Predation on two vole species by a shared predator: antipredatory response and prey preference

Popul Ecol (2008) 50:257 266 DOI 10.1007/s10144-008-0086-4 ORIGINAL ARTICLE Predation on two vole species by a shared predator: antipredatory response and prey preference Janne Sundell Æ Lenka Trebatická Æ Tarja Oksanen Æ Otso Ovaskainen Æ Marko Haapakoski Æ Hannu Ylönen Received: 24 July 2007 / Accepted: 31 March 2008 / Published online: 8 May 2008 Ó The Society of Population Ecology and Springer 2008 Abstract In prey communities with shared predators, variation in prey vulnerability is a key factor in shaping community dynamics. Conversely, the hunting efficiency of a predator depends on the prey community structure, preferences of the predator and antipredatory behavioural traits of the prey. We studied experimentally, under seminatural field conditions, the preferences of a predator and the antipredatory responses of prey in a system consisting of two Myodes species of voles, the grey-sided vole (M. rufocanus Sund.) and the bank vole (M. glareolus Schreb.), and their specialist predator, the least weasel (Mustela nivalis nivalis L.). To quantify the preference of the weasels, we developed a new modelling framework that can be used for unbalanced data. The two vole species were hypothesised to have different habitat-dependent vulnerabilities. We created two habitats, open and forest, to provide different escape possibilities for the voles. We found a weak general preference of the weasels for the grey-sided voles over the bank voles, and a somewhat stronger preference J. Sundell (&) O. Ovaskainen Metapopulation Research Group, Department of Biological and Environmental Sciences, University of Helsinki, P.O. Box 65, 00014 Helsinki, Finland e-mail: janne.sundell@helsinki.fi L. Trebatická M. Haapakoski H. Ylönen Konnevesi Research Station, Department of Biological and Environmental Science, University of Jyväskylä, P.O. Box 35, 40014 Jyväskylä, Finland T. Oksanen Department of Ecology and Environmental Sciences, Umeå University, 90187 Umeå, Sweden specifically in open habitats. The weasels clearly preferred male grey-sided voles over females, whereas in bank voles, there was no difference. The activity of voles changed over time, so that voles increased their movements immediately after weasel introduction, but later adjusted their movements to times of lowered predation risk. Females that were more active had an elevated mortality risk, whereas in the case of males, the result was the opposite. We conclude that, in vulnerability to predation, the species- or habitatspecific characteristics of these prey species are playing a minor role compared to sex-specific characteristics. Keywords Antipredatory behaviour Apparent competition Clethrionomys Prey preference Introduction Predators have the potential to shape the dynamics of their prey. Differential predation on different prey types or species can also modify the prey community structure. One way in which this happens is predator-mediated competition, in which the increase of one prey type has a negative effect on the other prey type indirectly via a shared predator (i.e., apparent competition sensu Holt 1977). In a predator-mediated apparent competition, predators can react numerically or behaviourally to density changes of one of the coexisting prey species, which increases predation also on the other alternative prey species. This is observable as negative effects between prey species, even without actual competition (Holt 1977; Holt and Kotler 1987). The strength of a predator s power to modify the prey community depends on its specialisation (generalists vs. specialists; e.g., Hanski et al. 1991), prey preferences and the vulnerability of different prey types (e.g., Abrams

258 Popul Ecol (2008) 50:257 266 et al. 1998), together with competitive interactions (Sih et al. 1985; Chase et al. 2002) and environmental factors (e.g., Menge and Sutherland 1976, 1987). According to Holt s (1977) analysis, the vulnerability of a given prey species depends on the ratio of its reproductive capacity (r) to the predator s searching efficiency (a). In principle, a prey can increase its r/a ratio in two ways; first, by being elusive and, thus, avoiding predation, which reduces the value of a, or by having high reproductive capacity, i.e., a high value of r. If predation risk is unpredictable and short-termed, the latter approach to a high-risk situation is unlikely to be optimal (Oksanen and Lundberg 1995), and the vulnerability of a given species can largely be assessed on the basis of its a value. Given the importance of this parameter for survival under predation risk, it is reasonable to ask why evolution has not resulted in the minimisation of a for all prey species. However, elusiveness has its costs. For example, it is impossible for a herbivore to be both extremely elusive and have the capacity to handle low-quality forage, as there is a trade-off between these two characteristics. Because of this apparent trade-off, herbivore species living in unproductive environments tend to have larger digestive tracks than their congeners living in environments with high-quality food (Oksanen 1992). The grey-sided vole, Myodes rufocanus Sund., and the bank vole, M. glareolus Schreb. (genus name Myodes, was previously Clethrionomys; nomenclature follows Wilson and Reeder 2005), form a suitable species pair for testing the above-mentioned conjecture on differential vulnerability, as they share the same ancestry, but the former species is adapted to less productive habitats than the latter. The grey-sided vole lives in a wide variety of habitats, ranging from forests and mire edges to open tundra (Christensen and Hörnfeldt 2006). The bank vole is also not restricted to any particular habitat type, and can be found commonly in forests and bushy edge habitats. Especially during the peak years of abundance, the two species can be found in the same habitats, and they are often the two most common species in southern Lapland (Henttonen and Hansson 1984). The grey-sided vole is slightly larger, relatively slow, eats more green vegetation and breeds in higher densities than in the other Myodes voles (Viitala 1977; Henttonen 2000). Thus, ecologically, the grey-sided vole resembles more species of another common genus of voles, Microtus voles, than other Myodes species (Hanski and Henttonen 1996). The Microtus voles are mainly folivorous, live in dense matrilinear aggregations (Pusenius et al. 1998) and are heavier and probably also slower in movements than Myodes. The bank vole, as a typical representative of Myodes, is more agile, partly arboreal and does not attain such high densities as Microtus voles. Because of its lower densities and better escape behaviour, weasel predation should not be targeted towards bank vole and it is considered as the alternative prey (Henttonen 1987; Hanski and Henttonen 1996). Although both Myodes and Microtus are primarily preyed upon by small mustelids, and especially by the least weasel (Mustela nivalis nivalis L.), Microtus voles are assumed to be the pivotal prey for breeding weasels and the key prey species in small mustelid vole interactions in cyclic vole populations of Fennoscandia (e.g., Korpimäki et al. 1991). Our purpose was to investigate the underlying factors behind potentially different vulnerabilities of the two prey species to weasel predation. Based on the characteristics of these different types of voles described above, and the characteristics of the hunting behaviour of the least weasel (Jedrzejewski and Jedrzejewska 1990), we hypothesised that the Microtus-like grey-sided vole would be more prone to predation by the least weasel than the typical Myodes, the bank vole. We expected that the less agile grey-sided voles would suffer from heavier predation by the mostly ground-dwelling least weasel and that the partly arboreal bank voles would survive better, especially in forested habitats, because of the escape possibilities to trees. Further, we aimed to investigate how behavioural traits such as space use and activity affect the vulnerability of the prey. According to general theory, increasing mobility should lead to a higher encounter rate with the predator and, consequently, to higher mortality (Lima and Dill 1990), but on the other hand, also, reduced movements with the accumulation of odorous waste products may expose prey to olfactorily hunting predators (Banks et al. 2000). Materials and methods Study design The study was conducted in eight 0.25-ha (50 9 50 m) enclosures in the vicinity of the Konnevesi Research Station of the University of Jyväskylä, Central Finland (62 37 0 N, 26 20 0 E) in September and October 2005. The walls of the enclosures were made of metal sheets reaching ca. 0.5 m below and 1.0 m above the ground. The enclosures were constructed on an old field in 1987. The main habitat is grassland, but bushes of Salix spp. cover about 1 10% of the enclosure areas. The terrain is generally flat with small dry ditches and some piles of stones. All of the enclosures had a few (ca. 10 per enclosure) saplings (2 6 m high) of birch (Betula spp.) and spruce (Picea abies). To create two different habitat types, we added young (3 5-m high) spruces by sticking them into the ground in four of the enclosures (to be called forest), while the remaining four were left as such (to be called

Popul Ecol (2008) 50:257 266 259 open). Spruces were erected within half of the area (0.125 ha) in each of the four enclosures, so that the total number of trees (added spruces and trees naturally growing in the enclosures) varied from 39 to 57 in the forest treatment. The trees in the forested parts of the enclosures were evenly distributed, with the distance between trees being never more than 7 m. The few trees naturally growing in the open enclosures and in the open half of the forested enclosures were treated by cutting off their lower branches. This procedure was evidently successful, as voles were never seen in these manipulated trees during the study. Small Salix bushes present in the open treatment were probably not suitable for climbing, as we did not observe in this study, nor have observed in our previous studies (e.g., Sundell et al. 2003), voles climbing on the thin and slippery branches mainly growing upwards. The aim of the habitat manipulation was to study the habitat-dependent vulnerability of the two vole species in the forest habitat providing arboreal escape possibilities. Both species have been considered to be able to climb on trees, but the observations of grey-sided voles in trees are largely anecdotal (e.g., Siivonen and Sulkava 1994). Bank voles are often seen collecting lichens for food from trees and are known to escape to trees when exposed to small mustelid predators (Jedrzejewski and Jedrzejewska 1990). The experiment was carried out in three subsequent runs, in which four enclosures were used simultaneously. Experimental animals For each enclosure in each replicate, eight adult mature voles ([19 g), four of each species, with sex ratio 1:1, were radio-collared (TW-4, Biotrack Ltd., Wareham, UK; ca. 1.5 g), marked subcutaneously with passive-induced transponders (Trovan, EID Aalten BV, Aalten, Holland) and released into the middle of a randomly assigned enclosure one day after marking and collaring. Experimental bank voles were captured near the Konnevesi Research Station or belonged to the first generation born in the laboratory or in the outdoor enclosures at the same place. All of the used bank voles had experienced natural conditions in enclosures as a part of an earlier experiment. Grey-sided voles were captured from Finnish, Swedish and Norwegian Lapland. They were brought to the Konnevesi Research Station before the experiment and were housed individually at the laboratory in standard mouse cages and fed with mouse pellets prior to the experiment. Even though the grey-sided vole is a slightly larger species than the bank vole (weight ranges from 20 to 50 g and from 7 to 40 g for grey-sided voles and bank voles, respectively, according to Jensen 1994), we wanted to control for the weight effect and minimise the proportional weight of the collar in this experiment by reducing the natural size difference between species. The mean weight was 24.9 g for bank voles (±2.4 g SD, range 20.0 30.0 g; for females 23.3 ± 1.6 g and for males 26.5 ± 1.9 g), and 25.9 g for grey-sided voles (±3.9 SD, 19.0 34.0 g; for females 25.0 ± 4.2 g and for males 26.9 ± 3.4 g). After 3 4 days habituation period in the enclosures, we started to monitor the space use of voles by locating them every second hour for a 24-h period. The possible use of trees was also recorded. Twenty trees were equipped with live traps (Ugglan Special, Grahnab AB, Hillerstorp, Sweden) to observe the use of trees between radio-trackings. After the habituation period of the voles, we released one least weasel per enclosure. The least weasels were wild-born or belonged to the first generation born in captivity. Altogether, 12 weasels, eight males and four females, were used in the experiment. The mean weights were 57.4 g (±4.4 SD, range 50.0 62.0 g) for males and 43.8 g (±1.2 SD, 43.0 46.0 g) for females. The radiocollared weasels (TW-4, Biotrack Ltd., Wareham, UK, ca. 2.2 g) were released in a randomly assigned enclosure s randomly assigned corner. Weasels were radio-located 3 4 times per 2 h. The weasels were let to hunt voles freely for 72 h or until they had captured four voles, after which, the experiment was terminated. Space use and activity To measure space use, we used the 100% minimum convex polygon method (MCP; Kenward 1987). We call in the sequel the area within the MCP as the home range, although in this kind of short-term enclosure experiment, the MCP reflects general space use rather than the size of an actual home range (Sundell et al. 2000). Home ranges were calculated with the program Ranges6 (Anatrack Ltd., Wareham, UK). We determined the activity of the experimental animals by measuring the mean distance moved between successive radiolocations (=interfix distance). We used data obtained during the entire experimental period or data collected within 24 h before and after the weasels were released, the latter for studying the effect of the predator on home ranges and activity patterns. The reason for restricting to a 24-h period after weasel release is a tendency for increasing home range size with increasing number of observations, especially when the number of observations is relatively small. The home range measures were log-transformed and proportions of habitat within a home range were arcsin-transformed to fulfil the assumptions of data for analysis of variance. Analyses were conducted using the MIXED procedure in the SAS Ò statistical software package, version 9.1. To control for variation in environmental conditions, the enclosures were nested within habitat types. Repeated measures analysis of variance was used when observations before and after

260 Popul Ecol (2008) 50:257 266 weasel release were compared. Linear regression analysis was used to analyse the dependence of vole activity on weasel activity (SPSS Ò statistical software package, version 12.0.1). Prior to the analysis of activity changes (r t = x t - x t-1 ), the data was log(x + 1) transformed for linearising the relationship. Cause and timing of death About half of the voles carried radio-collars with temperature sensors. Therefore, it was possible to determine the time of death with an accuracy of a few minutes, as the signal pulse gets faster with decreasing temperature. In the case of voles carrying ordinary radio-collars, we estimated the time of death using the following procedure. If the vole was inactive during three consecutive checks or if a weasel was observed near to its location, we tried to locate the body of the vole. The cause of death was deduced by searching for weasel bite marks from the vole and collar. Also, the location of the dead vole could be used to determine its cause of death, as voles killed by a weasel were usually hidden under the ground or amongst vegetation. All voles killed by weasels were left in the place where they were found. The timings of kills were used to investigate the killing order and preference of the weasels. For the sake of simplicity, here, we use the words preference and prefer, although the probability of an individual vole to fall as a prey for the weasels is likely to be the result of both the vulnerability of the vole and the weasel s active preference. The effect of different factors on the fate of voles (killed by weasel or survived) was examined with a generalised linear model (GLM) using the GENMOD procedure of the SAS Ò statistical software package, version 9.1. To examine the effect of the vole s activity on its vulnerability, we included as an explanatory variable the activity of the vole during the 24 h before weasel release. Since many of the voles died soon after the weasel release, the vole activity during a period of 24 h before the weasel release, in which all of the voles have approximately the same amount of fixes, was used as an explanatory variable. In the analysis, the enclosures were nested within habitat types and a logit link function with binary error distribution was assumed. Weasels hunting preference model The order in which the weasels killed voles was used to examine whether the weasels showed a preference to either prey species or to either sex within a species. As we did not replace the killed prey with new ones, the relative numbers of different prey types available for the weasels did not stay in balance during the experiment. For this reason, the use of preference indices developed earlier (Krebs 1999) would be problematic, and we develop here a new modelling framework that can be used for unbalanced data. We describe the modelling approach for the case of two prey types (e.g., species or sex, denoted by A and B), but it applies also for cases in which there are more types of prey. We assume that the predator moves around searching for prey. If the predator encounters an individual of type A, we assume that it has the probability q A of detecting and catching it, whereas the corresponding probability for type B is denoted by q B. Let us assume that there are currently n A individuals of type A and n B individuals of type B, and denote by x A = n A /(n A + n B ) the relative abundance of type A. Then, the probability p A that the predator will catch next an individual of type A follows the equation: p A ¼ x A ðq A þ ð1 q A Þp A Þþð1 x A Þð1 q B Þp A ð1þ The first term on the right-hand side represents the probability x A that the predator first encounters an individual of type A and either catches it right away (with probability q A ) or does not catch it on that occasion (with probability 1 - q A ), but still catches an individual of type A next (p A ). The second term represents the possibility that the predator encounters an individual of type B next (1 - x A ), but does not catch it on that occasion (1 - q B ), and catches an individual of type A next (p A ). Solving p A from Eq. 1 gives: x A z A p A ¼ ð2þ 1 x A þ x A z A where z A = q A /q B measures the relative preference for prey A. We transform z A into a logarithmic scale by defining the preference that the predator shows for type A over type B as w A = log 10 z A, so that w A = 0 corresponds to no preference, w A [ 0 to preference for type A and w A \ 0 to preference for type B (note that w A + w B = 0). Figure 1 illustrates how the probability p A depends both on the preference w A and on the relative numbers of types A and B. We used a Bayesian approach to relate the model given by Eq. 1 to the data assuming an uninformative (flat) prior for w A [ [-2, 2] (which covers already biologically unrealistically strong preferences) and zero prior probability for w [ 2. Let us denote by y ij the type of prey captured as the jth prey in the replicate i, so that y ij = 1 if the prey is of type A and y ij = 0 if the prey is of type B. The likelihood of the data is then given by: Pr ðyjp A Þ ¼ Yn Y k i i¼1 j¼1 p ij A yij A þ 1 pij A 1 y ij A ð3þ where n is the number of replicates, k i is the total number ij of prey caught in replicate i and p A is the probability of catching next a prey of type A (Eq. 1) in the situation where j - 1 prey were already caught in replicate i. To

Popul Ecol (2008) 50:257 266 261 Statistical power Pr[Pr(w > 0) > 0.95] Catching probability (p A ) 1.2 1.0 0.8 0.6 0.4 0.2 w = 0.1 w = 0.3 w = 0.5 0.0 0 2 4 6 8 10 12 14 16 Number of replicates 1.0 0.8 0.6 0.4 0.2 w = 1.0 w = 0.5 w = 0 w = -0.5 w = -1.0 0.0 0.0 0.2 0.4 0.6 0.8 1.0 Relative abundance of prey type A (x A ) Fig. 1 Upper panel The statistical power of the estimation method, measured as the probability that a data set contains high statistical support (posterior probability greater than 0.95) for the existence of a positive preference (w [ 0) at different levels of true values of the preference parameter w. Lower panel The model-based probability (p A ) that the predator will next catch an individual of type A (Eq. 1) with different preference levels (w) construct the posterior distribution for w A, we used a discrete grid approximation with grid size 1/50. We first tested the performance of the estimation scheme by performing a power analysis, in which we generated 100 replicate data sets for each of the values of n = 5, 10 and 15, with each experiment initially containing four individuals of both types A and B. We assumed that the preference of the predator was either mild (w A = 0.1), intermediate (w A = 0.3) or strong (w A = 0.5), and we let it catch four prey with the probabilities given by Eq. 1. We estimated the posterior distribution for the parameter w for each of the created data sets and calculated the posterior probability Pr(w A [ 0), by which the data suggests preference for prey type A. We then calculated the fraction of the 100 replicates for which Pr(w A [ 0) [ 0.95, in order to measure how often the results from a given experiment would give a strong signal for the preference for species A. We estimated the posterior distribution of the preference parameter w A for the real data set in two ways. First, we let the two species represent the types A and B, and estimated w A separately for the forest enclosures and for the open enclosures to see if the predator prefers either of the species and if the preference depends on the habitat type. Second, we examined whether there was a preference for either sex by treating the data for each species separately, but pooling the data for the two habitat types. In the real data set, there was an additional complication that some of the prey were recorded dead also for reasons other than predation by the weasels. We excluded such prey from the likelihood (Eq. 3), but accounted for the change that they caused to the relative numbers of the two prey types. One of the replicates (open habitat) was omitted from the analysis of preference, as, in this case, we were not able to determine reliably the killing order of the voles and as one of the male grey-sided voles was missing. Results Preference of weasels In both open and forest enclosures, the median estimates for w suggest that the weasels prefer the grey-sided voles over bank voles (Table 1 for the raw data and Table 2 for the results of analysis), the posterior probability for the pooled data being Pr(w [ 0) = 0.92. The estimated preference was stronger in open areas (w = 0.30) than in forests (w = 0.14). Comparing the two sexes, the weasels clearly prefer male grey-sided voles over female grey-sided voles (w =-0.39; Pr(w \ 0) = 0.97), whereas in the bank voles, there is no difference between the sexes (Table 2). As expected, the results of the power analysis (Fig. 1) showed that the possibility of obtaining a strong support for the existence of a preference increases with increasing number of replicates and with increasing strength of the preference. Assuming an intermediate strength for the preference (w A = 0.30), which was suggested by the model comparing the two species in the open enclosures, the probability of obtaining a result with Pr(w [ 0) [ 0.95 Table 1 Number of voles killed by weasels in the two enclosure types (forest or open) Grey-sided vole Bank vole Total Female Male Female Male Forest 3 9 5 5 22 Open 3 (5) 6 3 (4) 3 (4) 15 (19) Total 6 (8) 15 8 (9) 8 (9) 37 (41) The numbers in parentheses include also individuals for which the killing order was not possible to determine and which were, therefore, left out from the analysis of preference (see Materials and methods section). The total number of voles in the experiment was 96 (=12 9 8)

262 Popul Ecol (2008) 50:257 266 Table 2 Preference (w) of weasels for the grey-sided vole against the bank vole in different enclosure types (forest or open) and for females against males in different species 400 Forest w (95% HPDI) Pr(w [ 0) Preference (w) of weasels for the grey-sided vole against the bank vole Habitat type Forest 0.14 (-0.17 0.53) 0.78 Open 0.30 (-0.08 0.79) 0.90 Pooled data 0.20 (-0.03 0.50) 0.92 Preference (w) of weasels for females against males Prey species Grey-sided vole -0.39 (-0.76 0.02) 0.03 Bank vole 0.02 (-0.36 0.47) 0.53 The estimates shown for w are the median of the posterior distribution and the 95% highest posterior density interval (HPDI) is around half for the size of our data set (5 replicates in open areas). Performing 15 replicates would increase the probability to only three quarters, which illustrates that it is laborious to demonstrate the existence of a relatively mild preference. If the preference level would be stronger (w A = 0.5), then a set of five replicates would already produce a result with Pr(w [ 0) [ 0.95 with an 80% chance. As the preference of the weasels might depend on the size of the voles, we compared the weight of the voles that were killed by the weasels and those that survived, but no statistically significant difference was found (t-test: t = 1.60, df = 85, P = 0.114). Effect of weasels on space use and activity In the analysis of variance including all of the observations, the sizes of home ranges did not differ between species (F 1,80 = 0.01, P = 0.904), sexes (F 1,80 = 0.62, P = 0.434) or habitats (F 1,86 = 0.08, P = 0.777). All interactions were statistically non-significant and were excluded from the final model. Furthermore, when comparing home range sizes one day before and one day after the release of the weasels, no statistically significant differences were found (Fig. 2; Table 3). However, there was a tendency for larger home ranges in grey-sided voles (time 9 species, F 1,81 = 2.57, P = 0.113) after weasel release, especially in the forested enclosures. No statistically significant effects were found in the proportion of forest within the home ranges in forest enclosures (Table 3). In voles activity, measured as distances moved between consecutive fixes, no differences were found between species (ANOVA, F 1,74 = 0.88, P = 0.351), sexes (F 1,74 = 2.55, P = 0.114) or habitats (F 1,74 = 2.57, P = 0.114) when data from the entire experimental period were Home range MCP (m 2 ) 300 200 100 0 400 300 200 100 0 Open Grey-sided vole Bank vole Species Before After Fig. 2 Home range sizes of voles measured as minimum convex polygons (MCP) one day before weasel release and one day after the weasel release. Voles were located every second hour during a 24-h period. The upper panel shows the mean (+SE) home ranges in forest enclosures and the lower panel in open enclosures included. All interactions were statistically non-significant and were excluded from the final model. When the mean interfix distance moved during one day before and after the weasel release was examined, statistically significant differences were found. Generally, voles moved more after the weasel release (time, F 1,84 = 4.52, P = 0.036), but there was a difference between species (time 9 species, F 1,84 = 6.66, P = 0.012; Table 3) so that grey-sided voles moved longer distances when weasels were present (before 6.8 ± 0.7 m and during weasel presence 9.3 ± 0.8 m), while bank voles tended to move less (before 7.4 ± 0.7 m and during 7.1 ± 0.6 m). The fate of the voles (survived or killed by the weasels) was explained by their sex (GLM, v 2 = 7.07, df = 1, P = 0.008) and by their activity (sex 9 activity, v 2 = 5.48, df = 1, P = 0.019; Table 4); generally, more males were killed, but in females, more active voles were killed by weasels, while in males, more active ones survived (Fig. 3). As females moving long distances suffered higher mortality than females moving short distances, and as the pattern was the opposite in males, pooling the data

Popul Ecol (2008) 50:257 266 263 Table 3 Repeated measures ANOVA table on log-transformed home range sizes, arcsin-transformed proportions of forest in forest enclosures and interfix distances one day before and after the weasel release (time) Dependent variable Source of variation ndf ddf F P Home range Time (T) 1 81 1.64 0.204 Sex 1 80 1.25 0.268 Species (S) 1 80 0.20 0.656 Habitat (H) 1 80 0.91 0.342 Enclosure (habitat) 10 80 0.58 0.828 T 9 S 1 81 2.57 0.113 Proportion of forest in forest enclosures Time (T) 1 39 2.46 0.125 Sex 1 40 0.11 0.747 Species (S) 1 40 0.76 0.389 Enclosure 1 40 0.69 0.637 Interfix distance Time (T) 1 84 4.52 0.036 Sex 1 79 1.98 0.163 Species (S) 1 79 0.75 0.389 Habitat (H) 1 79 2.32 0.132 Enclosure (habitat) 10 79 2.44 0.014 T 9 S 1 84 6.66 0.012 Other factors are species (grey-sided vole or bank vole), sex, habitat (forest or open enclosure) and enclosure. In two of the models, enclosures were nested within habitat types to control the variation among enclosures (not in proportions of forest in forest enclosure). Non-significant interactions were not included in the final model, except in the model of home range, in which the interaction term closest to the 0.05 level of significance was included Table 4 Results of the generalised linear model (GLM) for the effects of activity (interfix distance before weasel release), sex, species (grey-sided vole or bank vole) and habitat (forest or open habitat enclosure) on the fates of the voles (survived or killed by the weasels) Source of variation df v 2 P Activity (A) 1 0.37 0.542 Sex 1 7.07 0.008 Species (S) 1 1.26 0.261 Habitat (H) 1 0.97 0.325 Enclosure (habitat) 10 5.48 0.857 A 9 sex 1 5.48 0.019 In the model, the enclosures were nested within habitat types to control variation among enclosures. Only the significant interactions are included in the model led to a non-linear relationship between the mortality and activity. The highest proportions of voles killed were found in the highest and the lowest distance categories, although voles that moved intermediate distances suffered substantial mortality by the weasels (Fig. 4). Voles activity increased considerably soon after the release of the weasels, but, later on, the voles tended to concentrate their movements to times of low weasel activity (linear regression R 2 = 0.166, F 1,35 = 6.76, P = 0.014; Fig. 5). Observations of voles in trees were rare. Before the weasel release, only one bank vole female was found in a tree twice. The same individual was observed in a tree five times after the release of the weasels. After the weasel release, altogether, three (3/24) grey-sided voles and five (5/24) bank voles were found in trees, with most of these on several occasions. Discussion Preference of weasels Largely based on different strategies and dissimilar r/a ratios (Holt 1977), it has been suggested that the grey-sided vole, as a representative of Microtus-type voles, would be more vulnerable to weasel predation than a typical Myodes vole the bank vole (Hanski and Henttonen 1996; Henttonen 2000). In line with this expectation, we found a slight preference of weasels for grey-sided voles over bank voles. We assumed that, in forested habitat, the bank vole, a good climber, would have gained from the availability of trees more than the grey-sided vole. In these species, the arboreal escape tactic has been observed previously only in bank voles (Jedrzejewska and Jedrzejewski 1990; Jedrzejewski et al. 1993; Sundell and Ylönen 2004). However, contrary to expectations, the preference for the grey-sided vole was stronger in open habitats than in forests. Our observations on climbing suggest that grey-sided voles also use arboreal escape, which may explain the unexpected result of only weak preference for grey-sided voles and a smaller difference in forested enclosures.

264 Popul Ecol (2008) 50:257 266 Interfix distance (m) 12 10 8 6 4 2 0 14 8 Female Grey-sided vole 7 11 Weasel Alive Male Male Female A factor not included in the present experiment is variation in body size. As we decided to control for size, partly to minimise the effect of the radio-collar s proportional weight (Kenward 1987), the natural weight difference between the species was reduced. In this experiment, within the size range variation present in the experimental animals, we did not find a statistically significant difference in weights between the killed voles and those that survived. Similarly, in a previous experiment (Sundell et al. 2003), the weight difference between experimental bank voles and field voles (Microtus agrestis L.) did not turn out to play any role, although it should be noted that, in that experiment, there was no possibility for arboreal escape. To obtain more conclusive evidence on the potential role of 8 13 8 Bank vole Fig. 3 Mean activity of voles (±SE) measured as interfix distance (distance moved between consecutive locations). Voles were located every second hour during the day before the release of weasels. The white bars correspond to voles which survived the entire experiment, while the black bars indicate the voles which were killed by the weasels. The numbers represent the sample sizes Proportion killed 1.0 Grey-sided vole Bank vole 0.5 0.0 0-2 2-4 4-6 6-8 8-10 10-12 12-14 14 < Interfix distance (m) Fig. 4 Proportions of grey-sided voles (black bars) and bank voles (white bars) killed by weasels as a function of the voles movement distances. Interfix distances (distance moved between consecutive locations) were averaged over the entire experiment 12 body size and individual behaviour, prey individuals of a wider range of sizes should be used. However, size is usually age-dependent, which further complicates a study focussing on the size effect only. We found that weasels preferred male grey-sided voles over females, while such a preference was not found in bank voles. This is interesting, as in previous studies, female voles have been found to be more vulnerable than male voles for mammalian predation in general, and especially for small mustelid predation (Klemola et al. 1997; Norrdahl and Korpimäki 1998; Sundell 2003), or no preference has been found (Ylönen et al. 2003). However, it should be noted that none of these studies have used grey-sided voles as a prey species. Instead, in previous studies, observations have been obtained in nature and on Microtus voles. In the field, Microtus voles are forming dense matrilinear kin-clusters during the breeding season (Pusenius et al. 1997), which are likely to attract predators and may, therefore, explain the preference for Microtus females in nature (Norrdahl and Korpimäki 1993). If the common view that female voles in breeding condition should be more prone to weasel predation is true (e.g., Cushing 1985), then our experiment may better describe the situation outside the breeding season. It might well be that the marking behaviour characteristic for Myodes males (e.g., Horne and Ylönen 1998), together with a greater mobility (Norrdahl and Korpimäki 1998), may expose them to weasel predation more than females, but it does not explain why we did not find such a preference for male bank voles. Effect of weasels on space use and activity Grey-sided voles moved longer distances between consecutive radiolocations after the weasels were released into the enclosure, but the same trend was not found in bank voles. Those female grey-sided voles that moved more were killed by the weasels, while in males, more mobile or active individuals survived better than less active individuals. The same tendency was evident also in bank voles. Another result of our study was that the activity rhythm of voles depended on the activity rhythm of weasels; voles were active when weasels were inactive and vice versa. Soon after the weasel release, voles became active and started to move longer distances. This observation might reflect an attempt to move away from the weasels. Our visual observations support this, as we saw, on several occasions, voles running along the fences after the weasel release. Unfortunately, we did not have controls for these observations, i.e., voles that would have stayed for the same period in the enclosures as in our weasel treatment but without weasels. However, the activity peak right after weasel release was the highest one we observed and the

Popul Ecol (2008) 50:257 266 265 Fig. 5 The mean activity of all voles and weasels measured as interfix distances (distance moved between consecutive locations). The animals were located every second hour during the entire experiment. The vertical line shows the time of the weasel release. The mean activity of the voles (±SE) is shown as the black line with dots, while the grey bars represent the mean activity of the weasels. The error bars for the weasel observations are omitted for the sake of clarity Interfix distance (m) 18 Weasels Before weasel After weasel Voles 16 14 12 10 8 6 4 2 0 20 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 20 0 4 8 12 16 Time (h) later peaks that coincided with weasels inactivity were not during the same time of the day in different days, which implies that they were likely to be direct reactions to weasel activity. The observation that activity patterns may change over time due to temporal variation in predation risk has also been made in earlier studies. For example, prey activity may first decrease, but if predation risk is present for a long period, the prey has to become active at some point to be able to, e.g., seek food (Lima and Bednekoff 1999). Previous studies on the activity of prey under predator presence have produced conflicting observations. Commonly, prey reduces activity, but, sometimes, an increase in activity has also been observed (e.g., Wooster and Sih 1995). Some of the differences can be caused by different measures used for activity, potentially reflecting very different types of activities (e.g., mobility, foraging, emigration). In general, reduced mobility should lead to lower encounter rates with predators and, consequently, to higher survival rates (Lima and Dill 1990). Norrdahl and Korpimäki (1998) found that voles moving long distances were killed more often than voles that moved less. This seemed to also be the case in females, but not in males, in the present experiment. Our observations of voles moving along the walls of the enclosures may indicate that they attempted to disperse longer distances than was possible in the closed enclosures. The U-shaped survival curve as a function of vole activity observed by Banks et al. (2000) was also apparent in our experiment if the two sexes were combined. When the sexes were investigated separately, the patterns of survival were mirror images. Banks et al. (2000) explained the phenomenon of nonlinearity in the predation risk so that reduced prey mobility leads to the accumulation of odorous waste products, which are used as cues by the predators. In the other extreme, very mobile individuals have an increased risk of random predator encounter. In our study, the same strategies to avoid weasel predation lead to a different success rates in female and male voles. In general, male voles have been observed to move more than females (Norrdahl and Korpimäki 1998). This is often explained by the mating system of voles, as male voles try to include as many female home ranges as possible within their own home ranges to be able to increase the fitness in terms of matings (Klemme et al. 2006). This implies that males are better adapted to longdistance and females to short-distance movements. In the light of this assumption, our results make sense, showing that females, adapted to move less, were vulnerable when they move more, while the opposite is evident for males. Conclusions We observed only weak evidence for the higher vulnerability of Microtus-type voles, here, the grey-sided vole, to small mustelid predation, and the evidence was unexpectedly stronger in open habitats than in forested habitats. Combined with observations of grey-sided voles in trees, this indicates that arboreal escape tactics are also used in this clumsy field-layer species. Thus, our results suggest that grey-sided voles may not be a true representative of Microtus-type voles, but, rather, an intermediate type between the two types. However, under natural conditions, grey-sided voles may have a higher vulnerability compared to bank voles due to their tendency to form relatively high-

266 Popul Ecol (2008) 50:257 266 density aggregations attracting predators. What part the generally larger size of the grey-sided vole compared to the bank vole plays in species-specific vulnerability remains to be tested with a wider size range of different sized individuals. Acknowledgments J. Gustafsson, A. Hämäläinen, J. Häyrinen, K. Juutilainen, E. Knuutila, S. Niskanen, A. Reckordt and H. Vihervaara are acknowledged for their assistance during the field work. We thank J.O. Wolff for the valuable comments on the manuscript and A.-L. Laine for helping with the statistics analysis. The study was funded by the Academy of Finland (project nos. 208478 to JS, 44878 to HY and MH, and 205376 to OO), Swedish Research Council to TO and CIMO to LT. The experiment complies with the current laws of Finland. References Abrams PA, Holt RD, Roth JD (1998) Apparent competition or apparent mutualism? Shared predation when populations cycle. Ecology 79:201 212 Banks PB, Norrdahl K, Korpimäki E (2000) Nonlinearity in the predation risk of prey mobility. Proc R Soc Lond B 267:1621 1625 Chase JM, Abrams PA, Grover JP, Diehl S, Chesson P, Holt RD, Richards SA, Nisbet RM, Case TJ (2002) The interaction between predation and competition: a review and synthesis. Ecol Lett 5:302 315 Christensen P, Hörnfeldt B (2006) Habitat preferences of Clethrionomys rufocanus in boreal Sweden. Landscape Ecol 21:185 194 Cushing BS (1985) Estrous mice and vulnerability to weasel predation. Ecology 66:1976 1978 Hanski I, Henttonen H (1996) Predation on competing rodent species: a simple explanation of complex patterns. J Anim Ecol 65:220 232 Hanski I, Hansson L, Henttonen H (1991) Specialist predators, generalist predators, and the microtine rodent cycle. J Anim Ecol 60:353 367 Henttonen H (1987) The impact of spacing behavior in microtine rodents on the dynamics of least weasels Mustela nivalis a hypothesis. Oikos 50:366 370 Henttonen H (2000) Long-term dynamics of the bank vole Clethrionomys glareolus at Pallasjärvi, northern Finnish taiga. Polish J Ecol 48(Suppl):87 96 Henttonen H, Hansson L (1984) Interspecific relations between small rodents in European boreal and subarctic environments. Acta Zool Fenn 172:61 65 Holt RD (1977) Predation, apparent competition, and the structure of prey communities. Theor Popul Biol 12:197 229 Holt RD, Kotler BP (1987) Short-term apparent competition. Am Nat 130:412 430 Horne TJ, Ylönen H (1998) Heritabilities of dominance-related traits in male bank voles (Clethrionomys glareolus). Evolution 52:894 899 Jedrzejewska B, Jedrzejewski W (1990) Antipredatory behaviour of bank voles and prey choice of weasels enclosure experiments. Ann Zool Fenn 27:321 328 Jedrzejewski W, Jedrzejewska B (1990) Effect of a predator s visit on the spatial distribution of bank voles: experiments with weasels. Can J Zool 68:660 666 Jedrzejewski W, Rychlik L, Jedrzejewska B (1993) Responses of bank voles to odours of seven species of predators: experimental data and their relevance to natural predator vole relationships. Oikos 68:251 257 Jensen B (1994) Suomen ja Pohjolan nisäkkäät. WSOY, Porvoo (in Finnish) Kenward R (1987) Wildlife radio tagging: equipment, field techniques and data analysis. Academic Press, London Klemme I, Eccard JA, Gerlach G, Horne TJ, Ylönen H (2006) Does it pay to be a dominant male in a promiscuous species? Ann Zool Fenn 43:248 257 Klemola T, Koivula M, Korpimäki E, Norrdahl K (1997) Small mustelid predation slows population growth of Microtus voles: a predator reduction experiment. J Anim Ecol 66:607 614 Korpimäki E, Norrdahl K, Rinta-Jaskari T (1991) Responses of stoats and least weasels to fluctuating food abundances: is the low phase of the vole cycle due to mustelid predation? Oecologia 88:552 561 Krebs CJ (1999) Ecological methodology, 2nd edn. Benjamin Cummings, Menlo Park, California Lima SL, Bednekoff PA (1999) Temporal variation in danger drives antipredator behavior: the predation risk allocation hypothesis. Am Nat 153:649 659 Lima SL, Dill LM (1990) Behavioral decisions made under the risk of predation: a review and prospectus. Can J Zool 68:619 640 Menge BA, Sutherland JP (1976) Species diversity gradients: synthesis of the roles of predation, competition, and temporal heterogeneity. Am Nat 110:351 369 Menge BA, Sutherland JP (1987) Community regulation: variation in disturbance, competition, and predation in relation to environmental stress and recruitment. Am Nat 130:730 757 Norrdahl K, Korpimäki E (1993) Predation and interspecific competition in two Microtus voles. Oikos 67:149 158 Norrdahl K, Korpimäki E (1998) Does mobility or sex of voles affect risk of predation by mammalian predators? Ecology 79:226 232 Oksanen L (1992) Evolution of exploitation ecosystems I. Predation, foraging ecology and population dynamics in herbivores. Evol Ecol 6:15 33 Oksanen L, Lundberg P (1995) Optimization of reproductive effort and foraging time in mammals: the influence of resource level and predation risk. Evol Ecol 9:45 56 Pusenius J, Viitala J, Marienberg T, Ritvanen S (1998) Matrilineal kin clusters and their effect on reproductive success in the field vole Microtus agrestis. Behav Ecol 9:85 92 Sih A, Crowley P, McPeek M, Petranka J, Strohmeier K (1985) Predation, competition, and prey communities: a review of field experiments. Annu Rev Ecol Syst 16:269 311 Siivonen L, Sulkava S (1994) Mammals of northern Europe, 5th edn. Otava, Helsinki (in Finnish) Sundell J (2003) Population dynamics of microtine rodents: an experimental test of the predation hypothesis. Oikos 101:416 427 Sundell J, Ylönen H (2004) Behaviour and choice of refuge by voles under predation risk. Behav Ecol Sociobiol 56:263 269 Sundell J, Norrdahl K, Korpimäki E, Hanski I (2000) Functional response of the least weasel, Mustela nivalis nivalis. Oikos 90:501 508 Sundell J, Eccard JA, Tiilikainen R, Ylönen H (2003) Predation rate, prey preference and predator switching: experiments on voles and weasels. Oikos 101:615 623 Viitala J (1977) Social organization in cyclic subarctic populations of the vole Clethrionomys rufocanus (Sund.) and Microtus agrestis (L.). Ann Zool Fenn 14:53 93 Wilson DE, Reeder DM (2005) Mammal species of the world, 3rd edn. Johns Hopkins University Press, Baltimore, Maryland Wooster D, Sih A (1995) A review of the drift and activity responses of stream prey to predator presence. Oikos 73:3 8 Ylönen H, Sundell J, Tiilikainen R, Eccard JA, Horne TJ (2003) Weasels (Mustela nivalis nivalis) preference for olfactory cues of the vole (Clethrionomys glareolus). Ecology 84:1447 1452