optimizing the positioning of wildlife crossing structures ...€¦ · these scenarios represent...
TRANSCRIPT
J Appl Ecol 20181ndash9 wileyonlinelibrarycomjournaljpe emsp|emsp1copy 2018 The Authors Journal of Applied Ecology copy 2018 British Ecological Society
Received12November2017emsp |emsp Accepted17January2018DOI 1011111365-266413117
R E S E A R C H A R T I C L E
Optimizing the positioning of wildlife crossing structures using GPS telemetry
Guillaume Bastille-Rousseau12 emsp|emspJake Wall2emsp|emspIain Douglas-Hamilton2emsp|emsp George Wittemyer12
1DepartmentofFishWildlifeandConservationBiologyColoradoStateUniversityFortCollinsCOUSA2SavetheElephantsNairobiKenya
CorrespondenceGuillaumeBastille-RousseauEmailgbrcolostateedu
Funding informationSavetheElephantsTheNatureConservancyNaturalSciencesandEngineeringResearchCouncilofCanada
HandlingEditorManuelaGonzaacutelez-Suaacuterez
Abstract1 Developmentoftransportationcorridorshasacceleratedgloballywithinfrastruc-tureprojects being implemented across remoteecosystems particularly in thetropicsSuchdevelopmentscanhavenegativeimpactsonwildlifeandtheireco-systemsThe importanceofwildlife crossing structures tomitigateadverseef-fectsofsuch features iswidely recognizedbut thesitingofand investment incrossing structures is contentiousDataonanimalmovementprovidevaluablehighly specific information for suchprocesses but canpresent analytical chal-lengesandremainunderutilizedinplanningmitigationefforts
2 Wedevelop two algorithmsbasedon Integer Linear Programming to prioritizecrossingpointsbasedonfrequencyofuseorbreadthofcoverageamongtrackedindividuals These scenarios represent metrics likely to guide the planning ofcrossingstructureswheretheformermayrelatetotheobjectiveofminimizingvehicle-animal collisions and the latter onmaintaining ecosystem connectivityWeexemplifythealgorithmsthroughapplicationonatrackingdatasetfromover150AfricanelephantslivingneartheproposedLamuPort-SouthSudan-Ethiopia-TransportcorridorWeexploretheinfluenceofsamplingbiasonoutcomesanddiscussconsiderationstoguidetheapplicationprocess
3 Given the generally open unfencednature of this ecosystem recordedmove-mentsoccurredthroughoutthesystemandathirdofthecorridorlengthintheecosystemwas intersectedby recordedelephantmovementsThe selectionofcrossingstructure locationsand their impactsonelephantsvariedwhetherweusedasubsampleofelephantrepresentativeoflocalpopulationdensityortotalsampleofmonitored individualsThetwoalgorithmsalsoselectedfordifferentcrossingstructurelocations
4 Synthesis and applicationsOurworkshowssomeofthechallengesofusingGlobalPositioningSystem telemetry indecidingwhere toput crossing structures anddemonstratestheneedtoidentifythetypeofconstraintsinthesystemandde-siredcrossingstructurecharacteristicsaprioriWerecommendmanagerscare-fullyevaluatethepresenceofpotentialbiasesintheirdataHigh-resolutiondatacombinedwithobjectiveprioritizationmethodsallowreasonedplanningactionsbut are often lacking during critical infrastructure planning stages Given the
2emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
1emsp |emspINTRODUC TION
With global development expanding the human footprint land-scape conversion and habitat loss are themost prevalent driversof population decline and key risk factors for species persistence(BrookSodhiampBradshaw2008)Landscapeconversionoccursinmany different shades from themost overt systematic alterationof the topology hydrology andvegetative structureof an area tomorenuancedchanges that leavesmuchof thenativecommunityin placeWhile nuanced developments tend to get less attentiontheycanhaveseriousecologicalramificationsLinearinfrastructuredevelopment including transportation and economic corridors inparticularisrecognizedtohaveoutsizedecologicalimpactsrelativeto their smallphysical footprints flagging these typesofdevelop-mentsasofincreasingconservationrelevanceandconcern(FahrigampRytwinski 2009) Thepotential impactsonwildlife species andecosystemfunctioningcausedbyhuman-made linearfeatures (LF)(roadsrailwayspowerlinesorpipelines)arearesultoffragmenta-tionofhabitatfacilitationofexoticspeciesinvasionandalterationofcommunityinteractions(FahrigampRytwinski2009)
InfrastructuredevelopmentandinparticularassociatedLFcaninterfere with the movement and behaviour of wildlife (EftestoslashlTsegayeHerfindal FlydalampColman 2014 Frair etal 2005) re-sulting inchanges inwildlifedistributionandabundance (FahrigampRytwinski2009)For someherbivores theedgeeffects andcan-opyopeningscausedbyLFcanenhancefoodavailabilityandaccess(GillNorrisGillampSutherland2001SmallampHunter1988)butmayalso enhance predator efficiency and alter critical interspecies in-teractions (JamesampStuart-Smith 2000Whittingtonetal 2011)LFcanincreasehumanaccesstoareaswhichcanhaveadverseef-fectsonwildlifepopulations(Northrupetal2012)andcandirectlyincreasemortalityofwildlifeandincreaseriskstohumansthroughvehiclecollisions(Neumannetal2012)Inadditiontodirectdemo-graphiceffectsLFcanactasimpermeablebarriersreducinggeneflowamongmeta-populationsbylimitingthedispersalof individu-alsorcreatingsmallpopulationdynamicsinisolatedpatches(RicoKindlmannampSedlaacuteček 2009)Mitigating thenegative impacts ofinfrastructure development on wildlife movement and ecosystemconnectivityiscriticaltothepersistenceofanimalpopulationsandthemaintenanceofkeyecosystemprocesses(Beyeretal2016)
Given these identified problems approaches to incorporateecologicaldatasets inplanningforthesitingandstructureofsuch
features are critical Crossing structures (eg overpass bridges orunderpasstunnels)arethefavouredmitigationmeasurestoreducewildlife-vehicle collisions and maintain connectivity (Clevengeramp Huijser 2011) However high-costs limit the number of thesecrossings given the limited budget allocated to these structuresIt is therefore critical that the location of these structures opti-mizestheireffectiveness inmitigatingthenegative impactsofde-velopmentsFine-scaledataonwildlifemovement suchasGlobalPositioningSystem(GPS)telemetrycanprovideaninvaluabletoolinidentifyingareasofimportanceforconnectivityalongaLF(HorneGartonKroneampLewis2007)butsurprisinglysuchdataarerarelyused during planning GPS telemetry can present analytical chal-lengeswhenappliedtoidentifyingwildlifecrossingsoneissuebeingthatthedegreetowhichthesampleofGPSmonitoredindividualsis spatially and behaviourally representative of the populationAdditionallythelimitedapplicationofGPStelemetrymayberelatedtothelackofobjectivetoolstooptimizecrossingstructureposition-ing(seeDownsetal2014LoraammampDowns2016)
In sub-SaharanAfrica over 53000kmof developments corri-dorshavebeenproposedorarebeingcreatedwiththegoalof in-creasing agricultural production natural resources extraction andoverall economic growth (Laurance SloanWeng amp Sayer 2015Wengetal2013)Theseproposedcorridorswillbisectover2000African protected areas and other areas containing high wildlifeabundances (Laurance etal 2015) making identification of keycrossing points of paramount importance to conservationmitiga-tioneffortsHerewedevelopapproachesusingGPStrackingdatatoquantifytheimportanceandoptimizetheselectionoflocationsalonginfrastructuredevelopmentsfortheestablishmentofwildlifecrossings We exemplify the utility of these approaches throughapplicationonanextensivetrackingdatasetfromover150AfricanelephantslivingneartheplannedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridoroneoftheproposedsub-Saharande-velopmentcorridorsinnorthernKenyaWedeveloptwoanalyticalapproaches forassisting in localizingpotentialcrossingstructuresthefirstidentifiestheheaviestusedcrossingpointsinagivenarea(ietherelativeuseandspatialdistributionoftheidentifiedcrossinglocations)andthesecondmaximizesaccesstocrossingstructuresbythelargestnumberofindividuals(iecoverage)Wecompareap-plicationoftheseapproachestotwodifferentelephantpopulationsamplingstrategiestoinvestigatehowsamplingbiasmayimpactre-sultsFinallywediscusstheutilityofouralgorithmswithinvarious
limitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevel-opments the tools developed and applied here can facilitate effective spatialplanning
K E Y W O R D S
AfricanelephantanimalmovementdevelopmentcorridorsGPStelemetryhabitatconnectivityintegerlinearprogrammingspatialplanningwildlifecrossingstructures
emspensp emsp | emsp3Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
managementcontextsWe includeall functionsdevelopedfor theanalysesinthefreerpackageldquowildxingrdquotofacilitatetheuseofourapproachbywildlifemanagersanddecisionmakers
2emsp |emspMATERIAL S AND METHODS
21emsp|emspStudy area
TheLaikipiaSamburuecosystem in northernKenya is inhabitedbythecountryrsquossecondlargestelephantpopulationaswellassev-eralotherspeciesofconcern including thecriticallyendangeredGrevyrsquoszebra (Equus Grevyrsquos)Theelephantpopulation isofcon-servationinterestandhasbeendesignatedoneoftheConventiononInternationalTradeinEndangeredSpecies(CITES)MonitoringofIllegalKillingofElephants(MIKE)sites(Wittemyeretal2014)TheMIKEsitedesignedtoenvelopthemajorityoftheelephantpopulations range is located at approximately 04degS to 2degN362degEto383degEcoveringanareaofapproximately34000km2 (Figure1)LanduseintheareavariesfromprivateandcommunalranchestogovernmentforestandnationalwildlifereservesTheareahasavarietyofhabitatsincludingcoolmoisthighlandforestsandsemi-aridsavannaLong-distancemovementsofbothwildlifeand livestock are common in order to access regions of highervegetation productivity driven by spatially stochastic rainfall
(Raizman Rasmussen King Ihwagi amp Douglas-Hamilton 2013WittemyerGetzVollrathampDouglas-Hamilton2007)
Our analysis is focused on a preliminary proposed section ofthe LAPSSET commerce corridor in and around the Samburu andBuffalo Springs National Reserves Complex (lying approximately05degN 375degE Figure1) The development of this transport andinfrastructure project will include the construction of railways tothecapitalsofSouthSudanandEthiopiaaroadnetworkoilpipe-linesairportsportsoilrefineriesandthreeresortcities(Figure1)Projectedrouteswillbisectthemiddleoftheecosystem
22emsp|emspData collection
We analysed GPS data collected since 1998 from 156 elephantsin this area as part of a long-term research project (WittemyerDaballenampDouglas-Hamilton2013)Duetocollarfailurestechni-caladvancesandprojectfundingcollaringeffortshavefluctuatedduringthisperiodElephantshaveacomplexgroupstructureinclud-ingstablefamiliesthatnormallytravelasagroupfromwhichmalesdisperseatmaturity(WittemyerDouglas-HamiltonampGetz2005)As a result GPS data collected from females represent a familyunitofbetween9and15individualsandmalestypicallyrepresenta single individualOnoneoccasionmore thanoneelephantwastrackedsimultaneouslyfromthesamefamilywhichwecontrolled
F IGURE 1emspTheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)CorridorprojectwillbisectcriticalwildlifeareasinnorthernKenyaWefocusedonasection(bold)oftherailwaycorridornearaConventiononInternationalTradeinEndangeredSpecies(CITES)MonitoringofIllegalKillingofElephants(MIKE)siteandnearthreenationalreserveswhereover150elephantsweretracked(representedasthinlines)
Kenya
Ethiopia
Tanzania
Somalia
Sudan
Uganda
0 340 680170 km
0 80 160 km40
LegendNational reserves
MIKE site
LAPPSET
4emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
forbyexcludingoneofthe individualsforthisanalysisErroneouslocationswerefilteredbyusingaspeed-filterof9kmhr(similartoWallWittemyerKlinkenbergLeMayampDouglas-Hamilton2013)Individual elephant tracking datasets averaged 22879 locations(range 179ndash142654) with a total sample of 3591945 locationsGPSlocationswereconvertedintotrajectorieswhereastepisthestraightlinebetweentwoconsecutivelocationsOnlystepswherethetimeintervalbetweentwolocationswaslessthan24hrwerein-cludedintheanalysissothatveryuncertainmovementstepswerenotincludedintheanalysisThesedatarepresentalargebutpoten-tially spatially biased samplebecausemany collarsweredeployedwithinnationalreserves
Wethereforeconductedaparallelanalysisonasubsetofthetotaldatasetcomprisedof40individualscollaredin2014ndash2015where collars were deployed systematically across the MIKEsiteTheseGPSlocationsarerepresentativeofelephantdensityintheecosystemascharacterizedbyaerialsurveydatacollectedin20022008and2012and locationsofnaturalmortalitycol-lectedaspartoftheMIKEecosystemwidemonitoringprogram(Ihwagi etal 2015) This dataset can be considered spatiallyunbiased
23emsp|emspQuantifying use of crossing sections of the infrastructure corridor
Toidentifycriticalcrossingpointsalongtheproposedrailwayswedivided the LAPSSET corridor into consecutive 200m long seg-ments(n=3165)evaluatedwhetheranelephantusedthesegmentandtalliedthenumberofstepsforeachelephantthatcrossed(in-tersected)thesegmentWealsocalculatedastandardizedcrossingintensitymetricCforeachsegment(s)following
whereCs isthestandardizedcrossingintensityforsegments xis isthenumberof steps from individual i thatcrossedsegments ti isthe time overwhich the individualwasmonitored in days andns isthenumberofdifferentindividualscrossingsegmentsWetestedif thecalculationof thismetricwassensitive to thenumberof in-dividuals using the segment and found no correlation (R=minus036df=1065 p=239) This standardizedmetric accounted for vari-ation in sampling frequencyand lengthofmonitoringamong indi-vidualsandcouldpotentiallyaccountfordifferentialspatialsamplingalongthecorridor(iemoreindividualscapturedandfollowedinagivenarea)Wetestedthisbylookingathowthecalculationofthecrossingintensitymetricdifferedwhetherusingthefulldatasetorthesystematicallysampledsubset
Finally we evaluated how the standardized crossing intensitymetriccomparedtocrossingpoints identifiedbycommunitygamescouts along a 40km section of the proposed LAPSSET corridorThelattercrossingpointsbasedonlocalknowledgewereassessedbydrivingtheroadwithlocalcommunityscoutsinJune2017Wecompared the segments identified by community scouts as being
usedornottothestandardizedcrossingintensityusingattestwithunequalvariancesTotestiflocalknowledgemorecloselyreflectedrecentcrossingbyelephantsorlongertermpatternsweperformedtheanalysisusingthetotaldatasetincludingonlythelatestyearofdata(May2016ndashJune2017)
24emsp|emspOptimization of crossing structure locations
Wedeveloped and applied two Integer Linear Programming algo-rithmstooptimizethepositioningofwildlifecrossingstructuresThefirstalgorithmisbasedonthestandardizedcrossingintensitymetricandthespatialdistributionofcrossingstructurelocations(hereafterreferredas thecrossing intensity [CI] algorithm)TheCIalgorithmruns an optimization procedure where the importance of a seg-mentrsquos relativeposition (spatial spread fromothercrossingpoints)vstheimportanceofintensityofusecanbespecifiedbytheuserin order to enhance the biological importance of identified cross-inglocationsandinsureselectedsegmentsaresufficientlyfarfromeachother(iewherebothprocessesareweightedequally identi-fiedcrossingswouldbeofhighuseandrelativelydistributedacrossthestudyarea)Thealgorithmwasdevelopedtoselectthemostim-portantcrossingpointsforelephants(iethoselocationsthatweremostheavilytrafficked)
Thesecondalgorithmsolvesatypicalmaximumcoverageloca-tionproblem(referredhereaftertoldquoMLCPalgorithmrdquoDownsetal2014)wheretheselectedsegmentswillldquocoverrdquothegreatestnumberofdifferentindividualsgiveneachsegmentcanserveuptoathresh-olddistanceThisalgorithmdiffersfromtheformerinthatselectedsegmentsdonothavetobeusedheavilybutcanbeincloseproxim-itytomultipleheavilyusedsegments(ieproximitytopathsismoreimportantthantheactualmovementsthathappenedacrossagivensegment) In additionbothalgorithmscanadjust theoptimizationroutinebyincorporatingnaturalcrossings(egabridgeoverariverthat is effectively a wildlife crossing point) or crossings specifieda priori for humaneconomic reasons to adjust the final optimiza-tionsolutionFurtherdetailsabouteachalgorithmareprovided inAppendixS1
Asanillustrationofeachalgorithmandtocomparehowtheycandifferwecreateddifferentscenariosthatcombinedtherep-resentative and global datasetswith each algorithm For all sce-narios we identified five sites (ie development plans budgetedtobuildfivecrossings)FortheCIalgorithmwegaveequalweighttothespatialspreadandintensityofuseFortheMCLPalgorithma thresholddistancefor thecoverageof5kmwasselectedThismeansthatsegmentswithin5km(oneachside)ofastructureareassumedtobecoveredbythisstructure(ieweassumedelephantwouldbewillingtotravelupto5kmtoreachacrossingstructuregiventhisisonthelowerendofaveragedailydisplacementsamongtheindividualstracked)Weevaluateddifferencesamongscenar-ios by calculating the minimum average distance between eachselectedcrossing(ieanalogoustotheEarthmoverrsquosdistancesta-tisticPottsAuger-MeacutetheacuteampLewis2014)WealsoranadditionalscenariosusingeachalgorithmwhicharepresentedinFiguresS1
Cs=
sumns
i=1
xis
ti
ns
emspensp emsp | emsp5Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
andS2 inAppendixS2All functions fromouranalysis (segmen-tationofLFshome-rangeintersectioncrossingintensityandop-timization algorithms) and an explanatory vignette are availableasanrpackage ldquowildxingrdquo (httpsgithubcomBastilleRousseauwildxing)
3emsp |emspRESULTS
Upto33ofthesegmentswerecrossedbyindividualsinthetotalsample containing movements of 156 individual elephants while18 of the 200m segments analysed along the LAPSSET cor-ridor were crossed by the representative subset of 40 elephants(Figure2)As calculated from the total sample each segmentwascrossed by an average of 324 different individuals (range=1ndash22Figure2d)anaverageof637times(range=1ndash88)Usingtherepre-sentativesubsetdataseteachsegmentwascrossedbyanaverageof145different individuals (range=1ndash7 Figure2b) an averageof264times(range=1ndash53)Theaveragestandardizedcrossinginten-sitywas0023(range=0002ndash0632Figure2c)forthetotalsampleand0019(range=0007ndash0186Figure2a)forthesubsetOverallthestandardizedcrossingintensitymetricestimatedfromthesub-sampledatasetshowedmoderatecorrelationwiththemetriccalcu-latedfromthetotalsample(R=37df=2411plt001)butstronger
correlationwith the number of individuals using a given segment(R=57df=2411plt001)
Comparisonbetweenthecrossingsegmentsidentifiedthroughthe optimization routine indicates that strong differentiation incrossingweighting occurredwith the different emphases of thedifferentroutines(Figure3)UsingtheCIalgorithmselectedloca-tionsshowedgreaterdifferencebetweentherepresentativesub-set and the total sample (average distance=36km Figure3ab)thanwhenusing theMCLPalgorithm (averagedistance=16kmFigure3cd) Applying each algorithm to the same dataset alsoprovideddifferentresultswithaveragedistancebetweenselectedcrossingof24and30kmfortherepresentative(Figure3ac)andglobal(Figure3bd)datasetsrespectivelyOveralltheMCLPpro-vided connectivity for a greater number of elephants (Figure3)Modifyingparametersofeachscenarioalsoresultedinadifferentsetofselectedlocations(FiguresS1andS2)indicatingthesensi-tivityofresultstoweighting
Crossing segments identified using local knowledge were notsignificantly aligned with tracking data-based crossing intensitymetrics given that the segments considered as used based onlocalknowledgedidnothaveahigherstandardizedcrossinginten-sity (t=minus074668 p=4602 Figure4a) Alignment was also notinfluenced by the timing of the tracking data (latest year of datat=minus0994p=348Figure4b)
F IGURE 2emspSubsectionoftheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)Corridorproject(a)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingarepresentativesampleof40elephants(seeSection2)(b)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingtherepresentativesampleofelephants(c)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingallmonitoredindividuals(d)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingallmonitoredindividuals
00
01 ndash 10
11 ndash 20
20 ndash 41
41 ndash 70
0000000
0000001 ndash 0001505
0001506 ndash 0002471
0002472 ndash 0005652
0005653 ndash 0018624
(b)(a) Number of individualsCrossing intensity
(c) (d)Number of individualsCrossing intensity
0000000 ndash 0000899
0000900 ndash 0003076
0003077 ndash 0004052
0004053 ndash 0011295
0011296 ndash 0063250
00 ndash 10
11 ndash 40
41 ndash 70
71 ndash 140
141 ndash 220
6emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
4emsp |emspDISCUSSION
Mitigatingtheimpactsofinfrastructurecorridorsonwildlifeiscriti-caltoreduceecologicalimpacts(ClevengerampHuijser2011)toen-hancepublicsafety(OlssonampWiden2008)andalleviatetheneedforcostlyretrofittingprojectsforfuturemitigationneedsTofacili-tateconservationplanninginrelationtothedevelopmentofinfra-structure projectswe developed simple and efficient approachesto analytically identify priority crossing locations fromGPS track-ingdataSuchdataare increasinglycollectedgloballyandprovidethehighestspatio-temporaldataonwildlifeusethatcanbepower-fulforunderstandingandmitigatingimpactsfromhumanland-usechanges
Weappliedthesealgorithmstoanextensiveradiotrackingdata-setcollectedonelephantsinhabitingtheLaikipiaSamburuecosys-temsubjecttoamajorinfrastructuredevelopmentprojectAfricanelephantswereourtargetspeciesgiventhattheyareaflagshipandan umbrella species whose protection can have significant posi-tiveimpactsonotherspecies(EppsMutayobaGwinampBrashares2011)andecosystemfunctioning(Haynes2012)Giventhegener-allyopenunfencednatureofthisecosystemrecordedmovementsoccurred throughout the system and roughly a third of the corri-dorlengthintheecosystemwasintersectedbyrecordedelephantmovements (ie elephants crossed over the proposed route) Theundevelopednatureofthestudyareaaffordsarareopportunityto
identifycriticalareaslikelytobeimpactedbydevelopmentaprioribut also presents a challenge given thatwildlifemovement is notconstrained andmost areas in the region are used bywildlifeAssuchrestrictingfocustoahandfulofcrossinglocationsissensitivetothedefinitionofonersquoscriteria
41emsp|emspApplication of crossing algorithms
WedevelopedtwoalgorithmsthathighlightdifferentfeaturesofthemovementdataandultimatelyaddressdifferentobjectiveswhenanalysingthesedataThefirstapproachtheCIalgorithmissuitedforsituationswherethespecificcrossingintensityisofinterestalonga LF This is applicable to scenarioswhen the immediate crossinglocationistheunitofinterestforplanningsuchaswhenattempt-ingtomitigatevehicle-animalcollisionsorwhencrossingstructureswillhavethebiggestimpactsbybeingplacedincurrentlyusedloca-tionsascanoccurwhenaLFwillremainunfencedandorrepresentasemi-permeablebarrierThesecondapproachusingtheMCLPal-gorithm issuited fora limited-accessscenario (eg fencing)wherefewcrossingstructureswillbelocatedatregulardistancesandcon-nectivityrestrictedinallotherlocationsIntheMCLPapproachtheimportanceoftheselectedsegmentisnotstronglyweightedratherthe focus ison identifyingthesegment thatprovides thegreatestcoverageforindividualsthatgenerallyuseanareaWeassumethisalgorithmwouldbeidealforscenarioswherefencingisusedalong
F IGURE 3emspOptimizationofcrossingstructureslocationalongtheLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridorrailwaysusingthecrossingintensityandmaximumcoveragelocationproblemalgorithms(seeSection2)(a)Thecrossingintensity(CI)algorithmselectedthetopfivecrossingsegmentsbasedonequallyweightingtheinfluenceofspatialspreadandintensityofuseintheoptimizationandusingthesubsampleofdata(b)Thescenarioin(a)appliedtothetotaldataset(c)Themaximumcoveragelocationalgorithm(MCLP)selectedthetopfivecrossingsegmentsbasedonathresholddistanceof5kmandusingtherepresentativesampleofindividuals(d)Thescenarioin(c)appliedtothetotaldataset
G
GGGG
G GG G
G
(b)(a)
elephants = 5 elephants = 43
G
GG
G
G
GG G
G
G
(c) (d)
elephants = 85 elephants = 96
emspensp emsp | emsp7Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
LFstofunnel individualsintoagivencrossingpointGiventhedif-ferentnatureofthesetwoapproachesandthedivergingresultstheyprovideitiscriticalthattheuseridentifiesthetypeofconstraintsinthesystemanddesiredcrossingstructurecharacteristics
42emsp|emspRepresentativeness of GPS telemetry
Overall theselectionofcrossingstructure locationsandtheir im-pactsonelephantsvariedwhetherweusedthesubsampleortotalsampleofmonitoredindividualsThisisconcerningfortherobust-nessof results reliantonGPSdata tooptimize crossing locationsStrong spatial bias in monitoring has the potential to impact theoveralloptimizationevenifalgorithmresultsarenotclusterednearthecoreof thespatiallybiasedsampleWerecommendmanagerscarefullyevaluatethepresenceofpotentialbiasesintheirdataandsubsampleorweightindividualdatainmannerthatoffersarepre-sentativedatasetTherunningofdifferentscenariosasexemplifiedherecanhelpdecisionmakersunderstandthesensitivityofresultstoimposeassumptionsHoweverthewidevariationinresultsinthestudyareaislikelyrelatedtothegeneralandwidelydisperseduseofthisunfencedecosystemOthersystemswithmoredevelopmentlikelyhavegreaterrestrictionsonpossiblecrossingpointsalongaLFwhichmayrequireexcludingsegmentsfromtheoptimizationalgo-rithm(bothoptimizationfunctionsavailableinthewildxingpackageincludethisoption)
We also compared elephant crossing locations obtained fromsurveyswith local communities to those identifiedusingGPS te-lemetrydatatocontrastoutcomesthatresultfromdifferenttypesof data The locations of key crossing locations were not wellmatched across these two sourcesWhere tracking and commu-nity identifiedcrossing locationscoincided thecrossing intensityasmeasuredby the trackingdatawas relatively lowThiswas ir-respectiveofthetimingoftheGPStrackingdatausedtoidentifycrossingsCommunity identified crossingsweremore likely tobefoundnearvillagesorlivestockpathsWhileGPStelemetrysuffersfromitsownissues(discussedinthepreviousparagraph)ouranaly-sissuggeststhatcommunity-baseddataareimportantwhereothersources are not available but subject to biases related to humanactivities and ideally would require more systematic surveyingEmpiricallycollecteddataonanimalactivitiesmaybemorereliablewhenavailable
43emsp|emspAdditional considerations
Our algorithms allowmultiple factors andmotivations to be con-sidered in selecting crossing locations Our algorithms combiningintensityofuseandnumberof individuals representedarean im-provement over existing approaches Previous approaches typi-callysummarized thegeneralcrossingbehaviourhabitatselectionorsimply record intensityofusewithoutsubsequentoptimization(Horne etal 2007 Sawyer Kauffman Nielson amp Horne 2009SchusterRoumlmerampGermain2013)Otherapproachesproposedop-timizationalgorithmsbutwithoutregardtosamplingbiasandspa-tialspread(Downsetal2014LoraammampDowns2016)Intensityof usemay not be the onlymetric of valuewhenquantifying theimportance of a segment forwildlifemovementApproaches thatconsider the importanceofa segment for specific typesofmove-ment(egmovementmodesGurarieetal2016MoralesHaydon
F IGURE 4emspElephantstandardizedcrossingintensityestimatedfromGPStrackingdatawascomparedwithcrossingpointsidentifiedbylocalcommunityscouts(a)Communitiesidentifiedcrossinglocations(dots)andstandardizedcrossingintensityfor200msegmentsoftheroadusingGPStelemetrycollectedbetween2001and2017(b)CommunityscoutidentifiedcrossinglocationsandstandardizedcrossingmetricsdevelopedusingGPStelemetrydatacollectedJune2016ndashJune2017
0000000
0000001 ndash 0003753
0003754 ndash 0006340
0006341 ndash 0010170
0010171 ndash 0019253
(a)
(b)Last year
Full dataset
0000000 ndash 0000691
0000692 ndash 0002250
0002251 ndash 0003951
0003952 ndash 0006352
0006353 ndash 0054055
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
2emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
1emsp |emspINTRODUC TION
With global development expanding the human footprint land-scape conversion and habitat loss are themost prevalent driversof population decline and key risk factors for species persistence(BrookSodhiampBradshaw2008)Landscapeconversionoccursinmany different shades from themost overt systematic alterationof the topology hydrology andvegetative structureof an area tomorenuancedchanges that leavesmuchof thenativecommunityin placeWhile nuanced developments tend to get less attentiontheycanhaveseriousecologicalramificationsLinearinfrastructuredevelopment including transportation and economic corridors inparticularisrecognizedtohaveoutsizedecologicalimpactsrelativeto their smallphysical footprints flagging these typesofdevelop-mentsasofincreasingconservationrelevanceandconcern(FahrigampRytwinski 2009) Thepotential impactsonwildlife species andecosystemfunctioningcausedbyhuman-made linearfeatures (LF)(roadsrailwayspowerlinesorpipelines)arearesultoffragmenta-tionofhabitatfacilitationofexoticspeciesinvasionandalterationofcommunityinteractions(FahrigampRytwinski2009)
InfrastructuredevelopmentandinparticularassociatedLFcaninterfere with the movement and behaviour of wildlife (EftestoslashlTsegayeHerfindal FlydalampColman 2014 Frair etal 2005) re-sulting inchanges inwildlifedistributionandabundance (FahrigampRytwinski2009)For someherbivores theedgeeffects andcan-opyopeningscausedbyLFcanenhancefoodavailabilityandaccess(GillNorrisGillampSutherland2001SmallampHunter1988)butmayalso enhance predator efficiency and alter critical interspecies in-teractions (JamesampStuart-Smith 2000Whittingtonetal 2011)LFcanincreasehumanaccesstoareaswhichcanhaveadverseef-fectsonwildlifepopulations(Northrupetal2012)andcandirectlyincreasemortalityofwildlifeandincreaseriskstohumansthroughvehiclecollisions(Neumannetal2012)Inadditiontodirectdemo-graphiceffectsLFcanactasimpermeablebarriersreducinggeneflowamongmeta-populationsbylimitingthedispersalof individu-alsorcreatingsmallpopulationdynamicsinisolatedpatches(RicoKindlmannampSedlaacuteček 2009)Mitigating thenegative impacts ofinfrastructure development on wildlife movement and ecosystemconnectivityiscriticaltothepersistenceofanimalpopulationsandthemaintenanceofkeyecosystemprocesses(Beyeretal2016)
Given these identified problems approaches to incorporateecologicaldatasets inplanningforthesitingandstructureofsuch
features are critical Crossing structures (eg overpass bridges orunderpasstunnels)arethefavouredmitigationmeasurestoreducewildlife-vehicle collisions and maintain connectivity (Clevengeramp Huijser 2011) However high-costs limit the number of thesecrossings given the limited budget allocated to these structuresIt is therefore critical that the location of these structures opti-mizestheireffectiveness inmitigatingthenegative impactsofde-velopmentsFine-scaledataonwildlifemovement suchasGlobalPositioningSystem(GPS)telemetrycanprovideaninvaluabletoolinidentifyingareasofimportanceforconnectivityalongaLF(HorneGartonKroneampLewis2007)butsurprisinglysuchdataarerarelyused during planning GPS telemetry can present analytical chal-lengeswhenappliedtoidentifyingwildlifecrossingsoneissuebeingthatthedegreetowhichthesampleofGPSmonitoredindividualsis spatially and behaviourally representative of the populationAdditionallythelimitedapplicationofGPStelemetrymayberelatedtothelackofobjectivetoolstooptimizecrossingstructureposition-ing(seeDownsetal2014LoraammampDowns2016)
In sub-SaharanAfrica over 53000kmof developments corri-dorshavebeenproposedorarebeingcreatedwiththegoalof in-creasing agricultural production natural resources extraction andoverall economic growth (Laurance SloanWeng amp Sayer 2015Wengetal2013)Theseproposedcorridorswillbisectover2000African protected areas and other areas containing high wildlifeabundances (Laurance etal 2015) making identification of keycrossing points of paramount importance to conservationmitiga-tioneffortsHerewedevelopapproachesusingGPStrackingdatatoquantifytheimportanceandoptimizetheselectionoflocationsalonginfrastructuredevelopmentsfortheestablishmentofwildlifecrossings We exemplify the utility of these approaches throughapplicationonanextensivetrackingdatasetfromover150AfricanelephantslivingneartheplannedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridoroneoftheproposedsub-Saharande-velopmentcorridorsinnorthernKenyaWedeveloptwoanalyticalapproaches forassisting in localizingpotentialcrossingstructuresthefirstidentifiestheheaviestusedcrossingpointsinagivenarea(ietherelativeuseandspatialdistributionoftheidentifiedcrossinglocations)andthesecondmaximizesaccesstocrossingstructuresbythelargestnumberofindividuals(iecoverage)Wecompareap-plicationoftheseapproachestotwodifferentelephantpopulationsamplingstrategiestoinvestigatehowsamplingbiasmayimpactre-sultsFinallywediscusstheutilityofouralgorithmswithinvarious
limitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevel-opments the tools developed and applied here can facilitate effective spatialplanning
K E Y W O R D S
AfricanelephantanimalmovementdevelopmentcorridorsGPStelemetryhabitatconnectivityintegerlinearprogrammingspatialplanningwildlifecrossingstructures
emspensp emsp | emsp3Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
managementcontextsWe includeall functionsdevelopedfor theanalysesinthefreerpackageldquowildxingrdquotofacilitatetheuseofourapproachbywildlifemanagersanddecisionmakers
2emsp |emspMATERIAL S AND METHODS
21emsp|emspStudy area
TheLaikipiaSamburuecosystem in northernKenya is inhabitedbythecountryrsquossecondlargestelephantpopulationaswellassev-eralotherspeciesofconcern including thecriticallyendangeredGrevyrsquoszebra (Equus Grevyrsquos)Theelephantpopulation isofcon-servationinterestandhasbeendesignatedoneoftheConventiononInternationalTradeinEndangeredSpecies(CITES)MonitoringofIllegalKillingofElephants(MIKE)sites(Wittemyeretal2014)TheMIKEsitedesignedtoenvelopthemajorityoftheelephantpopulations range is located at approximately 04degS to 2degN362degEto383degEcoveringanareaofapproximately34000km2 (Figure1)LanduseintheareavariesfromprivateandcommunalranchestogovernmentforestandnationalwildlifereservesTheareahasavarietyofhabitatsincludingcoolmoisthighlandforestsandsemi-aridsavannaLong-distancemovementsofbothwildlifeand livestock are common in order to access regions of highervegetation productivity driven by spatially stochastic rainfall
(Raizman Rasmussen King Ihwagi amp Douglas-Hamilton 2013WittemyerGetzVollrathampDouglas-Hamilton2007)
Our analysis is focused on a preliminary proposed section ofthe LAPSSET commerce corridor in and around the Samburu andBuffalo Springs National Reserves Complex (lying approximately05degN 375degE Figure1) The development of this transport andinfrastructure project will include the construction of railways tothecapitalsofSouthSudanandEthiopiaaroadnetworkoilpipe-linesairportsportsoilrefineriesandthreeresortcities(Figure1)Projectedrouteswillbisectthemiddleoftheecosystem
22emsp|emspData collection
We analysed GPS data collected since 1998 from 156 elephantsin this area as part of a long-term research project (WittemyerDaballenampDouglas-Hamilton2013)Duetocollarfailurestechni-caladvancesandprojectfundingcollaringeffortshavefluctuatedduringthisperiodElephantshaveacomplexgroupstructureinclud-ingstablefamiliesthatnormallytravelasagroupfromwhichmalesdisperseatmaturity(WittemyerDouglas-HamiltonampGetz2005)As a result GPS data collected from females represent a familyunitofbetween9and15individualsandmalestypicallyrepresenta single individualOnoneoccasionmore thanoneelephantwastrackedsimultaneouslyfromthesamefamilywhichwecontrolled
F IGURE 1emspTheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)CorridorprojectwillbisectcriticalwildlifeareasinnorthernKenyaWefocusedonasection(bold)oftherailwaycorridornearaConventiononInternationalTradeinEndangeredSpecies(CITES)MonitoringofIllegalKillingofElephants(MIKE)siteandnearthreenationalreserveswhereover150elephantsweretracked(representedasthinlines)
Kenya
Ethiopia
Tanzania
Somalia
Sudan
Uganda
0 340 680170 km
0 80 160 km40
LegendNational reserves
MIKE site
LAPPSET
4emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
forbyexcludingoneofthe individualsforthisanalysisErroneouslocationswerefilteredbyusingaspeed-filterof9kmhr(similartoWallWittemyerKlinkenbergLeMayampDouglas-Hamilton2013)Individual elephant tracking datasets averaged 22879 locations(range 179ndash142654) with a total sample of 3591945 locationsGPSlocationswereconvertedintotrajectorieswhereastepisthestraightlinebetweentwoconsecutivelocationsOnlystepswherethetimeintervalbetweentwolocationswaslessthan24hrwerein-cludedintheanalysissothatveryuncertainmovementstepswerenotincludedintheanalysisThesedatarepresentalargebutpoten-tially spatially biased samplebecausemany collarsweredeployedwithinnationalreserves
Wethereforeconductedaparallelanalysisonasubsetofthetotaldatasetcomprisedof40individualscollaredin2014ndash2015where collars were deployed systematically across the MIKEsiteTheseGPSlocationsarerepresentativeofelephantdensityintheecosystemascharacterizedbyaerialsurveydatacollectedin20022008and2012and locationsofnaturalmortalitycol-lectedaspartoftheMIKEecosystemwidemonitoringprogram(Ihwagi etal 2015) This dataset can be considered spatiallyunbiased
23emsp|emspQuantifying use of crossing sections of the infrastructure corridor
Toidentifycriticalcrossingpointsalongtheproposedrailwayswedivided the LAPSSET corridor into consecutive 200m long seg-ments(n=3165)evaluatedwhetheranelephantusedthesegmentandtalliedthenumberofstepsforeachelephantthatcrossed(in-tersected)thesegmentWealsocalculatedastandardizedcrossingintensitymetricCforeachsegment(s)following
whereCs isthestandardizedcrossingintensityforsegments xis isthenumberof steps from individual i thatcrossedsegments ti isthe time overwhich the individualwasmonitored in days andns isthenumberofdifferentindividualscrossingsegmentsWetestedif thecalculationof thismetricwassensitive to thenumberof in-dividuals using the segment and found no correlation (R=minus036df=1065 p=239) This standardizedmetric accounted for vari-ation in sampling frequencyand lengthofmonitoringamong indi-vidualsandcouldpotentiallyaccountfordifferentialspatialsamplingalongthecorridor(iemoreindividualscapturedandfollowedinagivenarea)Wetestedthisbylookingathowthecalculationofthecrossingintensitymetricdifferedwhetherusingthefulldatasetorthesystematicallysampledsubset
Finally we evaluated how the standardized crossing intensitymetriccomparedtocrossingpoints identifiedbycommunitygamescouts along a 40km section of the proposed LAPSSET corridorThelattercrossingpointsbasedonlocalknowledgewereassessedbydrivingtheroadwithlocalcommunityscoutsinJune2017Wecompared the segments identified by community scouts as being
usedornottothestandardizedcrossingintensityusingattestwithunequalvariancesTotestiflocalknowledgemorecloselyreflectedrecentcrossingbyelephantsorlongertermpatternsweperformedtheanalysisusingthetotaldatasetincludingonlythelatestyearofdata(May2016ndashJune2017)
24emsp|emspOptimization of crossing structure locations
Wedeveloped and applied two Integer Linear Programming algo-rithmstooptimizethepositioningofwildlifecrossingstructuresThefirstalgorithmisbasedonthestandardizedcrossingintensitymetricandthespatialdistributionofcrossingstructurelocations(hereafterreferredas thecrossing intensity [CI] algorithm)TheCIalgorithmruns an optimization procedure where the importance of a seg-mentrsquos relativeposition (spatial spread fromothercrossingpoints)vstheimportanceofintensityofusecanbespecifiedbytheuserin order to enhance the biological importance of identified cross-inglocationsandinsureselectedsegmentsaresufficientlyfarfromeachother(iewherebothprocessesareweightedequally identi-fiedcrossingswouldbeofhighuseandrelativelydistributedacrossthestudyarea)Thealgorithmwasdevelopedtoselectthemostim-portantcrossingpointsforelephants(iethoselocationsthatweremostheavilytrafficked)
Thesecondalgorithmsolvesatypicalmaximumcoverageloca-tionproblem(referredhereaftertoldquoMLCPalgorithmrdquoDownsetal2014)wheretheselectedsegmentswillldquocoverrdquothegreatestnumberofdifferentindividualsgiveneachsegmentcanserveuptoathresh-olddistanceThisalgorithmdiffersfromtheformerinthatselectedsegmentsdonothavetobeusedheavilybutcanbeincloseproxim-itytomultipleheavilyusedsegments(ieproximitytopathsismoreimportantthantheactualmovementsthathappenedacrossagivensegment) In additionbothalgorithmscanadjust theoptimizationroutinebyincorporatingnaturalcrossings(egabridgeoverariverthat is effectively a wildlife crossing point) or crossings specifieda priori for humaneconomic reasons to adjust the final optimiza-tionsolutionFurtherdetailsabouteachalgorithmareprovided inAppendixS1
Asanillustrationofeachalgorithmandtocomparehowtheycandifferwecreateddifferentscenariosthatcombinedtherep-resentative and global datasetswith each algorithm For all sce-narios we identified five sites (ie development plans budgetedtobuildfivecrossings)FortheCIalgorithmwegaveequalweighttothespatialspreadandintensityofuseFortheMCLPalgorithma thresholddistancefor thecoverageof5kmwasselectedThismeansthatsegmentswithin5km(oneachside)ofastructureareassumedtobecoveredbythisstructure(ieweassumedelephantwouldbewillingtotravelupto5kmtoreachacrossingstructuregiventhisisonthelowerendofaveragedailydisplacementsamongtheindividualstracked)Weevaluateddifferencesamongscenar-ios by calculating the minimum average distance between eachselectedcrossing(ieanalogoustotheEarthmoverrsquosdistancesta-tisticPottsAuger-MeacutetheacuteampLewis2014)WealsoranadditionalscenariosusingeachalgorithmwhicharepresentedinFiguresS1
Cs=
sumns
i=1
xis
ti
ns
emspensp emsp | emsp5Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
andS2 inAppendixS2All functions fromouranalysis (segmen-tationofLFshome-rangeintersectioncrossingintensityandop-timization algorithms) and an explanatory vignette are availableasanrpackage ldquowildxingrdquo (httpsgithubcomBastilleRousseauwildxing)
3emsp |emspRESULTS
Upto33ofthesegmentswerecrossedbyindividualsinthetotalsample containing movements of 156 individual elephants while18 of the 200m segments analysed along the LAPSSET cor-ridor were crossed by the representative subset of 40 elephants(Figure2)As calculated from the total sample each segmentwascrossed by an average of 324 different individuals (range=1ndash22Figure2d)anaverageof637times(range=1ndash88)Usingtherepre-sentativesubsetdataseteachsegmentwascrossedbyanaverageof145different individuals (range=1ndash7 Figure2b) an averageof264times(range=1ndash53)Theaveragestandardizedcrossinginten-sitywas0023(range=0002ndash0632Figure2c)forthetotalsampleand0019(range=0007ndash0186Figure2a)forthesubsetOverallthestandardizedcrossingintensitymetricestimatedfromthesub-sampledatasetshowedmoderatecorrelationwiththemetriccalcu-latedfromthetotalsample(R=37df=2411plt001)butstronger
correlationwith the number of individuals using a given segment(R=57df=2411plt001)
Comparisonbetweenthecrossingsegmentsidentifiedthroughthe optimization routine indicates that strong differentiation incrossingweighting occurredwith the different emphases of thedifferentroutines(Figure3)UsingtheCIalgorithmselectedloca-tionsshowedgreaterdifferencebetweentherepresentativesub-set and the total sample (average distance=36km Figure3ab)thanwhenusing theMCLPalgorithm (averagedistance=16kmFigure3cd) Applying each algorithm to the same dataset alsoprovideddifferentresultswithaveragedistancebetweenselectedcrossingof24and30kmfortherepresentative(Figure3ac)andglobal(Figure3bd)datasetsrespectivelyOveralltheMCLPpro-vided connectivity for a greater number of elephants (Figure3)Modifyingparametersofeachscenarioalsoresultedinadifferentsetofselectedlocations(FiguresS1andS2)indicatingthesensi-tivityofresultstoweighting
Crossing segments identified using local knowledge were notsignificantly aligned with tracking data-based crossing intensitymetrics given that the segments considered as used based onlocalknowledgedidnothaveahigherstandardizedcrossinginten-sity (t=minus074668 p=4602 Figure4a) Alignment was also notinfluenced by the timing of the tracking data (latest year of datat=minus0994p=348Figure4b)
F IGURE 2emspSubsectionoftheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)Corridorproject(a)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingarepresentativesampleof40elephants(seeSection2)(b)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingtherepresentativesampleofelephants(c)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingallmonitoredindividuals(d)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingallmonitoredindividuals
00
01 ndash 10
11 ndash 20
20 ndash 41
41 ndash 70
0000000
0000001 ndash 0001505
0001506 ndash 0002471
0002472 ndash 0005652
0005653 ndash 0018624
(b)(a) Number of individualsCrossing intensity
(c) (d)Number of individualsCrossing intensity
0000000 ndash 0000899
0000900 ndash 0003076
0003077 ndash 0004052
0004053 ndash 0011295
0011296 ndash 0063250
00 ndash 10
11 ndash 40
41 ndash 70
71 ndash 140
141 ndash 220
6emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
4emsp |emspDISCUSSION
Mitigatingtheimpactsofinfrastructurecorridorsonwildlifeiscriti-caltoreduceecologicalimpacts(ClevengerampHuijser2011)toen-hancepublicsafety(OlssonampWiden2008)andalleviatetheneedforcostlyretrofittingprojectsforfuturemitigationneedsTofacili-tateconservationplanninginrelationtothedevelopmentofinfra-structure projectswe developed simple and efficient approachesto analytically identify priority crossing locations fromGPS track-ingdataSuchdataare increasinglycollectedgloballyandprovidethehighestspatio-temporaldataonwildlifeusethatcanbepower-fulforunderstandingandmitigatingimpactsfromhumanland-usechanges
Weappliedthesealgorithmstoanextensiveradiotrackingdata-setcollectedonelephantsinhabitingtheLaikipiaSamburuecosys-temsubjecttoamajorinfrastructuredevelopmentprojectAfricanelephantswereourtargetspeciesgiventhattheyareaflagshipandan umbrella species whose protection can have significant posi-tiveimpactsonotherspecies(EppsMutayobaGwinampBrashares2011)andecosystemfunctioning(Haynes2012)Giventhegener-allyopenunfencednatureofthisecosystemrecordedmovementsoccurred throughout the system and roughly a third of the corri-dorlengthintheecosystemwasintersectedbyrecordedelephantmovements (ie elephants crossed over the proposed route) Theundevelopednatureofthestudyareaaffordsarareopportunityto
identifycriticalareaslikelytobeimpactedbydevelopmentaprioribut also presents a challenge given thatwildlifemovement is notconstrained andmost areas in the region are used bywildlifeAssuchrestrictingfocustoahandfulofcrossinglocationsissensitivetothedefinitionofonersquoscriteria
41emsp|emspApplication of crossing algorithms
WedevelopedtwoalgorithmsthathighlightdifferentfeaturesofthemovementdataandultimatelyaddressdifferentobjectiveswhenanalysingthesedataThefirstapproachtheCIalgorithmissuitedforsituationswherethespecificcrossingintensityisofinterestalonga LF This is applicable to scenarioswhen the immediate crossinglocationistheunitofinterestforplanningsuchaswhenattempt-ingtomitigatevehicle-animalcollisionsorwhencrossingstructureswillhavethebiggestimpactsbybeingplacedincurrentlyusedloca-tionsascanoccurwhenaLFwillremainunfencedandorrepresentasemi-permeablebarrierThesecondapproachusingtheMCLPal-gorithm issuited fora limited-accessscenario (eg fencing)wherefewcrossingstructureswillbelocatedatregulardistancesandcon-nectivityrestrictedinallotherlocationsIntheMCLPapproachtheimportanceoftheselectedsegmentisnotstronglyweightedratherthe focus ison identifyingthesegment thatprovides thegreatestcoverageforindividualsthatgenerallyuseanareaWeassumethisalgorithmwouldbeidealforscenarioswherefencingisusedalong
F IGURE 3emspOptimizationofcrossingstructureslocationalongtheLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridorrailwaysusingthecrossingintensityandmaximumcoveragelocationproblemalgorithms(seeSection2)(a)Thecrossingintensity(CI)algorithmselectedthetopfivecrossingsegmentsbasedonequallyweightingtheinfluenceofspatialspreadandintensityofuseintheoptimizationandusingthesubsampleofdata(b)Thescenarioin(a)appliedtothetotaldataset(c)Themaximumcoveragelocationalgorithm(MCLP)selectedthetopfivecrossingsegmentsbasedonathresholddistanceof5kmandusingtherepresentativesampleofindividuals(d)Thescenarioin(c)appliedtothetotaldataset
G
GGGG
G GG G
G
(b)(a)
elephants = 5 elephants = 43
G
GG
G
G
GG G
G
G
(c) (d)
elephants = 85 elephants = 96
emspensp emsp | emsp7Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
LFstofunnel individualsintoagivencrossingpointGiventhedif-ferentnatureofthesetwoapproachesandthedivergingresultstheyprovideitiscriticalthattheuseridentifiesthetypeofconstraintsinthesystemanddesiredcrossingstructurecharacteristics
42emsp|emspRepresentativeness of GPS telemetry
Overall theselectionofcrossingstructure locationsandtheir im-pactsonelephantsvariedwhetherweusedthesubsampleortotalsampleofmonitoredindividualsThisisconcerningfortherobust-nessof results reliantonGPSdata tooptimize crossing locationsStrong spatial bias in monitoring has the potential to impact theoveralloptimizationevenifalgorithmresultsarenotclusterednearthecoreof thespatiallybiasedsampleWerecommendmanagerscarefullyevaluatethepresenceofpotentialbiasesintheirdataandsubsampleorweightindividualdatainmannerthatoffersarepre-sentativedatasetTherunningofdifferentscenariosasexemplifiedherecanhelpdecisionmakersunderstandthesensitivityofresultstoimposeassumptionsHoweverthewidevariationinresultsinthestudyareaislikelyrelatedtothegeneralandwidelydisperseduseofthisunfencedecosystemOthersystemswithmoredevelopmentlikelyhavegreaterrestrictionsonpossiblecrossingpointsalongaLFwhichmayrequireexcludingsegmentsfromtheoptimizationalgo-rithm(bothoptimizationfunctionsavailableinthewildxingpackageincludethisoption)
We also compared elephant crossing locations obtained fromsurveyswith local communities to those identifiedusingGPS te-lemetrydatatocontrastoutcomesthatresultfromdifferenttypesof data The locations of key crossing locations were not wellmatched across these two sourcesWhere tracking and commu-nity identifiedcrossing locationscoincided thecrossing intensityasmeasuredby the trackingdatawas relatively lowThiswas ir-respectiveofthetimingoftheGPStrackingdatausedtoidentifycrossingsCommunity identified crossingsweremore likely tobefoundnearvillagesorlivestockpathsWhileGPStelemetrysuffersfromitsownissues(discussedinthepreviousparagraph)ouranaly-sissuggeststhatcommunity-baseddataareimportantwhereothersources are not available but subject to biases related to humanactivities and ideally would require more systematic surveyingEmpiricallycollecteddataonanimalactivitiesmaybemorereliablewhenavailable
43emsp|emspAdditional considerations
Our algorithms allowmultiple factors andmotivations to be con-sidered in selecting crossing locations Our algorithms combiningintensityofuseandnumberof individuals representedarean im-provement over existing approaches Previous approaches typi-callysummarized thegeneralcrossingbehaviourhabitatselectionorsimply record intensityofusewithoutsubsequentoptimization(Horne etal 2007 Sawyer Kauffman Nielson amp Horne 2009SchusterRoumlmerampGermain2013)Otherapproachesproposedop-timizationalgorithmsbutwithoutregardtosamplingbiasandspa-tialspread(Downsetal2014LoraammampDowns2016)Intensityof usemay not be the onlymetric of valuewhenquantifying theimportance of a segment forwildlifemovementApproaches thatconsider the importanceofa segment for specific typesofmove-ment(egmovementmodesGurarieetal2016MoralesHaydon
F IGURE 4emspElephantstandardizedcrossingintensityestimatedfromGPStrackingdatawascomparedwithcrossingpointsidentifiedbylocalcommunityscouts(a)Communitiesidentifiedcrossinglocations(dots)andstandardizedcrossingintensityfor200msegmentsoftheroadusingGPStelemetrycollectedbetween2001and2017(b)CommunityscoutidentifiedcrossinglocationsandstandardizedcrossingmetricsdevelopedusingGPStelemetrydatacollectedJune2016ndashJune2017
0000000
0000001 ndash 0003753
0003754 ndash 0006340
0006341 ndash 0010170
0010171 ndash 0019253
(a)
(b)Last year
Full dataset
0000000 ndash 0000691
0000692 ndash 0002250
0002251 ndash 0003951
0003952 ndash 0006352
0006353 ndash 0054055
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
emspensp emsp | emsp3Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
managementcontextsWe includeall functionsdevelopedfor theanalysesinthefreerpackageldquowildxingrdquotofacilitatetheuseofourapproachbywildlifemanagersanddecisionmakers
2emsp |emspMATERIAL S AND METHODS
21emsp|emspStudy area
TheLaikipiaSamburuecosystem in northernKenya is inhabitedbythecountryrsquossecondlargestelephantpopulationaswellassev-eralotherspeciesofconcern including thecriticallyendangeredGrevyrsquoszebra (Equus Grevyrsquos)Theelephantpopulation isofcon-servationinterestandhasbeendesignatedoneoftheConventiononInternationalTradeinEndangeredSpecies(CITES)MonitoringofIllegalKillingofElephants(MIKE)sites(Wittemyeretal2014)TheMIKEsitedesignedtoenvelopthemajorityoftheelephantpopulations range is located at approximately 04degS to 2degN362degEto383degEcoveringanareaofapproximately34000km2 (Figure1)LanduseintheareavariesfromprivateandcommunalranchestogovernmentforestandnationalwildlifereservesTheareahasavarietyofhabitatsincludingcoolmoisthighlandforestsandsemi-aridsavannaLong-distancemovementsofbothwildlifeand livestock are common in order to access regions of highervegetation productivity driven by spatially stochastic rainfall
(Raizman Rasmussen King Ihwagi amp Douglas-Hamilton 2013WittemyerGetzVollrathampDouglas-Hamilton2007)
Our analysis is focused on a preliminary proposed section ofthe LAPSSET commerce corridor in and around the Samburu andBuffalo Springs National Reserves Complex (lying approximately05degN 375degE Figure1) The development of this transport andinfrastructure project will include the construction of railways tothecapitalsofSouthSudanandEthiopiaaroadnetworkoilpipe-linesairportsportsoilrefineriesandthreeresortcities(Figure1)Projectedrouteswillbisectthemiddleoftheecosystem
22emsp|emspData collection
We analysed GPS data collected since 1998 from 156 elephantsin this area as part of a long-term research project (WittemyerDaballenampDouglas-Hamilton2013)Duetocollarfailurestechni-caladvancesandprojectfundingcollaringeffortshavefluctuatedduringthisperiodElephantshaveacomplexgroupstructureinclud-ingstablefamiliesthatnormallytravelasagroupfromwhichmalesdisperseatmaturity(WittemyerDouglas-HamiltonampGetz2005)As a result GPS data collected from females represent a familyunitofbetween9and15individualsandmalestypicallyrepresenta single individualOnoneoccasionmore thanoneelephantwastrackedsimultaneouslyfromthesamefamilywhichwecontrolled
F IGURE 1emspTheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)CorridorprojectwillbisectcriticalwildlifeareasinnorthernKenyaWefocusedonasection(bold)oftherailwaycorridornearaConventiononInternationalTradeinEndangeredSpecies(CITES)MonitoringofIllegalKillingofElephants(MIKE)siteandnearthreenationalreserveswhereover150elephantsweretracked(representedasthinlines)
Kenya
Ethiopia
Tanzania
Somalia
Sudan
Uganda
0 340 680170 km
0 80 160 km40
LegendNational reserves
MIKE site
LAPPSET
4emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
forbyexcludingoneofthe individualsforthisanalysisErroneouslocationswerefilteredbyusingaspeed-filterof9kmhr(similartoWallWittemyerKlinkenbergLeMayampDouglas-Hamilton2013)Individual elephant tracking datasets averaged 22879 locations(range 179ndash142654) with a total sample of 3591945 locationsGPSlocationswereconvertedintotrajectorieswhereastepisthestraightlinebetweentwoconsecutivelocationsOnlystepswherethetimeintervalbetweentwolocationswaslessthan24hrwerein-cludedintheanalysissothatveryuncertainmovementstepswerenotincludedintheanalysisThesedatarepresentalargebutpoten-tially spatially biased samplebecausemany collarsweredeployedwithinnationalreserves
Wethereforeconductedaparallelanalysisonasubsetofthetotaldatasetcomprisedof40individualscollaredin2014ndash2015where collars were deployed systematically across the MIKEsiteTheseGPSlocationsarerepresentativeofelephantdensityintheecosystemascharacterizedbyaerialsurveydatacollectedin20022008and2012and locationsofnaturalmortalitycol-lectedaspartoftheMIKEecosystemwidemonitoringprogram(Ihwagi etal 2015) This dataset can be considered spatiallyunbiased
23emsp|emspQuantifying use of crossing sections of the infrastructure corridor
Toidentifycriticalcrossingpointsalongtheproposedrailwayswedivided the LAPSSET corridor into consecutive 200m long seg-ments(n=3165)evaluatedwhetheranelephantusedthesegmentandtalliedthenumberofstepsforeachelephantthatcrossed(in-tersected)thesegmentWealsocalculatedastandardizedcrossingintensitymetricCforeachsegment(s)following
whereCs isthestandardizedcrossingintensityforsegments xis isthenumberof steps from individual i thatcrossedsegments ti isthe time overwhich the individualwasmonitored in days andns isthenumberofdifferentindividualscrossingsegmentsWetestedif thecalculationof thismetricwassensitive to thenumberof in-dividuals using the segment and found no correlation (R=minus036df=1065 p=239) This standardizedmetric accounted for vari-ation in sampling frequencyand lengthofmonitoringamong indi-vidualsandcouldpotentiallyaccountfordifferentialspatialsamplingalongthecorridor(iemoreindividualscapturedandfollowedinagivenarea)Wetestedthisbylookingathowthecalculationofthecrossingintensitymetricdifferedwhetherusingthefulldatasetorthesystematicallysampledsubset
Finally we evaluated how the standardized crossing intensitymetriccomparedtocrossingpoints identifiedbycommunitygamescouts along a 40km section of the proposed LAPSSET corridorThelattercrossingpointsbasedonlocalknowledgewereassessedbydrivingtheroadwithlocalcommunityscoutsinJune2017Wecompared the segments identified by community scouts as being
usedornottothestandardizedcrossingintensityusingattestwithunequalvariancesTotestiflocalknowledgemorecloselyreflectedrecentcrossingbyelephantsorlongertermpatternsweperformedtheanalysisusingthetotaldatasetincludingonlythelatestyearofdata(May2016ndashJune2017)
24emsp|emspOptimization of crossing structure locations
Wedeveloped and applied two Integer Linear Programming algo-rithmstooptimizethepositioningofwildlifecrossingstructuresThefirstalgorithmisbasedonthestandardizedcrossingintensitymetricandthespatialdistributionofcrossingstructurelocations(hereafterreferredas thecrossing intensity [CI] algorithm)TheCIalgorithmruns an optimization procedure where the importance of a seg-mentrsquos relativeposition (spatial spread fromothercrossingpoints)vstheimportanceofintensityofusecanbespecifiedbytheuserin order to enhance the biological importance of identified cross-inglocationsandinsureselectedsegmentsaresufficientlyfarfromeachother(iewherebothprocessesareweightedequally identi-fiedcrossingswouldbeofhighuseandrelativelydistributedacrossthestudyarea)Thealgorithmwasdevelopedtoselectthemostim-portantcrossingpointsforelephants(iethoselocationsthatweremostheavilytrafficked)
Thesecondalgorithmsolvesatypicalmaximumcoverageloca-tionproblem(referredhereaftertoldquoMLCPalgorithmrdquoDownsetal2014)wheretheselectedsegmentswillldquocoverrdquothegreatestnumberofdifferentindividualsgiveneachsegmentcanserveuptoathresh-olddistanceThisalgorithmdiffersfromtheformerinthatselectedsegmentsdonothavetobeusedheavilybutcanbeincloseproxim-itytomultipleheavilyusedsegments(ieproximitytopathsismoreimportantthantheactualmovementsthathappenedacrossagivensegment) In additionbothalgorithmscanadjust theoptimizationroutinebyincorporatingnaturalcrossings(egabridgeoverariverthat is effectively a wildlife crossing point) or crossings specifieda priori for humaneconomic reasons to adjust the final optimiza-tionsolutionFurtherdetailsabouteachalgorithmareprovided inAppendixS1
Asanillustrationofeachalgorithmandtocomparehowtheycandifferwecreateddifferentscenariosthatcombinedtherep-resentative and global datasetswith each algorithm For all sce-narios we identified five sites (ie development plans budgetedtobuildfivecrossings)FortheCIalgorithmwegaveequalweighttothespatialspreadandintensityofuseFortheMCLPalgorithma thresholddistancefor thecoverageof5kmwasselectedThismeansthatsegmentswithin5km(oneachside)ofastructureareassumedtobecoveredbythisstructure(ieweassumedelephantwouldbewillingtotravelupto5kmtoreachacrossingstructuregiventhisisonthelowerendofaveragedailydisplacementsamongtheindividualstracked)Weevaluateddifferencesamongscenar-ios by calculating the minimum average distance between eachselectedcrossing(ieanalogoustotheEarthmoverrsquosdistancesta-tisticPottsAuger-MeacutetheacuteampLewis2014)WealsoranadditionalscenariosusingeachalgorithmwhicharepresentedinFiguresS1
Cs=
sumns
i=1
xis
ti
ns
emspensp emsp | emsp5Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
andS2 inAppendixS2All functions fromouranalysis (segmen-tationofLFshome-rangeintersectioncrossingintensityandop-timization algorithms) and an explanatory vignette are availableasanrpackage ldquowildxingrdquo (httpsgithubcomBastilleRousseauwildxing)
3emsp |emspRESULTS
Upto33ofthesegmentswerecrossedbyindividualsinthetotalsample containing movements of 156 individual elephants while18 of the 200m segments analysed along the LAPSSET cor-ridor were crossed by the representative subset of 40 elephants(Figure2)As calculated from the total sample each segmentwascrossed by an average of 324 different individuals (range=1ndash22Figure2d)anaverageof637times(range=1ndash88)Usingtherepre-sentativesubsetdataseteachsegmentwascrossedbyanaverageof145different individuals (range=1ndash7 Figure2b) an averageof264times(range=1ndash53)Theaveragestandardizedcrossinginten-sitywas0023(range=0002ndash0632Figure2c)forthetotalsampleand0019(range=0007ndash0186Figure2a)forthesubsetOverallthestandardizedcrossingintensitymetricestimatedfromthesub-sampledatasetshowedmoderatecorrelationwiththemetriccalcu-latedfromthetotalsample(R=37df=2411plt001)butstronger
correlationwith the number of individuals using a given segment(R=57df=2411plt001)
Comparisonbetweenthecrossingsegmentsidentifiedthroughthe optimization routine indicates that strong differentiation incrossingweighting occurredwith the different emphases of thedifferentroutines(Figure3)UsingtheCIalgorithmselectedloca-tionsshowedgreaterdifferencebetweentherepresentativesub-set and the total sample (average distance=36km Figure3ab)thanwhenusing theMCLPalgorithm (averagedistance=16kmFigure3cd) Applying each algorithm to the same dataset alsoprovideddifferentresultswithaveragedistancebetweenselectedcrossingof24and30kmfortherepresentative(Figure3ac)andglobal(Figure3bd)datasetsrespectivelyOveralltheMCLPpro-vided connectivity for a greater number of elephants (Figure3)Modifyingparametersofeachscenarioalsoresultedinadifferentsetofselectedlocations(FiguresS1andS2)indicatingthesensi-tivityofresultstoweighting
Crossing segments identified using local knowledge were notsignificantly aligned with tracking data-based crossing intensitymetrics given that the segments considered as used based onlocalknowledgedidnothaveahigherstandardizedcrossinginten-sity (t=minus074668 p=4602 Figure4a) Alignment was also notinfluenced by the timing of the tracking data (latest year of datat=minus0994p=348Figure4b)
F IGURE 2emspSubsectionoftheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)Corridorproject(a)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingarepresentativesampleof40elephants(seeSection2)(b)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingtherepresentativesampleofelephants(c)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingallmonitoredindividuals(d)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingallmonitoredindividuals
00
01 ndash 10
11 ndash 20
20 ndash 41
41 ndash 70
0000000
0000001 ndash 0001505
0001506 ndash 0002471
0002472 ndash 0005652
0005653 ndash 0018624
(b)(a) Number of individualsCrossing intensity
(c) (d)Number of individualsCrossing intensity
0000000 ndash 0000899
0000900 ndash 0003076
0003077 ndash 0004052
0004053 ndash 0011295
0011296 ndash 0063250
00 ndash 10
11 ndash 40
41 ndash 70
71 ndash 140
141 ndash 220
6emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
4emsp |emspDISCUSSION
Mitigatingtheimpactsofinfrastructurecorridorsonwildlifeiscriti-caltoreduceecologicalimpacts(ClevengerampHuijser2011)toen-hancepublicsafety(OlssonampWiden2008)andalleviatetheneedforcostlyretrofittingprojectsforfuturemitigationneedsTofacili-tateconservationplanninginrelationtothedevelopmentofinfra-structure projectswe developed simple and efficient approachesto analytically identify priority crossing locations fromGPS track-ingdataSuchdataare increasinglycollectedgloballyandprovidethehighestspatio-temporaldataonwildlifeusethatcanbepower-fulforunderstandingandmitigatingimpactsfromhumanland-usechanges
Weappliedthesealgorithmstoanextensiveradiotrackingdata-setcollectedonelephantsinhabitingtheLaikipiaSamburuecosys-temsubjecttoamajorinfrastructuredevelopmentprojectAfricanelephantswereourtargetspeciesgiventhattheyareaflagshipandan umbrella species whose protection can have significant posi-tiveimpactsonotherspecies(EppsMutayobaGwinampBrashares2011)andecosystemfunctioning(Haynes2012)Giventhegener-allyopenunfencednatureofthisecosystemrecordedmovementsoccurred throughout the system and roughly a third of the corri-dorlengthintheecosystemwasintersectedbyrecordedelephantmovements (ie elephants crossed over the proposed route) Theundevelopednatureofthestudyareaaffordsarareopportunityto
identifycriticalareaslikelytobeimpactedbydevelopmentaprioribut also presents a challenge given thatwildlifemovement is notconstrained andmost areas in the region are used bywildlifeAssuchrestrictingfocustoahandfulofcrossinglocationsissensitivetothedefinitionofonersquoscriteria
41emsp|emspApplication of crossing algorithms
WedevelopedtwoalgorithmsthathighlightdifferentfeaturesofthemovementdataandultimatelyaddressdifferentobjectiveswhenanalysingthesedataThefirstapproachtheCIalgorithmissuitedforsituationswherethespecificcrossingintensityisofinterestalonga LF This is applicable to scenarioswhen the immediate crossinglocationistheunitofinterestforplanningsuchaswhenattempt-ingtomitigatevehicle-animalcollisionsorwhencrossingstructureswillhavethebiggestimpactsbybeingplacedincurrentlyusedloca-tionsascanoccurwhenaLFwillremainunfencedandorrepresentasemi-permeablebarrierThesecondapproachusingtheMCLPal-gorithm issuited fora limited-accessscenario (eg fencing)wherefewcrossingstructureswillbelocatedatregulardistancesandcon-nectivityrestrictedinallotherlocationsIntheMCLPapproachtheimportanceoftheselectedsegmentisnotstronglyweightedratherthe focus ison identifyingthesegment thatprovides thegreatestcoverageforindividualsthatgenerallyuseanareaWeassumethisalgorithmwouldbeidealforscenarioswherefencingisusedalong
F IGURE 3emspOptimizationofcrossingstructureslocationalongtheLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridorrailwaysusingthecrossingintensityandmaximumcoveragelocationproblemalgorithms(seeSection2)(a)Thecrossingintensity(CI)algorithmselectedthetopfivecrossingsegmentsbasedonequallyweightingtheinfluenceofspatialspreadandintensityofuseintheoptimizationandusingthesubsampleofdata(b)Thescenarioin(a)appliedtothetotaldataset(c)Themaximumcoveragelocationalgorithm(MCLP)selectedthetopfivecrossingsegmentsbasedonathresholddistanceof5kmandusingtherepresentativesampleofindividuals(d)Thescenarioin(c)appliedtothetotaldataset
G
GGGG
G GG G
G
(b)(a)
elephants = 5 elephants = 43
G
GG
G
G
GG G
G
G
(c) (d)
elephants = 85 elephants = 96
emspensp emsp | emsp7Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
LFstofunnel individualsintoagivencrossingpointGiventhedif-ferentnatureofthesetwoapproachesandthedivergingresultstheyprovideitiscriticalthattheuseridentifiesthetypeofconstraintsinthesystemanddesiredcrossingstructurecharacteristics
42emsp|emspRepresentativeness of GPS telemetry
Overall theselectionofcrossingstructure locationsandtheir im-pactsonelephantsvariedwhetherweusedthesubsampleortotalsampleofmonitoredindividualsThisisconcerningfortherobust-nessof results reliantonGPSdata tooptimize crossing locationsStrong spatial bias in monitoring has the potential to impact theoveralloptimizationevenifalgorithmresultsarenotclusterednearthecoreof thespatiallybiasedsampleWerecommendmanagerscarefullyevaluatethepresenceofpotentialbiasesintheirdataandsubsampleorweightindividualdatainmannerthatoffersarepre-sentativedatasetTherunningofdifferentscenariosasexemplifiedherecanhelpdecisionmakersunderstandthesensitivityofresultstoimposeassumptionsHoweverthewidevariationinresultsinthestudyareaislikelyrelatedtothegeneralandwidelydisperseduseofthisunfencedecosystemOthersystemswithmoredevelopmentlikelyhavegreaterrestrictionsonpossiblecrossingpointsalongaLFwhichmayrequireexcludingsegmentsfromtheoptimizationalgo-rithm(bothoptimizationfunctionsavailableinthewildxingpackageincludethisoption)
We also compared elephant crossing locations obtained fromsurveyswith local communities to those identifiedusingGPS te-lemetrydatatocontrastoutcomesthatresultfromdifferenttypesof data The locations of key crossing locations were not wellmatched across these two sourcesWhere tracking and commu-nity identifiedcrossing locationscoincided thecrossing intensityasmeasuredby the trackingdatawas relatively lowThiswas ir-respectiveofthetimingoftheGPStrackingdatausedtoidentifycrossingsCommunity identified crossingsweremore likely tobefoundnearvillagesorlivestockpathsWhileGPStelemetrysuffersfromitsownissues(discussedinthepreviousparagraph)ouranaly-sissuggeststhatcommunity-baseddataareimportantwhereothersources are not available but subject to biases related to humanactivities and ideally would require more systematic surveyingEmpiricallycollecteddataonanimalactivitiesmaybemorereliablewhenavailable
43emsp|emspAdditional considerations
Our algorithms allowmultiple factors andmotivations to be con-sidered in selecting crossing locations Our algorithms combiningintensityofuseandnumberof individuals representedarean im-provement over existing approaches Previous approaches typi-callysummarized thegeneralcrossingbehaviourhabitatselectionorsimply record intensityofusewithoutsubsequentoptimization(Horne etal 2007 Sawyer Kauffman Nielson amp Horne 2009SchusterRoumlmerampGermain2013)Otherapproachesproposedop-timizationalgorithmsbutwithoutregardtosamplingbiasandspa-tialspread(Downsetal2014LoraammampDowns2016)Intensityof usemay not be the onlymetric of valuewhenquantifying theimportance of a segment forwildlifemovementApproaches thatconsider the importanceofa segment for specific typesofmove-ment(egmovementmodesGurarieetal2016MoralesHaydon
F IGURE 4emspElephantstandardizedcrossingintensityestimatedfromGPStrackingdatawascomparedwithcrossingpointsidentifiedbylocalcommunityscouts(a)Communitiesidentifiedcrossinglocations(dots)andstandardizedcrossingintensityfor200msegmentsoftheroadusingGPStelemetrycollectedbetween2001and2017(b)CommunityscoutidentifiedcrossinglocationsandstandardizedcrossingmetricsdevelopedusingGPStelemetrydatacollectedJune2016ndashJune2017
0000000
0000001 ndash 0003753
0003754 ndash 0006340
0006341 ndash 0010170
0010171 ndash 0019253
(a)
(b)Last year
Full dataset
0000000 ndash 0000691
0000692 ndash 0002250
0002251 ndash 0003951
0003952 ndash 0006352
0006353 ndash 0054055
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
4emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
forbyexcludingoneofthe individualsforthisanalysisErroneouslocationswerefilteredbyusingaspeed-filterof9kmhr(similartoWallWittemyerKlinkenbergLeMayampDouglas-Hamilton2013)Individual elephant tracking datasets averaged 22879 locations(range 179ndash142654) with a total sample of 3591945 locationsGPSlocationswereconvertedintotrajectorieswhereastepisthestraightlinebetweentwoconsecutivelocationsOnlystepswherethetimeintervalbetweentwolocationswaslessthan24hrwerein-cludedintheanalysissothatveryuncertainmovementstepswerenotincludedintheanalysisThesedatarepresentalargebutpoten-tially spatially biased samplebecausemany collarsweredeployedwithinnationalreserves
Wethereforeconductedaparallelanalysisonasubsetofthetotaldatasetcomprisedof40individualscollaredin2014ndash2015where collars were deployed systematically across the MIKEsiteTheseGPSlocationsarerepresentativeofelephantdensityintheecosystemascharacterizedbyaerialsurveydatacollectedin20022008and2012and locationsofnaturalmortalitycol-lectedaspartoftheMIKEecosystemwidemonitoringprogram(Ihwagi etal 2015) This dataset can be considered spatiallyunbiased
23emsp|emspQuantifying use of crossing sections of the infrastructure corridor
Toidentifycriticalcrossingpointsalongtheproposedrailwayswedivided the LAPSSET corridor into consecutive 200m long seg-ments(n=3165)evaluatedwhetheranelephantusedthesegmentandtalliedthenumberofstepsforeachelephantthatcrossed(in-tersected)thesegmentWealsocalculatedastandardizedcrossingintensitymetricCforeachsegment(s)following
whereCs isthestandardizedcrossingintensityforsegments xis isthenumberof steps from individual i thatcrossedsegments ti isthe time overwhich the individualwasmonitored in days andns isthenumberofdifferentindividualscrossingsegmentsWetestedif thecalculationof thismetricwassensitive to thenumberof in-dividuals using the segment and found no correlation (R=minus036df=1065 p=239) This standardizedmetric accounted for vari-ation in sampling frequencyand lengthofmonitoringamong indi-vidualsandcouldpotentiallyaccountfordifferentialspatialsamplingalongthecorridor(iemoreindividualscapturedandfollowedinagivenarea)Wetestedthisbylookingathowthecalculationofthecrossingintensitymetricdifferedwhetherusingthefulldatasetorthesystematicallysampledsubset
Finally we evaluated how the standardized crossing intensitymetriccomparedtocrossingpoints identifiedbycommunitygamescouts along a 40km section of the proposed LAPSSET corridorThelattercrossingpointsbasedonlocalknowledgewereassessedbydrivingtheroadwithlocalcommunityscoutsinJune2017Wecompared the segments identified by community scouts as being
usedornottothestandardizedcrossingintensityusingattestwithunequalvariancesTotestiflocalknowledgemorecloselyreflectedrecentcrossingbyelephantsorlongertermpatternsweperformedtheanalysisusingthetotaldatasetincludingonlythelatestyearofdata(May2016ndashJune2017)
24emsp|emspOptimization of crossing structure locations
Wedeveloped and applied two Integer Linear Programming algo-rithmstooptimizethepositioningofwildlifecrossingstructuresThefirstalgorithmisbasedonthestandardizedcrossingintensitymetricandthespatialdistributionofcrossingstructurelocations(hereafterreferredas thecrossing intensity [CI] algorithm)TheCIalgorithmruns an optimization procedure where the importance of a seg-mentrsquos relativeposition (spatial spread fromothercrossingpoints)vstheimportanceofintensityofusecanbespecifiedbytheuserin order to enhance the biological importance of identified cross-inglocationsandinsureselectedsegmentsaresufficientlyfarfromeachother(iewherebothprocessesareweightedequally identi-fiedcrossingswouldbeofhighuseandrelativelydistributedacrossthestudyarea)Thealgorithmwasdevelopedtoselectthemostim-portantcrossingpointsforelephants(iethoselocationsthatweremostheavilytrafficked)
Thesecondalgorithmsolvesatypicalmaximumcoverageloca-tionproblem(referredhereaftertoldquoMLCPalgorithmrdquoDownsetal2014)wheretheselectedsegmentswillldquocoverrdquothegreatestnumberofdifferentindividualsgiveneachsegmentcanserveuptoathresh-olddistanceThisalgorithmdiffersfromtheformerinthatselectedsegmentsdonothavetobeusedheavilybutcanbeincloseproxim-itytomultipleheavilyusedsegments(ieproximitytopathsismoreimportantthantheactualmovementsthathappenedacrossagivensegment) In additionbothalgorithmscanadjust theoptimizationroutinebyincorporatingnaturalcrossings(egabridgeoverariverthat is effectively a wildlife crossing point) or crossings specifieda priori for humaneconomic reasons to adjust the final optimiza-tionsolutionFurtherdetailsabouteachalgorithmareprovided inAppendixS1
Asanillustrationofeachalgorithmandtocomparehowtheycandifferwecreateddifferentscenariosthatcombinedtherep-resentative and global datasetswith each algorithm For all sce-narios we identified five sites (ie development plans budgetedtobuildfivecrossings)FortheCIalgorithmwegaveequalweighttothespatialspreadandintensityofuseFortheMCLPalgorithma thresholddistancefor thecoverageof5kmwasselectedThismeansthatsegmentswithin5km(oneachside)ofastructureareassumedtobecoveredbythisstructure(ieweassumedelephantwouldbewillingtotravelupto5kmtoreachacrossingstructuregiventhisisonthelowerendofaveragedailydisplacementsamongtheindividualstracked)Weevaluateddifferencesamongscenar-ios by calculating the minimum average distance between eachselectedcrossing(ieanalogoustotheEarthmoverrsquosdistancesta-tisticPottsAuger-MeacutetheacuteampLewis2014)WealsoranadditionalscenariosusingeachalgorithmwhicharepresentedinFiguresS1
Cs=
sumns
i=1
xis
ti
ns
emspensp emsp | emsp5Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
andS2 inAppendixS2All functions fromouranalysis (segmen-tationofLFshome-rangeintersectioncrossingintensityandop-timization algorithms) and an explanatory vignette are availableasanrpackage ldquowildxingrdquo (httpsgithubcomBastilleRousseauwildxing)
3emsp |emspRESULTS
Upto33ofthesegmentswerecrossedbyindividualsinthetotalsample containing movements of 156 individual elephants while18 of the 200m segments analysed along the LAPSSET cor-ridor were crossed by the representative subset of 40 elephants(Figure2)As calculated from the total sample each segmentwascrossed by an average of 324 different individuals (range=1ndash22Figure2d)anaverageof637times(range=1ndash88)Usingtherepre-sentativesubsetdataseteachsegmentwascrossedbyanaverageof145different individuals (range=1ndash7 Figure2b) an averageof264times(range=1ndash53)Theaveragestandardizedcrossinginten-sitywas0023(range=0002ndash0632Figure2c)forthetotalsampleand0019(range=0007ndash0186Figure2a)forthesubsetOverallthestandardizedcrossingintensitymetricestimatedfromthesub-sampledatasetshowedmoderatecorrelationwiththemetriccalcu-latedfromthetotalsample(R=37df=2411plt001)butstronger
correlationwith the number of individuals using a given segment(R=57df=2411plt001)
Comparisonbetweenthecrossingsegmentsidentifiedthroughthe optimization routine indicates that strong differentiation incrossingweighting occurredwith the different emphases of thedifferentroutines(Figure3)UsingtheCIalgorithmselectedloca-tionsshowedgreaterdifferencebetweentherepresentativesub-set and the total sample (average distance=36km Figure3ab)thanwhenusing theMCLPalgorithm (averagedistance=16kmFigure3cd) Applying each algorithm to the same dataset alsoprovideddifferentresultswithaveragedistancebetweenselectedcrossingof24and30kmfortherepresentative(Figure3ac)andglobal(Figure3bd)datasetsrespectivelyOveralltheMCLPpro-vided connectivity for a greater number of elephants (Figure3)Modifyingparametersofeachscenarioalsoresultedinadifferentsetofselectedlocations(FiguresS1andS2)indicatingthesensi-tivityofresultstoweighting
Crossing segments identified using local knowledge were notsignificantly aligned with tracking data-based crossing intensitymetrics given that the segments considered as used based onlocalknowledgedidnothaveahigherstandardizedcrossinginten-sity (t=minus074668 p=4602 Figure4a) Alignment was also notinfluenced by the timing of the tracking data (latest year of datat=minus0994p=348Figure4b)
F IGURE 2emspSubsectionoftheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)Corridorproject(a)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingarepresentativesampleof40elephants(seeSection2)(b)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingtherepresentativesampleofelephants(c)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingallmonitoredindividuals(d)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingallmonitoredindividuals
00
01 ndash 10
11 ndash 20
20 ndash 41
41 ndash 70
0000000
0000001 ndash 0001505
0001506 ndash 0002471
0002472 ndash 0005652
0005653 ndash 0018624
(b)(a) Number of individualsCrossing intensity
(c) (d)Number of individualsCrossing intensity
0000000 ndash 0000899
0000900 ndash 0003076
0003077 ndash 0004052
0004053 ndash 0011295
0011296 ndash 0063250
00 ndash 10
11 ndash 40
41 ndash 70
71 ndash 140
141 ndash 220
6emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
4emsp |emspDISCUSSION
Mitigatingtheimpactsofinfrastructurecorridorsonwildlifeiscriti-caltoreduceecologicalimpacts(ClevengerampHuijser2011)toen-hancepublicsafety(OlssonampWiden2008)andalleviatetheneedforcostlyretrofittingprojectsforfuturemitigationneedsTofacili-tateconservationplanninginrelationtothedevelopmentofinfra-structure projectswe developed simple and efficient approachesto analytically identify priority crossing locations fromGPS track-ingdataSuchdataare increasinglycollectedgloballyandprovidethehighestspatio-temporaldataonwildlifeusethatcanbepower-fulforunderstandingandmitigatingimpactsfromhumanland-usechanges
Weappliedthesealgorithmstoanextensiveradiotrackingdata-setcollectedonelephantsinhabitingtheLaikipiaSamburuecosys-temsubjecttoamajorinfrastructuredevelopmentprojectAfricanelephantswereourtargetspeciesgiventhattheyareaflagshipandan umbrella species whose protection can have significant posi-tiveimpactsonotherspecies(EppsMutayobaGwinampBrashares2011)andecosystemfunctioning(Haynes2012)Giventhegener-allyopenunfencednatureofthisecosystemrecordedmovementsoccurred throughout the system and roughly a third of the corri-dorlengthintheecosystemwasintersectedbyrecordedelephantmovements (ie elephants crossed over the proposed route) Theundevelopednatureofthestudyareaaffordsarareopportunityto
identifycriticalareaslikelytobeimpactedbydevelopmentaprioribut also presents a challenge given thatwildlifemovement is notconstrained andmost areas in the region are used bywildlifeAssuchrestrictingfocustoahandfulofcrossinglocationsissensitivetothedefinitionofonersquoscriteria
41emsp|emspApplication of crossing algorithms
WedevelopedtwoalgorithmsthathighlightdifferentfeaturesofthemovementdataandultimatelyaddressdifferentobjectiveswhenanalysingthesedataThefirstapproachtheCIalgorithmissuitedforsituationswherethespecificcrossingintensityisofinterestalonga LF This is applicable to scenarioswhen the immediate crossinglocationistheunitofinterestforplanningsuchaswhenattempt-ingtomitigatevehicle-animalcollisionsorwhencrossingstructureswillhavethebiggestimpactsbybeingplacedincurrentlyusedloca-tionsascanoccurwhenaLFwillremainunfencedandorrepresentasemi-permeablebarrierThesecondapproachusingtheMCLPal-gorithm issuited fora limited-accessscenario (eg fencing)wherefewcrossingstructureswillbelocatedatregulardistancesandcon-nectivityrestrictedinallotherlocationsIntheMCLPapproachtheimportanceoftheselectedsegmentisnotstronglyweightedratherthe focus ison identifyingthesegment thatprovides thegreatestcoverageforindividualsthatgenerallyuseanareaWeassumethisalgorithmwouldbeidealforscenarioswherefencingisusedalong
F IGURE 3emspOptimizationofcrossingstructureslocationalongtheLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridorrailwaysusingthecrossingintensityandmaximumcoveragelocationproblemalgorithms(seeSection2)(a)Thecrossingintensity(CI)algorithmselectedthetopfivecrossingsegmentsbasedonequallyweightingtheinfluenceofspatialspreadandintensityofuseintheoptimizationandusingthesubsampleofdata(b)Thescenarioin(a)appliedtothetotaldataset(c)Themaximumcoveragelocationalgorithm(MCLP)selectedthetopfivecrossingsegmentsbasedonathresholddistanceof5kmandusingtherepresentativesampleofindividuals(d)Thescenarioin(c)appliedtothetotaldataset
G
GGGG
G GG G
G
(b)(a)
elephants = 5 elephants = 43
G
GG
G
G
GG G
G
G
(c) (d)
elephants = 85 elephants = 96
emspensp emsp | emsp7Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
LFstofunnel individualsintoagivencrossingpointGiventhedif-ferentnatureofthesetwoapproachesandthedivergingresultstheyprovideitiscriticalthattheuseridentifiesthetypeofconstraintsinthesystemanddesiredcrossingstructurecharacteristics
42emsp|emspRepresentativeness of GPS telemetry
Overall theselectionofcrossingstructure locationsandtheir im-pactsonelephantsvariedwhetherweusedthesubsampleortotalsampleofmonitoredindividualsThisisconcerningfortherobust-nessof results reliantonGPSdata tooptimize crossing locationsStrong spatial bias in monitoring has the potential to impact theoveralloptimizationevenifalgorithmresultsarenotclusterednearthecoreof thespatiallybiasedsampleWerecommendmanagerscarefullyevaluatethepresenceofpotentialbiasesintheirdataandsubsampleorweightindividualdatainmannerthatoffersarepre-sentativedatasetTherunningofdifferentscenariosasexemplifiedherecanhelpdecisionmakersunderstandthesensitivityofresultstoimposeassumptionsHoweverthewidevariationinresultsinthestudyareaislikelyrelatedtothegeneralandwidelydisperseduseofthisunfencedecosystemOthersystemswithmoredevelopmentlikelyhavegreaterrestrictionsonpossiblecrossingpointsalongaLFwhichmayrequireexcludingsegmentsfromtheoptimizationalgo-rithm(bothoptimizationfunctionsavailableinthewildxingpackageincludethisoption)
We also compared elephant crossing locations obtained fromsurveyswith local communities to those identifiedusingGPS te-lemetrydatatocontrastoutcomesthatresultfromdifferenttypesof data The locations of key crossing locations were not wellmatched across these two sourcesWhere tracking and commu-nity identifiedcrossing locationscoincided thecrossing intensityasmeasuredby the trackingdatawas relatively lowThiswas ir-respectiveofthetimingoftheGPStrackingdatausedtoidentifycrossingsCommunity identified crossingsweremore likely tobefoundnearvillagesorlivestockpathsWhileGPStelemetrysuffersfromitsownissues(discussedinthepreviousparagraph)ouranaly-sissuggeststhatcommunity-baseddataareimportantwhereothersources are not available but subject to biases related to humanactivities and ideally would require more systematic surveyingEmpiricallycollecteddataonanimalactivitiesmaybemorereliablewhenavailable
43emsp|emspAdditional considerations
Our algorithms allowmultiple factors andmotivations to be con-sidered in selecting crossing locations Our algorithms combiningintensityofuseandnumberof individuals representedarean im-provement over existing approaches Previous approaches typi-callysummarized thegeneralcrossingbehaviourhabitatselectionorsimply record intensityofusewithoutsubsequentoptimization(Horne etal 2007 Sawyer Kauffman Nielson amp Horne 2009SchusterRoumlmerampGermain2013)Otherapproachesproposedop-timizationalgorithmsbutwithoutregardtosamplingbiasandspa-tialspread(Downsetal2014LoraammampDowns2016)Intensityof usemay not be the onlymetric of valuewhenquantifying theimportance of a segment forwildlifemovementApproaches thatconsider the importanceofa segment for specific typesofmove-ment(egmovementmodesGurarieetal2016MoralesHaydon
F IGURE 4emspElephantstandardizedcrossingintensityestimatedfromGPStrackingdatawascomparedwithcrossingpointsidentifiedbylocalcommunityscouts(a)Communitiesidentifiedcrossinglocations(dots)andstandardizedcrossingintensityfor200msegmentsoftheroadusingGPStelemetrycollectedbetween2001and2017(b)CommunityscoutidentifiedcrossinglocationsandstandardizedcrossingmetricsdevelopedusingGPStelemetrydatacollectedJune2016ndashJune2017
0000000
0000001 ndash 0003753
0003754 ndash 0006340
0006341 ndash 0010170
0010171 ndash 0019253
(a)
(b)Last year
Full dataset
0000000 ndash 0000691
0000692 ndash 0002250
0002251 ndash 0003951
0003952 ndash 0006352
0006353 ndash 0054055
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
emspensp emsp | emsp5Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
andS2 inAppendixS2All functions fromouranalysis (segmen-tationofLFshome-rangeintersectioncrossingintensityandop-timization algorithms) and an explanatory vignette are availableasanrpackage ldquowildxingrdquo (httpsgithubcomBastilleRousseauwildxing)
3emsp |emspRESULTS
Upto33ofthesegmentswerecrossedbyindividualsinthetotalsample containing movements of 156 individual elephants while18 of the 200m segments analysed along the LAPSSET cor-ridor were crossed by the representative subset of 40 elephants(Figure2)As calculated from the total sample each segmentwascrossed by an average of 324 different individuals (range=1ndash22Figure2d)anaverageof637times(range=1ndash88)Usingtherepre-sentativesubsetdataseteachsegmentwascrossedbyanaverageof145different individuals (range=1ndash7 Figure2b) an averageof264times(range=1ndash53)Theaveragestandardizedcrossinginten-sitywas0023(range=0002ndash0632Figure2c)forthetotalsampleand0019(range=0007ndash0186Figure2a)forthesubsetOverallthestandardizedcrossingintensitymetricestimatedfromthesub-sampledatasetshowedmoderatecorrelationwiththemetriccalcu-latedfromthetotalsample(R=37df=2411plt001)butstronger
correlationwith the number of individuals using a given segment(R=57df=2411plt001)
Comparisonbetweenthecrossingsegmentsidentifiedthroughthe optimization routine indicates that strong differentiation incrossingweighting occurredwith the different emphases of thedifferentroutines(Figure3)UsingtheCIalgorithmselectedloca-tionsshowedgreaterdifferencebetweentherepresentativesub-set and the total sample (average distance=36km Figure3ab)thanwhenusing theMCLPalgorithm (averagedistance=16kmFigure3cd) Applying each algorithm to the same dataset alsoprovideddifferentresultswithaveragedistancebetweenselectedcrossingof24and30kmfortherepresentative(Figure3ac)andglobal(Figure3bd)datasetsrespectivelyOveralltheMCLPpro-vided connectivity for a greater number of elephants (Figure3)Modifyingparametersofeachscenarioalsoresultedinadifferentsetofselectedlocations(FiguresS1andS2)indicatingthesensi-tivityofresultstoweighting
Crossing segments identified using local knowledge were notsignificantly aligned with tracking data-based crossing intensitymetrics given that the segments considered as used based onlocalknowledgedidnothaveahigherstandardizedcrossinginten-sity (t=minus074668 p=4602 Figure4a) Alignment was also notinfluenced by the timing of the tracking data (latest year of datat=minus0994p=348Figure4b)
F IGURE 2emspSubsectionoftheproposedLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)Corridorproject(a)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingarepresentativesampleof40elephants(seeSection2)(b)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingtherepresentativesampleofelephants(c)StandardizedcrossingintensityofelephantwiththeLAPSSETcorridorusingallmonitoredindividuals(d)NumberofdifferentelephantscrossingeachsegmentoftheLAPSSETcorridorusingallmonitoredindividuals
00
01 ndash 10
11 ndash 20
20 ndash 41
41 ndash 70
0000000
0000001 ndash 0001505
0001506 ndash 0002471
0002472 ndash 0005652
0005653 ndash 0018624
(b)(a) Number of individualsCrossing intensity
(c) (d)Number of individualsCrossing intensity
0000000 ndash 0000899
0000900 ndash 0003076
0003077 ndash 0004052
0004053 ndash 0011295
0011296 ndash 0063250
00 ndash 10
11 ndash 40
41 ndash 70
71 ndash 140
141 ndash 220
6emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
4emsp |emspDISCUSSION
Mitigatingtheimpactsofinfrastructurecorridorsonwildlifeiscriti-caltoreduceecologicalimpacts(ClevengerampHuijser2011)toen-hancepublicsafety(OlssonampWiden2008)andalleviatetheneedforcostlyretrofittingprojectsforfuturemitigationneedsTofacili-tateconservationplanninginrelationtothedevelopmentofinfra-structure projectswe developed simple and efficient approachesto analytically identify priority crossing locations fromGPS track-ingdataSuchdataare increasinglycollectedgloballyandprovidethehighestspatio-temporaldataonwildlifeusethatcanbepower-fulforunderstandingandmitigatingimpactsfromhumanland-usechanges
Weappliedthesealgorithmstoanextensiveradiotrackingdata-setcollectedonelephantsinhabitingtheLaikipiaSamburuecosys-temsubjecttoamajorinfrastructuredevelopmentprojectAfricanelephantswereourtargetspeciesgiventhattheyareaflagshipandan umbrella species whose protection can have significant posi-tiveimpactsonotherspecies(EppsMutayobaGwinampBrashares2011)andecosystemfunctioning(Haynes2012)Giventhegener-allyopenunfencednatureofthisecosystemrecordedmovementsoccurred throughout the system and roughly a third of the corri-dorlengthintheecosystemwasintersectedbyrecordedelephantmovements (ie elephants crossed over the proposed route) Theundevelopednatureofthestudyareaaffordsarareopportunityto
identifycriticalareaslikelytobeimpactedbydevelopmentaprioribut also presents a challenge given thatwildlifemovement is notconstrained andmost areas in the region are used bywildlifeAssuchrestrictingfocustoahandfulofcrossinglocationsissensitivetothedefinitionofonersquoscriteria
41emsp|emspApplication of crossing algorithms
WedevelopedtwoalgorithmsthathighlightdifferentfeaturesofthemovementdataandultimatelyaddressdifferentobjectiveswhenanalysingthesedataThefirstapproachtheCIalgorithmissuitedforsituationswherethespecificcrossingintensityisofinterestalonga LF This is applicable to scenarioswhen the immediate crossinglocationistheunitofinterestforplanningsuchaswhenattempt-ingtomitigatevehicle-animalcollisionsorwhencrossingstructureswillhavethebiggestimpactsbybeingplacedincurrentlyusedloca-tionsascanoccurwhenaLFwillremainunfencedandorrepresentasemi-permeablebarrierThesecondapproachusingtheMCLPal-gorithm issuited fora limited-accessscenario (eg fencing)wherefewcrossingstructureswillbelocatedatregulardistancesandcon-nectivityrestrictedinallotherlocationsIntheMCLPapproachtheimportanceoftheselectedsegmentisnotstronglyweightedratherthe focus ison identifyingthesegment thatprovides thegreatestcoverageforindividualsthatgenerallyuseanareaWeassumethisalgorithmwouldbeidealforscenarioswherefencingisusedalong
F IGURE 3emspOptimizationofcrossingstructureslocationalongtheLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridorrailwaysusingthecrossingintensityandmaximumcoveragelocationproblemalgorithms(seeSection2)(a)Thecrossingintensity(CI)algorithmselectedthetopfivecrossingsegmentsbasedonequallyweightingtheinfluenceofspatialspreadandintensityofuseintheoptimizationandusingthesubsampleofdata(b)Thescenarioin(a)appliedtothetotaldataset(c)Themaximumcoveragelocationalgorithm(MCLP)selectedthetopfivecrossingsegmentsbasedonathresholddistanceof5kmandusingtherepresentativesampleofindividuals(d)Thescenarioin(c)appliedtothetotaldataset
G
GGGG
G GG G
G
(b)(a)
elephants = 5 elephants = 43
G
GG
G
G
GG G
G
G
(c) (d)
elephants = 85 elephants = 96
emspensp emsp | emsp7Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
LFstofunnel individualsintoagivencrossingpointGiventhedif-ferentnatureofthesetwoapproachesandthedivergingresultstheyprovideitiscriticalthattheuseridentifiesthetypeofconstraintsinthesystemanddesiredcrossingstructurecharacteristics
42emsp|emspRepresentativeness of GPS telemetry
Overall theselectionofcrossingstructure locationsandtheir im-pactsonelephantsvariedwhetherweusedthesubsampleortotalsampleofmonitoredindividualsThisisconcerningfortherobust-nessof results reliantonGPSdata tooptimize crossing locationsStrong spatial bias in monitoring has the potential to impact theoveralloptimizationevenifalgorithmresultsarenotclusterednearthecoreof thespatiallybiasedsampleWerecommendmanagerscarefullyevaluatethepresenceofpotentialbiasesintheirdataandsubsampleorweightindividualdatainmannerthatoffersarepre-sentativedatasetTherunningofdifferentscenariosasexemplifiedherecanhelpdecisionmakersunderstandthesensitivityofresultstoimposeassumptionsHoweverthewidevariationinresultsinthestudyareaislikelyrelatedtothegeneralandwidelydisperseduseofthisunfencedecosystemOthersystemswithmoredevelopmentlikelyhavegreaterrestrictionsonpossiblecrossingpointsalongaLFwhichmayrequireexcludingsegmentsfromtheoptimizationalgo-rithm(bothoptimizationfunctionsavailableinthewildxingpackageincludethisoption)
We also compared elephant crossing locations obtained fromsurveyswith local communities to those identifiedusingGPS te-lemetrydatatocontrastoutcomesthatresultfromdifferenttypesof data The locations of key crossing locations were not wellmatched across these two sourcesWhere tracking and commu-nity identifiedcrossing locationscoincided thecrossing intensityasmeasuredby the trackingdatawas relatively lowThiswas ir-respectiveofthetimingoftheGPStrackingdatausedtoidentifycrossingsCommunity identified crossingsweremore likely tobefoundnearvillagesorlivestockpathsWhileGPStelemetrysuffersfromitsownissues(discussedinthepreviousparagraph)ouranaly-sissuggeststhatcommunity-baseddataareimportantwhereothersources are not available but subject to biases related to humanactivities and ideally would require more systematic surveyingEmpiricallycollecteddataonanimalactivitiesmaybemorereliablewhenavailable
43emsp|emspAdditional considerations
Our algorithms allowmultiple factors andmotivations to be con-sidered in selecting crossing locations Our algorithms combiningintensityofuseandnumberof individuals representedarean im-provement over existing approaches Previous approaches typi-callysummarized thegeneralcrossingbehaviourhabitatselectionorsimply record intensityofusewithoutsubsequentoptimization(Horne etal 2007 Sawyer Kauffman Nielson amp Horne 2009SchusterRoumlmerampGermain2013)Otherapproachesproposedop-timizationalgorithmsbutwithoutregardtosamplingbiasandspa-tialspread(Downsetal2014LoraammampDowns2016)Intensityof usemay not be the onlymetric of valuewhenquantifying theimportance of a segment forwildlifemovementApproaches thatconsider the importanceofa segment for specific typesofmove-ment(egmovementmodesGurarieetal2016MoralesHaydon
F IGURE 4emspElephantstandardizedcrossingintensityestimatedfromGPStrackingdatawascomparedwithcrossingpointsidentifiedbylocalcommunityscouts(a)Communitiesidentifiedcrossinglocations(dots)andstandardizedcrossingintensityfor200msegmentsoftheroadusingGPStelemetrycollectedbetween2001and2017(b)CommunityscoutidentifiedcrossinglocationsandstandardizedcrossingmetricsdevelopedusingGPStelemetrydatacollectedJune2016ndashJune2017
0000000
0000001 ndash 0003753
0003754 ndash 0006340
0006341 ndash 0010170
0010171 ndash 0019253
(a)
(b)Last year
Full dataset
0000000 ndash 0000691
0000692 ndash 0002250
0002251 ndash 0003951
0003952 ndash 0006352
0006353 ndash 0054055
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
6emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
4emsp |emspDISCUSSION
Mitigatingtheimpactsofinfrastructurecorridorsonwildlifeiscriti-caltoreduceecologicalimpacts(ClevengerampHuijser2011)toen-hancepublicsafety(OlssonampWiden2008)andalleviatetheneedforcostlyretrofittingprojectsforfuturemitigationneedsTofacili-tateconservationplanninginrelationtothedevelopmentofinfra-structure projectswe developed simple and efficient approachesto analytically identify priority crossing locations fromGPS track-ingdataSuchdataare increasinglycollectedgloballyandprovidethehighestspatio-temporaldataonwildlifeusethatcanbepower-fulforunderstandingandmitigatingimpactsfromhumanland-usechanges
Weappliedthesealgorithmstoanextensiveradiotrackingdata-setcollectedonelephantsinhabitingtheLaikipiaSamburuecosys-temsubjecttoamajorinfrastructuredevelopmentprojectAfricanelephantswereourtargetspeciesgiventhattheyareaflagshipandan umbrella species whose protection can have significant posi-tiveimpactsonotherspecies(EppsMutayobaGwinampBrashares2011)andecosystemfunctioning(Haynes2012)Giventhegener-allyopenunfencednatureofthisecosystemrecordedmovementsoccurred throughout the system and roughly a third of the corri-dorlengthintheecosystemwasintersectedbyrecordedelephantmovements (ie elephants crossed over the proposed route) Theundevelopednatureofthestudyareaaffordsarareopportunityto
identifycriticalareaslikelytobeimpactedbydevelopmentaprioribut also presents a challenge given thatwildlifemovement is notconstrained andmost areas in the region are used bywildlifeAssuchrestrictingfocustoahandfulofcrossinglocationsissensitivetothedefinitionofonersquoscriteria
41emsp|emspApplication of crossing algorithms
WedevelopedtwoalgorithmsthathighlightdifferentfeaturesofthemovementdataandultimatelyaddressdifferentobjectiveswhenanalysingthesedataThefirstapproachtheCIalgorithmissuitedforsituationswherethespecificcrossingintensityisofinterestalonga LF This is applicable to scenarioswhen the immediate crossinglocationistheunitofinterestforplanningsuchaswhenattempt-ingtomitigatevehicle-animalcollisionsorwhencrossingstructureswillhavethebiggestimpactsbybeingplacedincurrentlyusedloca-tionsascanoccurwhenaLFwillremainunfencedandorrepresentasemi-permeablebarrierThesecondapproachusingtheMCLPal-gorithm issuited fora limited-accessscenario (eg fencing)wherefewcrossingstructureswillbelocatedatregulardistancesandcon-nectivityrestrictedinallotherlocationsIntheMCLPapproachtheimportanceoftheselectedsegmentisnotstronglyweightedratherthe focus ison identifyingthesegment thatprovides thegreatestcoverageforindividualsthatgenerallyuseanareaWeassumethisalgorithmwouldbeidealforscenarioswherefencingisusedalong
F IGURE 3emspOptimizationofcrossingstructureslocationalongtheLamuPort-SouthSudan-Ethiopia-Transport(LAPSSET)corridorrailwaysusingthecrossingintensityandmaximumcoveragelocationproblemalgorithms(seeSection2)(a)Thecrossingintensity(CI)algorithmselectedthetopfivecrossingsegmentsbasedonequallyweightingtheinfluenceofspatialspreadandintensityofuseintheoptimizationandusingthesubsampleofdata(b)Thescenarioin(a)appliedtothetotaldataset(c)Themaximumcoveragelocationalgorithm(MCLP)selectedthetopfivecrossingsegmentsbasedonathresholddistanceof5kmandusingtherepresentativesampleofindividuals(d)Thescenarioin(c)appliedtothetotaldataset
G
GGGG
G GG G
G
(b)(a)
elephants = 5 elephants = 43
G
GG
G
G
GG G
G
G
(c) (d)
elephants = 85 elephants = 96
emspensp emsp | emsp7Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
LFstofunnel individualsintoagivencrossingpointGiventhedif-ferentnatureofthesetwoapproachesandthedivergingresultstheyprovideitiscriticalthattheuseridentifiesthetypeofconstraintsinthesystemanddesiredcrossingstructurecharacteristics
42emsp|emspRepresentativeness of GPS telemetry
Overall theselectionofcrossingstructure locationsandtheir im-pactsonelephantsvariedwhetherweusedthesubsampleortotalsampleofmonitoredindividualsThisisconcerningfortherobust-nessof results reliantonGPSdata tooptimize crossing locationsStrong spatial bias in monitoring has the potential to impact theoveralloptimizationevenifalgorithmresultsarenotclusterednearthecoreof thespatiallybiasedsampleWerecommendmanagerscarefullyevaluatethepresenceofpotentialbiasesintheirdataandsubsampleorweightindividualdatainmannerthatoffersarepre-sentativedatasetTherunningofdifferentscenariosasexemplifiedherecanhelpdecisionmakersunderstandthesensitivityofresultstoimposeassumptionsHoweverthewidevariationinresultsinthestudyareaislikelyrelatedtothegeneralandwidelydisperseduseofthisunfencedecosystemOthersystemswithmoredevelopmentlikelyhavegreaterrestrictionsonpossiblecrossingpointsalongaLFwhichmayrequireexcludingsegmentsfromtheoptimizationalgo-rithm(bothoptimizationfunctionsavailableinthewildxingpackageincludethisoption)
We also compared elephant crossing locations obtained fromsurveyswith local communities to those identifiedusingGPS te-lemetrydatatocontrastoutcomesthatresultfromdifferenttypesof data The locations of key crossing locations were not wellmatched across these two sourcesWhere tracking and commu-nity identifiedcrossing locationscoincided thecrossing intensityasmeasuredby the trackingdatawas relatively lowThiswas ir-respectiveofthetimingoftheGPStrackingdatausedtoidentifycrossingsCommunity identified crossingsweremore likely tobefoundnearvillagesorlivestockpathsWhileGPStelemetrysuffersfromitsownissues(discussedinthepreviousparagraph)ouranaly-sissuggeststhatcommunity-baseddataareimportantwhereothersources are not available but subject to biases related to humanactivities and ideally would require more systematic surveyingEmpiricallycollecteddataonanimalactivitiesmaybemorereliablewhenavailable
43emsp|emspAdditional considerations
Our algorithms allowmultiple factors andmotivations to be con-sidered in selecting crossing locations Our algorithms combiningintensityofuseandnumberof individuals representedarean im-provement over existing approaches Previous approaches typi-callysummarized thegeneralcrossingbehaviourhabitatselectionorsimply record intensityofusewithoutsubsequentoptimization(Horne etal 2007 Sawyer Kauffman Nielson amp Horne 2009SchusterRoumlmerampGermain2013)Otherapproachesproposedop-timizationalgorithmsbutwithoutregardtosamplingbiasandspa-tialspread(Downsetal2014LoraammampDowns2016)Intensityof usemay not be the onlymetric of valuewhenquantifying theimportance of a segment forwildlifemovementApproaches thatconsider the importanceofa segment for specific typesofmove-ment(egmovementmodesGurarieetal2016MoralesHaydon
F IGURE 4emspElephantstandardizedcrossingintensityestimatedfromGPStrackingdatawascomparedwithcrossingpointsidentifiedbylocalcommunityscouts(a)Communitiesidentifiedcrossinglocations(dots)andstandardizedcrossingintensityfor200msegmentsoftheroadusingGPStelemetrycollectedbetween2001and2017(b)CommunityscoutidentifiedcrossinglocationsandstandardizedcrossingmetricsdevelopedusingGPStelemetrydatacollectedJune2016ndashJune2017
0000000
0000001 ndash 0003753
0003754 ndash 0006340
0006341 ndash 0010170
0010171 ndash 0019253
(a)
(b)Last year
Full dataset
0000000 ndash 0000691
0000692 ndash 0002250
0002251 ndash 0003951
0003952 ndash 0006352
0006353 ndash 0054055
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
emspensp emsp | emsp7Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
LFstofunnel individualsintoagivencrossingpointGiventhedif-ferentnatureofthesetwoapproachesandthedivergingresultstheyprovideitiscriticalthattheuseridentifiesthetypeofconstraintsinthesystemanddesiredcrossingstructurecharacteristics
42emsp|emspRepresentativeness of GPS telemetry
Overall theselectionofcrossingstructure locationsandtheir im-pactsonelephantsvariedwhetherweusedthesubsampleortotalsampleofmonitoredindividualsThisisconcerningfortherobust-nessof results reliantonGPSdata tooptimize crossing locationsStrong spatial bias in monitoring has the potential to impact theoveralloptimizationevenifalgorithmresultsarenotclusterednearthecoreof thespatiallybiasedsampleWerecommendmanagerscarefullyevaluatethepresenceofpotentialbiasesintheirdataandsubsampleorweightindividualdatainmannerthatoffersarepre-sentativedatasetTherunningofdifferentscenariosasexemplifiedherecanhelpdecisionmakersunderstandthesensitivityofresultstoimposeassumptionsHoweverthewidevariationinresultsinthestudyareaislikelyrelatedtothegeneralandwidelydisperseduseofthisunfencedecosystemOthersystemswithmoredevelopmentlikelyhavegreaterrestrictionsonpossiblecrossingpointsalongaLFwhichmayrequireexcludingsegmentsfromtheoptimizationalgo-rithm(bothoptimizationfunctionsavailableinthewildxingpackageincludethisoption)
We also compared elephant crossing locations obtained fromsurveyswith local communities to those identifiedusingGPS te-lemetrydatatocontrastoutcomesthatresultfromdifferenttypesof data The locations of key crossing locations were not wellmatched across these two sourcesWhere tracking and commu-nity identifiedcrossing locationscoincided thecrossing intensityasmeasuredby the trackingdatawas relatively lowThiswas ir-respectiveofthetimingoftheGPStrackingdatausedtoidentifycrossingsCommunity identified crossingsweremore likely tobefoundnearvillagesorlivestockpathsWhileGPStelemetrysuffersfromitsownissues(discussedinthepreviousparagraph)ouranaly-sissuggeststhatcommunity-baseddataareimportantwhereothersources are not available but subject to biases related to humanactivities and ideally would require more systematic surveyingEmpiricallycollecteddataonanimalactivitiesmaybemorereliablewhenavailable
43emsp|emspAdditional considerations
Our algorithms allowmultiple factors andmotivations to be con-sidered in selecting crossing locations Our algorithms combiningintensityofuseandnumberof individuals representedarean im-provement over existing approaches Previous approaches typi-callysummarized thegeneralcrossingbehaviourhabitatselectionorsimply record intensityofusewithoutsubsequentoptimization(Horne etal 2007 Sawyer Kauffman Nielson amp Horne 2009SchusterRoumlmerampGermain2013)Otherapproachesproposedop-timizationalgorithmsbutwithoutregardtosamplingbiasandspa-tialspread(Downsetal2014LoraammampDowns2016)Intensityof usemay not be the onlymetric of valuewhenquantifying theimportance of a segment forwildlifemovementApproaches thatconsider the importanceofa segment for specific typesofmove-ment(egmovementmodesGurarieetal2016MoralesHaydon
F IGURE 4emspElephantstandardizedcrossingintensityestimatedfromGPStrackingdatawascomparedwithcrossingpointsidentifiedbylocalcommunityscouts(a)Communitiesidentifiedcrossinglocations(dots)andstandardizedcrossingintensityfor200msegmentsoftheroadusingGPStelemetrycollectedbetween2001and2017(b)CommunityscoutidentifiedcrossinglocationsandstandardizedcrossingmetricsdevelopedusingGPStelemetrydatacollectedJune2016ndashJune2017
0000000
0000001 ndash 0003753
0003754 ndash 0006340
0006341 ndash 0010170
0010171 ndash 0019253
(a)
(b)Last year
Full dataset
0000000 ndash 0000691
0000692 ndash 0002250
0002251 ndash 0003951
0003952 ndash 0006352
0006353 ndash 0054055
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
8emsp |emsp emspenspJournal of Applied Ecology BASTILLE- ROUSSEAU ET AL
FrairHolsingerampFryxell2004)orprovideinsighttobroaderscaleconnectivityofalocation(egnetworkmetricsBastille-RousseauDouglas-HamiltonBlakeNorthrupampWittemyer2018WittemyerKeating Vollrath amp Douglas-Hamilton 2017) could providemorerefinedidentificationofcrossinglocationsinprioritizationschemesLikewise approaches assessing connectivity for multiple speciesmaybepreferablewherecommunityormultispeciesconservationgoalsareparamount(MimetClauzelampFoltecircte2016)Wefocusonasinglespeciesinthisstudygiventhelackofcomparabledatafromotherspecies
The increased pace of land-use change and ecosystem deg-radation poses serious threats to wildlife (Cardinale etal 2012)demandingnovelmitigationapproachesSolutionstomitigateland-usechangearechronicallyunderfunded (LindseyBalmeFunstonHenschel ampHunter 2016Miller 2014) requiring smart planningto maximize impact from limited resources (Kiesecker CopelandPocewicz amp McKenney 2010 Mandle etal 2016) Decades oflargemammal research have generated valuable data that can beapplied to such tasks High-resolution data combinedwith objec-tiveprioritizationmethodsallowreasonedplanningactionsbutareoftenlackingduringcriticalinfrastructureplanningstagesInsteadthesedecisionsarefrequentlymadesubjectivelywhichcanresultinsuboptimaluseoffundsdirectedtowardsconservation(KareivaGrovesampMarvier2014)Giventhelimitedbudgetalreadyallocatedtomitigationmeasuresinmostproposeddevelopmentstoolsthatfacilitatespatialplanningareofhighvalue
ACKNOWLEDG EMENTS
ElephantmovementdatacamefromtheSavetheElephantsTrackingAnimals forConservationProgramGBRwassupportedbySavetheElephants TheNatureConservancy and theNatural SciencesandEngineeringResearchCouncilofCanadaAnneMTrainordigi-tizedtheLAPSSETcorridorWethankCharlotteFNarrandAnneMTrainorforearlycommentsonthismanuscript
AUTHORSrsquo CONTRIBUTIONS
GW and JW initially thought of the crossing count approachGBRdevelopedthestandardizationtheoptimizationalgorithmand the analytical framework (wildxing package) GBR per-formedtheanalysesIDHandGWcollectedtheelephantdataGBR ledthewritingofthemanuscriptwithcontributionsfromGWAllauthorshavecommentedandapprovedthefinalversionofthemanuscript
DATA ACCE SSIBILIT Y
Elephanttrackingdatahavenotbeenarchivedgiventheirhighlysen-sitivenatureInterestedreaderscancontactthecorrespondingau-thordirectlyforinquiriesFunctionsandassociateddocumentation(whichincludeexamples)areavailablewithintherpackagewildxinghttpsdoiorg105281zenodo1158705(Bastille-Rousseau(2018)
ORCID
Guillaume Bastille-Rousseau httporcidorg0000-0001-6799-639X
R E FE R E N C E S
Bastille-Rousseau G (2018) First release of wildxing Zenodo Digital Repositoryhttpsdoiorg105281zenodo1158705
Bastille-RousseauGDouglas-HamiltonIBlakeSNorthrupJMampWittemyerG(2018)Applyingnetworktheorytoanimalmovementsto identify properties of landscape space Ecological Applicationshttpsdoiorg101002eap1697
Beyer H L Gurarie E Boumlrger L Panzacchi M Basille MHerfindal I hellip Matthiopoulos J (2016) ldquoYou shall not passrdquoQuantifying barrier permeability and proximity avoidanceby animals Journal of Animal Ecology 85 43ndash53 httpsdoiorg1011111365-265612275
BrookBWSodhiNSampBradshawC J (2008) SynergiesamongextinctiondriversunderglobalchangeTrends in Ecology amp Evolution23453ndash460httpsdoiorg101016jtree200803011
Cardinale B JDuffy J EGonzalezAHooperDU Perrings CVenail PhellipNaeem S (2012) Biodiversity loss and its impact onhumanityNature489326httpsdoiorg101038nature11373
ClevengerAPampHuijserMP(2011)Wildlifecrossingstructurehand-bookDesignandevaluationinNorthAmerica
DownsJHornerMLoraammRAndersonJKimHampOnoratoD(2014)Strategically locatingwildlifecrossingstructuresforFloridapanthersusingmaximalcoveringapproachesTransactions in GIS1846ndash65httpsdoiorg101111tgis12005
EftestoslashlSTsegayeDHerfindal IFlydalKampColmanJE(2014)Measuring effects of linear obstacles on wildlife movementsAccounting for the relationship between step length and cross-ing probability European Journal of Wildlife Research 60 271ndash278httpsdoiorg101007s10344-013-0779-7
Epps CWMutayoba BM Gwin L amp Brashares J S (2011) Anempirical evaluation of theAfrican elephant as a focal species forconnectivityplanning inEastAfricaDiversity and Distributions17603ndash612httpsdoiorg101111j1472-4642201100773x
FahrigLampRytwinskiT(2009)EffectsofroadsonanimalabundanceAn empirical review and synthesis Ecology and Society 14 art21httpsdoiorg105751ES-02815-140121
FrairJLMerrillEHVisscherDRFortinDBeyerHLampMoralesJM(2005)Scalesofmovementbyelk(Cervuselaphus)inresponsetoheterogeneity inforageresourcesandpredationriskLandscape Ecology20273ndash287httpsdoiorg101007s10980-005-2075-8
GillJANorrisKGillJampSutherlandWJ(2001)Whybehaviouralresponsesmaynot reflect thepopulation consequencesofhumandisturbance Biological Conservation 97 265ndash268 httpsdoiorg101016S0006-3207(00)00002-1
GurarieEBracisCDelgadoMMeckleyTDKojolaIampWagnerCM(2016)WhatistheanimaldoingToolsforexploringbehavioralstructureinanimalmovementsJournal of Animal Ecology8569ndash84httpsdoiorg1011111365-265612379
HaynesG(2012)Elephants(andextinctrelatives)asearth-moversandecosystemengineersGeomorphology157ndash15899ndash107httpsdoiorg101016jgeomorph201104045
HorneJSGartonEOKroneSMampLewisJS(2007)AnalyzinganimalmovementsusingBrownianbridgesEcology882354ndash2363httpsdoiorg10189006-09571
IhwagiFWWangTWittemyerGSkidmoreAKToxopeusAGNgene S Douglas-Hamilton I (2015)Using poaching levels andelephant distribution to assess the conservation efficacy of privatecommunalandgovernmentlandinnorthernKenyaPLoS ONE101ndash17
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117
emspensp emsp | emsp9Journal of Applied EcologyBASTILLE- ROUSSEAU ET AL
JamesARCampStuart-SmithAK(2000)DistributionofcaribouandwolvesinrelationtolinearcorridorsJournal of Wildlife Management64154ndash159httpsdoiorg1023073802985
Kareiva P Groves C amp Marvier M (2014) The evolving linkagebetween conservation science and practice at the nature con-servancy Journal of Applied Ecology 51 1137ndash1147 httpsdoiorg1011111365-266412259
Kiesecker J M Copeland H Pocewicz A ampMcKenney B (2010)DevelopmentbydesignBlendinglandscapelevelplanningwiththemitigationhierarchyFrontiers in Ecology and the Environment8261ndash266httpsdoiorg101890090005
LauranceWFSloanSWengLampSayerJA(2015)Estimatingtheenvironmental costs of Africarsquos massive ldquodevelopment CorridorsrdquoCurrent Biology 25 3202ndash3208 httpsdoiorg101016jcub201510046
LindseyPABalmeGAFunstonPJHenschelPHampHunterLTB(2016)LifeafterCecilChannellingglobaloutrageintofundingforconservationinAfricaConservation Letters9296ndash301httpsdoiorg101111conl12224
LoraammRWampDownsJA(2016)Awildlifemovementapproachtooptimally locatewildlifecrossingstructures International Journal of Geographical Information Science3074ndash88httpsdoiorg1010801365881620151083995
MandleLBryantBPRuckelshausMGenelettiDKieseckerJMampPfaffA (2016)Entrypointsforconsideringecosystemserviceswithin infrastructureplanningHowto integrateconservationwithdevelopmentinordertoaidthembothConservation Letters9221ndash227httpsdoiorg101111conl12201
MillerDC(2014)ExplainingglobalpatternsofinternationalaidforlinkedbiodiversityconservationanddevelopmentWorld Development59341ndash359httpsdoiorg101016jworlddev201401004
Mimet A Clauzel C amp Foltecircte J C (2016) Locatingwildlife cross-ings for multispecies connectivity across linear infrastructuresLandscape Ecology 31 1955ndash1973 httpsdoiorg101007s10980-016-0373-y
MoralesJMHaydonDTFrairJHolsingerKEampFryxell JM(2004) Extracting more out of relocation data Building move-mentmodelsasmixturesofrandomwalksEcology852436ndash2445httpsdoiorg10189003-0269
Neumann W Ericsson G Dettki H Bunnefeld N Keuler N SHelmers D P amp Radeloff V C (2012) Difference in spatiotem-poral patterns of wildlife road-crossings and wildlife-vehicle colli-sionsBiological Conservation14570ndash78httpsdoiorg101016jbiocon201110011
Northrup JMPitt JMuhlyTB StenhouseGBMusianiMampBoyceMS(2012)Vehicletrafficshapesgrizzlybearbehaviouronamultiple-uselandscapeJournal of Applied Ecology491159ndash1167httpsdoiorg101111j1365-2664201202180x
OlssonM PO ampWiden P (2008) Effects of highway fencing andwildlife crossingsonmooseAlcesalcesmovementsand spaceusein southwestern SwedenWildlife Biology14 111ndash117 httpsdoiorg1029810909-6396(2008)14[111EOHFAW]20CO2
Potts J Auger-MeacutetheacuteMamp LewisM (2014)A generalized residualtechnique for analyzing complex movement models using earth moverrsquos distanceMethods in Ecology and Evolution 5 1012ndash1022httpsdoiorg1011112041-210X12253
RaizmanEARasmussenHBKingLEIhwagiFWampDouglas-Hamilton I (2013)Feasibility studyon the spatial and temporalmovementofSamburursquoscattleandwildlifeinKenyausingGPSradio-tracking remote sensing andGISPreventive Veterinary Medicine11176ndash80httpsdoiorg101016jprevetmed201304007
Rico A Kindlmann P amp Sedlaacuteček F (2009) Can the barrier ef-fect of highways cause genetic subdivision in small mammalsActa Theriologica 54 297ndash310 httpsdoiorg104098jat0001-70510682008
Sawyer H Kauffman M J Nielson R M amp Horne J S (2009)Identifyingandprioritizingungulatemigrationroutesforlandscape-level conservation Ecological Applications19 2016ndash2025 httpsdoiorg10189008-20341
SchusterRRoumlmerHampGermainRR(2013)Usingmulti-scaledistri-butionandmovementeffectsalongamontanehighwaytoidentifyoptimal crossing locations for a large-bodiedmammal communityPeerJ1e189httpsdoiorg107717peerj189
SmallMFampHunterML(1988)ForestFragmentationandaviannestpredation in forest landscapes Oecologia 76 62ndash64 httpsdoiorg101007BF00379601
WallJWittemyerGKlinkenbergBLeMayVampDouglas-HamiltonI (2013) Characterizing properties and drivers of long distancemovements by elephants (Loxodonta africana) in the GourmaMali Biological Conservation 157 60ndash68 httpsdoiorg101016 jbiocon201207019
WengLBoedhihartonoAKDirksPHGMDixonJLubisMIampSayerJA(2013)Mineralindustriesgrowthcorridorsandagricul-turaldevelopmentinAfricaGlobal Food Security2195ndash202httpsdoiorg101016jgfs201307003
WhittingtonJHebblewhiteMDeCesareNJNeufeldLBradleyM Wilmshurst J amp Musiani M (2011) Caribou encounterswith wolves increase near roads and trails A time-to-event ap-proach Journal of Applied Ecology 48 1535ndash1542 httpsdoiorg101111j1365-2664201102043x
WittemyerGDaballenDampDouglas-HamiltonI(2013)Comparativedemographyofanat-riskAfricanelephantpopulationPLoS ONE8e53726httpsdoiorg101371journalpone0053726
WittemyerGDouglas-Hamilton IampGetzWM (2005)Thesocio-ecologyofelephantsAnalysisoftheprocessescreatingmultitieredsocial structures Animal Behaviour 69 1357ndash1371 httpsdoiorg101016janbehav200408018
WittemyerGGetzWMVollrathFampDouglas-HamiltonI(2007)Social dominance seasonal movements and spatial segregationin African elephants A contribution to conservation behaviorBehavioral Ecology and Sociobiology 61 1919ndash1931 httpsdoiorg101007s00265-007-0432-0
WittemyerGKeatingLMVollrathFampDouglas-HamiltonI(2017)Graphtheoryillustratesspatialandtemporalfeaturesthatstructureelephant rest locations and reflect risk perception Ecography 40598ndash605
Wittemyer G Northrup J M Blanc J Douglas-Hamilton I Omondi P amp Burnham K P (2014) Illegal killing for ivory drives global decline in African elephants Proceedings of the National Academy of Sciences of the United States of America 111 13117ndash13121 httpsdoiorg101073pnas 1403984111
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle
How to cite this articleBastille-RousseauGWallJDouglas-HamiltonIWittemyerGOptimizingthepositioningofwildlifecrossingstructuresusingGPStelemetryJ Appl Ecol 2018001ndash9 httpsdoiorg1011111365-266413117