UBC Faculty Research and Publications

Landscape-level movement patterns by lions in western Serengeti: comparing the influence of inter-specific… Kittle, Andrew M; Bukombe, John K; Sinclair, Anthony R E; Mduma, Simon A R; Fryxell, John M Jul 1, 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-40462_2016_Article_82.pdf [ 3.45MB ]
JSON: 52383-1.0368586.json
JSON-LD: 52383-1.0368586-ld.json
RDF/XML (Pretty): 52383-1.0368586-rdf.xml
RDF/JSON: 52383-1.0368586-rdf.json
Turtle: 52383-1.0368586-turtle.txt
N-Triples: 52383-1.0368586-rdf-ntriples.txt
Original Record: 52383-1.0368586-source.json
Full Text

Full Text

RESEARCH Open AccessLandscape-level movement patterns bylions in western Serengeti: comparing theinfluence of inter-specific competitors,habitat attributes and prey availabilityAndrew M. Kittle1,4*, John K. Bukombe2, Anthony R. E. Sinclair2,3, Simon A. R. Mduma2 and John M. Fryxell1AbstractBackground: Where apex predators move on the landscape influences ecosystem structure and function and istherefore key to effective landscape-level management and species-specific conservation. However the factorsunderlying predator distribution patterns within functional ecosystems are poorly understood. Predator movementshould be sensitive to the spatial patterns of inter-specific competitors, spatial variation in prey density, andlandscape attributes that increase individual prey vulnerability. We investigated the relative role of thesefundamental factors on seasonal resource utilization by a globally endangered apex carnivore, the African lion(Panthera leo) in Tanzania’s Serengeti National Park. Lion space use was represented by novel landscape-level,modified utilization distributions (termed “localized density distributions”) created from telemetry relocations ofindividual lions from multiple neighbouring prides. Spatial patterns of inter-specific competitors were similarlydetermined from telemetry re-locations of spotted hyenas (Crocuta crocuta), this system’s primary competitor forlions; prey distribution was derived from 18 months of detailed census data; and remote sensing data was used torepresent relevant habitat attributes.Results: Lion space use was consistently influenced by landscape attributes that increase individual preyvulnerability to predation. Wet season activity, when available prey were scarce, was concentrated nearembankments, which provide ambush opportunities, and dry season activity, when available prey were abundant,near remaining water sources where prey occurrence is predictable. Lion space use patterns were positivelyassociated with areas of high prey biomass, but only in the prey abundant dry season. Finally, at the broad scale ofthis analysis, lion and hyena space use was positively correlated in the comparatively prey-rich dry season andunrelated in the wet season, suggesting lion movement was unconstrained by the spatial patterns of their maininter-specific competitors.Conclusions: The availability of potential prey and vulnerability of that prey to predation both motivate lionmovement decisions, with their relative importance apparently mediated by overall prey abundance. With practicaland theoretical implications, these results suggest that while top carnivores are consistently cognizant of howlandscape features influence individual prey vulnerability, they also adopt a flexible approach to range use byadjusting spatial behaviour according to fluctuations in local prey abundance.Keywords: Crocuta crocuta, Panthera leo, Prey distribution, Prey vulnerability, Resource utilization, Seasonality, Spatialecology* Correspondence: akittle@uoguelph.ca1Department of Integrative Biology, University of Guelph, 50 Stone Road East,Guelph, Ontario N1G 2W1, Canada4Present address: The Wilderness &Wildlife Conservation Trust, 130 ReidAvenue, Colombo 04, Sri LankaFull list of author information is available at the end of the article© 2016 The Author(s). Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Kittle et al. Movement Ecology  (2016) 4:17 DOI 10.1186/s40462-016-0082-9BackgroundThe distribution and abundance of top predators on thelandscape can exert profound influence on the distribu-tion and abundance of their prey [1–3]. This in turn canimpact predator-prey population dynamics [4, 5] as wellas ecosystem structure as mediated by trophic cascades[6, 7]. Understanding the factors that drive apex preda-tor space use therefore provides valuable insights intocommunity structure and dynamics and is accordingly offundamental importance for the management and futureconservation of both top predators and the broader sys-tems of which they are a part.A basic precept of behavioral ecology is that naturalselection should favor organisms that use landscapes ina way that maximize fitness [8]. For predators a basiccomponent of fitness is the rate of individual prey cap-ture, which is assumed to be enhanced in areas of highprey density [9]. Hence, at the broadest (regional speciesrange) scale, the distribution of large carnivores isobviously determined by the availability of suitable prey[10, 11]. Within functional ecosystems however, themechanism governing predator distribution is more elusive,with space use either dictated by areas of the landscapewhere prey are particularly abundant or spatial locationswhere individual prey capture is more efficient [12].In multi-predator systems, the location of inter-specificcompetitors also can influence decisions on space use bycarnivores [13] often with subsequent impacts on popula-tion dynamics [14–16]. Exploitative and interference com-petition are particularly widespread among Africancarnivores and may be fundamental to shaping distribu-tion patterns [17]. Lions and hyenas are the most import-ant predators, functionally and numerically, in manyAfrican systems [18] and are potentially strong directcompetitors given that their diet and ecological range ex-tensively overlap [19–21]. As a result the two species typ-ically exhibit negative interactions in the form of directaggression [22] and kleptoparasitism [23] as they competefor the same suite of prey resources [24]. Adding furthercomplexity to this important inter-specific relationship,the relative status of these top carnivores is unclear withdominance appearing to be a function of prey availabilitywithin the shared ecosystem [25].Here we use landscape-level seasonal lion and hyenaspace use metrics as well as unusually comprehensiveprey abundance and distribution data to investigate thedrivers of space use by lions in a multi-prey, migratorysystem in the Western Corridor of Serengeti NationalPark, Tanzania. Specifically, we ask whether the use ofspace by lions is primarily influenced by 1) spatial nichepartitioning with their primary inter-specific competitorin the system, the spotted hyena, 2) attributes of thelandscape that increase individual prey vulnerability, or3) the direct availability of prey. Furthermore, sincespatial and temporal heterogeneity in resource availabil-ity partially underlies seasonal shifts in organism distri-bution patterns [26], we ask whether the factorsinfluencing predator space use differ between wet anddry seasons. Due to the annual migration of wildebeestacross the Greater Serengeti Ecosystem, prey availabilityvaries considerably in the Western Corridor [27]. Wedemonstrate that landscape features that increase indi-vidual prey’s vulnerability to predation are consistentlyinfluential predictors of lion range utilization whereasmeasures of general prey availability influence lionmovement patterns only when overall prey abundance inthe study area is high. We further show that at the broadscale of this analysis, lions are not employing spatialniche partitioning with respect to their primary competi-tor in the system, the spotted hyena.MethodsStudy areaThe 25 000 km2 Serengeti ecosystem includes three dis-tinct regions – the Serengeti Plains, the Western Corri-dor and the North [28]. The system structure isdominated by its rainfall regimen with the amount ofrainfall following a south-east (500 mm) to north-west(1100 mm) gradient [29]. The wet season runs fromNovember through May and dry season June throughOctober [27].This study was conducted in a 1440 km2 portion ofthe Western Corridor (Fig. 1), a geologically complex re-gion characterized by alluvial soil deposited by twomajor east-west oriented rivers, the Grumeti to thenorth and Mbalageti to the south, between which runs aseries of Precambrian banded ironstone hills [27]. TheCorridor is composed of a sparse woodland-grasslandmosaic interspersed with patches of dense woodland[28]. This is a transitory zone for the wildebeest migra-tion as it moves in a sweeping arc from the SerengetiPlains to the North, with the influx of migrating animalsarriving in the Corridor at the onset of the dry season(June – July) and typically passing through prior to com-mencement of the wet season [30]. In contrast to theSerengeti Plains, the Western Corridor has substantialpopulations of resident ungulates [28], including residentwildebeest [31]. Lions and hyenas are the dominantpredators in the Serengeti system occurring at densitiesof 0.12/km2 and 0.36/ km2 respectively [27] and ac-counting for ~85 % of large herbivore predation [20].Lion utilization distributionsBetween December 2009 and June 2011 we attachedGPS collars to a total of 6 adult female lions (TelonicsTGW-4500, Mesa, AZ, USA) from 5 separate, adjacentprides in the central portion of the Western Corridor(Fig. 1). Lions were immobilized by veterinarians fromKittle et al. Movement Ecology  (2016) 4:17 Page 2 of 18Tanzania National Parks (TANAPA) and Tanzania Wild-life Research Institute (TAWIRI). Five GPS-telemetrycollars were deployed at any one time and were pro-grammed to record GPS locations every 2 h. Collar GPSfix success rate was 95.2 % ensuring no bias in telemetryre-locations [32]. Although un-collared prides occurredto the east, west and south of these 5 collared prides,there were no known un-collared prides within the focalstudy area. Pride 5 split shortly after collaring and thisdisruption is reflected by the subsequent wide-rangingmovement of the collared lion from this pride (Fig. 1).This lion exhibited some overlap with an un-collaredpride occupying the area immediately southeast of prides1 and 2, as well as with other members of her originalpride occupying areas southeast of pride 5’s core terri-tory (Fig. 1).We determined utilization distributions (cell size =100x100m) for each individual for each season and yearusing fixed kernel density home range estimation (kde,adehabitatHR package in R) with reference bandwidth[33]. Although smoothing parameter or bandwidth selec-tion critically impacts kernel analysis [33], particularlythe determination of home range outer contours and toa lesser extent the estimation of the utilizationdistribution [34, 35], there is no single best method ofchoosing it a priori [36]. Unless data show a bivariatenormal distribution, the reference bandwidth can over-smooth resulting in an UD larger than necessary [33].However selecting the bandwidth that minimizes least-squares cross-validation [37], an alternative approach in-corporated into most statistical software packages, tendsto under-smooth when using large datasets (i.e. thou-sands of data points from telemetry re-locations) result-ing in an unnecessarily restricted UD [33]. Alternativemethods exist, such as “solve-the-equation plug in” [38]or reducing the reference bandwidth to a fixed [39] oreven flexible [16] proportion, the use of which can be se-lected based on analysis requirements [33]. A centralconsideration in the current analysis was to maximizeareas of overlap of inter-specific competitors, so the ref-erence bandwidth was selected with the tradeoff of ac-knowledging that the larger UDs resulting might be lesssensitive to detecting spatial patterns. We converted ker-nel UDs (kUDs) to volume UDs (vUDs) describing thepercentage of the total territory that needs to be utilizedfor a given cell to be included [40]. We subtracted vUDvalues from 100 to arrive at a more intuitive value foreach cell whereby low use cells get assigned low valuesFig. 1 Study area in Serengeti National Park’s Western Corridor showing prey road strip transect locations and 95 % kernel density utilizationdistribution extents for five neighbouring lion prides (1 – 5) in 2010 wet season. Main rivers flowing East to West are Raho (top), Grumeti (middle)and Mbalageti (bottom). Inset shows Greater Serengeti Ecosystem with location of present study area indicated by black rectangle. A = SerengetiNational Park, B = Ngorongoro Conservation Area, C = Loliondo Game Controlled Area, D =Masaai Mara National Reserve (Kenya), E = MaswaGame Reserve, F = Ikorongo Game Reserve and G = Grumeti Game ReserveKittle et al. Movement Ecology  (2016) 4:17 Page 3 of 18and high use cells, high values (i.e. a cell with vUD of75, which indicates that 75 % of the total range needs tobe used to ensure that a lion uses this cell, becomes 100- 75 = 25, which indicates that the probability of this cellbeing utilized by a lion is 0.25). To remove bias imposedby variation in territory size we integrated to 1 for eachindividual pride by dividing each cell value by the sumof all UD cell values for that pride.We then transformed this probability of use value intoa measure of localized density by multiplying cell valuesby lion group size for that season as determined duringregular monitoring. All lion age and sex classes were in-cluded in pride size counts. If a pride changed compos-ition during the course of a single season, verified byrepeated observations of a new pride size, the size forthat season was determined by averaging the distinctpride sizes weighted by the estimated seasonal durationof each pride size. These use layers represent lion spatialutilization but because they are weighted by group sizeare not true probability density functions and cannottherefore be termed utilization distributions in the for-mal sense [41]. Instead we consider these layers as local-ized density distributions (LDDs).One pride – the Grumeti pride – contained two collaredlions. These were spatially separated for prolonged periods(i.e. entire seasons) when one of the two females segregatedherself from the pride to give birth, whereas the other fe-male remained with the rest of the pride. We incorporateddata from both individuals in order to represent more com-pletely the full pride’s utilization of the landscape. To dothis we calculated seasonal LDDs for each individual in themanner illustrated above and then joined the LDDs to-gether by weighting each individual’s LDD values by theproportion of the total seasonal duration that it represented(i.e. if both individuals were tracked for the entire season,they were evenly weighted (50:50) but if one individual wastracked for 75 % of a given season and the other for the fullseason, weighting was 43:57). This method was followedfor both dry and wet seasons.Finally, we created seasonal landscape-level LDDs byamalgamating all individual pride LDDs for each season.Adjacent lion pride ranges often overlap [42], so if morethan one pride territory overlapped a single cell on thelandscape, we summed all values to determine the totalseasonal use value for each cell each year. We had wetseason data from two years (2010 and 2011) so in orderto arrive at an overall wet season LDD we averaged theindividual year LDDs by combining values across yearsand dividing by the number of years that a given cellwas used. The final landscape-level LDDs used 7983 dryseason relocations and 19164 wet season relocationsfrom 6 individuals representing 5 prides.The state (i.e. pregnant, hungry) as well as unique in-dividual behavioural characteristics of animals within apopulation can result in individual variation in space usepatterns [43] which may impact observed patterns at thepopulation level [44, 45]. By collaring only adult femalesand weighting individual UDs by group size weattempted to account for some of this potential biaswhich is inherent in studies that use individual patternsto scale up to infer population-level processes. Samplinga single age/sex class has the effect of minimizing poten-tial variation in behaviour driven by substantially differ-ent social roles and responsibilities [46]. Weightingindividual UDs by group size works to buffer the effect,without discounting it, of collared individuals whose be-havior, due to a specific state (e.g. pregnancy), circum-stance (e.g. post-pride split) or individual specialization[44], varies markedly from that of their age/sex class andsubsequently, social unit. Therefore weighting by groupsize should provide a more accurate reflection of thelandscape utilization of the larger population.Inter-specific competitionDuring the same period as the lion tracking, we de-ployed GPS radio collars on 6 adult female hyenas from5 separate clans that overlapped the home ranges of col-lared lion prides in the Western Corridor (LOTEK 3300and 4400, Newmarket, ON, Canada and Tellus 2A, Fol-lowit, Lindemark, Sweden). Landscape-level space uselayers were developed following the methods detailedabove with relocations every 2 h, except hyena UDs werenot weighted by clan size given that we were unable toeffectively determine group sizes for all collared clans.Final landscape-level dry season UD was from 4328 relo-cations of 4 individuals from 4 clans and wet season UDwas from 9669 relocations of 5 individuals from 5 clans.Prey availabilityAcross five days of every month for the duration of thestudy, we conducted a series of 6 transects throughoutthe study area between 6 am – noon. These encom-passed periods of increased activity for most availablepotential prey species in the system [47–50]. Transectsranged in length from 9.4 to 43.6 km and comprised atotal of 129 km of roads (Fig. 1). The distance of ob-served individuals from the transect line was determinedusing rangefinder binoculars. Beyond 100 m there is adecay in detection probability for ungulates in this sys-tem [51] so only those potential prey species (Table 1)[20, 52] within 100 m of the road were counted andtheir location along transects relative to the transectstart point recorded using the vehicle odometer. Detec-tion of individual animals was not affected by season[51] allowing for consistent comparisons between wetand dry seasons. The total area covered by thesemonthly transects was 129 km x 0.2 km = 25.8 km2. ThisKittle et al. Movement Ecology  (2016) 4:17 Page 4 of 18exceptional dataset allowed tracking of both spatial andtemporal prey distribution trends.To determine overall prey availability, transects wereoverlaid with a non-overlapping sequence of 645 quad-rats, each measuring 200 x 200 m. Quadrats wereassigned two separate measures for each season: prob-ability of occurrence of any prey species and averageprey biomass. Probability of prey occurrence was a pro-portional measure of the number of monthly transectsthat a given transect quadrat was occupied by any po-tential prey divided by the total number of transectsconducted for that season (i.e. if quadrat A was occupiedby potential prey of any species during 3 dry seasontransects, and there were 5 dry season transects con-ducted, the probability of prey occurrence for this quad-rat would be 3/5 = 0.6). This provides an indication ofthe reliability of a location in terms of the probabilitythat it will contain prey for lions. Average biomass foreach quadrat was calculated as the sum of all prey indi-viduals in that quadrat multiplied by their species-specific weight (Table 1) and then divided by the numberof transects conducted (i.e. if during the 5 dry seasontransects conducted, a given quadrat was detected tocontain a combined total of 6 zebra and 10 wildebeest,the average biomass for that quadrat would be ((6 x250 kg) + (10 x 170 kg))/5 = 640 kg). This measure pro-vides information about the gross seasonal distributionand abundance of prey on the landscape. The correlationbetween prey abundance (sum of all prey individuals ineach quadrat/number of transects conducted) and preybiomass was |r| > .9 for both seasons.Quadrats were also characterized by their compositionof four broad land cover categories based on Reed et al.’sphysiognomic classifications [53], as well as proximity towater sources and proximity to ranger posts and/ortourist lodges. Land cover classes were open grassland,wooded grassland, open woodland and dense woodland.Open grassland was composed of grassed areas (2–100 %) with < 20 % shrub cover and < 2 % tree coverwhereas wooded grasslands had similar shrub cover buttree cover between 2 and 19 %. Open woodland wascomprised of 20–49 % shrubs or trees and dense wood-land > 50 % shrub or tree coverage. Correlation analysiswas conducted to ensure that variable collinearity doesnot bias statistical inference (|r| < 0.7) [54].We used logistic regression appropriate for proportiondata (generalized linear models with binomial errorstructure and logit link function in R) to determine themodel that best explained the probability that quadratswere occupied by any prey [55, 56]. Hosmer and Leme-show goodness-of-fit and Likelihood ratio tests wereused to determine adequacy of model fit. To determinethe average biomass/cell we conducted log-linear model-ing using a negative binomial distribution and log link.A negative binomial distribution was chosen over Pois-son due to over-dispersion of the data [56]. Model as-sumptions were verified by plotting residuals vs. fittedvalues and creating normal QQ plots.Modeling of prey metrics was conducted by stepwisedeletion of predictor variables, starting from the fullysaturated (or global) model which included six variables– distance to water, distance to ranger posts and/ortourist lodges, proportion of open grassland, proportionof wooded grassland, proportion of open woodland, pro-portion of dense woodland plus a quadratic term (Dis-tance_water2). The quadratic was included based on theexpectation that many prey species have non-linear asso-ciations with water (e.g. need to be close to water todrink but not too close due to increased risk of preda-tion). Stepwise deletion was conducted using Likelihoodratio tests, which are appropriate to compare betweennested models [57]. Predictors were retained in the finalmodel when their P-values were <0.2 to prevent the in-advertent omission of important variables [58]. Thebackward stepwise variable elimination process, thoughwidely utilized in ecological modeling [57, 59] can beconsidered inferior to the protocol of determining alter-native plausible models and then challenging the datawith these models to see which the data best supports[60]. When considering competing models representingseparate underlying hypotheses stepwise approaches areflawed, however the current goal was to employ a set ofvariables carefully selected a priori based on biologicalconsiderations and determine the best available combin-ation of them to explain prey occurrence and abundancepatterns across the landscape. As such, we feel this ap-proach was justified.To verify adequacy of the resultant best models(Table 2), we tested for spatial autocorrelation in modelresiduals, first by creating a bubble plot (“sp” package inR) which plots model residuals vs. spatial coordinates, toTable 1 Average adult female weights of lion prey speciesdetected during monthly Western Corridor census surveys [98]Common name Scientific name Weight (kg)Giraffe Giraffa camelopardalis 800Buffalo Syncerus caffer 450Zebra Equus quagga 250Wildebeest Connochaetes taurinus 170Topi Damaliscus lunatus 120Waterbuck Kobus ellipsiprymnus 180Warthog Phacochoerus aethiopicusafricanus 60Grant’s gazelle Nanger granti 55Impala Aepyceros melampus 50Thomson’s gazelle Eudorcas thomsonii 20Olive baboon Papio anubis 20Kittle et al. Movement Ecology  (2016) 4:17 Page 5 of 18qualitatively evaluate whether similarly valued residualswere clumped [57] (Additional file 1: Figure S1). Wethen used variograms (“gstat” package in R; Additionalfile 2: Figure S2) to quantitatively verify that spatial auto-correlation was not an issue in the models [57]. As var-iograms assume isotrophy, we also plotted multi-directional variograms to verify this assumption [57](Additional file 3: Figure S3). The best models were thenused to map the probability of seasonal prey occurrenceand average prey biomass across the landscape. Outputcell size was 200 X 200 m to match the size of inputprey transect quadrats. We further evaluated model fitby plotting observed values for each of our prey avail-ability measures against the values projected from finalmodels and determining the resultant correlation coeffi-cients. These ranged from |r| = 0.11 for average prey bio-mass in the dry season to 0.28 for frequency ofoccurrence in the wet season indicating a weak effectsize and suggesting that some other, unquantified vari-able(s) were influencing prey distribution in the studyarea (Additional file 4: Figure S4 and Additional file 5:Figure S5). Finally, we conducted sensitivity analysis ofthese best models by plotting the projected seasonal aver-age biomass and frequency of occurrence against each in-dividual input variable comprising each top model andestimating correlation coefficients. This provides a visualmeans to indicate the relative influence of individual ex-planatory variables (Additional file 6: Figure S6, Additionalfile 7: Figure S7 and Additional file 8: Figure S8).Landscape attributesThe distance to drainage beds with clearly defined em-bankments and mean percentage of woody cover greaterthan 0.4 m high were used to characterize potential lionhunting cover [59, 61]. Embankments were defined byClasses 1 – 3 of the RiversV3 shapefile in the SerengetiDatabase www.serengetidata.org whereas cover wasbased on the average amount of woody cover calculatedfrom each of the 27 physiognomic land-cover classesidentified by Reed et al. [53] with the height based onminimum cover requirements for lions [62, 63]. We alsocharacterized the landscape in terms of the straight linedistance to nearest water sources as measured using GISanalysis tools. This included all rivers and ephemeralstreams in the wet season (Class 1 - 4) but since mostwater sources in the Western Corridor are highly sea-sonal, only distance to permanent water (Class 1 and 2)was measured in the dry season. Two permanent water-holes recently dug by Tanzania National Park(TANAPA) staff were added to GIS layers separately.The distance to permanent water sources and distanceto embankments were highly correlated (|r| > 0.7) so inthe dry season only the distance to water variable wasmaintained. Landscape topography can impact resourceselection for a variety of large mammals [64–67] so wecreated a digital elevation model (DEM) raster from theSerengeti contour layer from which we determined theaverage elevation and slope across the study area.Modeling lion space useWe created a rectangular grid of 5760 cells, each meas-uring 500 x 500 m, across the study area, as defined bythe outer margins of the largest 95 % landscape-levelseasonal LDD. Each grid cell was then populated withthe average lion use value from the landscape-level sea-sonal LDDs as well as the eight independent variablesrepresenting our three hypotheses (Table 3).Table 2 Best seasonal models explaining the frequency of occurrence and average biomass of preySeason Response Predictor variables θ SE P-valueDry Frequency of occurrence Distance to permanent water -2.49E-04 1.11E-04 <0.05(Distance to permanent water)2 4.67E-08 2.18E-08 <0.05Distance to rangerpost/lodge 1.66E-05 9.30E-06 <0.1Wooded grassland 2.31E-01 1.18E-01 <0.1Dry Average biomass Distance to rangerpost/lodge -7.91E-05 1.79E-05 <0.0001Wet Frequency of occurrence Distance to water 5.04E-04 2.02E-04 <0.05(Distance to water)2 -3.42E-07 1.10E-07 <0.01Distance to rangerpost/lodge -5.44E-05 7.88E-06 <0.0001Wooded grassland -1.60E-01 9.81E-02 <0.2Dense woodland -4.61E-01 2.05E-01 <0.05Wet Average biomass Distance to rangerpost/lodge -5.18E-05 1.77E-05 <0.01Open grassland 1.37E + 00 2.16E-01 <0.0001Open woodland 1.60E + 00 4.23E-01 <0.001Based on 200 x 200 m prey transect quadrats (n = 645). All models determined from backward stepwise elimination procedure using likelihood ratio tests, startingfrom full model (k = 7)Kittle et al. Movement Ecology  (2016) 4:17 Page 6 of 18The proportion of the seasonal lion LDDs included ineach analysis was constrained by the need to overlapwith the landscape-level hyena UDs from the same sea-son. This reduced the number of 500 x 500 m grid cellsincluded in the analysis from 1986 to 629 in the dry sea-son and from 3722 to 1405 in the wet season. We feltthis reduction was warranted in that we wanted to com-pare lion use directly to contemporary hyena utilizationas this parallel tracking of the region’s two primary pred-ators was one of the key attributes of the research.To account for the spatial autocorrelation in the re-sponse variable, we used generalized least squares (GLS)mixed effect regression models with an explicit correl-ation structure (the random effects) to determine the in-fluences of lion space use in our study area [57]. Welog-transformed the response variable (average lion LDDvalue), the inter-specific competition variable (averagehyena UD value) and average slope in order to complywith model assumptions. To determine the appropriatecorrelation structure for the data we ran a saturatedmodel (including all predictor variables) with differentcorrelation structures (our random effects) using the re-stricted maximum likelihood method (REML) [57]. Weused AIC to select the most appropriate correlationstructure (rational quadratic, corRatio in R) and vario-grams to verify that spatial autocorrelation was ad-equately accounted for [56].We then conducted two separate modeling procedures,the first to determine the best model for each of our threehypotheses (INTER = inter-specific competition, LAND=landscape attributes, PREY = prey availability) and the sec-ond to compare those best models, and their additive com-binations, to investigate the relative influence of each onlion landscape utilization. To determine the best model foreach hypothesis we created model sets of all hypothesis-specific potential predictor variables (Table 3) and used aninformation theoretic approach using ΔAIC to evaluate andrank models. Single parameter models with ΔAIC values <2were considered superior to multi-parameter top rankedmodels [68]. Model fit was verified by plotting normalizedresiduals against fitted values and investigating residual dis-tribution. Once we had determined the best model for eachhypothesis for each season, we created a suite of eight com-peting models including the null, the three best single hy-pothesis models and all additive combinations. Wecompared hypotheses using AIC and Akaike weights (wi) todetermine the weight of evidence in support of each [68].To directly compare the relative influence of our three hy-potheses we summed wi of all models in which each hy-pothesis was represented, ensuring equal representation forvalid comparisons [68]. To investigate the influence and as-sociation of individual parameters we re-ran all modelsusing REML to ensure unbiased parameter estimates foreach [57] and used model averaging, a form of multi-modelinference, to determine final unbiased estimates with un-conditional confidence intervals [68]. Model fit was furtherinvestigated by determining the correlation coefficient ofthe log of observed lion space use and the log of use pro-jected from final models, as well as from visual compari-sons of observed utilization maps and those projected frommodel output.Statistical and spatial analysis was undertaken using Rsoftware version 2.15.1 [69], ArcMap 10.1 [70] and Geo-spatial Modeling Environment [71].Direct lion observationsFrom January 2010 through June 2011 collared lionswere regularly re-located on the ground and observedfrom a jeep for a total of 649.5 h. This included 232 ob-servations < 30 min in duration and 198 monitoring pe-riods of individual lions where observation durationwas ≥ 30 min. Between June 2010 and June 2011 we con-ducted 177 individual follows of radio-collared lionsamounting to 607.5 h of monitoring. The average dur-ation of observations was 3.4 h (range 0.5 – 19.5) with332.5 h occurring during the day (7:00 – 18:00), 210.9 hat night (19:00 – 6:00) and 64.1 h during crepuscular pe-riods (6:00 – 7:00 and 18:00 – 19:00). Most nocturnalobservations occurred during 10 extended day-night fol-lows of collared individuals during the 48 h surroundingthe full moon. These extended follows were conductedmonthly between June 2010 and May 2011 with the ex-ception of December 2010 and January 2011. Lions wereobserved with the naked eye when moonlight was suffi-cient and otherwise with night-vision binoculars, occa-sionally supplemented with a hand-held, red-filteredspotlight. The seasonal breakdown saw 306.5 h of moni-toring in the dry season and 301 h in the wet season.ResultsDensity estimates based on monthly transect data clearlyshow the increased dry season availability of potentialTable 3 Independent variables representing each of the threehypotheses proposed to explain lion space useHypothesis Abbreviation VariablesInter-specificcompetitionINTER log(Hyena UD value)LandscapeattributesLAND distance to permanent water (dry)or all water (wet)distance to embankment(wet season only)cover %elevationlog(slope)Prey availability PREY average prey biomassprobability of prey useKittle et al. Movement Ecology  (2016) 4:17 Page 7 of 18lion prey species, particularly migrant wildebeest andThomson’s gazelles (Fig. 2). Landscape level lion densitydistribution maps reflect the increased importance ofpermanent water sources in this season, whereby lionrange utilization can be seen to contract in their vicinity(Fig. 3, left panels). This pattern is not observed forhyena utilization distributions (Fig. 3, right panels).Model evaluation resulted in single variable modelsrepresenting each hypothesis in both seasons (Table 4;See Additional file 9: Table S1 for full model competitionresults). This ensures that hypothesis comparisons arenot beholden to representative models with substantiallydifferent numbers of parameters, allowing greater confi-dence in the validity of comparisons.In the dry season, localized lion density was concen-trated in areas of the landscape that were also heavilyused by hyenas (Table 4). The model that best explainedlandscape-level lion space use during this season in-cluded elements of all three hypotheses. However, apositive association with hyenas indicates that lionspatial utilization patterns were not being influenced byspatial separation from their primary inter-specific com-petitor in this season, which was the expectation if inter-specific competition, acting through spatial niche parti-tioning, was driving lion space use. Therefore this hy-pothesis was eliminated from further consideration inthis season and analysis reduced to a direct competitionbetween prey resources and landscape attributes. Theresulting best dry season model included both preyavailability and landscape attributes (Table 5). Assess-ment of model fit also indicated that the combinedmodel showed the highest correlation between observedand projected lion use (Figs. 4 and 5). CumulativeAkaike weighting suggested that prey availability andlandscape attributes were almost equally associated withhow lions utilize space in this season (Fig. 6) althoughprey biomass exerted the greater influence in the topmodel (Additional file 10: Figure S9). Specifically, inaddition to areas of high hyena use, lion space use wasconcentrated during the dry season close to permanentwater where prey biomass is high (Table 4).The pattern was quite different in the wet season withcumulative model weighting suggesting landscape attri-butes influenced lion movement patterns more thantwice as much as either inter-specific competition orprey availability (Fig. 6). During this season, localizedlion density was disproportionately concentrated in areasin close proximity to embankments but their distribu-tion was not associated with prey availability or hyenaspatial utilization (Tables 4 and 5). Despite this, therewas a higher correlation between observed lion use anduse projected from the best prey availability model thanfrom either the landscape attribute or combined model(Fig. 7). The narrow range of projected lion use valueshowever suggests that additional, un-quantified factorsare influencing lion movement during this season ofprey scarcity. This is reflected in the maps of observedvs. projected use (Fig. 8).DiscussionIn Serengeti’s Western Corridor the massive influx ofmigrant herbivores arrives during the dry season so preyabundance for lions is considerably more plentiful thanduring the wet season (Fig. 2). This increased seasonalabundance is reflected in the movement patterns of indi-vidual lions, which undertake fewer long range (>500 m)Fig. 2 Seasonal density of selected prey species (#/km2) as determined from total animals observed during monthly (N = 18) road strip transects(129 km x 200 m). Assumes all animals within 100 m of transect were detectedKittle et al. Movement Ecology  (2016) 4:17 Page 8 of 18movements between 2-hourly telemetry relocations dur-ing the dry season than the wet season, both in the dayand at night (Additional file 11: Figure S10). Given thesheer mass of prey that enters the Western Corridor atthis time, it is perhaps not surprising that apex predatorsare cuing in on them and that during this season 71 %of all lion kills (n = 55) were wildebeest. Predictability isaided by the physiological limitations of the migrants,since grazers such as wildebeest, zebra and Thomson’sgazelle need to regularly drink and are therefore con-strained in their dry season distribution by the availabil-ity of water [72, 73]. During the dry season availablewater sources are limited as the vast majority of streamsand small water holes disappear, and even the two majorrivers – the Grumeti and Mbalageti - dry up into a seriesof unconnected, stagnating pools. These two factors –that migrating herbivores need to regularly drink andthat there are few places on the landscape where this ispossible – work in favour of the region’s top predators,allowing them to adopt area-restricted search behaviourto take advantage of aggregated prey [74]. Lions in thesemi-arid savanna in Zimbabwe similarly focus theirmovement in proximity to waterholes where they moveat slower speeds and use more tortuous paths [75]. Thusit seems that water is the “spatial anchor” (sensu [12])that allows predators to “win” the behavioural responserace during the dry season.The two lion prides that we were able to observe mostfrequently and for prolonged periods spent manyFig. 3 Seasonal landscape-level 95 % utilization distributions (UDs) for lions and hyenas. Serengeti National Park boundary is shown as a thickblack and white line and permanent rivers as blue lines. UDs transition from low use (green) to high use (red). Lion dry season UD (top left)represents 7983 relocations from 5 lions in 4 prides; lion wet season UD (bottom left) represents 19164 relocations from 6 lions in 5 prides; hyenadry season UD (top right) represents 4328 relocations from 4 hyenas in 4 clans; and hyena wet season UD (bottom right) represents 9669relocations of 5 hyenas from 5 clansTable 4 Model-averaged coefficient estimates with unbiasedstandard errors for final seasonal model variables for eachhypothesis. Model averaging utilized all models (N = 8) includedin the final model suiteSeason Hypothesis Variables θ SEDRY INTER log(hyenaUD)a 0.107592 0.046941LAND distance_permanent_h2oa -0.000193 0.000074PREY Average biomassa 0.003274 0.001067WET INTER log(hyenaUD) -0.014632 0.025151LAND distance_embankmenta -0.000102 0.000048PREY Frequency 0.460227 0.526398a = 95 % Confidence Intervals do not overlap 0Table 5 Model comparison table showing ΔAIC, Akaike weights(wi) and ranking for dry and wet seasonsDry season Wet seasonModel ΔAIC wi Rank ΔAIC wi RankNULL 13.16 0.0013 4 2.24 0.1059 4INTER NA NA NA 3.65 0.0522 6LAND 7.21 0.0249 3 0.00 0.3240 1PREY 5.41 0.0611 2 3.70 0.0511 7INTER + LAND NA NA NA 1.34 0.1660 3INTER + PREY NA NA NA 5.14 0.0248 8LAND + PREY 0.00 0.9128 1 1.14 0.1836 2INTER + LAND + PREY NA NA NA 2.51 0.0925 5INTER hypothesis is not considered in the dry season since associationbetween lion and hyena use was positive for this season rendering theinter-specific competition hypothesis untenableKittle et al. Movement Ecology  (2016) 4:17 Page 9 of 18daytime dry season hours in close proximity to the fewremaining water sources in their territories, presumablywaiting for the arrival of the migrant herds of wildebeestand zebra (Additional file 12: Figure S11). At thesetimes, pride members displayed little regard for conceal-ment, instead positioning themselves such that lionswere often placed directly between prey herds and thewater. On several occasions this strategy resulted inmultiple kills for each pride. Perhaps lions were simplytaking advantage of the physiological limitations of mi-grants that are most thirsty and therefore driven to drinkduring the heat of the day. However, herbivores canmake behavioural adjustments to reduce predation riskat important, spatially fixed resources like waterholes[76], so despite the apparent success of the predatorshere, perhaps the tendency for migrants to arrive atwater sources en masse in the middle of the day can re-duce per capita prey risk due to dilution effects [77]and/or predator confusion [78] during the time whenlions are typically least active [20] (Additional file 13:Figure S12).The larger watercourses in the Western Corridor rep-resent not only a source of drinking water for migrants,but also obstacles that must be crossed in order to con-tinue toward the main dry season grazing areas to thenorth [79]. Just as there were few accessible drinkingpools available to herbivores in the dry season there arealso limited river sections that can be easily crossed dueto thick vegetation and steep banks. Flat, shallow riversegments therefore become high density thoroughfaresfor migrant herbivores for a few crucial dry seasonweeks, encouraging predators to remain in closeproximity.Lions did not avoid those parts of the landscape uti-lized by their main inter-specific competitor, the spottedhyena, and were in fact strongly positively associatedFig. 4 Observed and model-projected dry season lion utilization distributions in Serengeti National Park’s western corridor. Top left shows observedlion use; top right shows lion use projected from the prey availability model; bottom left shows lion use projected from the landscape attributes model;and bottom right shows lion use projected from the prey availability + landscape attribute model. Dark line at top is the National Park boundary andblue lines are permanent rivers. Displayed lion UDs are only those portions of total lion ranges that overlap hyena UDsKittle et al. Movement Ecology  (2016) 4:17 Page 10 of 18with areas of high hyena utilization in the prey-rich dryseason. There are three plausible explanations for theobserved association between competitors: a) lions werecuing in on areas of high hyena use, b) areas of high lionuse were being tracked by hyenas, or c) lions and hyenaswere independently selecting the same locations of highprey availability. Hyenas are coursing predators so areunlikely to select for the same landscape features as lionsfor hunting purposes which hints at the probability thatthe observed positive association resulted from one spe-cies tracking the other. However, over the course of650 h of direct lion observation spread throughout theFig. 5 Correlation between observed and projected dry season lion use. Correlation between log of observed lion dry season space use (i.e. theprobability of occupancy of a quadrat) and (left) log of prey availability model-projected dry season lion space use (|r| = 0.33); (middle) log oflandscape attribute model-projected dry season lion space use (|r| = 0.26); and (right) log of prey availability + landscape attribute model-projecteddry season lion space use (|r| = 0.35)Fig. 6 The relative influence of each hypothesis on lion space use by season. Summed Akaike weights (wi) across all models representing eachhypothesis (INTER = inter-specific competition, LAND = landscape attributes, PREY = prey availability; n = 2 in dry season and 4 in wet season) andindicating the relative influence of each hypothesis on lion space use by season. INTER hypothesis is not considered in the dry season sinceassociation between lion and hyena use was positive for this season rendering the inter-specific competition hypothesis untenableKittle et al. Movement Ecology  (2016) 4:17 Page 11 of 18year, including long periods of sustained individual fol-lows, we observed only nine interactions with hyenas.Five of these were aggressive encounters, of which fourwere over kills. During these aggressive events, hyenassupplanted 2 – 3 female lions with cubs twice and singlemale lions supplanted groups of hyenas twice. The rela-tive paucity of aggressive interactions suggests that thesecompetitors were not actively tracking one another, sug-gesting that food is not a limiting resource for eitherspecies during the dry season. Both lions [13, 80] andhyenas [81] select for areas of high prey abundance atintermediate scales (i.e. 3rd order [82]). Perhaps inter-specific competition is of limited concern during the dryseason because migratory prey is plentiful, so top preda-tors independently utilized similar, prey-rich areas. How-ever, competition avoidance can be a more subtleprocess than a lack of obvious interactions. Both Hop-craft et al. [61] and Davidson et al. [80] in their analysesof lion predation events observed scale-dependent killsite selection with broader scale lion distribution influ-enced by prey abundance and finer scale prey utilization(i.e. 4th order [82]) predominantly influenced by habitatfeatures that increase prey vulnerability. Given that lionsand hyenas employ divergent hunting techniques, a finerscale of analysis could potentially detect more subtlespatial separation between these predators within thesewider shared regions that is not apparent here [83]. Nei-ther hyenas nor lions are clearly sub-ordinate to theother, with interaction outcomes typically dependent onrelative numbers and group composition [23]. Thereforeunlike subordinate cheetahs (Acinonyx jubatus) whichactively avoid both lions [13, 84] and hyenas [85], orleopards (P. pardus), which avoid dominant lions [13]and tigers (P. tigris) [86], avoidance behaviour is notentrenched here and in the absence of intense competi-tion, was not observed.In the wet season, overall prey biomass is much lowerin the Western Corridor and given the widespread avail-ability of both forage and water, less predictable spatially.At this time, with relatively few prey and few limiting re-sources, lions may be unable to effectively track whereprey are most abundant. Alternately, given that preyherds congregate in the open grasslands during this sea-son [51] (Table 2) it may be more difficult for lions toaccess individual prey there due to close grouping [87–89] and open habitat [90] promoting improved predatordetection. As a result, lions are cuing in on areas of thelandscape that should increase individual prey vulner-ability, disproportionately utilizing areas in proximity toembankments which allow effective concealment andFig. 7 Correlation between observed and projected wet season lion use. Correlation between log of observed lion wet season space use (i.e. theprobability of occupancy of a quadrat) and (left) log of prey availability model-projected wet season lion space use (|r| = 0.37); (middle) log oflandscape attribute model-projected wet season lion space use (|r| = 0.18); and (right) log of prey availability + landscape attribute model-projected wet season lion space use (|r| = 0.23)Kittle et al. Movement Ecology  (2016) 4:17 Page 12 of 18thus offer the potential to increase hunting success [61].During this season zebra comprised 48 % of lions kills(n = 21) with buffalo (19 %), wildebeest (19 %) and wart-hog (14 %) also important.We saw no correlation between lion and hyena spaceuse during the wet season, despite the decrease in avail-able prey. As coursing predators, hyenas are unlikely tobias their space use towards ambush features such asembankments which might de-couple their movementsfrom those of lions. Additionally, hyenas in the Serengetiare unusual (but see [91]), in that they can undertake ex-tended extra-territorial commutes to access areas of in-creased prey density [92]. One of the collared hyenas inthis study undertook such a commute in the late wetseason when prey availability in the Corridor was low,moving > 50 km southeast over the course of 19 days,presumably to access migrants on their way west (Fig. 9).Perhaps this ability to move beyond territorialboundaries relieves some of the burden of food acquisi-tion which might otherwise increase competitive interac-tions with inter-specifics, promoting co-existence.Overall, the observed variation in lion range use wasnot well captured by the best models, in the wet seasonin particular, as evidenced from the narrow range of pre-dicted lion utilization values (Figs. 5 and 7). This appearsto suggest that the model parameterization was subopti-mal or that other factors that were not the focus of thisstudy influence lion movement decisions.One potential shortfall in model parameterization mightstem from the employment of daytime prey transects. Sa-vanna ungulates have been observed to alter their habitatpreferences according to time of day [93] so the reliancehere on daytime transects might limit our ability to detectthe full range of lion prey distribution. Lions in the studyarea did hunt diurnally as well as nocturnally, with thehourly observed probability of a hunt, based on 52Fig. 8 Observed and model-projected wet season lion utilization distributions (500 m grid) in Serengeti National Park’s western corridor. Top left showsobserved lion use; top right shows lion use projected from the prey availability model; bottom left shows lion use projected from the landscapeattributes model; and bottom right shows lion use projected from the prey availability + landscape attribute model. Dark line at top is the National Parkboundary and blue lines are permanent rivers. Displayed lion UDs are only those portions of total lion ranges that overlap hyena UDsKittle et al. Movement Ecology  (2016) 4:17 Page 13 of 18observed hunting episodes, 0.078 in the day (7:00 –18:00), 0.109 during crepuscular periods (6:00 – 7:00 and18:00 – 19:00) and 0.057 at night (19:00 – 6:00). Whilethis partially validates the reliance on daytime transects,most nocturnal observations in this study were under-taken during full moon periods when hunting success, ifnot effort, is lower [94]. Additionally, lions displayed ahigher frequency of long range movement (>500 m) dur-ing the night than during the day across both wet and dryseasons (Additional file 11: Figure S10 and Additional file13: Figure S12), which is indicative of increased nocturnalactivity and presumably increased hunting effort. Further-more, lions in the dry season were more frequently fartherfrom water sources nocturnally than during the day (Add-itional file 12: Figure S11). At night ungulates are lesslikely to visit waterholes due to increased predation risk[76] forcing lions to similarly move away from these re-sources to access prey. These nocturnal behaviouralchanges have the potential to further alter prey distribu-tion patterns [1]. Therefore, it is recommended that futurework incorporate methods to determine nocturnal preydistributions in order to arrive at a more complete under-standing of the processes discussed here.A key element that might have weakened the observedrelationships between lion space use and the variablesthat were considered here is the behavioural state of in-dividual lions in the study. Of the 6 radio-collared fe-male lions, 5 of them had cubs during the course of theresearch. In lion society females typically retreat fromthe main pride to give birth and can stay separated fromtheir pride for several weeks after cubs are born [20].During this period these lions can alter space use deci-sions based on the prioritized need to keep cubs securefrom con- and inter-specifics. In our study area femalespost-partum ranged widely in behavior. One femalemoved to the periphery of her pride’s range andremained separate from her pride for several monthswhereas another, in the same pride, withdrew to a se-cluded location within the central portion of her pride’srange and resumed movement with the pride only a fewweeks after the birth of her cubs. Yet another femalegave birth shortly after her initial pride split and sheremained alone with her cubs, wandering widely for al-most a year before settling into a new pride near the endof the study period. In such a scenario it is likely thatmodelling each collared lion’s resource utilization wouldhave resulted in the detection of a wide range of influen-tial factors depending on the individual and a generalsynthesis of broad-scale space use patterns would nothave been possible. Using the admittedly more complexamalgamation method that we employed here allows in-dividual differences to be incorporated into model struc-ture but still manages to synthesize these differences anddetect, albeit weakly, selection processes that are drivingthe broad-scale patterns observed.A tradeoff here is that we use model outputs, whichalready represent approximations of the processes thatare the focus of those models, both as response andFig. 9 Extra-territorial movements of Grumeti Hill hyena during the late wet season 2010. Large cluster of black circles in upper left represents Wetseason range with 19 day “commute” visible to the Musabi Plains in the southeast. Point 1 = departure from usual range (May 07, 18:00),2 = first extra-territorial cluster 55 km from centre of usual range (May 8, 22:00 – May 12, 18:00), 3 = second extra-territorial cluster along Grumetiriver (May 13, 04:00 – May 24, 20:00), 4 = travel along boundary of protected area complex (May 25, 18:00), and 5 = return to usual range (May 26,02:00). Minimum travel distance based on summed step lengths = 160 kmKittle et al. Movement Ecology  (2016) 4:17 Page 14 of 18independent variables. The “noise” created by this strat-egy necessarily dilutes our ability to detect subtle effectsbut is compensated for by the ability to perceive theoverarching drivers of landscape-level use emergingfrom such a complex system, defined by multiple pridesof varying size and structure. Understanding these broadpatterns was the foremost goal of the research and thatmodel results concurred with expectations based on ex-tensive direct observations of lions within this systemserves to further validate the use of this procedure.One possible extension of this novel landscape-levelanalysis is the creation of a layer of predator space use thatcan spread beyond the boundaries of the original sourcearea (assuming new locations allow the accurate measure-ment of necessary co-variates) to predict predator distri-bution. This modelling process has been successfully usedin this way to link habitat quality to carnivore range size[95] and to create broad-scale predation risk layers towhich potential prey species respond [96, 97].ConclusionsOur results clearly suggest that both overall prey availabilityand landscape features that increase individual prey vulner-ability influence space use decisions by Serengeti lions inthe Western Corridor. Avoidance of main inter-specificcompetitors was not observed, suggesting that broad scalelion space use decisions are fundamentally shaped by theneed to locate, secure and capture prey. The relative contri-bution of these prey-based factors varies seasonally and ap-pears to hinge on the overall abundance of prey within theregion as well as its predictability, ensuring that when preyare scarce habitat features promoting hunting success be-come relatively more influential. This underscores the flex-ible approach to range use employed by top carnivores andhighlights the importance of investigating a multi-facetedsuite of ecological variables when the goal is to understandthe drivers of carnivore landscape utilization. In multi-predator assemblages where prey availability varies season-ally, as is the case in most tropical and sub-tropical systems,these results have important management implications. Fi-nally, top predators are essential in shaping the trophicstructure of the ecosystem but are some of the most im-periled of all species on earth [7]. These results show thatconservation of predators, and the whole trophic cascade,requires a knowledge of the fundamental factors that mo-tivate their utilization of the landscape.Additional filesAdditional file 1: Figure S1. Bubble plots of top model standardizedresiduals vs. spatial coordinates for A. average prey biomass (kg/km2) inthe dry season, B. average biomass (kg/km2) in the wet season, C.frequency of prey occurrence in the dry season, and D. frequency of preyoccurrence in the wet season. Residual values are distinguished by colourwith negative values in red and positive values in green. Excessiveclumping of similar values (i.e. red clumps vs green clumps) indicatespossible spatial autocorrelation, which appears absent from these plots.(GIF 70 kb)Additional file 2: Figure S2. Variograms showing variance of topmodel standardized residuals for A. average prey biomass (kg/km2) in thedry season, B. average biomass (kg/km2) in the wet season, C. frequencyof prey occurrence in the dry season, and D. frequency of preyoccurrence in the wet season. (GIF 40 kb)Additional file 3: Figure S3. Multi-directional variograms ofstandardized residuals for A. average prey biomass (kg/km2) in the dryseason, B. average biomass (kg/km2) in the wet season, C. frequency ofprey occurrence in the dry season, and D. frequency of prey occurrencein the wet season. Directions are indicated by degree values 0 = North-South, 45 = Northeast-Southwest, 90 = East-West and 135 = Southeast-Northwest. From the lack of strong spatial patters it appears thatisotrophy is a reasonable assumption. (GIF 64 kb)Additional file 4: Figure S4. Correlation between log of observed preybiomass (kg/km2) for each quadrat in the prey transects (N = 645) and(left) log of model-projected dry season prey biomass from the best dryseason model (|r| = 0.11) and (right) log of model-projected wet seasonprey biomass (|r| = 0.15) from the best wet season model. (TIFF 1235 kb)Additional file 5: Figure S5. Correlation between observed frequency(i.e. average probability that a quadrat is occupied by prey) for eachquadrat in the prey transects (N = 645) and (left) log of model-projecteddry season prey frequency from the best dry season model (|r| = 0.12)and (right) log of model-projected wet season prey frequency (|r| = 0.28)from the best wet season model. (TIFF 1235 kb)Additional file 6: Figure S6. Sensitivity analysis of input variablescomprising the best wet season prey biomass model. The left graphshows prey biomass (kg/km2) values projected from the top model(Distance to disturbance + Open grassland + Open woodland) against thedistance to disturbance input variable values (|r| = -0.46). The middlegraph shows the same projected biomass values against the proportionof open grassland input variable values (|r| = 0.76). The right graph showsthe same model output on the Y-axis against the proportion of openwoodland input variable values (|r| = 0.06). (TIFF 1235 kb)Additional file 7: Figure S7. Sensitivity analysis of input variablescomprising the best dry season prey frequency model. The left graph showsprey frequency (average probability of occurrence) values projected fromthe top model (Distance to permanent water + Distance to permanentwater2 + Distance to disturbance +Wooded grassland) against the distanceto permanent water input variable values (|r| = 0.06). The middle graphshows the same projected frequency values against the distance todisturbance input variable values (|r| = 0.38). The right graph shows thesame model output on the Y-axis against the proportion of woodedgrassland input variable values (|r| = 0.59). (TIFF 1235 kb)Additional file 8: Figure S8. Sensitivity analysis of input variablescomprising the best wet season prey frequency model. The far left graphshows prey frequency (average probability of occurrence) values projectedfrom the top model (Distance to all water + Distance to all water2 +Distance to disturbance +Wooded grassland + Dense woodland) againstthe distance to permanent water input variable values (|r| = -0.20). The insideleft graph shows the same projected frequency values against the distanceto disturbance input variable values (|r| = -0.87). The inside right graphshows the same model output on the Y-axis against the proportion ofwooded grassland input variable values (|r| = -0.08). The far right graphshows the same projected frequency values against the proportion of densewoodland input variable values (|r| = -0.35). (TIFF 1235 kb)Additional file 9: Table S1. Results of competition to determine bestmodel for each hypothesis (INTER, LAND and PREY) for each Season (DRYand WET). Best models as determined by ΔAIC are highlighted usingbold text. Where ΔAIC < 2 the model with the fewest parameters wasselected as the best model. (XLSX 12 kb)Additional file 10: Figure S9. Sensitivity analysis of input variablescomprising the best dry season lion space use model. The left graphshows log(lionuse) values projected from the top model (Average preyKittle et al. Movement Ecology  (2016) 4:17 Page 15 of 18biomass + Distance to permanent water) against the average preybiomass input variable values. The right graph shows the same modeloutput on the Y-axis against the distance to permanent water inputvariable values. Both the visual pattern and the Pearson correlationcoefficients (|r| = 0.89 and |r| = -0.33 respectively) indicate that averageprey biomass is the more influential variable. (TIFF 1235 kb)Additional file 11: Figure S10. Histograms showing 2 h step lengths(m) for all radio-collared lions (N = 6) in the study area between December2009 and June 2011. The y-axis is a measure of relative frequency ofoccurrence so that seasons with different numbers of step lengths can becompared. In both the day (6:00 – 18:00) and night (18:00 – 6:00) lionsmoved longer distances more frequently in the wet season than the dryseason with diurnal mean movement distance = 105 m during the dryseason and 135 m during the wet season (t = -5.1; P < 0.0001) and nocturnalmean movement distance = 449 m during the dry season and 604 m duringwet season (t = -10.5; P < 0.0001). During both diurnal and nocturnal periods2-h movement distances were usually below 500 m. (TIFF 367 kb)Additional file 12: Figure S11. Histograms showing average distance towater (m) for Kirawira pride between December 2009 and June 2011. The y-axis is a measure of the relative frequency of occurrence so that seasonswith different numbers or re-locations can be compared. In both the wetand dry seasons lions preferred to be in close proximity to water sources.This preference was stronger in the daytime than nocturnally in bothseasons, with the average dry season daytime distance = 463 m and averagenighttime distance = 549 m, t = -3.314, P < 0.001) and average wet seasondaytime distance = 341 m and average nighttime distance = 386 m, t= -4.632, P < 0.0001). Pride lions were within 100 m of water more than 2xas frequently in the dry season daytime that dry season night. (PNG 4 kb)Additional file 13: Figure S12. Histograms showing 2 h step lengths(m) for all radio-collared lions (N = 6) in the study area between December2009 and June 2011. The y-axis is a measure of relative frequency ofoccurrence so that seasons with different numbers of step lengths can becompared. In both the wet and dry seasons lions moved longer distancesmore frequently at night than during the day with wet season meanmovement distance = 135 m during the day and 604 m during the night(t = -52.1; P < 0.0001) and dry season mean movement distance = 105 mduring the day and 449 m during the night (t = -29; P < 0.0001). In bothseasons 2-h movement distances were usually below 500 m. (TIFF 367 kb)AcknowledgementsData collection was greatly assisted by Joseph Masoy, Ally Nkwabi, JohnMchetto and Anjali Watson. Tumaini Soyala provided invaluable initialguidance in the study area and Grant Hopcraft provided GIS expertise andan invaluable roof tent. Craig Packer provided welcome guidance and audioequipment necessary for hyena collar retrieval. Permits and logistical supportwas kindly provided by Tanzania National Parks (TANAPA) and TanzaniaWildlife Research Institute (TAWIRI). This work was supported throughDiscovery Grants to ARES and JMF from the Natural Sciences andEngineering Research Council of Canada as well as financial support fromthe Frankfurt Zoological Society. AMK was supported through an NSERCPostgraduate Fellowship.Authors’ contributionsAMK and JMF conceived and designed the study. AMK and JKB conductedthe fieldwork. AMK analyzed the data and wrote the manuscript. ARES andSARM provided logistical support and editorial advice on the manuscript. Allauthors read and approved the final manuscript.Competing interestsThe authors declare that they have no competing interests.Author details1Department of Integrative Biology, University of Guelph, 50 Stone Road East,Guelph, Ontario N1G 2W1, Canada. 2Tazania Wildlife Research Institute, P.O.Box 661, Arusha, United Republic of Tanzania. 3Biodiversity Research Centre,University of British Columbia, Vancouver, BC V6T 1Z4, Canada. 4Presentaddress: The Wilderness &Wildlife Conservation Trust, 130 Reid Avenue,Colombo 04, Sri Lanka.Received: 16 February 2016 Accepted: 17 May 2016References1. Lima SL, Dill LM. Behavioral decisions made under the risk of predation: areview and prospectus. Can J Zool. 1990;68:619–40.2. Ripple WJ, Beschta RL. Wolves and the ecology of fear: Can predation riskstructure ecosystems? BioScience. 2004;54:755–66.3. Valeix M, Loveridge AJ, Chamaillé-Jammes S, Davidson Z, Murindagomo F,Fritz H, Macdonald DW. Behavioral adjustments of African herbivores topredation risk by lions: spatiotemporal variations influence habitat use.Ecology. 2009;90:23–30.4. Shulman M. Recruitment of coral reef fishes: effects of distribution ofpredators and shelter. Ecology. 1985;66:1056–66.5. Lima S. Nonlethal effects in the ecology of predator-prey interactions.Bioscience. 1998;48:25–34.6. Elmhagen B, Ludwig G, Rushton S, Helle P, Linden H. Top predators,mesopredators and their prey: interference ecosystems along bioclimaticproductivity gradients. J Anim Ecol. 2010;79:785–94.7. Ripple WJ, Estes JA, Beschta RL, Wilmers CC, Ritchie EG, Hebblewhite M,Berger J, Elmhagen B, Letnic M, Nelson MP, Schmitz OJ, Smith DW, WallachAD, Wirsing AJ. Status and ecological effects of the world’s largestcarnivores. Science. 2014; doi: 10.1126/science.1241484.8. Morris DW. Toward an ecological synthesis: a case for habitat selection.Oecologia. 2003;136:1–13.9. Holling CS. The components of predation as revealed by a study of small-mammal predation of the European pine sawfly. Can Entomol. 1959;91:293–320.10. Carbone C, Gittleman JL. A common rule for the scaling of carnivoredensity. Science. 2002;295:2273–6.11. Karanth KU, Nichols JD, Kumar NS, Link WA, Hines JE. Tigers and their prey:Predicting carnivore densities from prey abundance. Proc Natl Acad Sci U SA. 2004;101:4854–8.12. Sih A. Predator-prey space use as an emergent outcome of a behavioralresponse race. In: Barbosa P, Castellanos I, editors. Ecology of predator-preyinteractions. Oxford: Oxford University Press; 2005. p. 240–55.13. Vanak AT, Fortin D, Thakar M, Ogden M, Owen C, Greatwood S, Slotow R.Moving to stay in place: behavioral mechanisms for coexistence of Africanlarge carnivores. Ecology. 2013;94:2619–31.14. Creel S, Creel NM. Limitation of African wild dogs by competition withlarger carnivores. Conserv Biol. 1996;10:526–38.15. Tannerfeldt M, Elmhagen B, Angerbjörn A. Exclusion by interferencecompetition? The relationship between red and arctic foxes. Oecologia.2002;132:213–20.16. Berger KM, Gese EM. Does interference competition with wolves limit thedistribution and abundance of coyotes? J Anim Ecol. 2007;76:1075–85.17. Caro TM, Stoner CJ. The potential for interspecific competition amongAfrican carnivores. Biol Conserv. 2003;110:67–75.18. Mills MGL. Large carnivores and biodiversity in African savanna ecosystems. In:Ray JC, Redford KH, Steneck RS, Berger J, editors. Large carnivores and theconservation of biodiversity. Washington DC: Island Press; 2005. p. 208–28.19. Kruuk H. The spotted hyena: A study of predation and social behaviour.Chicago: University of Chicago Press; 1972.20. Schaller GB. The Serengeti lion: A study of predator-prey relations. Chicago:University of Chicago Press; 1972.21. Hayward MW. Prey preferences of the spotted hyaena (Crocuta crocuta) anddegree of dietary overlap with the lion (Panthera leo). J Zool (Lond). 2006;270:606–14.22. Trinkel M, Kastberger G. Competitive interactions between spotted hyenasand lions in the Etosha National Park, Namibia. Afr J Ecol. 2005;43:220–4.23. Cooper SM. Optimal hunting group size: the need for lions to defend theirkills against loss to spotted hyaenas. Afr J Ecol. 1991;29:130–6.24. Watts HE, Holekamp KE. Ecological determinants of survival andreproduction in the spotted hyena. J Mamm. 2009;90:461–71.25. Périquet S, Fritz H, Revilla E. The Lion King and the Hyaena Queen: largecarnivore interactions and coexistence. Biol Rev. 2015;90:1197–214.26. Owen-Smith N. Spatial ecology of large herbivore populations. Ecography.2014;37:416–30.27. Sinclair ARE. Serengeti past and present. In: Sinclair ARE, Arcese P, editors.Serengeti II: Dynamics, management and conservation of an ecosystem.Chicago: University of Chicago Press; 1995. p. 3–30.Kittle et al. Movement Ecology  (2016) 4:17 Page 16 of 1828. McNaughton SJ. Serengeti grassland ecology: The role of compositeenvironmental factors and contingency in community organization. EcolMonogr. 1983;53:291–320.29. Mduma SAR, Sinclair ARE, Hilborn R. Food regulates the Serengetiwildebeest: A 40‐year record. J Anim Ecol. 1999;68:1101–22.30. Holdo RM, Holt RD, Fryxell JM. Opposing rainfall and plant nutritional gradientsbest explain the wildebeest migration in the Serengeti. Am Nat. 2009;173:431–45.31. Sinclair ARE, Mduma SAR, Brashares J. Patterns of predation in a diversepredator–prey system. Nature. 2003;425:288–90.32. Frair JL, Fieberg J, Hebblewhite M, Cagnacci F, DeCesare NJ, Pedrotti L.Resolving issues of imprecise and habitat-biased locations in ecologicalanalyses using GPS telemetry data. Philos Trans R Soc Lond B Biol Sci. 2010;365:2187–200.33. Kie JG, Matthiopoulos J, Fieberg J, Powell RA, Cagnacci F, Mitchell MS, GaillardJ-M, Moorcroft PR. The home-range concept: are traditional estimators stillrelevant with modern telemetry technology? Philos Trans R Soc Lond B BiolSci. 2010;365:2221–31.34. Seaman DE, Powell RA. An evaluation of the accuracy of kernel densityestimators for home range analysis. Ecology. 1996;77:2075–85.35. Fieberg J. Kernel density estimators of home range: smoothing and theautocorrelation red herring. Ecology. 2007;88:1059–66. doi:10.1890/06-0930.36. Worton BJ. Kernel methods for estimating the utilization distribution inhome-range studies. Ecology. 1989;70:164–8. doi:10.2307/1938423.37. Stone CJ. An asymptotically optimal window selection rule for kerneldensity estimates. Ann Stat. 1984;12:1285–97. doi:10.1214/aos/1176346792.38. Jones MC, Marron JS, Sheather SJ. A brief survey of bandwidthselection for density estimation. J Am Stat Assoc. 1996;91:401–7.doi:10.2307/2291420.39. Kie JG, Bowyer RT, Boroski BB, Nicholson MC, Loft ER. Landscapeheterogeneity at differing scales: effects on spatial distribution of mule deer.Ecology. 2002;83:530–44. doi:10.189/0012-9658.40. Calenge C. The package adehabitat for the R software: a tool for theanalysis of space and habitat use by animals. Ecol Model. 2006;197:516–19.41. Jennrich RI, Turner FB. Measurement of non-circular home range. J TheorBiol. 1969;22:227–37.42. Benhamou S, Valeix M, Chamaillé-Jammes S, Macdonald DW, Loveridge AJ.Movement-based analysis of interactions in African lions. Anim Behav. 2014;90:171–80.43. Bowler DE, Benton TG. Causes and consequences of animal dispersalstrategies: relating individual behaviour to spatial dynamics. Biol Rev. 2005;80:205–25.44. Bolnick DI, Svanbäck R, Fordyce JA, Yang LH, Davis JM, Hulsey CD, ForisterML. The ecology of individuals: incidence and implications of individualspecialization. Am Nat. 2003;doi:10.1086/343878.45. Araújo MS, Bolnick DI, Layman CA. The ecological causes of individualspecialization. Ecol Lett. 2011;14:948–58.46. Schoener TW. Resource partitioning. In: Kilkkawa J, Anderson DJ, editors.Community ecology pattern and process. Boston: Blackwell Scientific; 1986.p. 91–126.47. Jarman MV, Jarman PJ. Daily activity of impala. E Afr Wildl J. 1973;11:75–92.48. Walther FR. Round-the-clock-activity of Thomson’s gazelle (Gazellathomsoni Gunther 1884) in Serengeti National Park. Zool Tierpsychol.1973;32:75–105.49. Berry HH, Siegfried WR, Crowe TM. Activity patterns in a population of free-ranging wildebeest Connachaetes taurinus at Etosha National Park. ZoolTierpsychol. 1982;59:229–46.50. Owen-Smith N, Goodall V. Coping with savanna seasonality: comparativedaily activity patterns of African ungulates as revealed by GPS telemetry.J Zool (Lond). 2014;293:181–91.51. Bukombe J, Senzota RB, Fryxell JM, Kittle AM, Kija H, Hopcraft JGC, MdumaS, Sinclair ARE. Do animal size, seasons and vegetation type influencedetection probability and density estimates of Serengeti ungulates? Afr JEcol. 2015; doi:10.1111/aje.12255.52. Scheel D. Profitability, encounter rates and prey choice of African lions.Behav Ecol. 1993;4:90–7.53. Reed DN, Anderson TM, Dempewolf J, Metzger K, Serneels S. The spatialdistribution of vegetation types in the Serengeti ecosystem: the influence ofrainfall and topographic relief on vegetation patch characteristics. JBiogeogr. 2009;36:770–82.54. Dormann CF, Elith J, Bacher S, Buchmann C, Carl G, Carré G, Marquéz JRG,Gruber B, 845 Lafourcade B, Leitão PJ, Münkemüller T, McClean C, OsbornePE, Reineking B, Schröder B, Skidmore AK, Zurell D, Lautenbach S.Collinearity: a review of methods to deal with it and a simulation studyevaluating their performance. Ecography. 2013;36:27–46.55. Manly BFJ, McDonald LL, Thomas DL, McDonald TL, Erickson WP. Resourceselection by animals. 2nd ed. Dordrecht: Kluwer; 2002.56. Crawley M. The R, Book. Chichester: John Wiley & Sons, Ltd.; 2007.57. Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM. Mixed effects modelsand extensions in ecology with R. New York: Springer; 2009.58. Tabachnick BG, Fidell LS. Using multivariate statistics. 3rd ed. New York:Harper Collins; 1996.59. Hopcraft JGC, Anderson TM, Pérez-Vila S, Mayemba E, Olff H. Body size andthe division of niche space: food and predation differentially shape thedistribution of Serengeti grazers. J Anim Ecol. 2012;81:201–13.60. Burnham KP, Anderson DR, Huyvaert KP. AIC model selection andmultimodel inference in behavioral ecology: some background,observations, and comparisons. Behav Ecol Sociobiol. 2011;65:23–35.61. Hopcraft JGC, Sinclair ARE, Packer C. Planning for success: Serengeti lionsseek prey accessibility rather than abundance. J Anim Ecol. 2005;74:559–66.62. Elliot JP, Cowan IM, Holling CS. Prey capture by the African lion. Can J Zool.1977;55:1811–28.63. Scheel D. Watching for lions in the grass: The usefulness of scanning and itseffects during hunts. Anim Behav. 1993;46:695–704.64. Johnson BK, Kern JW, Wisdom MJ, Findholt SL, Kie JG. Resource selectionand spatial separation of mule deer and elk during spring. J Wildl Manage.2000;64:685–97.65. D’Eon R, Serrouya R. Mule deer seasonal movements and multiscaleresource selection using global positioning system radiotelemetry. J Mamm.2005;86:736–44.66. Dickson B, Jenness J, Beier P. Influence of vegetation, topography, androads on cougar movement in southern California. J Wildl Manage. 2005;69:264–76.67. Kittle AM, Fryxell JM, Desy GE, Hamr J. The scale-dependent impact of wolfpredation risk on resource selection by three sympatric ungulates.Oecologia. 2008;157:163–75.68. Burnham KP, Anderson DR. Model selection and multi-model interence: Apractical information-theoretic approach. 2nd ed. Berlin: Springer; 2002.69. R Core Team. R: A language and environment for statistical computing.Vienna: R Foundation for Statistical Computing. 2012; ISBN 3-900051-07-0,URL http://www.R-project.org/.70. ESRI. ArcGIS Desktop: Release 10.1. Redlands: Environmental SystemsResearch Institute; 2011.71. Beyer HL. Geospatial Modelling Environment (Version (software).2012. URL: http://www.spatialecology.com/gme.72. Redfern JV, Grant R, Biggs H, Getz WM. Surface water constraints on herbivoreforaging in the Kruger National Park, South Africa. Ecology. 2003;84:2092–107.73. Valeix M. Temporal dynamics of dry-season water-hole use by large Africanherbivores in two years of contrasting rainfall in Hwange National Park,Zimbabwe. J Trop Ecol. 2011;27:163–70.74. Kareiva P, Odell G. Swarms of predators exhibit “preytaxis” if individualpredators use area-restricted search. Am Nat. 1987;130:233–70.75. Valeix M, Loveridge AJ, Davidson Z, Madzikanda H, Fritz H, Macdonald DW.How key habitat features influence large terrestrial carnivore movements:waterholes and African lions in a semi-arid savanna of north-westernZimbabwe. Landsc Ecol. 2010;25:337–51.76. Crosmary W-G, Valeix M, Fritz H, Madzikanda H, Côté SD. African ungulatesand their drinking problems: hunting and predation risks constrain accessto water. Anim Behav. 2012;83:145–53.77. Foster WA, Treherne JE. Evidence for the dilution effect in the selfish herdfrom fish predation on a marine insect. Nature. 1981;295:466–7.78. Olson RS, Hintze A, Dyer FC, Knoester DB, Adami C. Predator confusion issufficient to evolve swarming behaviour. J R Soc Inter. 2013; doi:10:20130305 http://dx.doi.org/10.1098/rsif.2013.0305.79. Murray M. Specific nutrient requirements and migration of wildebeest. In:Sinclair ARE, Arcese P, editors. Serengeti II: Dynamics, management andconservation of an ecosystem. Chicago: University of Chicago Press; 1995. p.231–56.80. Davidson Z, Valeix M, Loveridge AJ, Hunt JE, Johnson PJ, Madzikanda H,Macdonald DW. Environmental determinants of habitat and kill siteselection in a large carnivore: scale matters. J Mamm. 2012;93:677–85.81. Kolowski JM, Holekamp KE. Ecological and anthropogenic influences onspace use by spotted hyaenas. J Zool (Lond). 2009;277:23–36.Kittle et al. Movement Ecology  (2016) 4:17 Page 17 of 1882. Johnson DH. The comparison of usage and availability measurements forevaluating resource preference. Ecology. 1980;61:65–71.83. Boyce MS. Scale for resource selection functions. Divers Distrib. 2006;12:269–76.84. Durant SM. Competition refuges and coexistence : an example fromSerengeti carnivores. J Anim Ecol. 1998;67:370–86.85. Broekhuis F, Cozzi G, Valeix M, McNutt JW, Macdonald DW. Risk avoidancein sympatric large carnivores: reactive or predictive? J Anim Ecol. 2013;82:1098–105.86. Odden M, Wegge P, Fredriksen T. Do tigers displace leopards? If so, why?Ecol Res. 2010;25:875–81.87. Powell GVN. Experimental analysis of the social value of flocking by starlings(Sturnus vulgaris) in relation to predation and foraging. Anim Behav. 1974;22:501–5.88. Fitzgibbon CD. Mixed-species grouping in Thompson’s and Grant’s gazelles:the antipredator benefits. Anim Behav. 1990;39:1116–26.89. Creel S, Winnie JA. Responses of elk herd size to fine-scale spatial andtemporal variation in the risk of predation by wolves. Anim Behav.2005;69:1181–9.90. Ripple WJ, Beschta RL. Wolf reintroduction, predation risk, and cottonwoodrecovery in Yellowstone National Park. For Ecol Manage. 2003;184:299–313.91. Jenner N, Groombridge J, Funk SM. Commuting, territoriality and variation ingroup and territory size in a black-backed jackal population reliant on aclumped, abundant food resource in Namibia. J Zool (Lond). 2011;284:231–8.92. Hofer H, East ML. The commuting system of Serengeti spotted hyenas: howa predator copes with migratory prey. I Social organization. Anim Behav.1993;46:547–57.93. Burkepile DE, Burns CE, Tambling CJ, Amendola E, Buis GM, Govender N,Nelson V, Thompson DI, Zinn AD, Smith MD. Habitat selection by largeherbivores in a southern African savanna: Relative roles of bottom-up andtop-down forces. Ecosphere. 2013;4:139. http://dx.doi.org/10.1890/ES13-00078.1.94. Packer C, Swanson A, Ikanda D, Kushnir H. Fear of darkness, the full moonand the nocturnal ecology of African Lions. PLoS One. 2011;6(7):e22285.doi:10.1371/journal.pone.0022285.95. Kittle AM, Anderson M, Avgar T, Baker JM, Brown GS, Hagens J, IwachewskiI, Moffatt S, Mosser A, Patterson BR, Reid DEB, Rodgers AR, Shuter J, StreetGM, Thompson ID, Vander Vennen LM, Fryxell JM. Wolves adapt territorysize, not pack size to local habitat quality. J Anim Ecol. 2015;84:1177–86.doi:10.1111/1365-2656.12366.96. Avgar T, Baker JM, Brown GS, Hagens J, Kittle AM, Mallon EE, McGreer M,Mosser A, Newmaster SG, Patterson BR, Reid DEB, Rodgers AR, Shuter J,Street GM, Thompson ID, Turetsky M, Wiebe PA, Fryxell JM. Space-usebehavior of woodland caribou based on cognitive movement model.J Anim Ecol. 2015;84:1059–70. doi:10.1111/1365-2656.12357.97. McGreer MT, Mallon EE, Vander Vennen LM, Wiebe PA, Baker JA, Brown GS,Avgar T, Hagens J, Kittle AM, Mosser A, Street GM, Reid DEB, Rodgers AR,Shuter J, Thompson ID, Turetsky MJ, Newmaster SG, Patterson BR, FryxellJM. Selection for forage and avoidance of risk by woodland caribou(Rangifer tarandus caribou) at coarse and local scales. Ecosphere. 2015; 6:art288. DOI:10.1890/ES15-00174.1.98. Estes RD. The behavior guide to African mammals: Including hoofedmammals, carnivores, primates. Berkeley: University of California Press; 1991.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:Kittle et al. Movement Ecology  (2016) 4:17 Page 18 of 18


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items