied between the four
modified landscapes and movement of livestock or game is individual cheetahs (Table 1) depending on the method of home
regularly managed, leading to regular human presence across the range analysis. Home range differences between individuals were
farms.Furthermore,infrastructurelikethepresenceofhomesteads expectedastheperiodofcollaringtimevariedbetweenindividuals.
(farm houses) and artificial barriers such as roads and fences will Based on existing knowledge of cheetahs seen on cameratraps in
also likely affect wildlife movement, even if they do not directly thearea,M1,M2andM4areconsideredresidentmalesandM3a
impedeit (BoastandHouser,2012;FormanandAlexander,1998; non-resident.Asoursamplesizewasverysmall,itwasnotpossible
Hayward and Kerley, 2012). Prey availability may also vary toconductanystatisticalanalysescomparinghomerangesbasedon
according to farm type. Game farms have higher numbers of residencystatus;however,itisclearthatthelargesthomerangewas
antelopespeciesthancattlefarms(BoastandHouser,2012),andthe that of the non-resident male despite having the shortest collaring
medium-sizedpreyspeciespreferredbycheetahs(Boastetal.,2016; period and the smallest number of fixes. The minimum convex
Haywardetal.,2006) mayalsovaryindensitydependingonfarm polygon (MCP) method produced much higher home range
type.Thechanginglandscapesinfarmingareasdueto presenceof estimates than either the kernel density estimator (KDE) or
livestock and resultant overgrazing, increases in food and water time-localconvexhull(T-LoCoH)methods.
availability,andreducedpressurefromcompetitors(Mcvittie,1979)
are very likely to result in changes in cheetah behaviour and Multilevelmeta-analysis
movementcomparedwiththosefoundinprotectedareas. All models showed asignificant amount ofresidual heterogeneity
The objective of this study was to examine how cheetahs may (highQstatistic,P<0.001)implyingthatthepredictorswetesteddo
haveadaptedtheirmovementinthesefarmlandsbyusinghigh-time not account for large amounts of between-individual variance in
resolutionGPScollarsthatcanrecordfinescalemovements(Hubel effectsizes.Despitecheetahscomingfromthesamepopulationin
et al., 2016a; Wilson et al., 2013). In particular, we set out to this study, there were large differences in effect sizes between
determinehomerangesizesofcheetahsinthesenon-protectedareas individuals.Estimatedcoefficientsforaveragedurationofvisitand
andevaluatefactorsthatmayinfluencesize.Wealsoaimedtoassess numberofvisitsforeachofthethreedistancepredictorsproduced
whether cheetahs selected or avoided particular natural resources, byeachmodelareshowninTable2.
artificial structures, locations within farms and farm type by
evaluating visitation and duration rates. Scent-marking trees Visitation
(hereafter marking trees) were selected as a natural resource as When analysed separately, the numberof visits were significantly
theyare highlysought out by cheetahs (Marker-Kraus and Kraus, affectedbydistancetomarkingtrees(z=−2.40,P=0.02),withthe
1994);homesteadlocationswereselectedasartificialstructures;the highest number of visits in areas close to marking trees (Fig. 1).
centrepointsoffarmswereusedasalocationvariableandwealso Therewassome degreeofresidual variabilitybetweensubjectsin
compared cattle and game farms. To determine selection or numberof visits at marking trees (τ2=1.48±1.22 s.d.). Distance to
avoidance we used a time-based local convex hull method that homesteads was significantly positively associated with visitation
can be used to determine visitation and duration rates from GPS rates (z=6.44, P<0.01). Cheetahs visited areas further away from
telemetrydata(Lyonsetal.,2013).Thefollowinghypotheseswere homesteadsmostoftenbutshowedvariabilitybetweenindividuals
tested:(a)cheetahhomerangesareaffectedbyresidencystatus;(b) in visits directly at homesteads (τ2=0.66±0.81 s.d.). Similarly,
preferrednaturalresourceswillbevisitedmoreoftenandforlonger distance to centre of farms was also positively associated with
periodsthanareasofhumanhabitation;(c)locationpreferenceswill visitation rates, but this was not significant (z=0.41, P=0.68).
have an effect more on duration than visitation rates, and (d) Residualvariabilitybetweenindividualsonthenumberofvisitsat
visitationfrequencyanddurationishigherongamefarmsthancattle thecentreoffarmswashigh(τ2=1.04±1.02s.d.). Ineachofthese
farms. cases, all heterogeneity between subjects in the slopes was
explained by the predictor as shown by τ2=0 in each case. In our
RESULTS global model analysis, where we incorporated all predictors
Home-rangeanalysis associated with distance, only distance to marking trees had a
Individualscheetahusedonaverage21±6s.d.farms(range13–39), significantnegativeeffect(z=−3.31,P<0.01).Therewasalsosome
withindividualM3rangingoverthegreatestnumber(39)offarms. degree of variability between individuals in visitation rates at
Averagefarmsizewas68.7km2±31s.d.(range12.4–272.5)witha markingtrees(τ2=0.9).
totalof64farmsbeingvisitedbyatleastonecollaredcheetah.Of Gamefarmswerevisitedmoreoftenbycheetahthancattlefarms,
these,20farmswerevisitedbyatleasttwoofthecollaredcheetahs. butthiseffectwasnotsignificant(z=1.58,P=0.11).Theamountof
The average proportion of game farms (0.25±0.05 s.e.m.) used unexplainedvariabilitybetweencheetahswasτ2=0.60(±0.78s.d.)
by the cheetahs was lower than the proportion of cattle farms for cattle farms and τ2=0.11 (±0.34 s.d.) for game farms, after
Table1.Homerangeestimates(km2)fromfourGPScollaredcheetahsAcinonyxjubatustrackedintheGhanzidistrictofwesternBotswana
MCP KDE* T-LoCoH‡
Subject Status Totalnumberoffixes Numberofdays 50% 95% 50% 95% 50% 95%
n
e
M1 Resident 32,197 165 554 1248 153 734 91 344
p
M2 Resident 35,080 189 152 615 68 425 84 390 O
M3 Non-resident 1445 77 1212 2270 725 3248 365 1008
y
M4 Resident 2906 82 209 328 94 419 42 166 g
o
Methodsusedwereminimumconvexpolygons(MCP),fixedkerneldensityestimators(KDE)andtime-localconvexhulls(T-LoCoH). l
o
*KDEbasedonasmoothingparameterof80%ofthehref. i
‡T-LoCoHcalculatedusingtheamethod. B
119
RESEARCHARTICLE BiologyOpen(2017)6,118-124doi:10.1242/bio.021055
Table2.Estimatedcoefficientsoftheinterceptandslope(β)foreachof
threetestedpredictorsontheresponsevariablesdurationandnumber
ofvisits
Duration Visitation
Variable Est.coef 95%CI Est.coef 95%CI
Markingtree
Intercept 13.63 7.38-25.15 10.44 3.16-34.44
β 1 0.99-1 0.99 0.99-1
Homestead
Intercept 13.92 8.25-23.47 5.04 2.27-11.20
β 1 0.99-1 1 1-1
Centreoffarm
Intercept 12.13 9.43-15.51 7.25 2.66-19.76
β 1 1-1 1 0.99-1
Coefficientsand95%confidenceintervals(CI)areback-transformedresults
Fig.2.Predicteddurationofvisitasafunctionofdistancetomarking
fromeachoftheindividualmodels.
trees,homesteadsandcentralpointsoffarmsforfourcollaredcheetahs
(Acinonyxjubatus)infarmlands,Botswana.Datawereanalysedusinga
accountingforfarmtype.Thelargestandarddeviationsimplylarge two-stagemultilevelmeta-analysis.Therewasnosignificanteffectofdistance
differencesbetweencheetahsinthenumberofvisits. toeachofthesefeaturesonduration(P>0.05).
Duration for this effect, but it was not due to sex or differences in prey
Periodsofdurationwerenotaffectedbyanyofthevariableswhen availability as all cheetahs’ utilised similar proportions of game
testedindependentlyinthisstudy(Fig.2).Distancetomarkingtrees and cattle farms. Despite our relatively short collaring periods,
(z=0.20, P=0.84), homesteads (z=−0.46, P=0.65) and centre of due to the high resolution collars used, home ranges were similar
farms (z=0.86, P=0.39) were all non-significant. In all cases the to those in previous studies in farming areas in Botswana
variability between individuals (τ2 below 0.4) and in the slopes (Houser et al., 2009a), smaller than those reported in Namibia
(τ2=0) was low. However, when incorporaating all variables, (Marker et al., 2008a), and larger than those in South Africa
distance to marking trees was significantly positively associated (MarnewickandCilliers,2006).Disparityinhomerangesizeacross
with duration rates (z=2.1, P=0.04), but neither distance to farming environments suggest that differing levels of available
homestead orcentreof farms had anyeffect. Variability washigh resources, such as water, prey and level of conflict, may be
betweenindividuals(τ2=1.36). responsible for large scale variability (Marker et al., 2008a), but
Duration periods were not affected by farm type (z=−1.04, within specific regions other factors such as status may still be
P=0.3), with cheetahs showing no preference for cattle or game important.
farms. There was a larger amount of remaining residual variation Non-protectedareassuchasfarmlandsprovideresourcesthatare
betweencheetahsongamefarms(τ2=0.16±0.4s.d.)thanoncattle importantforcheetahbehaviourandhabitatuse.Markingtreeswere
farms(τ2=0.02±0.13s.d.). highlyvisitedbyallcheetahswhentheywerewithinanaveragedaily
distanceofthese.Markingtreesareanimportantresourceformale
DISCUSSION cheetahs.Malesmarkthesetreesthroughurinationorclawingand
Cheetah home range sizes in this study may vary as a function thisisconsideredtobeawaytoadvertisearesident’sterritory(Eaton,
of residency status, with a single non-resident cheetah showing 1970; Marker-Kraus and Kraus, 1994; Marnewick et al., 2006).
asubstantiallylargerrangingsizecomparedwithresidentcheetahs. Although cheetahs regularly re-visited marking trees, duration
Our small sample size did not allow valid statistical conclusions periods were similar or even shorter at these sites compared with
areasfurtherawayfromthesesites.Similarly,durationperiodwasnot
influencedbydistancetohomesteads,butasweexpected,cheetahs
avoidedvisitinghomesteads.However,generalspatiallocationssuch
ascentralpointsoffarmsdidnotinfluenceeithervisitationorduration
periodsbycheetahs.Withinthestudyarea,cheetahsshowedaslight
preference for visiting game farms more often than cattle farms,
despite the much lower proportion of game farms used by these
cheetahs.Thispreferencewasnotsignificant,mostlikelyasaresultof
our small sample size and variability in visitation rates between
cheetahs. Game farms generally have greater species-richness than
cattle farms. Density and abundance of preferred prey species for
cheetahinthisarea,suchasduiker(Sylvicapragrimmia),steenbok
(Raphiceruscampestris)andgreaterkudu(Tragelaphusstrepsiceros)
n
(Boast and Houser, 2012; Boast et al., 2016) may vary depending
e
onfarmtype.Forexample,thedensityofduikerhasbeenfoundto p
O
be higher on cattle farms, whereas steenbok density was higher
Fig.1.ThepredictednumberofvisitsofcheetahsAcinonyxjubatus(n=4) y
ongamefarms,buttherewasnodifferenceforkudu(Kent,2011). g
inresponsetothedistancefrommarkingtrees,homesteadsandcentral
If prey availability is equally spread across both farm types, then o
pointsoffarms.Atwo-stagemultilevelmeta-analysiswithrandomeffectswas l
this mayexplain similar durationperiods foundbetween cattleand o
usedtomodelthedata.Visitstomarkingtreesandhomesteadswere i
significantlyaffectedbydistancetothesefeatures(P<0.05). gamefarms. B
120
RESEARCHARTICLE BiologyOpen(2017)6,118-124doi:10.1242/bio.021055
Other variables such as roads and fences are also likely to co-exist with predators. In this small-scale study, wewere able to
influence the movement of cheetah in farmlands. For example, showdetailedspatialmovementsinfarmlandsusinghighresolution
roadshavebeenshowntoinfluencethemovementofAfricanwild GPS collars. This technology is relatively new and has not
dogs (Lycaon pictus). During travelling periods, particularly in previously been used to detail fine scale movement patterns
dense environments, roads were actively used to enhance showing selection or avoidance of landscape features in these
movement, but avoided during resting periods (Abrahms et al., environments.Scent-markingtreeshavebeenshowntobeahighly
2015). However, roads have also been shown to negatively affect selectedresourceformalesasshownbyrevisitationrates.Similarly
animalmovementsthroughavoidancebysomespecies,increasing homesteadsareactivelyavoidedasshownbylowvisitationratesto
habitatfragmentationandmortality(FormanandAlexander,1998). these areas. We found a slight preference for visitation to game
Althoughtheextentofroadsmaybesimilarinfarmlandswiththose farms compared with cattle farms, but duration rates were not
in some protected areas, the additional construction of fences in affected by farm type. A better understanding of predator
farmlands is likely to provide another potential factor influencing movements in highly human-dominated landscapes is necessary
movement. In one study, a single fence was shown to affect lion to assist with developing novel conflict mitigation measures and
movement,butwasmorepermeabletospottedhyaena,wilddogand farmmanagementpractices,asameanstoconservethesespecies.
cheetah(Cozzietal.,2013).However,thehighdensityoffencesin
farmlandsandtheireffectoncheetahmovementisunknown,astoo MATERIALSANDMETHODS
fewfarmswithadequateroadandfencelocationswereavailableto Area
beinvestigatedinthisstudy.Recentstudieshaveshownimportant ThestudywasundertakenintheGhanzidistrict,Botswana,anareaconsisting
new insights into cheetah hunting strategies in protected areas ofcommunalandcommercialfarmsaswellaswildlifemanagementareas
(WMAs). The area is considered one of the most important areas linking
(Wilsonetal.,2013)andsimilarinformationisneededforcheetah
cheetah populations between Botswana and Namibia (Winterbach et al.,
in farmlands, where the landscape and habitat can vary
2015).Thestudyareafocussedonthecommercialfarmingblockmadeupof
considerably. Such information will provide valuable insight into
morethan200cattleorgamefarms.Thehabitatisprimarilymadeupoflow
understanding adapted hunting strategies in these environments,
treeandshrubsavannahandreceiveslowannualrainfall(400mm/year).Most
and highlight potential differences in hunting dynamics and kill waterintheareaisartificiallypumpedviaboreholesorfoundinnaturalpans
ratesbetweennaturalpreyandlivestock.Thisinturnmaybeuseful duringtherainyseasonwithnopermanentsurfacewater(ColeandBrown,
indevelopingmeasurestomitigatelivestockloss. 1976).ForfurtherdescriptionseeKent(2011)andHouseretal.(2009b).
Forcheetahsto co-exist in human-dominated environments,itis
likelythattheymayneedtoadapttheirmovementstoavoidhuman Animalcollaring
activitywhilststillutilisingareasthatareimportantfortheirdietary ApprovalfortheresearchstudywasprovidedbytheBotswanaMinistryof
and behavioural needs, as we have shown here. In order to help Environment, Wildlife and Tourism and Department of Wildlife and
conservecheetahandmitigatethehuman-wildlifeconflictthatishigh NationalParksonpermitnumberEWT8/36/4.Theworkwasapprovedby
theRVCEthicsandWelfareCommittee(LOCATE2013-1233).
infarmlands(Markeretal.,2003;Purchaseetal.,2007),itisvaluable
CheetahsintheGhanzifarmlandshavebeenmonitoredbyresearchersfor
to know what areas orresources mayattractordetercheetahs.The
several yearsthrough the useofcameratraps atidentified marking trees.
importance of marking trees to cheetahs has been exploited by
This long-term monitoring has allowed researchersto determinewhether
researchersassitesthatprovidevaluablemonitoringandpopulation
cheetahsobservedattheselocationsareresidentornon-residentindividuals.
assessment data (Marker et al., 2008b; Marnewick et al., 2008). Seven cheetahs were fitted with high-time resolution GPS collars by a
However, marking trees have also been associated with negative veterinaryspecialistinAugust2014withadropoffdateforFebruary2015.
attitudesandhavebeenusedbyfamersasawaytolocateandremove Data were collected for an average of 128days per individual (range
cheetah(Markeretal.,2003;Mcvittie,1979).Inthisstudywehave 77-189days).Twocheetahsweresinglemales,threewerefromacoalition
highlightedthatcheetahactivelyvisitthesetreesandthisinformation offourandtwowereindividualsfromcoalitionsoftwo.Fromtheseven
canassistinhelpingfarmingmanagementpractices,suchasadjusting collaredcheetahsonlyfourwereusedforanalysis;onesinglemale(M3),
twofromseparatecoalitionsoftwo(M1andM4)andonefromthecoalition
movementsoflivestockorconstructionofkraalstoavoidtheseareas.
offour(M2).Asthethreecollaredmalesfromthecoalitionoffourstayed
Similarly, as cheetah showed a preference for visiting game farms
within100mofeachotherfor80%oftheGPSfixes(Hubeletal.,2016a)
comparedwithcattlefarms,maintaininghighlevelsofsmallnatural
thesewereconsiderednon-independent;onecollarfromasinglemalefailed
preyonthesefarmsmayhelpreduceconflict.Astudyassessingprey
tocollectdataaftertwoweeks.
baseforcheetahacrossBotswanafoundthatmaintaininga20%wild
preybiomassinfarmingareascansubstantiallyreducethepotential
Collarconfiguration
for human-wildlife conflict (Winterbach et al., 2015). As non-
Royal Veterinary College (RVC) high-time resolution GPS collars were
protected areas such as farmlands provide valuable resources like
designedtocollectdatatostudymovementpatternsandhuntingbehaviour.
prey, water and marking trees for cheetahs, knowledge of spatial The collars contain a GPS receiver which provides highly accurate
movements can provide valuable information in understanding positional data. This enablesthe cheetahs’ movements and activity to be
cheetah habitat utilisation, which can assist with improved tracked over time and space in great detail. In order to manage power
measuresthathelppeopleco-existwiththesecarnivores.Therefore, consumptioneffectively,thecollarsweresettoswitchdynamicallybetween
further studies on farmland cheetahs are needed to provide three different operating states depending on the animal’s activity level
informationtomanageand protect highpopulations intheseareas. (Hubelet al., 2016b;Wilsonet al., 2013). Activity levels were basedon
accelerometermeasurements.Atlowactivity,whentheanimalwasdeemed n
This study has established some important characteristics of
e
toberesting,hourlyfixeswererecorded;withincreasedactivity,thesample
movement behaviour by cheetah, but there is still much to be p
rateincreasedto5minfixes.Whentheanimalwasjudgedtoberunningthe O
learnedtofacilitateco-existenceandprotectionofthespecies.
GPS rate increased to 5Hz. The datawereretrieved by downloading via
y
externaldownloadboxesplacedatmarkingtreesordirectlyfromthecollar g
Conclusions when retrieved following pre-timed automatic drop-off (Sirtrack Ltd., o
l
As cheetahs continue to utilise and rely on non-protected areas, Havelock North, New Zealand). In this study, the main objectivewasto o
i
human-wildlife conflict will persist until we are better able to studymovementpatternsutilisingGPSpointsrecordedbythecollars,but B
121
RESEARCHARTICLE BiologyOpen(2017)6,118-124doi:10.1242/bio.021055
notallofthehigh-samplerateofpointswasrequired.Assuch,thenumberof
pointsusedwereadjusteddependingontheparticularanalysisasdescribed
below.
Homerangeanalysis
Homerangeanalysisusingminimumconvexpolygons(MCP)(Jenrichand
Turner,1969)andfixedkerneldensityestimates(KDE)werecalculatedinR
(R Core Team, 2015) using the adehabitat package (Calenge, 2015).
CalculationofMCPsat50%and95%usedallGPSdatapointscollectedper
individual.ForKDEofhomerangeat50%and95%,weusedjustasingle
dailyGPSfixchosenclosestto12:00AMtomeettheindependenceofdata
assumption.SmoothingparametersforKDEanalysiswerefirsttestedusing
least-squarescrossvalidation(LSCV)whichgivestheleastbiasedestimates
(SeamanandPowell,1996),butastheydidnotconvergeinallcases,we
adjusted our analysis to using the reference bandwidth parameter (href).
However,asthehrefcanoftenoversmoothdata,wechosetouse80%ofthe
hrefvalueasabetterestimateofsmoothing(Kieetal.,2010).GPSlocations
were incorporated into ArcMap 10.3.1 (ESRI, Redlands, CA, USA) to
generatehomerangemaps.
Calculationofvisitationanddurationrates
TheR packageT-LoCoH(Lyonset al.,2013)wasusedtocalculatetwo
dependentvariables:visitationrate(numberofseparatevisitstoahull)and
duration of visits (mean numberof points per hull) using GPS locations
recordedbyeachcollar.AllGPSpointswereusedineachdatasettocheck
foroutliersandadjustedtolocaltime.GPSpointswerethenthinnedtoone
perhourtomaintainconsistencyofpointsacrosseachindividual.T-LoCoH
creates a minimum convex hull around each data point combining both
temporalandspatialdistancesbetweentwopoints.Atime-scaleddependent
(TSD)variablesisdeterminedbasedonbalancingbothtimeandspace.Low
valuesofsindicatetimehavinglittleimportance.Todeterminesweplotted
differentvaluesofsoverarangeofdifferenttimevalues(range12-96h).
s values were chosen at 24h periods as time and space were mostly
equalisedinallcases.Time-basedlocalconvexhullscanbecreatedusing
thekmethod(setnumberofnearestneighbourpoints),rmethod(nearest
Fig.3.GPSlocationsofonecollaredmalecheetah(M2)Acinonyx
neighbours in a radius r) or a method (both time and distance are jubatus,scent-markingtrees,homesteadsandthecentrepointoffarms,
incorporated).Wechosetheamethodasthiscantakeintoaccountareas
inafarmingareaofwesternBotswana.
wherepointsmaybethinandscattered,forexamplewhenanindividualmay
enter areas not commonly used (Lyons et al., 2013). Creation of local
convexhullswascompletedbyfollowingmethodsdescribedinLyonsetal. Dataanalysis
(2013)andtheassociatedTutorialandUsersGuide.Althoughcollarswere As GPS data generates multiple response outputs per individual, it is
settorecordatregulartime intervals, therewereirregularities insatellite necessarytoaccountforcorrelationindatapointsperindividual.Mixed-
acquisitionsbetweencollars,aswellaslengthofcollaringperiods,which effectmodelsusingsubjectasarandomfactorcanaccountforcorrelated
led to differences in the number of fixes per individual. Therefore we datapoints,howeveritisnotrecommendedforrandomeffectsinwhichthe
generated a values separately for each cheetah. Local convex hulls were numberoflevelsislessthanfive(Bolkeretal.,2009).Therefore,wechose
sortedbytime-usemetricsusinganinter-visitgap(IVG)periodof12hand touseasimplertwo-stageapproachthatiseffectiveinmodellingcorrelated
thiswasusedtocreateindependentvisitationanddurationrates.Time-based dataandapplicabletoourlargedatasets(Fiebergetal.,2010;Gelman,2005;
localconvexhullscanalsobeusedtogeneratehomeranges.Takinginto Murtaugh, 2007).In R we usedthe package lme4 (Bateset al., 2015)to
account the temporal nature of location datacan be argued to be a more perform linear (LM) and generalised linear (GLM) regression models to
suitablemethodforcorrelateddatapoints(FieburgandBörger,2012;Lyons analyse the relationship of duration and visitation rate to a set of three
etal.,2013),wealsogeneratedhomerangeestimatesforcomparisonwith distance-related predictor variables (distance to marking tree, distance to
traditionalestimatesusingMCPandKDEmethodsasdescribedabove. homestead,distancetocentreoffarm)andalsofarmtype.Aswehadsubset
ourdatatoamaximumof6kmdistances(forthedistancevariables),this
Incorporationoflandscapefeatures effectivelyremovedsomedatapointsintheothertwopredictors.Aswedid
Shapefilesthatrecordedpreviouslyidentifiedmarkingtrees,homesteads, not wish to lose valuable information per variable, we chose to run our
centrepointsoffarmsandfarmtypewereincorporatedintoArcMap(Fig.3) modelsseparatelyforeachfixedfactor.However,wealsoranaglobalmodel
asspatialvariablesofinterest.Otherpotentialinfluentialfeaturessuchas incorporatingallthedistance-relatedvariablestodeterminewhetherwestill
boreholesandroadswerenotavailableforthemajorityoffarmsusedand obtainedsimilarresultsdespitethelowersamplesize.Forstage1,eachof
were not included in this study. We calculated the distance of each GPS the fourcheetahs were modelled separately foreach of the fixed factors.
pointtothenearestofeachofthevariables.Aswedidnotknowallpossible Each model was checked for normality and deviances from
markingtreesorhomesteadsinthestudyarea,werestrictedthedatasettoa homoscedasticitywerecheckedbyvisualinspectionofresidualplots.The n
e
maximum range of 6km for each for these variables. This distance response variable duration was log transformed where necessary. As
p
represents an average daily travel distance for males in this area (Houser visitationisacountvariableweusedthePoissondistribution,withalog-link O
etal.,2009a),whichallowedustolookattheeffectofavoidanceorselection function.OverdispersionwascheckedusingtheAERpackage(Kleiberand
y
when cheetahs were within an average daily distance from each feature. Zeileis,2008)andinthesecaseswecorrectedthestandarderrorsusinga g
o
Mixedfarms(cattleandgame)representedonlyasmallfractionofthefarms quasi-GLMmodel.Wecreatedvectorsofmodelcoefficientsandablock
l
in the area and were removed from our analysis to increase power and variance-covariancematrixforeachofthetwopredictors.Instage2weused o
i
simplifytheanalysis. the package metafor (Viechtbauer, 2010) to conduct a multilevel meta- B
122
RESEARCHARTICLE BiologyOpen(2017)6,118-124doi:10.1242/bio.021055
analysisusingthemodelcoefficients(interceptandslope)againsteachof Broomhall,L.S.,Mills,M.G.L.andduToit,J.T.(2003).Homerangeandhabitat
the response variables. A random effect of subject of both intercept and usebycheetahs(Acinonyxjubatus)intheKrugerNationalPark.J.Zool.261,
slope was also added to the model. Coefficients were allowed to be 119-128.
Calenge, C. (2006). The package adehabitat for the R software: atool for the
heterogeneous and correlated using an unstructured variance-covariance
analysisofspaceandhabitatusebyanimals.Ecol.Model.197,516-519.
matrix.TheQstatisticwasusedtotestwhethereffectsizesbetweensubjects Caro, T. (1994). Cheetahs of the Serengeti Plains - Group living in an Asocial
arehomogenoususingthechi-squaredistribution.Aτ2statisticisproduced Species.Chicago:UniversityofChicagoPress.
accountingfortheamountofvariabilityamongeffectsizesnotaccounted Cole,M.M.andBrown,R.C.(1976).ThevegetationoftheGhanziareaofwestern
forbythepredictorvariable.τ2valuesequaltozeroinmixed-effectmodels Botswana.J.Biogeogr.3,169-196.
Cozzi,G.,Broekhuis,F.,McNutt,J.W.andSchmid,B.(2013).Comparisonofthe
implythatheterogeneityoftrueeffectsizesareduetothepredictorinthe
effectsofartificialandnaturalbarriersonlargeAfricancarnivores:implicationsfor
model(ViechtbauerandCheung,2010).Modelswerecheckedbyplotting
interspecificrelationshipsandconnectivity.J.Anim.Ecol.82,707-715.
standardisedresidualsandcheckingforinfluentialoutliersusingthemethod Cristescu,B.,Bernard,R.andKrause,J.(2013).Partitioningofspace,habitat,
described by Habeck and Schultz (2015), whereby values whose andtimingofactivitybylargefelidsinanenclosedSouthAfricansystem.J.Ethol.
standardisedresidualsweregreaterthanthreeandhatvaluesgreaterthan 31,285-298.
two times the mean hat value were considered as potential outliers. Durant,S.M.(2000).Livingwiththeenemy:avoidanceofhyenasandlionsby
cheetahsintheSerengeti.Behav.Ecol.11,624-632.
SignificancewassetatP<0.05andmeansarereportedwithstandarderrors
Durant,S.M.,Kelly,M.andCaro,T.M.(2004).Factorsaffectinglifeanddeathin
unlessspecifiedelsewhere.Inlogtransformedmodelsandmodelsusingthe
Serengeticheetahs:environment,age,andsociality.Behav.Ecol.15,11-22.
Poissondistribution,coefficientsand95%confidenceintervalswereback- Eaton, R. L. (1970). Group interactions, spacing and territoriality in cheetahs.
transformed,andtheseresultsareusedinallgraphsandtablespresented. Z.Tierpsychol.27,481-491.
Fieberg,J.,Matthiopoulos,J.,Hebblewhite,M.,Boyce,M.S.andFrair,J.L.
Acknowledgements (2010). Correlation and studies of habitat selection: problem, red herring or
opportunity?Philos.Trans.R.Soc.BBiol.Sci.365,2233-2244.
TheauthorswouldliketothankthegovernmentofBotswanaforgrantingpermission
Fieburg,J.andBörger,L.(2012).Couldyoupleasephrase“homerange”asa
toundertakethestudy.TheywouldalsoliketothankR.KleinfromCheetah
question?J.Mammol.93,890-902.
ConservationBotswana(CCB)forfacilitatingthefieldworkandtoallthefarmerswho
Forman,R.T.T.andAlexander,L.E.(1998).Roadsandtheirmajorecological
providedaccesstotheirfarmsduringcaptureandcollaringperiods.Wewouldliketo
effects.Annu.Rev.Ecol.Syst.29,207-231.
thankJ.LoweandS.Amosforfabricatingandprogrammingcollars;K.Roskillyfor Gelman,A.(2005).Two-stageregressionandmultilevelmodeling:acommentary.
GPS/IMUfusioncode;M.Bing,J.SharpeandA.SchaeferfromVetswanafor Polit.Anal.13,459-461.
additionalveterinaryservice;A.R.Wilsonandananonymousreviewerforhelpful Habeck,C.W.andSchultz,A.K.(2015).Community-levelimpactsofwhite-tailed
commentsonthemanuscript,andH.Reeves,A.Davies,P.Seele,P.Phale, deer on understorey plants in North American forests: a meta-analysis. AoB
C.Sebati,U.MolefeandR.Phillipsforfieldassistance. Plants7,plv119.
Hayward,M.W.andKerley,G.I.H.(2012).Fencingforconservation:restrictionof
Competinginterests evolutionarypotentialoraripostetothreateningprocesses?Biol.Conserv.142,
Theauthorsdeclarenocompetingorfinancialinterests. 1-13.
Hayward, M. W., Hofmeyr,M., Brien, J. O. and Kerley, G. I. H. (2006). Prey
preferences of the cheetah (Acinonyx jubatus) (Felidae: Carnivora):
Authorcontributions
morphologicallimitationsortheneedtocapturerapidlyconsumablepreybefore
A.M.W.andT.Y.H.wereinvolvedinconceivinganddesigningthemethodsand
kleptoparasitesarrive?J.Zool.270,615-627.
study.R.M.andJ.H.wereinvolvedinorganisinganddirectingthefieldprogram,
Houser,A.,Somers,M.J.andBoast,L.K.(2009a).Homerangeuseoffree-
capture,collaringandmonitoringthestudypopulation.J.S.wasinvolvedinanimal rangingcheetahonfarmandconservationlandinBotswana.S.Afr.J.Wildl.Res.
collaring.L.K.V.d.W.analysedandinterpretedthedataandwrotethemanuscript 39,11-22.
withinputfromalltheauthors. Houser, A. M., Somers, M. J. and Boast, L. K. (2009b). Spoor density as a
measureoftruedensityofaknownpopulationoffree-rangingwildcheetahin
Funding Botswana.J.Zool.278,108-115.
FundingwasprovidedbytheEngineeringandPhysicalSciencesResearchCouncil Hubel,T.Y.,Shotton,J.,Wilshin,S.D.,Horgan,J.,Klein,R.,McKenna,R.
(EP/H013016/1),BiotechnologyandBiologicalSciencesResearchCouncil(BB/ Wilson,A.M.(2016a).Cheetahreunion-thechallengeoffindingyourfriends
J018007/1)andEuropeanResearchCouncil(323041)andcharitabledonationsto again.PLoSONE11,e0166864.
CCB. Hubel,T.Y.,Myatt,J.P.,Jordan,N.R.,Dewhirst,O.P.,McNutt,J.W.andWilson,
A. M. (2016b). Energy cost and return for hunting in African wild dogs and
cheetahs.Nat.Commun.7,11034.
Dataavailability
Jenrich, R. and Turner, F. (1969). Measurement of non-circular home range.
Thedatasetscollectedandanalysedduringthecurrentstudyareavailablefromthe J.Theor.Biol.22,227-237.
correspondingauthoronreasonablerequest.Rscriptsforanalysishavebeen Kelly,M.J.andDurant,S.M.(2007).ViabilityoftheSerengeticheetahpopulation.
providedforthismanuscriptinsupplementarydatainformationfiles. Conserv.Biol.14,786-797.
Kent,V.T.(2011).TheStatusandConservationPotentialofCarnivoresinSemi-Arid
Supplementaryinformation Rangelands,BotswanaTheGhanziFarmlands:ACaseStudy.Doctoralthesis,
Supplementaryinformationavailableonlineat DurhamUniversity.
http://bio.biologists.org/lookup/doi/10.1242/bio.021055.supplemental Kie,J.G.,Matthiopoulos,J.,Fieberg,J.,Powell,R.A.,Cagnacci,F.,Mitchell,
M.S.,Gaillard,J.andMoorcroft,P.R.(2010).Thehome-rangeconcept:are
traditional estimators still relevant with modern telemetry technology? Philos.
References
Trans.R.Soc.BBiol.Sci.365,2221-2231.
Abrahms,B.,Jordan,N.R.,Golabek,K.A.,McNutt,J.W.,Wilson,A.M.and
Kleiber, C. and Zeileis, A. (2008). Applied Econometrics with R. New York:
Brashares, J. S. (2015). Lessons from integrating behaviour and resource
Springer-Verlag.
selection: activity-specific responses of African wild dogs to roads. Anim.
Klein,R.(2007).StatusreportforthecheetahinBotswana.CATNewsspecial
Conserv.19,247-255.
issue3,14-21.
Bates,D.,Mächler,M.,Bolker,B.M.andWalker,S.(2015).Fittinglinearmixed-
Lyons,A.J.,Turner,W.C.andGetz,W.M.(2013).Homerangeplus:aspace-time
effectsmodelsusinglme4.J.Stat.Softw.67,1-48. characterizationofmovementoverreallandscapes.BMCMov.Ecol.1,2.
Bissett,C.,Parker,D.M.,Bernard,R.T.F.andPerry,T.W.(2015).Management- Marker,L.L.,Mills,M.G.L.andMacdonald,D.W.(2003).Factorsinfluencing
inducednicheshift?Theactivityofcheetahsinthepresenceoflions.Afr.J.Wildl. perceptionsofconflictandtolerancetowardcheetahsonNamibianfarmlands.
Res.45,197-203. Conserv.Biol.17,1290-1298. n
Boast,L.K.andHouser,A.(2012).Densityoflargepredatorsoncommercial Marker,L.L.,Dickman,A.J.,Mills,M.G.L.,Jeo,R.M.andMacdonald,D.W. pe
farmlandinGhanzi,Botswana.S.Afr.J.Wildl.Res.42,138-143. (2008a). Spatial ecology of cheetahs on north-central Namibian farmlands. O
Boast,L.,Houser,A.,Horgan,J.,Reeves,H.,Phale,P.andKlein,R.(2016).Prey J.Zool.274,226-238.
preferencesoffree-rangingcheetahsonfarmland:scatanalysisversusfarmers’ Marker, L. L., Fabiano, E. and Nghikembua, M. (2008b). The use of remote gy
perceptions.Afr.J.Ecol.54,424-433. camera traps to estimate density of free-ranging cheetahs in north-central o
Bolker,B.M.,Brooks,M.E.,Clark,C.J.,Geange,S.W.,Poulsen,J.R.,Stevens, Namibia.CATNews49,22-24. ol
M.H.H.andWhite,J.-S.S.(2009).Generalizedlinearmixedmodels:apractical Marker-Kraus, L. and Kraus, D. (1994). The Namibian free-ranging cheetah. i
B
guideforecologyandevolution.TrendsEcol.Evol.24,127-135. Environ.Conserv.21,369-370.
123
RESEARCHARTICLE BiologyOpen(2017)6,118-124doi:10.1242/bio.021055
Marnewick, K. and Cilliers, D. (2006). Range use of two coalitions of male RCoreTeam(2015).R:Alanguageandenvironmentforstatisticalcomputing.
cheetahs, in the Thabazimbi district of the Limpopo Province, South Africa. Vienna,Austria:RFoundationforStatisticalComputing.
S.Afr.J.Wildl.Res.36,147-151. Seaman,D.E.andPowell,R.A.(1996).Anevaluationoftheaccuracyofkernel
Marnewick,K.A.,Bothma,J.P.andVerdoorn,G.H.(2006).Usingcamera- densityestimatorsforhomerangeanalysis.Ecology77,2075-2085.
trappingtoinvestigatetheuseofatreeasascent-markingpostbycheetahsinthe Viechtbauer,W.(2010).Conductingmeta-analysesinRwiththemetaforpackage.
Thabazimbidistrict.S.Afr.J.Bot.36,139-145. J.Stat.Softw.36,1-48.
Marnewick,K.,Funston,P.andKaranth,K.U.(2008).Evaluatingcameratrapping Viechtbauer,W.andCheung,M.W.(2010).Outlierandinfluencediagnosticsfor
asamethodforestimatingcheetahabundanceinranchingareas.S.Afr.J.Wildl.
meta-analysis.Res.Synth.Methods1,112-125.
Res.38,59-65.
Welch,R.J.,Bissett,C.,Perry,T.W.andParker,D.M.(2015).Somewhereto
Mcvittie,R.(1979).ChangesinthesocialbehaviourofSouthWestAfricancheetah.
hide:homerangeandhabitatselectionofcheetahsinanarid,enclosedsystem.
Madoqua11,171-184.
J.AridEnviron.114,91-99.
Murtaugh,P.(2007).Simplicityandcomplexityinecologicaldataanalysis.Ecology
Wilson, A. M., Lowe, J.C., Roskilly,K., Hudson, P. E., Golabek, K. A. and
88,56-62.
Purchase,G.K.andduToit,J.T.(2000).Theuseofspaceandpreybycheetahsin McNutt,J.W.(2013).Locomotiondynamicsofhuntinginwildcheetahs.Nature
MatusadonaNationalPark,Zimbabwe.S.Afr.J.Wildl.Res.30,139-144. 498,185-189.
Purchase, G., Marker, L., Marnewick, K., Klein, R. and Williams, S. (2007). Winterbach,H.E.K.,Winterbach,C.W.,Boast,L.K.,Klein,R.andSomers,M.J.
Regional assessment of the status, distribution and conservation needs of (2015).Relativeavailabilityofnaturalpreyversuslivestockpredictslandscape
cheetahsinsouthernAfrica.CATNewsspecialissue344-46. suitabilityforcheetahsAcinonyxjubatusinBotswana.PeerJ3,e1033.
n
e
p
O
y
g
o
l
o
i
B
124