E. Reimers, Department of Biology, University of Oslo, PO Box 1066, Blindern, 0316 Oslo, Norway. Tel.: +47 95242874; fax: +47 22854726; e-mail: firstname.lastname@example.org
Knowledge about changes in behavioural traits related to wildness and tameness is for most mammals lacking, despite the increased trend of using domestic stock to re-establish wild populations into historical ranges. To test for persistence of behavioural traits of wild reindeer (Rangifer tarandus L.) exposed to hunting, we sampled DNA, vigilance and flight responses in wild reindeer herds with varying domestic ancestry. Analyses of 14 DNA microsatellite loci revealed a dichotomous main genetic structure reflecting their native origin, with the Rondane reindeer genetically different from the others and with least differentiation towards the Hardangervidda reindeer. The genetic clustering of the reindeer in Norefjell-Reinsjøfjell, Ottadalen and Forollhogna, together with domestic reindeer, supports a predominant domestic origin of these herds. Despite extensive hunting in all herds, the behavioural measures indicate increasing vigilance, alert and flight responses with increasing genetic dissimilarity with domestic herds. Vigilance frequency and time spent vigilant were higher in Rondane compared to Hardangervidda, which again were higher than herds with a domestic origin. We conclude that previous domestication has preserved a hard wired behavioural trait in some reindeer herds exhibiting less fright responses towards humans that extensive hunting has, but only slightly, altered. This brings novel and relevant knowledge to discussions about genetic diversity of wildlife in general and wild reindeer herds in Norway in specific.
Behavioural traits in mammals certainly evolve, and the domestication of animals was essential for the development of modern human societies. Selection towards domestication supported tameness for improved handling and maintenance of livestock (Price, 1984), whereas premodern hunting likely selected for traits increasing a species survival abilities through increased escape behaviours. Interestingly, knowledge about changes in behavioural traits related to wildness and tameness is lacking, often due to the extinction of the wild parent stock (Clutton-Brock, 1987). Reindeer were only recently domesticated by humans, with extensive control of specific herds first evolving during the 16th and 17th centuries (Mirov, 1945). Today, almost 50% of the approximate 3 million reindeer in the Old World are wild animals, and wild and domestic herds are managed in close coexistence (Reimers & Klein, 1979).
With an increasing trend towards conservation of wild reindeer, similar to many animals around the world (Price, 1999), populations have been established using domestic stock to re-establish wild populations into historical ranges. This provides a unique opportunity to analyse the change and/or persistence in behavioural traits such as vigilance, flight and fright behaviour in wild reindeer herds with varying domestic ancestry. Assuming domestication selects for and supports tameness, we were interested in whether wild reindeer with some domestic stock and exposed to hunting remain tame or adopt wild behaviour. The question has implications both evolutionary and from a conservation perspective. In the past, restocking of previously wild reindeer habitats has occurred through the release of reindeer from domestic herds under the assumption that ‘a reindeer is a reindeer’. It was also assumed, as in other reintroduction projects using domestic or semi-domestic wildlife around the world, that following their reintroduction, the introduced animals will adopt behaviour similar to wild animals. In addition to the behavioural aspects, genetic mixing with domestic gene pools was rarely considered and, to our knowledge, never compared and tested in connection with the reintroduction of domesticated animals. In Norway, wild reindeer are managed through hunting in several separate herds with different ancestries, from assumed pure wild herds to herds of more domestic origin (Fig. S1). Recent molecular genetic analyses revealed genetic distinctions between wild and domestic herds, providing a tool for analysing the degree of domestic ancestry (Røed et al., 2008, 2011). Preliminary studies (Reimers et al., 2000; Reimers & Svela, 2001) indicated a reduced vigilance, alert and flight responses among hunted wild reindeer with an assumed domestic origin compared to herds of more wild origin. To improve the comparison of behaviour traits among wild populations subjected to hunting, genetic analyses of the study populations were included with sampled behavioural characteristics to confirm and compare their wild or domestic ancestry. Through combined analyses of behaviour traits and genetic markers in wild reindeer herds, we were able to test the change and/or persistence of vigilance and fright behaviour related to the degree of domestic ancestry of a herd.
Materials and methods
We studied five reindeer herds in southern Norway: Rondane (1441 km2), Hardangervidda (8064 km2), Forollhogna (1822 km2), Ottadalen North and South (4550 km2) and Norefjell-Reinsjøfjell (308 km2) (Table 1 and Fig. S1). These are mainly alpine terrain at altitudes of 1000–1500 m. Large sections of the first four areas are national parks under strict management control regarding human infrastructure in terms of cabins, tourist resorts, hydroelectric development, power lines and roads. Recreational activities are extensive in Norefjell-Reinsjøfjell, but moderate in the four other areas both within and outside the national parks due to more remote location, topography and less developed hiking facilities and other human infrastructure (Table 1). Based on overnight visitors in 2004 to tourist cabins, central areas were visited by < 500 persons that year in Ottadalen (Torsbu) and < 3000 persons in Hardangervidda (Sandhaug) and Rondane (Grimsdalshytta) (Den norske turistforening, unpublished). Norefjell-Reinsjøfjell and Forollhogna have no centrally located tourist cabins, and the number of annual visitors to the areas is estimated by the lead author on the basis of cabin areas, trail systems and other human infrastructures to more than 100 000 in the former and < 1000 in the latter.
Table 1. Winter herd density (animals km−2) during the study period, hunting start, estimated sustainable harvest rate (% of winter herd) since hunting began, study period and distribution of area without infrastructure (%) in 2005. The herds are arranged from top in increasing distance from domestic stock.
Hunting is allowed in all areas (also within the national parks) and is the only important mortality factor, as wolves (Canis lupus L.) are essentially absent from the areas and wolverine (Gulo gulo L.), golden eagle (Aquila chrysaetos L.) and lynx (Lynx lynx L.), although permanently present or present as stragglers, exert minor predatory influence.
Rondane reindeer are considered to be mainly of wild origin due to their distinct genetic variation when compared with domestic reindeer (Røed, 1985, 1987; Røed et al., 2008), whereas Hardangervidda reindeer appear to be of mixed origins. The Hardangervidda herd is reported to have gone through substantial temporal genetic alterations during the period from early medieval times to the present. This appears to be related to the introgression of domestic reindeer into the wild gene pool in the 20th century during periods when reindeer husbandry was practised in this mountain region (Røed et al., 2011). The Ottadalen herd originates from animals bought from the domestic reindeer company that closed in 1964 (Reimers, 1972). Continuous domestic reindeer herding in Forollhogna from 1784 up to 1914 with herd sizes up to 10 000 animals (Frøstrup & Vigerstøl, 2006) has probably displaced the wild reindeer. The present herd is assumed to mainly originate from animals that escaped from neighbouring domestic herding groups during the 1950s. The Norefjell-Reinsjøfjell herd originates from some 30 reindeer that escaped slaughter in 1968 when the reindeer herding company closed (Reimers et al., 2009). However, the degree of wild or domestic ancestry of the present Ottadalen, Forollhogna and Norefjell-Reinsjøfjell herds may be questioned because the extent of remnant wild animals present during the founding of these populations and/or possible immigrations from neighbouring wild populations are unknown.
Group behaviour characterizes wild reindeer throughout the year in all the areas. When moving or lying, individuals are rarely more than a few metres apart, whereas during grazing they may be somewhat more spread out.
Sampling procedure: genetic data
We obtained tissue samples of reindeer from Ottadalen (n = 36), Forollhogna (n = 44), Norefjell-Reinsjøfjell (n = 37) and Rondane (25 from Snøhetta and 31 from Sølnkletten). DNA was isolated using DNeasy kit (Qiagen GmbH, Germany), following the manufacturer’s guidelines. All samples were analysed for 14 reindeer-specific microsatellites (NVHRT-01, NVHRT-03, NVHRT-16, NVHRT-21, NVHRT-24, NVHRT-31, NVHRT-48, NVHRT-73, NVHRT-76 (Røed & Midthjell, 1998) and RT-1, RT-5, RT-6, RT-9, and RT-27 (Wilson et al., 1997). The amplification was performed on a GeneAmp PCR System 9700 (Applied Biosystems, Foster City, CA, USA) as previously described (see Methods in Røed et al. (2002)). PCR products were electrophoresed using an abi prism 3100 Genetic Analyzer (Applied Biosystems). Use of these markers in a previous study has given evidence of low scoring errors (< 5%) due to stutter bands, allelic dropout or null alleles (Røed et al., 2008).
The genotypes were pooled together with previous reported scoring of the same microsatellites in the wild reindeer herds from the Hardangervidda region (pooled Nordfjella, Hardangervidda and Setesdal-Ryfylke, n = 83) and the Rondane/Dovre region (pooled Snøhetta, Knutshø, Sølnkletten, Rondane Nord and Rondane Sør, n = 211) and the domestic reindeer from southern Norway (pooled Filefjell, Vågå and Røros, n = 106) (Røed et al., 2008).
We video-recorded vigilance in grazing animals from a hidden position and defined grazing as the act of ingesting forage with the muzzle down. Individuals were recorded preferably for 10 min or until the reindeer lied down, were hidden by neighbouring animals or moved out of sight. We defined a vigilance bout as the act of interrupting feeding to lift the head above the shoulders (Frid, 1997) and observe the surroundings for ≤ 10 s before returning to feeding (Bøving & Post, 1997). Vigilance bouts lasting more than 10 s frequently ended up in nonvigilance behaviour like scratching, urinating and licking before returning to grazing or moving on to another grazing spot and hence were disregarded as vigilance bouts. Position of the filmed animal would frequently change from internal to peripheral and vice versa in a moving grazing group with normal speed of approximately 3 km h−1. The interindividual distance between neighbouring grazing animals is short, frequently < 5 m, and individuals were not sampled more than once on the same day.
While filming, we registered group size, group structure [males, females and yearlings, mixed (all ages and both sexes)], sex and age of the video-recorded animal: lactating female (female with calf at heel), barren female (≥ 1 year old), male (≥ 1 year old) or calf, wind speed following the Beaufort Wind Scale (calm, < 1 m s−1; light/gentle breeze, 1.6–5.4 m s−1; moderate/fresh breeze, 5.5–10.7 m s−1 or gale, 10.8–17.1 m s−1), weather (sunny/partly sunny, cloudy, rain/snow or foggy) and topography of the surrounding area (i.e. level or rugged).
We played back the videotapes on a 27” plasma TV monitor following individual reindeer grazing throughout the observation period and registered (all registrations made by the lead author) the number and duration of vigilance bouts. Animals were selected by stopping the film at random intervals and recording the closest animal that was grazing. Slow walking between vegetation hot spots with head down was included in total grazing time. Slow walking with head up was excluded from total grazing time.
We recorded vigilance for 2934 reindeer during daylight hours in the four areas in 3 years (2004–2006) during three periods: March–April (winter), June–July (summer before the insect season) and October (rut after the insect season) – Rondane (n =683), Hardangervidda (n =1024), Ottadalen (n =663) and Norefjell-Reinsjøfjell (n =564). All recordings were made in the alpine zone above timberline. Observation period per animal averaged 176 s. Mean and median group size in the herds were 176 and 120 in Norefjell-Reinsjøfjell, 505 and 175 in Ottadalen, 330 and 250 in Hardangervidda and 554 and 500 in Rondane. Animals were randomly selected for video-recording from 24 different groups in Norefjell-Reinsjøfjell, 23 groups in Ottadal, 34 groups in Hardangervidda and 25 groups in Rondane.
Flight responses were measured in the same four areas over a longer time period – Rondane (1993, 2003–2006), Hardangervidda (2004–2006), Ottadalen (1992, 2004–2006), Norefjell-Reinsjøfjell (1992, 2002–2006) – and in a fifth area Forollhogna (1996). When a group of reindeer was first sighted, the observer measured start distance and then five additional parameters: group size, group composition (i.e. mixed, all ages and both sexes; males, yearlings, and older; females and calves), dominant activity of the group when first sighted (i.e. lying, grazing or standing), wind direction relative to the observer (i.e. tail wind or into the wind including crossways to the wind) and topography (terrain ruggedness) of the surrounding area (i.e. level or rugged).
The observer then approached the centre of the group directly at a constant speed of 1.1 m s−1 with ≤ 6 s stops to measure alert distance (the distance when the reindeer exhibited an increased alert response by grouping together) and flight initiation distance (the distance from the directly approaching observer to the reindeer when the reindeer initially took flight). The observer continued until reaching the position where the reindeer were located at the start of the approach and then measured escape distance (the shortest straight-line distance from where the reindeer took flight to where the reindeer resumed grazing or bedded down). All distances were measured with laser binoculars or monocular (Leica Geovid 7 × 42 BDA or Leica Rangemaster 1200 Scan (Leica, Wetzlar, Germany); 1-m accuracy at 1000–1200 m). Longer escape distances (> 1000 m) were measured in metre from maps.
Assessment time was time elapsed from alert to flight initiation estimated from measured distances and assuming a constant observer speed of 1.1 m s−1. We then divided assessment time into two classes according to established dichotomous procedure (Reimers et al., 2009, 2011) to estimate assessment probability: > 1 s, which we classified as animals assessing the observer, and ≤ 1 s, which we classified as animals not assessing. For the summer, rut and winter seasons combined, we approached groups of reindeer 671 times in the five areas: Rondane (n =104, average group size 207; median 148), Hardangervidda (n =65, 132; 55), Ottadalen (n =139, 342; 80), Forollhogna (n =83, 122; 30) and Norefjell-Reinsjøfjell (n =280, 132; 342).
The flight response data from Forollhogna, Norefjell-Reinsjøfjell and Hardangervidda are previously published in other contexts (Reimers et al., 2006, 2009, 2010) and are included here to make a comprehensive review of vigilance and flight behaviour in wild reindeer herds in southern Norway in the light of the genetic analyses. As the differences in flight responses in the prehunt and hunt periods in Norefjell-Reinsjøfjell were small (Reimers et al., 2009), we pooled the flight response data from the two periods in the current presentation. The flight response data from Forollhogna (Reimers et al., 2006) are integrated in the total sample and are given a more updated statistical treatment.
Statistical analyses: genetic data
The amount of genetic variation is expressed as mean number of alleles, allele richness and gene diversity (Nei, 1987) in each herd across loci using fstat 2.9.3 (Goudet, 2001). This program was also used to assess the genetic structure among herds (FST). Significance levels for the FST tests were corrected for type 1 and type 2 errors according to the false discovery rate procedure of Benjamini & Hochberg (1995).
Genetic structure at an individual level was analysed by the Bayesian assignment approach as implemented in the software structure (Pritchard et al., 2000). The log likelihood of our data [ln Pr(X│K)] was estimated given different numbers of genetic clusters (K ∈ [1,10] using an admixture model with uniform priors (α = 1, αmax = 50), correlated allele frequencies, 50 000 burn-in cycles and 500 000 MCMC iterations. All analyses were run without prior herd information and were repeated 10 times for each K value.
Statistical analyses: vigilance
To assess the variation in the frequencies of vigilance bouts among the four areas, we fitted the observed vigilance data to a Poisson model. Assuming that the expected number of vigilance bouts observed for an individual was proportional to the time it was observed, we included ln(time observed) as a fixed offset in the log-linear model for the number of vigilance bouts. As the areas and groups were not surveyed on the same days, we included the random variation among days and individuals within groups and days as variance components in the model to account for weather effects and to facilitate robust inferences about the differences among the areas and the herds. Hence, we assumed the number of vigilance bouts for an individual i observed over τi seconds during day d(i) to be Poisson-distributed with the log-linear expectation
where δd(i) is normally distributed, δd(i)∼ N(0, σ2), and xiβ models the difference between the areas and log-linear effects of covariates (xi). This model was fitted with the function ‘lmer’ in the ‘lme4’ package in r version 2.11.0 (http://www.r-project.org/).
Besides effects of area, we evaluated the main effects of group size (i.e. small: < 20 animals, medium: 20–50 animals and large: > 50 animals), group structure (mixed sex groups or single sex groups of either females or males), functional category of the observed animal (lactating female, barren female or male), season, year, weather, wind force and topography.
To compare the mean duration of the vigilance bouts among areas and groups of individuals, a log-normal linear model (including all the fixed and random variables in the model above) was fitted to the duration of the first observed vigilance bouts per individual. Inspection of the residuals showed that the model fitted well to the data. To facilitate the interpretation of time spent vigilant while grazing, we multiplied vigilance frequencies with vigilance duration in the four areas (Tables S2 and S3) for presentation in Fig. 3b.
Statistical analyses: fright responses
To assess fright responses, we transformed response distances into their natural logarithms prior to analysis. Because we approached some reindeer groups repeatedly on the same day (range = 1–9, median = 1, = 1.5), we analysed data with linear mixed-effects models LME (Pinheiro & Bates, 2000) and generalized linear mixed-effects models GLMM (Woods, 2006) when response variables were continuous and binomial, respectively. We included reindeer group as a random effect (intercept) to account for dependency associated with observing the same group more than once and number of approaches per group per day as a numerical fixed effect. Number of approaches per day provided a test of habituation and sensitization in responses.
We analysed alert, flight initiation and escape distances with LME. For tests of fixed effects in LMEs, we used marginal F-tests (Pinheiro & Bates, 2000). We selected variables using backwards elimination (Crawley, 2005) by starting with full models containing all biologically plausible main effects, including area, herd density, season, start distance (ln-transformed and centred at the mean), group size (i.e. small: < 20 animals, medium: 20–50 animals and large: > 50 animals), group composition, wind direction, topography, activity type, number of approaches per day (centred at 1) and two 2-way interactions: season x start distance and wind direction x topography. The rationale for the inclusion of the interaction terms was the seasonal influence on the start distance (higher visibility in winter) and the topographical influence on wind direction (more variation in wind direction in rugged topography). We removed variables with the highest P-value and repeated this procedure until only variables with P <0.05 were retained except for area, which we always retained in all models. We do not report results of marginal F-tests; we present only parameter estimates of the final models. We fitted LME models using the library nlme (Pinheiro & Bates, 2000) implemented in r version 2.11.0. We analysed probability of reindeer assessment time (delay flight after being alert) with a GLMM, using the function ‘lmer’ in the R library ‘lme4’. We classified the binomial response as one if assessment time was > 1 s and zero if assessment time was ≤ 1 s (i.e. immediate flight after alert).
The level of genetic variation was explicitly lower in the Norefjell-Reinsjøfjell reindeer as compared to the other herds with regard to the number of alleles observed, allele richness and gene diversity (Table 2), suggesting that some genetic loss has occurred during and after the founding of the Norefjell-Reinsjøfjell herd.
Table 2. Mean values of genetic variation across 14 microsatellite loci in herds of reindeer from southern Norway. The herds are arranged from top in increasing distance from domestic stock. Values are number of individuals analysed (N), mean number of alleles per locus (A), mean allele richness (Ar) per locus and mean genetic diversity (H).
There was substantial genetic differentiation among the herds as expressed by the FST (Table S1). All herds were significantly differentiated with FST ranging from 0.009 between the Ottadalen herd and the domestic reindeer to 0.133 between the Norefjell-Reinsjøfjell herd and the Rondane reindeer. The Rondane reindeer showed high differentiation to all the other herds (mean FST ± SD of 0.109 ± 0.027), but with least differentiation towards the Hardangervidda reindeer (FST = 0.065), which again was more similar to the domestic reindeer than Rondane reindeer. Both Ottadalen and Forollhogna herds were genetically most similar to the domestic reindeer.
The unrooted genetic distance (DA) tree corroborated these results (Fig. 1). The Rondane reindeer clustered distinctly different from the others. Although the long genetic distance of the Ottadalen, Forollhogna and the Norefjell-Reinsjøfjell herds illustrates their genetic distinction, the branching gives a significant closer genetic distance of these herds towards the domestic herds as compared to the Hardangervidda and Rondane herds, supporting their mainly domestic origin.
The structure algorithm showed a significant increase in log-likelihood values from K =1 to K =4, after which the values were more unstable (Fig. S2). The increase was particularly significant from K =1 to K =2, suggesting a dichotomy as the main genetic structure with a proportionate memberships of most Rondane reindeer in a separate cluster from all the others (Fig. 2). At K =3, most of the Hardangervidda reindeer separated out leaving most reindeer with an assumed domestic origin in its own cluster. This suggests a close ancestry of the reindeer in Norefjell-Reinsjøfjell, Ottadalen and Forollhogna towards the domestic reindeer also at the individual level. The proportionate memberships of reindeer in the fourth cluster (K =4) were consistently Norefjell-Reinsjøfjell reindeer (Fig. 2), illustrating a genetic distinctness.
The number of vigilance bouts per minute grazing in summer was lower in Norefjell-Reinsjøfjell (0.11) and Ottadalen (0.14) than in Hardangervidda (0.21) and Rondane (0.37) (Table 3, Fig. 3a). Adults were more vigilant in winter and during rut than during summer and more vigilant than calves during rut and winter (no vigilance measure of calves in summer) (Fig. 3a, Table S2). The animals expressed maximum vigilance in the rutting season, medium in winter and least in summer (Fig. 3a). Females with calves at foot in summer were not more vigilant than barren females (females that had lost their calves or that had not given birth). Vigilance decreased with increasing group size (Tables 3 and S2). Animals in small groups were 33% more vigilant (95% CI: 6–68%) than animals in large groups. Total time spent vigilant during grazing was similar during rut and winter and 2.3–2.2 times longer in these two seasons than in summer (Fig 3b, Table S3). Reindeer in Norefjell-Reinsjøfjell and Ottadalen spent least time vigilant, whereas Rondane reindeer spent most time vigilant and Hardangervidda reindeer were intermediate.
Table 3. Generalized linear model for predicting the number of vigilance bouts per minute (ln-transformed) during grazing and linear mixed-effects model for predicting alert, flight initiation and escape distances (ln-transformed) of groups of wild reindeer disturbed by an approaching person in five areas (Norefjell-Reinsjøfjell, Ottadalen, Forollhogna, Hardangerviddda and Rondane) in southern Norway during three seasons (summer, rut and winter). We used generalized linear mixed-effects model for predicting probability of reindeer groups assessing an observer before fleeing (1 = assessing). Reference levels for categorical variables are provided in the table (the level after ‘vs.’). NAs denote variables that did not enter the model. See Tables S2–S7 for details. The start model for predicting vigilance included area, group size, group structure, functional category of the observed animal (lactating female, barren female or male), season, year, weather, wind force and topography. The start model for predicting response distances included area, season, group size, ln(start distance), number of disturbance on the same group, group structure, wind direction in relation to observer, topography, activity and the interaction terms season : ln(start distance) and wind direction in relation to observer : topography. The model for predicting assessment included ln(alert distance) in addition.
Prob. of assessing
†0.001 < P ≤0.01.
‡0.01 < P ≤0.05.
Area: Hardangervidda vs. Rondane
Area: Ottadalen vs. Rondane
Area: Forollhogna vs. Rondane
Area: Norefjell vs. Rondane
Season: rut vs. winter
Season: summer vs. winter
Group size: small vs. large
Topography: level vs. rugged
No. of approaches
Alert and flight initiation distances increased with increasing start distance, indicating that when we approached reindeer from farther distances, they responded to the person at farther distances. Start distances 200 and 400 m predict flight initiation distances at 58 and 72 m in Norefjell-Reinsjøfjell and 297 and 367 m in Hardangervidda (Table S4). Controlling for other factors, groups became alert at shorter distance in Norefjell-Reinsjøfjell (118 m) compared to Forollhogna (180 m), Ottadalen (196 m), Hardangervidda (332 m) and Rondane (325 m) (Fig. 4, Tables 3 and S4). Alert distances were shorter in summer and rut compared to winter and longer for small groups (< 20 animals) compared to large groups (> 50 animals) (Tables 3 and S4). Alert distances were shorter during rut than in summer (t158=−3.292, P =0.001). Number of approaches to the same group on the same day did not influence alert distances, which were shorter in level compared to rugged terrain (Tables 3 and S4).
Flight initiation distances were longer in winter than in summer and rut and, within areas, longer in Rondane (335 m) and Hardangervidda (338 m) than in Norefjell-Reinsjøfjell (66 m), Ottadalen (142 m) and Forollhogna (151 m) (Fig. 4, Table 3). Flight distances were shorter during rut than in summer (t167 =−3.089, P =0.002). Small- and medium-sized groups had longer flight initiation distances than did large groups, and distances were farthest in rugged terrain (Tables 3 and S5). When we approached the same groups repeatedly on the same day, they initiated flight at shorter distances, a flight decrease that ranged from 4 m per attempt in Norefjell-Reinsjøfjell, 9 m in Ottadalen and Forollhogna to 21 m in Hardangervidda and Rondane (Tables 3 and S5).
Winter and summer escape distances were similar in the five areas and longer than in rut (Table 3, Fig. 4). Escape distances in winter were shorter among groups in the three herds originating mostly from domestic reindeer: 166, 405 and 242 m, respectively, in Norefjell-Reinsjøfjell, Ottadalen and Forollhogna vs. 1495 m in Hardangervidda and 1064 m in Rondane (Table 3, Fig. 4). Small- and medium-sized groups showed longer escape distances than large groups, and terrain ruggedness did not influence escape distance. When we approached the same groups repeatedly on the same day, the escape distance decreased by 12% (95% CI: 19.8–3.5%) per attempt (Tables 3 and S6).
Probability of assessing the observer before fleeing was lower in winter than in summer or during rut (Table 1, Fig. 4). In summer, the probability to assess the observer exceeded 75% in Norefjell-Reinsjøfjell, Ottadalen and Forollhogna herds vs. 27% in Hardangervidda and 49% in Rondane. Assessment probability was lower when reindeer were encountered in rugged compared to level terrain (Tables 3 and S7).
The genetic analyses show that Rondane reindeer maintain a genetic makeup different from Hardangervidda animals and distinctly different from the herds in Norefjell-Reinsjøfjell, Ottadalen and Forollhogna. Vigilance and fright responses support a corresponding pattern, with the highest vigilance rate in Rondane, medium in Hardangervidda and lowest in Norefjell-Reinsjøfjell and Ottadalen. This likely reflects a genetic impact scale from previous domestic reindeer herding in the areas. Although the long genetic distance of the Ottadalen, Forollhogna and the Norefjell-Reinsjøfjell herds illustrates their genetic distinction, the branching gives a significant closer genetic distance for these herds towards the domestic herds as compared to the Hardangervidda and Rondane herds, supporting their mainly domestic origin. The Norefjell-Reinsjøfjell reindeer dominated the fourth cluster (K =4) in the assignment analyses, supporting a genetic distinctness likely evolved through bottlenecks and genetic drift since the founding of this herd in 1968. The reduced amount of genetic variation in the Norefjell-Reinsjøfjell reindeer together with the long genetic distance to the Hardangervidda herd suggests that no gene flow has occurred from the Hardangervidda to Norefjell-Reinsjøfjell herd. A mainly domestic origin of the Norefjell-Reinsjøfjell, Ottadalen and Forollhogna herds, together with the reduced vigilance and flight responses in these reindeer, indicates that previous domestication has preserved a hard wired behavioural trait and that extensive hunting since 1956 in Forollhogna, 1967 in Ottadalen and 1992 in Norefjell-Reinsjøfjell has, but only slightly, altered.
As all five herds are extensively hunted, there are influential factors other than their differences in genetic origin that cause differences in fright responses among the areas. As predator pressure is low for all herds, recreational activities being extensive in Norefjell-Reinsjøfjell (Reimers et al., 2009) and low in the other areas (Andersen & Hustad, 2004) are a relevant candidate. It is well established that frequency of encounters with humans is an important behavioural interacting factor in ungulates; see reviews by Tarlow & Blumstein (2007) and Stankowich (2008). Overall, there was a weak yet robust habituation response of ungulates to increased human activity (Stankowich, 2008). Ungulates in areas with frequent contact with humans show reduced flight responses compared to those in areas where human encounters are rare (Cassirer et al., 1992; Burkowski, 2001; Colman et al., 2001). Two recent studies (Reimers et al., 2010, 2011) confirm Colman et al.’s (2001) findings that regular exposure to humans in a nonconsumptive recreational context can buffer short bursts of hunting and other fright-provoking activities. Hunting may increase vigilance, for example (Benhaiem et al., 2008; Jayakody et al., 2008). In his review, Stankowich (2008) found that hunting had a weak, yet robust effect on flight responses. However, this effect was highly heterogeneous, and several studies report little or no effect of hunting (Donadio & Buskirk, 2006; Reimers et al., 2009). The absence of an increase in (i.e. selection for) vigilance and/or flight responses in these studies despite hunting may be a result of modern, less intrusive hunting techniques and behaviour of modern hunters as discussed in the study by Colman et al. (2001). Although the behavioural response is innate, learning plays an important role in the manner and degree to which ungulates respond to humans and predators (Geist, 1971; Kloppers et al., 2005). The lower vigilance rate and flight response distances in Norefjell-Reinsjøfjell compared to Ottadalen, in spite of comparable domestic origin and extensive hunting, may reflect habituation to the high level of recreational activities in the former, supporting Stankowich & Blumstein (2005) who found a general cross-species effect of repeated approaches on flight behaviour. Ottadalen is a larger area, less accessible, with limited recreational activities (Table 1), hence counteracting habituation to humans because of a lower encounter rate.
Our observations that flight initiation distance and escape distance decreased with number of approaches on the same group on the same day support these and other studies (Thomson, 1972). The amount of decrease in escape distance upon repetitive provocations of the same groups was approximately 12%. Flexibility in behavioural responses might help animals cope with increased predation risk, and also with human encounters, without suffering excessive costs. Rather than making redundant investment in antipredator behaviour, animals that already are safe enough can make greater investment in foraging (Frid, 1997).
Declining individual vigilance efforts with increasing group size has been widely reported for both mammals and birds (Elgar, 1989; Lima & Dill, 1990; Lima, 1995). Wild reindeer conform to this strategy. However, contrary to expectation, alert, flight initiation and escape distances decreased with increasing group size. In their review, Stankowich & Blumstein (2005) found that whereas fish tolerated closer approach when in larger schools, other taxa had greater flight initiation distance when in larger groups. It looks as individuals through a coordinated shoaling and compacting response gained an increased perception of safety when aggregated. It may be that the compacting behaviour seen in Rangifer in response to disturbance gives the individuals a corresponding increased perception of safety.
High vigilance rates will in most cases compromise feeding time as suggested by Laundre et al. (2001), although vigilance is not necessarily mutually exclusive with processing food (chewing and swallowing) (Fortin et al., 2004). The low scan frequency and total scan duration found in reindeer in this study should not compromise feeding efficiency in any of the study herds. However, a vigilance level two and three times higher in Hardangervidda and Rondane, respectively, compared with Norefjell-Reinsjøfjell and Ottadalen may indicate a generally higher level of alertness among reindeer in the two former areas. As shown for roe deer (Capreolus capreolus) (Weiner, 1977) and moose (Alces alces) (Regelin et al., 1981), even a small change in a standing posture more than doubled the energy cost of standing over lying. The low probability of assessment and the long response distances in Rondane and Hardangervidda may add to these costs in terms of both loss of grazing time and an increase in energy expenditure if animals are exposed to frequent disturbances on a weekly or seasonal basis. A possible relationship between vigilance and flight response behaviour and the lower total body mass of adult female reindeer in Hardangervidda and Rondane (54–66 kg) compared to the substantially higher weights recorded in Norefjell-Reinsjøfjell, Ottadalen and Forollhogna (74–77 kg) (Reimers, 1997) may add to an otherwise general pasture quality explanation for body weight differences among Norwegian wild reindeer populations.
Contingent on confirmation of genetic origin, we expected that wild reindeer with mainly domestic origin and exposed to extensive hunting would re-develop increased vigilance and fright responses similar to reindeer with a historically wild origin. Molecular genetic analyses provided a clear genetic structure among our study herds mainly reflecting an either domestic or wild ancestry. However, and contrary to our expectations, hunting pressure was not associated with high vigilance or longer fright responses in our study areas. Efficient predator control and habituation/sensitization resulting from area differences in recreational activities likely add to this effect. Habituation would apply to the relaxed reindeer herd in the recreational hotspot Norefjell-Reinsjøfjell, despite hunting, compared to the far less recreationally impacted other herds. Higher alertness levels and longer response distances in Hardangervidda and Rondane could easily translate into energy budget consequences that may contribute to reindeer body weight differences recorded in these herds.
We thank S. Eftestøl., K. Flydal, Ø. Flaget, E. Lurås, C. Egeland, T.K. Tveitehagen, B. Dahle, M. Jenstad, A. Muniz and L. Dervo for excellent field assistance; L.E. Loe, T. Ergon and D.T. Alemu for statistical advice; and D.T. Blumstein, T. Stankowich and an anonymous referee for critical comments to an earlier version of the manuscript. Financial support was provided from The Norwegian Science Foundation, Trygve Gotaas Fond, Direktoratet for Naturforvaltning, Stiftelsen Thomas Fearnley, Heddy og Nils Astrup.