Intrinsic and Climatic Determinants of Population Demography: the Winter Dynamics of Tundra Voles

Ecology, 83(12), 2002, pp. 3449±3456 ᭧ 2002 by the Ecological Society of America

INTRINSIC AND CLIMATIC DETERMINANTS OF POPULATION DEMOGRAPHY: THE WINTER DYNAMICS OF TUNDRA VOLES

JON AARS1,3 AND ROLF A. IMS2 1Department of Zoology, University of Aberdeen, Tillydrone Avenue, Aberdeen AB24 2TZ, Scotland, UK 2Department of Ecology, Institute of Biology, University of Tromsù, N-9037 Tromsù, Norway

Abstract. The relative impacts of intrinsic factors (e.g., density dependence) and ex- trinsic factors (e.g., climate) on winter demography may be critical for the generation of different population dynamic patterns (including cyclicity) in northern vole and lemming populations. However, little is known about winter demography because studies with temporal and spatial replication at the population level and an adequate sample of individuals with known fates within each population are rare. In this study, we monitored the winter demography of 48 local tundra vole populations introduced to experimentally enclosed plots the preceding spring for four years in Norway. The rate of population change over the winter (November±May) was density dependent due to recruitment. However, the large variation in the rate of change between the different winters was due to a density-independent, and most likely a climatically driven, variation in survival rate. In particular, mild weather that led to the formation of ice on the ground seemed to be detrimental for winter survival. We predict that if increased frequency of such events arose, due to climate change, normal cyclic dynamics of northern small rodent populations would be disrupted. We found support for the hypothesis that voles adjusted their body mass toward a certain mean during the winter so as to maximize winter survival. The survival rate of males was lower than that of females, possibly due to their larger body mass, and this resulted in female-biased population sex ratios in the spring. This result suggests a link between sexual selection (responsible for the sexual size dimorphism) and natural selection (operating though size-dependent winter survival) with implications for the demographic structure of the population. Key words: body mass; density dependence; Microtus oeconomus; Norway; population cycles; sex ratio; survival rate; tundra vole; winter climate.

INTRODUCTION population ¯uctuations, often with clear evidence of The population dynamics of small mammals can be cyclic dynamics (Hansson and Henttonen 1988, Hanski simultaneously determined by density-dependent and et al. 1993, Stenseth and Ims 1993, Bjùrnstad et al. density-independent factors (Leirs et al. 1997, Karels 1995, Turchin and Hanski 1997). Winter has been sus- and Boonstra 2000). Climatic factors, which normally pected to play a crucial role in the generation of the underlie the density-independent component of popu- population dynamics pattern in northern regions (Hans- lation change, are expected to exert their in¯uence most son and Henttonen 1988, Stenseth 1999, Yoccoz and strongly during certain seasons (e.g., the summer dry Ims 1999). Although recent analyses of time series have period in the tropics or the winter cold period in the indicated that the rate of population change during the arctic). Density-dependent processes are more likely to winter is strongly density dependent (Stenseth et al. shape population dynamics year round. However, the 1998, Hansen et al. 1999), the demographic mecha- speci®c mechanism of the density dependence may dif- nisms underlying the rate of change have not yet been fer depending on the season (Ostfeld et al. 1993, Ost- elucidated. To unravel the demographic mechanisms, feld and Canham 1995, Hansen et al. 1999). Knowing capture±recapture data monitoring the fates of the in- the relative strength of density-dependent and density- dividuals over the winter will be required. Unfortu- independent processes during all seasons of the year is nately, demographic studies of northern rodents con- necessary to understand the great variety of multi-an- ducted on a multi-annual time scale are generally scarce nual population dynamics found in small mammals (Yoccoz et al. 1998). In particular, studies covering the (Turchin and Ostfeld 1997, Stenseth 1999). critical winter period are almost nonexistent and thus Small mammal populations in regions with long and badly needed (Stenseth 1999). Here we analyze winter demography of the tundra snowy winters have become famous for their violent vole (Microtus oeconomus) in grassland habitats based Manuscript received 14 November 2001; revised 18 April on data covering four consecutive winters at Evenstad 2002; accepted 22 April 2002. Research Station, in southeastern Norway. Evenstad 3 E-mail: [email protected] has a continental winter climate (mean temperature in 3449 3450 JON AARS AND ROLF A. IMS Ecology, Vol. 83, No. 12

January: Ϫ10.7ЊC), with snow normally covering the and boreal meadow and mire habitats in Fennoscandia ground from November to late April. Our analyses were (Tast 1966). The population densities in the enclosures facilitated by spatial replication of experimentally en- were within the range of what has been observed in closed populations and by a large number of individ- natural populations. Details regarding the maintenance ually marked animals at the onset of the winter, of of the plots and the procedure for establishment of the which a sample suf®cient to allow statistical analyses experimental populations can be found in Aars and Ims was recaptured in the spring. The experimental setting (1999, 2000), Aars et al. (1999), and Gundersen et al. was repeated over four winters so that the study in- (2001). cluded among-year variation in winter weather. The populations grew freely over the summer, and We focused on two aspects of winter demography. were monitored by 3-d live-trapping sessions (capture± First, at the population level, we quanti®ed the effects mark±recapture trapping) every 18 d throughout the of density-dependent and density-independent factors snow-free period. A high trap density in the habitat on the rate of population change during the winter, and patches and 6 trap checks during the 3 d in each trap- we estimated the contribution of survival and recruit- ping session ensured close to 100% trapping rate all ment to these effects. Second, based on the well-es- years (see Aars et al. 1999, Aars and Ims 2000, Gun- tablished fact that most northern small mammals de- dersen et al. 2001). Thus, we could ignore capture prob- press their body size from summer to winter (Iverson ability when estimating demographic rates. Live trap- and Turner 1974, Whitney 1976, Merritt and Merritt ping was terminated when permanent snow cover was 1978, Hansson 1990, 1991, 1992), we tested the hy- established in November. Trapping was resumed im- pothesis that mass loss over the winter is an adaptive mediately after the snow had melted in late April or adjustment to maximize winter survival (Stenseth early May. During this spring trapping session, all an- 1978, Hansson 1990). Changes in mass in small rodents imals were removed and the plots were left empty until have been shown to be induced by photoperiod (Dark new populations were established by the release of new 1983), thus indicating that it is an adaptive preparation laboratory-raised animals. The vole-proof fences ef®- for the winter (Iverson and Turner 1974, Malcolm and ciently prevented dispersal between enclosures also Brooks 1993). We also evaluated possible links be- over the winter, despite drifted snow in some short tween individual-level life history tactics and popu- periods that reached the fence tops. No marked animal lation-level dynamics based on the results of our anal- was ever recorded to have moved between enclosures. ysis. Analyses of the summer demography in the experimental populations are found in Aars et al. (1999), MATERIALS AND METHODS Aars and Ims (1999, 2000), and Gundersen et al. The data was obtained from experiments conducted (2001). These analyses focus on the effects of dispersal during the years 1994 to 1998 at Evenstad Research on demography and population genetics. Here, we use Station in southern Norway (61Њ12Ј N, 11Њ06Ј E). New data from the two trapping sessions at the onset and experimental populations of tundra voles were estab- termination of the winter, respectively, to highlight de- lished in spring/early summer every year by releasing mographic processes during winter. laboratory-raised tundra voles on six fence-enclosed The rate of population change from autumn (Nautumn) experimental plots. The animals in the laboratory were to spring (Nspring) at the patch level was analyzed by outbred as new animals from the ®eld were added to ®tting the following statistical model to the data (Le- the breeding stock every year. Each plot was 0.5 ha breton 1991): (50 ϫ 100 m) and contained two 750-m2 (20 ϫ 37.5 N ϭ N ϫ exp[␣ϩ␤ϫN ] ϩ␧ m) meadow patches (i.e., habitat) separated by 50 m spring,i autumn,ijjautumn,ii of barren ground that was treated with herbicide during where ␤ is the strength of density dependence (i.e., the the growing season. As effective dispersal between the slope), ␣ a constant, j is year (i.e., winter), and ␧ is an two patches was mostly restricted to the early summer error term speci®c to each population i. The model was season (Aars et al. 1999), each enclosure effectively ®tted to the data using a logarithmic link, i.e., log consisted of two populations both in the autumn and (E[Nspring]), with log(Nautumn) as an offset term, and the during the winter. The meadow patches in the experi- error term was quasi-Poisson distributed due to over- mental plots were fertilized every spring to standardize dispersion (residual deviance/df ϭ 3.32). The in¯uence habitat quality among the years. This gave rise to a of year was tested (Wald ␹2) with respect to the ␣- dense vegetation cover that formed a thick mat of wilt- (additive year effect) and ␤-parameters (year ϫ Nautumn ed grass and herbs during the winter period. We chose interaction effect). to burn this debris every spring as to avoid a gradual Recruitment was analyzed with Poisson regression, change in the substrate that would increase the envi- and survival with logistic regression. These regression ronmental variability between habitat patches in time models ®tted the data (i.e., no overdispersion), and the and space. Tundra voles are common in natural and signi®cance of the predictor variables could be eval- seminatural meadow patches in the surroundings of the uated with ordinary likelihood ratio tests and Akaike's experimental plots at Evenstad and elsewhere in alpine Information Criterion (AIC) values. December 2002 WINTER DEMOGRAPHY OF VOLES 3451

following spring (Nrecruits ϭ exp[1.65 Ϫ 0.03 ϫ Nautumn], likelihood ratio test [LRT] ␹2 ϭ 16.70, df ϭ 1, P Ͻ 0.0001). Adding year as an extra term in this model did not improve its ®t (year effect: LRT ␹2 ϭ 2.10, df ϭ 3, P ϭ 0.35). Winter survival rates differed markedly among years, and a logistic regression model applied to the patch-speci®c proportion of individuals present in No- vember that were recaptured six months later in the spring gave the following estimates of year-speci®c survival probabilities: 1994: 0.02 (95% CI ϭ 0.01± 0.05), 1995: 0.25 (95% CI ϭ 0.17±0.34), 1996: 0.15 FIG. 1. Change in population density (number of tundra voles per habitat patch) from autumn to spring during the (95% CI ϭ 0.09±0.24). There was no evidence for den- four years of this study at the Evenstad Research Station, sity-dependence in the survival rate (Wald ␹2 ϭ 0.52, Norway. df ϭ 1, P ϭ 0.48). The yearly estimates of winter survival rates correlated negatively with the mean winter temperature at Evenstad (Pearson product-moment RESULTS correlation r ϭ 0.97, n ϭ 4, P ϭ 0.034). The mildest winters with the lowest survival rates had many days Population level parameters with temperatures Ͼ0ЊC during the midwinter period The habitat patches constitute the natural scale for (Fig. 3). This corresponds well with our observations analyses of density-dependent effects on population pa- in the ®eld in the spring, especially in 1995 and 1998, rameters, because the densities frequently differed when vole habitats were extensively covered by ice as markedly between the two patches within enclosures. a result of alternating melting and freezing events dur- There was no tendency for patch-speci®c population ing the winter. densities to be similar within enclosures (enclosure ran- Altogether, these analyses of recruitment and sur- dom effect on log[spring density], F15,18 ϭ 1.00, P ϭ vival show that the highly variable winter decline rate 0.49). had a large among-year component due to a highly The rate of decline in patch-speci®c population den- variable winter survival rate, most likely due to vari- sities (number of animals per habitat patch) over the able winter climate. The winter demography also had winter varied strikingly among the four years (Fig. 1). a considerable density-dependent within-year compo- The declines were most severe during the winters of nent due to a sharply declining recruitment rate with 1994 and 1997. The fairly low-density autumn popu- increasing autumn densities. lations in 1997 (on average 17 Ϯ 8 individuals per patch Individual level parameters [mean Ϯ 1 SD]) were all extinct by the following spring. In 1994, four of the 12 patches, which generally had Turning to the relationship between the autumn body very high autumn densities (77 Ϯ 23 individuals per mass of the individuals and their survival probability, patch in November), went extinct, and the eight extant we found that a logistic regression model with autumn spring populations were only 6% of their autumn size. The autumn populations of 1995 and 1996 had very similar average sizes (1995: 31 Ϯ 13 and 1996: 29 Ϯ 13 individuals per patch), and all were extant in the spring, with 36% (1995) and 22% (1996) of their autumn numbers, respectively. We had to exclude the year 1997 from the analysis because there were no extant populations left in the spring. In addition to a signi®cant partial effect of year (Wald ␹2 ϭ 10.20, df ϭ 2, P ϭ 0.006), there was also evidence for a negative density-dependent rate of change over the winter (␤ϭϪ0.020 Ϯ 0.009 [mean Ϯ 1 SE], Wald ␹2 ϭ 5.12, df ϭ 1, P ϭ 0.024) (Fig. 2). This model accounted for 64% of the total variation (i.e., deviance). On average, 26.5% of the animals in the spring were unmarked due to recruitment over the winter period FIG. 2. Rate of population change in tundra voles from (November±April). A Poisson regression gave evi- autumn to spring (r ϭ log[Nspring/Nautumn]) as a function of dence for a strong negative effect of population density autumn population density for years with extant populations in the autumn on the number of recruits in a patch the in the spring. 3452 JON AARS AND ROLF A. IMS Ecology, Vol. 83, No. 12

small and large animals (Fig. 4), but also a change in body mass from autumn to spring for those animals that survived the winter (Fig. 6). Large individuals, in particular large females (Ͼ29 g), lost mass from autumn to spring, while small individuals (especially males) grew (Fig. 6). The small individuals (Ͻ20 g) in the autumn were juveniles that had been recently weaned, while some of the largest individuals (Ͼ35 g) were still reproducing. There were no signs of reproductive activity in the spring populations.

DISCUSSION The rate of population change over winter had both density-dependent and density-independent compo- nents in this study. In previous studies, the relative

FIG. 3. Yearly winter tundra vole survival rate (with 95% con®dence intervals) plotted against the number of days with temperatures above 0ЊC during midwinter (December±Feb- ruary). Mean winter temperature and the year are denoted above the survival rate estimates. body mass as a second order polynomial (mass ϩ mass2:LRT␹2 ϭ 19.55, df ϭ 2, P Ͻ 0.0001) and with different intercepts for the two sexes (LRT ␹2 ϭ 11.28, df ϭ 1, P ϭ 0.001) and the three years (LRT ␹2 ϭ 167.49, df ϭ 2, P Ͻ 0.0001) explained the variation in survival best in terms of the model selection criterion AIC (we found no evidence for any interaction between year and mass: LRT ␹2 ϭ 1.30, df ϭ 2, P ϭ 0.58). The predicted survival function according to body mass was modal with a maximum survival probability for animals weighing 25 g in the autumn in each year (Fig. 4). Females survived generally better than males (odds- ratio females:males: 1.93, 95% CI ϭ 1.61±2.32), causing the sex ratio to change from 55% females in the autumn to 70% in the spring. A variable denoting whether an individual had been recorded as reproductive or not during the summer did not improve the survival model described above (LRT ␹2 ϭ 0.35, df ϭ 1, P ϭ 0.55). However, reproductive activity and autumn mass were confounded (R2 ϭ 53%, mean autumn mass reproductive: 43.5 Ϯ 8.6 [mean Ϯ 1 SD], non- reproductive: 24.5 Ϯ 7.2), so their relative effects on winter survival were hard to assess. Nonetheless, autumn body mass appeared more important than reproductive history, since a survival model with reproductive history substituting the polynomial mass term gave a less appropriate model (⌬AIC ϭ 5). Moreover, within the group of animals that had reproduced there was a signi®cant negative effect of body mass (LRT ␹2 ϭ 5.92, df ϭ 1, P ϭ 0.015), while this was not the case for those animals that had not reproduced over the course of the summer (LRT ␹2 ϭ 0.61, df ϭ 1, P ϭ FIG. 4. Winter survival probability predicted from body 0.43). mass in the autumn for males (solid circles) and females (open circles) in the three years with extant spring populations of The variance of the body mass distribution shrank tundra voles. Observed survival and deaths (both sexes com- quite substantially from autumn to spring (Fig. 5). This bined) are shown as small black squares at the top and the change partly re¯ects the lower survival rate of very bottom of each panel, respectively. December 2002 WINTER DEMOGRAPHY OF VOLES 3453

FIG. 5. Female and male tundra vole body mass distributions in autumn and spring. impacts of environmental stochasticity and density de- Tast 1984), and most often during winters preceding pendence on the rate of population change over the cyclic peaks in summer population density. However, winter in vole and lemming populations have mostly it has been an unresolved issue whether population den- been inferred from time series data of population density or other circumstances are the triggering factors sity (Reid and Krebs 1996, Stenseth et al. 1998, Hansen (Krebs 1993, Hansson 1984). Although the four winters et al. 1999, Stenseth 1999). However, the underlying in this study differed much in terms of the rate of demographic mechanisms have remained elusive due population change and the mortality rate, there was no to a lack of suf®cient capture±recapture data in terms independent effect of year on recruitment rate, thus of spatial and temporal replication. In this study, we pointing to population density as the most important had access to an extensive data set comprising indi- determinant of winter breeding. vidually marked animals distributed on a large number of population replicates allowing for inferences about determinants of winter demography, both at the level of individuals and populations. We found that the density-dependent component of population change worked through recruitment, but we found no evidence for a density-dependent effect on survival. This result is consistent with previous demographic studies of tundra voles and other Microtus species in the normal breeding season (spring and summer), which showed that recruitment is the most strongly density-dependent demographic parameter (Ostfeld et al. 1993, Ostfeld and Canham 1995, Ims and An- dreassen 1999). However, studies on winter demography are rare, and in particular, winter breeding is a poorly understood phenomenon in northern populations of small mammals (Hansson 1984, Krebs 1993). While responsiveness to photoperiod at the onset of reproductive activity has been shown to exhibit indi- FIG. 6. Change in body mass from autumn to spring for vidual variation under laboratory conditions (Spears individual males (solid circles) and females (open circles) and Clarke 1988, Bronson and Kerbeshian 1995), it is plotted against autumn body mass. Female regression line (solid line): growth ϭ 19.6 Ϫ 0.68(autumn mass), r2 ϭ 0.76, unclear how this could produce population-level var- n ϭ 111, P Ͻ 0.001. Male regression line (dashed line): iation in nature. Recruitment in the winter seems to growth ϭ 25.3 Ϫ 0.62(autumn mass), r2 ϭ 0.53, n ϭ 52, P take place only occasionally in voles (Hansson 1984, Ͻ 0.001. 3454 JON AARS AND ROLF A. IMS Ecology, Vol. 83, No. 12

Although density-dependence in¯uenced the winter dreassen 2000) are absent during the winter period. dynamics through recruitment (at least for those pop- Mustelides were ef®ciently kept out by the predator ulation that did not go extinct), population density at fence and trapping. No tracks of mammalian predators the onset of the winters was a poor predictor of the were observed on the snow cover, even though the fate of a population over the winter when considering enclosures were visited several times during the winter all four winters included in this study. For example, months. the most severe population crashes (causing extinc- Evenstad belongs to a geographic region (the north- tions) took place after the two autumns with maximum ern boreal coniferous forest zone), where the previous (1994) and minimum (1997) population densities, re- population cycle of Microtus voles has vanished during spectively. Indeed, the demographic parameter most the last two decades (LindstroÈm and HoÈrnfeld 1994, responsible for the highly variable winter dynamics in Hanski and Henttonen 1996, Steen et al. 1996). Small this study, namely winter mortality, showed no evi- rodent population ¯uctuations in this region now are dence of density dependence. characterized by declines during the winter leading to The experimental setting at Evenstad was similar low spring population densities (Steen et al. 1996, across the years. Mammalian predators were excluded Hansson 1999). The changes in dynamics have been with fences and removal trapping, and the conditions particularly pronounced for Microtus voles in grassland within the enclosures were standardized by introducing habitats (Hanski and Henttonen 1996, Steen et al. new laboratory-raised founder populations and by 1996). Although climate has been suggested as a po- burning and fertilizing the habitat patches every spring. tential cause that either directly (i.e., due to lack of Thus, potential in¯uences of delayed density-depen- snow or presence of ice; Yoccoz and Ims 1999) or dent factors can most likely be ruled out. Still, the indirectly (i.e., due to predation; LindstroÈm and HoÈrn- population dynamics in the experimental populations feld 1994, Hansson 1999) regulates winter survival, of tundra voles at Evenstad were dominated by large the mechanism of the increased mortality rate during annual variation, both during the winter and during the the winter should be studied in greater detail. summer, as shown elsewhere (Aars et al. 1999, Ims and Some clues to the mechanism by which winter cli- Andreassen 2000). Moreover, just like what is the case mate in¯uences population demography can be ob- for winter dynamics, it is differential mortality that tained by identifying individual determinants of winter drives the between-year variation in summer dynamics mortality. It has been hypothesized that body mass is (Ims and Andreassen 2000). Due to detailed monitoring an important determinant of winter survival (Stenseth of radio-tagged individuals during the summer, it has 1978, Hansson 1990, 1995). This hypothesis is based been demonstrated that the main mortality factor was on empirical studies of northern small mammals show- predatory birds (Ims and Andreassen 2000). Although ing that individuals seem to adjust their body size to we do not have equivalent information on the fates of what is presumably a physiological optimum in terms individuals during the winter, the strong correlation of winter survival (e.g., Iverson and Turner 1974, Mer- between winter temperature and winter survival rate suggests that the winter demography to a large extent ritt and Merritt 1978, Hansson 1990, 1991). In this was climatically driven. In particular, it seems that win- study, a similar adjustment was seen in terms of chang- ters with mild weather are detrimental because of the es in mass in animals that survived the winter: large formation of ice on the ground. When inspecting the animals lost mass, while small animals gained mass. meadow patches at Evenstad during snow melt in the However, to the best of our knowledge this is the ®rst spring, we recorded extensive formation of ice covering demonstration of the expected link between body mass the ground in these two years. Such ice formations will and winter survival. Individuals that were either large both reduce thermal insulation and accessibility of food or very small at the onset of the winter paid a cost in resources and, moreover, causes ¯ooding during spring terms of low winter survival probability. Such indi- thaw. It is possible that the ¯at, homogeneous habitats viduals may have died because they failed to adjust in our experimental areas exaggerated the in¯uence of their mass, or they may have paid a cost following mass such climatic events. However, for natural populations adjustment that compromised survival. inhabiting heterogeneous landscapes, similar critical We suggest that the higher mortality rate in males weather episodes in winter and early spring have also than in females was due to the fact that males over- been suggested to dramatically reduce winter survival wintered at a larger size than did females. Tundra voles in small mammals (e.g., Merritt and Merritt 1978, are polygynous and sexually dimorphic with respect to Boonstra and Rodd 1983). Such episodes introduce a body size, with breeding males ϳ30% larger than fe- strong stochastic element in the demography that may males (Bondrup-Nielsen and Ims 1990, Ims 1997). dominate the population dynamics of some northern Since the large size of males is due to sexual selection, small-mammal populations (Yoccoz and Ims 1999). we speculate that the cost males pay in terms of low Predation can be ruled out as an important factor of winter survival is offset by a reproductive advantage winter survival in our study. The species of birds of in the winter if winter breeding takes place or in the prey that in¯uenced summer survival (Ims and An- early part of the next breeding season (Hansson 1995), December 2002 WINTER DEMOGRAPHY OF VOLES 3455 or by a survival advantage due to their large size in most likely climatically driven, and our results support the summer (Boonstra and Krebs 1979). the idea that climate change, due to an increased fre- The individual determinants of winter survival may quency in mild weather during the winter, can disrupt interact with population processes in several ways. the normal cyclic dynamics of northern populations of First, different winter survival rates of males and fe- voles and lemmings. males leads to a seasonally changing sex ratio in the population. In the tundra vole, the higher mortality rate ACKNOWLEDGMENTS in males than in females (which we think is related to This research was supported by various grants from the the larger body size in males) produced a more female- Research Council of Norway. We thank the following people for constructive comments on the manuscript: Harry P. An- biased sex ratio in the spring than in the autumn. In dreassen, Ottar Bjùrnstad, Torbjùrn H. Ergon, Karen Hodges, some other small rodent species, the sexual size di- Xavier Lambin, and two anonymous referees. morphism is reversed, i.e., breeding females are larger than males (Bondrup-Nielsen and Ims 1990). There- LITERATURE CITED fore, given the kind of size-dependent winter mortality Aars, J., and R. A. Ims. 1999. The effect of habitat corridors rate demonstrated here, we may expect seasonal sex on rates of transfer and interbreeding between vole demes. Ecology 80:1648±1655. ratio changes opposite to what we found in the tundra Aars, J., and R. A. Ims. 2000. Population dynamic and ge- vole. Indeed, both in the wood lemming Myopus schis- netic consequences of spatial density-dependent dispersal ticolor (Bondrup-Nielsen et al. 1993) and in several in patchy populations. American Naturalist 155:252±265. populations of Clethrionomys spp. (Kalela 1971, Yoc- Aars, J., E. Johannesen, and R. A. Ims. 1999. Demographic consequences of movements in subdivided root vole pop- coz and Mesnager 1998), for which females are larger ulations. Oikos 85:204±216. than males (Bondrup-Nielsen and Ims 1990), the sex Bjùrnstad, O. N., N. C. Stenseth, and W. Falck. 1995. A ratio at the onset of the breeding season in the spring gradient in small mammal density ¯uctuations: a statistical is often more male biased than it is in autumn. modelling approach. Proceedings of the Royal Society of Second, the mean size of animals in the fall may London, Series B 262:127±133. Bondrup-Nielsen, S., and R. A. Ims. 1990. Reversed sexual depend on factors such as reproductive activity, age size dimorphism in microtines: are females larger than distribution, and population density. In particular, peak males or are males smaller than females? Evolutionary density animals in cyclic populations are typically larg- Ecology 4:261±272. er than animals in other phases of the population cycle Bondrup-Nielsen, S., R. A. Ims, R. Fredriksson, and K. Fred- ga. 1993. Demography of the wood lemming (Myopus (Chitty 1952, Krebs and Myers 1974, Boonstra and schisticolor). Pages 493±507 in N. C. Stenseth and R. A. Krebs 1979). Therefore, given that winter survival is Ims, editors. The biology of lemmings. Academic Press, dependent on body size, one could expect peak density London, UK. animals to have lower winter survival rates than ani- Boonstra, R., and C. J. Krebs. 1979. Viability of large- and mals in other phases of the cycle. Note, however, that small-sized adults in ¯uctuating vole populations. Ecology 60:567±573. the mean mass of the animals at the onset of winter in Boonstra, R., and F. H. Rodd. 1983. Regulation of breeding 1994 (the year with the highest population density in density in Microtus Pennsylvanicus. Journal of Animal the present study) was not higher than in the other years Ecology 52:757±780. (29.5 g vs. 30.6 g). Thus, population mean body mass Bronson, F. H., and M. C. Kerbeshian. 1995. Reactions of reproductively photoresponsive versus unresponsive mead- could not in this instance explain the large yearly dif- ow voles to simulated winter conditions. Canadian Journal ferences in survival rate over the winter. of Zoology 73:1479±1488. Chitty, D. 1952. Mortality among voles (Microtus agrestis) CONCLUSION at Lake Vyrnwy, Montgomeryshire, in 1936±9. Philosoph- Due to a standardized study setting with new ex- ical Transaction of the Royal Society of London, Series B 236:505±552. perimental populations of tundra voles introduced to Dark, J. 1983. Short photoperiods reduce winter energy re- the same 12 habitat patches over four years, we were quirements of the meadow vole, (Microtus pennsylvanicus). able to highlight winter demography. This has until Physiology and Behavior 31:699±702. now been a little explored, but potentially very im- Gundersen, G., J. Aars, H. P. Andreassen, and R. A. Ims. 2001. Inbreeding in the ®eld: an experiment on root vole portant, aspect of the ecology of northern small-mam- populations. Canadian Journal of Zoology 79:1901±1905. mal populations. We identi®ed both intrinsic and ex- Hansen, T. F., N. C. Stenseth, and H. Henttonen. 1999. Mul- trinsic determinants of winter demography. Density- tiannual vole cycles and population regulation during long dependent recruitment and body size-dependent sur- winters: an analysis of seasonal density dependence. Amer- vival probability over the winter contributed to the size ican Naturalist 154:129±139. Hanski, I., and H. Henttonen. 1996. Predation on competing and demographic structure of the populations at the rodent species: a simple explanation of complex patterns. onset of the breeding season in the spring. However, Journal of Animal Ecology 65:220±232. in terms of yearly variation in rate of population Hanski, I., P. Turchin, E. KorpimaÈki, and H. Henttonen. 1993. change, these intrinsic factors were relatively unim- Population oscillations of boreal rodents: regulation by mustelid predators leads to chaos. Nature 364:232±235. portant compared to events with catastrophic, density- Hansson, L. 1984. Winter reproduction of small mammals in independent mortality that sometimes led to population relation to food conditions and population dynamics. Pages extinctions. These population crashes in winter were 225±234 in J. F. Merritt, editor. Winter ecology of small 3456 JON AARS AND ROLF A. IMS Ecology, Vol. 83, No. 12

mammals. Carnegie Museum of Natural History, Pitts- editors. The biology of lemmings. Academic Press, San burgh, Pennsylvania, USA. Diego, California, USA. Hansson, L. 1990. Ultimate factors in winter weight depres- Merritt, J. F., and J. M. Merritt. 1978. Population ecology sion of small mammals. Mammalia 54:397±404. and energy relationships of Clethrionomys gapperi in a Col- Hansson, L. 1991. Regional and individual variation in body orado subalpine forest. Journal of Mammalogy 59:576± growth in winter of bank voles Clethrionomys glareolus. 598. Acta Theriologica 36:357±362. Ostfeld, R. S., and C. D. Canham. 1995. Density-dependent Hansson, L. 1992. Fitness and life history correlates of processes in meadow voles: an experimental approach. weight variation in small mammals. Oikos 64:479±484. Ecology 76:521±532. Hansson, L. 1995. Size dimorphism in microtine rodent pop- Ostfeld, R. S., C. D. Canham, and S. R. Pugh. 1993. Intrinsic ulations: characteristics of growth and selection against density-dependent regulation of vole populations. Nature large-sized individuals. Journal of Mammalogy 76:867± 366:259±261. 872. Reid, D. G., and C. J. Krebs. 1996. Limitations to collared Hansson, L. 1999. Intraspeci®c variation in dynamics: small lemming population growth in winter. Canadian Journal of rodents between food and predators in changing land- Zoology 74:1284±1291. scapes. Oikos 86:159±169. Spears, N., and J. R. Clarke. 1988. Selection in ®eld voles Hansson, L., and H. Henttonen. 1988. Rodent dynamics as (Microtus agrestis) for gonadal growth under short photo community processes. Trends in Ecology and Evolution 3: periods. Journal of Animal Ecology 57:61±70. 195±200. Steen, H., R. A. Ims, and G. A. Sonerud. 1996. Spatial and Ims, R. A. 1997. Determinants of geographic variation in temporal patterns of small-rodent population dynamics at growth and reproductive traits in the root vole. Ecology a regional scale. Ecology 77:2365±2372. Stenseth, N. C. 1978. Energy balance and the Malthusian 78:461±470. parameter, m, of grazing small rodents. Oecologia 32:37± Ims, R. A., and H. P. Andreassen. 1999. Effects of experi- 55. mental habitat fragmentation and connectivity on root vole Stenseth, N. C. 1999. Population cycles in voles and lem- demography. Journal of Animal Ecology 68:839±852. mings: density dependence and phase dependence in a sto- Ims, R. A., and H. P. Andreassen. 2000. Spatial synchroni- chastic world. Oikos 87:427±461. zation of vole population dynamics by predatory birds. Na- Stenseth, N. C., O. B. Bjùrnstad, and T. Saitoh. 1998. Sea- ture 408:194±196. sonal forcing on the dynamics of Clethrionomys rofocanus: Iverson, S. L., and B. N. Turner. 1974. Winter weight dy- modeling geographic gradients in population dynamics. Re- namics in Microtus pennsylvanicus. Ecology 55:1030± searches on Population Ecology 40:85±95. 1041. Stenseth, N. C., and R. A. Ims. 1993. The biology of lem- Kalela, O. 1971. Seasonal trends in the sex ratio of the grey- mings. Linnean Society Symposium Series. Academic sided vole, Clethrionomys rufocanus (Sund). Annales Zool- Press, London, UK. ogici Fennici 8:452±455. Tast, J. 1966. The root vole, Microtus oeconomus (Pallas), Karels, T. J., and R. Boonstra. 2000. Cocurrent density de- as an inhabitant of seasonally ¯ooded land. Annales Zool- pendence and independence in populations of arctic ground ogici Fennici 3:127±171. squirrels. Nature 408:460±463. Tast, J. 1984. Winter success of root voles, Microtus oec- Krebs, C. J. 1993. Are lemmings large Microtus or small onomus, in relation to population density and food con- reindeer? A review of lemmings after 25 years and rec- ditions at KilpisjaÈrvi, Finnish Lapland. Pages 59±66 in J. ommendations for future work. Pages 247±260 in N. C. F. Merritt, editor. Winter ecology of small mammals. Car- Stenseth and R. A. Ims, editors. The biology of lemmings. negie Museum of Natural History, Pittsburgh, Pennsylva- Academic Press, London, UK. nia, USA. Krebs, C. J., and J. H. Myers. 1974. Population cycles in Turchin, P., and I. Hanski. 1997. An empirically based model small mammals. Advances in Ecology Research 8:267± for latitudinal gradients in vole population dynamics. 399. American Naturalist 149:842±874. Lebreton, J.-D. 1991. Modeling density dependence, envi- Turchin, P., and R. S. Ostfeld. 1997. Effects of density and ronmental variability, and demographic stochasticity from season on the population rate of change in the meadow population counts: an example using Wytham Wood great vole. Oikos 78:355±361. Whitney, P. 1976. Population ecology of two sympatric spe- tits. Pages 89±102 in J. Blondel, A. Gosler, J.-D. Lebreton, cies of subarctic microtine rodents. Ecological Monographs and R. McCleery, editors. Population biology of passerine 46:85±104. birds. Springer-Verlag, Berlin, Germany. Yoccoz, N. G., and R. A. Ims. 1999. Demography of small Leirs, H., N. C. Stenseth, J. D. Nichols, J. E. Hines, R. Ver- mammals in cold regions: the importance of environmental hagen, and W. Verheyen. 1997. Stochastic seasonality and variability. Ecological Bulletins 47:137±144. nonlinear density-dependent factors regulate population Yoccoz, N. G., and S. Mesnager. 1998. Are alpine voles larger size in an African rodent. Nature 389:176±180. and more sexually dimorphic because adults survive better? LindstroÈm, E., and B. HoÈrnfeld. 1994. Vole cycles, snow Oikos 82:85±98. depth and fox predation. Oikos 70:156±160. Yoccoz, N. G., K. Nakata, N. C. Stenseth, and T. Saitoh. 1998. Malcolm, J. R., and R. J. Brooks. 1993. The adaptive value The demography of Clethrionomys rufocanus: from math- of photoperiod-induced shape changes in the collared lem- ematical and statistical models to further ®eld studies. Re- ming. Pages 311±328 in N. C. Stenseth and R. A. Ims, searches on Population Ecology 40:107±121.