Introduction

Roe deer (Capreolus capreolus) is the most abundant European ungulate with a distributional range spanning from the south of Spain to northern Norway and from the British Isles to the Caucasus (Andersen et al. 1998; Apollonio et al. 2010). Roe deer occupy a wide variety of natural and human-altered habitats (Andersen et al. 1998) and face diverse environmental and climatic conditions, which influence their reproduction—the major component of population dynamics (Andersen and Linnell 2000). The most important parameter of reproductive potential in roe deer females is litter size. Studies on local populations have shown that the number of offspring in a litter varies with individual female quality (body size and body mass; Kjellander et al. 2004; Hamel et al. 2009; Flajšman et al. 2017a), population density and demographic structure (Andersen and Linnell 2000, Nilsen et al. 2009), weather conditions, landscape structure, habitat quality (Nilsen et al. 2004; Toïgo et al. 2006), and the genetic characteristics of individual females within populations (Hewison 1997).

Body mass has been reported to be a strong determinant of female fertility (probability of pregnancy) in red deer (Cervus elaphus), moose (Alces alces), caribou (Rangifer tarandus), white-tailed deer (Odocoileus virginianus), and roe deer (Saether and Haagerund 1985; Gaillard et al. 1992; Pachkowski et al. 2013; Simard et al. 2014; Borowik et al. 2016). In roe deer, litter size varies from 1 to 5 (Danilkin 1996; Flajšman et al. 2017b). Heavier females have larger litters and produce more offspring (Danilkin 1996; Andersen and Linnell 2000; Hewison and Gaillard 2001; Focardi et al. 2002; Macdonald and Johnson 2008; Flajšman et al. 2014, 2017a), though becoming pregnant and then carrying more than one embryo generally require the reaching of a threshold body mass (Focardi et al. 2002; Flajšman et al. 2014). Furthermore, body mass of ungulates often varies latitudinally (Bergmann’s rule; Langvatn and Albon 1986; Sand et al. 1995), so this needs to be taken into account in any meta-analysis at biogeographical scale.

In ungulates, higher population densities can indirectly affect reproduction through social stress and lower per capita availability of food (Hewison 1996; Body et al. 2011; Couturier et al. 2012). This causes a decrease in the number of pregnant females in the population, leads to smaller litters, and postpones the age of the first breeding (Focardi et al. 2002; Bonenfant et al. 2009). The negative effect of high population density on the reproductive rates of ungulates has been reported both in a temporal context in one population (long-term demographic data; e.g. Mysterud et al. 2007) and in a spatial context through the comparison of different populations (Andersen et al. 1998; Andersen and Linnell 2000; Focardi et al. 2002). The inverse relationship between population density and fertility is due to the deterioration of female body condition caused by intraspecific competition for food resources (Stewart et al. 2005; Tavecchia et al. 2005).

To date, studies on roe deer reproduction have been performed in single local populations or have compared a relatively small number of contrasting populations (Hewison and Gaillard 2001; Gaillard et al. 2003; Macdonald and Johnson 2008). In this paper, we aimed to review locally obtained findings from the species’ European range. Drawing on a survey of the scientific literature from 14 European countries, we collated data on roe deer population density, mean body mass of females, and in utero litter size. Based on the findings from local studies, we expected that: (1) litter size would be positively related to female body mass, (2) higher population densities would negatively affect litter size, and (3) these effects would be stronger in subadult than in adult females. The latter prediction is based on studies on large ungulates that showed that fertility of subadult primiparous females is more affected by body mass than that of adult females (Gaillard et al. 2000; Borowik et al. 2016).

Material and methods

We collected published data on roe deer in utero litter size by searching scientific databases (namely ISI Web of Science, Scopus, and ResearchGate), using the following keywords: roe deer, Capreolus capreolus, reproduction, reproductive parameters, litter size, and corpora lutea. Apart from one set of data from Slovenia (Flajšman et al. 2015), we used only peer-reviewed papers on free-ranging populations. For our analyses, we searched for the number of corpora lutea or number of embryos per reproducing female in all females (when age was not defined), and subadult (12–24 months old) and adult females (> 24 months old). The age classes were determined by examination of the tooth eruption pattern (Aitken 1975). Based on previous findings (Borg 1970; Strandgaard 1972; Flajšman et al. 2017b), we assumed that the embryonic loss in the first period of gestation was negligible. Therefore, the data on the number of corpora lutea and embryos were analysed together. In total, 27 studies reporting on 59 local populations in 14 European countries were included in the analysis (Fig. 1, Table S1 in Supplementary material).

Fig. 1
figure 1

Locations of studies on roe deer (Capreolus capreolus) litter size in Europe, conducted in 1948–2015. Black points are data used in the linear mixed-effects models (LMMs). Open points are data on litter size only used for calculating the mean values (in Table 1). The list of data and sources is given in Table S1 (Supplementary material)

Full size image

For each of the studied sites (when available), we gathered information on population density (all age and sex classes combined; number of individuals/km2) and mean body mass of females in a population (kg). Roe deer population densities were estimated by a diverse array of methods (Table S1, Supplementary material), including drive-counts, capture-mark recapture, vantage-point observations, pellet-group counts, and recalculation from harvest data controlled by pellet-group counts. Three sources did not provide information on the counting method. Although the use of diverse methods inevitably created additional variation in the dataset, we do not believe this caused any systematic error that could have led to a false litter size—population density relationship. Also, the range of the reported density variation (the minimum and maximum values differed by an order of magnitude) was far beyond the possible errors of the counting methods.

The majority of the studies reported dressed or eviscerated body mass. In cases where body mass of live animals was reported, we multiplied live body mass by 0.75 to obtain the eviscerated body mass (after Krže 2000). Google Maps (http://maps.google.co.uk) was used to determine the coordinates of the approximate midpoints of the study sites. We tested for differences in body mass and litter size between subadult and adult females with a Mann-Whitney-Wilcoxon test.

We used linear mixed-effects models (LMMs,“lme4” package; Bates and Sarkar 2006) to separately analyse the effects of body mass and population density on the litter size of roe deer females. In the analyses, we included only those populations where data on either density (all females—30 populations, subadults—19, and adults—27) or body mass of females (all females—38 populations, subadults—29, and adults—30) were available (Fig. 1). Only a few populations had information on both body mass and population density (all females—13 populations, subadults—11, adults—11), hence the sample size was too small to analyse the combined effect of these factors on roe deer litter size.

For some populations, we had a data series from different years; thus, we assigned population identification (ID) as a random factor in the mixed modelling framework (Zuur et al. 2009). Finally, we ran six models, which—for all females, subadults, and adults—tested the effects of body mass or population density on litter size. We checked the normality and homoscedasticity in the distribution of the models’ residuals by inspecting their diagnostic plots. In the model testing the effect of body mass on litter size in all females, the diagnostic plot indicated the presence of an influential outlier, which was removed from further analyses. Applied variograms did not show significant spatial autocorrelation in model residuals.

In addition, we tested if female body mass increased with latitude by fitting a segmented regression (“segmented” package; Muggeo 2016) to each of the analysed datasets. For each subset, segmented regression indicated the breakpoint latitudes, the significance of which was tested with Dories’ test for a change of slope. We found that above 56° N female body mass increased northwards, in which case the positive effect of body mass on litter size could have been explained by Bergmann’s rule. We verified this hypothesis by dividing each set of body mass data (all females, subadults, and adults) into southern and northern subsets and tested the effect of body mass on litter size for the two latitudinal subsets separately. All statistical analyses were conducted in R version 3.3.2 (R Development Core Team 2016).

Results

Across study sites, roe deer population density ranged from 4.5 to 73.5 individuals/km2 (Table 1). There was on average 1.79 (± 0.03 SE) embryo or corpora lutea per pregnant female, and the mean litter size was significantly lower in subadult (1.63 ± 0.05) than in adult females (1.90 ± 0.04) (Mann-Whitney-Wilcoxon test, W = 1293, P < 0.001). Higher population densities negatively affected litter size; in a sample of all females, mean litter size declined from 2.0 to 1.4 (P = 0.004; Table 2, Fig. 2). Population density did not affect mean litter size of subadult and adult females (Table 2).

Table 1 Population density, female body mass, and litter size in roe deer populations studied in 1948–2015 throughout Europe. Distribution of the study sites is shown in Fig. 1, and the list of data and sources in Table S1 (Supplementary material). Litter size, number of embryos or corpora lutea per pregnant female; N, number of populations; All females, roe deer populations where no information on female age was available
Full size table
Table 2 The effect of population density on litter size of roe deer throughout Europe. Results of linear mixed-effects models (LMMs). R 2, likelihood-ratio based on pseudo-R-squared; N, number of populations; All females, roe deer populations where no information on female age was available
Full size table
Fig. 2
figure 2

Upper panel: Influence of population density (both sexes and all age classes combined) on litter size of roe deer females in Europe. Full results of the linear mixed-effects models (LMMs) and estimates of model parameters are given in Table 2. Trend line is drawn for the significant effect only. Lower panel: Influence of mean body mass of females on litter size of roe deer in Europe. Full results of linear mixed-effects models (LMMs) and estimates of model parameters are given in Table 3

Full size image

The mean (± SE) body mass was 16.5 (± 0.3) kg and was significantly lower in subadult (15.4 ± 0.4 kg) than in adult females (17.0 ± 0.4 kg) (Mann-Whitney-Wilcoxon test, W = 732, P = 0.001; Table 1). The change in the body mass−latitude relationship at the breakpoints of 56°47′ to 57°49 ′N (see below) was highly significant (Dories’ test for a change of slope, P from 0.001 to 0.008). Between 46° and 56° N, body mass of female roe deer did not show a significant trend of increase contrary to our prediction (P from 0.41 to 0.55), whereas between 56° and 63°N, it increased with latitude, and this pattern appeared in subadult, adult, and all females which supported our prediction (R 2 from 0.75 to 0.83; P from 0.01 to 0.05; Fig. 3).

Fig. 3
figure 3

Latitude-related changes in the eviscerated body mass of females in roe deer populations. Body mass breakpoints were estimated using segmented regression. All females, R 2 = 0.52, P < 0.001; subadults, R 2 = 0.49, P = 0.29; adults, R 2 = 0.45, P = 0.14. Body mass increase with latitude is only significant for northern regions (56°–63°N; P for slopes from 0.01 to 0.05)

Full size image

Female body mass had a significant and strong positive effect on litter size both in subadult (P = 0.001) and adult females (P = 0.001) as well as in the sample of all females (P = 0.004; Table 3). An increase body mass (subadults: from 11.2 to 20.8 kg; adults: from 12.1 to 22.4 kg) resulted in an increase in litter size by 0.88 for subadults, and 1.0 for adults (Fig. 2). In other words, in female roe deer, near doubling of body mass increased the number of embryos by almost 1.

Table 3 The effect of female body mass on litter size in roe deer populations throughout Europe. Results of linear mixed-effects models (LMMs). R 2, likelihood-ratio based on pseudo-R-squared; N, number of populations; All females, roe deer populations where no information on female age was available
Full size table

As female body mass varied latitudinally (see Fig. 3), we analysed the litter size—body mass relationship separately for two subsamples of data (as indicated by the results of the segmented regression): (1) all females—populations located below and above 56°49′ N, (2) subadult females—below and above 57°49′ N, and (3) adult females—below and above 57°26′ N. The pattern of heavier females producing larger litters appeared at both higher and lower latitudes, as predicted. However, due to small sample sizes, this relationship was only significant for subadult and adult females in the southern populations (P = 0.04, R 2 = 0.35 and P = 0.02, R 2 = 0.48, respectively), whereas this relationship was only significant for all females in the northern populations (P = 0.01, R 2 = 0.75; Fig. 4). To sum up, roe deer populations with high body mass were more likely to have larger litters even when taking into account Bergmann’s rule.

Fig. 4
figure 4

Influence of female body mass on litter size of roe deer in European populations located below and above: (1) 56°49′ N for all females, (2) 57°49′ N for subadults, and (3) 57°26′ N for adult females. Trend lines are drawn for significant effects. Breakpoints for the divisions into two latitudinal subsets of populations were shown by the results of the segmented regressions (see Fig. 3)

Full size image

Discussion

Our meta-analysis of factors shaping roe deer litter size at a continental scale is—to our best knowledge—the first such macro-ecological study of this species. Compared to investigations of local roe deer populations, our study operated on values averaged for populations (and not on variation among individuals), so it ignored intra-population variation in reproductive parameters. On the other hand, in such a large-scale study, the variation in both extrinsic and intrinsic factors affecting reproduction is substantially larger than could be expected in any local population, which gives the opportunity to cover a broad spectrum of a species’ response to that variation.

Generally, we showed that individual and population variables that influence the reproductive potential of roe deer females in local populations had a similar effect at a much broader biogeographical scale: litter size declined with growing population density, and it increased with higher body mass of females. Moreover, the body mass−litter size relationship observed at the biogeographical scale was not the sole effect of Bergmann’s rule, as it also held true at lower latitudes, where female body mass was not correlated to latitude. At the continental scale, mean litter size of subadult females was 15% smaller than that of adult females. An age-related difference in roe deer litter size was discernible across the whole range of population density and body mass variation.

The main pathway by which density-dependence affects reproduction is via the body mass of females. Individuals in good condition attain the threshold size, reproduce earlier, and have higher fecundity than females in poor condition (Bonenfant et al. 2009). This pattern has been observed in many species of ungulates, including red deer (Albon et al. 1983; Borowik et al. 2016), reindeer (Gerhart et al. 1997), Alaskan caribou (R. tarandus granti) (Adams and Dale 1998), moose (Testa and Adams 1998; Garel et al. 2009), Scottish Blackface sheep breed (Gunn and Doney 1975), and wild boar (Sus scrofa) (Servanty et al. 2009; Jelenko et al. 2014). Heavier females of roe deer had larger litters and produced more offspring than lighter ones (Andersen and Linnell 2000; Focardi et al. 2002; Flajšman et al. 2014, 2017a; this study).

Melis et al. (2009) found that—in the absence of large carnivores—food availability was the major factor shaping population density of roe deer in Europe and proposed that this was caused by competition for food driving density-dependence in population growth rates. Density-dependent effects in the life-history traits of large mammals most often occur at high densities, when populations approach carrying capacity (Fowler 1981; Gaillard et al. 2000). This is in line with the long-term (over 100 years) data series from eastern Poland showing clear density-dependence in the roe deer population (Jędrzejewska et al. 1997). In general, an increase in population density leads to unequal partitioning of food resources among individuals (as a result of contest competition; Łomnicki 1978) and a reduction in reproductive output (from ovulation to weaning) and increase in mortality (reviewed in: Bonenfant et al. 2009). Indeed, our meta-analysis of population data at the biogeographic scale clearly showed that litter size decreased with an increase in roe deer density.

Although we analysed the effects of body size and population density on litter size separately, both variables emerged as significant driving mechanisms in roe deer populations, similarly to other species of ungulates (e.g. moose: Gingras et al. 2014; Soay sheep Ovis aries: Forchhammer et al. 2001; saiga antelope Saiga tatarica: Coulson et al. 2000). Yet, in our models the two predictor variables explained 54–56% of the variation in litter size. Therefore, although body mass and population density may drive litter size of roe deer at a biogeographical scale, other factors such as variation in local climatic conditions and vegetation productivity, life-history traits, predation pressure, different hunting, and management regimes as well as evolutionary differences in life history traits and adaptations among various phylogenetic lineages of the species may confound the explanatory power of these predictors. Pan-European study on variation in roe deer population density (Melis et al. 2009) showed that the presence of large predators especially the Eurasian lynx (Lynx lynx) is the crucial factor shaping roe deer population response to food resources (approximated by habitat productivity and forest cover), whereas different hunting regimes did not significantly affect the observed variation in densities. Recent studies have shown that in the eastern part of its range (east of central Poland and Hungary), roe deer populations carry a large proportion of mtDNA haplotypes of the ‘Siberian’ lineage—the result of past hybridization with Siberian roe deer (C. pygargus) (Lorenzini et al. 2014; Matosiuk et al. 2014; Olano-Marin et al. 2014). It remains unknown whether the ‘Siberian’ lineage differs from the ‘native’ lineages of European roe deer in life history traits and reproductive biology. Due to the availability of published material, the dataset used in this paper mostly comprised local populations free of lynx and wolf (Canis lupus) predation (comp. Jędrzejewski et al. 2011) and located beyond the western limits of the ‘Siberian’ lineage. Thus, the potential bias resulting from these two factors was negligible in our dataset. However, it should be kept in mind that the patterns revealed in our meta-analysis may only hold true in the areas of Europe covered by the studied data points, in populations not subject to predation and in ‘pure’ lineages of European roe deer.