Oecologia DOI 10.1007/s00442-014-3004-9
Behavioral ecology - Original research
Testing the risk of predation hypothesis: the influence of recolonizing wolves on habitat use by moose Kerry L. Nicholson · Cyril Milleret · Johan Månsson · Håkan Sand
Received: 22 April 2013 / Accepted: 25 June 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract Considered as absent throughout Scandinavia for >100 years, wolves (Canis lupus) have recently naturally recolonized south-central Sweden. This recolonization has provided an opportunity to study behavioral responses of moose (Alces alces) to wolves. We used satellite telemetry locations from collared moose and wolves to determine whether moose habitat use was affected by predation risk based on wolf use distributions. Moose habitat use was influenced by reproductive status and time of day and showed a different selection pattern between winter and summer, but there was weak evidence that moose habitat use depended on predation risk. The seemingly weak response may have several underlying explanations that are not mutually exclusive from the long term absence of nonhuman predation pressure: intensive harvest by humans during the last century is more important than wolf predation as an influence on moose behavior; moose have not adapted to recolonizing wolves; and responses may include other behavioral adaptations or occur at finer temporal and spatial levels than investigated.
Communicated by Ilpo Kojola. Electronic supplementary material The online version of this article (doi:10.1007/s00442-014-3004-9) contains supplementary material, which is available to authorized users. K. L. Nicholson (*) · C. Milleret · J. Månsson · H. Sand Department of Ecology, Grimsö Wildlife Research Station, Swedish University of Agricultural Sciences, 730 91 Riddarhyttan, Sweden e-mail: [email protected] C. Milleret Faculty of Forestry and Wildlife Management, Hedmark University College, Evenstad, 2480 Koppang, Norway
Keywords Anti-predator behaviour · Foraging tradeoffs · Habitat shift · Landscape of risk · Trophic cascade · Resource selection
Introduction The risk of predation influences prey to make decisions that decrease their vulnerability (Lima and Dill 1990). Behavioral changes by prey reflect the need to balance the cost that an anti-predatory behavior can ultimately have with acquisition of necessary resources. Thus, there is significant interest in understanding the influence of predator presence and perceived predation risk on prey behavior. Behavioral adaptations include prey changing their use of space (distribution patterns; Creel et al. 2008; Lima and Dill 1990), vigilance to detect predators (Abramsky et al. 2002; Laundré et al. 2001), changing group sizes (Caro 2005; Lima 1995), or temporal activity patterns (Lima and Dill 1990). These behavioral changes may be costly to the animals as they are allocating time to wariness and increasing security, rather than to foraging, offspring, or mate choice (Abramsky et al. 2002; Edwards 1983; Laundré et al. 2001). Isolation from predators should result in selection against such costly anti-predator behavior (Blumstein and Daniel 2005) and, therefore, as isolation persists, prey may lose or change their anti-predator behavior (Byers 1998). Because predation risk is a dynamic spatio-temporal process, where certain habitats or time periods may pose more of a risk than others (Brown 1992; Brown et al. 1999), variation in prey behavior will be an expected response. Furthermore, individuals that are more subject to predation are expected to express a higher response to predation risk (Creel and Christianson 2008). Therefore, the cost of an anti-predator behavior is likely to result in individual
13
Oecologia
variation in terms of fitness (Lima and Dill 1990) that have consequences for population dynamics (Creel and Christianson 2008). The risk and intensity of response may vary according to age, sex, reproductive status, and the level of nutritional stress of the prey (Bjørneraas et al. 2011; Dussault et al. 2005; Edwards 1983; Vijayan et al. 2012). For example, females with offspring are likely at greater risk because predators often target the calves or young. Vulnerability can also increase due to external influences such as severe winters affecting locomotion and the accessibility of forage (Mech et al. 2001) or during night when activity of prey and predators peak (Eriksen et al. 2011; Sand et al. 2005; Winnie and Creel 2007). A number of studies have investigated the effect of the re-introduction of large predators on the re-establishment of anti-predatory behavior of prey related to habitat selection (Creel et al. 2005; Edwards 1983; Hernández and Laundré 2005; Mao et al. 2005), the dynamics of prey population structure, and abundance (Hebblewhite et al. 2002), and the subsequent impacts on trophic cascades (Berger et al. 2001; Fortin et al. 2005; Ripple and Beschta 2004; Smith et al. 2003). In some studies with prolonged predator absences, prey populations swiftly regained their former anti-predatory behavior (Berger 1999; Berger et al. 2001; Creel and Christianson 2008; Hunter and Skinner 1998; Laundré et al. 2001). In contrast, others have found no or weak effects of predator recolonization (Kauffman et al. 2010) as predation risk may not induce strong effects on prey behavior, demography, or trophic cascades (Thaker et al. 2011). Moose (Alces alces) are browsers, and their habitat selection is dependent on seasonal forage availability (Cederlund and Okarma 1988; Olsson et al. 2011; Van Beest et al. 2010), time of day, sex, and reproductive status of the cow (Bjørneraas et al. 2011, 2012). Moose in wolf (Canis lupus)-free areas of Scandinavia select coniferous forest that provide good forage and cover. Bjørneraas et al. (2012) suggested that this is likely a behavioral response to predation risk by humans. In Scandinavia, the wolf was extensively persecuted in the nineteenth and twentieth centuries and was exterminated and absent from Scandinavia by the mid-to-late 1800s (Lönnberg 1934). They were classified as functionally extinct from Sweden by the late 1960s (Haglund 1968). During this >100 year absence, the moose population grew, and hunting by humans replaced most of the natural mortality (Cederlund and Sand 1991; Sæther et al. 1996; Stubsjøen et al. 2000). Wolves began to recolonize central Scandinavia in the early 1980s (Wabakken et al. 2001) resulting in an increased risk of predation mortality, and thus a unique opportunity to study numerical and behavioral effects of wolf re-establishment. With wolves restored to the system,
13
moose now face substantially greater risk of predation, which potentially could prompt them to alter their habitat use patterns. Thus, our objective was to examine the influence of predation risk on individually based habitat use of moose. We hypothesized that predation risk would be an important factor affecting the pattern of habitat use of moose in addition to season, time of day, sex, and reproductive status. Therefore, we expected that exposure to a wolf should be a significant predictor variable of habitat use among individual moose. We expect moose to use less risky habitat (i.e., dense coniferous forests; Kunkel and Pletscher 2000) compared to more risky habitats (i.e., clear-cuts, young forest plantations; Gervasi et al. 2013). Because resources, and thus habitat use, are variable among seasons, we also expect moose to respond differently to the risk of predation between seasons. Similarly, we expect wolf predation risk to be greater during the night than during the day, as previous studies indicate wolf predation events occur primarily at night (Sand et al. 2005). Finally, we expect wolf risk to vary across sex and reproductive status because wolves select calves first, and females (f) before male (m) moose (Sand et al. 2005, 2008, 2012), with females and calves at heel (fc) showing the strongest response in changing habitat due to predation risk.
Materials and methods Study area We conducted the study in the southern boreal zone in south-central Sweden which encompassed the Grimsö Wildlife Research Area (59–60°N and 15–16°E). Elevation ranged from 100 to 150 m. Topography was composed of low, flat ridges with till and boulders interspersed with bogs. The annual mean temperature was 5 °C, and the monthly temperature averaged between −4 and −6 °C in winter (i.e., December to February; Vedin 1995). Total annual mean precipitation was 600–700 mm, of which 30 % (180–210 mm) fell as snow, and normally covered the ground from December to late March; the mean snow depth was 20–30 cm (Alexandersson and Andersson 1995). The period of vegetative growth (number of days with a mean temperature >5 °C) was 160 days (Alexandersson and Andersson 1995). The area consisted of forests (78 %), bogs (8 %), lakes and rivers (6 %), and meadows and farmland (8 %). Forest stands are dominated by Scots pine (Pinus sylvestris) and Norway spruce (Picea abies), and are sometimes mixed with birch (Betula pubescens and B. pendula), aspen (Populus tremula), and willows (Salix). Forest stand size ranged from 0.5 to 64 ha with a mean size of 6 ha. The
Oecologia
forest was intensively logged for several centuries for timber and pulp. Meadows and fields consisted primarily of dwarf shrubs like bilberry (Vaccinium myrtillus) and lingonberry (Vaccinium vitis-idaea) on the forested land, with dwarf birch (Betula nana) and heather (Calluna vulgaris) in the bogs. Clear-cuts were dominated by grasses, particularly wavy hair grass (Deschampsia flexuosa). Aerial counts of moose in 2002 and 2006 yielded densities of 1.2 and 0.8 moose/km2, respectively (Rönnegård et al. 2008). The moose population was harvested annually, and hunting was the main mortality factor (Rönnegård et al. 2008). The only other ungulate within the study area was roe deer (Capreolus capreolus), whose population densities ranged between one and five roe deer/km2 (Rönnegård et al. 2008). The other established large predator in the study area was the Eurasian lynx (Lynx lynx) whose main prey is roe deer Liberg et al. (2010). In 2003, wolves naturally re-established in the study area and the population was named the Uttersberg pack. From 2003 to 2005, this pack was monitored by snow tracking, and by snow tracking and Global Positioning System (GPS) collars from 2005 to 2011 (for description of census techniques and population development, see Liberg et al. 2010, 2011). In 2009, this pack was replaced by the Hedbyn pack (Wabakken et al. 2011). Capture and handling Adult moose were captured in March 2007 (n = 24) and 2010 (n = 15). We tranquilized moose by dart gun from a helicopter (Arnemo et al. 2003) and fitted them with GPS/ Global System for Mobile Communications (GSM) collars (GPS/GSM Plus 4D; Vectronic Aeorospace GmbH, Berlin, Germany) set to acquire locations every 2 h. Wolves were immobilized in winter from the air (Arnemo et al. 2007) and fitted with GPS collars (GPS/GSM Plus 1D; Vectronic Aeorospace). Over this time, four wolves (the alpha pairs) were collared and continuously monitored except during a 4-month period when the Uttersberg pack was replaced by the Hedbyn pack. Handling protocols fulfilled the ethical requirements for research on wild animals in Sweden (decision C315/6). Defining covariates We explored moose habitat use in relation to the variables of season, time of day, sex, reproductive status, and wolf risk. Mean calving date for females was the 20th of May (n = 31; range 11 May–1 June). Gestating females may modify their movement and habitat use 1 month before parturition (Ciuti et al. 2006); therefore, females with a calf at heal later during the season were classified as female with calf before their parturition (from 1 May). To monitor
reproductive status of the cow, field personnel surveyed the moose three times per year: in May for calving success, July–August for summer survival, and April for winter survival. When a female lost her calf between two survey occasions (i.e., unknown date of death) or when field check was impossible because of non-working VHF-beacon collar device, the individual was excluded from the analysis. The year was divided into winter and summer. Because the numbers of days with snow cover ≥10 cm affect moose habitat use (Månsson 2009), the period with ≥10 cm snow depth was defined as winter. For our study this period ranged from 28 November ± 18 days to 24 March ± 10 days; therefore, winter was set as 1 November to 31 March each year. Snow depth data was obtained from the Spannarboda weather station, 4 km southwest of the study area (Swedish Meteorological and Hydrological Institute, http://www.smhi.se). The summer season was defined as 1 May to the 31 August, which includes periods of vegetative growth, but excludes the rutting season (Olsson et al. 2011). Location data were further categorized into day/night according to mean monthly sunset and sunrise times. Habitat and topographic variables We used the Swedish CORINE Land Cover map (SMD) and reclassified landforms into eight habitats (Online Resource Table 1). Because logging activity is intense in the study area, we created vegetation maps for each year to account for yearly changes due to clear-cutting and successional development. For each year of the study, new clearcuts (obtained from the Swedish Forestry Board) were updated, and previous years’ clear-cuts (starting after 2007) were transitioned into the young forest stage. Because we were concerned with designating biologically significant habitat for moose, available water habitat was limited to a 20-m buffer inside from the water edge. This effectively removed the center or deepest part of the lakes, yet retained water as a habitat available for use. Moose may use water in the summer for foraging or as escape habitat and in the winter as corridors for crossing between habitat patches. We obtained elevation from a 50 × 50 m digital elevation map (Geographical Data Sweden; GSD, National Land Survey of Sweden). We calculated terrain ruggedness (TRI; Sappington et al. 2007) in ArcGIS 9.3.1TM (© 2009 ESRI Inc.) with a Vector Ruggedness Measure extension with a neighborhood size of three. We then converted elevation and TRI rasters into a 25 × 25 m grid cell size. We calculated the Normalized Difference Vegetation Index (NDVI) from the RED (band 2) and Near Infrared (band 3) of the IRS-P6_LISS-3 satellite imagery taken on 14 July 2010 (www.saccess.lantmateriet.se, Accessed 05 March 2012). This imagery indicates net primary above-ground
13
Oecologia
production and is often used as a proxy for vegetation density or as a greenness index (Pettorelli et al. 2005). We derived maps of distances to anthropogenic structures (lowand high-traffic roads, houses, and settlements) from digital data of Sweden (GSD-Översiktskartan, Lantmäteriet, Sweden). Data handling We screened moose GPS-data for position errors following the non-movement method developed by Bjørneraas et al. (2010) in program R (R Development Core Team 2010). Distance parameters were set as Δ = 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/h and turning angle θ = −0.97 (Bjørneraas et al. 2010) to identify spurious locations that formed a spike. These error locations (n = 125, ranged between 0 and 10 per individual) were removed from the analysis. All locations from the 7 days post-capture were excluded to avoid the effect of immobilization on moose behavior a
b
c
d
Fig. 1 Summer (a, c) and winter (b, d) habitat selection of 35 moose in south-central Sweden within their home ranges 2007–2011. a, b Variable loadings on the first two factorial axes (axis 1: x axis; axis 2: y axis). c, d The marginality vectors of individuals after recentering on each individual home-range composition (i.e., the origin of the
13
(Neumann et al. 2011). Finally, moose with
Behavioral ecology - Original research
Testing the risk of predation hypothesis: the influence of recolonizing wolves on habitat use by moose Kerry L. Nicholson · Cyril Milleret · Johan Månsson · Håkan Sand
Received: 22 April 2013 / Accepted: 25 June 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract Considered as absent throughout Scandinavia for >100 years, wolves (Canis lupus) have recently naturally recolonized south-central Sweden. This recolonization has provided an opportunity to study behavioral responses of moose (Alces alces) to wolves. We used satellite telemetry locations from collared moose and wolves to determine whether moose habitat use was affected by predation risk based on wolf use distributions. Moose habitat use was influenced by reproductive status and time of day and showed a different selection pattern between winter and summer, but there was weak evidence that moose habitat use depended on predation risk. The seemingly weak response may have several underlying explanations that are not mutually exclusive from the long term absence of nonhuman predation pressure: intensive harvest by humans during the last century is more important than wolf predation as an influence on moose behavior; moose have not adapted to recolonizing wolves; and responses may include other behavioral adaptations or occur at finer temporal and spatial levels than investigated.
Communicated by Ilpo Kojola. Electronic supplementary material The online version of this article (doi:10.1007/s00442-014-3004-9) contains supplementary material, which is available to authorized users. K. L. Nicholson (*) · C. Milleret · J. Månsson · H. Sand Department of Ecology, Grimsö Wildlife Research Station, Swedish University of Agricultural Sciences, 730 91 Riddarhyttan, Sweden e-mail: [email protected] C. Milleret Faculty of Forestry and Wildlife Management, Hedmark University College, Evenstad, 2480 Koppang, Norway
Keywords Anti-predator behaviour · Foraging tradeoffs · Habitat shift · Landscape of risk · Trophic cascade · Resource selection
Introduction The risk of predation influences prey to make decisions that decrease their vulnerability (Lima and Dill 1990). Behavioral changes by prey reflect the need to balance the cost that an anti-predatory behavior can ultimately have with acquisition of necessary resources. Thus, there is significant interest in understanding the influence of predator presence and perceived predation risk on prey behavior. Behavioral adaptations include prey changing their use of space (distribution patterns; Creel et al. 2008; Lima and Dill 1990), vigilance to detect predators (Abramsky et al. 2002; Laundré et al. 2001), changing group sizes (Caro 2005; Lima 1995), or temporal activity patterns (Lima and Dill 1990). These behavioral changes may be costly to the animals as they are allocating time to wariness and increasing security, rather than to foraging, offspring, or mate choice (Abramsky et al. 2002; Edwards 1983; Laundré et al. 2001). Isolation from predators should result in selection against such costly anti-predator behavior (Blumstein and Daniel 2005) and, therefore, as isolation persists, prey may lose or change their anti-predator behavior (Byers 1998). Because predation risk is a dynamic spatio-temporal process, where certain habitats or time periods may pose more of a risk than others (Brown 1992; Brown et al. 1999), variation in prey behavior will be an expected response. Furthermore, individuals that are more subject to predation are expected to express a higher response to predation risk (Creel and Christianson 2008). Therefore, the cost of an anti-predator behavior is likely to result in individual
13
Oecologia
variation in terms of fitness (Lima and Dill 1990) that have consequences for population dynamics (Creel and Christianson 2008). The risk and intensity of response may vary according to age, sex, reproductive status, and the level of nutritional stress of the prey (Bjørneraas et al. 2011; Dussault et al. 2005; Edwards 1983; Vijayan et al. 2012). For example, females with offspring are likely at greater risk because predators often target the calves or young. Vulnerability can also increase due to external influences such as severe winters affecting locomotion and the accessibility of forage (Mech et al. 2001) or during night when activity of prey and predators peak (Eriksen et al. 2011; Sand et al. 2005; Winnie and Creel 2007). A number of studies have investigated the effect of the re-introduction of large predators on the re-establishment of anti-predatory behavior of prey related to habitat selection (Creel et al. 2005; Edwards 1983; Hernández and Laundré 2005; Mao et al. 2005), the dynamics of prey population structure, and abundance (Hebblewhite et al. 2002), and the subsequent impacts on trophic cascades (Berger et al. 2001; Fortin et al. 2005; Ripple and Beschta 2004; Smith et al. 2003). In some studies with prolonged predator absences, prey populations swiftly regained their former anti-predatory behavior (Berger 1999; Berger et al. 2001; Creel and Christianson 2008; Hunter and Skinner 1998; Laundré et al. 2001). In contrast, others have found no or weak effects of predator recolonization (Kauffman et al. 2010) as predation risk may not induce strong effects on prey behavior, demography, or trophic cascades (Thaker et al. 2011). Moose (Alces alces) are browsers, and their habitat selection is dependent on seasonal forage availability (Cederlund and Okarma 1988; Olsson et al. 2011; Van Beest et al. 2010), time of day, sex, and reproductive status of the cow (Bjørneraas et al. 2011, 2012). Moose in wolf (Canis lupus)-free areas of Scandinavia select coniferous forest that provide good forage and cover. Bjørneraas et al. (2012) suggested that this is likely a behavioral response to predation risk by humans. In Scandinavia, the wolf was extensively persecuted in the nineteenth and twentieth centuries and was exterminated and absent from Scandinavia by the mid-to-late 1800s (Lönnberg 1934). They were classified as functionally extinct from Sweden by the late 1960s (Haglund 1968). During this >100 year absence, the moose population grew, and hunting by humans replaced most of the natural mortality (Cederlund and Sand 1991; Sæther et al. 1996; Stubsjøen et al. 2000). Wolves began to recolonize central Scandinavia in the early 1980s (Wabakken et al. 2001) resulting in an increased risk of predation mortality, and thus a unique opportunity to study numerical and behavioral effects of wolf re-establishment. With wolves restored to the system,
13
moose now face substantially greater risk of predation, which potentially could prompt them to alter their habitat use patterns. Thus, our objective was to examine the influence of predation risk on individually based habitat use of moose. We hypothesized that predation risk would be an important factor affecting the pattern of habitat use of moose in addition to season, time of day, sex, and reproductive status. Therefore, we expected that exposure to a wolf should be a significant predictor variable of habitat use among individual moose. We expect moose to use less risky habitat (i.e., dense coniferous forests; Kunkel and Pletscher 2000) compared to more risky habitats (i.e., clear-cuts, young forest plantations; Gervasi et al. 2013). Because resources, and thus habitat use, are variable among seasons, we also expect moose to respond differently to the risk of predation between seasons. Similarly, we expect wolf predation risk to be greater during the night than during the day, as previous studies indicate wolf predation events occur primarily at night (Sand et al. 2005). Finally, we expect wolf risk to vary across sex and reproductive status because wolves select calves first, and females (f) before male (m) moose (Sand et al. 2005, 2008, 2012), with females and calves at heel (fc) showing the strongest response in changing habitat due to predation risk.
Materials and methods Study area We conducted the study in the southern boreal zone in south-central Sweden which encompassed the Grimsö Wildlife Research Area (59–60°N and 15–16°E). Elevation ranged from 100 to 150 m. Topography was composed of low, flat ridges with till and boulders interspersed with bogs. The annual mean temperature was 5 °C, and the monthly temperature averaged between −4 and −6 °C in winter (i.e., December to February; Vedin 1995). Total annual mean precipitation was 600–700 mm, of which 30 % (180–210 mm) fell as snow, and normally covered the ground from December to late March; the mean snow depth was 20–30 cm (Alexandersson and Andersson 1995). The period of vegetative growth (number of days with a mean temperature >5 °C) was 160 days (Alexandersson and Andersson 1995). The area consisted of forests (78 %), bogs (8 %), lakes and rivers (6 %), and meadows and farmland (8 %). Forest stands are dominated by Scots pine (Pinus sylvestris) and Norway spruce (Picea abies), and are sometimes mixed with birch (Betula pubescens and B. pendula), aspen (Populus tremula), and willows (Salix). Forest stand size ranged from 0.5 to 64 ha with a mean size of 6 ha. The
Oecologia
forest was intensively logged for several centuries for timber and pulp. Meadows and fields consisted primarily of dwarf shrubs like bilberry (Vaccinium myrtillus) and lingonberry (Vaccinium vitis-idaea) on the forested land, with dwarf birch (Betula nana) and heather (Calluna vulgaris) in the bogs. Clear-cuts were dominated by grasses, particularly wavy hair grass (Deschampsia flexuosa). Aerial counts of moose in 2002 and 2006 yielded densities of 1.2 and 0.8 moose/km2, respectively (Rönnegård et al. 2008). The moose population was harvested annually, and hunting was the main mortality factor (Rönnegård et al. 2008). The only other ungulate within the study area was roe deer (Capreolus capreolus), whose population densities ranged between one and five roe deer/km2 (Rönnegård et al. 2008). The other established large predator in the study area was the Eurasian lynx (Lynx lynx) whose main prey is roe deer Liberg et al. (2010). In 2003, wolves naturally re-established in the study area and the population was named the Uttersberg pack. From 2003 to 2005, this pack was monitored by snow tracking, and by snow tracking and Global Positioning System (GPS) collars from 2005 to 2011 (for description of census techniques and population development, see Liberg et al. 2010, 2011). In 2009, this pack was replaced by the Hedbyn pack (Wabakken et al. 2011). Capture and handling Adult moose were captured in March 2007 (n = 24) and 2010 (n = 15). We tranquilized moose by dart gun from a helicopter (Arnemo et al. 2003) and fitted them with GPS/ Global System for Mobile Communications (GSM) collars (GPS/GSM Plus 4D; Vectronic Aeorospace GmbH, Berlin, Germany) set to acquire locations every 2 h. Wolves were immobilized in winter from the air (Arnemo et al. 2007) and fitted with GPS collars (GPS/GSM Plus 1D; Vectronic Aeorospace). Over this time, four wolves (the alpha pairs) were collared and continuously monitored except during a 4-month period when the Uttersberg pack was replaced by the Hedbyn pack. Handling protocols fulfilled the ethical requirements for research on wild animals in Sweden (decision C315/6). Defining covariates We explored moose habitat use in relation to the variables of season, time of day, sex, reproductive status, and wolf risk. Mean calving date for females was the 20th of May (n = 31; range 11 May–1 June). Gestating females may modify their movement and habitat use 1 month before parturition (Ciuti et al. 2006); therefore, females with a calf at heal later during the season were classified as female with calf before their parturition (from 1 May). To monitor
reproductive status of the cow, field personnel surveyed the moose three times per year: in May for calving success, July–August for summer survival, and April for winter survival. When a female lost her calf between two survey occasions (i.e., unknown date of death) or when field check was impossible because of non-working VHF-beacon collar device, the individual was excluded from the analysis. The year was divided into winter and summer. Because the numbers of days with snow cover ≥10 cm affect moose habitat use (Månsson 2009), the period with ≥10 cm snow depth was defined as winter. For our study this period ranged from 28 November ± 18 days to 24 March ± 10 days; therefore, winter was set as 1 November to 31 March each year. Snow depth data was obtained from the Spannarboda weather station, 4 km southwest of the study area (Swedish Meteorological and Hydrological Institute, http://www.smhi.se). The summer season was defined as 1 May to the 31 August, which includes periods of vegetative growth, but excludes the rutting season (Olsson et al. 2011). Location data were further categorized into day/night according to mean monthly sunset and sunrise times. Habitat and topographic variables We used the Swedish CORINE Land Cover map (SMD) and reclassified landforms into eight habitats (Online Resource Table 1). Because logging activity is intense in the study area, we created vegetation maps for each year to account for yearly changes due to clear-cutting and successional development. For each year of the study, new clearcuts (obtained from the Swedish Forestry Board) were updated, and previous years’ clear-cuts (starting after 2007) were transitioned into the young forest stage. Because we were concerned with designating biologically significant habitat for moose, available water habitat was limited to a 20-m buffer inside from the water edge. This effectively removed the center or deepest part of the lakes, yet retained water as a habitat available for use. Moose may use water in the summer for foraging or as escape habitat and in the winter as corridors for crossing between habitat patches. We obtained elevation from a 50 × 50 m digital elevation map (Geographical Data Sweden; GSD, National Land Survey of Sweden). We calculated terrain ruggedness (TRI; Sappington et al. 2007) in ArcGIS 9.3.1TM (© 2009 ESRI Inc.) with a Vector Ruggedness Measure extension with a neighborhood size of three. We then converted elevation and TRI rasters into a 25 × 25 m grid cell size. We calculated the Normalized Difference Vegetation Index (NDVI) from the RED (band 2) and Near Infrared (band 3) of the IRS-P6_LISS-3 satellite imagery taken on 14 July 2010 (www.saccess.lantmateriet.se, Accessed 05 March 2012). This imagery indicates net primary above-ground
13
Oecologia
production and is often used as a proxy for vegetation density or as a greenness index (Pettorelli et al. 2005). We derived maps of distances to anthropogenic structures (lowand high-traffic roads, houses, and settlements) from digital data of Sweden (GSD-Översiktskartan, Lantmäteriet, Sweden). Data handling We screened moose GPS-data for position errors following the non-movement method developed by Bjørneraas et al. (2010) in program R (R Development Core Team 2010). Distance parameters were set as Δ = 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/h and turning angle θ = −0.97 (Bjørneraas et al. 2010) to identify spurious locations that formed a spike. These error locations (n = 125, ranged between 0 and 10 per individual) were removed from the analysis. All locations from the 7 days post-capture were excluded to avoid the effect of immobilization on moose behavior a
b
c
d
Fig. 1 Summer (a, c) and winter (b, d) habitat selection of 35 moose in south-central Sweden within their home ranges 2007–2011. a, b Variable loadings on the first two factorial axes (axis 1: x axis; axis 2: y axis). c, d The marginality vectors of individuals after recentering on each individual home-range composition (i.e., the origin of the
13
(Neumann et al. 2011). Finally, moose with
