*Present address and correspondence: Luc A. Wauters, Department of Structural and Functional Biology, University of Insubria, Varese, Via J.H. Dunant 3 I-21100 Varese (VA), Italy. E-mail: email@example.com
1Social organization and dispersal of red squirrels (Sciurus vulgaris L.) differ between sexes, and intrasexual competition is intense. Therefore, we predicted that demographic parameters should be gender-specific: that is density-dependent factors will be more strongly related to density of the same sex than to density of the opposite sex. We studied the relative importance of within- and between-sex density-dependent factors and of density-independent factors (habitat type, food abundance, winter temperature) on different demographic parameters, in two populations in northern Belgium.
2Spring density of males was positively correlated with tree-seed abundance in the previous year, but this was not the case for females. None of the population parameters we measured differed between habitats, indicating that the same density-dependent and density-independent mechanisms prevailed in coniferous and deciduous habitat.
3Within each sex, we found several demographic parameters that were dependent on the densities of the same sex; however, none of these parameters was found to be dependent on the density of the opposite sex.
4Reproductive rate increased with food abundance and decreased with female density. Adult survival of females decreased with female density in autumn–winter, while survival of adult males in spring–summer increased with the size of the previous year's seed crop.
5Immigration rate of males was higher in spring than in autumn, and spring immigration increased with food abundance. Male recruitment rate, in both seasons, increased with food abundance, but was male density dependent. However, spring–summer loss rates also increased when food supplies were good, suggesting that despite high food availability, emigration of juvenile and subadult males increased when intrasexual competition was intense. Recruitment rate of females decreased with increasing female density. After a good seed crop, more subadult females dispersed, but their settlement success (recruitment) was lower at high female density.
6Seed crop size positively affected red squirrel densities through increased reproduction, immigration and adult survival of males, but density-dependent reproduction and within-sex density-dependent recruitment of locally born juveniles and dispersing subadults limit the fluctuations in numbers and regulate densities in winter–early spring, as well as in summer.
Recent synthesis on population regulation emphasizes that both exogenous factors (that have no dynamic feedback with population density) and endogenous factors (that represent dynamic feedbacks affecting population numbers, possibly involving time lags) affect population change, and that their relative contribution will vary among different population systems (Turchin 1999). The effects of exogenous factors, such as food abundance or weather conditions, can represent important biological processes affecting population change and thus make an interesting subject for study (Julliard et al. 1999; Turchin 1999; Dobson & Oli 2001). Regulating factors, by definition, have a density-dependent effect, involving negative feedback in response to changes in population size (Montgomery 1989; Dobson 1995; Leirs et al. 1997; Turchin 1999; Hixon, Pacala & Sandin 2002). Consequently, a regulated population will persist for many generations with constrained fluctuations, and showing the propensity to increase when small and decrease when large (Turchin 1999; Hixon et al. 2002).
The first purpose of this study was to investigate the relationships between exogenous and endogenous factors and population dynamics, and to unravel the interactions (within and between sexes) in the context of population regulation. Eurasian red squirrels (Sciurus vulgaris L.) are ideal model organisms for such studies. They occur in several habitat types that differ in the overall amount of food resources (Wauters & Lens 1995; Kenward et al. 1998) and within-habitat variation in resource availability can be patchy, resulting in fine-grained patterns of habitat heterogeneity (Lurz, Garson & Wauters 1997, 2000; Wauters & Dhondt 1995; Wauters et al. 2001). Further, the distribution and availability of critical resources (high-energy tree-seeds) can be measured accurately, both at population level and within an individual's home range (Wauters & Dhondt 1992, 1995; Wauters & Lens 1995). Finally, the animal's relatively large body size, diurnal habits and average lifetime allow detailed analysis of their life-history traits and social organization (Wauters & Dhondt 1989, 1992, 1995; Lurz et al. 1997, 2000).
In many species of small mammals, social organization, aggressiveness and space-use patterns differ between males and females, and polygynous or promiscuous mating systems result in differential use of critical resources by the two sexes to enhance fitness (e.g. Ostfeld 1985; Boyce & Boyce 1988; Lambin 1994). In Eurasian red squirrels, for example, the distribution and temporal variation in food resources might affect males and females differently, because: (i) the timing and the amount of energy allocated to reproductive investment differ between the sexes (Wauters & Dhondt 1989; Wauters, Dhondt & De Vos 1990); and (ii) males have larger home ranges (and core areas) than females, which, on average, gives them access to more high-quality food patches (Wauters & Dhondt 1989, 1992). Also, when intrasexual competition is more pronounced than competition between the sexes, the density of resident males and females in the population might have different effects on behaviour and subsequent dynamics of the same sex with respect to the other sex. Male Eurasian red squirrels behave aggressively during most encounters with other males, but rarely in encounters with resident females, and dominant females chase all other females, but are aggressive to males only when near their nest with young: thus aggressive behaviour is mainly directed towards animals of the same sex (Wauters & Dhondt 1989, 1993). The home-range core areas of males overlap strongly with those of other males and with those of females, but adult females defend exclusive core areas against other females (intrasexual territoriality, Wauters & Dhondt 1992). Finally, in both sexes, recruitment into the population by juveniles and immigrants depends on body mass and on the degree of intrasexual competition for space, while competition with the other sex is not important (Wauters & Dhondt 1993; Wauters, Bijnens & Dhondt 1993). Despite these gender-specific life-history traits, which can affect the relationships between food resources, behaviour and population dynamics in different ways, we know of no long-term studies on small mammals that have addressed sex-related variation in factors (including sex-specific densities) influencing the different population parameters of reproduction, dispersal, recruitment and survival. However, in polygynous species with conspicuous sexual dimorphism, such as red deer (Cervus elaphus L.), where morphological and behavioural traits that affect fitness differ markedly between males and females, long-term population studies have demonstrated that the dynamics and the mechanisms of population regulation differ between the sexes (e.g. Clutton-Brock, Major & Guinness 1985; Clutton-Brock et al. 1987, 2002; Albon et al. 2000). Given the strong a priori reasons to believe that male and female squirrels in our study populations would also respond differently to fluctuations in resources, we decided to use a similar approach, analysing population processes (i.e. immigration, recruitment, losses including emigration and juvenile/subadult mortality, and adult survival) for males and females separately. These analyses allow us to test for the effects of food availability and of density of each single sex on changes in density and on seasonal and annual variation in the various population parameters of animals of the same and of the opposite sex. In other words, we investigated whether we can improve our understanding of populations dynamics and regulation by considering how demographic parameters may be influenced differently according to sex. Therefore, we monitored annual fluctuations in tree-seed productivity, the squirrels’ major food source (Wauters & Lens 1995), and analysed the changes in squirrel numbers (density), and annual and seasonal variations in reproductive rate, dispersal rate and adult survival rate, over a 9-year period (1984–93) in two populations of Eurasian red squirrels.
Materials and methods
Red squirrels were studied in two study plots of 30 ha each, which were part of larger forests. One area, hereinafter called coniferous habitat (212 ha, N. Belgium, 51°08′ N, 4°43′ E), was mainly coniferous woodland, dominated by mature (>50 years old) and young stands (15–40 years old) of Scots pine, Pinus sylvestris L., and Corsican pine, Pinus nigra Arnold (together 80% of trees on the study area), with some oak, Quercus robur L. (7%), and beech, Fagus sylvatica L. (3%) occurring in small patches and in rows along forest roads. Poor-quality patches consisted of larch, Larix decidua Mill. (7%), young Norway spruce, Picea abies Karst (1%), and birch, Betula sp. (2%). To the south, the study plot was bordered by meadows and a road that were never crossed by resident squirrels. Farmland and less suitable patches of wood occurred to the west. Along the north and east sides of the study plot suitable habitat continued. The other area, hereinafter called deciduous habitat (150 ha, N. Belgium, 51°16′ N, 4°29′ E), was mainly mature deciduous woodland, dominated by oak (53%) and beech (20%) with some chestnut, Castanea sativa Mill. (5%). A 1-ha plot consisted of mature (>100 years old) Scots pine and a few Corsican pine (6%). Poor-quality patches had some larch (3%), red oak, Quercus rubra L. (5%), and birch (7%). It was bordered by houses and a road to the north and by meadows and a motorway along the south and south-west; these borders were never crossed by residents. Along the east side of the study plot the forest continued (Wauters & Dhondt 1993; Wauters & Lens 1995). In both study areas an edge correction was used to estimate squirrel density, by assuming that the area trapped extended 100 m (average home range radius of female red squirrel, Wauters & Dhondt 1992) beyond the edges of the trapping grid where the woodland continued.
trapping and handling squirrels
Trapping was carried out bimonthly, for at least 5 days, from October 1984 to January 1993 (deciduous area) or January 1994 (coniferous area). Wooden box traps and Tomahawk ‘squirrel’ traps were placed on the ground or against the tree trunk in a 70 × 70 m2 grid and baited with sunflower seeds and hazelnuts. Each trapped squirrel was marked individually using numbered metal ear-tags (type 1003 S National Band and Tag Co, Newport, KY), and weighed to the nearest 5 g using a Pesola spring balance. The length of the right hind foot (without the nail) was measured (0·5 mm) with a thin ruler (Wauters & Dhondt 1989). Sex, age and reproductive condition were recorded. All squirrels first caught as juveniles (6 weeks to 4 months old, body mass below 250 g, Wauters et al. 1993) were defined as locally born offspring (see also Wauters & Dhondt 1995). Litter size at weaning of successfully reproducing females was determined by marking young in the nest, or by trapping and observing young in the vicinity of the nest-tree (Wauters & Dhondt 1995; Wauters & Lens 1995). To ensure that immigrants were not missed during dispersal periods, we set the traps for more than 5 days, continuing until no additional new squirrels were trapped for at least 3 consecutive days (Wauters & Dhondt 1993).
The size of the seed crop of Scots pine, Corsican pine, oak and beech was used as a measure of food abundance. Food abundance was calculated by counting fallen seeds (including the remains of food items consumed before they had fallen) on 1-m2 plots, with an average of one counting plot per 0·5 ha. Overall food abundance was expressed in 104 kJ ha−1, using energy contents of Scots pine seeds, acorns and beechnuts (L. Wauters unpublished data). To analyse effects of annual food abundance on population dynamics, a ‘squirrel-year’ was defined from July (calendar year t) to June (calendar year t + 1), as described in Wauters & Lens (1995). Thus, food abundance in year t (i.e. the seed crop produced in July–August of year t) was used to test the effect of food availability on summer–autumn reproduction, survival and dispersal in year t, and winter–spring survival, spring reproduction and dispersal in year t + 1. Overall food abundance was, in most years, much higher in the deciduous than in the coniferous habitat (Wauters & Lens 1995); thus to allow pooling data for both study areas, estimates of overall food abundance were normalized within each study area (in a set of normalized data, the mean = 0 and SD = 1). The mean daily temperature (°C) from November to February was used as a measure for winter cold.
Eurasian red squirrels in the Belgian populations can have two litters per year. Seasonal changes in numbers are caused by spring reproduction and dispersal, resulting in an increase in numbers in April–May, followed by mortality of juveniles and dispersers over summer. In September–October, summer reproduction and autumn dispersal result in a second increase phase, which again is followed by a decrease in numbers through autumn–winter mortality of locally born juveniles and dispersers (Wauters & Lens 1995). We used the minimum number of animals known to be alive (MNA), from trapping, radio-tracking or observations, during each trapping period as our estimate of population size. Since high trapping success allowed us to tag all squirrels in both study sites, we feel confident that these estimates realistically represent red squirrel densities. Residents were recorded as the number of adults and subadults present in March (spring density), July (summer density) or October (autumn density) of each year that had been marked previously. Red squirrels with established home ranges show high site fidelity in our study areas (Wauters & Dhondt 1992). Therefore, disappearances of adults from the study area were considered to represent incidences of mortality. Seasonal adult survival estimates (6-month period) were made from October year t to March year t + 1 (winter survival), and from April year t to September year t (summer survival). To allow a staggered entry design (Krebs 1999), we used the Kaplan–Meier estimate of finite survival rate for each 6-month period
SK = 1 − (di/ri),
where di= number of adult squirrels that disappeared between time i and time i + 1, and ri = number of adult squirrels alive at time i.
Since reproduction and dispersal occurred in two distinct periods, spring and summer–autumn, and mortality differed between seasons, all population parameters were analysed seasonally. Juvenile squirrels start to disperse around 4–5 months old, while immigrants generally settle or disappear within 2 months of entering the study area (Wauters & Dhondt 1993). Therefore, locally weaned offspring were considered recruits if they were still present when 6 months old; immigrants were considered to be recruits if they were still present at least 3 months after their first detection on the study plot. Thus, red squirrels that were born or had immigrated in spring (March–May) and were no longer present on the study plot on 1 September had emigrated or died. Those born or having immigrated in summer–autumn (July–October) that were no longer present on the study plot on 1 January had also emigrated or died. We defined as losses the animals that disappeared because of mortality or emigration. In the coniferous woodland, 11 male and 15 female radio-tagged immature squirrels that were not recruited on the study area were intensively monitored throughout the dispersal period (for at least 3 months), producing estimates of mortality and ‘real’ emigration. Of these 26 juveniles and subadults (immigrants), 16 emigrated (65%) and 10 died (35%). There was no sex difference in the proportion that effectively emigrated (7 out of 11 males, 63%, and 9 out of 15 females, 67%, Fisher exact test P = 0·85).
Seasonal reproductive rate was calculated as the ratio of the number of locally weaned juvenile red squirrels, of both sexes, on the number of resident females present in March (spring breeding) or July (summer breeding). Immigration rate was estimated as the number of immigrants divided by the number of residents in March (spring immigration) or July (summer immigration). Recruitment rate was calculated as the number of recruits, locally born juveniles and immigrants, per resident for spring (residents present in March) and summer (residents present in July), respectively. Losses, emigration and mortality of immigrants and locally born juveniles were expressed as number of squirrels lost divided by number of residents in March–April (spring losses) and September–October (autumn losses). Immigration rate, recruitment rate, losses and survival rate were analysed for each sex separately.
Population growth rate over the summer of year t (1 April to 30 September) was calculated as
rsum = ln Nautumn − ln Nspring,
while over the winter of year t to t + 1 (1 October year t to 31 March year t + 1) it was calculated as
Each of the population processes (reproductive rate, immigration rate, recruitment rate, losses, adult survival rate) was used as the dependent variable in a multiple ancova model. We tested for effects of study area (coniferous or deciduous) and season (spring or summer–autumn) as categorical variables, density of male red squirrels, density of female red squirrels, normalized food abundance and winter cold as continuous independent variables, and for the interactions between male density and study area, female density and study area, male density and season, female density and season, food abundance and study area, and the interaction between food abundance and winter temperature (see Gurnell 1996). All continuous variables were normally distributed (Shapiro–Wilk's test for normality of data, all P > 0·05, SAS 1989). For all tests with multiple independent variables, models were selected through stepwise removal of non-significant terms in reverse order of significance, and interactions were removed before main effects (PROC GLM or PROC REG, SAS 1989). If not otherwise indicated, data are presented as means, with the coefficient of variance (%CV) expressed as a percentage of the mean.
To evaluate the relative contribution of the different demographic parameters on variation in population growth, we tested the following linear regression models for each sex separately:
Winter population growth rate (October−March) = a + b1 (winter survival) + b2 (autumn recruitment rate) + b3 (autumn−winter loss rate).
A stepwise backward procedure (PROC REG, SAS 1989) was used to eliminate non-significant variables.
fluctuations in food abundance and squirrel density
The size of the tree-seed crops, expressed in MJ ha−1 (or 103 kJ ha−1), fluctuated three-fold, from 496 to 1722 MJ ha−1 over 10 years in the coniferous woodland (mean 1086 MJ ha−1, CV 38%), but nearly nine-fold over 9 years in the deciduous habitat (mean 10 839 MJ ha−1, CV 62%, range 2386–20 000). Annual fluctuations in red squirrel densities were less variable. Spring (prebreeding) densities ranged from 0·83 to 1·17 ha−1 (mean 0·97, CV 12%) in the coniferous and from 0·70 to 1·00 ha−1 (mean 0·97, CV 33%) in the deciduous habitat, except for one year when density in March (1991) increased to 1·7 squirrels ha−1, after a mast-crop of beech. Peak autumn (postbreeding) densities fluctuated between 0·87 and 1·70 ha−1 in the coniferous (mean 1·26, CV 19%), and between 0·77 and 1·63 ha−1 (mean 1·15, CV 26%) in the deciduous habitat. Thus, the variance in population density, in spring as well as in autumn, was greater in the deciduous habitat, while in most years densities were slightly higher in the coniferous habitat. In the latter, population variability appeared to be greater in autumn than in spring (CV = 19 vs. 12%), but this was not the case in the deciduous habitat (CV = 26 vs. 33%).
Effects of habitat, food abundance, winter temperature, the food by winter temperature interaction and the habitat by food interaction on density of male and female squirrels in spring (low density) and in autumn (high density), were investigated with multiple ancova models. There were no significant effects of habitat in any of the four models (all P > 0·2). Spring density of red squirrel males increased with food abundance (food effect F1,14 = 10·2, P = 0·007), and male densities were higher when a rich seed crop coincided with a mild winter (food by winter temperature interaction F1,14 = 7·14, P = 0·018, Fig. 1a). Both variables together explained 51% of annual variation in spring density of male red squirrels in both habitats. Male autumn density was not significantly correlated with food abundance (F1,16 = 0·14, P = 0·71), but it tended to be higher when the previous winter had been a mild one (effect of winter temperature F1,17 = 5·82, P = 0·03, R2 = 0·25). The relation of female spring density with food abundance (F1,14 = 4·17, P = 0·06) differed between habitats (habitat by food interaction F1,14 = 8·50, P = 0·01): in the deciduous habitat density increased with food (Fig. 1b, F1,6 = 6·40, P = 0·045), but this was not so in the coniferous habitat (Fig. 1b, F1,7 = 0·81, P = 0·40). However, this habitat effect was due to a single outlier (high density of 0·83 females ha−1 in spring 1991, following masting of beech, Fig. 1b); excluding this outlier from the analysis there was no significant effect of the food by habitat interaction (F1,11 = 1·73, P = 0·21), or of food abundance (F1,12 = 0·13, P = 0·72) on female density in spring. There was no significant effect of any of the variables tested on autumn density of female red squirrels (all P > 0·05).
The mean values and variance of the different population processes and of population growth are summarized in Table 2 for each habitat and season separately. For males, parameters related to dispersal, immigration, recruitment and loss rate were notably higher in spring than in autumn; there was less of a seasonal bias in females, although they tended to have higher autumn than spring immigration rates (Table 1). Annual variation in the parameters related to breeding and dispersal were much stronger than variation in adult survival rate (Table 1).
Table 2. Selected ancova models, and significant independent variables and classes, for the different population parameters
Male red squirrels Independent variables
Female red squirrels Independent variables
(a) Immigration rate (n = 34)
F1,31 = 28·9, P = 0·0001
No significant variables
F1,31 = 9·01, P = 0·005
Selected ancova model
F2,31 = 17·5, P = 0·0001, R2 = 0·53
No model selected
Linear regression per season for male immigration rate (Mim)
Spring: F2,14 = 6·24, P = 0·012, R2 = 0·47, Mim = 0·77 (± 0·16) + 0·14 (± 0·04) food − 0·03 (± 0·01) male density
Table 1. Population parameters for Eurasian red squirrels, broken down by sex, season and habitats. Mean and variance (coefficient of variance in percentage of mean) of each parameter per season and throughout the entire study period
n = 9 years in the coniferous habitat, n = 8 years in deciduous habitat; except for recruitment rate and loss-rate where n = 8, and n = 7, respectively, for the autumn season.
Male population growth
Female population growth
Male immigration rate
Female immigration rate
Male recruitment rate
Female recruitment rate
Male loss rate
Female loss rate
Adult male survival rate
Adult female survival rate
Reproductive rates did not differ significantly between study areas or seasons (Table 1). Reproductive rate increased with food abundance, and decreased with the density of female squirrels (food F1,31 = 24·0, P = 0·0001; female density F1,31 = 7·53, P = 0·01), both variables explaining 49% of temporal variation in reproductive rate (Fig. 2). Male density had no significant effect on reproductive rate (F1,29 = 0·03, P = 0·87).
Male red squirrel immigration rate varied seasonally and increased with food abundance (Table 2a). This rate was significantly higher in spring than in autumn (Table 1, Table 2a). In spring, immigration rate of male red squirrels increased with food abundance, but this change appeared to be density dependent, decreasing when the number of male red squirrels in spring was high (Table 2a, Fig. 3). In autumn, fewer males immigrated and variation in seed crop size explained 23% of annual variation in immigration rate of male red squirrels. There was no effect of female density on male immigration rate (F1,29 = 0·65, P = 0·43). None of the variables tested significantly affected female immigration rate (Fig. 3, all P > 0·10).
Recruitment rate of male red squirrels tended to be higher in spring than in autumn, but not significantly so (Table 2b). Variation in male density and food abundance explained 35% of seasonal and annual variation in recruitment rate of red squirrel males (Fig. 3, Table 2b). Recruitment rate in female red squirrels decreased when female density increased, and was positively, but not significantly related to food abundance (Fig. 3, Table 2b). However, variation in female density explained only 13% of variation in recruitment rate of female red squirrels. When variation in food abundance was added, the regression model explained 22% of seasonal and annual variation in female recruitment rate (F2,29 = 4·01, P = 0·03, R2 = 0·22, female recruitment rate = 0·51 (± 0·11) − 0·54 (± 0·23) female density + 0·054 (± 0·030) food). Density of the other sex did not affect male or female recruitment rate (all P > 0·8).
Rate of losses in male red squirrels were significantly higher in spring than in autumn and, in both seasons, increased with food abundance (Table 2c). Season and food together explained 57% of variation in the loss rate of male red squirrels (Fig. 4). Although loss rate of males increased with male density (r = 0·42, d.f. = 30, P = 0·017), adding male density in the ancova model did not explain any additional annual or seasonal variation in loss rate. Loss rate in female red squirrels did not differ with season, but increased with female density and with the size of the tree-seed crop (Table 2c), both factors explaining 50% of variation in loss rate of female red squirrels (Fig. 4). Density of the other sex did not significantly affect loss rate of males (F1,27 = 3·00, P = 0·09) or females (F1,27 = 0·45, P = 0·51).
There was a significant effect of season on survival rate in both sexes (Table 2d). The significant effect of the season–male density interaction term indicated that the effect of male density on survival rate of male red squirrels changed with season (Table 2d). In spring–summer (April to September), the regression coefficient of male survival on male spring density was positive (31 ± 24), but not significant (linear regression, effect of male density F1,14 = 1·61, P = 0·23), and survival of male red squirrels increased significantly with food abundance (Table 2d, Fig. 5a). In autumn–winter (October–March), survival rate of male red squirrels was positively related with food abundance and negatively with male density, but neither variable was statistically significant (linear regression, effect of food F1,14 = 2·36, P = 0·15, b = 4·3 ± 2·8; effect of male density F1,14 = 3·79, P = 0·072, b=−34 ± 18). In contrast, survival rate of female red squirrels in spring was not significantly correlated with either food abundance or female density (all P > 0·15), but was density dependent, decreasing when female density increased, in autumn (Table 2d, Fig. 5b). However, male and female autumn densities were strongly correlated (r = 0·69, d.f. = 15, P = 0·002), and a regression of female autumn–winter survival rate on male density also was statistically significant (F1,15 = 5·81, P = 0·03, b =−0·36 ± 15, R2 = 0·28).
population growth of males and females
We tested the effects of study area (habitat), food abundance, male density and female density on seasonal population growth in each sex with a multiple ancova model. Spring–summer population growth rate, both of male and female red squirrels, was positive in all years except one, in each of the two habitats (1986 in the coniferous, n = 9, 1991 in the deciduous habitat, n = 8). Population growth in autumn–winter, in contrast, always was negative for female red squirrels in both habitats, while for males it was positive in one year (1986) in the coniferous and in two years (1984, 1990) in the deciduous habitat. Mean population growth rate (Table 1) did not differ between study areas in any season or for either sex (ancova model, effect of study area, all P > 0·3).
There was no significant univariate model explaining variation in spring–summer population growth rate of males. However, a model containing male density and food had a significant male density effect (male density F1,14 = 4·81, P = 0·046, food F1,14 = 3·91, P = 0·07, R2 = 0·29, growth rate = 0·7 (± 0·2)–0·86 (± 0·39) male density + 0·11 (± 0·05) food). When the food effect was removed from the model, the male density effect became non-significant as well (P = 0·23). Spring–summer population growth rate of female squirrels showed density dependence with female density (F1,15 = 10·4, P = 0·006, R2 = 0·41, growth rate = 0·9 (± 0·2)−1·33 (± 0·41) female density), while food abundance had a nearly significant, positive effect on population growth (F1,14 = 4·21, P = 0·059, estimate 0·10 ± 0·05). In both sexes, autumn–winter population growth was negatively correlated with male density (male population growth F1,15 = 4·55, P = 0·05, R2 = 0·23, Y = 0·20 (± 0·19)−0·65 (± 0·30) male density; female population growth F1,15 = 14·8, P = 0·002, R2 = 0·50, Y = 0·19 (± 0·12)−0·70 (± 0·18) male density). None of the other variables was significantly related to autumn–winter population growth (all P > 0·1). When female density was forced into the model and male density removed, there was a negative, but non-significant, relationship of female density on female population growth in autumn–winter (F1,15 = 3·82, P = 0·07, R2 = 0·20, Y = 0·03 (± 0·14)−0·46 (± 0·23) female density). Thus, autumn–winter decrease in the number of male and female red squirrels was stronger when peak (autumn) densities of males were high, and less pronounced at low autumn male densities. We must remember, however, that autumn densities of males and females were strongly correlated.
When we examined the relationship between population growth rate and the different demographic parameters, there were no differences between the study areas (all interaction terms with study area: P > 0·4). Variation in annual reproductive rate had the strongest contribution to summer population growth of males (R2 = 0·61, Table 3a), while reproductive rate and female survival rate in summer together explained 64% of variation in summer growth rate of females (Table 3a). For both sexes, variation in winter population growth rate was best explained by winter survival of adults and autumn recruitment rate (Table 3b), both factors explaining 53% and 51% of variation for male and female red squirrels, respectively (Table 3b).
Table 3. Stepwise backward linear regression models testing the effects of demographic parameters on within-sex population growth rate. Spring–summer rates of demographic parameters for spring–summer population growth, and autumn or autumn–winter rates of demographic parameters for autumn–winter population growth
This study produced three major results. (1) Fluctuations in red squirrel numbers were affected by density-dependent factors as well as density-independent factors, namely the abundance of tree-seeds, the squirrel's major food supply from summer to late winter–early spring of the next year. (2) Density dependence occurred in most population parameters, and always in relation to density of animals of the same sex, never of the opposite sex, except for autumn–winter survival of females, which decreased when female as well as male autumn densities were high. Moreover, patterns were consistent in both habitats. (3) Density-dependent parameters affected the dynamics of female squirrels (the territorial sex; Wauters & Dhondt 1992) more strongly than those of males. Male densities affected only male autumn recruitment, while female densities also had an effect on female loss rates, female survival and reproductive rates.
fluctuations in squirrel densities
There are very few long-term studies on the dynamics of tree squirrels: in some populations, densities fluctuate in parallel with the size of the tree-seed crop, while in other habitats population size remains stable. Fluctuations in numbers of North American red squirrels (Tamiasciurus hudsonicus Erxleben) were synchronous over large areas of south- and central Canada (Kemp & Keith 1970) and increased when conifers produced large crops (Smith 1968) or when food conditions were improved experimentally (Klenner & Krebs 1991). Population densities in jackpine (Pinus banksiana Lamb.) or high-quality white spruce (Picea glauca Moench) forests in north-west and west Canada, where individuals defend year-round territories around one or more middens (larderhoards containing several thousands of cones), remained stable (jackpine) or fluctuated little, irrespective of large annual fluctuations in the size of white spruce cone crops (Larsen & Boutin 1994; Humphries & Boutin 2000). In temperate, mixed deciduous forests, densities of eastern grey squirrels (Sciurus carolinensis Gmelin) and fox squirrels (Sciurus niger L.) remain relatively stable, or fluctuate annually, and are positively correlated with the size of the previous year's seed (acorn) crop (e.g. Thompson 1978; Nixon, Hansen & Havera 1986; Gurnell 1996).
In our study populations, autumn density fluctuated two-fold, around an equilibrium density of about 1·30 and 1·15 squirrels ha−1 in the coniferous and the deciduous areas, respectively (see also Wauters & Lens 1995). Similar fluctuations were found in populations, occurring at slightly lower densities, from temperate mixed woodlands in England and northern Italy (Kenward & Holm 1993; Kenward et al. 1998; Wauters et al. 2001). Long-term gamebag records of red squirrels in Denmark (Strandgaard & Asferg 1980) fluctuated only two-fold over 25 years, and population indices tended to be correlated with the seed crop of beech. Populations in conifer plantations in northern England, Swedish boreal forests and Central-European montane spruce–beech forests, characterized by cold and/or wet winters and large annual fluctuations in the size of conifer cone crops, occur at low densities and undergo marked annual fluctuations in population size. In these habitats, red squirrel densities varied 10-, 14- and 5-fold, respectively, over 4- to 10-year periods, with maximum densities between 0·16 and 0·44 squirrels ha−1 (Moller 1986; Andrén & Lemnell 1992; Lurz, Garson & Rushton 1995; Münch 1998; Lurz et al. 2000). Winter densities were correlated with the Norway spruce crop produced 1·5 years before (Andrén & Lemnell 1992), or increased after a mast crop of spruce and beech (Münch 1998). Thus red squirrel populations from montane or boreal habitats with high annual variation in primary food resources occur at much lower densities and undergo stronger fluctuations in numbers than those occurring in more stable, temperate mixed woodlands.
density regulation within each sex
The red squirrel populations we studied have persisted over more than a century, thus for at least 20 generations (unpublished hunting records and personal communication). During our study, numbers fluctuated two-fold around an equilibrium density, and autumn numbers in both populations increased eight out of nine times they were below these average densities, and decreased seven out of eight times when above the equilibrium (see also Fig. 2 in Wauters & Lens 1995). Thus, population regulation seems to be at work in our study populations. Summer (April–September) is the period of time when population size normally increases, and for both sexes during this time, growth was most strongly affected by reproductive rate. Reproductive rate in turn had a density-dependent as well as a density-independent (food) component. For females, summer growth rate also increased with their spring–summer survival rate, a parameter not significantly related to any of the variables tested. Annual variation in population growth rates during autumn–winter (October–March), which was generally negative, was, in both sexes, best explained by winter survival of adults and autumn recruitment rate. Since male recruitment rate was density dependent and increased with food availability, the winter decrease in male numbers was strongest when food supplies were poor and summer densities high. The winter decrease in female numbers was strongly density dependent, since both parameters (recruitment and autumn–winter surival) decreased when female summer densities were high and improved when they were low.
tree squirrel demography and food abundance
Positive effects of food availability on squirrel numbers and some demographic parameters have been revealed by long-term or experimental studies on other Holarctic tree squirrels. In North American red squirrels, supplemental food resulted in a strong increase of density, but the decline in density when food was removed was density dependent (Klenner & Krebs 1991). In white spruce forests, average litter size of North American red squirrels differed significantly between years, and the annual proportion of litters from which at least one offspring survived to 18–36 days of age (range 70–>95%) was positively correlated with the corresponding change in cone supply (Humphries & Boutin 2000). Annual variation in tree-seed abundance and the interaction of food abundance with an index of winter cold explained 77% of fluctuations in the summer density of grey squirrels in oak–hazel woods (optimal habitat) in southern England (Gurnell 1996). However, both at very high or very low densities, density-dependent net reproductive rate (combined result of reproduction and dispersal) drove population size quickly back to average levels (Gurnell 1996).
Relatively short-term studies in mixed broadleaf and conifer woods in Belgium and England have already demonstrated the importance of both spatial and temporal variation in food availability, with rich seed crops positively affecting reproductive success and population density (Wauters & Dhondt 1989; Wauters et al. 1990; Lurz et al. 1997, 2000; Kenward et al. 1998). Although these studies showed a positive relationship between fluctuations in the size of tree-seed crops and annual changes in Eurasian red squirrel numbers, they were too short to investigate the effects of different demographic parameters on density fluctuations, or possible mechanisms of density regulation. A common mechanism in these relatively high-density red squirrel populations was that density and reproductive rate correlated positively with tree-seed crops, but reproduction was reduced at high densities of adult squirrels (Kenward et al. 1998). Similarly to these studies, we found that in red squirrels, reproductive rate, largely determined by the proportion of females that weaned offspring and, to a lesser extent, by the number of weaned young per successful female (Wauters & Lens 1995), increased with seed crop size, but was also density dependent with female density.
Some of our findings, however, were somewhat different from those described for British squirrel populations. Only spring (prebreeding) density of males was significantly related to the size of tree-seed crops, but not autumn density, while female spring density fluctuated with seed crop size only in the deciduous habitat. Intrasexual territoriality among adult females, being more pronounced in the coniferous woodland (Wauters & Dhondt 1992), was probably responsible for winter–spring densities of females in this habitat remaining constant in most years (0·4–0·5 ha−1 in 8 out of 9 years, see Fig. 1b). However, a closer look at the food–female density relationship in the deciduous woodland showed that the positive correlation was mainly due to the extremely high spring density in 1991, after a mast crop of beech, while during the other years the food–density relationship had a nearly similar pattern as in the coniferous habitat (see Fig. 1a). The fact that autumn densities of either sex were not significantly correlated with food abundance probably reflects a combination of: (i) a time effect, as the tree-seeds produced in summer had only been available (and consumed) since July–August, thus for a short period of time, making a strong effect on autumn (September–October) densities unlikely; and (ii) the effect of density-regulating mechanisms, since between April and September, most of the density-dependent processes occur (reproduction, spring–summer recruitment and dispersal).
In North American red squirrels, the addition of supplemental food resulted in a strong increase in recruitment of juvenile immigrants in autumn, but both recruitment rate onto food-supplemented grids and the decline in density when food was removed was density dependent (Klenner & Krebs 1991). The authors argued that territorial behaviour was limiting local recruitment of juveniles and immigrants, and regulated numbers at a level at which excess food is still available (Klenner & Krebs 1991; Klenner 1991). We found positive effects of food abundance on male red squirrel immigration rates (particularly in spring), recruitment rates, loss rates and spring–summer survival rates. Annual variation in food abundance also directly affected female loss rate, while the relationship with recruitment rate of female red squirrels was not significant (P = 0·08, see Table 2b). Since for most of these parameters, effects of food occurred together with density-dependent effects, their relative importance is discussed below.
density-dependent processes regulating red squirrel numbers
When density dependence occurred in one or more of the demographic parameters, it was always related to density of the same sex and not that of the opposite sex: in male red squirrels, recruitment rate and spring immigration rate were male density dependent; in female red squirrels, recruitment rate, loss rate and autumn–winter survival rate were female density dependent. Only the latter, autumn–winter survival of females, was also significantly and negatively correlated with male autumn density, probably because autumn densities of both sexes tended to fluctuate in parallel. Finally, reproductive rate, thus production and survival of both male and female offspring until weaning, was density dependent with female density, the sex with the highest reproductive investment.
Density dependence in reproductive rate
Wauters & Lens (1995) showed that density-dependent reproduction in these red squirrel populations was caused by two mechanisms: a direct density effect of more non-territorial floaters at high female density, and an effect of habitat heterogeneity, with more females occupying poor territories at high densities, resulting in low breeding success. This is confirmed here by the density dependence with female and not with male density. In both our populations, and for both sexes, annual variation in reproductive rate contributed most strongly to annual variation in population growth rate during summer. Hence, summer population growth was mainly determined by the combined effects of good spring and summer breeding, which both increased when food supplies were good, but were reduced at high female densities. This provided a mechanism of population regulation where, from spring to autumn, densities of both sexes showed a marked increase at initially low densities and when tree-seeds were abundant, while this increase was limited when densities were already high. In contrast, when food abundance decreased and/or initial (spring) densities were high, the number of squirrels decreased.
Density dependence in immigration and recruitment rate
Increased immigration rate of (mainly) subadult males occurred in the spring following a rich seed crop, and thus coincided with high levels of spring reproduction. Both effects can be explained by rich food supplies increasing winter–early spring body mass of red squirrels. Subadults in good condition are more likely to survive winter and disperse (immigrate), and adult females in good condition are more likely to reproduce successfully (Wauters & Dhondt 1989, 1993). However, immigration rate of male red squirrels was also inversely related with male density, suggesting that more males immigrated at low densities.
It is unclear why neither food abundance nor female density affected immigration rate in the female red squirrels in our study. However, females’ space-use differs from that of males (Wauters & Dhondt 1992), and proximate factors causing females to disperse (and show up as immigrants) might be more subtle. Indeed, subadult females have two options when leaving the natal home range: they can move around over a large area, checking several territories in order to find a vacant one (floaters); or they can become resident as a subordinate female at the edges of the home range of two or more dominant females, waiting for a territory to become vacant (Wauters & Dhondt 1992). As such, female dispersal, which was more intense in autumn, will depend more strongly on changes in local female densities (at a scale of an area of 5–10 ha) than on female density in the population as a whole.
Recruitment rate, of both males and females, decreased with increasing density of the same sex, as predicted from patterns of recruitment at the individual level, with red squirrel immigrants being more likely to establish residency (thus being recruited) in an area where an adult of the same sex disappeared (Wauters & Dhondt 1993). Likewise, the probability of recruitment of locally born juveniles was higher when they were in good condition (high body mass at weaning) and weaned early in the breeding season, giving them a better chance of finding a vacant home range (females) or an area with low density of the same sex (males) (Wauters et al. 1993). Similarly, the behaviour of resident adult grey squirrels seemed to regulate recruitment, and in turn population density, through interactions with juveniles and subadults (Thompson 1978), while in fox squirrels, recruitment of young improved when female density was reduced (Hansen & Nixon 1985; Nixon et al. 1986). Apparently, Eurasian red squirrels differ from other tree squirrels in: (i) strict philopatry occurring very rarely, resulting in neighbours of the same sex not being closely related (Wauters & Dhondt 1993); and (ii) the importance of intrasexual competition for space in both males and females, causing the dispersal/recruitment pattern of each sex to depend mainly on the density of squirrels of the same sex (Wauters & Dhondt 1989, 1992, 1993).
Density dependence in loss rate and survival rate
Loss rate was positively correlated with density: for males it tended to increase with male density, for females with female density. But in both habitats, loss rates were also positively correlated with food abundance, and tended to increase when rich seed crops had caused a strong increase of population size through immigration and production of young, both in early summer (April–May), and in autumn (September–October). At high male and female densities in May and September, loss rates of each sex increased more than at lower densities, and this causes the typical reduction of squirrel density in June–August and again in November–January. Seasonal variation in losses can be explained by changing levels of aggressive encounters, in particular between dominant, resident males and unknown males (immigrants) and juveniles of both sexes, and between dominant resident females and subadult (immigrant) females or unrelated juveniles (Wauters & Dhondt 1989; Wauters et al. 1990). This seasonal increase of intrasexual competition between established residents and juveniles or dispersing subadults trying to settle, causing emigration of subordinate squirrels (65% of non-recruited radio-tagged juveniles and subadults), will be more pronounced when resident densities are already high, explaining the inverse density dependence of loss rate in both sexes. Frequent aggressive interactions may also reduce the body condition of juveniles and subadults, increasing predation risk or mortality through diseases, and mortality accounted for 35% of losses of radio-tagged young squirrels. That density dependence was more pronounced in females, agreed with them being the more territorial sex (Wauters & Dhondt 1992, 1993, 1995).
Survival rates of adults of both sexes were high in most years, and adult mortality generally was higher in spring–summer, when tree-seeds become depleted and squirrels forage on a large variety of secondary food items, than in autumn–winter, when squirrels forage nearly exclusively on tree-seeds (Wauters 2000). Spring–summer survival of males was positively correlated with the abundance of tree-seeds as estimated the previous autumn. This was not the case for females, where autumn–winter survival was female density dependent. These sex differences in survival may be explained by slightly different feeding strategies, especially in spring and summer, with breeding females recovering cached seeds to avoid strong body mass loss, but also foraging on lower-energy seasonal food items, minimizing time away from the nest (Wauters et al. 1995; Wauters 2000). Moreover, density-dependent survival of adult females in autumn–winter could be explained by the presence of more floaters and non-territorial females at high densities, which were less likely to survive winter when all female ‘territories’ are occupied (Wauters & Dhondt 1992).
For red squirrels populations in temperate, so-called ‘stable’ habitats (see Lurz et al. 1997), tree-seed abundance is an important limiting factor, affecting their population dynamics. More juvenile animals of both sexes are weaned and more immigrant and locally born males are recruited after a rich than after a poor seed crop, resulting in a higher spring–summer population growth rate when food supplies are high. However, density dependence in reproductive rate and spring–summer recruitment in both sexes will result in constrained population fluctuations (boundedness; Turchin 1999), with population growth decreasing at high densities even if food is abundant. Moreover, the typical population decline throughout late autumn–winter, is mainly determined by adult survival, which is density dependent in females, and by within-sex density-dependent autumn–winter recruitment in both sexes. So-called mast crops, i.e. extremely rich seed crops, can temporarily drive numbers up to densities that are twice the equilibrium density, but within 15 months, squirrel numbers will return to normal levels (Wauters & Lens 1995). The importance of density dependence also explains why there were only weak or non-significant correlations between annual variation in the size of tree-seed crops and male or female red squirrel densities in spring and autumn. Thus, female density-dependent reproduction and within-sex density-dependent recruitment are the most important demographic processes regulating red squirrel population size, around a long-term equilibrium winter density of 0·50 males and 0·46 females ha−1 in coniferous, and 0·47 males and 0·40 females ha−1 in deciduous woodland.
We recommend that further studies of demography in red squirrels, and other small mammals, should aim to understand the relative importance of gender-specific density-independent and density-dependent factors on seasonal and annual variation in demographic rates.
We thank the families Stoelen, Bittebier and Van Havre and the city of Antwerp for allowing us to work on their estates. K. Larsen reviewed the manuscript and offered helpful suggestions, and comments by R. Smith and an anonymous referee helped to improve the manuscript. D. Preatoni made the figures. Research was supported by a Concerted Action of the Belgian Ministry of Education, and by the European Community (EC-Step-0040 project).