Introduction

Changes in population density affect most life-history traits of vertebrates (Fowler 1987; Stearns 1992). Among them, body mass plays a fundamental role in population dynamics (Stubbs 1977; Begon et al. 1990). In ungulates, for example, body mass is a good proxy for individual performance (Clutton-Brock 1991). As a general rule, the heaviest individuals survive better at all ages (Gaillard et al. 1997, 2000) and, starting earlier in life, they produce a larger number of offspring than lighter conspecifics (Albon et al. 1983; Hewison 1996). Among cervids living in temperate areas, there is a consistent inverse relationship between birth weight or early growth and population density (Skogland 1983; Hewison et al. 2002), and birth weight is in turn related to adult body mass (Dooney and Gunn 1981; Albon et al. 1987; Clutton-Brock et al. 1987). Thus, the level of intra-specific competition while in utero and during early growth can determine adult body mass and, hence, the early survival and subsequent reproductive rate of cohorts (Albon et al. 1987; Skogland 1990; Rose et al. 1998; Gaillard et al. 2003a for a review on deer).

Inter-annual variation in climate is also known to influence early growth and hence result in the reproduction of large herbivores (Albon et al. 1983; Coulson et al. 2001), possibly through effects on the spatial and temporal quantity and quality of forage (Langvatn et al. 1996). In fact, there is increasing evidence for the delayed effects of conditions during the spring of birth on subsequent life-history traits in temperate vertebrates (see Lindström 1999; Beckerman et al. 2002). Therefore, high population density or severe climatic conditions at birth may have persistent long-term effects on population dynamics because it may not be possible for individuals to compensate later in life (Pettorelli et al. 2002).

Previous studies of cohort effects in ungulates have been restricted to single population studies [e.g., Rose et al. (1998) for red deer (Cervus elaphus), Gaillard et al. (1993b) for roe deer (Capreolus capreolus) and Forchhammer et al. (2001) for Soay sheep (Ovies aries)], without the possibility to make comparisons between contrasting populations over the same density range. To ensure that population studies identify general trends and increase our understanding of biological patterns, it is valuable to adopt a comparative approach. Indeed, a recent analysis using data from 43 previously published meta-analyses revealed that individual ecological and evolutionary studies generally explain a low proportion of variance (mean amount of variance explained was <6%, Møller and Jennions 2002; but see Peek et al. 2003). A comparative approach also ensures that methods are standardized between studies. Here, in this context, we aim to evaluate the influence of population density on fawn cohort body mass in two contrasting roe deer populations.

The roe deer, a small-sized ungulate (adult body mass 20–30 kg; Andersen et al. 1998), is an income breeder (sensu Jönsson 1997), relying on current resource intake for survival and reproduction (Andersen et al. 2000). Environmental factors such as density and weather can have powerful effects on yearly variation in adult body mass of ungulates that stock body reserves (Festa-Bianchet et al. 1996, 1998 in bighorn sheep (Ovis canadensis), Mysterud et al. 2002 in red deer). However, these factors are not expected to affect annual variation in adult mass of income breeders such as roe deer to the same extent. In roe deer, adult body mass of individuals varies very little over years since they do not store large fat reserves to compensate for changes in per capita food availability (Andersen et al. 2000). Even so, such fat reserves may provide a greater proportion of energy needs for a population of a given species in more northern and seasonal environments (Lindstedt and Boyce 1985, for roe deer see Holand et al. 1998). Certainly, selection may act to favor those individuals which store extra fat in harsh climates through decreased winter mortality, particularly in years of severe snow conditions (Cederlund and Lindström 1983). We thus predicted that: (1) body mass of growing roe deer fawns should be highly sensitive to changes in environmental conditions (population density and climatic conditions), and should thereby show marked among-year variation, (2) fawn cohorts in the southern population are expected to show a stronger response to density than those in the north due to increased predictability and higher productivity of the growing season with increasing latitude (Geist 1987; Loe et al. 2005) and (3) roe deer fawns in the more seasonal northern population should be heavier both in absolute terms and relative to their body size than their conspecifics in the southerly population, simply because of higher levels of fat deposited (Lindstedt and Boyce 1985). We tested these predictions using data collected in two long-term studies of marked animals in Sweden (15 years at Bogesund) and France (23 years at Chizé) where density was manipulated experimentally.

Study area

Bogesund

The study was conducted during 1989–2003 at Bogesund, a 1,200-ha research area located a few kilometers north of Stockholm, (59°23′N, 18°15′E) in the hemiboreal zone in east central Sweden. The entire research area constitutes the major part of a peninsula surrounded by water on all sides except to the north. The area is rather flat with low interspersed ridges. A few people live in small houses, and there are three larger farms associated with the open agricultural areas.

A mosaic of forested and field habitats characterize the landscape. Approximately 65% of the area is forested and 25% is agricultural, while rocky areas and bogs constitute 10%. The forests are dominated by two coniferous species, Norway spruce (Picea abies) and Scots pine (Pinus sylvestris), whereas deciduous species including birch (Betula spp.), willow (Salix spp.) and common oak (Quercus robur) are also common in most forest types (Guillet et al. 1996). Associated with the open areas are numerous small meadows and former fields replanted with birch or spruce. Some fields are resown in early autumn or spring with, for example, rape (Brassica napus), oats (Avena sativa) or wheat (Triticum spp.). Dwarf shrubs, primarily bilberry (Vaccinium myrtillus), are patchily distributed, while numerous herbs are abundant in most habitats.

The climate is influenced by the Baltic Sea. The warmest month is July with an average temperature of 16.7°C, and the coldest is January with −3.7°C. Annual precipitation is approximately 550 mm, with most falling in July, and the least in February. Snow cover normally lasts for 80 days, usually from late December to early March, with a maximum snow depth of 13 cm in February. During the years of 1990–1992, winters were unusually mild, with little or no snow, and deviated markedly from the long-term means.

Chizé

Chizé is a 2,614-ha reserve of enclosed forest in western France (46°05′N, 0°25′W). The climate is oceanic with mild winters (mean daily temperature in January is 5.5°C) and hot, dry summers (mean daily temperature in July is 20.4°C and total rainfall in July–August is 98 mm). The soils of Chizé are shallow, calcareous, and not very fertile. Three broad vegetation associations of varying quality for roe deer are found at Chizé (Pettorelli et al. 2003a). The richest habitat for roe deer is an oak forest with mainly hornbeam coppice (covering 1,046 ha). About a third of the area (815 ha) is poor roe deer habitat consisting of beech forest on limestone with virtually no coppice. The understory is dominated by woodfalse-brome (Brachypodium sp.), butcher’s broom (Ruscus aculeatus), ivy (Hedera helix), wild madder (Rubia peregrina) and wood melick (Melica uniflora). Neither woodfalse-brome nor butcher’s broom are favored forages of roe deer. Ivy, which is highly selected by roe deer in winter, decreased following a peak in the size of the roe deer population in 1983–1984. The remainder of the Chizé forest (758 ha) is a habitat of intermediate quality consisting of oak forest in which the coppice is mainly Montpellier maple (Acer monspessulanum). The productivity of the entire forest is quite low (long-term average of 3.77 m3 of wood produced/ha/year versus 5.92 m3 for the same forest type in east France, data from Inventaire National Forestier, unpublished data).

Materials and methods

Roe deer measurements and population estimates

Bogesund

Roe deer were captured since 1988 using trap boxes, nets or as newborns. Live-captured individuals were marked, weighed (kg) and body measurements were recorded (Table 1). In this paper we used total body mass and right-hind foot length from 139 males and 136 females, killed or living deer of known age, shot or caught between November and April at the age of 6–11 months (hereafter called fawns).

Table 1 Cohort-specific body mass (kg±1 SE) of male and female fawn roe deer at Bogesund, Sweden and Chizé, France measured during winter
Full size table

Roe deer density was determined by using the Petersen-Lincoln estimate (Caughley 1977) where the capture-mark-recapture procedure constituted observations of the ratio of marked:unmarked animals in a population marked with radio-collars (collar size 50–300 g; Televilt International AB, Lindesberg, Sweden) during the study. In total, 405 individuals were followed by radio tracking, on average constituting 41% of the population during the study (varying between 10% and 70% among years). Visual field observations of deer were made in September each year from 1989 and onwards. The protocol for “field observations” was performed in the same way over the study period. A minimum of two persons performed animal observations at dawn and dusk during on average 13 days (ranging between 7 and 20 days between years) until the pre-monitoring target number of animal observations was obtained (varying between 70 and 372 observations). The observation effort was spread evenly over 20 zones of the study area. Prior to the start of the Bogesund study, the roe deer population was regularly hunted. Then, at the start of the study in 1988, a ban on all hunting was introduced, allowing deer density to increase rapidly. Subsequently, hunting was reinstated from October 1992 to March 1994 to reduce population density once again. Thereafter (1995 and onwards) we allowed population density to increase slowly (Fig. 1a).

Fig. 1
figure 1

Population density of roe deer estimated at a Bogesund in September and b Chizé in March. Bars indicate max and min for 95% CI

Full size image

Chizé

The roe deer population at Chizé has been intensively monitored for more than 20 years (from 1977 onwards). Each year between 8 and 12 days of captures are organized in January and February. We used data from 629 male and 606 female fawns that were caught and weighed at least once using an electronic balance (Table 1). For each individual that was captured, year of birth, body mass, sex and hind foot length were recorded. The size of Chizé population has been monitored since 1978 by using recent developments of capture-mark-recapture models (see e.g., Gaillard et al. 2003b). The roe deer population fluctuated markedly during the study period, increasing from 350 individuals >1-year-old in the study area in 1979 to about 550 in 1983, and then decreasing to less than 200 between 1993 and 1996 (Gaillard et al. 2003b). Since 1996, the population size has increased to densities similar to those of 1979 (Fig. 1b). The population increased because few roe deer were removed annually and decreased due both to high annual culling and density-dependent responses of population parameters (Gaillard et al. 1993). Recruitment decreased from 1.7 to 1.3 fawns per doe following the population peak in 1983 (Boutin et al. 1987). Most does first bred as 2-year-olds, but does born during the population peak first bred as 3- or 4-year-olds (Gaillard et al. 1992b). Without removals, the annual finite rate of increase (λ) would have averaged 1.25 over the study period (Gaillard et al. 1992a).

Data analysis

Because roe deer show some sexual dimorphism in size (Andersen et al. 1998), we analyzed cohort body mass in relation to sex to test our first prediction of widespread cohort effects in both populations. Between-year (i.e., cohort, n=15 and 23 respectively, for Bogesund and Chizé) variation in fawn body mass was analyzed by standard analysis of variance (ANOVA). To test our second prediction of higher density-dependence at Chizé than at Bogesund, accurate estimates of population density based on the capture-mark-recapture (Gaillard et al. 1993) or capture-mark-resighting monitoring were available between 1980 and 2002 at Chizé and 1989–2003 at Bogesund. We tested for possible effects of density on yearly variation in fawn body mass using the observed sex-specific mass of each roe deer fawn cohort with linear models (one per site), including climatic variables as covariates (to account for possible confounding effects), sex as a factor (to account for the weak sexual size dimorphism of roe deer). To control for fawn growth over the winter catch season (Pettorelli et al. 2001; Hewison et al. 2002), we included the months of capture in the models (6 and 2 modalities at Bogesund and Chizé, respectively). In addition, we controlled for known effects of habitat quality at Chizé (i.e., fawns are heavier in the rich oak-hornbeam forest stand than in the oak-maple and beach forest stands, Pettorelli et al. 2001) and mast availability at Bogesund (i.e., fawns are heavier during mast years, Kjellander 2000) by fitting habitat quality (three classes) and mast occurrence (two classes), respectively, while in utero as additional factors of variation in fawn body mass. Acorn mast production was recorded by a commercial seedling producer (Svenska Skogsplantor AB) at several different locations in southern Sweden, but not in the study area itself. Mast harvest varied between 0 and 500 kg in a poor year and 1.5–4 tonnes in true mast years. Mast years occur concomitantly across the entire Swedish range of oak (Hamilton, Swedish National Board of Forestry, unpublished Data). In temperate areas, both winter severity and summer weather have been found to correlate with temporal variation in demographic traits of ungulates (Sæther 1997; Gaillard et al. 2000). In order to assess density-dependent responses of fawn body mass without any confounding effect of climatic variation, we used proxys of both winter (at Bogesund only because winters are very mild at Chizé) and summer (at both sites) conditions. To index winter conditions, we used the number of days per year with snow lying deeper than 10 cm. We did not use the winter NAO often reported to have a strong influence on several life-history traits, including body mass (Post and Stenseth 1999; Coulson et al. 2001; Mysterud et al. 2002) because NAO did not correlate to local climatic factors at our study sites. To index summer conditions, we used mean daily temperature in June–July because it has been previously reported to negatively affect winter fawn body mass of roe deer fawns (Gaillard et al. 1996). Interactions between covariates were fitted by using the product of interacting covariates of their main effects (see Portier et al. 1998 for further explanations). We then tested various effects of explanatory variables using Wald tests. The best model was selected using the Akaiké Information Criterion (AIC) following Burnham and Anderson (1998). Then, to measure accurately the population-specific response of fawn body mass to changes of density, we standardized our density estimates for both populations so that density had a mean of zero and a standard deviation (SD) of 1 in order to explore density-dependent responses over a density range common to both sites. Lastly, to test our third prediction that fawns should be heavier at Bogesund than at Chizé for a given size, we fitted allometric relationships between fawn body mass and hind-foot length. We regressed body mass on hind-foot length after log-transformation and tested for the effects of sex and site, as well as their interaction, on both the slope and the intercept of the regression using Wald tests. We used R software (R-Development Core Team 2004) for all the statistical analyses.

Results

Cohort differences

At Chizé, when we accounted for possible confounding effects of month of capture (male and female fawns were on average 0.221 and 0.285 kg heavier in February than in January) and habitat type (male and female fawns were on average 1.023 and 1.108 kg heavier in the oak-hornbeam stands than in the beech stand), there was significant variation among cohorts in mean body mass for both sexes during the study period (males from 13.7 kg in 1984 and 1990 to 19.3 kg in 1995: F 22, 606=10.339, P<0.0001, females from 12.9 kg in 1985 and 1986 to 16.7 kg in 1995: F 22, 583=9.987, P<0.0001; Table 1). At Bogesund, when we accounted for possible confounding effects of month of capture (male fawns were on average 1.94 kg heavier in November than in February and female fawns were on average 1.85 kg heavier in November than in December), we found significant differences among cohorts in females only, although the range of cohort variation observed in males was similar to that observed at Chizé (males from 14.4 kg in 1992 to 17.9 kg in 1995: F 14, 119=1.66, P=0.073, females from 12.3 kg in 2003 to 17.4 kg in 1998: F 14, 116=1.96, P=0.027; Table 1). The lack of a significant difference among cohorts in male fawns at Bogesund was likely a consequence of small sample size in some years (see Table 1).

Effects of density and climate

At Bogesund, models including two-way interactions performed less well than models including only main effects (Table 2a). From the additive model, the main effects of sex, density, and mast occurrence were significant and accounted for 56.3% of the variation observed in cohort-specific fawn body mass. Male fawns were 0.97 kg (SE=0.307) heavier than female fawns, fawn body mass decreased by 0.064 kg (SE=0.022) per increase of 1 roe deer/100 ha, and fawns were 0.82 kg (SE=0.362) heavier following mast years. Although the relationships were in the expected direction, the additive effects of snow conditions (cohort-specific body mass tended to decrease by 0.019 kg (SE=0.014) per increase of one snow day) and of summer temperature (cohort-specific body mass tended to decrease by 0.173 kg (SE=0.135) per increase of 1°C) were not retained in the final model (Table 2a).

Table 2 Model selection procedure for the analysis of the effects of population density, climatic conditions, and sex on cohort-specific body mass
Full size table

Likewise at Chizé, models including two-way interactions performed less well than models only including main effects (Table 2b). The model including additive effects of sex and density was selected and accounted for 49.8% of the variation observed in cohort-specific fawn body mass (Table 2b). Male fawns were 0.823 kg (SE=0.282) heavier than female fawns and fawn body mass decreased by about 0.207 kg (SE=0.034) per increase of 1 roe deer/100 ha. As previously reported in this population, fawns tended to be lighter following hot summers (cohort-specific body mass decreased by 0.225 kg (SE=0.128) per increase of 1°C).

Sites combined

To combine our data sets, we used cohort-specific values that had first been corrected for those confounding environmental factors which we had shown to significantly affect fawn body mass variation (mast occurrence at Bogesund). When we standardized population density (mean of 14.18 roe deer/km2 and SD of 6.55) to assess the site-specific response of body mass to given absolute changes of roe deer density, the model including the three-way interaction between site, sex and density was not retained (Table 3). However, the between-site difference in slope (0.604, SE=0.285) was significant (Wald test, P=0.04), indicating that the density-dependent response of fawn body mass differed between the study sites. This shows that the density-dependent response of roe deer at Chizé was greater than that observed at Bogesund (Fig. 2), as expected from our second prediction. The strength of the density-dependent response of body mass over this standardized density range did not differ between sexes (difference of slope = 0.061, SE=0.268, Wald test, P=0.820). The selected model (Table 3) accounted for 50.0% of the variation observed in cohort-specific fawn body mass. Male fawns were 0.883 kg (SE=0.235) heavier than female fawns.

Table 3 Model selection procedure for the analysis of the effects of standardized population density, sex and site on cohort-specific body mass
Full size table
Fig. 2
figure 2

Standardized population density in the year of birth versus mean fawn cohort body mass in male (filled symbols) and female (open symbols) roe deer at Bogesund (squares, broken lines) and Chizé (circles, unbroken line). Population density was standardized to have a mean of zero and a SD of 1. For each site the upper regression line corresponds to males while the lower one corresponds to females

Full size image

Allometric relationships

Data inspection revealed that while the lightest fawns were found at Chizé (11.1 and 8 kg at Chizé vs. 13.1 and 11.1 kg at Bogesund for males and females, respectively), the smallest measurements of hind-foot length were reported from Bogesund (275 and 253 mm at Bogesund vs. 294.5 and 279.5 mm at Chizé for males and females, respectively). Including these individuals in the analysis led to a poor fit of the allometric model (skewed distribution of residuals and marked underestimation of the allometric slope with estimates of 0.983 and 1.869 for males and females, respectively, compared to the expected value of 3 based on the equation-dimension, i.e., volume/length ratio, Peters 1983). Hence we removed the smallest males (N=3) and female (N=1) measured at Bogesund to perform the analysis. We did not find any statistical evidence for a two-way interaction between sex and site on the slope of the allometric relationship between hind-foot length and body mass (Table 4). Further, we did not find any support for a model including interaction terms (Table 4). There was, as expected, a strong allometric relationship between hind-foot length and body mass. The allometric exponent (2.735±0.154 (SE)) did not differ significantly from the expected coefficient of 3. The intercept of the allometric relationship differed highly between sites (difference of 0.079, SE=0.013 in favor of Bogesund) but did not differ between the sexes (difference of 0.010, SE=0.012 in favor of males). For a given size, fawns at Bogesund reached a higher body mass than at Chizé, as expected from our third prediction. Finally, no difference was found between males or females; i.e., male and female fawns followed the same allometric relationship (difference of slope = 0.078, SE=0.306, Fig. 3, Table 4).

Table 4 Model selection procedure for the analysis of the effects of sex and site (as well as of their interactions) on the intercept and slope of the allometric relationship between fawn body mass and hind foot length at Bogesund and Chizé
Full size table
Fig. 3
figure 3

The allometric relationship between body size (metatarsus length) and body mass at Bogesund, Sweden (open squares) and Chizé, France (filled circles) in male and female roe deer fawns

Full size image

Discussion

We found support for all our three predictions. In both study sites, when we accounted for other possible sources of environmental variation, a clear and consistent effect of population density occurred on cohort-specific body mass of fawns such that fawns born at high density weighed less during their first winter compared to those born at low density (Prediction 1). In support of our second prediction (Prediction 2), we found a much stronger density-dependent response at Chizé than at Bogesund. Lastly, as expected from the third prediction (Prediction 3), fawns at Bogesund were heavier than at Chizé, both in absolute (i.e., higher body mass) and relative (i.e., heavier for a given body size) terms, due to a higher degree of fat accumulation in northern environments. Below we discuss these findings in further detail.

Marked cohort effects on fawn body mass

The body mass of fawns was strongly dependent on the year of birth, with mean weights varying by between 3.5 and 5.6 kg depending on sex and site (Fig. 2). This represents a between-cohort difference in mean body mass of up to ±29%. There was a clear pattern in both sites whereby cohorts that were born in years of high density were relatively lighter during their first winter than those born during years with lower density. Such effects of density on juvenile size have been reported in other species of large mammals (Fowler 1987; see Gaillard et al. 2000 for a review on ungulates) and could be caused by differential early growth rates, both in utero and during the first summer of life (Skogland 1983; Clutton-Brock et al. 1992; Cook et al. 1996; Gaillard et al. 1996; Hewison et al. 1996, 2002). In addition to population density, we confirmed the negative effect of summer temperature on body mass previously reported at Chizé for cohorts born before 1993 (Gaillard et al. 1996). However, the absence of a climatic influence on fawn body mass at Bogesund indicates that changes in population density rather than climatic variations predominate in shaping fawn body mass of roe deer.

Population-specific responses of fawn body mass to changes in density

Either when considering the whole range of density observed in the two populations, or when standardizing the between-site comparison to the same relative range of density, the density-dependent response of fawn body mass was more marked at Chizé than at Bogesund. Increasing roe deer density by 10 animals per km2 led to a decrease in mean cohort body mass of 0.64 kg at Bogesund versus 2.07 kg at Chizé. This pattern is striking and, if not a coincidence, requires explanation.

Density-dependent effects are typically related to the interaction between population density and habitat quality (e.g., Sinclair 1989 for a review; Pettorelli et al. 2003b for a case study on roe deer). Roe deer seem to fit that pattern closely. Roe deer at the relatively high latitude of Bogesund have access to high quality habitat during spring and summer. At high latitude, resources should be richer and more predictable during the growing season than at low latitude (Geist 1987; Loe et al. 2005) and, in the absence of severe limitation during winter (as suggested by the absence of an effect of snow conditions), roe deer at Bogesund are expected to experience highly favorable conditions for body growth. In contrast, at Chizé the availability of resources is limited by summer drought, is more spread over the year (less seasonality compared to Bogesund) and is less synchronized with the short time during which roe deer fawns require high energy input (May–June, Mauget et al. 1999). Therefore, habitat quality during spring and summer time is likely superior at Bogesund in comparison to Chizé and, for a given density, we expect the per capita resources to be higher in the more productive habitat. Thus, the higher productivity during late spring and summer time at Bogesund compared to Chizé will support more individuals. This demonstrates the fact that absolute density has no biological meaning by itself. Only the density relative to the habitat quality is expected to influence population performance (see Van Horne 1984). We report here that the marked between-site differences in the density-dependent response of fawn body mass to identical changes in population density allows us to reject any social regulation suggested to occur in roe deer (Strandgaard 1972) which would be independent of habitat quality. Hence, as Andersen et al. (1998), our results support the view that habitat quality rather than social pressure is the regulating mechanism for roe deer populations.

Between-population differences in absolute and relative mass

Roe deer in the northern Bogesund (Sweden) study site were heavier than those in the more southern Chizé (France) study site (Fig. 2). This result seems to support the predicted pattern (Prediction 3) of higher accumulation of body fat in the northern population, even if fawns were also larger in absolute terms. Indeed, comparative studies of roe deer mass across the European range also support such a pattern (Andersen et al. 1998), although it may reflect adaptation to increasingly seasonal environments towards the north rather than to cold climates per se (Lindstedt and Boyce 1985). Further, fawns at Bogesund were heavier for a given body size than at Chizé, but at each site both sexes had the same mass for a given body size. This result is in agreement with the low level of sexual selection exhibited by roe deer (Andersen et al. 1998). However, it is not possible to distinguish the causes of the between-site differences observed in size and body mass (Fig. 3). Lesage et al. (2001) have previously shown that body mass and size of white-tailed deer (Odocoileus virginianus) fawns were higher in populations experiencing severe winter conditions. Further, large animals are able to accumulate a disproportionately larger amount of fat than smaller individuals (Peters 1983), and since individuals with the largest fat reserves and fasting endurance will suffer least winter mortality, there will be strong selection towards larger body size in northern areas (Lindstedt and Boyce 1985). Alternatively Geist’s 1987 suggestion that food resources in northern latitudes are of generally higher quality than in more southern temperate areas might account for our findings, especially because we are concerned here with growth during the short summer period when mothers require high food quality for lactation. As a consequence, roe deer may use the vegetation flush in northern latitudes to increase growth rate and thereby reach higher body mass and larger size. A reverse pattern is expected during winter because of the high degree of seasonality in more northern areas. Summer conditions should therefore be more limiting at Chizé than at Bogesund, while winter conditions should limit access to high-quality food at Bogesund, e.g., because of snow conditions. Both these seasonal effects should operate through an effect on body mass or growth (Albon et al. 1987; Gaillard et al. 1996; Sand et al. 1996).