The problem of wolverine Gulo gulo predation on free-ranging sheep Ovis aves in mountain areas is well documented from Norway (Kvam et al. 1988, Overrein & Fox 1989, Loen 1991). Usually some sheep herds are severely hit, and in some years lamb losses exceed 40%. At present, the Norwegian government pays more than NOK 6 million (USD 945,000) per year in compensation for wolverine predation on sheep, and uses considerable amounts in the administration of compensation and on preventive measures. It has been difficult to document any reduced sheep losses as an effect of these preventive efforts. Sustainable management of viable populations of wolverine is an official policy in Norway (Miljøvern-departementet 1992). However, the conflict created by wolverine predation on sheep and domestic reindeer Rangifer tarandus results in strong local pressure for licenses to kill wolverines and in illegal killing. This poses a threat to the sustainable management of wolverines, which often occur in small populations.

The wolverine lives in a boreal/mountain ecosystem where reindeer, small game, small rodents (Rodentia) and birds are important prey species (Myhre & Myrberget 1975, Pulliainen 1968, Magoun 1987, Landa et al. in prep.). Sheep, usually completely unattended, are in the system throughout the short summer period when the general food supplies are highest. However, the importance of sheep as prey for wolverines is not known.

The behavioural mechanisms involved when wolverines prey on sheep, both the behaviour of sheep and wolverines, are poorly known. Research to find ways for wolverines and sheep to coexist has received little attention so far and there has been no previous research on ways to alter wolverine predatory behaviour. Several research programs have been conducted on different carnivores with respect to altering their behaviour towards and preventing predation on domestic animals. These programs included fencing (Nass & Theade 1988), sound (Bomford & O'Brien 1990), toxic sheep collars (Connoly et al. 1978, Savarie & Sterner 1979), taste and olfactory chemicals (Cringan 1972, McColloch 1972, Conover & Kessler 1994) and guarding dogs (Green et al. 1984, Black & Green 1984). A commercial ‘assault spray’, containing oleoresin capsicum, has been shown to be effective in driving away grizzly/brown bears after sudden encounters (Hunt 1985).

The literature on smell and taste agent experiments on carnivores is relatively limited, but some is available on dogs Canis familiaris and coyotes Canis latrans. Donovan (1967) supposed that secretion from a dog's anal gland could be effective in frightening other dogs. Linhart et al. (1977) identified a number of potential ‘anticoyote’ agents that had been tested by different researchers. With conditioned food aversion (CFA), sheep carcasses or sheep bait packages are treated with the emetic agent lithium chloride (LiCl). In theory, the predator becomes sick after ingesting the bait, develops an aversion to the taste, and subsequently avoids killing sheep (Conover et al. 1977, Forthman Quick et al. 1985). Two large-scale field evaluations on coyotes have been carried out (Bourne & Dorrance 1982, Gustavsen et al. 1982, Jelinski et al. 1983), but opposite conclusions were reached, and Conover & Kessler (1994) found that the method had not been applied in practical use because it is not sufficiently effective. The free-ranging summer pasturing system of Norwegian sheep husbandry would probably also prohibit use of such agents because of the possibilities of detrimental effects on non-target scavenging species such as golden eagle Aquila chrysaëtos, carrion crows Corvus corone, ravens Corvus corax, red fox Vulpes vulpes and arctic foxes Alopex lagopus.

To find potential olfactory aversive agents for wolverines, tests with many non-lethal chemicals and oils have been performed under controlled conditions in captivity (Landa et al. 1993, Landa & Tømmerås in press). After first having tested the aversion effect of each individual substance, a mixture of tiger balsam, ylang-ylang oil, tea-tree oil, peppermint oil, pelargonium oil, methyl salicylate, p-dichlorobenzene and amyl acetate was found to be most potent as a repellent and was used for field tests.

The aims of our study were: 1) to measure the extent to which sheep was included in the diet of the wolverine and 2) to test whether the repellent agents identified earlier could be attached to sheep and thereby cause a reduction in wolverine predation.

Study area and methods

The Snøhetta region in central South Norway, was recolonised by wolverines in 1976–79 (Landa & Skogland 1995). The wolverine is presently the only ‘large’ predator in this area and the population has been monitored by snow-tracking almost every year since 1979 (Kvam 1979, 1980, Kvam & Sørensen 1981, 1983, Overskaug et al. 1986, Sørensen & Kvam 1986, Kvam et al. 1987, Røskaft 1988, Landa & Skogland 1989, Loen 1991). Since 1990, annual monitoring has been supplemented with radiotracking of wolverines. All sightings or indications of wolverine hunting or scavenging activities on different prey have been recorded. All dens and observations of females with cubs, or lone cubs, observed from May to September, have been recorded annually since 1979, when the first denning activity was recorded. Observations reported by the public were evaluated by interviewing of the observers.

To study the possible influence of sheep as prey on wolverine population dynamics, we tested if there were any correlations between sheep losses and sheep population size and between these variables and the recorded number of wolverines and wolverine reproductive output. Wolverines are known to store food (Haglund 1966). To test any possible effects of stored sheep carcasses, we tested if there were any correlations between the number of sheep claimed to have been lost due to wolverine predation during the autumn and the recorded wolverine numbers the following winter or wolverine reproductive output the following year. Since 1989, we collected scats and prey remains from all known dens in June. At this time of the year the dens are no longer used by wolverines. Scats were analysed using the frequency method (Myhre & Myrberget 1975) and identifiable bone, hair or feather fragments were classified as either reindeer, small rodent, bird, sheep or hare.

The evaporation rates of the different agents to which captive wolverines responded (Landa & Tømmerås in press) were first tested in the laboratory. We developed low density polyethylene dispensers with a release rate of 0.8 mg to 3.0 mg per day at 20°C. The most powerful agents from the captivity experiments were portioned in separate dispensers in the 1993 field test of evaporation rates under ambient temperatures throughout the grazing season. The dispensers were attached to a specially designed wool clip on an elastic collar. The wool clip was constructed to keep the dispenser in position in the upper neck region because wolverines almost invariably kill sheep with bites in this part of the body (Myrberget & Sørensen 1981, Landa et al. 1986). During the 1994 field tests, we used a mixture of all the agents in one polyethylene dispenser. The amount of each substance was 0.25 ml. With seven agents used, the total amount of liquid in each dispenser was thus 1.75 ml. In 1994, the dispenser was sewn onto the elastic collar and the wool clip was improved. We also tested a set up where the dispenser was enclosed in a cloth bag and attached to a conventional ear tag. A set up where the dispenser was taped onto the top edge of the bell collar of ewes was also employed. We performed a small-scale test in sheep pens to ensure that the agents did not alter ewe/lamb interactions.

Table 1.

Recorded numbers of wolverines, dens, successful breeding attempts, sheep remains in wolverine scats and total numbers of free-ranging and lost sheep in the Snøhetta area, 1979–1995.

Figure 1.

Relationship between the number of free-ranging and lost sheep in organised grazing districts within the summer range at Sn0hetta, 1979–1994.

Sheep flocks used in field tests were chosen according to the following criteria: 1) high documented wolverine predation rates (min. 10% losses during the previous two years), 2) minimum of 150 animals in flocks, and 3) the presence of precise herd records. We also prioritised flocks that grazed in areas frequented by radio-fitted wolverines.

Table 2.

Average spring and autumn weights and weight increase (kg) in marked and unmarked lambs in two sheep flocks, one with dummy dispensers and one with active dispensers, 1993.

In 1993, we chose five sheep flocks in the Snøhetta area. In 1994, we included three additional flocks in the Trollheimen area. In 1993 we also tested if there were any effects of dispensers without chemical agents by equipping two flocks with empty dispensers. These flocks were chosen according to the same criteria as above, but were as geographically distinct from the others as possible. Although difficult to control, this was done to ensure that wolverines that had experienced real repellents should not find the dummy control group. Each test flock was divided into two groups, one control group not fitted with dispensers and one test group fitted with dispensers. All test animals were chosen randomly, but if possible we fitted whole family groups, that is, a ewe and her lambs, with dispensers. The sheep were fitted with dispensers when they were released on their summer ranges, varying from the end of May until the end of June. During summer the flocks ranged freely without supervision by shepherds, and they were gathered from the beginning of September until October.

All experimental flocks were monitored by trained personnel throughout the grazing season. In areas frequented by radio-fitted wolverines, the personnel carried receivers. Post-mortem analyses were performed on sheep carcasses using the method described by Myrberget & Sørensen (1981) and Sørensen et al. (1984).

Table 3.

Number of ewes and lambs released in spring and number of lost ewes and lambs in test flocks, 1993.

Results

Field test of repellents in 1993

In the flocks with dummy dispensers in 1993, lamb loss was significantly higher in the untreated (7.6%) than in the treated group (3.2%), (χ² = 4.03, df = 1, P = 0.044, Table 3). Among experimental flocks marked with active dispensers, the untreated group showed much higher losses (11.8%) than the treated group (1.8%), (χ² = 17.46, df = 1, P = 0.000, see Table 3). In flock no 3, which grazed in the same area as flock no 2, we found one marked and three unmarked lambs that were killed by wolverines. In flock no 4 no carcasses were found.

The elastic collar proved to be too tight on the largest lambs, but not critically so. The collars tended to collect twigs and coniferous needles, which reduced the quality of the wool in the neck region somewhat. About 20–25% of the wool clips were either lost or broken. Of the remaining partly intact marks, more than 40% of the wool clips had loosened from the attachment area so that the repellents were out of position. Only about 3% of the dispensers were completely intact, because many of the tubes containing chemicals had opened during the grazing season.

Table 4.

Numbers of ewes and lambs released in spring and the total number of lost sheep in flocks with active dispensers in the Snehetta and Trollheimen areas, 1994.

Figure 2.

Geographical distribution of summer pasture for the experimental flocks in the Snehetta area (1–5), and the Trollheimen area (1–3).

Discussion

During the more than 70 years for which wolverines were absent from the Snøhetta area, the sheep management system changed from a sheep herding system to a free-ranging and unattended system, where sheep are released in spring and collected in the autumn. With the return of the wolverine, increased sheep losses, mainly lambs, due to wolverine predation have been documented annually (e.g. Kvam et al. 1988). Between 10% and 20% of the sheep are found and wolverine predation accounts for 50–85% of the mortality (Mortensen 1995, Børset 1995). The amount of meat required for the basal metabolism of an adult wolverine has been calculated to be 100 – 180 gram daily (Iversen 1972, Kvam & S0rensen 1981), which means that one wolverine would not need more than 3 – 5 lambs during the summer-autumn period. The number of lambs killed by wolverines (70–100 lambs/wolverine/season) is therefore much higher than what should be expected from the wolverine's metabolic needs.

However, in spite of the problems with wolverine predation in the area, and in spite of the annual compensation for more than 1,200 sheep paid by the government at present, the number of sheep in organised grazing districts in the area has increased more than fivefold since 1979. Our analysis demonstrates that sheep mortality is increasing in a density-independent fashion. Although the wolverine is an important predator on sheep, we have no proof that sheep is an essential part of wolverine diet. Landa et al. (in prep.) found that the abundance of small rodents was the most important factor correlated with wolverine breeding success and that reindeer, mostly from scavenging, was the most frequent food during winter. Considering the small wolverine population size, the annual claims for kill permits, and the danger of illegal actions reducing the wolverine population, combined with few management options to solve the problems, we feel justified in trying methods to alter wolverine behaviour towards sheep.

Although a high proportion of technical failures occurred in both years of field experiments, there was a consistent and statistically significant pattern of lower losses in treated groups than in untreated groups, both in dummy experiments and repellent experiments. Since all marked family groups were chosen randomly, such differences should not be expected unless wolverines selectively avoided marked sheep. The largest effects were found in active dispenser experimental flocks, but the dummy experiments indicate that also dummy dispensers helped reduce wolverine predation. Thus, simple mechanic arrangements could be effective. Because lambs from both marked and unmarked groups pastured together in smaller groups, this could have resulted in a lower loss also in the control group. The random marking of whole families could have influenced our results even though we had a high degree of technical failures, as our captivity experiments showed that wolverines responded to small amounts of the chemical agents. Therefore, one explanation could be that, in spite of technical failures, some of the volatile agents were present throughout most of the pasturing season even if an ampoule was open.

Many factors influence the detection of dead lambs and the determination of the cause of death, including personnel experience, pasturing pattern, vegetation, intensity and time of predation. There is an element of uncertainty in concluding that losses below the normal expectation in our test flocks are caused by lowered wolverine predation, when wolverine predation only was documented a few times within, or close to, our test flocks. But this is unavoidable in experiments like this. The observation that treated lambs with intact repellents were killed and the variation in lamb losses indicate that both predation pressure and individual variation in wolverine predatory behaviour and shyness could be important.

In conclusion, repellents do seem to work, but essential questions have not been answered by our tests. What will happen when all the sheep within a large area are marked and the wolverines no longer are given a choice between marked and unmarked sheep? Will wolverines simply switch back to their natural prey or habituate to the volatile agents? Large-scale tests must be performed before it can be determined whether the use of averting agents may allow wolverines to persist in areas with large numbers of untended sheep grazing on open range.