1. Macroparasites may be a major factor shaping animal behaviour. Tundra ecosystems inhabited by caribou and reindeer (Rangifer tarandus) are known for large concentrations of ectoparasites including mosquitoes (Culicidae) and black flies (Simuliidae), as well as endoparasitic oestrid flies (Oestridae).
2. Increased intensity and duration of insect harassment because of climatic warming is hypothesized as a potential factor in recent declines of Rangifer across the circumpolar north. Although there is a well-observed relationship between insect harassment and caribou/reindeer behaviour, the influence of ecto- relative to endoparasitic species is unclear. Climatic changes may favour the activity patterns, distribution or abundance of certain insect species; thus, understanding differential effects on the behaviour of Rangifer is important.
3. We recorded caribou behaviour using group scan and focal sampling methods, while simultaneously trapping insects and recording weather conditions on the postcalving/summer range of the Bathurst barren-ground caribou herd in Northwest Territories and Nunavut, Canada, during 2007–2009.
4. We developed statistical model sets representing hypotheses about the effects of insects, weather, habitat/location, and date/time on caribou behaviour. We used multinomial logistic regression models to explore factors affecting the relative dominance of behaviour types within groups of caribou and fractional multinomial logistic regression models to determine factors influencing time allocation by individual caribou. We examined changes in feeding intensity using fractional logistic regression.
5. Relative dominance of insect avoidance behaviour within caribou groups and time allocation to insect avoidance by individual caribou increased when oestrid flies were present or black flies were active at moderate–high levels. Mosquito activity had relatively little effect on caribou behaviour. Time spent feeding was reduced by the greatest degree when all three insect types were present in combination. Feeding intensity was influenced to a greater extent by the accumulation of growing degree days over the course of the postcalving/summer season than by insect activity. Changes in Arctic systems that increase the activity/abundance of ecto- and endoparasites could have implications for the productivity of Rangifer populations.
To address these knowledge gaps, we systematically trapped insects and monitored weather conditions in the vicinity of caribou groups while concurrently recording caribou behaviour over the 24-h period on the postcalving/summer range of the Bathurst barren-ground caribou (Rangifer tarandus groenlandicus Linnaeus) herd in Northwest Territories and Nunavut, Canada. Our goal was to define fine-scale functional relationships between caribou behaviour, activity/abundance of parasitic flies and environmental/temporal variables (i.e. weather, habitat and time/date). Specific objectives were to determine effects of the different families of parasitic flies, levels of insect activity, weather, habitat and time on: (i) the relative dominance of types of behaviour within caribou groups, (ii) time allocation by individual caribou and (iii) feeding intensity.
Materials and methods
We monitored weather conditions, trapped insects and recorded behavioural observations for caribou of the Bathurst herd over intensive sampling sessions during the 2007–2009 postcalving/summer seasons (Fig. 1). Groups of caribou were located based on positions of satellite-collared females. We used portable weather and light meters to record environmental conditions in the vicinity of groups of caribou under observation (Kestrel 4500 on Kestrel Portable Vane Mount; Nielsen Kellerman, Boothwyn, PA, USA; EA30 light meter; Extech, Waltham, MA, USA; data-logging light meter; Sper Scientific, Scottsdale, AZ, USA). We collected insects using modified Malaise traps baited with carbon dioxide (Anderson, Nilssen & Hemmingsen 2001). Trap catch and weather data were averaged over 2-h intervals and linked to all caribou observations occurring within the interval. Weather stations and insect traps were generally located 25 m–2 km from caribou groups, although in some instances, caribou came within a few meters of the traps during the natural course of their movements.
We used both group scan and focal sampling methods to collect behavioural data (Altmann 1974). We observed caribou for the 24-h daily period using spotting scopes and classified behaviour as feeding, walking, running, lying, standing/other and insect avoidance. Insect avoidance was considered as a hierarchical behaviour following the classification of one of the other five behaviour types. For focal sampling, feeding was further broken down into eating and searching (modified from Russell, Martell & Nixon 1993; Griffith et al. 2001). Eating involved ingesting forage, while searching encompassed times when caribou had their muzzle near the ground but did not ingest vegetation. Feeding intensity was calculated as the ratio of time eating to total time eating and searching (Griffith et al. 2001). Behaviours identified as insect avoidance included: ear flicking, tail wagging, head tossing, body shaking, foot stamping, biting, sneezing, kicking, rearing, bucking, alarm posture, rapid erratic running and ‘animal stance’ (stationary, with the head touching or close to the ground, and remaining motionless for a period of time) (Downes, Theberge & Smith 1986). Although increases in movement in the absence of the specific avoidance behaviours mentioned above could also be responses to insect harassment, these behaviours were simply classified as walking or running.
We performed instantaneous scan sampling of a randomly selected group of caribou c. every 30 min. We estimated group size to the nearest 100 for groups of 250–1000 animals and the nearest 1000 for groups >1000 animals. For groups of <250 animals, we recorded the number of caribou per group engaged in each of the behaviour categories. When visibility or movement made this prohibitive, and for group sizes >250, we noted the approximate percentage of animals engaged in the behaviours.
We performed focal sampling on randomly selected adult female caribou for intervals of up to 30 min, during which we recorded the behavioural sequence and amount of time spent engaged in each behaviour type (HP 100 LX Palmtop PC; Hewlett Packard, Palo Alto, CA, USA; Gillingham 2008). For both group scan and focal sampling, we recorded date, time and location. We noted when caribou reacted to human observers, other human disturbance, or predators, and we excluded those observations from further analysis.
We examined the effects of insect harassment, weather, time, date and habitat on three aspects of caribou behaviour: relative dominance of behaviour type within caribou groups, time allocation of individual caribou and feeding intensity of individual caribou. The goal of the first analysis was to understand factors affecting increases or decreases in less prevalent behaviours of caribou, such as insect avoidance. For each group scan, we assigned a percentile value to each of the six behaviours (feeding, lying, standing, walking, running, insect avoidance) by comparing the percentage of the caribou group engaged in a given behaviour to the observed range of engagement in the behaviour across all scans from the 2007–2009 data (Appendix S1). The percentile values of the six behaviours were ranked within each group scan, and the behaviour with the highest value was scored as the relatively most dominant. We chose this approach because feeding, walking and lying tend to dominate caribou activity budgets (Roby 1978; Downes, Theberge & Smith 1986; Colman et al. 2003). Rare behaviours with a low absolute prevalence may be masked by more common behaviours even when biologically important changes from typical activity budgets occur. We used multinomial logistic regression (mlogit; Long & Freese 2001) to model the effects of biotic and abiotic variables on the relative dominance of behaviour types. Each binary comparison within the mlogit examined the effect of environmental variables on the probability of a particular behaviour exhibiting relative dominance compared to another.
In the second analysis, we examined factors affecting time allocation by individual caribou using data from focal sampling. We used fractional multinomial logistic regression (fmlogit; Buis 2008) to model behaviour choices while recognizing time as a limited resource (Ye & Pendyala 2005). Fmlogit models examined the effects of environmental variables on a caribou’s proportional allocation of time to each type of behaviour relative to other behaviours for all possible comparisons.
Finally, we used fractional logistic regression (flogit; Papke & Wooldridge 1996) to explore the effects of environmental variables on feeding intensity. This analysis was restricted to 2008–2009 focal samples; feeding intensity data were not recorded in 2007. In all three modelling approaches, we used a robust clustering technique to account for potential autocorrelation among behavioural observations at a given site (Nielsen et al. 2002).
We developed explanatory models that included variables from three sets of factors thought to influence caribou behaviour: weather/insects, habitat/location and time of day/year (Appendix S2). Weather-related variables included temperature, wind speed and light intensity. We also modelled variables for activity/abundance of mosquitoes and black flies, and presence/absence of oestrid flies (both nose-bot, Cephenemyia trompe Modeer, and warble flies, Hypoderma tarandi Linnaeus). Habitat-related variables included vegetation type and location on the Bathurst range. We used sunrise/set times (National Research Council Canada 2009) to create three categories (morning, afternoon and dawn/dusk/night) representing time of day. We parameterized categorical variables using deviation coding to contrast the effect of each level against the overall effect of the categorical variable (Menard 2001).
As a measure of seasonality, we used a growing degree day (gdd) variable based on temperature accumulation above 0 °C beginning at the snow-free date (Karlsen, Elvebakk & Johansen 2005). Growing degree days reflected conditions relevant to the developmental biology of green vegetation, allowing us to explore whether caribou may have modified their behaviour as vegetative phenology changed over the course of the postcalving/summer season. We included a variable for year to determine whether there were annual variations in caribou behaviour not captured by other environmental variables. All models of relative dominance of behaviour from scan samples contained a group-size variable to account for variations in behaviour because of the number of caribou in a group. We did not use focal observations <60 s in length and included a variable representing duration of the focal observation in all models of individual behaviour and feeding intensity to correct for biases in this sampling method. We used variance inflation factors to assess collinearity among independent variables (Menard 2001).
Model selection and predictive ability
We used a priori knowledge to develop model sets representing biologically plausible hypotheses. For all three behavioural analyses, we classified models according to: insect/weather, time/date/habitat, and combinations of insects and time/date/habitat. Models within the insect/weather theme tested the effects of mosquitoes, black flies and oestrids on caribou behaviour. In the time/date/habitat theme, we created models to determine the influence of circadian and annual cycles, as well as habitat type, on caribou behaviour. We organized models within the insect/weather and time/date/habitat themes to test whether one of these two broad themes might be a predominant driver of caribou behaviour as compared to the other. An additional hypothesis was that the indirect effects of weather on insect activity would have a greater influence on caribou behaviour than direct effects of meteorological variables. We developed a third ‘combination’ theme of models including explanatory variables related to insect activity and time/date/habitat. We made comparisons among models in each of the themes, but this does not imply that we captured the full range of model possibilities.
We based model selection on an information theoretic model comparison approach using Akaike’s Information Criteria for small sample sizes (AICc) and Akaike weights (w) to select the most parsimonious model (Anderson, Burnham & Thompson 2000). We interpreted w as approximating the probability that a given model was the best within a model set. When two or more top models had a difference in AICc < 2, we considered these models to be of near equal parsimony (Burnham & Anderson 1998). For the best models, we generated β-coefficients and 95% confidence intervals for each parameter. We calculated Pearson’s standardized residuals to determine the difference between observed and predicted values. We used the area under the receiver operating characteristic (ROC) curve (AUC) and a withheld data set of 20% of the observations to evaluate the predictive ability of mlogit models (Swets 1988). For flogit and fmlogit, we used Spearman’s correlations to assess the relationship between the withheld observations and predicted proportions.
We performed a total of 198 scans in 2007, 450 scans in 2008 and 257 scans in 2009 on groups containing one to >500 caribou (219 ± 532 SD). During focal sampling, we observed 271 (cumulative observation time 2614 min), 257 (2214 min) and 172 (1689 min) individual caribou in 2007, 2008 and 2009, respectively. Focal observations ranged from one to 30 min in length (9·7 min ± 7·8 SD). In all years, feeding and walking were the most common behaviours (Fig. 2).
Relative dominance of behaviour within caribou groups
Two models of relative dominance of behaviour within caribou groups had nearly identical AICc scores (Table 1, Appendix S3). These models were from the combination theme and contained covariates for insect activity levels, time, easting, northing, gdd, year and group size. Predictive abilities of the top models were reasonable to good: using independent data, ROC scores for the set of binary logistic regressions representing all possible comparisons of relative dominance between behaviours ranged from 0·62 to 1·00 for the top model, and 0·70–1·00 for the second-ranked model. Both models clearly identified when insect avoidance and running were the dominant behaviours (AUC > 0·90). The models were least predictive when distinguishing between feeding and walking (AUC: 0·62–0·74), and feeding relative to lying (AUC: 0·70–0·72). In all cases except the comparison between feeding and walking in the top model (AUC = 0·62), these ROC scores were still considered ‘reasonable’ (Swets 1988).
Table 1. Top multinomial logistic regression models of relative dominance of behaviour type within caribou groups on the Bathurst caribou postcalving/summer range, 2007–2009
Number of parameters (K), log-likelihood, Akaike’s Information Criterion (AICc) scores, differences in AICc scores (ΔAICc) and AICc weights (w) for the best models in each theme (insect/weather, time/date/habitat, and combination). Gdd (growing degree days).
Mosquito + black fly + oestrid + group size
Black fly + oestrid + group size
Vegetation + easting + northing + time + gdd + year + group size
Vegetation + easting + northing + group size
Black fly + oestrid + time + easting + northing + gdd + year + group size
Mosquito + oestrid + time + easting + northing + gdd + year + group size
Mosquito + black fly + oestrid + time + easting + northing + gdd + year + group size
Effects of insect activity on the relative dominance of behaviour within caribou groups differed depending on the family and activity level of insects (Figs 3 and 4; Appendix S4). Oestrid flies had the largest effect on caribou behaviour. The likelihood of dominance of insect avoidance increased relative to all other behaviours when oestrids were present; the relationships relative to lying and walking were significant.
Relative dominance of behaviours varied depending on time of day (Figs 3 and 4; Appendix S4). The likelihood of insect avoidance becoming the dominant behaviour increased during morning. Feeding and walking were other behaviours likely to dominate in morning. Insect avoidance and standing were most likely to dominate during afternoon. During dawn/dusk/night, the probability of insect avoidance dominating decreased significantly relative to all other behaviours. Lying was the behaviour most likely to dominate at this time.
Time allocation by individual caribou
The two top models describing time allocated by individual caribou to behavioural classes differed by only 0·90 points (Table 2, Appendix S5). These models were from the insect/weather theme and contained covariates for insect activity levels and duration of focal observations. For both models, Spearman’s correlations indicated weak to moderate (rs = 0·13–0·54; P ≤ 0·05) correlations between the observed and predicted proportions of time allocated to the six behaviours. The models were most successful at predicting proportions of time caribou spent engaged in insect avoidance and lying. Using independent data, the correlation between observed and predicted proportion of time devoted to insect avoidance was 0·54 in the top model and 0·38 in the second-ranked model. Correlation coefficients for lying were 0·44 and 0·45, respectively. The models had difficulty predicting proportions of time caribou spent feeding and running. Correlation coefficients for feeding were 0·20 and 0·13 and for running were 0·14 and 0·13, for the top and second-ranked models. Residual analysis indicated that both models performed poorly when caribou engaged in a single behaviour for the majority of a focal observation; all models underpredicted the proportion of time caribou engaged in the predominant behaviour.
Table 2. Top fractional multinomial logistic regression models of time allocation of individual caribou from focal sampling on the Bathurst caribou postcalving/summer range, 2007–2009
Number of parameters (K), log-likelihood, Akaike’s Information Criterion (AICc) scores, differences in AICc scores (ΔAICc) and AICc weights (w) for the best models in each theme (insect/weather, time/date/habitat and combination). Gdd (growing degree days).
Black fly + oestrid + duration
Mosquito + oestrid + duration
Mosquito + black fly + oestrid + duration
Vegetation + easting + northing + duration
Year + duration
Oestrid + time + easting + northing + gdd + vegetation + year + duration
Mosquitoes had a variable effect on the behaviour of caribou (Figs 5 and 6, Appendix S6). Counterintuitively, time spent in insect avoidance decreased relative to all other behaviours when mosquito activity was high. Time allocated to insect avoidance increased when black flies were present at moderate to high levels (Figs 6 and 7, Appendix S6). At moderate black fly activity, this increase was significant relative to feeding, lying and standing. At high black fly activity, insect avoidance increased significantly relative to all behaviours except running. When oestrid flies were present, the proportion of time caribou spent engaged in insect avoidance increased significantly relative to feeding, lying, walking and running; standing also increased relative to all behaviours except insect avoidance (Figs 5–8; Appendix S6).
Feeding intensity of individual caribou was best explained by a model from the time/date/habitat theme that contained covariates for year, gdd and duration (Table 3, Appendix S7). Predictive ability of the top model was fair. Using withheld data, we found a moderate positive correlation (rs = 0·44, P < 0·001) between observed and predicted records. The most parsimonious model suggested that feeding intensity decreased in 2009 relative to 2008 (β = −0·644; 95% CI: −1·064 to −0·225), decreased as gdd accumulated over the course of the summer (β = −0·005; 95% CI: −0·008 to −0·002) and was higher in focal samples of longer duration (β = 0·001; 95% CI: 0·0002–0·001).
Table 3. Top fractional logistic regression models of feeding intensity of individual caribou from focal sampling on the Bathurst caribou postcalving/summer range, 2008–2009
Number of parameters (K), log-likelihood, Akaike’s Information Criterion (AICc) scores, differences in AICc scores (ΔAICc) and AICc weights (w) for the two best models per theme (insect/weather, time/date/habitat and combination). Gdd (growing degree days).
Insect avoidance behaviours by Rangifer are widely reported (Pruitt 1960; Kelsall 1968; Russell, Martell & Nixon 1993; Hagemoen & Reimers 2002); however, in most cases, the effects of different parasitic fly families have not been clearly differentiated. Understanding the degree to which different insects affect Rangifer behaviour is particularly important as parasitic flies may exhibit dissimilar responses to climate change. As summer temperatures rise, the activity levels of black flies and oestrids are predicted to increase on the Bathurst postcalving/summer range (Witter 2010). Although temperature positively affects mosquitoes, mosquito activity may ultimately decline because of changes in other meteorological variables such as wind speed and relative humidity (Witter 2010).
The presence of even a single oestrid fly caused larger and more consistent behavioural responses by caribou than either mosquito or black fly activity. This supported reports of alertness and stress spreading through caribou herds when only a few individual caribou were directly attacked by oestrids (Roby 1978; Hagemoen & Reimers 2002). We found that both the relative dominance within caribou groups and the proportion of time individuals spent in insect avoidance and standing behaviours increased when oestrids were present. These behaviours increased at the expense of lying, feeding, walking and running; however, lying was reduced to a greater degree than the other behaviours. This trade-off was also observed for caribou in Alaska and reindeer in Norway (Russell, Martell & Nixon 1993; Hagemoen & Reimers 2002).
There is little documentation on the effects of black flies on caribou behaviour. In the few cases where black flies were observed on Rangifer postcalving/summer ranges, they were present in low numbers, assumed to have little effect and/or not considered separately from mosquitoes (Roby 1978; Anderson, Nilssen & Hemmingsen 2001; Hagemoen & Reimers 2002). We found, however, that caribou increased time allocation towards insect avoidance and running when black flies were active at moderate to high levels. These effects were notable even though the absolute number of black flies was relatively low. This stands in contrast to mosquitoes which were active in much greater numbers but had noticeably smaller effects on caribou.
Caribou exhibited little increase in stereotypical insect avoidance behaviours when mosquitoes were active. When caribou reacted to mosquitoes, it was by increasing time spent walking. Although this response was weak, it corresponded to reports of increased walking and increased rate of movement by reindeer in Norway and caribou in Alaska when mosquitoes were present (Dau 1986; Mörschel & Klein 1997; Hagemoen & Reimers 2002). While increases in walking may have energetic implications for caribou, mosquitoes did not seem to be a major stressor of Bathurst caribou on the postcalving/summer range at the activity levels observed during this study. Mosquito activity/abundance, however, may vary across the ranges of different herds, and harassment could be a larger factor in postcalving/summer-range ecology in other areas.
The behavioural responses of caribou to particular insect families do not occur in isolation. Alterations in activity budgets are likely magnified when some combination of mosquitoes, black flies and oestrids is present concurrently. On the Bathurst range, mosquito activity peaked earlier in the season when compared to black flies and oestrids (Witter 2010). Mosquito activity also increased during dawn, dusk and night, whereas black flies and oestrids were more active during morning and afternoon hours. The relationships we observed between dominance of behaviour within caribou groups and time of day were likely related to these diel patterns of insect activity. Other studies have also reported more severe Rangifer responses to insect harassment during mid-day (Roby 1978; Anderson & Nilssen 1998; Anderson, Nilssen & Hemmingsen 2001). The relative dominance of feeding did not increase noticeably at dawn/dusk/night, suggesting caribou may not have used this time period to compensate for lost foraging opportunities during morning and afternoon. Colman et al. (2003) also failed to observe grazing compensation at night. Whether or not compensation is necessary for Rangifer to maintain adequate forage intake during periods of insect harassment may depend on a variety of factors including reproductive status and the quality and quantity of available forage (Downes, Theberge & Smith 1986; Fancy 1986; Colman et al. 2003). The relative dominance of lying did increase during dawn/dusk/night. This has been observed in other herds (Colman et al. 2001; Loe et al. 2007) and could be due to decreased levels of insect harassment or to intrinsic physiological cues (Colman et al. 2001).
There is some debate over the degree to which weather conditions directly affect caribou/reindeer as opposed to indirect effects via the influence of weather on insect activity (Downes, Theberge & Smith 1986; Mörschel & Klein 1997; Anderson & Nilssen 1998; Skarin et al. 2004). In our analysis, models of caribou behaviour that contained weather variables did not perform as well as those containing covariates related to insect activity. This suggests that the indirect effects were larger than direct effects of weather on caribou behaviour. Other studies have also observed little response by Rangifer to weather variables such as temperature, light and precipitation in the absence of parasitic insects (Hagemoen & Reimers 2002). In some instances, however, caribou/reindeer were found to respond directly to high temperatures by decreasing feeding time (Mörschel & Klein 1997) or altering habitat use by moving to snow patches or higher elevations for thermoregulation (Downes, Theberge & Smith 1986; Anderson & Nilssen 1998; Skarin et al. 2004).
Insect activity levels changed over the course of the summer (Witter 2010), and we also observed trends in caribou behaviour as gdd accumulated over time. Predominance of lying, feeding and insect avoidance increased, while walking and running decreased over the postcalving/summer season. Increased insect avoidance could have been due to greater black fly and oestrid activity in mid-to-late summer. The trend towards decreased movement and increased lying and feeding may have been due to variation in factors such as forage quality that were not included in our study. Feeding intensity also declined as gdd accrued, likely also due to factors distinct from insect harassment. Models of feeding intensity that contained covariates representing insect activity did not score as well during model selection as those containing temporal covariates. Although both the predominance of and time allocation towards feeding decreased when oestrids were present or black fly activity was moderate–high, changes in feeding intensity did not appear as drastic. The accumulation of gdd should reflect changes in vegetative phenology throughout the postcalving/summer season. As vegetation senesces later in the summer, caribou may spend more time searching out remaining patches of new green vegetation that is higher in nutrients and lower in fibre and phenolic content (Kelsall 1968; Skoog 1968; White et al. 1975; Russell, Martell & Nixon 1993); thus, feeding intensity might decline. A shift to forage of lower quality or higher fibre content might also explain the increased dominance of lying later in the summer as longer rumination bouts may be required to facilitate digestion (White et al. 1975; Trudell & White 1981; Robbins 1993).
Through the combination of rigorous methods for insect collection and observation of caribou behaviour, we refined the knowledge of Rangifer response to insect harassment. Field-based behavioural studies are difficult (Altmann 1974; Martin & Bateson 1993), however, and interpretation of results and extrapolation to other herds requires caution. There were times when caribou may have reacted to oestrid flies active in the environment, but traps did not record oestrid presence. This could have led to errors of omission and reduced the predictive ability of behavioural models. Additionally, some types of behaviour affect the duration or number of individuals observed (Altmann 1974; Fragaszy, Boinski & Whipple 1992). Caribou lying down may be less visible within groups, and predominance of lying could have been underrepresented in group scan observations. Furthermore, caribou walking or running were difficult to observe for extended durations.
Variations in the quality and quantity of forage are likely major determinants of caribou behaviour and habitat use during the postcalving/summer season (Roby 1978; Russell, Martell & Nixon 1993; Skarin et al. 2008). Poor range quality and severe insect harassment may have cumulative effects on caribou behaviour and body condition, while good forage conditions could mitigate the negative effects of parasitic insects. The lack of information on forage availability and quality, variation in forage conditions and diet composition represents a critical gap in our understanding of the Bathurst herd’s postcalving/summer-range ecology.
Differences in forage quality/quantity, availability of insect relief terrain and identities of parasitic species present mean that each Rangifer herd has a unique set of circumstances driving postcalving/summer-range dynamics. The relative severity of response to mosquitoes, black flies and oestrids, however, broadly illustrates the type of behavioural trade-offs faced by caribou experiencing insect harassment and may be applicable to Rangifer in other areas. We documented increases in the time spent by caribou in insect avoidance and other energetically costly behaviours in response to harassment by black flies and oestrids. Climatic warming to date has increased the proportion of the postcalving/summer season during which conditions are favourable for black fly and oestrid activity on the Bathurst range (Witter 2010). Increased understanding of behavioural reactions can be paired with ongoing body condition and population monitoring (Adamczewski et al. 2009), as well as energetics modelling (Fancy 1986; Russell, White & Daniel 2005), to illuminate potential population-level responses of Rangifer to changing levels of insect activity/abundance in the context of climatic variation and industrial development in the circumpolar north. When combined with other stressors, behavioural modifications in response to insect harassment could drive Rangifer into a negative energy balance during the postcalving/summer season, to the detriment of population productivity (Fancy 1986; Russell, Martell & Nixon 1993).
Behavioural choices made by individuals have repercussions for survival and reproduction that ultimately translate into consequences at the level of populations (Clutton-Brock, Guiness & Albon 1982; Richner 1998; Rubenstein 1998; Namgail, Fox & Bhatnagar 2007). The methods presented here for exploring trade-offs may be relevant to a variety of questions in behavioural ecology. In particular, these methods provide robust treatment of complex species-response data that allow one to explore the biological relevancy of hypotheses that identify specific factors that may influence animal behaviour. Information theoretic approaches in combination with detailed process-driven generalized linear models and corresponding measures of the predictive relevancy of results are adept at representing complex behavioural responses to numerous biotic and abiotic factors, as we observed. Using such approaches to better understand the influences of human disturbances and environmental variation on animal behaviour will become increasingly important to the development of effective conservation and management strategies in the context of global change.
Our research was financed by the CircumArctic Rangifer Monitoring and Assessment Network (CARMA), Government of the Northwest Territories Department of Environment and Natural Resources (GNWT ENR), Natural Sciences and Engineering Research Council of Canada (NSERC), and University of Northern British Columbia (UNBC). Logistical support was provided by Jan Adamczewski, Kirstin Mahler and Steve Matthews. Thanks to Shawn Ranahan and Sage Suzuki for their skill and patience during flights to locate caribou. Tess Alain, Dave Dewar and Kirstin Mahler assisted in the field. Lisa Poirier and Don Russell provided comments on earlier drafts.