Biological Conservation 173 (2014) 1–9

Contents lists available at ScienceDirect

Biological Conservation

journal homepage: www.elsevier .com/ locate /biocon

Brown bear circadian behavior reveals human environmentalencroachment

http://dx.doi.org/10.1016/j.biocon.2014.03.0060006-3207/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Department of Ecology and Natural ResourceManagement, Norwegian University of Life Sciences, Postbox 5003, NO-1432 Ås,Norway. Tel.: +47 34689987062.

E-mail address: andres.ordiz@gmail.com (A. Ordiz).

Andrés Ordiz a,b,⇑, Jonas Kindberg c, Solve Sæbø d, Jon E. Swenson a,e, Ole-Gunnar Støen a,c

a Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, Postbox 5003, NO-1432 Ås, Norwayb Grimsö Wildlife Research Station, Department of Ecology, Swedish University of Agricultural Sciences, SE-730 91 Riddarhyttan, Swedenc Department of Wildlife, Fish, and Environmental Studies, Swedish University of Agricultural Sciences, SE-901 83 Umeå, Swedend Department of Chemistry, Biotechnology and Food Science, Norwegian University of Life Sciences, Postbox 5003, NO-1432 Ås, Norwaye Norwegian Institute for Nature Research, Postbox 5685 Sluppen, NO-7485 Trondheim, Norway

a r t i c l e i n f o

Article history:Received 13 September 2013Received in revised form 6 March 2014Accepted 8 March 2014

Keywords:BehaviorConservationLarge carnivoresManagementRoadsUrsus arctos

a b s t r a c t

Large carnivores adjust their daily movement patterns in response to environmental factors and/or humandisturbance, and often respond differently across their distribution range. Whether such behavioralplasticity is due to environmental or anthropogenic factors has not yet been fully clarified. Beyond largecarnivore conservation and management, understanding behavioral changes in the movement patternsof these elusive species may prove useful to evaluate anthropogenic influences on ecosystems. We used696 318 GPS locations from 105 radio-collared brown bears in 3 study areas in Sweden to construct dailybear movement patterns, calculating the distance traveled by the bears every 30 min. We used a Bayesianapproach to analyze whether human and/or road density around bear locations could explain observeddifferences in bear movement patterns among the areas. Proximity to settlements, a proxy of the generallylow human density in Scandinavian bear range, did not influence circadian bear movements. However,bears moved most in the nocturnal and twilight hours and less during daytime in areas with higher roaddensity, compared to roadless areas. Human-caused behavioral changes in large carnivores may havepotential ecosystem-level consequences, given the key ecological role that these species can play inecosystems. Limiting the creation and use of roads is necessary to maintain large carnivore distributionranges and movement corridors, reduce human-caused mortality, and minimize human-induced distur-bance that modifies carnivore behavior.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Animal behavior plays an important role in shaping ecologicalprocesses, including species’ distribution, abundance, and popula-tion dynamics (Sih et al., 2012). Behavioral responses, like changesin movement patterns or habitat use, are often the first measurablereactions that animals show to human-induced environmentalchanges, and can help determine a species’ capacity to adapt to thesechanges (Tuomainen and Candolin, 2011; Sih et al., 2011, 2012).Therefore, changes in daily movement patterns of elusive species,such as large carnivores, may be used as an indicator of the degreeof environmental stress caused by anthropogenic influence onecosystems (Seryodkin et al., 2013).

Variation in large carnivore behavior across continents, such asdifferences in daily activity and movement patterns, has beendiscussed in relation to their history of human persecution. Whereashuman persecution of large carnivores has lasted many centuries inEurope, it has been more recent, intense and efficient in NorthAmerica (Frank and Woodroffe, 2001). At a transcontinental scale,North American brown bears (Ursus arctos) and wolves (Canis lupus)are primarily diurnal (Munro et al., 2006; Mech, 1992), but theyadjust their spatio-temporal use of areas with higher human activity(Hebblewhite and Merrill, 2008). However, European brown bearsand wolves, the same species as in North America, show twilightor nocturnal activity periods, probably to avoid humans (Vilá et al.,1995; Ciucci et al., 1997; Theuerkauf et al., 2003; Ordiz et al., 2012,2013a). Other large carnivores also adjust their movements inrelation to humans. For example, mountain lions (Puma concolor)became more nocturnal when human activity increased (Van Dykeet al., 1986) and coyotes (Canis latrans) resumed diurnality afterhuman persecution ceased (Kitchen et al., 2000).

2 A. Ordiz et al. / Biological Conservation 173 (2014) 1–9

Nevertheless, some authors argue that factors like temperatureor prey activity, not humans, may cause nocturnal behavior inNorth American and European wolves (Mech, 1992; Theuerkaufet al., 2007); and others ask for evidence of the causal relationshipbetween large carnivore nocturnal behavior and levels of distur-bance faced by individual animals (Kaczensky et al., 2006). There-fore, clarifying this issue is important in a behavioral ecologycontext. If human density has a major effect on large carnivorebehavior, we may be able to use it as an indicator of the degreeof anthropogenic influence on the environment.

Woodroffe (2000) found positive associations between humandensity and large carnivore extinctions. However, nocturnalbehavior may have allowed large carnivores to survive in quitehumanized European areas, whereas government-sponsoredpersecution eradicated them in some areas with few people(Woodroffe, 2000). Thus, management policy may be more impor-tant than human density to ensure large carnivore persistence(Linnell et al., 2001).

As an alternative to human density, distance to roads or roaddensity have been proposed as the best proxy for the effects ofhuman land use on wildlife, e.g. resource extraction and exporta-tion, and/or increased human presence (Trombulak and Frissell,2000). Species with large movement ranges, low reproductive rates,and low densities, all typical characteristics of large carnivores, areexpected to respond negatively to roads (Fahrig and Rytwinski,2009). Some large carnivores, such as mountain lions or wolves,can use dirt or small roads for traveling (e.g., Dickson et al., 2005).However, negative effects of roads in terms of spatial avoidanceand reduced survival have been reported for a variety of species,including mountain lions (Belden and Hagedorn, 1993; Dicksonet al., 2005), wolves (Whittington et al., 2005), jaguars (Pantheraonca) (Colchero et al., 2011), brown bears (e.g., Mace et al., 1996;Northrup et al., 2012), and tigers (Panthera tigris) (Kerley et al.,2002). Distinguishing the demographic effects of roads, such asincreased mortality, from behavioral responses is still needed(Fahrig and Rytwinski, 2009) and this is of special interest, consid-ering the role that large carnivores play in some ecosystems (e.g.Estes et al., 2011; Ordiz et al., 2013b; Ripple et al., 2014).

Brown bears, like many other large carnivores, are generallythreatened by human-caused mortality, habitat loss, and fragmen-tation (Servheen et al., 1999) and avoid human activities throughouttheir range (e.g., Mace et al., 1996; Nellemann et al., 2007). The bearpopulation in Sweden has been growing (Kindberg et al., 2011), but,as is also common among large carnivores,�90% of bear mortality iscaused by people (Bischof et al., 2009). We analyzed the daily move-ment patterns of 105 GPS-collared brown bears inhabiting areaswith different human and road densities in Sweden. We hypothe-sized that bears would travel longer distances during daytime inareas with fewer people and/or roads, whereas they would movemost at twilight-nocturnal hours elsewhere. This study, with highlydetailed spatial and temporal data, may help document the influ-ence of human factors on variations in large carnivore behavior,both at local and continental scales. This may illustrate the utilityof behavioral studies to measure human-induced environmentalstress on large mammals and the ecosystems they inhabit.

2. Material and methods

2.1. Study areas

We used data from three study areas in Sweden (Fig. 1). Thesouthern study area is separated from the two northern areas by600 km. The southern area (hereafter, ‘‘South’’) (61�N, 15�E) has arolling landscape of coniferous forest, mainly Scots pine (Pinussylvestris) and Norway spruce (Picea abies), with elevations from

200 to 1000 m. The northeastern area (‘‘Northeast’’) (67�N,17�E) is similar, whereas the ‘‘Northwest’’ area reachedaltitudes of 2000 m, partially included Sarek National Park, andalso has a subalpine forest of birch (Betula pubescens) and willows(Salix spp.).

In 2011, human density was 4–7 habitants/km2 in the South,and 0.3–1.2 habitants/km2 in the northern areas (StatisticsSweden, 2012). Logging is intense in the coniferous forests,including the South and Northeast study areas, with many roads(1 ± 0.5 km/km2 –mean and SD–, range 0–4.6 km/km2), whereasthe western part of northern Sweden has very few roads (Fig. 1).Indeed, there are no houses or roads in and around Sarek NationalPark (Fig. 1b). Husbandry of free-ranging, semidomestic reindeer(Rangifer tarandus) is a major human activity in the northern areas.Bears are hunted in Sweden with annually established quotas, butare legally protected inside national parks.

2.2. Bear data and statistics

We used GPS data recorded from 2008 to 2011 from 39 malesand 66 female brown bears: 34 males and 44 females in the South,5 males and 8 females in Northeast, and 14 females in Northwest.Thirty-two of the females had dependent cubs in some years. Bearshad GPS–GSM collars (VECTRONIC Aerospace GmbH, Berlin,Germany) and a VHF transmitter implant (IMP 400L, Telonics,USA). Details on capturing and marking are available in Arnemoet al. (2011). GPS receivers have accuracy within 5 m of the trueposition under open sky conditions, and within 10 m under closedcanopies (Wing et al., 2005).

We used GPS positions, recorded every 30 min, to constructdaily movement patterns by calculating the distance traveledbetween consecutive positions during the 24 h. We used data fromJuly to September, i.e. the hyperphagia season when bears feedcopiously to gain fat before hibernation. Hibernation starts earlier(October) and finishes later (May–June) in the north than in thesouth (Manchi and Swenson, 2005). Mean summer temperaturesin the 3 study areas were similar (11–12.5 �C, Statistics Sweden,2012).

We analyzed the data with a Bayesian approach for threereasons: (1) The models were quite complex, including bothrandom effects (bear effects) and autoregressive terms (temporaleffects), (2) the large number of missing values (calculating30-min distance traveled by bears was not possible when GPS loca-tions were missing) are handled elegantly by data augmentation(treating missing values as unknowns to be predicted), and (3)the straightforward availability of estimates of derived parameters(any function of the model parameters) from the Markov ChainMonte Carlo (MCMC) runs (details below). These issues couldprobably be handled in a non-Bayesian way, but this would requireeither simplification of data, data imputation, or implementing theEM-algorithm (Sundberg, 1974; Dempster et al., 1977) to deal withmissing values. However, this would not be a straightforward taskfor the complex linear models used here. We used a linear modelto explore how the response variable, y (square root of distancetraveled by bears; we transformed the data using the square rootto make them normally distributed), was influenced by the factorsbear ID, age, time interval (48 levels, every 30 min), study area (3levels), proximity to settlements, road density, sex class (3 levels),and daylight.

We included proximity to settlements and road densities in themodel as continuous variables. The distance to settlements aroundbear locations was calculated as the Euclidean distance from everyGPS location to the edge of the closest village or town within 50 kmusing Geographic Information Systems (GIS). We calculated roaddensity as the average length of roads per km2 with a movingwindow and the topographic map (GSD-vägkartan, National Land

Fig. 1. (a) Brown bear study areas in Northwest, Northeast and Southern Sweden (national and county borders are shown), as defined by the locations of the 105 GPS-radio-collared bears included in the study; gray patches are bear GPS positions in the Northeast, black patches are bear GPS positions in the Northwest and South; (b) and (c) roads(thin lines) in the northern (b) and southern (c) study areas (road density is highest in the south). Diagonal lines are National Parks in the north. A gray-shading scale reportsnumber of resident people; there were 3–1100 people (median = 10) in each 1 � 1 km square (0.25 � 0.25 km squares inside towns).

A. Ordiz et al. / Biological Conservation 173 (2014) 1–9 3

Survey of Sweden, license i2012/901). We used ArcGIS 10.1 (ESRI)for all geographic analyses. Hunting regulations allow shootingmales and solitary females, but protect females with cubs, whichare more diurnal than solitary individuals (Ordiz et al., 2007,2012). Therefore we included these 3 levels of sex classes in themodel (1 = male, 2 = solitary female, 3 = female with cubs). We in-cluded the potential effect of daylight on bear activity through thestudy period, accounting for latitude differences between northernand southern study areas.

The linear model used in the analyses can be expressed as fol-lows, where y is the distance traveled (in square root meters) every30 min for a given individual bear. We modeled the effect of hu-man density in three different ways, yielding three differentmodels:

(1) Categorical effect of study area as a proxy for humandensity:

yijklm ¼ kij þ ak þ gm þ b1jxday þ b2xage þ 2ijklm

where the parameter kij is the effect of daily half-hour interval j(j = 1, 2, . . . , 48) in study area i (i = 1, 2, 3). Further, ak is the ran-dom effect of individual k (k = 1, 2, . . . 105) assumed to be dis-tributed as Nð0;r2

bearÞ, gjm is the time dependent effect of sex-class m, xday is the continuous measure of daylight hours (meancentered) with regression coefficient b1j being dependent on thetime interval j, xage is the age of the bear with regression coeffi-cient b2, and � is the noise term accounting for unexplained var-iation in the traveled distances and assumed distributed as

4 A. Ordiz et al. / Biological Conservation 173 (2014) 1–9

Nð0;r2j Þ, that is, normal distributed with time of day dependent

variance.(2) Proximity to settlements as a continuous proxy of human

density:yijklm ¼ kj þ ak þ gjm þ b1jxday þ b2xday þ b3jxhum þ �ijklm

where xhum is the Euclidian distance to the closest human settle-ment, as described above, with time of day dependent regressioncoefficient b3j.

(3) Road density as a continuous proxy for human disturbance:yijklm ¼ kj þ ak þ gjm þ b1jxday þ b2xage þ b4jxroad þ �ijklm

where xroad is the length of roads per km2 with regression coef-ficient b4j, also time dependent.For both Models 2 and 3, the time of day effect kj is independent

of study area i, because the study area information is not includedin these models, as we assumed that human density and road den-sity accounted for the extra difference between study areas, whendaylight effect had been accounted for. For all models, we modeleda temporal dependence between the effect levels due to time ofday in the following way:

kj ¼ m � kj�1 þ �j for j ¼ 2; . . . ;48k1 � Nð0;1000Þ�j � Nð0; s2Þ

where m is the autoregressive coefficient, and the variance compo-nent s2 can be regarded as a smoothing parameter. A large valuewill induce low degree smoothing, whereas a small value will givehigh level smoothing of the time effect. Due to the large amount ofdata, the time effects were quite accurately estimated in this study,hence a low degree of smoothing was necessary and s2 = 1000 wasused in all analyses.

For all regression coefficients of continuous effects and theparameter m, we used vague normal distributions N(0, 1000) asprior distributions. For the sex class effect, we set the effect ofthe first level (male) to zero, whereas the vague normal priors wereused for the two remaining levels. For all variance components inthe model, we assumed a priori that the inverse variance (the pre-cision) was gamma distributed with Ga(0.0001, 0.0001), which is adistribution with expectation 1 but with a large variance.

Fig. 2. Daily brown bear movement patterns in 3 study areas in Sweden. Time of day in a(midpoint of our study period). Curves represent movement for a male bear of average agtime effects kij (solid lines) and the 95% credible intervals (dashed lines) are shown.

We estimated the unknown model parameters by Bayesian pos-terior means using Markov Chain Monte Carlo (MCMC) methods,implemented in WinBUGS (Lunn et al., 2000). All models were wellformulated and convergence was obtained within a burn-in of5000 iterations (visually confirmed from trace plots). Upon conver-gence, we ran a further set of 15,000 iterations with a thinning of10, yielding 1500 samples for each parameter as basis for posteriorestimation.

The estimated posterior distributions for the model parametersprovided point estimates (mean) and credible intervals (lower 2.5%and upper 97.5% percentiles of the estimated distribution). Weconsidered effects to be statistically significant if the credibleintervals of the corresponding parameters did not contain the zerovalue (see Figs. 2–6) (e.g., Kruschke, 2011). The MCMC approach forparameter estimation is an iterative process allowing the missingvalues to be predicted by the given model and the current estimatesof the unknown model parameters (data augmentation). Anotheradvantage of adopting a Bayesian approach with the MCMCestimation method is the straightforward possibility to obtainposterior mean estimates and uncertainty intervals for anycombination of the main model parameters. For Model 1 we usedthis to estimate time dependent contrasts (with credible intervals)between the study areas to study how the differences between theareas depend on the time of day.

3. Results

In total, we used 696 318 GPS locations from 105 bears during2008–2011 to calculate distance traveled by the bears every30 min. The average distance traveled during each half-hour was238 meters for males, 211 for solitary females, and 179 for femaleswith cubs. The results from Model 1 showed differences in the bears’daily movement patterns among the 3 study areas (Fig. 2), withbears traveling longer distances during daytime in the Northwest,longer distances at night and early morning in the Northeast, andlonger distances at twilight hours in the South (estimates ofpairwise differences of time effects kij � ki0j between study areas iand i0 are given in Fig. 3 for each time j). In the analyses includinghuman-density variables, proximity to settlements (in Model 2)did not significantly affect daily movement patterns (Fig 4a).

ll Figures is GMT time, with sunrise at �03:30 and sunset at �18:30 by mid-Auguste on a day with average daylight length in each of the study areas. The mean of the

Fig. 3. Compared differences in distance traveled (in square root meters) by brown bears at 30-min intervals during 24 h in 3 study areas in Sweden. Differences at 30-minintervals were considered significant when the posterior mean (central line) and the 95% credible intervals (external lines) were all above or below the 0 line. (a) Curvesrepresent differences between the mean of the distance traveled (and 95% credible intervals) between bears in the Northwest relative to bears in the South, i.e. bears weremore active during daytime in the Northwest than in the South. (b) Difference in the mean of the distance traveled (and 95% credible intervals) between bears in theNorthwest relative to bears in the Northeast; i.e. bears were more active during daytime in the Northwest. (c) Curves represent differences in the mean of the distancetraveled (and 95% credible intervals) between bears in Northeast relative to bears in South, i.e. bears were more active during night and early morning in Northeast than inSouth.

A. Ordiz et al. / Biological Conservation 173 (2014) 1–9 5

However, road density had a negative effect during daytime andmost bear movements occurred in the nocturnal and twilight hoursin areas with higher road density (Model 3; see Fig 4b and c). Males

and solitary females had similar movement patterns, whereas fe-males with cubs moved more during the middle of the day and lessduring night (Fig. 5), as previously reported (see Section 2.2. in

Fig. 4. Curves represent the posterior mean estimate (central line) and the 95% credible intervals (external lines) for the time-dependent linear effects of (a) proximity tosettlements ðb̂3jÞ, and (b) road density ðb̂4jÞ on expected brown bear movement across the three study areas in Sweden. Proximity to settlements had no significant effect onexpected bear movement, because the zero-line is well inside the credible intervals for all time points. Hence, the variable xhum appeared to have no effect at any time of theday. (b) For time points where all curves are positive, increasing road density ðxroadÞ gives a significant increase in expected bear movement. Thus, (c) with increasing roaddensity across the study areas, bears moved most during late evening-night and early morning and less during daytime. Curves represent movement for a male bear ofaverage age (8 years old) on a day with average daylight length (963 min).

6 A. Ordiz et al. / Biological Conservation 173 (2014) 1–9

Methods, or Ordiz et al., 2007, 2012). The model predicted that bearswould become more diurnal with the shortening daylight length asthe season advanced (estimates of the negative values of b1j inFig. 6).

4. Discussion

We found that brown bear movement was mostly restricted tonocturnal and twilight hours in areas with higher road densities,

Fig. 5. Time-of-day effect estimates on brown bear movements for males, solitary females, and females with cubs (all areas combined). The curves represent the estimatedposterior mean of the distance traveled and the 95% credible intervals (corrected for the other model effects).

Fig. 6. Curves represent the negative value of the posterior mean estimate (central line) and 95% credible intervals (external lines) for the time-dependent linear effects of thedaylight length on brown bear daily movements. A positive value means that, as daylight lengths decrease, the bears become more active at that given time of the day, andvice versa.

A. Ordiz et al. / Biological Conservation 173 (2014) 1–9 7

compared to roadless areas (Fig. 4b and c). Daily movement pat-terns based on distance traveled between GPS locations closelyresembled bear daily activity patterns (activity sensors in the col-lars indicate if a bear is active or resting) in the same area (see‘‘South’’ in Fig. 2 and Moe et al., 2007). This suggests that bearmovement and activity patterns are proxies of each other. Usingone or the other in a given study may be a choice dependent onthe specific research goals and data availability.

Latitude-related differences could explain part of the observedvariation in bear movement patterns between the southern andthe two northern study areas. However, the largest differenceswere between the two northern areas. It has also been arguedfor different bear species that seasonal changes in diet, thermo-energetic regulation related with high-caloric food (hard mast),and/or human disturbance may explain changes in daily move-ment patterns throughout the year (e.g. Beckmann and Berger,2003; Hwang and Garshelis, 2007; Ordiz et al., 2012). Indeed, itis important to apply ecologically sound definitions of seasons inbiological studies (Basille et al., 2013). In an attempt to reduce

confounding factors, such as food and other seasonal changes,e.g. variation in climatic conditions among study areas, weanalyzed bear movements only during the hyperphagia season, ahomogeneous ecological period when Scandinavian brown bearsrely on berries, with no hard mast available (Dahle et al., 1998;Persson et al., 2001).

Regarding disturbance, the lack of influence of human proxim-ity might be a consequence of the generally low human densityand low variation throughout most of the bear range in Sweden.Road density (Fig. 4) appeared to be a better proxy of the influenceof human activities on the behavior of this large carnivore.Forestry, reindeer husbandry, and many outdoors activities, e.g.hunting, fishing, and berry picking, are common in Sweden. Roadsare used extensively for all of these activities, which seems to beessential in explaining why daily bear movement patterns weremost different between the two northern areas. The high roaddensity in the Northeast provides ready human access to bearhabitat, and likely explains why the nocturnal-crepuscular behav-ior of bears is similar in the Northeast and South, compared to the

8 A. Ordiz et al. / Biological Conservation 173 (2014) 1–9

roadless Northwest, where bears were more active during daytime(Fig. 3a). High road density may increase bears’ vulnerability byproviding ready access to hunters, as has been documented inCanada (McLellan and Shackleton, 1988). Thus, high road densityseemed to counteract the switch towards more diurnal behavioras the season advanced (Fig. 6). It has also been shown that bearsreact to the start of hunting seasons by becoming more nocturnal(Ordiz et al., 2012), as bears also do after encountering people(Ordiz et al., 2013a). In the Russian Far East, Seryodkin et al.(2013) also reported a switch to nocturnal activity by brown bearsin areas with higher human activity that was not related toseasonal changes in bear diet.

Human activities and actions to promote large carnivoreconservation, e.g. management policy and its enforcement, maynot only explain carnivores’ persistence better than human densityalone (Linnell et al., 2001), but also appear to drive large carnivorebehavioral reactions. The bears’ increased nocturnal activity in theNortheast may have helped compensate for the reduced activityduring daytime. Bears must gain enough fat to survive the winterbefore hibernation starts, which occurs earlier and lasts longer innorthern than in southern Sweden (see Methods). Thus, reducedmovement during daytime due to human activities, together withlatitude-related factors that shorten the active season of bears andlimit landscape productivity in the north, may make weight gainmost difficult in the Northeast. Bears there were more active atnight, probably to compensate for less daytime feeding (Figs. 2and 3c).

Large carnivores often occur in areas far from roads and withlow human density, and bears are no exception (Nellemannet al., 2007; Falcucci et al., 2009). Areas of high road densities areavoided by wolves (Whittington et al., 2005), constrain homeranges of mountain lions and movements of jaguars (Belden andHagedorn, 1993; Colchero et al., 2011), and decrease survivorshipand reproductive success of tigers (Kerley et al., 2002). We havedocumented that bears show spatio-temporal avoidance of humanactivity during daytime (Ordiz et al., 2011, 2012, 2013a; thisstudy), as do wolves (e.g. Hebblewhite and Merrill, 2008).Sometimes, large carnivores can cope with, and even benefit from,intermediate levels of human disturbance that favor their stapleprey species. That seems to be the case for Eurasian lynx (Lynx lynx)preying on roe deer (Capreolus capreolus), but lynx also avoidedareas with the highest human and road densities (Basille et al.,2009).

This plasticity in large carnivore behavior may have positiveand negative consequences in terms of distribution, survival, andperformance of their ecological function as top predators.Temporal avoidance of human activity may allow large carnivoresto maintain their distribution range, and even expand it, in increas-ingly human-dominated landscapes (e.g. the case for Eurasian lynx,Basille et al., 2009; and brown bears, Kindberg et al., 2011, inrecent decades in Scandinavia). Proximity to humans may alsoshield vulnerable animals from conspecifics, e.g. female brownbears with cubs use areas closer to humans (but not to road infra-structure) to avoid adult males during the mating season (Steyaertet al., 2013). However, most large carnivore mortality is caused byhumans, often in attractive sink habitats close to people (e.g.Delibes et al., 2001). Furthermore, large carnivore avoidance ofhumans can lead to behaviorally induced trophic cascades as aresult of prey using humans as shelter against predation, whichtherefore limits the ecological role of large carnivores, whosedistribution is constrained by people (Hebblewhite et al., 2005;Berger, 2007; Hebblewhite and Merrill, 2008). Also at thepopulation level and adding a historical and global perspective,repeated encounters with humans may not only help explain thedistribution of large carnivores avoiding humans at the landscapescale, but may also explain why large carnivores are mainly diurnal

in remote areas, whereas they are crepuscular and nocturnal inhuman-dominated areas (Woodroffe, 2000; Kolowski et al., 2007;Ordiz et al., 2012; and references therein). This also may amplifythe effects of global warming on conservation by limiting theefficiency with which animals forage on their most important foodsources (Ordiz et al., 2013a). All of these results reinforce the roleof animal behavior in shaping ecological processes, such as distri-bution and abundance (Sih et al., 2012), and suggests that changesin the circadian activity patterns of large carnivores is indeeduseful as a proxy of anthropogenic influences on ecosystems(Seryodkin et al., 2013).

4.1. Conclusions and management recommendations

In a worldwide scenario, where even remote areas are beingencroached by human development, appropriate management tolimit the creation and use of roads and/or to close them after theyhave served their original purpose is necessary to retain high-quality habitat for large carnivores. In addition to managing thedirect impacts of human development on carnivores, i.e., reducingmortality, maintaining movement corridors and genetic connectiv-ity (Whittington et al., 2005; Roever et al., 2010; Proctor et al.,2012), managers should focus on minimizing human-induceddisturbance that modifies large carnivore behavior (e.g., Northrupet al., 2012; Coleman et al., 2013). Such behavioral reactions mayalso cause fitness disadvantages for the affected animals. Theseissues require future research, because changed large carnivorebehavior can potentially influence the ecosystems in whichlarge carnivores play key ecological roles.

Acknowledgments

The Scandinavian Brown Bear Research Project (SBBRP) isfunded by the Swedish Environmental Protection Agency, TheNorwegian Environment Agency, Swedish Association for Huntingand Wildlife Management, the Austrian Science Foundation, andthe Research Council of Norway (RCN). We also thank the ReindeerPredation project that, in collaboration with the Wildlife DamageCentre, is funded by the Swedish Ministry of Rural Affairs. A.O.was funded by RCN. The captures of the bears were approved bythe Swedish Environmental Protection Agency (permit Dnr412-7327-09 Nv) and by the Ethical Committee, Djuretiskanämnden, in Uppsala, Sweden (permit C 47/9). This is paper no.161_ from the SBBRP.

References

Arnemo, J.M., Evans, A., Fahlman, Å., 2011. Biomedical Protocols for Free-rangingBrown Bears, Wolves, Wolverines and Lynx. Hedmark University College,Evenstad, Norway. <http://wolverinefoundation.org/wp-content/uploads/2011/02/bpfrws11.pdf>.

Basille, M., Herfindal, I., Santin-Janin, H., Linnell, J.D.C., Odden, J., Andersen, R.,Hogda, K.A., Gaillard, J.M., 2009. What shapes Eurasian lynx distribution inhuman dominated landscapes: selecting prey or avoiding people? Ecography32, 683–691.

Basille, M., Fortin, D., Dussault, C., Ouellet, J.-P., Courtois, R., 2013. Ecologically baseddefinition of seasons clarifies predator–prey interactions. Ecography 36, 220–229.

Beckmann, J.P., Berger, J., 2003. Rapid ecological and behavioural changes incarnivores: the responses of black bears (Ursus americanus) to altered food. J.Zool. 261, 207–212.

Belden, R.C., Hagedorn, B.W., 1993. Feasibility of trans–locating panthers intonorthern Florida. J. Wildlife Manage. 57, 388–397.

Berger, J., 2007. Fear, human shields and the redistribution of prey and predators inprotected areas. Biol. Lett. 3, 620–623.

Bischof, R., Swenson, J.E., Yoccoz, N.G., Mysterud, A., Gimenez, O., 2009. Themagnitude and selectivity of natural and multiple anthropogenic mortalitycauses in hunted brown bears. J. Anim. Ecol. 78, 656–665.

Ciucci, P., Boitani, J., Francisci, F., Andreolli, G., 1997. Home-range, activity andmovements of a wolf pack in central Italy. J. Zool. 243, 803–819.

A. Ordiz et al. / Biological Conservation 173 (2014) 1–9 9

Colchero, F., Conde, D.A., Manterola, C., Chávez, C., Rivera, A., Ceballos, G., 2011.Jaguars on the move: modeling movement to mitigate fragmentation from roadexpansion in the Mayan Forest. Anim. Conserv. 14, 158–166.

Coleman, T.H., Schwartz, C.C., Gunther, K.A., Creel, S., 2013. Grizzly bear and humaninteraction in Yellowstone National Park: an evaluation of bear managementareas. J. Wildlife Manage. 77, 1311–1320.

Dahle, B., Sorensen, O.J., Wedul, E.H., Swenson, J.E., Sandegren, F., 1998. The diet ofbrown bears Ursus arctos in central Scandinavia: effect of access to free-rangingdomestic sheep Ovies aries. Wildlife Biol. 4, 147–158.

Delibes, M., Gaona, P., Ferreras, P., 2001. Effects of an attractive sink leading intomaladaptive habitat selection. Am. Nat. 158, 277–285.

Dempster, A.P., Laird, N.M., Rubin, D.B., 1977. Maximum likelihood from incompletedata via the EM algorithm. J. Roy. Stat. Soc. B 39, 1–38.

Dickson, B.G., Jenness, J.S., Beier, P., 2005. Influence of vegetation, topography, androads on cougar movement in southern California. J. Wildlife Manage. 69, 264–276.

Estes, J.A. et al., 2011. Trophic downgrading of planet earth. Science 333, 301–306.Fahrig, L., Rytwinski, T., 2009. Effects of roads on animal abundance: an empirical

review and synthesis. Ecol. Soc. 14, 21.Falcucci, A., Ciucci, P., Maiorano, L., Gentile, L., Boitani, L., 2009. Assessing habitat

quality for conservation using an integrated occurrence–mortality model. J.Appl. Ecol. 46, 600–609.

Frank, L.G., Woodroffe, R., 2001. Behaviour of carnivores in controlled and exploitedpopulations. In: Gittleman, J.L., Wayne, R.K., Macdonald, D.W., Funk, S.M. (Eds),Carnivore Conservation. Cambridge University Press.

Hebblewhite, M., Merrill, E., 2008. Modelling wildlife–human relationships forsocial species with mixed-effects resource selection models. J. Appl. Ecol. 45,834–844.

Hebblewhite, M., White, C., Nietvelt, C., McKenzie, J., Hurd, T.E., Fryxell, J.M., Bayley,S.E., Paquet, P.C., 2005. Human activity mediates a trophic cascade caused bywolves. Ecology 86, 2135–2144.

Hwang, M.H., Garshelis, D.L., 2007. Activity patterns of Asiatic black bears (Ursusthibetanus) in the Central Mountains of Taiwan. J. Zool. 271, 203–209.

Kaczensky, P., Huber, D., Knauer, F., Roth, H., Wagner, A., Kusak, J., 2006. Activitypatterns of brown bears (Ursus arctos) in Slovenia and Croatia. J. Zool. 269, 474–485.

Kerley, L., Goodrich, J.M., Miquelle, D.G., Smirnov, E.N., Quigley, H.B., Hornocker,M.G., 2002. Effects of roads and human disturbance on Amur tigers. Conserv.Biol. 16, 97–108.

Kindberg, J., Swenson, J.E., Ericsson, G., Bellemain, E., Miquel, C., Taberlet, P., 2011.Estimating population size and trends of the Swedish brown bear Ursus arctospopulation. Wildlife Biol. 17, 114–123.

Kitchen, A.M., Gese, E.M., Schauster, E.R., 2000. Changes in coyote activity patternsdue to reduced exposure to human persecution. Can. J. Zool. 78, 853–857.

Kolowski, J.M., Katan, D., Theis, K.R., Holekamp, K.E., 2007. Daily patterns of activityin the spotted hyena. J. Mammal. 88, 1017–1028.

Kruschke, J.K., 2011. Doing Bayesian data analysis. A tutorial with R and Bugs.Academic Press–Elsevier, Burlington, USA, ISBN 978-0-12-381485-2.

Linnell, J.D.C., Swenson, J.E., Andersen, R., 2001. Predators and people: conservationof large carnivores is possible at high human densities if management policy isfavourable. Anim. Conserv. 4, 345–349.

Lunn, D.J., Thomas, A., Best, N., Spiegelhalter, D., 2000. WinBUGS – a Bayesianmodelling framework: Concepts, structure, and extensibility. Stat. Comput. 10,325–337.

Mace, R.D., Waller, J.S., Manley, T.L., Lyon, L.J., Zuring, H., 1996. Relationships amonggrizzly bears, roads and habitat in the Swan mountains, Montana. J. Appl. Ecol.33, 1395–1404.

Manchi, S., Swenson, J.E., 2005. Denning behaviour of Scandinavian brown bearsUrsus arctos. Wildlife Biol. 11, 123–132.

McLellan, B.N., Shackleton, D.M., 1988. Grizzly bears and resource–extractionindustries: effects of roads on behaviour, habitat use and demography. J. Appl.Ecol. 25, 451–460.

Mech, L.D., 1992. Daytime activity of wolves during winter in north–easternMinnesota. J. Mammal. 73, 570–571.

Moe, T.F., Kindberg, J., Jansson, I., Swenson, J.E., 2007. Importance of diel behaviorwhen studying habitat selection: examples from female Scandinavian brownbears (Ursus arctos). Can. J. Zool. 85, 518–525.

Munro, R.H.M., Nielsen, S.E., Price, M.H., Stenhouse, G.B., Boyce, M.S., 2006. Seasonaland diel patterns of grizzly bear diet and activity in west-central Alberta. J.Mammal. 87, 1112–1121.

Nellemann, C., Støen, O., Kindberg, J., Swenson, J., Vistnes, I., Ericsson, G., Katajisto, J.,Kaltenborn, B., Martin, J., Ordiz, A., 2007. Terrain use by an expanding brownbear population in relation to age, recreational resorts and human settlements.Biol. Conserv. 138, 157–165.

Northrup, J.M., Pitt, J., Muhly, T.B., Stenhouse, G.B., Musiani, M., Boyce, M.S., 2012.Vehicle traffic shapes grizzly bear behaviour on a multiple-use landscape. J.Appl. Ecol. 49, 1159–1167.

Ordiz, A., Rodríguez, C., Naves, J., Fernández, A., Huber, D., Kaczensky, P., Mertens, A.,Mertzanis, Y., Mustoni, A., Palazón, S., Quenette, P.Y., Rauer, G., Swenson, J.E.,2007. Distance-based criteria to identify minimum number of brown bearfemales with cubs in Europe. Ursus 18, 158–167.

Ordiz, A., Støen, O.-G., Delibes, M., Swenson, J.E., 2011. Predators or prey? Spatio-temporal discrimination of human-derived risk by brown bears. Oecologia 166,59–67.

Ordiz, A., Støen, O.-G., Sæbo, S., Kindberg, J., Swenson, J.E., 2012. Do bears know theyare being hunted? Biol. Conserv. 152, 21–28.

Ordiz, A., Støen, O.-G., Sæbø, S., Sahlén, V., Pedersen, B.E., Kindberg, J., Swenson, J.E.,2013a. Lasting behavioural responses of brown bears to experimental humanencounters. J. Appl. Ecol. 50, 306–314.

Ordiz, A., Bischof, A., Swenson, J.E., 2013b. Saving large carnivores, but losing apexpredators? Biol. Conserv. 168, 128–133.

Persson, I.-L., Wikan, S., Swenson, J.E., Mysterud, I., 2001. The diet of brownbear Ursus arctos in the Pasvik Valley, northeastern Norway. Wildlife Biol. 7,27–37.

Proctor, M.F. et al., 2012. Population fragmentation and inter-ecosystemmovements of grizzly bears in western Canada and the northern UnitedStates. Wildlife Monogr. 180, 1–46.

Ripple, W.J. et al., 2014. Status and ecological effects of the world’s largestcarnivores. Science 343, 1241484. http://dx.doi.org/10.1126/science.1241484.

Roever, C.L., Boyce, M.S., Stenhouse, G.B., 2010. Grizzly bear movements relative toroads: application of step selection functions. Ecography 33, 1113–1122.

Servheen, C., Herrero, H., Peyton, B., 1999. Bears: Status Survey and ConservationAction Plan. IUCN/SSC Bear and Polar Bear Specialists Groups.

Seryodkin, I.V., Kostyria, A.V., Goodrich, J.M., Miquelle, D.G., 2013. Daily activitypatterns of brown bear (Ursus arctos) of the Sikhote Alin mountain range(Primorskiy Krai, Russia). Russ. J. Ecol. 44, 50–55.

Sih, A., Ferrari, M.C.O., Harris, D.J., 2011. Evolution and behavioural responses tohuman-induced rapid environmental change. Evol. Appl. 4, 367–387.

Sih, A., Cote, J., Evans, M., Fogarty, S., Pruitt, J., 2012. Ecological implications ofbehavioural syndromes. Ecol. Lett. 15, 278–289.

Statistics Sweden, 2012. <http://www.scb.se/Pages/SubjectAreaList____250626.aspx>.

Steyaert, S., Swenson, J.E., Kindberg, J., Brunberg, S., Zedrosser, A., 2013. Malereproductive strategy explains spatiotemporal segregation in brown bears. J.Anim. Ecol. 82, 836–845.

Sundberg, R., 1974. Maximum likelihood theory for incomplete data from anexponential family. Scand. J. Stat. 1, 49–58.

Theuerkauf, J., Rouys, S., Jedrzejewski, W., 2003. Selection of den, rendezvous, andresting sites by wolves in the Bialowieza Forest. Poland. Can J. Zool. 81,163–167.

Theuerkauf, J., Gula, R., Pirga, B., Tsunoda, H., Eggermann, J., Brzezowska, B., Rouys,S., Radler, S., 2007. Human impact of wolf activity in the Bieszczady Mountains,SE Poland. Ann. Zool. Fenn. 44, 225–231.

Trombulak, S.C., Frissell, C.A., 2000. Review of ecological effects of roads onterrestrial and aquatic communities. Conser. Biol. 14, 18–30.

Tuomainen, U., Candolin, U., 2011. Behavioural responses to human-inducedenvironmental change. Biol. Rev. 86, 640–657.

Van Dyke, F.G., Brocke, R., Shaw, H., Ackerman, B., Hemker, T., Lindzey, F.G., 1986.Reactions of mountain lions to logging and human activity. J. Wildlife Manage.50, 95–102.

Vilá, C., Urios, V., Castroviejo, J., 1995. Observations on the daily activity patterns inthe Iberian wolf, in Ecology and Conservation of Wolves in a Changing World.In: Carbyn, L.N., Fritts, S.H., Seip, D.R. (Eds), Canadian Circumpolar Institute,Alberta, Occasional Publication No. 35, pp. 335–340.

Whittington, J., St. Clair, C.C., Mercer, G., 2005. Spatial responses of wolves to roadsand trails in mountain valleys. Ecol. Appl. 15, 543–553.

Wing, M.G., Ecklund, A., Kellogg, L.D., 2005. Consumer-grade Global PositioningSystem (GPS) accuracy and reliability. J. Forest. 103, 169–173.

Woodroffe, R., 2000. Predators and people: using human densities to interpretdeclines of large carnivores. Anim. Conserv. 3, 165–173.