How does local weather predict red deer home range size at different temporal scales?


Correspondence author. E-mail:


1. There is a rapidly growing literature on how climate affects populations of vertebrates. For large herbivorous mammals, most attention has been paid to demographic responses to climate variation. Much less information is available regarding how climate affects animal behaviour, i.e. the climate mechanisms. Further, the appropriate measurement scale of climate variables remains debated. Here, we investigate how local climate variables determine home range sizes at four temporal scales using the Börger-method on GPS telemetry data from 47 female red deer Cervus elaphus L. in Norway.

2. If local climate operates directly on the immediate activity level of the animal, we predict home range sizes to show season-specific variation on short temporal scale (weekly-daily) related to temperature and precipitation. If local climate operate indirectly through plant growth, we rather predict variation in home range sizes to be apparent on longer time scales (biweekly-monthly), and during summer only.

3. At all time scales home range size was positively correlated with temperature during winter and negatively during summer, while the effect of precipitation was season- and scale-specific, except when accumulating as snow. Extensive snow cover decreased home range size, indicating direct effects of climate.

4. The effects of local climate was weaker at the shortest time scales (weekly-daily) compared to the longest time scales (monthly-biweekly), while the effects of day length on home range size was only apparent on the monthly and daily scale. At the longest time scales variation in local climate had a large effect on home range size. This is consistent with climatic variables operating indirectly through plant growth, but we cannot exclude a certain direct effect even at longer time scales.

5. We show how local climate-home range size correlations measured over different temporal scales can be used to infer direct and indirect climate mechanisms. Insight on the behavioural basis of responses to climate enables more accurate predictions of possible nonlinear relationships to future global warming.


The impacts of climate on vertebrate populations are well documented across marine and terrestrial ecosystems. Most attention has been paid to quantifying demographic responses (e.g. Grosbois et al. 2008), as these determine population dynamics and therefore whether or not species are able to persist or not under climate change. However, the study of the timing of reproduction in spring of predator and prey (match-mismatch hypothesis) has revealed the importance of understanding in detail the proximate climate links to successfully predict (nonlinear) population responses (Stenseth & Mysterud 2002; Visser & Both 2005). Studies of large mammals have revealed strong direct effects of winter weather on survival (Coulson et al. 2001), as well as indirect effects operating through plant quality on individual growth (Mysterud et al. 2001; Pettorelli et al. 2005). Behavioural responses of ungulates to adverse weather include seeking forest cover for relief (Moen 1976), reducing activity (Beier & McCullough 1990), migrating to lower elevations to avoid deep snow (Mysterud 1999) and range displacement during icing events (Stien et al. 2010), but studies of local climate effects on movements are few and typically scale-specific. Indeed, animal needs may differ profoundly depending on the size of the temporal scale used when estimating and investigating the behavioural response (Börger et al. 2006b).

Spatial movement patterns in mammals are closely linked to energetic requirements (Ford 1983; Tufto, Andersen & Linnell 1996). Across species, home range sizes typically increase with body size (McNab 1963; Harestad & Bunnell 1979). Within species, home range size may decrease with habitat productivity (Kie et al. 2002; Anderson et al. 2005; Börger et al. 2006b) and show large seasonal variation (Georgii 1980; Georgii & Schroder 1983; Börger et al. 2006b). Apart from the well studied seasonal effects on home range, the more short-term, less predictable changes in climate may require individuals to regulate their home range use due to thermal stress and influence on forage availability and energy requirements (Parker, Robbins & Hanley 1984; Van Soest 1994; Börger et al. 2006b). We still have a limited theoretical framework to predict home range-climate responses to movements at different temporal scales. Indeed, studies investigating and comparing factors influencing intraspecific variation in home range size at different temporal scales are scarce (Börger et al. 2006b). Using the Börger-method (a mixed-model approach; Börger et al. 2006b) and red deer Cervus elaphus L. as model species, we aim to quantify spatial and temporal factors influencing home range size at different temporal scales through the year. This method allows the local climatic variables temperature, snow depth and precipitation to be included in the models as deviations (residuals) from mean values for a specific time of year. Thus the effect of climate leaves only the unpredictable environmental changes in the models. We tested the following hypotheses and predictions:

H1. Animals are expected to reduce activity (and movement) at high temperatures during summer and low temperatures during winter (e.g. Beier & McCullough 1990). We therefore predict home ranges sizes to (H1a) increase with temperature during winter and (H1b) decrease with temperature during summer, with the largest effects at short temporal scales. At longer time scales, seasonal variation in pelage insulation is expected to reduce correlations.

H2. Precipitation is expected to have direct negative effects on activity and thereby movement at short temporal scales, predicting a negative correlation with home range size (Parker 1988). During winter, snow depth is expected to reduce home range size, but likely at longer time scales than precipitation, as snow cover often persist for an extended period of time.

H3. Indirect factors like plant growth are likely to influence home ranges estimated on long temporal scales, but not short scales. We therefore take a correlation between temperature (or derivations such as growing degree days) and precipitation during summer and home range size measured over longer time scales as evidence of indirect effects of climate (H3a). Light levels may be viewed as a crude estimate of plant productivity, and a correlation between the number of daylight hours and home range size is also interpreted as evidence of indirect effects of climate (H3b).

H4. Home range sizes in animals are known to be related to energetic requirements (McNab 1963; Harestad & Bunnell 1979). At all temporal scales we expect home range size to vary depending on the dominant habitat type within the home range. Individuals with less forage-rich habitat types dominating their home range should have larger home ranges than individuals with forage-rich habitats dominating their home range, as they are able to cover their nutritional needs within a smaller area.

Materials and methods

Study area

The study was conducted in the county of Sogn og Fjordane, Norway (Supporting Information, Fig. S1). The vegetation on the west coast of Norway is mostly in the boreonemoral zone. Natural forests are dominated by deciduous (mainly birch Betula spp. L. and alder Alnus incana L.) and pine forest (Pinus sylvestris L.), with juniper (Juniperus communis L.), bilberry (Vaccinium myrtillus L.) and heather (Calluna vulgaris L.). Norway spruce Picea abies L. has been planted on a large scale, and is an important winter habitat for red deer. On flatter and more fertile grounds in the bottom of valleys, areas have been cleared for cultivation. These areas are used mostly as pastures and meadows for grass production. The topography is mountainous, and is characterized by steep hills, valleys, streams and fiords, and the slope increase from coast to inland. In general, temperature and precipitation decline from coast to inland, while snow depth and duration of snow cover increases (Mysterud et al. 2000).

Red deer data

Data derive from 47 female red deer caught by darting on winter feeding sites during 2005–2007, after a procedure approved by the Norwegian national ethical board for science (‘Forsøksdyrutvalget’, The deer were fitted with Televilt Basic ‘store-on-board’ GPS (Global Positioning System) collars or Televilt Basic GPS collars with GSM (Global System for Mobile Communications) option for transfer of data via cell phone network (Televilt TVP Positioning AB, Lindesberg, Sweden). The GPS collars were programmed to record hourly positions, and to release a drop-off mechanism after approximately 10 months (tracking period ranged from 6 to 12 months). Locations taken during the first 24 hours of tracking and all locations where the animal had moved at a speed of >40 km per hour or a distance >10 km between fixes were removed, as these are most likely erroneous. In total, only 0·5% of the locations were removed as outliers. As some locations (very few) were preceded and succeeded by missing locations, the distance or speed rule was not able to identify these as outliers. To remove these we performed a visual inspection of locations from each individual, to see if any obvious outliers were still retained (N = 27, 0·000098 of locations). If ever in doubt to remove a location based on visual inspection, the location was retained. As the home range concept applies only to stationary space use patterns, and also because migration behaviour is very different from the normal home range behaviour, we removed all locations within and partly overlapping with the migration periods. Deer in our area either stay year round in one area or migrate quickly in spring and autumn between clearly separated seasonal home ranges, with intermediate strategies being rare. We plotted for each individual the distance from the first location by time (Julian date). Individuals displaying distinct migration periods in this plot were classified as migratory, and the rest were classified as stationary (see Supporting Information, Fig. S2). We used piecewise regression (library ‘segmented’ in the statistical software R; Muggeo 2008; R Development Core Team. 2008) to identify migration periods. The aim of the method is applying least-squares methods to identify breakpoints by fitting broken-line relationships between the response and predictor variable. Using Julian date as predictor and distance from first location as the response, we identified when migration started and ended in spring and autumn for each migratory individual and removed data from these periods.

Error estimation

Location error – defined as the distance between estimated and true position – is a common source of error for GPS data. To correct for this, we used the method presented by Börger et al. (2006b). Details on this can be found in the Supporting Information.

Local climate variables

Data on temperature, precipitation and snow depth were taken from meteorological stations located within the study area ( Temperature was available from seven stations, and precipitation and snow depth from 18 stations (Supporting Information, Fig. S1). For each GPS location recorded from the animals, the closest meteorological station was identified and the daily precipitation, snow depth and temperature value from this station was assigned to the location. We calculated growing degree days (GDD) from hourly temperature measures, which were available from five of the seven temperature stations, as it is expected to be a more direct determinant of plant productivity (Woodward 1987). We estimated GDD according to Molau & Mølgaard (1996), by summing up hourly temperatures for each day when the temperature was above a threshold of 5 °C, and dividing the sum by 24. Day lengths in the study area were downloaded from the U.S. Naval Observatory (

Habitat data

We evaluated the availability of four different habitat types within the red deer home ranges. The dominant habitat type within each home range was determined by using digital land resource maps provided by the Norwegian Forest and Landscape Institute, with scale 1:5000 and a resolution of 50 × 50 m. The digital resource maps were divided into five habitat types, by merging of habitat classes from the original maps (as in Godvik et al. 2009). ‘Forest of high productivity’ and ‘pastures’ are considered forage-rich, and ‘forest of low productivity’ and areas deficient in shelter and forage plants (mainly mountains and marshes, defined as ‘other’ in figures and tables; Godvik et al. 2009) are considered forage-poor. The fifth category contained lakes, sea and uncharted areas, and was thus not of interest. In any case that this was the dominant habitat type, we used the second most abundant habitat type in the analysis.

Home range estimation

To be able to evaluate and compare the results, we estimated home ranges by two different methods. The methods were fixed-kernel (the currently most popular method) using the reference method for calculation of the smoothing factor h (Kernohan, Gitzen & Millspaugh 2001) and minimum convex polygon (MCP, traditionally the most used method). MCP home ranges were estimated using the library ‘adehabitat’ (Calenge 2006) implemented in the statistical software R version 2.10.1 (R Development Core Team. 2009), and kernel home ranges were estimated using the Animal Movement extension in ArcView GIS 3.3 (ESRI, USA). We only estimated home ranges if the individual had at least 95% coverage of the given time interval, and at least 16 GPS locations. Home ranges were estimated at four different temporal scales; daily, weekly, two weeks (biweekly) and monthly. To investigate if the effect of a factor varied for different home range density isopleths (i.e. affected ‘core’ area more than full home range), we always estimated the 90%, 70% and 50% home range for each individual.

Statistical analysis

To examine variation in home range size we used the method developed by Börger et al. (2006b). We fitted linear mixed-effects models (Pinheiro & Bates 2004), using the library ‘nlme’ (Pinheiro et al. 2009) implemented in R (R Development Core Team. 2009). The response variable in the model was log-transformed home range size (ha), and only individuals with repeated home range estimates during the particular temporal scale were used in the analysis. Five covariates were fitted as fixed effects; temperature, precipitation, day length, dominant habitat type and spatial GPS error (Supporting Information, Table S1), including all combinations of two-way interactions. The climatic variables temperature and precipitation were fitted as residuals obtained from a linear regression against day length to remove the seasonal trend. We fitted separate models for the winter season, defined as 1st December–31st March, where snow depth (residuals obtained in the same way as above) was included as a covariate. We fitted separate models for summer, where growing degree days (GDD) was included as a covariate. Summer was defined differently for stationary and migratory individuals. For stationary individuals summer was defined as 1st May–31st August. If migratory individuals started spring migration after 1st May, the date of arrival was used, and the start date of autumn migration defined the end of summer if this was before 31st August.

As random terms we fitted random intercepts for each individual. When dealing with a random sample of individuals with varying numbers of corresponding home range estimates, coefficient estimates will be biased towards individuals with the largest amount of estimates and/or individuals with extreme home range sizes (Follmann & Lambert 1989; Pinheiro & Bates 2004). If a random intercept is needed, this implies variation in home range size within individuals is smaller than between individuals. The random intercept added to the model accounts for this individual variability by adjusting the overall average home range size, allowing inference to be extended to the entire population (Neter et al. 1996).

The models were checked for unequal variance structures of the within-group errors by investigation of relevant model diagnostic plots (plots of residuals vs. fitted values for the relevant model and variable; Pinheiro & Bates 2004) and by comparing models with and without different variance functions, using likelihood ratio tests. If selected, we implemented variance functions in the models, as according to Pinheiro & Bates (2004). The variance functions tested for were either a general function specifying the fitted values as variance covariates, a function with different variance parameters for each level of the variable specifying the dominant habitat type, or with different variance parameters for migratory and stationary individuals. These were fitted using the ‘varPower’-function in library ‘nlme’, which models cases where there is an increase or decrease in variance with the absolute value of the covariate (Pinheiro et al. 2009). We also checked for any remaining dependencies among the within-group errors after the fixed and random effects were fitted. If present, these were modelled using correlation structures. We tried fitting either a spatial or a temporal correlation function. Spatial autocorrelation between home ranges was corrected by using the mean coordinates of the home ranges, and temporal autocorrelation was corrected by giving the home range estimates for each individual continuous integers starting with 1 for the first home range estimate. Both spatial and temporal correlation were for all models fitted inside an exponential correlation structure using Euclidean distances as distance metric (function ‘corExp’ in the library ‘nlme’; Pinheiro et al. 2009), as this provided the best fit based on investigation of diagnostic plots (plots of residuals vs semivariogram) and likelihood ratio tests of models without and including the different spatial correlation structures (‘corStruct’-classes) available in the ‘nlme’-library (see Pinheiro & Bates 2004, pp 226–249). As it is not possible to fit spatial and temporal correlation structures together in the same model, we proceeded with the most parsimonious model based on diagnostic plots and likelihood ratio tests. Spatial correlation structure was best for 92% of all models. The likelihood ratio tests and diagnostic plots of the three models selecting temporal correlation over spatial correlation structures showed no distinct improvement, suggesting that the inclusion of a temporal correlation structure is not critical. Therefore, if a correlation structure was needed, we always used a spatial correlation structure.

We always checked, based on likelihood ratio tests, if a random intercept was required to enhance model fit. If the mixed-effects model was not significantly better than a linear model without a random intercept, we checked for unequal variances and dependencies among within-group errors, which can be corrected for by utilizing generalized least squares (GLS) models, but without having to add complexity by including unnecessary random effects to the model (Pinheiro & Bates 2004). In no case was such correction needed (thus GLS models were never required) so in cases where no random intercept was required, we fitted a basic linear model (LM).

When the initial model structure was determined and model assumptions met, fixed effects model selection was conducted by backwards selection of variables from the full model (Murtaugh 2009). Model comparison between the reduced and the more complicated model was by likelihood ratio tests (Pinheiro & Bates 2004). During comparison of mixed-effects models with different fixed effects, parameter values must be obtained using maximum likelihood estimation (Pinheiro & Bates 2004). After model selection, the final model was fitted using restricted maximum likelihood estimates.

All these analyses were performed for full year, summer and winter models, both home range estimation methods, all home range density isopleths, and all temporal scales. Kernel and MCP estimates of home range size were generally in agreement. The models showed the same trends, but with MCP estimates being somewhat larger than the kernel estimates. Also, the different models run for each of the three density isopleths (90%, 70% and 50%) showed similar results. For a summary of descriptive statistics of the data set, home range sizes at the different temporal scales, effects of individual differences, spatial autocorrelation and variance explained in the models, see Supporting Information, Tables S2–S4.

We report results obtained from models fitted on home range sizes estimated by the kernel method, due to both kernel and MCP models showing similar trends, but also due to earlier studies comparing these two methods of home range estimation concluding with kernel methods performing better (Worton 1987; Börger et al. 2006a). If not explicitly stated, the results refer to all density isopleths, and for predictions made with the dominant habitat type ‘forest of high productivity’ as reference, as this is the most common dominant habitat type. Except for monthly winter models (70% and 50% density isopleths), models including random intercepts for individuals were always more parsimonious. The monthly 70% and 50% kernel winter models were fitted as basic linear models, as no correlation structures or variance functions were necessary. Final models are in many cases quite complicated, with several interactions retained (Supporting Information Tables S5–S12 and S15). When presenting the results, we focus on the interactions considered biologically significant, not presenting results including the variable ‘error’ and interactions with this variable, as this variable is included in the models for corrective purposes only (see Materials and Methods).


Effect of day length and climate variables

Temperature explained a significant amount of variance in the seasonal pattern in home range size (the interaction day length × temperature in Table 1, P < 0·05 for all scales). As predicted by hypothesis H1a-b home range size increased when the temperature was higher than expected during mid-winter (minimum day length), while the inverse relationship was apparent during mid-summer (maximum day length; t > 2·2, P < 0·03 for all scales, see Fig. 1 and Supporting Information, Tables S5–S8). During winter, a 10 °C increase in temperature (which is within our observed range of values) predicted a 70% increase in monthly home range size, and an equivalent increase was apparent in summer when there was a 10 °C decrease in temperature (estimates from 90% kernels; see Supporting Information, Table S5).

Table 1. F-values for the fixed effects retained in the most parsimonious full year mixed-effects models for all temporal scales and all density isopleths (90%, 70% and 50%). Bold indicates P < 0·01, otherwise P < 0·05. NS = not significant and NR = not retained in the reduced model
Temporal scaleF-value
 Day lengthNS 6·41 4·57
 TemperatureNSNS 4·90
 Day length × temperature 8·1514·5821·39
 Temperature × precipitationNRNSNR
 Temperature × habitatNR 3·55NR
 Precipitation × error 6·52 5·10NR
 Precipitation × habitat 3·74NRNR
 Error × habitatNRNR 4·34
 Day lengthNSNSNS
 Precipitation 9·42 9·0811·88
 Day length × temperature 5·54 7·42 8·31
 Temperature × precipitation 4·63NRNR
 Error × habitat 5·59NR 4·78
 Day lengthNRNSNS
 Precipitation14·1010·31 7·67
 Error12·9212·24 6·14
 HabitatNSNS 3·52
 Day length × temperatureNRNR 5·27
 Day length × precipitationNR 4·37 5·08
 Precipitation × errorNRNR 3·91
 Error × habitat15·06 9·51 6·45
 Day lengthNS18·24 8·65
 Habitat35·40 7·06 4·14
 Day length × temperature65·0360·5949·32
 Day length × precipitation 9·33 7·83 7·51
 Day length × error24·7125·0928·26
 Temperature × precipitationNSNRNR
 Temperature × error24·7113·5614·26
 Temperature × habitatNRNR 2·92
 Precipitation × errorNRNRNS
 Error × habitat52·9238·6334·63
Figure 1.

 Plot of predicted log-transformed home range sizes (ha) in relation to temperature in °C, measured as residuals from a regression against day length. Estimates for all temporal scales are made for the most common habitat type in the data set, ‘forest of high productivity’, for the mean of the error and precipitation variables, and for the minimum, mean and maximum value of day length, corresponding to winter, spring/autumn and summer respectively. Lines show predicted values for 90%, 70% and 50% kernel estimates, in black, blue and green respectively. Points in corresponding colours are raw residuals.

Precipitation also explained a significant amount of variation on all scales and density isopleths (P < 0·01, Table 1). On the longer temporal scales precipitation showed a consistent positive relationship with home range and no seasonal effect (t > 2·3, P < 0·02 for all monthly models, 70% and 50% biweekly models and 90% weekly models, see Fig. 2 and Supporting Information, Tables S5–S7). For 90% monthly kernels a 10 mm increase in precipitation predicted a 90% increase in home range size (see Supporting Information, Table S5). On shorter temporal scales (weekly and daily) the effect of precipitation on home range size changed with season, showing positive correlation with home range size during winter and negative during summer (t > 1·9, P < 0·05, see Fig. 2 and Supporting Information, Tables S7–S8), which for summer is in agreement with hypothesis H2. For the winter season, snow depth explained further variance in home range size (P < 0·01 for all scales and all density isopleths except 50% monthly kernels, Table 2). As predicted (H2), home range size generally decreased as snow depth exceeded expected levels (t > 2·1, P < 0·04 for all temporal scales excluding weekly and 90% biweekly models; see Fig. 3 and Supporting Information, Tables S9–S12). In mid-February, a 10 cm decrease in snow depth predicted a doubling of 90% monthly home range size (see Supporting Information, Table S9).

Figure 2.

 Plot of predicted log-transformed home range sizes (ha) in relation to precipitation in mm, measured as residuals from a regression against day length. Estimates for all temporal scales are made for the most common habitat type in the data set, ‘forest of high productivity’, for the mean of the error and temperature variables, and for the minimum, mean and maximum value of day length, corresponding to winter, spring/autumn and summer respectively. Lines show predicted values for 90%, 70% and 50% kernel estimates, in black, blue and green respectively. Points in corresponding colours are raw residuals.

Table 2. F-values for the fixed effects retained in the most parsimonious winter only mixed-effects or basic linear models for all temporal scales and all density isopleths (90%, 70% and 50%). Bold indicates P < 0·01, otherwise P < 0·05. NS = not significant and NR = not retained in the reduced model
Temporal scaleF-value
 Day length471·2379·4845·09
 Snow depth 46·10 8·72NS
 Error83·68 6·29NS
 Day length × temperature10·06NRNR
 Day length × precipitation13·70NSNS
 Day length × snow depthNS 9·77NS
 Day length × error190·2461·2640·70
 Temperature × precipitation 9·52NRNR
 Temperature × snow depth 6·2120·5118·41
 Temperature × errorNSNRNR
 Temperature × habitatNSNRNR
 Precipitation × snow depthNS 6·09NR
 Precipitation × error32·84NRNR
 Snow depth × error 6·14NRNR
 Day lengthNS 4·51NS
 Snow depth83·3151·0840·69
 Habitat 3·51 3·27 4·01
 Day length × temperature 6·71 4·95NS
 Day length × snow depthNSNS 4·22
 Day length × error 8·48 9·08 7·28
 Temperature × error15·6611·98 8·22
 Temperature × habitat11·20 9·45 3·20
 Snow depth × error 4·32 6·38 9·11
 Snow depth × habitat 5·66 6·62 3·81
 Error × habitatNRNRNS
 Day length42·5324·9622·62
 Snow depth52·4023·2814·83
 Day length × temperature 8·16NRNR
 Day length × precipitationNR 5·45NR
 Day length × error21·4516·8017·95
 Day length × habitat 3·13 2·70NR
 Temperature × precipitationNR 4·24NR
 Temperature × snow depth12·5413·2411·40
 Error × habitat 4·28NRNR
 Day length42·0442·2328·48
 Snow depth24·6320·0926·17
 HabitatNSNS 3·01
 Day length × error15·5118·0724·83
 Day length × habitat 4·78NSNR
 Temperature × habitatNR 2·64NR
 Error × habitat 6·71 4·62 7·72
Figure 3.

 Plot of predicted log-transformed home range sizes (ha) in relation to snow depth in cm, measured as residuals from a regression against day length. Estimates for all temporal scales are made for the most common habitat type in the data set, ‘forest of high productivity’, for the mean of the error, temperature and precipitation variables, and for day length corresponding to mid-winter (February 19th). Lines show predicted values for 90%, 70% and 50% kernel estimates, in black, blue and green respectively. Points in corresponding colours are raw residuals.

The effect of the local climatic variables was strongest (steeper slope; Fig. 4) on home ranges measured over longer temporal scales, consistent with the indirect effects hypothesis (H3a). The effects were always significantly larger on monthly scale than on daily scale (t > 1·6, P < 0·05, see Fig. 4; Supporting Information, Table S13), but the effect was also consistently larger on longer temporal scales when comparing other scales (although not always significant; Supporting Information, Table S13). Due to interactions in the model, the assessment of the scale effect (Fig. 4) is sensitive to the choice of reference values for which the prediction is conducted. The predicted effects are for the mean of all continuous variables (climate variables and error variable), and for ‘forest of high productivity’ as dominant habitat type. Summer models are predicted at maximum and winter models for minimum of day length. As an exception, when predicting the effect of snow depth, the reference value used for light represents mid-February, which is the peak of the Norwegian winter. Predicted effects of temperature and precipitation were robust to alterations in reference values (Fig. 4a) while snow depth and day length were sensitive to alterations in the reference values when changing the value of interacting covariates ±1 SD away from the mean (Fig. 4b).

Figure 4.

 Plots of estimated slopes and corresponding 95% confidence limits of the predicted lines of (a) temperature and precipitation (from Figs 1 and 2) and (b) snow depth and hours of day light (from Fig. 3 and Supporting Information Fig. S3) shown for each of the temporal scales day, week, biweek and month. The slopes and confidence limits shown are estimated from the same reference values used to show the predicted lines in Figs 1–3 and Supporting Information Fig. S3, which are the mean of the error and climate variables, ‘forest of high productivity’ as dominant habitat type and, for temperature and precipitation plots, maximum and minimum of day length representing summer and winter respectively. For snow depth, the reference value used for light represents mid-February. 90%, 70% and 50% kernel estimates are shown in black, blue and green respectively. The horizontal bars show how the slope changes when altering reference values of interacting covariates ±1 SD away from the mean. The dotted line illustrates slope = 0, meaning the variable exhibits no significant effect on home range size.

The seasonal effect showed that home ranges were always larger during winter than during summer, but significantly so only on the daily and monthly 70% and 50% kernels (consistent with hypothesis H3b; P < 0·05, the effect of day length in Table 1 and Supporting Information Fig. S3). In summer models, growing degree days (GDD) was rarely retained in the reduced models, and only showed significance (P < 0·05) in 90% biweekly models (1 out of 12 models). The effect of GDD on home range size changed with day length (t = 2·49, P < 0·02, Supporting Information, Table S15), showing a negative relationship between GDD and home range size during mid-summer [hypothesis H3a; maximum day length (June)], and a positive relationship during late summer [minimum day length (late August)]. The effect of GDD also depended on dominant habitat type (= 2·82, P < 0·04; see Supporting Information, Table S15). At the peak of summer, the relationship between GDD and home range size changed from negative for animals having ‘forest of high productivity’ as dominant habitat type to positive when ‘other’ was more abundant within the home range. Temperature was retained in more models than GDD and had a significant effect on home range size in 9 out of the 12 summer models (t > 2·2, P < 0·03). In all cases home ranges decreased when temperatures increased above normal.

Effect of dominant habitat type

Home range size varied with the dominant habitat type within the home range on the daily, weekly and biweekly scale, but no strong pattern was found on the monthly scale (Fig. 5; Supporting Information, Table S14a,b). Thus, the results presented here relates to the daily, weekly and biweekly scales. Home ranges dominated by the habitat type ‘other’, which is assumed to be deficient in shelter and forage plants, were generally larger than home ranges dominated by the forage-rich habitat types ‘forest of high productivity’ and ‘pastures’, as predicted by hypothesis H4 (Fig. 5; Supporting Information, Table S14a,b). For 90% biweekly kernels, home ranges dominated by the habitat type ‘other’ were predicted to be 30% larger than home ranges dominated by ‘forest of high productivity’, and 60% larger than ‘pastures’ on the 50% density isopleth. However, the difference in size was not always significant (t > 1·9, P < 0·05 in 11 of 16 cases, estimates are made for the mean of all other variables in the models; see Supporting Information, Table S14a,b). Deviations from the expected pattern were apparent on the daily scale, where home ranges dominated by ‘pastures’ in the majority of cases were larger than home ranges dominated by the other habitat types (t > 3·8, P < 0·01 in 5 of 6 significant cases, see Fig. 5 and Supporting Information, Table S14b). Home ranges dominated by ‘forest of low productivity’ varied somewhat more over the various temporal scales in relation to the forage-rich habitat types, with no clear pattern emerging (Fig. 5; Supporting Information, Table S14a,b). Interactions between habitat type and other variables are listed under ‘Other interactions’.

Figure 5.

 Plot of predicted log-transformed home range sizes (ha) in the four habitat types (‘forest of high productivity’, ‘forest of low productivity’, ‘pastures’ and ‘other’). Estimates for all temporal scales are made for the mean of the error and climate variables, and for maximum day length (summer). The points show predicted values for 90%, 70% and 50% kernel estimates, with decreasing line-thickness as the size of the density isopleths decrease. The size of the points represent the relative amount of observations within each habitat type.

Other interactions

The interactions between day length and temperature/precipitation were the only biologically significant interactions consistent over density isopleths and/or scales. Other interactions were retained for certain scales and isopleths, but with no consistent pattern (Table 1). The dominant habitat type interacted with temperature in the 70% monthly and 50% daily kernel models and with precipitation in the 90% monthly kernel model (P < 0·05, Table 1; Supporting Information, Tables S5 and S8). In the winter models the effect of snow depth depended on temperature on monthly and biweekly scale (P < 0·05, Table 2; Supporting Information, Tables S9 and S10) and with habitat on the biweekly scale (P < 0·05, Table 2, Supporting Information, Table S10). As neither of these interactions was predicted from our original hypotheses, and all terms are inconsistent across scales, they are not elaborated in any further detail.

Variation explained

The observed variance in home range size explained by the full year models ranged between 19% and 46%, and from 16% to 99% in the winter models (Supporting Information, Table S4). For both full year and winter models, there was generally a gradual increase in variance explained from the shorter to the longer temporal scales, and from the smaller to the larger density isopleths. For all models except 70% and 50% monthly winter models, the inclusion of a random intercept for individual identity improved model fit (P < 0·05). The random effects explained between 19·8% and 44·0% of the variance in home range size in full year models, and between 19·6% and 44·4% in winter models, and the same pattern of increase over scales and density isopleths as found above was found here (Supporting Information, Table S4). For model fit, the fitted home range size values were always within the range of the observed values.


This study is one of the first to show that depending on the temporal scale chosen, the influence of local climate on home range size differs. We propose to use patterns of home range size variation on four temporal scales to infer regarding direct and indirect effects of local climate on animal movement and activity. Direct effects of local climate (except snow depth) influencing home range size through activity are likely to operate more strongly on shorter temporal scales (weekly-daily), while indirect climatic effects likely are operating more strongly through plant growth on longer temporal scales (biweekly-monthly). We found results consistent with both direct and indirect effects of local climate, consistent with H1-H3. Climatic effects on home range were larger on longer temporal scales, indicating indirect effects being relatively stronger than the direct effects. Further, in consistence with the importance of available energy for home range size (H4), home ranges dominated by habitat types with low forage density were larger than home ranges dominated by forage-rich habitat types.

Temporal scale and direct vs. indirect effects

Different biological processes may operate on a particular temporal or spatial scale (Senft et al. 1987; Wiens 1989; Levin 1992). Few studies have investigated the factors determining home range size in ungulates down to daily scales (Russo, Massei & Genov 1997; Kamler, Jedrzejewska & Jedrzejewski 2007), and studies comparing home ranges at multiple temporal scales are rare (Börger et al. 2006b). We suggest such a comparison over scales can be used for inference regarding direct and indirect effects of climate. For ectotherm vertebrates, the effects of, e.g. spring temperatures during embryogenesis are obvious examples of direct effects of climate (Massot, Clobert & Ferriére 2008). For large herbivores, the relative role of direct and indirect (trophic) effects of climate is more complicated (Mysterud et al. 2008). By removing the seasonal pattern in this study, only unpredictable climatic events remain. We therefore expected temperature and precipitation to have a larger effect on shorter temporal scales, where the individuals’ immediate and transient responses to direct effects of local climate should be more easily captured (Wingfield 2005). The lack of a response to such variation on shorter temporal scales may occur if the animals shift their entire home range into areas where the climate is more stable and closer to the expected seasonal values, e.g. by moving to a habitat where snow depth is less deep (Ramanzin, Sturaro & Zanon 2007).

We found that the effects of climatic variables were strongest on the longest temporal scale (monthly), which may be viewed as evidence of indirect effects of local climate on home range size, operating through vegetation development and abundance. The North Atlantic oscillation (NAO) is well known to affect local weather patterns during winter (Hurrell 1995), and in turn red deer performance along the south-west coast of Norway (e.g. Mysterud et al. 2000, 2001, 2008). Although mainly the NAO winter index has been used, it is known to affect autumn body mass of red deer indirectly through snow accumulation patterns (Mysterud et al. 2000) affecting soil moisture pattern (Kettlewell et al. 2006) and in turn determining plant growth (Pettorelli et al. 2005). Plant quality and quantity are likely to influence animal movement and activity through the degree of which the available forage fulfill the animal’s energetic needs (McNab 1963; Harestad & Bunnell 1979). Our results are therefore consistent with the suggested indirect effect of local climate operating through plant growth, and provide a behavioural pattern consistent with what analysis of autumn body mass has concluded (Mysterud et al. 2008).

If indirect effects operating through plants are the main mechanism of home range size variation, we would expect growing degree days (GDD) to be a better alternative to temperature residuals, as GDD theoretically provide a more direct measure of temperature-driven variation in plant productivity. However, the variable only showed a significant effect on home range size in one model, with the effect depending on day length and habitat type, while temperature residuals had a significant negative effect on home range size in the majority of the models. In our case, there was a strong correlation between GDD and the temperature residuals (Pearson r ranged 0·84–0·89 across scales). The use of GDD is likely more sensitive to exact measures of absolute temperature levels, as it involves a strict threshold at 5 °C. Although weather stations are fairly close by the individuals, the topography in the area is expected to produce some variation in absolute levels between stations and home range area of individual deer. This suggests that temperature, fitted as deviations from normal values (residuals from a regression against day length) is a better measure of plant productivity for our system. We clearly cannot exclude the possibility that temperature residuals also capture some direct effects even at the longer temporal scales, making it a superior variable relative to GDD.

Temperature, precipitation and snow depth effects

Direct and indirect effects are clearly not mutually exclusive at any temporal scale, although the relative importance is likely to change as we predict. A negative relationship between snow depth and home range size is indicative of a direct effect of local climate (see also Schmidt 1993; Grignolio et al. 2004). Snow cover is expected to accumulate and persist for longer time periods, and therefore acts as a direct effect also on the longest temporal scales. Movement in snow cause higher energy expenditure for large mammals as snow depth increases (Parker, Robbins & Hanley 1984), and the negative relationship with snow depth is thus probably a result of the animals conserving energy by restricting movement when snow depth exceeds normal levels. The pattern of weight loss of young cervids during winter is commonly related to snow depth, and body weight is in turn related to over-winter survival (Loison, Langvatn & Solberg 1999).

The effect of temperature varied depending on season. Home range size increased when temperature was higher than expected during winter, while the inverse relationship was apparent during summer. In summer, heat stress or insect stress during periods when temperatures are higher than normal may cause the decrement in home range size. This negative relationship between movement and temperature during warm periods is consistent with what was found in earlier studies on cervids (seasonal scale, Hayes & Krausman 1993). The positive relationship during winter could be interpreted as an energy-conserving strategy, with decreased activity when temperatures sink below average (Cuyler & Oritsland 1993; Arnold et al. 2004). The effect of precipitation on home range size was positive and consistent through the year on the longer temporal scales (biweekly-monthly), including summer. On shorter temporal scales (weekly-daily), the effect of precipitation was dependent on season, and changed in the same manner as temperature, with a negative relationship between home range size and precipitation in summer and positive relationship during winter. Precipitation is known to increase heat loss in ungulates (Barrett 1981; Parker 1988), thus we expected individuals to decrease home range size to conserve energy when precipitation exceeded normal levels in summer. The relationship found on short temporal scales in summer was therefore consistent with our predictions of direct effects of climate (H2). On shorter temporal scales the increase in home range size with increased precipitation found in winter could be explained by the need for shelter (Staines 1976). If e.g. a number of daily home ranges consist mainly of habitats poor in shelter, animals may have to move over longer distances to obtain this. However, this is not plausible on longer time scales, as animals are likely to have more heterogeneous home ranges. This subject therefore requires further attention in future studies.

Day length and habitat effects

Day length is commonly reported in studies as a factor influencing home range size. This factor is assumed to be a reflection of the available energy and forage within the home range (Kjellander et al. 2004; Anderson et al. 2005; Ramanzin, Sturaro & Zanon 2007), and therefore has an indirect effect on movement and activity. Börger et al. (2006b) found day length, along with habitat type, to be the most important factor influencing home range size in roe deer Capreolus capreolus L. on biweekly, seasonal and half-year scale. In our study, day length was negatively correlated to home range size, but not consistent across scales (on the monthly and daily time scale only). In summer, when home ranges are smallest, forage is highly nutritious and abundant enabling the animals to cover their energy demands within a smaller area than during winter, when resources are scarce and more spread. Energy requirements are shown to be higher during summer, due to calving and lactation (Hanwell & Peaker 1977; Clutton-Brock et al. 1982). Coupled with forage quality and quantity this may counteract the expected difference in home range size through the year, explaining the lack of a more consistent effect of day light. The relationship between available energy and activity may also be transferred to how habitat types influence home range size through resource properties of varying qualities and quantities (McNab 1963; Harestad & Bunnell 1979; McLoughlin & Ferguson 2000). Consistent with patterns found in earlier studies (Tufto, Andersen & Linnell 1996; Relyea, Lawrence & Demarais 2000; Said et al. 2005), we found that home ranges dominated by habitat types considered as resource-poor were generally larger than home ranges dominated by more forage-rich habitats, or habitats expected to provide both adequate forage and cover. This was apparent on all temporal scales but monthly, providing further evidence that habitat differences likely related to variation in resource abundance and quality are main determinants of home range size in large herbivores.

Home range size as a space use metric

Animal movement ecology is a field of rapid development (Nathan et al. 2008). The combination of higher resolution data from GPS collars and increased methodological sophistication (enabled by increasing computational capacity) contributes largely to these advances. The usefulness of home range size as a space use metric has been heavily debated since its introduction (review in Börger, Dalziel & Fryxell 2008). Home ranges can be viewed as emergent patterns generated by individual movement paths (Börger, Dalziel & Fryxell 2008). Movement paths are in turn determined by complex interactions between individuals and their environment (Nathan et al. 2008). Several methods are available for estimating home ranges (see e.g. Millspaugh & Marzluff 2001), each with corresponding advantages and disadvantages. The minimum convex polygon method (MCP) is criticized because of high sensitivity to outliers, sample size and sampling strategy (Seaman et al. 1999; Börger et al. 2006a). Future studies should therefore consider using other methods in addition to the MCP method, and recently, the local convex hull polygon method (LoCoH; Getz & Wilmers 2004; Getz et al. 2007) is gaining increasing support as a promising new home range estimation method. Our results were robust whether the MCP or kernel method was used, suggesting that choice of specific home range metric was not a key issue.

Mechanistic modelling suggests that site fidelity obtained by memory effects is a crucial component for stable home ranges to emerge from movement paths (Gautestad & Mysterud 2005; Van Moorter et al. 2009). The distinct division and return to seasonal home ranges in our case suggests memory effects to play a role in our system, and that the home range size is a useful metric to obtain biological insight of the dynamic responses of animals to environmental variation. That does not exclude the possibility that other approaches may yield additional insight. Recently, state-space models have been used to analyze individual movement paths more directly to identify different behavioural states, which in turn can be coupled to environmental variables (Morales et al. 2004; Patterson et al. 2008). State-space models have been developed mainly to study fine-scale movement processes, rather than comparing patterns of space use over multiple temporal scales. A pragmatic view indeed embraces the fact that several methods (each with pros and cons) are currently applicable for answering slightly different questions regarding space use, although we will likely see and welcome further integration of approaches in near future (Börger, Dalziel & Fryxell 2008).


Using an established multi-scale approach (Börger et al. 2006b) in a novel way, we were able to shed light on both the direct (temperature effect on activity level) and indirect (plant growth) effects of climate in determining home range size. Predictions of future animal distributions under climate change are often done with methods such as mechanistic niche modelling (Kearney & Porter 2009). Such approaches often ignore indirect climate effects operating on herbivores through trophic interactions. Our analysis suggests a role for climate-driven variation in plant production as well as direct effect in particular of snow depth on space use of red deer at the northern distribution limit. Understanding the effect of climate on behaviours such as movement is important to enable more accurate predictions of possible nonlinear relationships, when future global warming may bring local weather variables outside the current range.


We are grateful to Luca Börger for providing scripts from his American Naturalist paper and help in understanding them, and to Dan Nussey and two anonymous referees for valuable comments. This study was funded by the Research Council of Norway (‘Natur og næring’-program; project no. 179370/I10).