Ivar Herfindal, Department of Biology, Norwegian University of Science and Technology, N-7491 Trondheim, Norway. Tel.: +47 73596253. Fax: +47 73596100. E-mail: Ivar.Herfindal@bio.ntnu.no
1A general problem in population ecology is to predict under which conditions stochastic variation in the environment has the stronger effect on ecological processes. By analysing temporal variation in a fitness-related trait, body mass, in 21 Norwegian moose Alces alces (L.) populations, we examined whether the influence of temporal variation in different environmental variables were related to different parameters that were assumed to reflect important characteristics of the fundamental niche space of the moose.
2Body mass during autumn was positively related to early access to fresh vegetation in spring, and to variables reflecting slow phenological development (low June temperature, a long spring with a slow plant progression during spring). In contrast, variables related to food quantity and winter conditions had only a minor influence on temporal variation in body mass.
3The magnitude of the effects of environmental variation on body mass was larger in populations with small mean body mass or living at higher densities than in populations with large-sized individuals or living at lower densities.
4These results indicate that the strongest influence of environmental stochasticity on moose body mass occurs towards the borders of the fundamental niche space, and suggests that populations living under good environmental conditions are partly buffered against fluctuations in environmental conditions.
The fundamental niche is a combination of environmental factors that allow a population of a species to persist (Hutchinson 1957), i.e. to have a population growth rate λ larger than 1. This simple definition ignores the effects of density dependence [see Holt, Knight & Barfield (2004) for a more general approach], but whatever the mechanisms for variation in body size, we believe that body size is a fitness-related trait that can be used as an index of the localization of the population in the multidimensional niche space. Accordingly, we suggest that populations with small relative body sizes are located more towards the border of the realized niche than populations with larger body sizes (Holt & Gaines 1992; Holt & Gomulkiewicz 1997).
Here our aim is to examine whether there is a relationship between temporal variation in a fitness-related trait (body mass) within populations and differences in mean body mass among populations. We consider the mean body mass to reflect the position of the population within the multidimensional niche space (Hutchinson 1957) and hence the suitability of the prevailing environmental conditions for the moose. We examine two hypotheses for such a relationship. One hypothesis suggests a positive relationship between temporal variation in body mass and the suitability of the habitat because few constraints affect the positive effects of environmental factors. Alternatively, if individuals are not able to buffer the environmental effects under harsh conditions, we expect larger temporal variation in populations with smaller individuals. These hypotheses were evaluated by examining temporal variation in body mass of moose Alces alces (L.) calves and yearlings across Norway.
Data were collected from 21 Norwegian municipalities covering most of the geographical range of moose distribution in Norway (Garel et al. 2006). For details regarding the study area and municipalities included in the analyses, see Garel et al. (2006). Data from hunter killed moose was collected from 1982 to 2002. For each moose, we had information on carcass mass (body mass from here), sex, age and date of kill. Only calves and yearlings killed in September and October were included in the analyses to avoid large variation in body mass during the autumn. We only included municipalities in which data were available for more than 10 individuals for 10 or more years in each sex- and age-group. Body growth from calf to yearling was estimated as Gt = Wy,t −Wc,t−1, where Wy,t and Wc,t is yearling and calf body mass, respectively, in year t. Owing to small sample sizes for yearling females, temporal variation in body growth was only analysed in 14 time series of males.
We used satellite-derived variables based on the normalized difference vegetation index (NDVI) from the GIMMS data set (Zhou et al. 2001) to describe environmental phenology. The GIMMS data are 15 days maximum composite of NDVI values with a spatial resolution of 8 × 8 km2, and covering the period 1982 till present (Zhou et al. 2001). The NDVI is an index of the relationship between reflected red and near-infrared radiation from the ground, and is found to represent the greenness of the vegetation, or the photosynthetic activity (Myneni et al. 1995). We calculated the following parameters based on the annual curve of NDVI values: onset of spring, length of growing season, derived spring NDVI, peak value, peak time, length of spring, integrated NDVI (Table 1). For more details about these variables and the GIMMS data set and processing, see Reed et al. (1994), Zhou et al. (2001) and Pettorelli et al. (2005a). All parameters were calculated annually for each pixel in the GIMMS data set, and mean values were calculated for municipalities, using pixels with centre inside municipality and below the tree limit. Summer temperature, a variable that not necessarily is reflected by the plant phenology curve, may also influence plant quality (Deinum 1984; Langvatn et al. 1996). Thus, we also included mean temperatures during May and June in our analyses. We assigned values from closest weather station representing similar climate type (e.g. coastal, continental) in those municipalities in which weather data were not available. As a measure of winter severity, we used mean winter temperature and the North Atlantic Oscillation winter index (NAO; Hurrell 1995). The winter temperature was calculated as the mean of monthly temperature means for December, January and February. The NAO for the study period was retrieved from http://www.cgd.ucar.edu/cas/jhurrell/indices.data.html#naostatdjfm (read 10 May 2005). A high positive NAO index is generally associated with relatively warm winters and high precipitation in the northern Atlantic coastal Europe, whereas low values of the index tend to result in cold winters and low precipitation (Hurrell 1995).
Table 1. Description of the explanatory variables used to analyse the effects of environmental variation on moose body mass (βi) and proportion of variance in moose body mass explained by the environmental variable i (p in eqn 1). Variables abbreviations are given in parentheses
Relative moose density (RPD)
Moose killed during hunt/municipality area size below tree limit, divided by the annual above-ground vegetational biomass
1-year lag in relative moose density (RPD1)
Relative moose density previous year
May temperature (MT)
Mean temperature for May
June temperature (JT)
Mean temperature for June
Winter temperature (WT)
Mean temperature for December, January and February. Relates to winter harshness
Onset of spring (OS)
Week number in the year when the NDVI value increase above the value that represent birch leaf burst. Indicates when green vegetation becomes available as forage
Length of growing season (LGS)
Number of weeks between onset of spring and onset of autumn (measured as week number in the year where NDVI value drops below the same threshold value used to calculate onset of spring). Indicates how long green vegetation is available as forage
Peak time (PT)
Week number in summer where NDVI value reaches its highest value
Peak value (PV)
The NDVI value at peak time. Relates to the biomass productivity when it is on top during growing season
Length of spring (LS)
Number of weeks between onset of spring and peak time. Indicates how long fresh vegetation with high nutritious value is available
Derived spring NDVI (DSN)
The NDVI value at onset of spring – NDVI value the previous 15 days composite image. Indicates how fast the plant develop during early spring
Integrated NDVI (IN)
Sum of the NDVI values through the plant growing season. Relates to the production of foliage during growing season
Winter (December–March) North Atlantic Oscillation index. Correlates with temperature and precipitation during winter.
For each municipality, we used the number of hunter killed moose per square kilometre (below tree line) as an index of population density. However, because the number of individuals per unit area does not necessarily reflect the number of individuals per unit food, we divided the number of moose shot per square kilometre of forest with the above-ground biomass estimated for the municipality [for further details regarding this index of population density, see Garel et al. (2006) and Solberg et al. (2006)]. This relative population density (RPD) represents the moose density relative to the biomass of vegetation in an area. Because there may be delayed effects of population density in ungulate population dynamics (e.g. Solberg et al. 1999), we also included relative population density the previous year in our analyses (RPD1).
As a measure of the suitability of the habitat, we used the mean body mass of calves and yearlings in the population (PMBM). The PMBM was calculated by averaging the body mass for each age and sex group within a municipality, standardized these for all municipalities, and use the mean of these standardized values for each age and sex group within a municipality. Moreover, because the effect of environmental condition can be modified by population density (Portier et al. 1998; Weladji & Holand 2003), we used the mean relative population density (MRPD) within municipality during the study period, standardized among populations, as an index of intraspecific food competition. Populations with on average heavy or light individuals are referred to as large- or small-size populations, whereas populations with high or low relative density are referred to as high- or low-density populations, respectively,
Environmental conditions and body mass were calculated as mean values within a municipality, grouped on year, age and sex. These values were standardized within municipality, age and sex.
To reduce the influence of autocorrelation in the time series, we analysed the first order differentiates (Chatfield 1989) of body mass and body growth. Accordingly, in the subsequent analyses body mass and body growth refers to the first order differentiates and not body mass and growth per se. We calculated the first order differentiates as ΔWk,t = Wk,t − Wk,t−1 where W is body mass in year t, and k is one of the four categories male calf, female calf, male yearling or female yearling. Similarly, we also analysed the first order differentiate of annual variation in body growth from calf to yearling ΔGt, and for all the environmental variables (Table 1). Let βj,i denote the effects of environmental variable i (see Table 1) on ΔWt or ΔGt in population j. Following Sæther, Sutherland & Engen (2004) and assuming that the environmental covariates can be introduced as random effects, the fraction of the environmental variance in body mass explained by covariate i in population j is
( eqn 1)
where βj,i is the effect size of environmental variable i in population j, uj,i represents the covariate i (Table 1) in population j and is the residual component of the environmental variance in population j that is not explained by any of the environmental covariates i. We calculated βj,i and pj,i for each age and sex separately. We then analysed the variation in βi and pi in relation to the population characteristics (mean body mass and mean relative density of the populations) with linear models. Age and sex as well as the interactions between age and sex, age and population characteristics, and sex and population characteristics were included as independent variables, except in the analysis of body growth from calf to yearling ΔGt, for which only male data were available. The selection of the best models explaining the variation in βi and pi was based on AICc values and AICc weights w (Burnham & Anderson 2002).
The mean body mass among populations was 66·93 kg ± 6·08 SD, 63·42 kg ± 5·90 SD, 139·69 kg ± 10·77 SD and 130·18 kg ± 9·48 SD for male calves, female calves, male yearlings and female yearlings, respectively. However, there were large differences in the range of body mass between the populations (male calf: 59·06–75·49 kg, female calf: 55·93–71·08 kg, male yearling: 127·59–154·60 kg, female yearling: 122·77–141·55 kg).
temporal variation in body mass
Overall, higher body masses were found for all age- and sex-classes after summers with an early start of vegetation growth (βOS = −0·164, t = −3·91, d.f. = 57, P < 0·001), after a spring that started with a slow progression in plant development [derived spring NDVI; yearlings only (βDSN = −0·295, t = −5·02, d.f. = 26, P < 0·001)], after a spring that lasted long before the peak of the primary production (βLS = 0·108, t = 2·38, d.f. = 57, P = 0·020), after long growing seasons (βLGS = 0·119, t = 2·40, d.f. = 57, P = 0·020), and after summers with low temperatures in May (βMT = −0·149, t = −3·36, d.f. = 57, P = 0·001) and June (βJT = −0·220, t = −6·05, d.f. = 57, P < 0·001). Thus, higher body masses were found in years with a slow phenological development of plants. In addition, higher body masses were associated with cold winters (βWT = −0·081, t = −2·02, d.f. = 57, P = 0·048). A two-way anova with age and sex as fixed factors revealed no significant (P > 0·10) sex differences in the effects of any environmental variable on body mass. However, yearling body mass was associated with a slow plant progression in early spring (derived spring NDVI), whereas calf body mass did not show any significant relationship with this variable (anova, F1,56 = 8·43, P = 0·005). None of the other sets of βi showed age-specific effects (P > 0·10).
June temperature explained the highest proportion of the temporal variation in body mass (pJT = 0·180 ± 0·204 SD), followed by the length of spring (pLS = 0·158 ± 0·175 SD), derived spring NDVI (pDSN = 0·156 ± 0·186 SD), and May temperature (pMT = 0·145 ± 0·141 SD), whereas NAO and relative population density explained a smaller proportion of the variation in body mass (pNAO = 0·106 ± 0·124 SD and pRPD = 0·093 ± 0·135 SD).
High body growth from calf to yearling was associated with years with low June temperature (−0·241, t = −3·08, d.f. = 13, P = 0·009), and June temperature also explained the highest proportion of variance in growth from calf to yearling (pJT = 0·170 ± 0·187 SD), followed by winter temperature and NAO (pWT = 0·156 ± 0·177 SD and pNAO = 0·106 ± 0·128 SD).
environmental effects in relation to population characteristics
The effect of peak time (time in summer when photosynthetic activity reaches its highest level) and length of spring on the variation in body mass decreased with increasing mean body mass in the population (Table 2, Fig. 1a). In small-sized populations the effect was positive, whereas in large-sized populations there was barely any effect (Fig. 1a). Similarly, in small-sized populations there were negative effects of high winter temperatures and NAO, but as the mean body mass in a population increased, these effects more or less disappeared (Table 2, Fig. 1a). These patterns were similar for all age- and sex-classes. The other sets of βi did not covary in any degree with mean body mass between populations (Table 2, Fig. 1a). The negative effect of June temperature on growth from calf to yearling was stronger in small-sized populations than in large-sized, populations (estimate of slope = −0·141 ± 0·075 SD).
Table 2. The best models for the relationship between the regression coefficient βi (the effect of environmental covariate i on mean body mass ΔWt) and (a) the population mean body mass (PMBM) or (b) mean relative population density (MRPD). The effects of age and sex are included in the models. Only models with ΔAICc less than two compared with the best model (ΔAICc = 0) is presented. X indicates variables included in the model. An interaction between two explanatory variables is denoted with *, and w is the AICc-weights. For abbreviations of dependent variables i, see Table 1
The effects of length of growing season and integrated NDVI (variables related to biomass quantity) on the variation in body mass were higher in high-density populations than in low-density populations (Table 2, Fig. 1b), but regarding the length of growing season, only present for yearlings (Table 2, Fig. 1b). No relationship was found between the effects of environmental variation on growth from calf to yearling and mean relative population density.
The overall pattern in proportion of variance in body mass explained by the environmental variables (p) was a decrease with increasing mean body mass in the population, and an increase with increasing mean relative population density. Thus, the environmental variables explained more of the variation in body mass for small-sized populations and populations with high relative densities, as expected when looking at how the β-values varied with these two population characteristics (Figs 1a,b). Moreover, the p was higher for yearlings than for calves and the decrease in p with increasing mean body mass and decreasing relative density in populations was more evident for yearlings than for calves.
In temperate ungulates, quantity, accessibility and quality of forage are important for body growth and development (White 1983; Andersen & Sæther 1992). During summer, access to food is seldom restricted, but quantity and quality may vary considerably (Sæther & Heim 1993). Our results suggest that during spring and summer, variation in forage quality is more important than quantity for temporal variation in moose autumn body mass. Variables related to biomass production (particularly integrated NDVI and peak value) showed little effect on the body mass. In contrast, variables related to forage quality, particularly during spring and early summer (derived spring NDVI, May and June temperature) had large effect on the body mass and growth from calf to yearling. Furthermore, large body masses were found in years that gave early access to fresh vegetation after the winter (Fig. 1). This indicates that forage quality, rather than quantity, is most important for moose body mass. This has previously been suggested to occure both during summer (Sæther 1985; Sæther et al. 1996; Hjeljord & Histøl 1999; Solberg et al. 1999) and winter (Andersen & Sæther 1992; Sæther et al. 1996). Such a relationship between variation in a life-history trait, e.g. the body mass, and quality of food seems to be a general pattern in temperate and arctic ungulates (e.g. for red deer Cervus elaphus L., Langvatn & Albon 1986; Albon & Langvatn 1992; Langvatn et al. 1996; Mysterud et al. 2001a; Pettorelli et al. 2005b; for reindeer and caribou Rangifer tarandus spp., Reimers, Klein & Sørumgård 1983; White 1983; Post & Klein 1999; for sheep Ovis aries L., Mysterud et al. 2001b; Steinheim et al. 2004; for roe deer Capreolus capreolus L., Gaillard et al. 1996). This supports the theory of Klein (1970) that foraging conditions in spring and early summer is most important for variation in demographic traits and population dynamics of northern ungulates. Because this occurs during the period of the year with highest body growth rate, even small variation in the forage quality can have large impact on body mass and development through a multiplier effect (White 1983; Cook et al. 2004). The duration of access to high-quality forage will also affect the period for rapid development of body tissue and fat reserves (Hjeljord & Histøl 1999; Ericsson et al. 2002), hence the positive effect of the length of the growing season in all populations (Fig. 1).
In this study, the effect of environmental variation increased with decreasing mean body mass and increasing relative density of the population. This can have several explanations. First, the variance in the independent variable affects the regression coefficient and hence the proportion of variance explained by the independent variable (Sokal & Rohlf 1995, pp. 464–465). If the variance in the environmental variables (Table 1) is correlated with regional variation in mean body mass, this covariation can explain the relationship between βi or pi, and mean body mass (Table 2, Fig. 1a). In fact, such a confounding correlation with mean body mass was present for some of the environmental variables, although mainly due to the effect of four populations. However, after removing these four populations, similar relationships were found between βi or pi and mean body mass and density of populations as in Table 2 and Fig. 1. Thus, the results and conclusions were not affected by the confounding relationship between the variance in the environmental variables and mean body mass.
Secondly, differences in body mass among Norwegian moose population may be influenced by gradients in environmental factors influencing forage abundance or quality (Sæther et al. 1996; Herfindal et al. unpublished data), e.g. if small-sized populations are more likely to be localized further away from the optimum in the fundamental niche space than large-sized populations. The position in the niche will also be affected by density-dependent effects. However, these effects are rarely included in traditional niche definitions (Hutchinson 1957; Holt & Gaines 1992; but see Holt et al. 2004). Furthermore, because intraspecific competition will be less intense at low population densities relative to the available biomass (Côte et al. 2004; Stewart et al. 2005), individuals may be better buffered against temporal variation in climate. Such buffers, being larger body mass or better access to critical resources, can prevent individuals from utilizing important body tissue during periods of food shortage, and make them better able to retain a stable body condition (Reimers 1984; Sæther & Gravem 1988; Cederlund et al. 1991). Accordingly, individuals that lose more of their body tissue during winter may be more dependent on high-quality forage during summer for compensating their losses than larger individuals. A similar pattern was suggested for wild reindeer, for which recruitment rates were less affected by climatic harshness in populations living under favourable environmental conditions than in populations in poorer environments (Skogland 1985).
Thirdly, body mass-dependent responses to temporal environmental variation may also be related to the large differences in body growth rates recorded among Norwegian moose populations (Garel et al. 2006). Accordingly, large-sized moose grew faster than smaller-sized moose (Garel et al. 2006), and thus may have reached a higher proportion of the adult body mass as yearlings. Because the period of early growth and development is the most sensitive for environmental variation in mammals (Lindström 1999), smaller-sized moose may need longer periods of body growth, which could result in higher sensitivity to fluctuations in environmental conditions. However, although larger-sized moose grow faster than moose of smaller size, they also grew for a longer period of time (Garel et al. 2006). Thus, the difference in proportion of adult body mass gained as yearling did not differ among populations. In addition, if there was a smaller effect of environmental variation on individuals close to adult body mass, the environmental variables should explain a higher proportion of the variation in body mass for calves than yearlings because yearlings are closer to adult body mass. This was not the case in the present study because the proportion of variance in body mass explained by the environmental variables was higher for yearlings than for calves, and in particular for smaller-sized moose or for moose living at high densities (see ‘Environmental effects in relation to population characteristics’ in Results). This indicates that there is a buffer of resources available for larger-sized moose during the first winter when forage shortage often leads to loss of body tissue and reduction in body condition (Sæther & Gravem 1988).
To summarize, our results suggest that the relative influence of environmental fluctuations on body mass is less in populations with large individuals or in populations with low densities than in populations with smaller individuals or higher densities because the individuals are more likely to be buffered against environmental stochasticity when the resource availability is large. This concurs with previous studies showing stronger effects of density-independent processes in populations already weakened by density dependence or other factors (Skogland 1985; Sæther 1997; Hallet et al. 2004). Furthermore, although the ecological mechanisms influencing life-history traits (e.g. body mass) may vary considerably over relatively short geographical distances (Mysterud et al. 2001b), this suggests that we still can be able to predict, based on knowledge of basic population characteristics, under which conditions the effects of environmental stochasticity are expected to be most prominent.
We are grateful to comments from J. Fryxell and A. Loison on previous versions of the manuscript. We also thank Compton J. Tucker at Goddard Space Flight Center, USA, for providing us the GIMMS data set, and the thousands Norwegian moose hunters, and local moose managers for providing the moose data. This project was funded by the Directorate for nature management and the Research Council of Norway (programs NORKLIMA and Changing Landscapes).