Combined effects of long-term feeding, population density and vegetation green-up on reindeer demography

Large herbivores living in seasonal environments are generally food-limited through density dependence and climatic factors. This may result in starvation and increased mortality in winter and reduced fecundity the following summer. Variability in winter conditions has the potential to create persistent fitness differences at the cohortand the individual-level in iteroparous species. Using a 9-year dataset from two herds of individually marked free-ranging reindeer we assessed whether population growth rates, somatic allocation (female body mass) and reproductive allocation (reproductive success and calf body mass) were affected by supplementary feeding, population density and the timing of the onset of spring (i.e., vegetation onset). The supplementary fed population had a higher population growth rate, the females were more likely to reproduce and their calves were heavier than in the control population. Female body mass was negatively related to timing of vegetation green-up in both herds. Since both populations increased in the last decade we found support to our prediction that density-dependence negatively affected our study herds. Indeed, density negatively affected growth rates, female body mass, reproductive success and calf body mass in both populations and, as expected, this effect was more marked in the control herd. We suggest that food supplemented females may, at least partially, be able to compensate for the energetic costs of negative density-dependence following late vegetation green-up while control females may not. Our findings reveal that late winter conditions have an important limiting role in the study area and that density-dependent food limitation in late winter/early summer acts as a main factor affecting our reindeer population.


INTRODUCTION
Population dynamics and life-history traits of large herbivores can be greatly influenced by stochastic environmental variation, density-dependence and vegetation productivity (Gaillard et al. , 2000. Moreover, the influence of these extrinsic factors on reproduction and survival operate mainly through effects on body mass (Jorgenson et al. 1993, Saether 1997, Gaillard et al. 2003. Body mass is therefore an important phenotypic characteristic with a strong influence on individual fitness (e.g., Saether 1997, Gaillard et al. 2000.
In temperate environments large herbivores experience large contrasts in climate and food availability between seasons (e.g., Fauchald et al. 2004a, Bå rdsen et al. 2008. Body mass has been found to be affected by both snow cover and food accessibility in winter (Schmidt andHoi 2002, Solberg et al. 2004) and food quality and quantity in spring/summer (Pettorelli et al. 2005a). Increased competition for food due to increased density can considerably affect body mass further modulating reproduction and neonatal survival (Forchhammer et al. 2001, Pettorelli et al. 2007, Bå rdsen et al. 2011. Furthermore, it has been suggested that the negative effects of harsh climatic conditions are exacerbated by population density (Coulson et al. 2001, Bonenfant et al. 2009) and that poor early life development conditions can have long-term consequences on a cohort's life history (Lindströ m 1999, Lummaa and Clutton-Brock 2002, Solberg et al. 2004. In reindeer, Rangifer tarandus tarandus, Tveraa et al. (2003) have demonstrated that body mass at the end of the winter is of critical importance to survival and reproduction. Besides the fact that late winter conditions can directly affect body mass development and survival, harsh winters can affect reindeer indirectly through modifications of their reproductive allocation the following summer (e.g., Bå rdsen et al. 2008). During summer, females with an offspring have to balance the amount of resources spent on their own gain in mass and investment in reproduction, which may affect their survival probability the following winter. The effects of reproductive decisions are observed on the following autumn body mass due to the lost opportunity for individuals with calves to gain mass in summer (Bå rdsen et al. 2010). Thus, body mass plays a central role for reindeer life histories and population dynamics. The effects of density and environmental fluctuations on reindeer are mostly food mediated, and supplementary feeding has been used to improve late winter condition (Bå rdsen et al. 2008(Bå rdsen et al. , 2009). In fact, supplementary feeding of ungulates is a common management practice used to buffer the effect of environmental stochasticity on body mass by protecting animals from winter starvation while maintaining high population densities (Fauchald et al. 2004a, Putman andStaines 2004).
Most long-term studies on wild large herbivores are observational and experimental manipulations are largely lacking ). Indeed, long-term studies (of marked individu-als) combining an experimental manipulation are needed to unravel the relationship between population parameters and variables such as density and environmental factors ). Here we report the long-term population level effects of food supplementation, population density and vegetation green-up on winter body mass and reproductive success of two reindeer populations in northern Norway using data from 2002-2010. In the present study, feeding conditions were manipulated as one population received supplementary feeding during late winter/early summer (treatment) while the other was grazing only on natural pastures (control). We expected supplementary feeding to affect population dynamics by increasing the ecological carrying capacity and the intrinsic population growth rate of the treatment population. Furthermore, we predicted supplementary feeding to result in elevated body masses, improved fecundity and a reduced effect of reindeer density and environmental variation on average calf and female body masses as well as female fecundity. Onset of spring was chosen as our measure of environmental conditions as it has been shown to be a main climate associated predictor in other studies of northern large herbivore populations (Pettorelli et al. 2005a, b, Helle and Kojola 2008, Post and Forchhammer 2008. Accordingly, we predicted that supplementary feeding would reduce the effect of the timing of the onset of spring. Finally, since late gestation and early postnatal life are critical periods for developmental growth, we expected calves to be more affected by environmental variation than adults.

Reindeer herding in Norway
The semi-domestic reindeer husbandry in Norway is organized in a system of owners, siidas and districts. Each siida represents a group of owners, and one or several siidas form a district that utilizes an appointed territory defined by law and agreements, both in space and time. To avoid mixing of herds, summer siida pastures in most districts of Finnmark, Northern Norway, are well defined by a system of fences and natural barriers. In addition, the animals are marked with owner specific ear-cuts.
v www.esajournals.org In winter, the siida-herds are separated by continuous herding on a defined share of common pastures. If siida-herds get mixed, the animals are gathered and the herds separated.

Study area and reindeer data
The present study was conducted on two different semi-domestic reindeer populations in Finnmark (Fig. 1). The two herds are members of the same summer siida and from May to October (post-rut), they share an oceanic summer pasture area of ;400 km 2 strictly defined by fences and natural barriers (Fig. 1). This area is dominated by rugged terrain rising from sea-level to the highest peaks at ;1000 m above sea levels (a.s.l). During autumn, they gradually migrate to the continental winter pastures further inland, with rolling terrain, wood clad valleys at ;300 m a.s.l and bare hilltops at ;500 m a.s.l. In December, the animals are gathered and the two populations are separated by herding on the common winter pastures. During winter, both herds share the same pastures (i.e., the same habitat conditions) until they start migrating towards the summer pastures in April.
Since 2000 the population subjected to experimental manipulation has been kept within an enclosure from mid/late April until late May/ early June (depending on snow and pasture conditions). In the enclosure each individual of the population was supplementary fed with 800 g per day of reindeer pellets (Poron Herkku Raisio, Finland; calorific value: 9.9 MJ/kg). The control population was not subjected to any foraging manipulation since animals were using natural pastures only (Fig. 1). To secure enough natural pasture during calving, the enclosure (;5 km perimeter) was moved annually within a larger area close to the border of the summer pasture until spring green-up started. Calves from the treatment population were marked in late May/early June and released on the summer pastures where they mixed with the control population. The control herd reached the calving ground inside the summer pasture area in early May, distant from approximately 15 km of the supplemented herd, and offspring were marked during a collective gathering in early July.
Calf predation was monitored closely in the two herds during summers 2002 and 2003 (Fauchald et al. 2004b). In the treatment popula-tion the mortality inside the enclosure during May was 3%. Free ranging in the summer pastures, the manipulated population suffered a loss of 1.5% from June-September while the control population suffered a loss of 3% from July-September. Based on these records and the fact that predators are actively removed by humans in the study area, predation was considered to be low and herders' management and harvesting strategies appeared as the main top-down processes affecting population growth rates within both populations.
A sample of 476 adult females (older than 2 years; treatment: N ¼ 355, control: N ¼ 121) were included in the study between winter 2002 and winter 2010. Each study animal was individually marked with an ear tag and/or a collar for identification. Presence of a calf was registered at the owners' calf markings by observing if females had a calf on foot in late May early June in the supplementary fed population and in July in the control population. In winter (December), female and calf body mass were registered to the nearest 0.2 kg using an electronic balance (Avery Berkel, Birmingham, UK). However, no calf body mass was recorded in winter 2002. We used the data available from official counts made in the spring each year as an index of reindeer density, i.e., the number of males, females and calves, counted in the spring divided by the size of the summer pasture area (km 2 ). Official counts of harvested animals on the 31st of March have been used to estimate the ratio between the number of harvested animals and the total number of reindeer for each year.

Vegetation data: spring green-up
Timing of the onset of spring has been shown to affect the performance of ungulates inhabiting highly seasonal environments: earlier green-up, which is related to higher quantity and/or quality of food resources, positively affecting body masses (Pettorelli et al. 2005a). Furthermore, spring green-up is an important predictor of environmental conditions in our study area (Bå rdsen and Tveraa 2012). Although other climate indices and environmental factors encountered during other seasons might also affect our population, there were not considered in this study.
We used a double logistic model to describe v www.esajournals.org the Enhanced Vegetation Index (EVI) values from the Moderate Resolution Imaging Spectroradiometer (MODIS) to determine spring onset (see Beck et al. 2006). The EVI was developed to optimize the vegetation signal with improved sensitivity in high biomass regions and improved vegetation monitoring compared to the Normalized Difference Vegetation Index (NDVI, Huete et al. 2002). By taking into account the blue reflectance, the EVI improves vegetation monitoring sensitivity through a reduction in atmosphere influences and a de-coupling of the canopy background (e.g., soil wetness) in, for example, open canopies like tundra and wetlands (Huete et al. 2002

Statistical analyses
To estimate the intrinsic growth rate (r) and the ecological carrying capacity (K ) we used the Ricker-model, which is a discrete generation model suitable for the understanding of population dynamics (Forsyth and Caley 2006). We followed Morris and Doak (2002), using log (k t ) as the response and population density (D t ) as the only potential predictor in the Ricker model: v www.esajournals.org The model was fitted using nonlinear least-square regressions to the time series data for the control and the manipulated herds separately using the nls function (Pinheiro and Bates 2000). To control for harvesting strategies in the two herds we used the ratio between the number of slaughtered animals and the total number of reindeer, log transformed, in the growth rate analysis.
We used linear models (LM) to analyze the effect of supplementary feeding, population density and spring onset on growth rates and body mass of both the females and their calf. Population, spring onset, density and calf sex were used as potential predictors. In all models, we started by fitting a global model containing all predictors and biologically meaningful interactions. From this model, we formed a set of a priori candidate models with reduced number of parameters where all covariates and interactions were removed sequentially (see Appendix: Tables A1, A2 and A3). From this pool of models we selected the most parsimonious model with the lowest Akaike's information criterion (AIC or AICc in the growth rate analysis only) value (Burnham and Anderson 2002; Appendix: Tables A1, A2 and A3). When the difference between two models was 2, we retained the simplest model according to the principle of parsimony (Burnham and Anderson 2002).
In order to analyze the effect of the previous explanatory variables on reproductive success, i.e., data on calf presence (1) or absence (0) for each individual female, we used generalized linear models (GLM) with logit link-function and a binomial distribution (i.e., logistic regression model). Similarly to the analyses of body mass we used the same model selection procedure as for the linear models above (see Appendix: Table A4 for details). All statistical analyses were performed in R 2.14.1 (R Development Core Team 2011), and the null hypothesis was rejected at an a-level of 0.05.

RESULTS
Reindeer density increased from 2002 to 2010, notably in the food supplemented group (R 2 ¼ 0.86, F 2,13 ¼ 39.95, P , 0.01, Fig. 2A). The average body mass of females and calves decreased v www.esajournals.org throughout the period for both herds (Fig. 2B, C). Finally, reproductive success was higher in the supplemented group compared to the control group (Fig. 2D).

Growth rates
The supplementary fed population showed a more positive growth rate compared to the control population (main effect of treatment: 0.33, P , 0.01, Table 1A), but the growth rates in both populations were negatively affected by density (main effect of density: À0.77, P , 0.01, Table 1A). Neither spring onset nor any of the interactions involving manipulation affected growth rates. Intrinsic growth rate and carrying capacity were higher in the supplemented population compared to the control one (r ¼ 0.62 vs. 0.39 and K ¼ 1.77 vs. 1.68, Table 2) and as reindeer density increased, the annual growth rate of both populations declined (Fig. 3A). Harvesting rates did not differ between the two populations (ANOVA: F ¼ 2.36, P ¼ 0.14) and variation in harvest rates did not have a significant effect on the population growth rates in the populations (main effect of harvest: 0.13, P ¼ 0.32, Table 1A).

Female body mass
Reindeer belonging to the supplementary fed population were not heavier compared to animals in the control population (main effect of treatment: À0.61 kg, P ¼ 0.18, Table 1B). Density had a negative effect on female body mass (main effect of density: À7.29 kg, P , 0.01, Table 1B). Moreover, the interaction between the manipulation and density was positive and significant meaning that the supplementary fed population was less affected by negative density-dependence compared to the control one (Treatment 3 Density interaction: 4.90 kg, P ¼ 0.01, Table 1B and Fig. 3B). Spring onset had a negative effect on female body mass (main effect of spring onset: À0.16 kg, P , 0.01, Table 1B) and as the interaction between manipulation and spring onset was not significant (Treatment 3 Spring onset interaction: 0.10 kg, P ¼ 0.09, Table 1B) the effect of spring onset was similar across the two populations. Nevertheless, the treatment population was less affected by the negative effect of late spring green-ups on female body mass (Table 1B, Fig. 4).

Calf body mass
Calves in the supplementary fed population had a higher body mass than calves belonging to the control group (main effect of treatment: 1.56 kg, P ¼ 0.02, Table 1C). Males were significantly larger than females (Calf sex: 2.37 kg, P , 0.01, Table 1C). Calf body mass was negatively affected by density in both populations (main effect of density: À12.79 kg, P , 0.01, Table 1C and Fig. 3C). The effect of spring onset on calf body mass was not significant. There was no evidence for an influence of supplementary feeding on the effect of spring onset and density on calf body mass.

Reproductive success
The supplementary fed population had a higher reproductive success than the control one (probability of 0.91 vs. 0.85 of having a calf for treated vs. control females keeping all other predictors at their average values, P , 0.01; Table  3 and Fig. 3D). Reproductive success was negatively related to reindeer density (main effect of density: À0.27, P , 0.01, Table 3). As expected, reproductive success in the supplementary fed population was less affected by negative density-dependence than the control herd (Treatment 3 Density interaction: 0.29, P , 0.01, Table 3). The estimated effect of spring onset on reproductive success was also negative but non-significant (Table 3).

DISCUSSION
The present quasi-experimental study allowed us to assess the effects of supplementary feeding, onset of spring and density on reindeer fitnessrelated traits. Our results support the hypothesis that late winter conditions are of critical importance for reindeer and that supplementary winter feeding acts as a buffer against negative environmental conditions and high population density since: (1) supplementary feeding had a positive effect on both population growth rates and carrying capacity, calf body mass and reproductive success; (2) variation in onset of spring had a stronger effect on population growth rates, body masses and reproductive success in the control population than in the supplementary fed population; and (3) the negative effect of density on adult and calf body v www.esajournals.org mass as well as female reproductive success tended to be more pronounced in the control population.

Effects of supplementary feeding
Supplementary feeding in late winter/early spring did not significantly affect female body mass the subsequent autumn. The latter result could be due to quantitative and/or qualitative (i.e., duration) insufficiencies of the treatment during this period of high energy requirement. Although it would have been informative to conduct the supplementary feeding at other critical time periods (e.g., throughout autumn), the herding practices (i.e., gathering schedules of the animals) did not allow it. Our results are nonetheless in line with previous findings showing no difference in body condition (i.e., body fat) between supplementary fed and free-ranging elk in winter (Bailey 1999). Similarly, in reindeer, positive effects of winter supplementary feeding were reported on spring but not summer or autumn body masses (Bå rdsen et al. 2009).
We observed positive effects of supplementary feeding on growth rate, calf body mass and reproductive success. Following the risk sensitive reproductive strategy hypothesized in reindeer (e.g., Bå rdsen et al. 2008), we showed that insuring food predictability during late winter and early spring (i.e., prior to breeding) increases, on a long-term basis, the willingness of females to allocate resources into reproduction. Indeed, the experimentally improved food conditions during winter at the peak of nutritional demand in late gestation (Clutton-Brock et al. 1989) might allow females to increase their reproductive allocation in the fetus. As a result, this may lead to heavier newborns as birth mass is related to winter nutrition in many herbivores (Skogland 1983). Namely, Keech et al. (2000) have Note: Spring onset, density and harvest were centered, i.e., subtracting the average, to make the intercept biologically meaningful (it represents the predicted value for the animals belonging to the control group with a value for the average of these three predictors). shown that birth mass of moose calves was positively correlated to body fat reserves of their mothers in late winter. Individuals born under favorable conditions can in turn reproduce earlier, reach a higher body mass and have a higher reproductive success compared to individuals born under less favorable conditions (Gaillard et al. 2000, Bonenfant et al. 2009). Consequently, calves of mothers receiving supplementary feeding should benefit from greater   v www.esajournals.org maternal care via for example higher milk quantity and quality (richer fat content), as shown in mule deer (Tollefson et al. 2011) and elk (Cook et al. 2004), allowing them to maximize their growth (e.g., in red deer: Clutton-Brock et al. 1982;elk: Smith et al. 1997) and overwinter survival (e.g., in three species of mountain ungulates: Pettorelli et al. 2007).

Density-dependence
Density-dependent processes were the main factors negatively affecting population growth rate, adult and calf body mass as well as reproductive success. Our results are in accordance with previous findings where densitydependent food-limitation has been suggested as the main mechanism regulating reindeer populations in Fennoscandia (Skogland 1985, Tveraa et al. 2007). The combination of high densities and harsh winter conditions results in episodes of starvation affecting female and calf body condition and/or adult and neonatal survival (e.g., Solberg et al. 2001, Tveraa et al. 2003. Indeed, density-dependence may negatively affect female body mass via an increased food competition in summer and autumn preventing reindeer from acquiring the amount of fat reserves needed for the following winter (see also Bå rdsen et al. 2010, Simard et al. 2010, Bå rdsen and Tveraa 2012. Moreover, autumn body mass is known to act as an insurance against winter starvation (e.g., Fauchald et al. 2004a). Accordingly, high density may not only impair female body mass but also reproduction.
The food supplemented population tended to compensate, at least partially, the negative effect of density on reproductive success. However, we found a negative effect of density on reproductive success in the control population, which mirrors previous results from other populations (e.g., red deer: Coulson et al. 2000; white-tailed deer: Simard et al. 2010) where increasing density reduced female fecundity. Density-dependence experienced by individuals in early life may have long-term effects on life-histories and on population dynamics in general, through delayed effects on reproduction and reproductive strategies (Pettorelli et al. 2001, Gaillard et al. 2003, Rodriguez-Hidalgo et al. 2010. In fact, females with an offspring have to partition the amount of resources spent investing in their calf, which compromises their survival the next winter, and investing in their own mass gain, which will enhance their survival probability (Bå rdsen et al. 2011). As a result, we can expect a higher proportion of adult females with no calf at the onset of winter, in the control population in particular, allowing them to spare the energetic costs associated with lactation (Clutton-Brock et al. 1989) and to enter winter with higher body fat reserves compared to years with maternal commitments. Calf body mass was more affected by high density compared to adult female body mass. Indeed, juveniles are generally expected to be more vulnerable than adults to high population density, presumably affecting vegetation quantity and quality (Gaillard et al. 2000, Bonenfant et al. 2009), since offspring have to pay a direct cost of density-dependence through body mass loss early in life and an indirect cost through a decrease in maternal reproductive allocation in utero (Forchhammer et al. 2001).

Effects of spring onset
Foraging conditions in spring and early summer are important for variation in demographic traits and population dynamics of northern large herbivores (Herfindal et al. 2006). Indeed, this period of particularly high energy requirement (late gestation and lactation) corresponds to the peak of nutritional demand for female ungulates and will influence future maternal care and body condition (Clutton-Brock et al. 1989, Pettorelli et al. 2006). Here we report that control females increased their summer body mass in years with early onset of spring, whereas supplementary fed females did not respond to spring green-ups. Moreover, the fact that reproductive success was not significantly related to spring onset in both populations indicates that females did not adopt an optimal reproductive strategy to increase reproductive allocation when food was available earlier in the season for certain years. These results mirror findings from previous studies stating that female reindeer have adopted a risk sensitive reproductive allocation strategy (e.g., Bå rdsen et al. 2008).
Calf body mass is usually related to weather parameters which are known to influence ungulates through their effects on plant phenology and forage quality (e.g., Langvatn et al. 1996). However, our results show that calf body mass v www.esajournals.org during winter was not affected by the onset of spring in any of the populations, which contrasts with previous findings on roe deer (Pettorelli et al. 2006) and reindeer (Pettorelli et al. 2005a) where plant productivity in the spring had a marked effect on winter body mass for the youngest ones. A possible explanation could be that spring green-up was influenced by density, via habitat deterioration, hiding a potential positive effect of the onset of spring (see also Côté et al. 2004). Furthermore, as reported by Pettorelli et al. (2006) in roe deer, similar responses to spring green-up and density were found for both male and female calves in both herds (lower rank model, not presented) indicating no difference between the sexes.
In conclusion, our results suggest that supplementary feeding acts as a buffer against stochastic variation in environmental conditions since the control population experienced more constraints compared to the fed population. Moreover, we have shown that food supplementation has an effect on the population level since feeding: (1) tends to enhance the carrying capacity of populations; and (2) enhances both direct and indirect effects at the individual-level. Feeding directly affects body mass (calves) and reproductive success, and indirectly affects how density-dependence and vegetation green-up impacts individuals. Body mass of the fed animals was, however, negatively affected by increasing densities. This suggests that supplementary feeding in winter/spring does not compensate for density-dependent effects operating during summer. Therefore, one can expect that continued population growth in response to improved access to food, through e.g., man induced actions like supplementary feeding, will eventually increase vulnerability to density-independent climatic effects the following winter. This finding has important implications both for reindeer owners and for the future management of this socio-economic system. Table A2. The set of candidate models used in the analyses of female body mass (Table 1B). The best model presents the lowest Akaike's information criterion (AIC) value, a DAIC equal to zero and the highest AIC weight (wAIC). The predictors included in the models are marked with an ''X''.