Predation is one of the key factors potentially influencing the structuring components (e.g. age, sex and individual quality) in natural populations (Sih 1987, Lima & Dill 1990, Stearns 1992, Huggard 1993). Although two very different predator species may be functionally identical, it is important to know the specific effect on prey for each predator, as earlier studies (Sih et al. 1998, and references therein) have demonstrated that multiple predators have effects that cannot be predicted simply by summing the effects of each predator species. While it has long been assumed that predators select ‘the old, weak and sick’ members of prey populations (Mech 1970, Curio 1976) and that the degree of selection depends on the predators' hunting technique (Schaller 1972), modern day recreational hunters often select large trophy animals. Despite the fact that humans are the dominant ungulate predator in most temperate ecosystems, there have been few attempts to compare their selectivity with that of mammalian carnivores within the same ecosystem (see Ginsberg & Milner-Gulland 1994 for a tropical example).

The European roe deer Capreolus capreolus is the most widespread wild ungulate in Europe today, and it occurs mainly in areas where large mammalian predators are presently absent. However, in an increasing number of areas of sympatry predator-prey interactions between the Eurasian lynx Lynx lynx and the European roe deer are common. Although > 30 species are found in the diet of lynx (Jędrzejewska & Jędrzejewski 1998), roe deer constitutes up to 90% of the biomass in lynx diet (Jędrzejewski et al. 1993), and each lynx is estimated to kill 30-70 roe deer annually (Jobin et al. 2000; R. Andersen, unpubl. data). Earlier studies in Switzerland (Molinari-Jobin et al. 2002) indicated that roe deer does were most often killed, while studies in Poland (Okarma et al. 1997) indicated that roe deer fawns and adults were killed in proportion to their occurrence in the population. In the latter study, the body condition of lynx-killed roe deer (measured as marrow fat content in 70 animals) was described as normal, indicating no selection for undernourished prey.

In our study we compare the age, sex and condition of roe deer killed by lynx and recreational hunters, and relate this to estimates of the status of the standing population. More specifically we predict that roe deer killed by lynx will represent a random section of the population (i.e. age, sex and body condition of prey will not be different from that of the standing population), whereas hunters will select adult males, i.e. a non-random section of the population.

Material and Methods

Estimation of Age and Body Condition of Predated and Harvested Animals

Lynx-killed roe deer were found by either 1) snow-tracking unmarked lynx until a roe deer kill was located, or 2) radio-tracking lynx until the movement pattern indicated that a kill had been made. The kills by radio-tracked lynx were not approached until the lynx had left, so as not to disturb the lynx. Only animals located during autumn and winter were used in the present analysis. Lynx-killed roe deer were collected in both the Hedmark and Nord-Trøndelag study areas. A total of 153 roe deer killed by lynx was found between September 1994 and May 1996. Of these, 151 could be aged (< 1 year old or > 1 year old) and sexed. In order to determine the exact age class of roe deer that were > 1 year old, we sectioned the incisors (Reimers & Nordby 1968 ) in 44 roe deer killed by lynx (during 1994-1996) and 53 roe deer killed by hunters (in 1995 only).

The femur was collected from 79 lynx-killed roe deer. From 23 of these animals we also collected the lower jaw. A portion of the bone marrow was removed from the white portion of the femur, avoiding the haemopoietic end arts of bones. Body condition was expressed as percentage fat and was analysed volumetrically using the Babcock method (Helrich 1990). As has previously been found for moose (Bergstrøm et al. 1986) and semi-domestic reindeer (Kjelvik 1997), we also found a close correlation between bone marrow fat content in the lower jaw and the femur in 23 lynx-killed animals (F  =  52.92, df  =  1,21, P < 0.001, R²  =  0.72). Consequently, we transformed values of fat content from the lower jaw in 46 animals harvested between October and December 1996, to femur values using the formula:

Age and sex structure of hunter-killed animals was collected from two sources. First, we collected harvest bag statistics from all the municipalities within both study areas. From these we were able to sex and age 6,074 roe deer shot during 1988-1997. In addition, jawbones were collected from hunter-shot animals so we could accurately age animals > 1 year old for a more detailed look at age structure. These jawbones were collected from both the Hedmark and Nord-Trøndelag study areas.

Results

Effect of Age and Sex

The age and sex structure of lynx-killed roe deer had a non-significant tendency to differ from that observed in the standing population (χ²  =  7.333, df  =  3, P  =  0.062; Table 1). The minor discrepancy in the structure could be ascribed, entirely, to a larger proportion of juvenile females (20.5 vs 11.4% in the standing population) and, to a lesser extent, to a smaller proportion of adult males (23.8 vs 32.7% in the standing population) in the lynx-killed sample.

Table 1

Standing age and sex distribution of roe deer revealed by proportions seen on artificial feeding places during November-April (N  =  246; in addition there were 21 juveniles which could not be sexed), and proportion killed by lynx (N  =  148) and hunters (N  =  6,074) in 13 municipalities in central and southeastern Norway.

Recreational hunters, on the other hand, shot a higher proportion of males (58.5%) and of adults (72.9%) than available in the population (45.0 and 69.4%, respectively; χ²  =  16.98, df  =  1, P < 0.001 and χ²  =  8.82, df  =  1, P  =  0.003, respectively). This was an effect of the fact that most animals were shot during the buck-only season, as the proportion of males shot during the ordinary hunting period was not significantly different from the standing sex distribution (χ²  =  0.86, df  =  1, P  =  0.354). Among > 1-year-old animals for which exact ages were available there were no significant differences in age distribution between those killed by lynx and those killed by hunters (Kolmogorov-Smirnov 2-sample test: Z  =  0.973, P  =  0.30; Fig. 1).

Figure 1

Sex and age distribution of roe deer killed by lynx (A; during the winters of 1994/95 and 1995/96) and hunters (B; in autumn 1995). Eighty roe deer killed by lynx were classified as fawns or adults, and 44 adults could be classified to exact age classes. The proportion of fawns killed by hunters was calculated from hunting records from the study area (N  =  153), whereas the age distribution of adults was calculated from samples sent in by hunters from the study area (N  =  53).

Discussion

According to our predictions we found that lynx killed a random section of the prey population. There was no selection of any sex or age category, and the body condition of lynx-killed and hunter-killed roe deer did not differ. Furthermore, among adult animals, specific age did not differ between roe deer killed by lynx and roe deer killed by hunters, however, hunters killed a disproportionally large proportion of adults, especially males.

Despite large environmental variation (e.g. snow depth), there was no temporal variation in prey selectivity. The mobility of roe deer is affected by snow depths of > 40-50 cm (Mysterud et al. 1997), but roe deer in the Hedmark study area did not change their habitat use, despite large annual variation in snow depth. Artificial feeding sites were heavily used in both years by all sex and age groups. Although roe deer reach about 70% of adult body size during their first winter (Andersen et al. 1998), we would expect the mobility of the youngest age class to be most affected by high snow accumulation. Although it is hypothesised that predators preferentially select young animals because they are less able to escape predation (Mech 1970, Curio 1976, Vitale 1989, Karanth & Sunquist 1995), optimal foraging theory predicts that predators ought to choose the most profitable prey (MacArthur & Pianka 1966, Charnov 1976). For large predators the most profitable prey type would be the largest available prey that can safely be killed (Sunquist & Sunquist 1989). All adult lynx can safely kill any age and sex class of roe deer. Hence, if lynx were able to select specific categories of roe deer, adult animals would be overrepresented. However, we will hypothesise that a stalking lynx will select its prey based on the actual position of the animal (e.g. lynx try to shorten the distance to the prey before attacking), and do not have time to pick out a particularly vulnerable animal like coursing predators (Fuller 1989, Bowyer 1987). This can explain why there was no temporal variation in prey selectivity (i.e. no difference between the years with different snow depth).

The age and sex structure of lynx prey was quite similar to the standing age and sex distribution (except for a higher occurrence of female fawns in lynx-killed roe deer), and the body condition did not differ between animals killed by lynx and animals killed by hunters. As it is unlikely that hunters will be able to judge the body condition of an individual animal, we expect that lynx select a random sample of animals within each sex and age group. The fact that there was no significant difference in the actual age of animals killed by lynx and hunters supports this conclusion. In contrast, in predator-prey studies involving coursing predators like wolves Canis lupus, it has been shown that the median ages for adult male and female prey were somewhat greater than for hunter-killed deer (Fuller 1989). The lack of such selectivity in this study can be attributed both to the fact that lynx is a stalking predator, i.e. select the prey that could be most easily approached, and to the facts that there are neither body size dimorphism or further age-related increases in body size in roe deer > 2 years old (Andersen et al. 1998), making it impossible to select on either age or sex of animals based on body size alone. Thus, the pattern of lynx predation found in our study is broadly similar to those from the Jura mountains in Switzerland (Molinari-Jobin et al. 2002) and from Białowieża in Poland (Okarma et al. 1997).

Male biased predation has been found in several predation studies (e.g. Karanth & Sunquist 1995, Fitzgibbon 1990, Fitzgibbon & Fanshawe 1989, Poulle et al. 1993). Several explanations for the observed pattern are presented. Males are often solitary, i.e. no group-shared vigilance (Taylor 1984), or in species that group specific spacing behaviours of the males may render them more vulnerable to predators, as observed in gazelles (Fitzgibbon 1990), even though the overrepresentation of male gazelles in the diet of African wild dogs Lyacon pictus was hypothesised to be caused by lower flight distances (Fitzgibbon & Fanshawe 1989). Furthermore, males roam widely during the rut, which increases the risk of encountering predators, or males may be hampered during escape because of their antlers or may suffer from intra-specific aggression-related injuries or post-rutting loss of condition (Karanth & Sunquist 1995). During winter, roe deer bucks are included in groups of animals surrounding artificial feeding places, where often matrilineal groups are developed (Holand et al. 1998), moreover, bucks have no antlers in winter. Consequently, bucks do not roam, are not solitary and share the same group vigilance as other animals. Furthermore, the highest level of territorial aggression in roe deer is in May; thus, aggression-related injuries in winter are less likely to affect escape from predators. This may explain the lack of male-biased predation in our study.

Results from studies of large carnivores reporting body condition of their prey are varied. Indications of heavy predation on old animals and animals in poor physical condition have been observed in some studies, e.g. African wild dogs preying on impala Aepyceros melampus (Pole et al. 2003), lynx preying on red deer fawns (Okarma 1984), or coyotes Canis latrans preying on mule deer Odocoileus hemionus and white-tailed deer O. virginianus (O'Gara & Harris 1988). In contrast, Poulle et al. (1993) found that coyote killed deer > 6 years old, however, most deer killed were in good physical condition. Mountain lions Puma concolor preying on mule deer and white-tailed deer often killed individuals in prime condition (O'Gara & Harris 1988, Huseeman et al. 2003).

The relative vulnerability of ungulate prey species to solitary stalking predators may depend both on prey anti-predator behaviours and on prey size (Temple 1987). It is not surprising that a stalking predator like the mountain lion, nearly as large as its prey, can kill prey in good physical condition (Pierce et al. 2000), whereas lynx in Poland, being ⅓ the body size of red deer, select substandard individuals. At the same time, cooperative hunting, such as e.g. in coyotes, makes predators able to select animals in good body condition (Pierce et al. 2000). Hence, the size of the stalking predator relative to its prey, and cooperative hunting in many coursing predators, may reverse the general patterns usually found in coursing and stalking predators.

In contrast to lynx, recreational hunters in our study showed clear selectivity for adult animals, especially males. This is in accordance with a study by Lentle et al. (2000), reporting that recreational red deer hunters in New Zealand shot a larger proportion of males (most bucks in the rutting season) and harvested a significantly greater number of older males, than compared to a random sample. Under most environmental conditions, both the mean and the variance of fitness components greatly differ among age and sex classes in populations of large herbivores (Gaillard et al. 2000), and the components most potentially influential for changing population growth show buffered temporal variation (Gaillard & Yoccoz 2003). While most populations can sustain quite high predation rates on juveniles, those populations decline quite rapidly when the annual survival rate of prime-age animals, especially females, decreases substantially (e.g. Kinley & Apps 2001, McLoughlin & Owen-Smith 2003). This indicates that recreational hunters could have the potential to affect population growth per se more than lynx, although their focus on males may buffer this impact. While extremely skewed sex ratios (low male percentage) have been shown to affect the reproduction of female moose (Sæther et al. 2003), there is no evidence that these effects occur in roe deer under the existing sex ratios. Given the current preference of recreational hunters to shoot trophy males it is actually lynx which kill the highest proportion of adult females (see also Molinari-Jobin et al. 2002). The fact that both lynx and recreational hunters selected animals in good physical condition, indicate that their effect on roe deer mortality is likely to be additive, i.e. increasing their potential effect on population growth rate. Because predator avoidance can potentially be costly, the manner in which roe deer make space use decisions to balance the relative risks of hunter versus lynx predation is an additional issue deserving attention before we can truly evaluate the relative numerical and selection impacts of lynx and hunters.