Mobility of moose—comparing the effects of wolf predation risk, reproductive status, and seasonality

Abstract In a predator–prey system, prey species may adapt to the presence of predators with behavioral changes such as increased vigilance, shifting habitats, or changes in their mobility. In North America, moose (Alces alces) have shown behavioral adaptations to presence of predators, but such antipredator behavioral responses have not yet been found in Scandinavian moose in response to the recolonization of wolves (Canis lupus). We studied travel speed and direction of movement of GPS‐collared female moose (n = 26) in relation to spatiotemporal differences in wolf predation risk, reproductive status, and time of year. Travel speed was highest during the calving (May–July) and postcalving (August–October) seasons and was lower for females with calves than females without calves. Similarly, time of year and reproductive status affected the direction of movement, as more concentrated movement was observed for females with calves at heel, during the calving season. We did not find support for that wolf predation risk was an important factor affecting moose travel speed or direction of movement. Likely causal factors for the weak effect of wolf predation risk on mobility of moose include high moose‐to‐wolf ratio and intensive hunter harvest of the moose population during the past century.


| 8871
WIKENROS Et al. concentrated movements can be expected for prey species in habitats with high availability of food versus longer and more directed movements as animals move through areas with low availability of resources (Fryxell et al., 2008) or when fleeing from predators (Wikenros, Sand, Wabakken, Liberg, & Pedersen, 2009). In particular, increased movement rate is likely an advantageous behavior in escaping predators once detected or to minimize the time spent near predators (Gude, Garrott, Borkowski, & King, 2006;Mitchell & Lima, 2002). The presence of predators may also result in lower travel speed as a response to increased vigilance (Berger, 1999;White & Berger, 2001) and thereby suppresses mobility (Lima & Dill, 1990). Such a reduction in movement can also be a beneficial antipredator behavior, because moving animals are generally more easily detected by a predator than are inactive animals (Lima & Dill, 1990). Because these antipredator behaviors are costly, for example, prey species need to trade reduced risk with reduced consumption, theory predicts that prey in systems with low risk or absence of predators should result in a loss of costly antipredator behavior (Blumstein & Daniel, 2005). Likewise, re-establishment of large predators may result in a resumption of a formerly lost antipredator behavior by prey (Berger, Swenson, & Persson, 2001).
Wolves have been absent from south-central Scandinavia after extermination for more than 100 years but started to recolonize this region in the 1980s (Wabakken, Sand, Liberg, & Bjärvall, 2001). The population size was estimated to be 289-325 wolves in the winter of 2010/2011 . During the period without wolves, hunter harvest replaced predation as the main mortality source for moose (Lavsund & Sandegren, 1989;Lavsund et al., 2003;Stubsjoen, Saether, Solberg, Helm, & Rolandsen, 2000) likely preventing adaptation of antipredator behavior. For example, moose have not expressed behavioral adjustments to lower wolf hunting success even in territories occupied by wolves for more than 20 years Sand et al., 2006a;Wikenros et al., 2009). The situation in Scandinavia is therefore interesting because it contrasts with the conditions normally found in protected areas (e.g., national parks) where predation from large predators has not been replaced by hunter harvest. Studies on behaviorally mediated effects on prey by large predators carried out within national parks may not be representative of predator-prey interactions outside national parks (Mech, 2012). Large predators are predicted to have less behaviorally mediated effects on other species in areas where anthropogenic changes have a large impact on several trophic levels (Eriksen et al., 2011;Kuijper et al., 2016;Mech, 2012;Nicholson, Milleret, Månsson, & Sand, 2014;Ripple et al., 2014). Studies in contrasting environments like those found in Scandinavia are therefore important as anthropogenic impact occurs in most parts of the world's wolf range.
We examined the effects of recolonizing wolves on the movement patterns of moose at different spatiotemporal scales. More specifically, we tested whether mobility in terms of travel speed and direction of movement of female moose decreased at a course spatiotemporal scale (annual or seasonal) in order to avoid detection and increased at a finer spatial scale (when wolves are nearby moose) in order to flee from an approaching predator. Ungulate prey are known to adapt to changing environmental conditions affecting resource availability, induced by seasonality as well as reproductive status (Eriksen et al., 2011). Moose calves are the main prey for wolves in Scandinavia year round (Sand et al., 2005. Therefore, we included time of the year and reproductive status as explanatory variables. We predicted that females with calves would be most likely to change movement patterns in response to wolf predation risk.

| Study area
The study was conducted in the surroundings of Grimsö Wildlife Research Area (59-60°N, 15-16°E), located in the boreal zone of south-central Sweden (Rönnegård, Sand, Andrén, Månsson, & Pehrson, 2008) with a study area of approximately 1,000 km 2 . The topography of this rugged plateau is characterized by flat ridges, boulders, and swampy areas with the elevation ranging between 100 and 150 m (digital elevation model, Geographical Data Sweden, GSD, National Land Survey of Sweden). The main land cover type in the area is forest (72%), bogs (18%), lakes and rivers (7%) as well as meadows (3%; Björkhem & Lundmark, 1975). Intensive forest management dominates within the area, with average stand rotation periods of 80-100 years. The main tree species are Scots pine (Pinus silvestris), Norway spruce (Picea abies), and birches (Betula pubescens and B. pendula; Månsson, Andrén, Pehrson, & Bergström, 2007). The climate is characterized by continental climate with average temperatures of −5°C in January and 15°C in July (Vedin, 1995). The ground is usually snow-covered between late November and early April with a mean snow depth of 20 cm in mid-January (Dahlström, 1995).
During the study period (2007)(2008)(2009)(2010), wolves were continuously present in the study area. The territorial pair (named Uttersberg) established its territory during the winter of 2003/2004 (Wabakken, Aronson, Sand, Strømseth, & Kojola, 2004). Reproduction was con-  (Wabakken et al., 2010). Wolf territory sizes reached a maximum of 1,000 km 2 during the study period . Moose home ranges jointly covered a total area of 410 km 2 that was partly outside the Uttersberg territory and totally surrounded by the Hedbyn territory.

| GPS locations
Wolves and moose were immobilized by darts from helicopters (see Sand, Wikenros, Wabakken, & Liberg, 2006b;Månsson, Andrén, & Sand, 2011 for details). Handling protocols fulfilled the ethical requirements for research on wild animals in Sweden (decision C281/6 and C315/6). Over the study period, four wolves (the territorial pairs) were collared and the packs were continuously monitored except during a 3-month period when the Uttersberg pack was replaced by the Hedbyn pack. Female moose were collared in March 2007 (n = 20) with an additional 10 females collared in 2010. Wolf GPS collars were programmed for locations with 12-hr intervals, whereas the GPS collars of the moose took locations every second hour. For this study, we used locations of both species from four consecutive years (2007)(2008)(2009)(2010) We screened moose GPS data for location errors following the nonmovement method developed by Bjørneraas, Van Moorter, Rolandsen, and Herfindal (2010). We set the distance parameters as ∆ = 100 km and μ = 10 km (three successive locations moving back and forth with high speed limit), and the speed limit was set as ∝ = 1.5 km/hr and turning angle θ = −0.97 (Bjørneraas et al., 2010)  locations that formed a spike. We excluded all locations from the 7-day postcapture to avoid the effect of immobilization on moose behavior (Neumann, Ericsson, Dettki, & Arnemo, 2011). We screened the data using the package Adehabitat (Calenge, 2006)

| Wolf predation risk
Predation risk is the probability of being killed per unit time by the predator (Lima & Dill, 1990), but the mere presence of a predator could be equally important for predicting risk (Hebblewhite & Merrill, 2007). We used three methods to calculate wolf predation risk at different spatiotemporal scales.
First, we calculated an annual predation risk index as the annual home range overlap (%) between all moose home ranges and the wolf territory. We used locations from one of the adult wolves at the time to estimate wolf territory range (n = 4) during 2007-2010 (the male in the Uttersberg territory and the female in the Hedbyn territory).
This was based on the assumption that the movement and activity of a pair is highly synchronized, with the exception of the pup rearing period (Alfredéen, 2006;Eriksen et al., 2011). We estimated annual territories for wolves and annual moose home ranges using both the 100% minimum convex polygon (MCP; Mohr, 1947) and the 95% fixed kernel (Kernel; Worton, 1987) with the reference technique (href) to calculate the smoothing factor h (Kie et al., 2010). We calculated home ranges using the R library AdehabitatHR (Calenge, 2006).

| Moose mobility
We estimated travel speed (m/hr) and direction of movement (linearity) within all individual seasonal moose home ranges and used these as indices of mobility. We used (1) (Eriksen et al., 2011). We calculated movement parameters of each study animal with R library "AdehabitatLT" (Calenge, 2006). In order to meet the assumption of normally distributed residuals, travel speed was transformed by ln(x + 1) and direction of movement by exp(arcsin(√x)) (Eriksen et al., 2011).

| Moose reproductive status
Female moose were checked for reproduction in terms of the number of newborn calves at heel in the spring (12 May-4 July), and reproductive individuals were again checked in late summer (26 August-9 September) and at the end of winter the following year (1 April-29 April). We classified females as with or without a calf in each of the four seasons.

| Analyses
We conducted all analyses in R version 3.2.2 (R Development Core We also included the interactions between season and reproductive status, season and wolf predation risk, and reproductive status and wolf predation risk. We repeated the analyses using a subset of the data where distances between moose and wolves were ˂1 km (representing wolves being nearby moose; n = 104), 10.5-11.49 km (representing the average distance between wolves and moose; n = 1,698), and 20.5-21.49 km (representing wolves being far away from moose: n = 272), and the explanatory variable wolf predation risk was used as a three-level categorical variable.
Moose ID was used as random effect to account for multiple observations of the same individual in all models. We used Akaike information criterion corrected for small sample sizes (AIC c ) to rank models. Models with ∆AIC c = 0-2 were considered to have equally strong support and models with ∆AIC c = 4-7 to have some support (Burnham & Anderson, 2002).

| RESULTS
The

| Travel speed
Travel speed of moose averaged 54.8 m/hr ± 1.5 SE during all seasons. The best model explaining variation of travel speed included season and reproductive status as explanatory variables (Table 1).
Travel speed was highest during the calving and postcalving seasons and lowest during the low activity season and precalving season (Table 2, Figure 1). Females with calves had a lower travel speed compared to females without calves (Table 2, Figure 1). No support was shown for models including annual wolf predation risk (Table 1).
The ΔAIC c for models combining seasonal wolf predation risk with season and reproductive status ranged between 5.1 and 6.2 (Table 1).
The models including seasonal wolf predation risk thereby did not provide any net reduction in ΔAIC c (Arnold, 2010) compared to the top models only including season and reproductive status (ΔAIC c = 0, Table 1). The results were independent of the home range estimation method (Table 1). Travel speed between consecutive locations averaged 49.9 m/hr ± 0.6 SE, and the best model included the interaction effect between season and reproductive status (Tables 2 and 4).
No support was shown for models including the instantaneous wolf predation risk (Table 4). Neither was any support shown for models including the instantaneous wolf predation risk as a categorical variable while using a subset of the data (Table 4).

| Direction of movement
Direction of movement averaged 0.79 ± 0.003 SE during all seasons. The best model explaining variation in directions of movement included the interaction effect between season and reproductive status (Table 3). Movements with a low degree of directionality during the calving season were most pronounced for females with calves (Table 2, Figure 2). No support was shown for models including annual or seasonal wolf predation risk (Table 3) irrespective of the home range estimation method used (Table 3). Direction of movement between consecutive locations averaged 0.76 ± 0.002 SE, and the best model included the interaction effect between season and reproductive status (Tables 2 and 4). No support was shown for models including the instantaneous wolf predation risk (Table 4). Neither was any support shown for models including the instantaneous wolf predation risk as a categorical variable while using a subset of the data (Table 4).

| DISCUSSION
This study gave no support for the hypothesis that the re-establishment of wolves in Sweden has affected mobility in terms of either travel speed or direction of movement of female moose. However, both travel speed and direction of movement were affected by seasonal changes and reproductive status. Travel speed of females was highest during the calving season and the postcalving season, and reduced during the rest of the year. This is in line with previous studies that show that moose movement rates peak sometime during May to September (Cederlund, 1989;Eriksen et al., 2011;Vander Wal & Rodgers, 2009) and gradually reduce from October through November (Eriksen et al., 2011) to the lowest around February (Cederlund, 1989). This variation in movement rates follows seasonal changes because activity patterns are highly correlated with food quality and availability (Cederlund, 1989;Cederlund, Bergström, & Sandegren, 1989;Renecker & Hudson, 1986 variation in the directionally of movements of female moose between seasons, except for a reduction in June, which they explained were restricted movements due to the limited movement abilities of newborn calves. We found no support for the prediction that females with calves should be more prone to changes in their mobility in relation to wolf predation risk than females without calves. Although behaviorally mediated effects have been examined for several traits in Scandinavian moose as a response to wolf recolonization, no significant effect on moose behavior has so far been shown. Behavioral effects investigated include moose defense behavior against attacking wolves Sand et al., 2006aSand et al., , 2006bWikenros et al., 2009), daily and seasonal activity patterns between wolves and moose (Eriksen et al., 2008(Eriksen et al., , 2011, and moose habitat selection (Nicholson et al., 2014). The results from our present study add further support for the view that recolonization of a predator to previously inhabited areas does not always and universally lead to changes in the behavior of their prey species. In our study area, wolves first established a territory 3 years prior to the GPS collaring of moose and the data collected span over a 3-to 6-year period after the establishment of wolves. An alternative interpretation of our results is therefore that the time of exposure of wolves has been too short to initialize a behavioral response in the moose population. However, there are a number of studies that have reported rapid behavioral responses of prey as a result of resumed levels of predation risk (Berger, 1999;Berger et al., 2001;Hunter & Skinner, 1998;Laundré et al., 2001), suggesting that behaviorally mediated effects may show a rapid manifestation in the prey population.
Our results contrast with results from a number of studies in North America, showing that prey species, for example, elk (Cervus elaphus), bison (Bison bison), and moose, can show behavioral adaptation toward the presence of wolves with adjusting habitat (Stephens & Peterson, 1984), increased vigilance (Berger, 1999;Berger et al., 2001;White & Berger, 2001), shift in feeding and birthing sites (Edwards, 1983, or aggressive behavior (Mech & Peterson, 2003 In this perspective, our results may be considered as rather surprising. However, not all studies investigating behaviorally mediated effects on prey as a result of resumed predation risk by wolves have been able to confirm that these effects always exist even in North America (Kauffman, Brodie, & Jules, 2010). We suggest that there are several causal factors that together may explain why we do not find support for the hypothesis that the recolonization of wolves will lead to behaviorally mediated effects on prey in Scandinavia. The moose population in Scandinavia has experienced a strong decline during the 18th to early 20th century (Niedziałkowska et al., 2015), which to some extent was contemporary to the extinction of wolves from central Scandinavia in the mid-19th century. As a result, the rebounding of the moose population starting in the late 19th century became mainly regulated by hunter harvest and has since then been by far the most important mortality factor of the moose population (Lavsund et al., 2003).
Although wolves have been present in some areas in Scandinavia for more than 20 years Sand et al., 2006a) and have been shown to prey mainly on moose with relatively high kill rates (Sand et al., 2005Zimmermann, Sand, Wabakken, Liberg, & Andreassen, 2015), hunter harvest typically remains the main mortality factor even within most Scandinavian wolf territories (Wikenros et al., 2015). Also, the relatively large current size of wolf territories  and high density of moose (Sand et al., 2012) both contribute to create low ratios of wolf-to-moose (Eriksen et al., 2011;Sand et al., 2012) in Scandinavia. A direct consequence of the low wolf-to-moose ratio is that the frequency of encounters between wolves and any particular moose individual is low (Eriksen et al., 2008) that is also confirmed in this study where the distance between moose and wolves on average was 11 km.
In contrast to the predator-prey system in Scandinavia, most of the studies in North America have been carried out in protected areas such as national parks (Mech, 2012). Therefore, one factor that may explain these variable results may be the degree of anthropogenic impact on the ecosystem and on prey populations in particular. The impact of hunter harvest as an evolutionary force relative to predation by large predators has so far received little attention but is limited mainly by the access to empirical long-term data on morphology and behavior (Darimont et al., 2009;Fenberg & Roy, 2008). Further, because most of the prey populations now exposed to recolonizing populations of large predators (Chapron et al., 2014) is, and for a long time has been, under strong anthropogenic influence, the results received from the current and previous studies in Scandinavia on behaviorally mediated effects on prey may be more the norm as compared to studies carried out in protected areas.
We conclude that in a moose population where hunter harvest is the main mortality factor, the movement pattern of female moose was mainly influenced by external factors and reproductive status, and not by the return of their long absent natural predator, the wolf.