Landscape EcolDOI 10.1007/s10980-015-0292-3RESEARCH ARTICLEA multi-scale assessment of population connectivityin African lions (Panthera leo) in response to landscapechangeSamuel A. Cushman . Nicholas B. Elliot . David W. Macdonald .Andrew J. LoveridgeReceived: 8 May 2015 / Accepted: 28 September 2015Ó Springer Science Business Media Dordrecht 2015AbstractContext Habitat loss and fragmentation are amongthe major drivers of population declines and extinction, particularly in large carnivores. Connectivitymodels provide practical tools for assessing fragmentation effects and developing mitigation or conservation responses. To be useful to conservationpractitioners, connectivity models need to incorporatemultiple scales and include realistic scenarios basedon potential changes to habitat and anthropogenicpressures. This will help to prioritize conservationefforts in a changing landscape.Objectives The goal of our paper was to evaluatedifferences in population connectivity for lionsSpecial issue: Multi-scale habitat modeling.Guest Editor: K. McGargial and S. A. Cushman.Electronic supplementary material The online version ofthis article (doi:10.1007/s10980-015-0292-3) contains supplementary material, which is available to authorized users.S. A. CushmanUSDA Forest Service, Rocky Mountain Research Station,2500 S Pine Knoll Dr., Flagstaff, AZ 86001, USAN. B. Elliot D. W. Macdonald (&) A. J. LoveridgeWildlife Conservation Research Unit, Department ofZoology, Recanati-Kaplan Centre, University of Oxford,Tubney House, Abingdon Road, Tubney,Oxfordshire OX13 5QL, UKe-mail: [email protected](Panthera leo) across the Kavango-Zambezi Transfrontier Conservation Area (KAZA) under differentlandscape change scenarios and a range of dispersaldistances.Methods We used an empirically optimized resistance surface, based on analysis of movement pathways of dispersing lions in southern Africa to calculateresistant kernel connectivity. We assessed changes inconnectivity across nine landscape change scenarios,under each of which we explored the behavior of lionswith eight different dispersal abilities.Results Our results demonstrate that reductions inthe extent of the protected area network and/or fencingprotected areas will result in large declines in theextent of population connectivity, across all modeleddispersal abilities. Creation of corridors or erection offences strategically placed to funnel dispersersbetween protected areas increased overall connectivity of the population.Conclusions Our results strongly suggest that themost effective means of maintaining long-termpopulation connectivity of lions in the KAZAregion involves retaining the current protected areanetwork, augmented with protected corridors orstrategic fencing to direct dispersing individualstowards suitable habitat and away from potentialconflict areas.Keywords Connectivity Fragmentation FRAGSTATS Resistant kernel Scale UNICOR123

Landscape EcolIntroductionHabitat loss frequently results in small, isolatedpopulations, which have increased vulnerability tolocal extinctions due to environmental and demographic stochasticity (Winterbach et al. 2013). Inaddition, subsequent loss of genetic diversity mayincrease disease susceptibility (e.g., Trinkel et al.2011) and decrease reproductive success (e.g.,Packer et al. 1991). Management actions thatfacilitate the movement of individuals betweenisolated populations have the potential to increaseeffective population size, thereby reducing the risksassociated with inbreeding depression (Schwartz andMills 2005). In order to aid this decision-makingprocess, there has been a proliferation in connectivity models (for reviews see Sawyer et al. 2011;Zeller et al. 2012; Cushman et al. 2013) publishedwhich assess functional connectivity given a particular population size, dispersal ability and landscaperesistance pattern.Given that connectivity is scale, species and systemdependent (Cushman 2006), it is difficult to reliablypredict population connectivity (Rudnick et al. 2012;Cushman et al. 2013). For a particular species in aparticular system, population connectivity is the resultof the combined effects of the distribution and densityof the population, composition and configuration ofthe landscape, and species specific dispersal characteristics including sex and age differences, effects ofdifferent landscape features on movement, and howthese combine to shape the dispersal kernel. In mostpopulations, however, there is substantial uncertaintyabout species distributions and densities (Cushman2006), how different landscape features affect movement (Zeller et al. 2012), and limited understanding ofspecies dispersal abilities (Elliot et al. 2014a). Theseuncertainties combine to limit the reliability ofconnectivity modeling results. In a few cases (e.g.,Elliot et al. 2014b) these parameters have beenincluded in the development of resistance surfaces,which are the foundation of most contemporarymethods for predicting connectivity (Zeller et al.2012).In order to develop effective conservation strategiesto reduce habitat loss and fragmentation it is essentialto assess how population connectivity varies according to landscape change and the dispersal ability of thefocal species. Throughout Africa, landscape change is123occurring at a rapid and accelerating rate and isimpacting the amount of land set aside for wildlife.Human and livestock populations are growing, coinciding with an increased demand for land. In addition,African economies are expanding and land that iscurrently set aside for wildlife may be perceived asbeing more profitable if converted to agriculture ormining. These factors, combined with recent nationwide and partial bans on trophy hunting in Botswanaand Zambia respectively, have led to concerns thatland which is not officially protected by governmentsmay be converted into land uses that exclude wildlife.Growing human and livestock populations,together with an increased demand for land, hasresulted in increased human-wildlife conflict (Woodroffe and Frank 2005). In the case of the AfricanLion (Panthera leo), human-wildlife conflict andhabitat loss are thought among the primary driversof their recent declines (Bauer et al. 2012). Inaddition, edge effects have significant negativeimpacts on protected lion populations (Loveridgeet al. 2010). Packer et al. (2013a) conducted ameta-analysis across 42 lion populations and relatedlion density to management practices. They demonstrated that in fenced reserves lions are closer totheir carrying capacities and cost less to manageand protect compared to unfenced reserves in whichlion populations occur at lower densities relative totheir potential densities and are more likely todecline to extinction. This led the authors toconclude that fencing and other measures to mitigate edge effects are highly effective managementstrategies for conserving lions. Nevertheless extensive fencing of lion populations would drasticallyincrease population fragmentation and eliminateexisting connectivity and dispersal between subpopulations, leaving lions, and other species, ingenetic isolation. It is therefore imperative that theeffects of landscape change and potential management solutions are empirically tested with data fromdispersing individuals so that sound decisions canbe made.African lion populations have suffered an estimated75 % range reduction in the last 100 years (Riggioet al. 2013) and are increasingly divided into small andisolated remnant populations within protected areas.Bjorklund (2003) used a population genetics model toshow that a minimum of 50–100 prides, with no limits

Landscape Ecolto dispersal, is required to maintain long-term geneticdiversity. Very few remnant populations provide nearthis number of prides and some populations havesuffered declines in genetic diversity, which has beenshown to decrease reproductive performance (Packeret al. 1991) and increase susceptibility to disease(Trinkel et al. 2011). Thus, given the insufficient sizeof many extant protected areas to provide sufficientpopulation size for long term viability, dispersalamong remnant populations may be critical to maintain long-term genetic health and provide demographic rescue of regional lion populations.However, the degree to which remnant lion populations are functionally isolated and the factors that mayfacilitate gene flow between them through dispersalare largely unknown.In this paper we used a resistance surface that wasempirically optimized based on analysis of the movement pathways of dispersing lions in southern Africa(Elliot et al. 2014b) to calculate resistant kernelconnectivity (Compton et al. 2007) across theKavango-Zambezi Trans-frontier Conservation Area(KAZA) for each of nine landscape change scenariosand eight lion dispersal abilities. Our work wasdesigned to evaluate five hypotheses relating to theeffects of landscape change on population connectivity for lions in southern Africa.Hypothesis 1 Loss of protected area status forWMAs would result in very large decreases inpopulation connectivity across all dispersal abilities.Hypothesis 2 Extensive fencing of protected areaswould likewise lead to very large increases in population subdivision and isolation.Hypothesis 3 Designation and protection of corridors or fences strategically placed to funnel dispersingindividuals between protected habitats would increasepopulation connectivity.Hypothesis 4 An increased human populationwould result in large decreases in population connectivity across all dispersal abilities.Hypothesis 5 The effects of all landscape changescenarios would be scale-dependent, with largereffects on connectivity when lions disperse overshorter distances and smaller relative effects whendispersal is less geographically limited.Materials and methodsStudy extentThe study extent (&1.5 million km2) encompasses theentire Kavango Zambezi Transfrontier ConservationArea (KAZA-TFCA) and traverses sections ofAngola, Botswana, Namibia, Zambia and Zimbabwe(Fig. 1). Approximately 31 % (&458,520 km2) of thestudy extent is managed for wildlife, including 26national parks (145,570 km2), 297 Forest Reserves(52,776 km2) and 117 wildlife management areas(WMAs) (260,273 km2). National parks, forestreserves and WMAs are all gazetted wildlife areas.This extensive area is of great conservation importance for lions as it contains 13 ‘Lion ConservationUnits’ (IUCN 2006) and the Okavango-Hwangeecosystem is one of Africa’s 10 remaining lion‘strongholds’ (Riggio et al. 2013).Resistance surfaceWe used a resistance surface that was empiricallyoptimized by Elliot et al. (2014b) who collectedGlobal Positioning System data over 10 years from50 African lions Panthera leo (11 male nataldispersers, 20 adult males and 19 adult females)and used a path level analysis to parameterizedemographic-specific resistance surfaces for theKavango Zambezi Transfrontier Conservation Area(KAZA) in Southern Africa. Elliot et al. (2014b)used path-level (e.g., Cushman et al. 2010a, b;Cushman and Lewis 2010) randomization and multiscale mixed-effects conditional logistic regression topredict landscape resistance to dispersal as functionsof land use, land cover, human population densityand roads. As such the analysis was multi-scale,single level. The analysis was multi-scale in spacebut not time, since the shift was a spatial shift ofpaths that were constant in time (30 day longmovement segments). The scales that were evaluated included shifts of 0, 12.5, 25, and 50 km (5scales), and each variable was evaluated across thefive scales in a univariate conditional logisticregression, with the scale with the lowest AIC valuechosen for inclusion in multivariate models.The disperser data used in Elliot et al. (2014b) wasparsed into 997 segments corresponding to 30 day123

Landscape EcolFig. 1 Study areaorientation mapmovement periods. The average displacement acrossthose 30 day segments was 11.7 km with standarddeviation of 13.6 km. Based on a normal distributionwe would expect 99.7 % of path segments have lessthan 50 km displacement, 97.8 % to have less than25 km displacement, and 52 % to have less than12.5 km displacement. Thus the range of scalesconsidered spans the range of available habitat thatcould be reached by a lion in a 30 day movement boutand spans from relatively small extent of availablehabitat to relatively large, given the movement abilityof the species.Elliot et al. (2014b) found that lion path selectionvaried according to demographic grouping: adultfemales were most averse to risky landscapes suchas agro-pastoral lands, towns, areas of high humandensity and highways. Male natal dispersers were theleast-risk averse suggesting they are potentially themost prone demographic to human–lion conflict.Adults of both sexes selected bushed grassland andshrubland habitats and avoided woodland. Male nataldispersers displayed the opposite trend suggestingcon-specific avoidance and/or suboptimal habitat use.We used the resistance map for dispersing sub-adultmale lions in this paper, since genetic exchange amongpopulations of lions is primarily mediated by123dispersal. The resistance surface was created with a500 m cell size, with 2384 rows and 2618 columns,corresponding to an extent of 1,192 km by 1,309 km,or 1.56 million square kilometers.Landscape change scenariosOur analysis evaluated nine different scenarios ofchanging landscape resistance, which we divided intofive groups (Table 1; Fig. 2).The first group of landscape change scenariosinvolved conversion and reduction of the extent ofprotected lands. An increasing human populationdensity and potential economic growth in the regionis certain to put increasing pressure on protected areassuch that it is conceivable that in the future some areascurrently protected for wildlife could be developed.Scenario 1 evaluated the current situation (as predictedby Elliot et al. 2014b). In scenario 2 we held landscaperesistance the same as under the current situation, butreassigned all WMAs in Botswana and Zimbabwe(142,912 km2) to unprotected agro-pastoral lands andrecalculated their landscape resistance using theequation in Elliot et al. (2014b). This scenario wasevaluated as an extreme outcome of proposed miningactivity in the WMAs of Zimbabwe and a trophy

Landscape EcolTable 1 Five hypotheses relating to change in landscape connectivity for African lions (Panthera leo) were evaluated for each ofnine landscape change scenarios at eight dispersal s11. Current situationIf the status quo can be maintained2. No wildlife managementareas in Botswana andZimbabweConcerns that a ban of trophy hunting and mining activitycould reduce land available for wildlife1, 2H13. National parks onlyExtreme scenario of human pressure for land resulting in theconversion of all non-National Parks to agro-pastoral land3H14. Designation of protectedcorridorsPeace Parks Foundation facilitated the establishment ofKAZA and have expressed interest in establishing corridors4H35. Erection of funneling fencesaFences could be used to ensure lions do not disperse into areaswhere there is no ecological value of doing so5, 6, 7, 8, 9H36. Fenced National ParksaFences could be used to protect National Parks from humanencroachment and prevent animals from leaving5, 6, 7, 8, 9H27. Botswana veterinary fencesand fenced boundary of HNPaExisting veterinary fences in Botswana could be enhanced anda new fence erected in HNP5, 6, 7, 8, 9H248. Doubling of humanpopulationHuman populations are growing rapidly in Sub-SaharanAfrica—Africa’s population is expected to double by 20503H459. Only fenced national parksremainaExtreme scenario where all wildlife areas have been lostexcept National Parks, which are then fenced5, 6, 7, 8, 9H1, H223Predictions were formulated based on hypotheses and findings from the following sources: (1) Lindsey et al. (2006); (2) Lindsey et al.(2012); (3) Population Reference Bureau (2013); (4) PPF (2013); (5) Packer et al. (2013a); (6) Creel et al. (2013); (7) Packer et al.(2013b); (8) Woodroffe et al. (2014); (9) Pfeifer et al. (2014)aAll fences were made impermeable to lion movementhunting ban in Botswana, both of which couldpotentially result in a landscape dominated by anthropogenic activity, such as in agro-pastoral lands. Inscenario 3 we reassigned all areas not designated asNational Parks (313,050 km2) to unprotected agropastoral lands and recalculated their landscaperesistance.In the second group of landscape change scenarioswe evaluated the effect of fencing and corridors onlandscape permeability. Designation of corridors andbuilding of fences to direct dispersing individualstoward suitable habitat and away from areas of highmortality risk are approaches proposed to improvepopulation connectivity and mitigate for habitat loss(Beier et al. 2008). Scenario 4 investigated the effectof designating and protecting four wide dispersalcorridors across breaks between protected areas andassigning them resistance equivalent to national parks,while scenario 5 evaluated the effect of buildingimpermeable fences to funnel dispersing individualstoward protected areas and away from areas of highhuman population density (Fig. 2).In the third group of scenarios, we evaluated twoscenarios related to fencing. Recently, Packer et al.(2013a) suggested that one option to mitigate againstthe edge effects associated with rapid human population and economic growth, is to fence lion populations.In scenario 6, we evaluated the effects of fencingexisting National Parks, while keeping all protectionoutside the parks as existing currently. Scenario 7evaluated the potential impact that existing veterinaryand forestry fencing coupled with potential newfencing along the eastern boundary of HwangeNational Park might have if made impenetrable tolion dispersal.There was a single scenario in each of the fourthand fifth groups. Scenario 8 evaluated a doubling ofthe human population, while keeping all currentprotected lands in their present status. Scenario 9combined aspects of both reduction of the extent ofprotected land and use of fences. Specifically, thisscenario involved the transfer of all non-national parklands to agro-pastoral use and fencing of NationalParks.123

Landscape EcolFig. 2 Visual depiction ofthe nine landuse changescenariosEvaluation of influences of a range of dispersalabilitiesEmpirical optimization such as presented by Elliot et al.(2014b) can provide robust estimates of the relativeresistance costs of different landscape conditions.However, information on relative cost of movementacross a landscape does not enable reliable prediction ofpopulation connectivity without knowledge of thedispersal ability of the species, in terms of how muchcumulative cost a disperser is able to traverse. For most123organisms information on dispersal ability is absent oranecdotal. However, recent studies of lion dispersalhave vastly improved our knowledge on movement(Elliot et al. 2014a), connectivity (Elliot et al. 2014b)and survivorship of dispersing individuals (Elliot et al.2014c). This uncertainty in dispersal ability is particularly important given recent work which showed thatdispersal ability may often have a larger effect onpopulation connectivity than differences in the resistance map itself (e.g., Cushman et al. 2012, 2013). Weevaluated the sensitivity of connectivity predictions to

Landscape EcolTable 2 Description of eight dispersal abilities evaluatedDispersalabilityscenarionumberDispersal ability in costunits of Elliot et al.(2014b) resistance mapExpected maximumdispersal distancethrough protectednational park 000291.784,000,000333.3dispersal ability by evaluating eight different dispersalabilities, designated in cost units (Table 2). We soughtto bracket the plausible range of lion dispersal from av