Reindeer habitat selection under the risk of brown bear...

November 2016 v Volume 7(11) v Article e015831 v www.esajournals.org

Reindeer habitat selection under the risk of brown bear predation during calving season

Therese R. Sivertsen,1,† Birgitta Åhman,1 Sam M. J. G. Steyaert,2,3 Lars Rönnegård,4 Jens Frank,5 Peter Segerström,5 Ole-Gunnar Støen,2,6 and Anna Skarin1

1Department of Animal Nutrition and Management, Swedish University of Agricultural Sciences, P.O. Box 7024, 750 07 Uppsala, Sweden

2Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, P.O. Box 5003, 1432 Ås, Norway3Department of Environmental and Health Studies, University College of Southeast Norway, NO-3800 Bø, Norway

4Section of Statistics, School of Technology and Business Studies, Dalarna University, 791 88 Falun, Sweden5Grimsö Wildlife Research Station, Department of Ecology, Swedish University of Agricultural Sciences, 730 91 Riddarhyttan, Sweden

6Department of Wildlife, Fish and Environmental Studies, Swedish University of Agricultural Sciences, 901 83 Umeå, Sweden

Citation: Sivertsen, T. R., B. Åhman, S. M. J. G. Steyaert, L. Rönnegård, J. Frank, P. Segerström, O.-G. Støen, and A. Skarin. 2016. Reindeer habitat selection under the risk of brown bear predation during calving season. Ecosphere 7(11):e01583. 10.1002/ecs2.1583

Abstract. The depredation of semi- domesticated reindeer by large carnivores reflects an important human–wildlife conflict in Fennoscandia. Recent studies have revealed that brown bears (Ursus arctos) may kill substantial numbers of reindeer calves (Rangifer tarandus tarandus) in forest areas in Sweden. Several authors have suggested that predation risk is an important driver of habitat selection in wild Rangifer populations where predation is a limiting factor, but little is known about these mechanisms in semi- domesticated populations. We examined the habitat selection of female reindeer in relation to spatial and temporal variations in brown bear predation risk on the reindeer calving grounds and evaluated the simultaneous responses of brown bears and reindeer to landscape characteristics. We used GPS data from 110 reindeer years (97 individuals) and 29 brown bear years (19 individuals), from two reindeer herding districts in the forest area of northern Sweden. Our results did not indicate that reindeer alter their behav-ior in response to spatiotemporal variation in brown bear predation risk, on the scale of the calving range. Instead, we suggest that spatiotemporal behavioral adjustments by brown bears were the main driver of prey–predator interactions in our study system. Contrasting responses by brown bears and reindeer to clear- cuts and young forest indicate that forestry can influence species interactions and possibly yield negative consequences for the reindeer herd. Even if clear- cuts may be beneficial in terms of calf survival, logging activity will eventually cause greater abundance of young regenerating forest, reducing available reindeer habitats and increasing habitat preferred by brown bears. Domestication may have made semi- domesticated reindeer in Fennoscandia less adapted to cope with predators. Areal restrictions, limiting the opportunity for dispersion and escape, possibly make the calves more susceptible to predation. Also, a generally higher population density in semi- domesticated herds compared to wild populations can make dispersion a less efficient strategy and the reindeer calves easier prey. Overall, the lack of ability of the reindeer females to reduce brown bear encounter risk on the scale of the calving range is probably an im-portant reason for the high brown bear predation rates on reindeer calves documented in our study areas.

Key words: brown bears; habitat selection; predation risk; predator–prey interactions; Rangifer; resource selection functions.

Received 25 July 2016; accepted 26 September 2016. Corresponding Editor: Eric M. Gese. Copyright: © 2016 Sivertsen et al. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.† E-mail: [email protected]

http://dx.doi.org/10.1002/ecs2.1583

http://creativecommons.org/licenses/by/3.0/


SIVERTSEN ET AL.

IntroductIon

Knowledge of habitat requirements, along with the mechanisms behind habitat selection pat-terns in animals, is essential for understanding how environmental variation and human activ-ity affect animal populations, as well as inter-actions between species (Wiens et al. 1993, Sih et al. 2012). Habitat selection involves tradeoffs between fulfilling demands for feeding, mating, and parental care, while at the same time reduc-ing the risk of predation and harmful encounters (Sih 1980, Rosenzweig 1991). In a heterogeneous landscape, certain areas and habitat features can be coupled to a higher risk of predation (Laundré et al. 2010). Also, predator activity and the risk of predation may vary over time (Lima and Bednekoff 1999). Spatial and temporal variations in predation risk generate a dynamic landscape of fear, and natural selection should favor prey behaviors that maximize reduction of predation risk in space and time, against related costs such as reduced forage quality (Brown et al. 1999, Lima and Bednekoff 1999, Laundré et al. 2010). Indeed, spatiotemporal adjustments to predation risk have been documented across a broad range of study systems and species (e.g., Holbrook and Schmitt 1988, Heithaus and Dill 2002, Valeix et al. 2009, Laundré 2010, Latombe et al. 2014, Bonnot et al. 2016).

In Fennoscandia, the majority of the reindeer (Rangifer tarandus tarandus) are domesticated, however free- ranging within the herding district borders throughout most of the year. The num-bers of large carnivores in Fennoscandia (i.e., wolves [Canis lupus], brown bears [Ursus arctos], lynx [Lynx lynx], and wolverines [Gulo gulo]) have increased over the last few decades, creating a challenge for the herding industry (Hobbs and Andrén 2012, Åhman et al. 2014). Depredation of reindeer results in both economical and emo-tional strain for the reindeer herders, and man-agement must constantly balance the interests of reindeer husbandry and large carnivore conser-vation (Swenson and Andrén 2005). The brown bear range in Sweden largely overlaps with the reindeer herding area (Appendix S1: Fig. S1). During the ungulate calving season (i.e., spring), ungulate neonates can be an important compo-nent of the bear diet (Adams et al. 1995, Linnell et al. 1995, Nieminen 2010, Karlsson et al. 2012),

with recent studies revealing that brown bears may kill substantial numbers of reindeer calves in forest areas in Sweden (Karlsson et al. 2012). This has triggered a need for greater insight into habitat use and behavioral interactions of semi- domesticated reindeer and brown bears during the reindeer calving and post- calving season in Fennoscandia.

In the presence of large carnivores, Rangifer (caribou and reindeer) females must balance high energy requirements with predator avoidance during the calving period (Bergerud and Page 1987, Barten et al. 2001, Leblond et al. 2016). Most studies of Rangifer parturient behavior and their spatial interactions with predators have been of caribou in North America (e.g., Bergerud et al. 1990, Fancy and Whitten 1991, Rettie and Messier 2001, Gustine et al. 2006, Leclerc et al. 2012). A wide array of studies have reported that cari-bou reduce predation risk by selecting habitats with lower densities of predators and alternate prey species. This includes coarse- scale migra-tory movements to calving grounds above the tree line (Heard et al. 1996, Bergerud et al. 2008, Bastille- Rousseau et al. 2015). Also, the more sedentary forest- dwelling woodland caribou, with ranges overlapping those of their predators throughout the year, have been shown to choose habitat types carrying a lower risk of encounter-ing predators during calving (Rettie and Messier 2000, Mahoney and Virgl 2003, McLoughlin et al. 2005, Latham et al. 2011, Pinard et al. 2012).

Human- caused land- use changes, such as for-est harvesting and road construction, can influ-ence the habitat use of both Rangifer and their predators and, in turn, the vulnerability to pre-dation. For example, forest- dwelling woodland caribou avoid roads and regenerating stands, which are associated with a higher risk of wolf and bear encounters (Rettie and Messier 2000, Hins et al. 2009, Leclerc et al. 2012, Leblond et al. 2016). In fact, recent studies from North America have shown that the responses of woodland caribou to forest management can have fitness consequences as a result of effects on calf vul-nerability to black bear predation (Dussault et al. 2012, Leclerc et al. 2014). Effects of human- caused land- use changes on predator–prey interactions have also been documented for other ungulates, one example being that moose cows with calves in Yellowstone National Park select areas closer


SIVERTSEN ET AL.

to roads to reduce the risk of encountering bears (Berger 2007a).

It is still uncertain, however, how forest- living semi- domesticated reindeer in Fennoscandia, which are subject to herding activities and areal restrictions, respond to predation risk together with anthropogenic disturbance. Moreover, little information exists on the habitat use of sympat-ric populations of semi- domesticated reindeer and brown bears within reindeer calving ranges located in forested areas.

In this study, we used GPS location data to model female reindeer and brown bear habitat selection on the reindeer calving range. The study was carried out in two forest reindeer herding dis-tricts in Northern Sweden, with varying degrees of impact from forest harvesting and roads, and where brown bear predation apparently was the foremost cause of death among reindeer calves during the study period (Karlsson et al. 2012). In our study areas, brown bear predation on rein-deer calves occurred within the first 4–6 weeks following birth (Karlsson et al. 2012). Also, pre-dation risk depended on the time of day within the predation period, with higher predation rates at night (Karlsson et al. 2012).

Our overall objective was to investigate whether and how forest- living semi- domesticated female reindeer alter their habitat selection in response to spatial and temporal variations in brown bear predation risk on their calving grounds. Moreover, we wanted to evaluate the simulta-neous responses of brown bears and reindeer to landscape characteristics, and in particular, to human- caused land- use changes (i.e., roads and forest harvesting). In our study system, active herding and fences limit reindeer movements and thus their potential to use large, landscape- scale antipredator tactics. We therefore restricted our analyses to the calving range level of selec-tion. With respect to the temporal variation in brown bear predation risk as documented in our area, we subdivided the data on a seasonal basis into the predation and post- predation period, and on a daily basis into hours of high and low brown bear predation risk. We modeled resource selection functions (RSF) separately for reindeer and brown bears for the seasonal and daily sub-division, with time period as an interaction term, and quantified spatial overlap for each species–time period combination.

The landscape of fear and the risk alloca-tion hypotheses predict that prey will respond to spatial and temporal variation in predation risk by selecting habitats with lower predation risk and that this response is more pronounced when predation risk is high (Lima and Bednekoff 1999, Laundré 2010). On the other hand, a predator should have a stronger preference for prey habitat when actively hunting prey (Sih 2005). These opposing behavioral responses likely operate simultaneously in a predator–prey system, and several factors may influence the dominant response (Sih 2005). Importantly though, the omnivorous nature of brown bears will likely cause them to select habitats rich in food resources other than reindeer calves, even during calving season (Dahle et al. 1998). Also, brown bears in Scandinavia generally respond negatively to human activity, either by avoid-ance or by seeking shelter (Nellemann et al. 2007, Ordiz et al. 2011). Together with the distinct tem-poral patterns in brown bear predation activity observed in our study area, this can create spatio-temporal variation in brown bear encounter risk on the reindeer calving range.

Accordingly, we hypothesized that female rein-deer in our study areas alter their habitat selec-tion in response to spatial and temporal variation in brown bear predation risk and expected (1) that the spatial overlap between reindeer and brown bears would be lower (due to stronger avoidance behavior in reindeer) in the predation period (compared to the post- predation period) and in high predation hours (compared to low predation hours), and (2) to observe shifts in reindeer habitat selection patterns correspond-ing to brown bear avoidance in response to the temporal variation in predation risk.

Methods

Study areaThe two study areas were centered on the

spring and summer ranges of Udtja (1210 km2, 66.2° N, 19.4° E) and Gällivare (1641 km2, 66.6° N, 21.4° E) forest reindeer herding districts (i.e., dis-tricts where the reindeer remain in forested areas all year round) located in Norrbotten County, northern Sweden (Fig. 1). We restricted the study areas to the reindeer calving range in the two herding districts. More specifically, the study


SIVERTSEN ET AL.

areas were defined as the 100% minimum convex polygon (MCP) encompassing all reindeer GPS positions within a predefined area, delineated by a combination of the reindeer herder’s defini-tions of the reindeer calving range, formal herd-ing district borders, and landscape features (i.e., rivers, roads, and railways). The vegetation is dominated by Norway spruce (Picea abies) and Scots pine (Pinus sylvestris), interspersed with bogs, lakes and, at the highest elevations, subal-pine birch (Betula pubescens) forest (Appendix S1: Fig. S2). The topography is characterized by an undulating forested landscape with elevations ranging from 187 to 714 m a.s.l. in Udtja and 38 to 528 m a.s.l. in Gällivare. In Udtja in particular, the seasonal movements of the reindeer from the winter areas to the calving ranges correspond to the elevation range following a south–north gra-dient, with higher elevations in the north.

The Udtja spring and summer ranges are mainly located within a closed military missile range, with the main human activities in the area being military training activities. Since 1995, a large part of the area has also been a nature reserve with no logging activity allowed. In Gällivare, logging activities are more intense and the road density is higher. The densities of small roads (mainly gravel roads) and major roads (public roads with regu-lar traffic) were 0.25 and 0.02 km/km2 in Udtja, and 0.38 and 0.06 km/km2 in Gällivare, respec-tively. The reindeer basically move freely within the borders of the herding districts, but are occa-sionally subject to herding activities. Movements outside the borders, or into other groups within the herding district, are restricted by natural bar-riers such as rivers, herding using snowmobiles, and in some cases, fences. In Udtja, the herding district is fenced toward the north.

Fig. 1. Map of the study areas with road networks, on the calving ranges of Udtja (lower left) and Gällivare (upper right) forest reindeer herding districts. ©Lantmäteriet i2014/764.


SIVERTSEN ET AL.

Norrbotten County (total area 97,257 km2) is sparsely populated with approximately 2.6 inhabitants per km2 in 2013, of which about two- thirds are concentrated in and around the county capital Luleå, on the Baltic Sea coast (Norrbotten County Administrative Board 2014). The total brown bear population in Norrbotten was esti-mated to be 713–1152 individuals in 2011 (Tyrén 2011). Brown bears are hunted during the annual hunting season in the autumn (21 August–15 October or until quotas are reached). In Udtja and Gällivare, the estimated brown bear popu-lation in 2010 was 62–96 and 53–75, respectively (Karlsson et al. 2012). There are no wolves in the study area, and population densities of lynx and wolverines are low (Tyrén 2011). The reindeer densities in Udtja and Gällivare are approxi-mately 110/100 km2. During the study period (2010–2012), we documented large losses of rein-deer calves to brown bear predation within the two herding districts during the reindeer calving season (Karlsson et al. 2012).

Study periodLocation data from brown bears and reindeer

were collected between 2010 and 2012 in Udtja, and in 2011 and 2012 in Gällivare. Within each year, we restricted the study period to 10 May (beginning of reindeer calving season) until 30 June (beginning of reindeer calf marking). All except three (332 of 335 calves) of the calves pre-dated by GPS- collared brown bears during the study were killed between 10 May and 9 June (Karlsson et al. 2012). Based on this information, we subdivided the study period into the predation period (10 May–9 June) and the post- predation period (10–30 June). Further, brown bears in the study area mainly killed calves between 18:00 hours and 06:00 hours (Karlsson et al. 2012); accordingly, we classified the data into high preda-tion hours (18:00 to 06:00) and low predation hours (06:00 to 18:00) within the predation period.

GPS location dataIn total, we used GPS data from 110 adult fem-

ale reindeer years and 29 brown bear years, rep-resenting 97 individual reindeer females (Udtja: 67; Gällivare: 30) and 19 individual brown bears (Udtja: 11; Gällivare: 8). GPS- collared reindeer females were mainly so- called leading females, considered to be most representative of the herd

movements. GPS locations from the female reindeer were obtained every 2 h (Telespor AS, Tromsø, Norway; Followit AB, Stockholm, Sweden). Brown bears were captured and equipped with GPS- GSM collars (VECTRONIC Aerospace GmbH, Berlin, Germany) within pre-defined areas on the reindeer spring and summer ranges in the two herding districts. See Arnemo et al. (2011) for details on capturing and marking. The brown bear GPS collars were programmed to record locations every 30 min.

Environmental dataEnvironmental data used in the analyses

included land cover types, elevation, terrain rug-gedness, and distance to the nearest large and small roads. All environmental parameters (Table 1) were extracted using ArcGIS 10.0–10.3 software (ESRI Inc., Redlands, California, USA ©2010–2015). Land cover data were mainly obtained from Geographical Data Sweden vege-tation data (Vegetationskartan, Geographical Data Sweden, National Land Sur vey of Sweden). These data are in the form of a vegetation map in vector format that includes 70 main land cover types with 300 variants, based on aerial photo-graphs (1:60,000) and field classification carried out between 1978 and 1991, with a minimum mapping unit of 3 ha across our study area. We grouped the original categories into five classes (excluding water and exploited areas [industry and settlements], which we considered as unavail-able habitat for both brown bears and reindeer): coniferous moss forest, coniferous lichen forest, deciduous forest, wetland, and other open habi-tats (cultivated land, grassland, bare rocks). The data from the vegetation map were combined with updated information on clear- cuts (derived from 1:10,000 satellite images, taken from 2000 to 2012; Utförd avv erkning, Swedish Forest Agency 2015) and data on older forestry activities, settle-ments, and cultivated land from the Swedish land cover map with a 25 × 25 m grid, with a minimum mapping unit of 1 ha (SMD Corine Land Cover Data 2000). By integrating recent data on clear- cuts with satellite forestry data from the year 2000, we defined three categories representing forestry activity: recent clear- cuts (0–5 yr), old clear- cuts (6–12 yr or <2 m in the year 2000), and young forest (2–5 m in the year 2000). In Udtja, old and recent clear- cuts were merged into a


SIVERTSEN ET AL.

common clear-cut category (Cl), due to a low pro-portion of recent clear- cuts (0.001). The final land cover map was rasterized into a 50- m grid. Road data were obtained from a road map in vector for-mat (Vägkartan, Geographical Data Sweden, National Land Survey of Sweden). We classified roads into smaller roads (mainly gravel roads) and major roads (public roads with regular traf-fic) and calculated the Euclidean distance to the nearest road for each 50 × 50 m raster cell in the study area. As the responses of large mammals to local landscape features likely deteriorate at larger distances from the feature, we transformed the “distance to road” variable (measured in meters) to exponential decays of the form 1 − eαd, where d is the distance to the feature and α was set to 0.002 (approximate effect zone <1500 m), following the approach of Nielsen et al. (2009). Exponential road distance decays ranged from 0 at the feature to 1 at very large distances. We excluded large roads from the analyses in Udtja, because (1) distances to large roads were highly correlated with elevation when exceeding an effect zone of approximately 2000 m, and (2) due to a very low density and skewed distribution of large roads in Udtja, the proportion of the study area covered by a road effect zone of ≤2000 m was too small to be included in the models (Fig. 1). Elevation was obtained from a digital elevation model with 50 m resolution and a vertical accu-racy of ±2 m (Geographical Data Sweden, National Land Survey of Sweden). Slope steepness (°) was calculated using ArcGIS. Terrain ruggedness was

calculated using the Vector Ruggedness Measure tool (VRM) (Sappington et al. 2007) in SAGA GIS (http://www.saga-gis.org) with the neighborhood parameter set to five cells. Elevation and terrain ruggedness were standardized to facilitate com-parability of regression coefficients. Slope steep-ness was removed from the analyses due to a variance inflation factor (VIF) >2; for the remain-ing set of covariates, multicollinearity was not a problem as all had a variance inflation factor (VIF) ≤2 (Zuur et al. 2010).

Statistical analysisWe used a total of 49,528 reindeer positions

(n = 97, μ = 510, SD = 219) and 24,092 brown bear positions (n = 19, μ = 1268, SD = 1337) to model brown bear and reindeer resource selection func-tions (RSF) within the reindeer calving ranges using binary logistic regression (Lele and Merrill 2013). To account for structured errors due to repeated measurements, we used generalized lin-ear mixed models (GLMMs) with the individual animal as a random effect on the intercept of the models (Zuur et al. 2009). GPS relocations repre-sented “resource use,” and a random sample of points within the calving range represented “resource availability.” For each model, the num-ber of random points equaled the number of GPS positions for each individual. We linked all locations to the environmental variables in ArcGIS. Resource use versus resource availability was the binomial response variable, whereas all environmental variables and time period (i.e.,

Table 1. Overview of environmental variables used in the analyses of reindeer and brown bear resource selec-tion, with range (min–max) and mean values for continuous variables (DEM, ruggedness, distance to road) and percentage coverage of each land cover class in the two study areas.

Environmental variables (abbreviation, unit of measurement) Udtja herding district Gällivare herding district

Digital Elevation Model (DEM, m a.s.l.) (187–714, μ = 492) (38–528, μ = 327)Ruggedness (VRM) (0–0.01, μ = 0.01) (0–0.01, μ = 0.005)Small road distance (SmRoad, m) (0–12,384, μ = 2387) (0–5402, μ = 936)Large road distance (LaRoad, m) (0–12,602, μ = 4646)Land cover classes (%)

Coniferous moss forest (Mo) 34.6 36.9Wetland (We) 27.2 36.0Coniferous lichen forest (Li) 21.0 10.2Recent clear- cut (Rc) 0.1 0.8Old clear- cut (Oc) 4.7 5.3Young forest (Yo) 2.6 8.5Deciduous forest (De) 3.0 0.5Other open habitats (Op) 3.4 1.0

http://www.saga-gis.org


SIVERTSEN ET AL.

predation/post- predation or high/low predation hours) were considered as fixed effects. We mod-eled resource selection separately for reindeer and brown bears in each study area. As we were inter-ested in how habitat selection changed between time periods, we included interactions between time period and environmental variables in all models. Moreover, we created separate models for the seasonal subdivision of time periods (pre-dation/post- predation period) and the daily sub-division of time per i ods (high/low predation hours), as opposed to a nested design (including all subdivisions in the same model). This was to simplify model interpretation and because data from the post- predation period were too scarce to be subdivided further into high/low predation hours. Following the information theory approach, we fitted four candidate models a priori (Burnham and And erson 2002). The model sets encompassed a “full model” including all environmental vari-ables, a “road and topography model” including road distance and topographic variables, a “land cover and topography” model including land cover and topographic variables, and a null model (Appendix S1: Table S1). We selected the final set of models using a combination of two criteria: (1) Determine the most parsimonious model within each model set, using second- order Akaike’s information criteria (AICc), considering the model with the lowest number of parameters with ΔAICc < 2 as the most parsimonious model to fit the data (Arnold 2010), and (2) keep the set of covariates in the models within each study area constant, to enable comparison of parameters across models within study areas. To validate the models, we used k- fold cross- validation, follow-ing the approach of Boyce et al. (2002).

We used resource selection maps obtained from the RSF models as a basis to quantify and compare spatial overlap between reindeer and brown bears, as a function of temporal variation in risk. In these maps, each pixel value represents the relative probability of selection by reindeer or brown bears during each time period. First, we determined the level of spatial autocorrelation within the RSF maps using Gaussian- fitted semi-variograms and considered the average semivar-iogram range of the RSF maps as the distance in which locations become spatially independent (see Hiemstra et al. 2009 for a detailed descrip-tion of theory and methodology). To do this, we

used the “automap” package in R (Hiemstra et al. 2009). Based on the semivariograms, we set the average range distance (Udtja: 925 m; Gällivare: 3118 m) as a sampling criterion to obtain spa-tially independent random locations within the two study areas. We obtained RSF values for each species–time period combination (extracted from the respective resource selection maps), using 171 and 1419 spatially independent points in Gällivare and Udtja, respectively. Finally, we used Pearson product- moment correlation to quantify the correlation between the RSF values obtained for brown bears and reindeer for each study area and time period, as a measure of spatial overlap.

Further, we wanted to investigate more closely the responses of reindeer and brown bears to the specific habitat characteristics, and how these changed with temporal variation in predation risk on a seasonal and daily level. To illustrate responses to environmental variables by brown bears and reindeer across different time periods, we calculated predicted probabilities of selection and the 95% confidence interval (CI). We calcu-lated predicted probabilities for a given predictor variable while keeping the other predictor vari-ables constant (at their mean values). For contin-uous predictor variables, we averaged predicted probabilities across land cover categories. The coefficients in a logistic regression are log odds ratios, estimated from the model:

where w(x) is the log odds ratio estimated from the logistic regression model, p is, in the context of resource selection functions, the probability of selection ranging from 0 to 1, and Zv represents the random intercept in mixed effect models. The predicted probability is thus calculated using the following equation:

We inferred that there was no active selection or avoidance and use proportional to the habitat area, with the predicted probability not differing from 0.5 (i.e., 95% confidence limits overlap 0.5), positive selection with predicted probability >0.5, and avoidance with predicted probability <0.5. The coefficients estimated from logistic regression were

w(x)= ln[ p1−p

]

=β0+β1x1+…+βnxn+Zv+ε

p=exp(β0+β1x1+…+βnxn+Zv+ε)

1+exp(β0+β1x1+…+βnxn+Zv+ε)


SIVERTSEN ET AL.

considered as statistically significant when 95% CI was not overlapping with zero. All statistical anal-yses were carried out in R 3.0.2 (R Development Core Team 2013). GLMMs were fit using the glmer function in the “lme4” package (Bates et al. 2014).

results

The full model, including all land cover types, road distance, and topographic variables, was the best- supported model (most parsimonious) across nearly all model sets in both study areas. The exception was the model set encompassing

reindeer and high/low predation hours in Udtja, where the full model had the lowest AICc, but with ΔAICc < 2 between the full model and the “land cover–topography” model. As we wanted a constant set of coefficients across models in each study area, we selected the full model in all model sets (Appendix S1: Table S2). Model vali-dation indicated high predictive performance for all models (Tables 2 and 3).

Spatial overlap between reindeer and brown bearsTo compare the spatial overlap in reindeer and

brown bear habitat selection between the

Table 2. Selection coefficients (log odds, β) and 95% confidence interval (CI) of resource selection functions from mixed model logistic regression for reindeer and brown bears, comparing predation/post-predation period in each reindeer herding district.

Variable

Udtja GällivareReindeer Bear Reindeer Bear

β 95% CI β 95% CI β 95% CI β 95% CI

Intercept 0.47 0.40, 0.54 −0.34 −0.46, −0.22 −0.10 −0.29, 0.10 −1.80 −2.11, −1.50Land cover†

Op −0.03 −0.13, 0.07 0.36 0.21, 0.50 0.48 0.20, 0.75 −0.10 −0.60, 0.40Mo −0.63 −0.68, −0.58 0.61 0.53, 0.69 −0.13 −0.24, −0.03 0.71 0.58, 0.85De −0.45 −0.56, −0.33 0.25 0.08, 0.42 −0.51 −0.95, −0.08 1.45 1.07, 1.84Cl 0.08 −0.01, 0.16 −0.11 −0.27, 0.05Rc 1.02 0.75, 1.29 −1.13 −1.83, −0.43Oc 0.89 0.74, 1.03 −0.10 −0.33, 0.12We −0.57 −0.62, −0.51 −0.44 −0.53, −0.35 −0.53 −0.64, −0.42 −0.78 −0.92, −0.63Yo −0.75 −0.88, −0.62 0.24 0.07, 0.42 −0.66 −0.81, −0.51 0.83 0.66, 1.00

DEM 0.08 0.06, 0.10 0.14 0.10, 0.18 0.39 0.36, 0.42 0.04 0.00, 0.08SmRoad −0.10 −0.18, −0.02 0.14 0.01, 0.27 0.70 0.59, 0.82 0.60 0.45, 0.75LaRoad −0.22 −0.39, −0.05 1.19 0.91, 1.48VRM −0.19 −0.22, −0.17 −0.05 −0.08, −0.02 −0.39 −0.43, −0.35 −0.23 −0.28, −0.18Period‡ −1.07 −1.22, −0.92 −0.91 −1.15, −0.67 −0.66 −0.96, −0.35 0.23 −0.31, 0.77Op × Period 0.64 0.44, 0.83 −0.45 −0.80, −0.11 1.19 0.74, 1.65 0.31 −0.57, 1.19Mo × Period 0.66 0.55, 0.77 0.26 0.11, 0.41 0.79 0.59, 0.99 0.02 −0.24, 0.28De × Period 0.60 0.37, 0.84 −0.27 −0.64, 0.11 0.92 0.20, 1.63 0.50 −0.14, 1.14Cl × Period 0.03 −0.17, 0.23 0.48 0.21, 0.75Rc × Period 1.34 0.89, 1.78 −1.36 −3.49, 0.77Oc × Period 0.46 0.20, 0.73 1.18 0.82, 1.55We × Period 1.17 1.07, 1.28 0.42 0.25, 0.60 1.77 1.57, 1.97 −0.10 −0.39, 0.19Yo × Period 0.35 0.06, 0.65 0.62 0.30, 0.93 1.40 1.15, 1.66 1.22 0.92, 1.52DEM × Period 0.11 0.06, 0.16 −0.71 −0.78, −0.64 −0.03 −0.09, 0.02 −0.16 −0.23, −0.08SmRoad × Period 0.41 0.25, 0.58 0.69 0.44, 0.95 0.47 0.28, 0.66 −0.02 −0.29, 0.24LaRoad × Period −1.01 −1.28, −0.74 −0.59 −1.08, −0.10VRM × Period −0.33 −0.41, −0.26 0.32 0.26, 0.38 −0.20 −0.28, −0.13 0.15 0.08, 0.22Spearman r̄s 0.99 0.92 0.96 1.00

Notes: The interaction term “Period” represents the subdivision into time periods according to temporal variation in brown bear predation risk for the whole study period: the predation period versus the post- predation period. Characters in bold in-dicate significance (95% CI not overlapping with zero). DEM, Digital Elevation Model; VRM, Vector Ruggedness Measure tool; SmRoad, small road distance; LaRoad, large road distance; Mo, coniferous moss forest; We, wetland; Cl, clear-cut; Rc, recent clear- cut; Oc, old clear- cut; Yo, young forest; De, deciduous forest; Op, other open habitats.

† Reference category: Coniferous lichen.‡ Binomial response (predation/post- predation) with reference category: Predation period.


SIVERTSEN ET AL.

predation/post- predation period and high/low predation hours, we quantified the correlation in habitat selection within each time period. The results contradicted our prediction that the spatial overlap between reindeer and brown bears would be lower (due to stronger avoidance behavior in reindeer) in the predation period (compared to the post- predation period) and in high predation hours (compared to low predation hours). The correlation between reindeer and brown bear hab-itat selection (e.g., spatial overlap) in the two study areas was significantly higher in the pre-dation period compared to the post- predation

period. Similarly, in Udtja, the correlation was sig-nificantly higher in high predation hours compared to low predation hours. The same trend was apparent, but not significant, in Gälli vare between the high and low predation hours (Fig. 2).

Habitat selection—predation and post- predation period

Both brown bears and reindeer clearly changed their responses to land cover between the preda-tion and post- predation period in both study areas. These changes in land cover selection con-firmed the higher spatial overlap observed

Table 3. Selection coefficients (log odds, β) and 95% confidence interval (CI) of resource selection functions from mixed model logistic regression for reindeer and brown bears, comparing high/low predation hours in each reindeer herding district.

Variable

Udtja GällivareReindeer Bear Reindeer Bear

β 95% CI β 95% CI β 95% CI β 95% CI

Intercept 0.46 0.36, 0.55 0.21 0.04, 0.37 0.05 −0.22, 0.31 −1.70 −2.11, −1.29Land cover†

Op −0.04 −0.18, 0.09 0.46 0.25, 0.66 0.27 −0.11, 0.65 0.06 −0.60, 0.71Mo −0.66 −0.73, −0.59 0.49 0.38, 0.59 −0.29 −0.44, −0.15 0.40 0.22, 0.57De −0.44 −0.60, −0.28 −0.03 −0.28, 0.21 −0.78 −1.36, −0.20 0.75 0.25, 1.26Cl 0.07 −0.05, 0.19 −0.03 −0.23, 0.18Rc 0.61 0.24, 0.98 −0.98 −1.85, −0.11Oc 0.76 0.56, 0.96 −0.03 −0.32, 0.27We −0.66 −0.73, −0.59 −0.39 −0.51, −0.27 −0.79 −0.93, −0.64 −0.74 −0.93, −0.55Yo −0.65 −0.83, −0.47 −0.01 −0.26, 0.23 −0.78 −0.98, −0.58 0.31 0.09, 0.54

DEM 0.10 0.07, 0.13 0.22 0.17, 0.27 0.46 0.41, 0.50 0.05 0.00, 0.11SmRoad −0.07 −0.18, 0.04 −0.43 −0.61, −0.25 0.73 0.57, 0.90 0.40 0.20, 0.60LaRoad −0.23 −0.47, 0.00 1.48 1.09, 1.86VRM −0.19 −0.23, −0.16 −0.06 −0.10, −0.02 −0.36 −0.42, −0.31 −0.24 −0.30, −0.17Hour‡ 0.00 −0.14, 0.14 −1.27 −1.52, −1.02 −0.41 −0.79, −0.03 −0.27 −0.89, 0.35Op × Hour 0.01 −0.19, 0.21 −0.20 −0.50, 0.10 0.47 −0.08, 1.02 −0.40 −1.46, 0.66Mo × Hour 0.05 −0.05, 0.16 0.27 0.11, 0.42 0.35 0.14, 0.56 0.77 0.49, 1.04De × Hour −0.01 −0.24, 0.23 0.55 0.21, 0.89 0.61 −0.28, 1.49 1.59 0.78, 2.39Cl × Hour 0.01 −0.16, 0.19 −0.23 −0.55, 0.10Rc × Hour 0.96 0.41, 1.51 −0.37 −1.86, 1.13Oc × Hour 0.32 0.03, 0.61 −0.19 −0.66, 0.29We × Hour 0.19 0.08, 0.29 −0.11 −0.29, 0.07 0.55 0.33, 0.77 −0.07 −0.37, 0.24Yo × Hour −0.22 −0.48, 0.04 0.50 0.14, 0.86 0.24 −0.06, 0.54 1.14 0.80, 1.48DEM × Hour −0.04 −0.09, 0.01 −0.13 −0.21, −0.06 −0.05 −0.11, 0.02 −0.04 −0.12, 0.04SmRoad × Hour −0.07 −0.23, 0.09 1.36 1.09, 1.63 −0.07 −0.30, 0.17 0.42 0.12, 0.72LaRoad × Hour 0.08 −0.27, 0.42 −0.64 −1.22, −0.06VRM × Hour −0.02 −0.07, 0.03 0.04 −0.02, 0.09 −0.11 −0.19 −0.02 0.02 −0.07, 0.11Spearman r̄s 0.92 0.89 0.88 0.98

Notes: The interaction term “Hour” represents the subdivision into time periods according to temporal variation in brown bear predation risk on a daily basis within the predation period: high predation hours versus low predation hours. Characters in bold indicate significance (95% CI not overlapping with zero). DEM, Digital Elevation Model; VRM, Vector Ruggedness Measure tool; SmRoad, small road distance; LaRoad, large road distance; Mo, coniferous moss forest; We, wetland; Cl, clear-cut; Rc, recent clear- cut; Oc, old clear- cut; Yo, young forest; De, deciduous forest; Op, other open habitats.

† Reference category: Coniferous lichen.‡ Binomial response (high/low predation hours) with reference category: high predation hours.


SIVERTSEN ET AL.

between brown bears and reindeer during the predation period and suggested that brown bears increased their preference for land cover types used by reindeer at this time. Reindeer showed little adjustments in relation to brown bears (Table 2, Fig. 3a–d). Reindeer mainly selected open areas and recent clear- cuts and avoided young forest throughout the study period. Further, reindeer switched from selecting coniferous lichen forest and old clear- cuts in the predation period to selection of wetlands in the post- predation period. Brown bears mainly selected moss forest and young forest and avoided recent clear- cuts throughout the study period. Indications of higher preference for reindeer habitat included noticeably greater selection of open areas and deciduous forest in the predation period by brown bears in Udtja, and greater selection of recent clear- cuts in Gällivare. Also, in both areas, brown bear selec-tion of lichen forest was greater and selection of young forest reduced in the predation period (Table 2, Fig. 3a–d). The increased spatial overlap during the predation period was particularly

evident from the seasonal trends in brown bear and reindeer responses to elevation and rugged-ness (Table 2, Fig. 4). Here, reindeer selection remained relatively constant throughout the study period, whereas the responses by brown bears changed markedly between the predation and post- predation period. Particularly in Udtja, brown bears showed a distinct seasonal switch from selection of less rugged terrain and higher elevations in the predation period to more rug-ged terrain and lower elevations in the post- predation period. Reindeer and brown bears in Gällivare avoided small roads throughout the study period. In Udtja, reindeer showed a weak preference and brown bears a weak avoidance to small roads in the predation period, whereas both species avoided roads in the post- predation period. Reindeer showed a weak selection for areas closer to large roads in the predation period, and this selection got stronger in the post- predation period, whereas brown bears generally avoided areas close to larger roads throughout the study period (Table 2).

Habitat selection—high and low predation hoursReindeer habitat selection was nearly constant

between high and low predation hours. In con-trast, brown bears changed patterns in land cover selection at the daily basis (Table 3, Fig. 3e–h). These temporal changes supported the increased spatial overlap between reindeer and brown bears in high predation hours, with brown bear selection more closely resembling reindeer in high compared to low predation hours. In Udtja, this was further supported by the temporal change in brown bears’ response to small roads, as they selected areas closer to small roads dur ing high predation hours and chose to be far ther from small roads in low predation hours (Table 3).

dIscussIon

Our study demonstrates marked differences in habitat selection between forest- living female reindeer and brown bears on the calving ranges. However, we did not find support for the hypothesis that reindeer alter their behavior in response to spatiotemporal variations in the risk for brown bear predation. Rather, the results indicate that spatiotemporal behavioral adjust-ments by brown bears were the main driver of

Fig. 2. Correlations in resource selection tested with Pearson product- moment correlation between reindeer and brown bears, comparing the predation (Pred) and post- predation period (Post), and high and low predation hours (High, Low), in Udtja reindeer herding district (a, b) and Gällivare reindeer herding district (c, d). The figure shows correlation coefficients (Pearson’s r) and 95% confidence intervals.


SIVERTSEN ET AL.

prey–predator interactions in our study system. Notably, an important assumption for the subdi-vision into time periods, and generally, the hab-itat selection models for the respective species, is that our data are representative of the study populations.

Consequently, the spatial overlap between reindeer and brown bears was higher during the predation period and in high predation hours, compared to the post- predation period and low predation hours, respectively. These patterns were mainly caused by changes in brown bear habitat selection. During the pre-dation period, brown bear habitat selection more closely resembled reindeer habitat selec-tion, with greater selection of higher elevation and less rugged terrain, compared to the post- predation period. Also, brown bears reduced selection of young forest and had greater selec-tion of recent clear- cuts, lichen- rich forest, and open areas in the predation period compared to the post- predation period. A similar pattern was apparent when comparing high and low

predation hours. Overall, this strongly suggests that brown bears actively hunted reindeer calves in our area. Interestingly, in comparison with North American study systems, this finding con-tradicts previous reports of black bears as being mainly opportunistic predators of woodland caribou calves (Bastille- Rousseau et al. 2011). We suggest that an important reason for these possible diverging foraging strategies between Scandinavian brown bears and North American black bears might be the very large difference in neonate densities between these two study sys-tems (reindeer in our study areas: 110 animals per 100 km2; woodland caribou population: 3.3 animals per 100 km2). With a low density of neo-nates, active searching for patches rich in veg-etation and opportunistic predation is probably most beneficial (Bastille- Rousseau et al. 2011), whereas it is reasonable to believe that the higher density of reindeer calves in our study area pro-motes active hunting behavior by brown bears.

The shifts in land cover selection by reindeer between the predation and post- predation period

Fig. 3. Predicted probabilities of habitat selection in relation to land cover with 95% confidence intervals during the predation (solid circles) and post- predation period (open circles), and during high (solid squares) and low (open squares) predation hours within the predation period, for (a) and (e) brown bears in Udtja, (b) and (f) reindeer in Udtja, (c) and (g) brown bears in Gällivare, and (d) and (h) reindeer in Gällivare. The horizontal dashed gray line indicates predicted probability of 0.5; values above this line indicate selection for a particular habitat characteristic, and values beneath indicate selection against the habitat characteristic. Mo, coniferous moss forest; Li, coniferous lichen forest; We, wetland; Cl, clear-cut; Rc, recent clear-cut; Oc, old clear-cut; Yo, young forest; De, deciduous forest; Op, other open habitats.


SIVERTSEN ET AL.

appeared relatively unrelated to the temporal shift in brown bear predation risk and seem best explained by forage availability. That reindeer preferred coniferous lichen forest in the predation period and then shifted to a stronger selection of wetlands in the post- predation period corre-sponds closely to the changes in vegetation that occur over the same time. In early spring, snow first disappears on the dry pine heaths in the conif-erous lichen forests, making lichen on the ground more available for reindeer, while green protein- rich vegetation gradually appears on the wet-lands later in the season (Helle 1981). Although there were few indications that reindeer altered their habitat selection to reduce the probability of brown bear encounters, the observed selection patterns may still partly reflect an adaptation to predation risk. The fact that brown bears had a preference for young regenerating forest, com-bined with a reduced predator detection ability of reindeer in this habitat type, could be important in explaining the avoidance of young forest by reindeer throughout the study period. Selection of recent clear- cuts could be beneficial in terms

of both forage and predation risk (Dussault et al. 2012, Leblond et al. 2016). Recent clear- cuts can be rich in grasses and forbs (Dussault et al. 2012), which are important forage for reindeer during spring (Helle 1981, Skogland 1984). In addition, they provide good visibility, concealment cover near the ground, and lower risk of bear encoun-ters (Dussault et al. 2012). Yet, on the scale of the calving range, brown bear habitat selection seems to mainly override any attempts by reindeer to reduce brown bear encounter risk in our study areas. This suggests that the reindeer might be maladapted to cope with brown bear predation risk on their calving grounds. Maladaptive habitat selection can arise when anthropogenic activities, such as logging activities, introduce a mismatch between cues used by the animal for selecting a habitat and the actual habitat quality (Hollander et al. 2011). Dussault et al. (2012) found that older clear- cuts are more attractive to black bears and thereby represent higher predation risk for cari-bou calves, and thus, selection for clear- cut areas can in a longer perspective have negative conse-quences for fitness. In our study area, reindeer

Fig. 4. Predicted probabilities of habitat selection in relation to elevation and ruggedness variables during the predation (solid line) and post- predation period (dashed line) with 95% confidence intervals for (a) and (e) brown bears in Udtja, (b) and (f) reindeer in Udtja, (c) and (g) brown bears in Gällivare, and (d) and (h) reindeer in Gällivare. The horizontal dashed gray line indicates predicted probability of 0.5; values above this line indicate selection for a particular habitat characteristic, and values beneath indicate selection against the habitat characteristic.


SIVERTSEN ET AL.

selected, while brown bears avoided or showed no selection, for recent and old clear- cuts during the predation period. Thus, if female reindeer exhibit high fidelity to calving areas, as has been observed in female woodland caribou (Faille et al. 2010), such selection can become detrimental when clear- cuts eventually turn into regenerat-ing forest (Dussault et al. 2012). Also, prey may display dysfunctional antipredator strategies fol-lowing recent predator re- introduction or rapid growth of predator populations (Berger 2007b, Leblond et al. 2016). The near- absence of large carnivores in Scandinavia over the last century (until a few decades ago) could have caused such effects in reindeer, as has been shown for moose in Scandinavia lacking antipredator responses to newly restored wolf populations (Berger et al. 2001). Furthermore, domestication may have influ-enced behavioral interactions with predators and vulnerability to predation of semi- domesticated reindeer in various ways. Reindeer in a forest herding community originate from the tundra reindeer subspecies, which is the most common domesticated subspecies of reindeer (Bjørklund 2013). Records exist of animals migrating to calve in the mountains if they managed to escape the herding district borders. Thus, reindeer females may be restricted by the herding system in their movements and calve in areas for which they are not fully adapted. Further, general predator avoidance theory states that, for animals living in forested habitats, the best way to reduce the risk of predation is to remain scattered at low densi-ties, forcing the predator to search large areas for prey (Bergerud and Page 1987). This behavior is commonly observed in forest- dwelling woodland caribou in North America and taiga reindeer in Eurasia during calving (Baskin 1986, Bergerud et al. 1990). Similarly, the reindeer females in our study areas have been observed to disperse during calving. However, areal restrictions and a strong herd behavior as a result of the domesti-cation process (Skarin and Åhman 2014) limit the dispersion and, in combination with larger pop-ulation densities than seen in most wild popula-tions, probably make the calves more susceptible to predation.

The lack of variation in habitat selection between day and night among female reindeer in our study area further suggests that they may be maladapted to the risk of brown bear

predation. In many wild populations exposed to predation risk, habitat selection patterns dif-fer greatly between day and night (Tolon et al. 2009, Tambling et al. 2015). In contrast, the daily patterns of habitat selection among reindeer in our study more closely resemble the behavior of ungulates living in a predator- free environ-ment, although our study populations face a nocturnally active predator (Tambling et al. 2015, Bonnot et al. 2016).

The contrasting responses by brown bears and reindeer to the different forest growth stages indicate that forestry activity can influence rein-deer–brown bear behavioral interactions, and highlight the importance of considering both prey and predators when assessing ecological effects of human activity. Although selection for clear- cuts may reduce the predation risk on calves, and thus have positive consequences for female fitness in the short term (Dussault et al. 2012, Leblond et al. 2016), logging activity will, eventually, increase the abundance of young regenerating forest and thus reduce suitable habitats for the reindeer. Further, the effects of clear- cuts can depend on the level of intensity of human disturbance. A higher proportion of clear- cuts within the home range may have detrimen-tal effects on calf survival if the local density of clear- cuts is high (Leclerc et al. 2014). Also, older clear- cuts may have negative consequences for female fitness (Losier et al. 2015, Leblond et al. 2016). Moreover, our results suggest that forestry activity on the calving ranges may have negative consequences for the reindeer herd in the long term. We therefore suggest that forestry activities should be minimized on the main reindeer calv-ing ranges in forest reindeer herding districts. Also, Bastille- Rousseau et al. (2011), studying black bear predation on caribou calves, showed that the risk of bear predation became higher in areas fragmented by young forest when young forest was selected by bears. This is because black bears typically move frequently between vegetation- rich patches to optimize foraging efficiency, which can result in relatively high encounter rates with ungulate neonates, even when bears are not actively hunting (Bastille- Rousseau et al. 2011). Like black bears, brown bears have an omnivorous nature, inhabit a patchy environment, and have a large body size that necessitates high intake rates and short


SIVERTSEN ET AL.

searching times for forage (Wiens 1976, Welch et al. 1997, Hertel et al. 2016). A similar foraging behavior as in black bears can thus be expected in brown bears. Hence, comparable effects from forestry on brown bear–reindeer interactions might arise in Scandinavia. This is an important concern that should be investigated further.

The hierarchical habitat selection theory pos-tulates that in populations where predation is the primary limiting factor, predator avoidance will occur at the largest scale possible (Rettie and Messier 2000). In wild Rangifer populations with calving ranges overlapping with predators, females select habitats with a lower predator encounter risk within the calving range (Rettie and Messier 2000). We could not document such behavior in our study populations. It is possible, however, that reindeer antipredator behavior still occurred on other spatial scales. One may expect that if predator avoidance fails at the larger spatial scale, reindeer should con-tinue to select habitats that minimize predation risk at finer spatial scales (Rettie and Messier 2000). For example, in woodland caribou, pred-ator avoidance behavior has been documented within the home range (Hins et al. 2009) and at fine- scale calving site selection (Leclerc et al. 2012). Nevertheless, our results imply that in our study area, the reindeer are losing the prey–predator behavioral game (sensu Sih 2005 and Laundré 2010: the behavioral response race and landscape of fear) and that brown bear behavior dominates the predator–prey interac-tions at the scale of the reindeer calving range. Overall, the lack of ability of female reindeer to reduce the probability of brown bear encounters within their calving range is probably an import-ant reason for the high brown bear predation rates on reindeer calves in these two reindeer herding districts.

AcknowledgMents

This study was part of the project “Reindeer calving in pens and zoning of brown bears—Effects on preda-tion,” with project number L2012/2817, funded by the Swe dish Government, which is a collaboration between Udtja and Gällivare reindeer herding dis-tricts, the Swedish Wildlife Damage Center, Swedish University of Agricultural Sciences (SLU), the Scandinavian Brown Bear Research Project (SBBRP),

and the Dep artment of Animal Nutrition and Management, SLU. Work by Sivertsen and Skarin was supported by the Faculty of Veterinary Medicine and Animal Science, SLU. The SBBRP is funded by the Norwegian Environ ment Agency, the Swedish Environmental Protection Agency, the Research Council of Norway, and the Swedish Association for Hunting and Wildlife Man agement. We acknowledge the support of the Center for Advanced Study in Oslo, Norway, that funded and hosted the research project “Climate effects on harvested large mammal popula-tions” during the academic year 2015–2016. Finally, we would like to thank Martin- Hugues St- Laurent and one anonymous revi ewer for valuable and construc-tive comments that helped to improve the manuscript. This is scientific paper no. 223 from the Scandinavian Brown Bear Research Project.

lIterAture cIted

Adams, L. G., F. J. Singer, and B. W. Dale. 1995. Cari-bou calf mortality in Denali National Park, Alaska. Journal of Wildlife Management 59:584–594.

Åhman, B., K. Svensson, and L. Rönnegård. 2014. High female mortality resulting in herd collapse in free- ranging domesticated reindeer (Rangifer tarandus tarandus) in Sweden. PLoS ONE 9:e111509.

Arnemo, J. M., A. Evans, and Å. Fahlman. 2011. Bio-medical protocols for free-ranging brown bears, wolves, wolverines and lynx. Hedmark University College, Evenstad, Norway. http://www1.nina.no/RovviltPub/pdf/Biomedical%20Protocols%20Car nivores%202011.pdf

Arnold, T. 2010. Uninformative parameters and model selection using Akaike’s information criterion. Journal of Wildlife Management 74:1175–1178.

Barten, N. L., R. T. Bowyer, and K. J. Jenkins. 2001. Habitat use by female caribou: tradeoffs associated with parturition. Journal of Wildlife Management 65:77–92.

Baskin, L. M. 1986. Differences in the ecology and be-haviour of reindeer populations in the USSR. Ran-gifer Special Issue 1:333–340.

Bastille-Rousseau, G., D. Fortin, C. Dussault, R. Cour-tois, and J.-P. Ouellet. 2011. Foraging strategies by omnivores: Are black bears actively searching for ungulate neonates or are they simply opportunistic predators? Ecography (Cop) 34:588–596.

Bastille-Rousseau, G., J. P. Potts, J. A. Schaefer, M. A. Lewis, E. H. Ellington, N. D. Rayl, S. P. Mahoney, and D. L. Murray. 2015. Unveiling tradeoffs in resource selection of migratory caribou using a mechanistic model of availability. Ecography 38:1049–1059.

Bates, D., M. Maechler, B. Bolker, and S. Walker. 2014. lme4: linear mixed-effects models using Eigen and

http://www1.nina.no/RovviltPub/pdf/Biomedical%2520Protocols%2520Carnivores%25202011.pdf




SIVERTSEN ET AL.

S4. R package 1.1-7. http://cran.r-project.org/pack age=lme4

Berger, J. 2007a. Fear, human shields and the redistri-bution of prey and predators in protected areas. Biology Letters 3:620–623.

Berger, J. 2007b. Carnivore repatriation in holarctic prey: narrowing the deficit in ecological effective-ness. Conservation Biology 21:1105–1116.

Berger, J., J. E. Swenson, and I.-L. Persson. 2001. Recolonizing carnivores and naive prey: conserva-tion lessons from Pleistocene extinctions. Science 291:1036–1039.

Bergerud, A. T., R. Ferguson, and H. E. Butler. 1990. Spring migration and dispersion of wood-land caribou at calving. Animal Behaviour 39: 360–368.

Bergerud, A. T., S. N. Luttich, and L. Camps. 2008. McGill-Queen’s native and northern series: return of caribou to Ungava. McGill-Queen’s University Press, Montreal, Quebec, Canada.

Bergerud, A. T., and R. E. Page. 1987. Displacement and dispersion of parturient caribou at calving as antipredator tactics. Canadian Journal of Zoology 65:1597–1606.

Bjørklund, I. 2013. Domestication, reindeer husb andry and the development of Sami pastoralism. Acta Borealia 30:174–189.

Bonnot, N. C., N. Morellet, A. J. M. Hewison, J.-L. Martin, S. Benhamou, and S. Chamaillé-Jammes. 2016. Sitka black- tailed deer (Odocoileus hemio-nus sitkensis) adjust habitat selection and activity rhythm to the absence of predators. Canadian Jour-nal of Zoology 94:385–394.

Boyce, M. S., P. R. Vernier, S. E. Nielsen, and F. K. A. Schmiegelow. 2002. Evaluating resource selection functions. Ecological Modelling 157:281–300.

Brown, J. S., J. W. Laundré, and M. Gurung. 1999. The ecology of fear: optimal foraging, game theory, and trophic interactions. Journal of Mammalogy 80:385–399.

Burnham, K. P., and D. R. Anderson. 2002. Model selec-tion and multimodel inference: a practical informa-tion-theoretic approach. Second edition. Springer, New York, New York, USA.

Dahle, B., O. J. Sørensen, E. H. Wedul, J. E. Swenson, and F. Sandegren. 1998. The diet of brown bears Ursus arctos in central Scandinavia: effect of access to free- ranging domestic sheep Ovis aries. Wildlife Biology 4:147–158.

Dussault, C., V. Pinard, J.-P. Ouellet, R. Courtois, and D. Fortin. 2012. Avoidance of roads and selection for recent cutovers by threatened caribou: Fitness- rewarding or maladaptive behaviour? Proceed-ings of the Royal Society B: Biological Sciences 279:4481–4488.

Faille, G., C. Dussault, J.-P. Ouellet, D. Fortin, R. Courtois, M.-H. St-Laurent, and C. Dussault. 2010. Range fidelity: The missing link between caribou decline and habitat alteration? Biological Conservation 143:2840–2850.

Fancy, S. G., and K. R. Whitten. 1991. Selection of calv-ing sites by Porcupine herd caribou. Canadian Journal of Zoology 69:1736–1743.

Gustine, D. D., K. L. Parker, R. J. Lay, M. P. Gillingham, and D. C. Heard. 2006. Calf survival of woodland caribou in a multi- predator ecosystem. Wildlife Monographs 165:1–32.

Heard, D. C., T. M. Williams, and D. A. Melton. 1996. The relationship between food intake and preda-tion risk in migratory caribou and implications to caribou and wolf population dynamics. Rangifer Special Issue 9:37–44.

Heithaus, M. R., and L. M. Dill. 2002. Food availability and tiger shark predation risk influence bottlenose dolphin habitat use. Ecology 83:480–491.

Helle, T. 1981. Studies on wild forest reindeer (Rangifer tarandus fennicus Lonn.) and semi- domestic rein-deer (Rangifer tarandus tarandus L.) in Finland. Acta Universitatis Ouluensis Series A 107:Biol 12.

Hertel, A., S. M. J. G. Steyaert, A. Zedrosser, A. Myster-ud, H. K. Lodberg-Holm, H. W. Gelink, J. Kindberg, and J. E. Swenson. 2016. Bears and berries: species- specific selective foraging on a patchily distribut-ed food resource in a human- altered landscape. Behavioral Ecology and Sociobiology 70:831–842.

Hiemstra, P. H., E. J. Pebesma, C. J. W. Twenhofel, and G. B. M. Heuvelink. 2009. Real- time automatic int-erpolation of ambient gamma dose rates from the Dutch Radioactivity Monitoring Network. Com-puters and Geosciences 35:1711–1721.

Hins, C., J.-P. Ouellet, C. Dussault, and M.-H. St-Laurent. 2009. Habitat selection by forest- dwelling caribou in managed boreal forest of eastern Canada: evidence of a landscape configu-ration effect. Forest Ecology and Management 257: 636–643.

Hobbs, N. T., and H. Andrén. 2012. Native predators reduce harvest of reindeer by Sámi pastoralists. Ecological Applications 22:1640–1654.

Holbrook, S. J., and R. S. Schmitt. 1988. The combined effects of predation risk and food reward on patch selection. Ecology 69:125–134.

Hollander, F. A., H. Van Dyck, G. San Martin, and N. Titeux. 2011. Maladaptive habitat selection of a migratory passerine bird in a human- modified landscape. PLoS ONE 6:e25703.

Karlsson, J., et al. 2012. Björnpredation på ren och potentiella effekter av tre förebyggande åtgärder. Rapport från Viltskadecenter 6. Sveriges Lant-bruksuniversitet SLU, Grimsø forskningsstation,

http://cran.r-project.org/package=lme4

http://cran.r-project.org/package=lme4


SIVERTSEN ET AL.

Riddarhyttan, Sweden. [In Swedish.] http://pub.epsilon.slu.se/13047/1/karlsson_j_etal_160301.pdf

Latham, A. D. M., M. C. Latham, and M. S. Boyce. 2011. Habitat selection and spatial relationships of black bears (Ursus americanus) with woodland caribou (Rangifer tarandus caribou) in northeastern Alberta. Canadian Journal of Zoology 89:267–277.

Latombe, G., D. Fortin, and L. Parrott. 2014. Spatio- temporal dynamics in the response of woodland caribou and moose to the passage of grey wolf. Journal of Animal Ecology 83:185–198.

Laundré, J. W. 2010. Behavioral response races, predator- prey shell games, ecology of fear, and patch use of pumas and their ungulate prey. Ecol-ogy 91:2995–3007.

Laundré, J. W., L. Hernandez, and W. J. Ripple. 2010. The landscape of fear: ecological implications of being afraid. Open Ecology Journal 3:1–7.

Leblond, M., C. Dussault, J.-P. Ouellet, and M.-H. St-Laurent. 2016. Caribou avoiding wolves face increased predation by bear – Caught between Scylla and Charybdis. Journal of Applied Ecology 53:1078–1087.

Leclerc, M., C. Dussault, and M.-H. St-Laurent. 2012. Multiscale assessment of the impacts of roads and cutovers on calving site selection in woodland car-ibou. Forest Ecology and Management 286:59–65.

Leclerc, M., C. Dussault, and M.-H. St-Laurent. 2014. Behavioural strategies towards human disturb-ances explain individual performance in woodland caribou. Oecologia 176:297–306.

Lele, S., and E. Merrill. 2013. Selection, use, choice and occupancy: clarifying concepts in resource selection studies. Journal of Animal Ecology 82:1183–1191.

Lima, S. L., and P. A. Bednekoff. 1999. Temporal vari-ation in danger drives antipredator behavior: the predation risk allocation hypothesis. American Naturalist 153:649–659.

Linnell, J., R. Aanes, and R. Andersen. 1995. Who killed Bambi? The role of predation in the neonatal mortality of temperate ungulates. Wildlife Biology 1:209–223.

Losier, C. L., S. Couturier, M.-H. St-Laurent, P. Drapeau, C. Dussault, T. Rudolph, V. Brodeur, J. A. Merkle, and D. Fortin. 2015. Adjustments in habitat selection to changing availability induce fitness costs for a threatened ungulate. Journal of Applied Ecology 52:496–504.

Mahoney, S., and J. Virgl. 2003. Habitat selection and demography of a nonmigratory woodland caribou population in Newfoundland. Canadian Journal of Zoology 81:321–334.

McLoughlin, P., J. S. Dunford, and S. Boutin. 2005. Relating predation mortality to broad- scale habitat selection. Journal of Animal Ecology 74:701–707.

Nellemann, C., O.-G. Støen, J. Kindberg, J. E. Swenson, I. Vistnes, G. Ericsson, J. Katajisto, B. P. Kaltenborn, J. Martin, and A. Ordiz. 2007. Terrain use by an expanding brown bear population in relation to age, recreational resorts and human settlements. Biological Conservation 138:157–165.

Nielsen, S. E., J. Cranston, and G. B. Stenhouse. 2009. Identification of priority areas for grizzly bear con-servation and recovery in Alberta, Canada. Journal of Conservation Planning 5:38–60.

Nieminen, M. 2010. The impact of large carnivores on the mortality of semi- domesticated reindeer ( Rangifer tarandus tarandus L.) calves in Kainuu, southeastern reindeer- herding region. Rangifer 30:79–88.

Norrbotten County Administrative Board. 2014. Facts about Norrbotten. Norrbotten County Admin-istrative Board, Luleå, Sweden. http://www.lans styrelsen.se/norrbotten/SiteCollectionDocuments/Sv/publikationer/om%20lansstyrelsen/Engels ka%20-%20Fakta%20om%20Norrbotten.pdf

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

Pinard, V., C. Dussault, J.-P. Ouellet, D. Fortin, and R. Courtois. 2012. Calving rate, calf survival rate, and habitat selection of forest- dwelling caribou in a highly managed landscape. Journal of Wildlife Management 76:189–199.

R Development Core Team. 2013. R: a language and environment for statistical computing. R Founda-tion for Statistical Computing, Vienna, Austria. https://www.r-project.org/

Rettie, W. J., and F. Messier. 2000. Hierarchical habitat selection by woodland caribou: its relationship to limiting factors. Ecography (Cop) 23:466–478.

Rettie, W. J., and F. Messier. 2001. Range use and move-ment rates of woodland caribou in Saskatchewan. Canadian Journal of Zoology 79:1933–1940.

Rosenzweig, M. L. 1991. Habitat selection and popula-tion interactions: the search for mechanism. Amer-ican Naturalist 137:5–28.

Sappington, J., K. M. Longshore, and D. B. Thompson. 2007. Quantifying landscape ruggedness for an-imal habitat analysis: a case study using bighorn sheep in the Mojave Desert. Journal of Wildlife Management 71:1419–1426.

Sih, A. 1980. Optimal behavior: Can foragers bal-ance two conflicting demands? Science 210: 1041–1043.

Sih, A. 2005. Predator-prey space use as an emergent outcome of a behavioral response race. Pages 240–255 in P. Barbosa and I. Castellanos, editors. Ecology of predator-prey interactions. Oxford Uni-versity Press, Oxford, UK.

http://pub.epsilon.slu.se/13047/1/karlsson_j_etal_160301.pdf

http://pub.epsilon.slu.se/13047/1/karlsson_j_etal_160301.pdf

http://www.lansstyrelsen.se/norrbotten/SiteCollectionDocuments/Sv/publikationer/om%2520lansstyrelsen/Engelska%2520-%2520Fakta%2520om%2520Norrbotten.pdf




https://www.r-project.org/


SIVERTSEN ET AL.

Sih, A., J. Cote, M. Evans, S. Fogerty, and J. Pruitt. 2012. Ecological implications of behavioural syndromes. Ecology Letters 15:278–289.

Skarin, A., and B. Åhman. 2014. Do human activity and infrastructure disturb domesticated reindeer? The need for the reindeer’s perspective. Polar Biol-ogy 37:1041–1054.

Skogland, T. 1984. Wild reindeer foraging niche orga-nization. Ecography (Cop) 7:345–379.

Swenson, J. E., and H. Andrén. 2005. A tale of two coun-tries: large carnivore depredation and compensation schemes in Sweden and Norway. Pages 323–339 in R. Woodroffe, S. Thirgood, and A. Rabinowitz, ed-itors. People and wildlife: Conflict or co-existence? Cambridge University Press, Cambridge, UK.

Tambling, C. J., L. Minnie, J. Meyer, E. W. Freeman, R. M. Santymire, J. Adendorff, and G. I. H. Kerley. 2015. Temporal shifts in activity of prey following large predator reintroductions. Behavioural Ecolo-gy and Sociobiology 69:1153–1161.

Tolon, V., S. Dray, A. Loison, A. Zeileis, C. Fischer, and E. Baubet. 2009. Responding to spatial and tempo-ral variations in predation risk: space use of a game species in a changing landscape of fear. Canadian Journal of Zoology 87:1129–1137.

Tyrén, H. 2011. Rovdjursinventeringen 2010-2011 i Norrbottens län. Länsstyrelsens rapportserie nr

9/2011 (In Swedish). http://www.lansstyrelsen.se/norrbotten/SiteCollectionDocuments/Sv/publika tioner/djur%20och%20natur/Rovdjur/Rovdjursin ventering2011webb.pdf

Valeix, M., A. J. Loveridge, S. Chamaillé-Jammes, Z. Davidson, F. Murindagomo, H. Fritz, and D. W. MacDonald. 2009. Behavioral adjustments of Afri-can herbivores to predation risk by lions: Spatio-temporal variations influence habitat use. Ecology 90:23–30.

Welch, C., J. Keay, K. C. Kendall, and C. T. Robbins. 1997. Constraints on frugivory by bears. Ecology 78:1105–1119.

Wiens, J. A. 1976. Population responses to a patchy environment. Annual Review of Ecology and Sys-tematics 7:81–120.

Wiens, J., N. C. Stenseth, B.Van Horne, and R. A. Ims. 1993. Ecological mechanisms and landscape ecolo-gy. Oikos 66:369–380.

Zuur, A. F., E. N. Ieno, and C. S. Elphick. 2010. A pro-tocol for data exploration to avoid common statis-tical problems. Methods in Ecology and Evolution 1:3–14.

Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. Mixed effect models and exten-sions in ecology with R. Springer, New York, New York, USA.

supportIng InforMAtIon

Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2.1583/full

http://www.lansstyrelsen.se/norrbotten/SiteCollectionDocuments/Sv/publikationer/djur%2520och%2520natur/Rovdjur/Rovdjursinventering2011webb.pdf




http://onlinelibrary.wiley.com/doi/10.1002/ecs2.1583/full

http://onlinelibrary.wiley.com/doi/10.1002/ecs2.1583/full

Reindeer habitat selection under the risk of brown bear...

Documents

Transcript of Reindeer habitat selection under the risk of brown bear...