\4Je observed the hunting behavior of the owl until ir killed ... killed one of the voles. ..... facilitic! for ari.uv expcriments, and Xai Norrdahl, Dalid (nrrie, Burr.
BehavioralEcologr \ol I No.3:261-2t6
Refuge sites of voles under owl predation risk: priority of dominant individuals? Vesa l(oivunen, Erkki I{orpinnki, and Harri Hakkarainen Laboratory of Ecology and Animal Systematics,Department of Biology, University of Turku, FIN-20014Turku, Finland
In an aviary experiment, we studied whether body size or habitat familiarity of field voles (Miootus agrestis\affected predation risk by Tengmalm's owls (Aegoliusfunereus). In the field, we compared the bodv size offield voles snaptrapped in good (covered) and poor (open) habitats in 1992 and 1994 to deterrnine whcther therc were habitat-related differences in the body size of voles. In the aliary, large individuals occupied the good habitat signilicantly more than small indiriduals boti in the control (owl not present) and experimental treaffnents (owl present). Furthermore, habitat-familiar voles inhabited tie good habitat more than habitat-unfamiliar voles did when an owl was present. Our field data were consistent with our aviary data: larger field voles were more frequently found in good habitats than in poor habitars. In the aviary, Tengmalm's owl predation risk was higher for small and hatritat-unfamiliar voles. This suggests that large field voles may have priority to sheltered habitats. Furthermore, habitat familiaritv mav play a central role in avoiding risky habitats. Kcy uords: dominance relationships, field voles, habitat familiariry, predation risk, prey choice, rodents, Tengmalm's owls. [Behav Ecol 9:261-266 (1998)]
Jr is generallv expected that predarors selectively consurne I prey items based on optimum energetic inuke in rclarion to the costs of searching for and handling food (Stepbens and X.rebs, 1987)- This means that predators should prefer large individuals as prey items because such a prey could be assumed to be energetically the most prolitable. It has been shown, however, that predators may diverge from the predictions of optimalilv models. For example, many studies on avian predators fceding mainly on small mammals, such as voles, inice, and shrews, have shown that small prey individuals are the most vulncrable prey category (e.g., Halle, 1988; Koiunen et al., 199fia,b; Mappes et al., 1993; Marti and Hogue, 1979). A mechanistic approach to prey selection proposes that prey choice is influenced by the behavior ofboth the predator and the prey-not only by predators as optimaliry models assume (Sih, 1993; Sih and Moore, 1990). Therefore, the refuges, cr,lpticity, and moring activity of the prey may also influence the encounter rate between the prey and predator S€lection of some prev items is simply a result of varfng cap ture success,rrther than a result ofactive decision of the predator to attack or not ro attack (Sih. 1993: Sih and Moore, 1990). Among small rodent-eating avian predators, selective predation is usually assumed to be a result of hunring habirar selection, or of behavioral differences between the subgroups in a prey population (see Dickman, 1992; Dickman et al,, l99l). Little attention. however. has been focused on identi- . fiing exact mechanisms causing the lulnerability of small prey indiriduals. This may be because experiments concerning beharioral differences in small mammal populations are laboriou5. erpecialll rn rhr field. $,here ir is diaficulr ttr conrrol for all the behalioral and ecological factors that may act in the prey population (Dickman, 1992; Dickman et al., 1991). However, much information can be obtained using aviary experiments, which pro\ide opportunities to examine the behavior
of small mammals under avian oredation and without the risk of avian predation. In Fennoscandia, small mammals, such as the field vole (Microtus agrestisl, the bank vole (Clzthionomls glareolus), and the common shrew (Sore.raraneus) are the main prey items of many arian predators (Erlinge et al., 1983; Ko unen et al., 1996a,b). Field voles mainlv occupy open, grassyhabitats (Myl\'rnaki, 1977), and their densities fluctuate lG to l0Gfold during 1 to ,fyear population cycles in western Finland (e.g., . Korpimiki and Norrdahl. 1991). These small mammals are the itaple prey of Tengmalm's owl (Aegtlius funtrats\ , which is a small, nocturnal, predatory bird in coniferous forests of northern Europe (Korpimeki, 1981, 1988). The main purpose of this sturlr uas to cxaminc. in an aviarv. whether body size or habitat familiarity affect the lulnerability of field voles to predation by Tengmalm's owls. On the basis of our earlier lield studies (Koiwnen et al., 1996a,b), we predicted that small field voles could be more !'ulnerable to oul predation than large ones, and that this difference could be due to behal'ioral dissimilarities. Because an earlier studv on meadow 'iofes I MicrotusI1?nn\tuanifut has shorn thar habitat-familiar indMduals may have a distinct advantage in es caping predators (Ambrose, 1972), we also predicted that habitat-unfamiliar field voles could be more vulnerable to owl oredation than habitat-familiarvoles.In addition. we oresentfield dau rtmparing the bod) size of field voles snaptrapped in good (sheltered) and poor (open) habitats. As large body size has been shown to reflect intmspecific dominance among small mammal populations (e.g., Grant, 1972), we predicted that large (i.e., dominzurt) field voles occupy good habitas with low owl predation risk more often than small (i.e., sut! dominant) voles. MATERIAIS
AND METIIODS
Aviary experiments Address correspondence to E. KorpimakiReceived 20 January 1997; first revisnn 26 N{ar 1997: second revition 2 October 1997; accepted I October 199?. :-il 1998 Inrernational Societl ior Bcharioral Ecolog.r
The prey-choice experiment was carried out in August 1993 in an a\iar,v (3.0X4.8 m, height 2.2 m) at the Satakunta Environmental Rcsearch Center of Turku University, in Pori, western Finland (61'30' N, 2l'30' E). The field voles (n : 68) used in the experiment were caught in July and August
262
1993 in the vicinity of Pori and Turku (60"30' N, 22'30' E) using Swedish Ugglan live traps. Between trials, we kept individual voles in separate cages (l x1 m, height 0.5 m), where they had unlirnited accessto food and water. Tengmalm's owls [11 adults (age >l year) and 6 yearlings (age (l year)] were captured during their autumn migmtion in Koklola (64" N, 23" E), on the coast of the Baltic Sea by mist-nets and playback technique (see Korpimiki and Hongell, 1986). Owls were trapped and held in captivity with the permission of Finnish Ministry of the Environment. After the experiment, tley were released in the same area where they were d-apped. In an aviary, there were two identical uncovered pens (2.3X1.6 m, height 0.5 m) located qrnmetrically around a perch (height 1.4 m) for the hunting owl. Both pens were partitioned into two halves. which were connected to each other by two plastic tubes (diam 5 cm, length 10 cm) so that a vole was free to move between the two halves of the Den. An agricuhural en!irunmenr was simulated by placing a-3-4 cm deep layer of ha,vin one half of each pen (a good habitat, sheltered), and a 0-5-l cm deep layer of hay in the other half of both pens (a poor habitat, open). Our earlier study showed that 3-.1 cm and 0.5-1 cm deep layers of hay can be used to simulate good and poor habitats of lield voles (Korpimiiki et al., 1996). The two halves of each pen also contained a feeding place for voles (pellets for labor-atory mice and water). ln the aviary, the light level was the same as it would have been in nature during the normal hunting time of Tengmalm's owls in summer (note that nights in midsummer are light in the main range of Tengmalm's owls; see Korpimeki, 1981; Mikkola, 1983). We recorded the behavior of voles from a hide located on the wall 3.5 m above the bottom of the aviary. In this study, we were interested to know whether body size and habitat familiaritl aftected the !'ulnerabiliq of neld voles to Tengmalm's owl predation. Therefore we used large (i.e., dominant), small (i.e., subdominant), habitat-familiar and habitat-unfamiliar voles in the exDerimental trcatments. In the control Eeatments (r? : 17, Tengmalm's owl absent), we used three field voles of the same sex. A laree and small vole were introdu(t'd tu a randomly selected pen. and a habiralfamiliar lole to the other pen for l5 min (mean body weight of large voles,35.8 g, SD : 6.5; small voles,21.3 g, SD : 2.9). Thereaftet we recorded the behavior of voles for 15 min to confirm that they were familiar with the pen and that they did nol behave abnormalll {i.e.. \oles werr mu\rng, staying. eating, or grooming), and an additional 15 min to record behavior for the data analyses. During the firsr rwo 15-min periods, voles probably had enough time to explore both halves of the pen and to become acquainred with dre pen (Korpimeki et al., 1996). In Tengmalm's owl tre atments (n : 17\, a habitat-unfamiliar volc (same body size and sex as the habitat-familiar one) was released to the same pen with the habirar-familiar vole (small and large voles were in the other pen), and almost immediately thereafter, an owl was released into the aviary. \4Jeobserved the hunting behavior of the owl until ir killed one of the voles. Owls werc not fed during the 24 h before the trials to rnsurc that hunting behavior would begin ^frer an owl was released into the aviary. After a release, the owl usuall) flew around the aliary for l-2 min, scanned from the perch at a distance of 2-4 m from voles, and attacked and killed one of the voles. The perch height (1.4 m) was not unnatural because in the field Tengmalm's owls use low prrches tBye er al.. 1992: Koirtnen i. Hakkarainen H. unpublished radio-tracking data). We used each owl and surviving vole only once in the trial. We collected behavioral data fiom voles in two trial pens bv the lime sampling merhod {Alrmann, 197.1r:we recorded rhi behavior at I min interlals durins the l5-min trial. Vole be-
Behavioral Ecology VoI. 9 No. 3
havior *as classified as follows: ( 1) moving, when the vole was mobile in the pen, (2) staying, when the vole was immobile, (3) eating, when the vole was feeding on pellets and drinking water, and (.1) cleaning, when the vole was observed groomi.g. Snaptrapping
data
The snaptrapping data were gathered in the decrease (1992) and increase (199'l) vole years in the Kauhau region (63'N, 23o E), western Finland, in an area mainly consisting of pine and spruce foress (467o), but with high proportions of agricultural land (27Vo) and peatland bogs (237o) (Korpimnki, 1 9 8 1 .1 9 8 8 ) . To study the body weight of field voles in the field, snap trappings were made during lnight periods in l7 sampling plots (each I km:) in 1992 liom the end of March to the end of May, and in 10 sampling plots (each 1 km') in 1994 from early April to earlyJune. We conducted snaptrappings using the short-line method (see Koi.larnen et al., 1996b). The trap lines were set in each sampling plot in relation to the pr(} portion of four main habitat tlpes (coniferous forest, grassland, culti\ated farmland, and clear-cut area). For example, if the proportion of coniferous forest was 30%, that of grassland 30%, cultivated farmland 30%, and clear-cut;uea 107o, then three tmplines \tere erected in coniferous forests, three trap lines in grassland, three traplines in cultilated farmland and one traDline in clear-cut area. (;rassland has the densest and deepesigrass la,ver,and voles are less visible to alian predators there than are individuals in the other habitat t!'Des. There_against fore, grassv harfields mav ofler the best pror er ril,n avian predators. The ground cover ofthe coniferous forest mainly consisted of blueberries (Vaccinium mlrtiLlus), lingonbelies (V ritis-idaz), and mosses.Coniferous forest was classified as the second best habitat according to ground cover, followed by cultivated farmland (mainly spring crops, such as oat and barley, were cultivated) and cleaHut areas (the most open habitao. Each sampling plot included l0 lines consisting of 10 snaF traps at l0-m intervals. The traps were baited with mixed€rain bread (i.e., bread made of a mixture of rye and wheat flour). We checkcd snaptraps evert morning, and trapped animals were identified, sexed. and measured (body mass to the nearest 0.1 g and length to the nearest I mm) according to Siivonen (1974). To estimate the physiological condition of voles, the amount of internal fat was estimated visually from dissectedanimals:0: no fat, 1: little,2: intermediate, and 3 : much (Koilunen et al., 1996b). Statistical analyseswere performed using the SYSTAT statistical package (see Wilkinson, 1989), apart from logit analyses, which were performed using PROC C,ATMOD of SAS software (SAS Institute, 1990). All statistical tests were two-tailed. RXSUtiTS Aviary experiments Large field voles occupied the good habitat significantly rnore than small voles did in both the control (Tengmalm's owl al> sent; Figure la) and experimental (an owl present;Figure lb) $eatment (Mann Whitney Lrtest, U = 212.0, n = 17, p = .91 and U: 201.5, n:17, P : .04, respectively).Furthermore, habitat-familiar individuals inhabited the good habitar significantlv morc than unfamiliar voles did when an owl was presenr in rhr ^vj,aJytlJ - 199.5.n -17. p - .Ob)In addition to the differences in habitat utilization, large indir.rdualsmoved lesi than small one\ when owls were abscnt (Table l).'whereas there were no differences in movement
Koivunen et al. . Voles and owl predation
risk
a
26t
Table I The mean numb€r of obserr.ations of movemenrs p€r replicate Treatrnent and vole Tengmalm's owl absenr Large voles Small voles
E E
o Habitat-familiar
F
vol€s
Tengmalm s owl pres€nr Large voles Small voles
2
Large
Small
Familiar voles Unfamiliar voles
3
Familiar
Mean
SD
3.2 6.1 U = 85.0 P=.o3 2.6 3.8
n
t7 t7 17
1.6 1.9 U = 125.5 P=.5 0.5 2.6 2.4 U = 52.0 P= 001
Differences b€t$een lole status were tested by Mann,Whitney U
b significandy more than habitat-famitiar voles (Xz = 3.b6, df =
1'P = .oot.
Snaptrapping
ro F
1 Large
2 Small
3 4 Familiar Unfamiliar
Iigur€ f The mean (aSD) dme periodper reptiiareIn t7, 15 nL, ea,h) ot hetd toteso(cupvinggood (6ltedbar.r and poor (openb sl habitatsfor large,small,and habitar-familiarvoiesin (j) control rrearminr: ihl lhe \ame but tor Iarge,smdll.lrabirar-familiar, and habitat-unfaniliarvol€sin Tengmalin.sowl rreatment.Note rhar a largenDd. rdllrol. werein onc p.n, and a hdbirar-famitiar dnd habirat-unlamiliar \ule in thc orhir pen.
when owls were.present (Table l). We did not observe any eating and drinking in the trial pens, but rhis may be because voleshad unlimired a( cessto fo;d and waler In holding caqes betweentrials. Habirat-lamiliarindividualsmoved sisniFcan"rlv less than unfamiliar ones when owls were present if"lt" t).' Tengmalm's owls captured 2 of lT large voles, g of lZ small voles, I of 17 habitat-familiar voles, and6 of l? habitat-unfamiliar voles. To tesr the effects of vole sratus (large/small and habitat familiar/unfamiliar) on rhe caprure fiequency of Tengmalm's owl, logit analysis was used (see Collett;1990j. tn this analvsis.lhc eftecrs ot bodv size and deqree ul habital tamrlrarrrvon caplurc probabilin rvole killed ur n.}r Lilled) were tested in the same analvsis, but we were not interested in the interaction term because the voles of these two cate_ gories were in separate pens and could not intemct with each other. Small voles were more iulnerable to Tengmalm's owl prrdari.'n than large rt,lcs tlj - 4.48. dl - L p - .03). and habitat-unfamiliar voles were exposed to owl predation almost
data
A tl{ o-way ANOVA showed that the body weights of field voles snaptrapped in 1992 significantly differed among rhe four main habitat qpes, but there was no obvious sex-related differences in body weight (Table 2). In addition, no inreraction effect between habiur t,vpe and sex on body weights of field voles was found. A posreriori Tukey rests indicar;d tiat field voles in the grassland habitat were siqnificanrly heavier than those in coniferous loresr. cultirated iarmland, and clearcut areas (, = .009, .002, and .002, respectivcly; Figure 2a). In 1994, there r,rasno obr,ious habitar-related diflerence in the body weight of field voles (Ikuskal-Wallis test: H = b.4, n = 44, p : .14: Figure 2a). In a three-way ANOVA for rhe pooled data from 1992 and 1994, there was a significant diffeience in the body weishts of field voles between gmssland and culrivated farmlandi voles were heavier in rhe grassland than in rhe cultivated farmland (Table 3, Figure 2a). In addition, on rhese two habitat tlpes. field voles were hcavier in 1994 rhan in 1992 tTahle 3. Fi'gpre 2a). No intersexual differences were found, bur a thre;av interacrion bewren hahital, sex, and year was signrficanr1Tible 3). A twcway ANOVA revealed rhat the body lengths of field voles snaptrapped in 1992 significantly differed among the four main habitat tJpes, but no obvious inrersexual diffeience was found (Table 2). A posteriori Tukey tests indicated that field voles wcre longer in grasslandrian in forrsr, rultirated farmland, and clear