Study area and the wolf pack

The wolf territory of this study was situated at the border between Sweden and Norway, and the wolves spent almost all their time on the Swedish side in the province of Värmland (position of territory core area: 60°30′N, 12°50′E), in south-central Scandinavia. We estimated the territory to hold a mean number of five individuals during the study. The pack size varied from two to five individuals. In addition, more or less throughout the whole study period, there were one or a few loners in the area which seemed loosely connected to the pack. As we dealt with the only existing family group of wolves on the Scandinavian Peninsula at the time, consequently no neighbouring territories existed. We cannot rule out the possibility that we were actually dealing with two subsequent alpha pairs within the same area. Reproduction took place each year, which we verified by visual observations, presence of cub hair in scats (see below), and tracks on the first snow in autumn. Litter sizes were unknown, but during the entire study period we never recorded more than two cubs to be alive in the autumn. Territory size during winter was approximately 600 km², determined from snow tracking, and it had more or less the same distribution throughout the whole study period.

The study area is hilly with altitudes ranging from 120 to 600 m a.s.l., and mainly covered with boreal forests of Norway spruce Picea abies and Scots pine Pinus sylvestris. Deciduous trees, e.g. birch Betula spp., aspen Populus tremula, alder Alnus incana, and willow Salix spp., occur mainly at rivers, lakes, human settlements, and as primary vegetation in the numerous clear cuttings. Logging activities are extensive and forest gravel roads penetrate the entire landscape.

Most precipitation falls during summer and temperatures range between 10° and 20°C. During winter, temperature normally ranges between 0° and -25°C, and the snow cover often exceeds one metre. However, throughout this study winters were unusually mild, the snow depth was much below normal and varied little between years. Only in the winter of 1990/91 did the depth exceed 0.5 m from mid-January to mid-March (snow depth was measured weekly at one mid-altitude spot in the territory throughout the study). However, the snow cover varied considerably with altitude within the territory.

Roe deer have undergone a dramatic population increase in Scandinavia during the last decades, and have relatively recently spread to the study area (Liberg, Cederlund & Kjellander in press). Here they live more or less on the edge of their possible range, restricted by deep snow and partly dependent on artificial feeding. The moose population peaked in the late 1970s and early 1980s, and thereafter stabilised at a somewhat lower density (Cederlund & Markgren 1987, Hornberg 1995). The moose population is strongly influenced by human hunting, and the annual harvest is often between ⅓ and ½ of the total population.

The population of lynx Lynx lynx, which mainly feeds on roe deer and therefore is a potential competitor for the wolf, has increased relative to the increase in roe deer (Liberg & Glöersen 1995). Red fox Vulpes vulpes, an important predator on roe deer fawns (Linnell, Aanes & Andersen 1995), was almost totally absent during the study period due to sarcoptic mange (Lindström, Andrén, Angelstam, Cederlund, Hornfeldt, Söderberg, Lemnell, Martinsson, Sköld & Swenson 1994). Brown bear Ursus arctos occurs only sporadically in the study area.

Diet investigation by scat analysis

Unlike many other studies only using kills observed in the snow for analyses of winter diet (e.g. Peterson 1977), we chose to investigate wolf diet by scat analysis all year round, and hence we avoided methodological differences between summer and winter. By regularly patrolling forest gravel roads by car, we collected 684 wolf scats from March 1988 through June 1992. After rinsing and drying the scats, hair from prey species were identified by microscopic examination of (i) impression of scale structure, and (ii) cellular structure in cross sections; the latter being the only way of distinguishing moose hair from roe deer hair. Several tufts of hair were examined from each scat, and if macroscopic examination suggested presence of more than one type of hair, additional samples were examined. Comparisons were made to reference collections (Lund University) and the literature (Wildman 1954, Appleyard 1960, Brunner & Coman 1974, Debrot, Fivaz, Mermod & Weber 1982, Blazej 1989, Teerink 1991). Several blind tests were made, especially to establish reliability in distinguishing between moose and roe deer (several parts of the body of both juvenile and adult individuals of both species were checked) and no errors were detected.

In autumn, moose calves and roe deer fawns moult their reddish juvenile fur to grey/brownish fur similar to that of adults, and therefore these two types of hair are the only age categories distinguishable when analysing scats. Consequently, in winter, hair classified as coming from 'adult' moose or roe deer (see e.g. Table 1), still may come from young-of-the-year.

As prey species differ in volume/surface ratio, large species leave less hair in wolf scats per unit mass consumed than small species (Mech 1970). Consequently, by only using frequency of occurrence of hair in scats as an estimate of species composition in the diet, large prey species will be overestimated by number and underestimated by mass. Therefore, we corrected the data using a linear regression equation (Y= 0.439 + 0.008X) between prey mass of each species per scat and the total mass of an average-sized individual of each prey species (Weaver 1993, see also Floyd, Mech & Jordan 1978). The estimated mean masses of the different prey species, X in the equation were the following (the numbers in parenthesis represent the mean mass estimated to have been utilised by wolves, subsequently used for back-calculation for proportion by number of consumed prey): moose with adult fur 200 (100) kg, adjusted for predominance of females and young, see below; moose with juvenile fur taken before 15 July 30 (25) kg; moose with juvenile fur taken after 15 July 70 (50) kg; roe deer with adult fur 25 (17) kg; roe deer with juvenile fur taken before 15 July 7 (5) kg; roe deer with juvenile fur taken after 15 July 12 (10) kg; badger Meles meles 12 (9) kg; beaver Castor fiber 15 (12) kg; hare Lepus timidus 3.2 (3) kg; bird Aves 1.0 (1.0) kg; domestic sheep 30 (20) kg. Voles and lemmings Arvicolinae were not adjusted because their low mass is outside the range on which Weaver's (1993) regression equation was based; instead they were assumed to have been consumed in one piece and their mass was estimated to be 25 grams.

One must bear in mind that the scat analysis method has its limits in correctly reflecting the importance of different prey species (see Ciucci, Boitani, Pelliccioni, Rocco & Guy 1996). For example, the size and sex of the individual prey is not determined and one has to estimate the mass of the consumed prey. Moreover, although Weaver's (1993) equation covered the size range of the prey animals of this study (except voles and lemmings), it was partly based on other prey species. Two important species in this study, roe deer and badger, were not included in the feeding experiments leading to the equation.

Sex and age of moose killed by wolves vs hunters

During 1985–1992, we found 65 moose carcasses, the majority during winter, that were defined as having been killed by wolves. We often examined carcasses several days, or even weeks, after killing in order not to disturb the wolves. Consequently, it was sometimes difficult to determine the cause of death, but we only included kills where tracks and/or injuries suggested that the moose had been killed by wolves. The lower jaw was collected for age determination, which we made by counting annual cementum layers of the first molar (Ml) (Markgren 1969). Sex was determined by presence or absence of antlers (or rudiments).

For comparison we also investigated age and sex in a sample of moose killed by hunters. This sample contained all moose (N = 155) shot in hunting areas owned by ‘Stora Skog AB’ within parts of the wolf territory during two years, 1989 and 1991. In general there were no restrictions on the age and sex classes which could be killed, but hunters were generally recommended to take calves and to avoid shooting large bulls, which are very rare. Hunters generally also tended to avoid shooting cows with calves and to prefer males. Hence, there may have been a significant surplus of females in the population.

Ungulate census

During the first winter, 1987/88, we found tracks of the alpha pair over an area of 523 km² on the Swedish side of the border, the area approximately corresponding to the wolf territory. In this area we made an annual ungulate census each spring (1988–1992) by fecal pellet group counts (see Neff 1968) in 60 permanent squares of 1 km². Of these, 25 and 35 squares were randomly distributed above and below the median altitude (390 m a.s.L), respectively. The sampling frequency at lower altitudes was higher because of an expected higher density of ungulates there in winter. In each of these squares we walked along two parallel transects, 500 m apart, and every 50 m we counted pellet groups in circular spots with the size of 25 m² for moose and 10 m² for roe deer. The spot size for the roe deer census was smaller in order to minimise the risk of overlooking the smaller-sized pellet groups (Smith 1968).

To estimate the mean density, N, of moose and roe deer during each winter we used the formula

where D is the number of pellet groups found, A the total area sampled, T the number of days during which the pellets accumulated (from defoliation to time for pellet groups counts which were made immediately after snow melting; 247 and 229 days above and below median altitude, respectively), and F the average number of defecations per day and individual; 14 for moose (K. Wallin, pers. comm.); and 21 for roe deer (modified from Mitchell, Rowe, Rat cliffe & Hinge (1985) and O. Liberg, pers. comm.).

However, these estimates of ungulate densities do not give relevant measures of variance and therefore, variance is not presented in Figure 4. For statistical analysis of differences between years, more reliable results are achieved by comparing mean number of pellet groups in winter, which indirectly indicates ungulate density. This was done using the likelihood ratio test, according to method and statistical software (PELANAL) described and provided by White & Eberhardt (1980). To do this we first used the goodness of fit test to determine whether our pellet group data fitted the negative binomial distribution. For moose, we pooled the number of pellet groups found in the 20 spots in each transect and used each transect as a unit (resulting in N = 120, 500 m² each). For roe deer, we used the pooled number of pellet groups found on both transects in each square, i.e. 40 spots, as a unit (resulting in N = 60, 400 m² each), to obtain the best fit to the negative binomial distribution. The negative binomial distribution is described by two parameters: the mean number of pellet groups (m), and a positive exponent (k) (measure of contagion); for details see White & Eberhardt (1980).

Diet investigation by scat analysis

Moose, roe deer, and badger dominated the diet of the wolves, as hair from these three species were found in 52%, 50%, and 19% of the scats, respectively (Table 1; note that hair from more than one species could be found in a single scat). Remains from beavers, hares, and birds were each found in less than 5% of scats. Hair from voles and lemmings were found in 11% of scats (see Table 1), but normally as secondary content and therefore not important as prey for the wolves. Hair from sheep, the only domestic animal we found remains from, occurred in 0.4% of scats (see Table 1). In 4% of scats we found hair from wolf cubs, normally as secondary content. The majority (75%) of these were collected in May each year (the first scat was collected 28 April, the last 16 June), showing that reproduction had occurred. We found remains from more than one prey species in 35% of scats and on average there were remains from 1.39 (S.D. 0.59) species in each scat. Moreover, in the 359 and 341 scats containing hair from moose and roe deer, respectively, adults and juveniles of both species occurred together in 11%.

Table 1.

Occurrence of hair (or feathers) from various prey species in wolf scats given as proportions (%) and total number (N). Hair from more than one prey species could occur in the same scat.

There were no changes in the composition of prey species between summer and winter (see Table 1). Nevertheless, there were significant differences between summer and winter in the occurrence of hair from juvenile and adult moose, and juvenile roe deer, but not in adult roe deer (see Table 1). However, presented proportions of ‘adult’ moose and roe deer during winter in Table 1, also includes unknown proportions of young-of-the-year (see Methods).

We found significant between-year differences in the composition of hair from various prey species in the scats (x2₉ = 24.09, P < 0.01; Fig. 1). However, the frequency of occurrence of moose and roe deer hair was stable throughout the study period, although there was a tendency for moose to be more common the first year (see Fig. 1). The smaller proportion of hair from ‘other’ prey species (e.g. beaver, hare, rodents, birds), and the higher proportion of hair from badger in the last year of the study contributed most to the significant difference between years (see Fig. 1).

Our calculations suggested that the 684 scats represented a total mass of 915 kg food ingested, all prey species pooled (area/volume ratio adjustments according to Weaver 1993). The three largest species, moose, roe deer, and badger constituted 97% of the mass. Among these three, moose was most important, constituting 66%, whereas roe deer and badger constituted 27% and 8%, respectively (Fig. 2). When back-calculating from assumed prey weights (only estimated mass consumed included, see Methods) we found roe deer to be the most commonly consumed species by number, constituting 52%, whereas moose and badger were about equally common, constituting 25% and 23%, respectively (see Fig. 2).

We used these proportions by number in an attempt to assess the impact of the wolves on the ungulate populations. However, we have no data on consumed biomass of the wolves in this study. The literature suggests a range of 2 to 8.4 kg of consumed mass/wolf/day (Kolenosky 1972, Peterson 1977, Fuller & Keith 1980, Fritts & Meach 1981, Hollerman & Stephenson 1981, Carbyn 1983, Peterson, Woolington & Bailey 1984, Messier & Crete 1985, Ballard et al. 1987), assuming a mean mass of 40 kg of the wolves in this study (Olsson & Wirtberg, unpubl. data). Restricting the interval to 2.5 to 5 kg/wolf/day, our assessment is that the five wolves annually killed 90–180 roe deer, 45–90 moose, and 40– 80 badgers, plus a smaller number of other species. The majority of these wolf-killed ungulates were young animals, because (i) about one third of the roe deer and half of the moose were those with juvenile (reddish) fur killed during summer (judging from proportions of hair in scats and taking prey mass at different age into account), and (ii) in the remaining ones, with 'adult' fur (see Methods), the proportion of young-of-the-year still was high, 50% in moose, judging from the kills found (see next section).

Sex and age of moose killed by wolves vs hunters

During a period of eight years (1985–1992) we found 65 moose killed by wolves; only seven kills of roe deer were found and most of these were so thoroughly consumed that sex and age could not possibly be determined. Being more easily detectable under snow conditions, 80% of moose kills were found during winter (November-April). Sex could be determined in 58 (89%) of the moose kills. Of these, 76% were females (all age classes pooled), which was significantly more (χ², = 9.37; P < 0.01) than the proportion of females (53%) in the sample of moose killed by hunters (total N = 155; all age classes pooled). Moreover, wolves and hunters killed different proportions of the sexes depending on the age of the moose. In the age class 0–2 years, 67% of the wolf-killed moose were females (N = 42, sexes pooled) versus 39% killed by hunters (N = 95, sexes pooled), which is a significant difference (χ², = 8.97; P < 0.01). Of moose > 3 years old, females dominated in both wolf (100%; N = 16) and hunter (77%; N = 57, sexes pooled) kills, but the higher proportion of females killed by wolves was significant (χ², = 4.69; P < 0.05). Accordingly, while wolves killed most females in both young and old moose, hunters changed from males in young to females in old moose.

Figure 1.

Observed proportions of hair from prey species in wolf scats collected during 1988–1992 in northern Värmland, south- central Scandinavia. The bars show the yearly proportion of moose, roe deer, badger and 'others' based on scats collected from 10 May to 9 May the following year. To keep the data uniform within the 12-month periods 44 and 92 scats collected before the former and after the latter date, respectively, were excluded.

Figure 2.

Estimated proportions by number and consumed biomass of the three most important prey species of wolves in south-central Scandinavia. Data are derived from the scat analysis and adjustments are made to different area/volume ratios in animals of different body-size according to Weaver (1993).

Of the 65 kills by wolves 51% were calves (<1 year old), whereas 1- and 2-year-old moose together constituted 25% (Fig. 3). The age interval of 3–7 years was poorly represented (5%) in the wolf diet, whereas the interval of 8–12 years was more common (15%). We only found three moose >12 years old to have been killed by wolves, one (1.5%) was 14 years old, and two (3%) were 20 years old (see Fig. 3).

We found no significant differences in the proportion of moose of different age categories killed by wolves and hunters (χ² ₂ = 1.47; P = 0.48) (moose were pooled in three age classes: 0–2, 3–11, and >11 years to achieve sufficient numbers for the analysis). To make the comparison more adequate we excluded hunter-killed males that were more than two years old because no moose in this age category were killed by wolves.

We also made a general assessment of disease or deformity in animals killed by wolves and hunters, but found no obvious signs of neither in the two samples.

Ungulate census

The estimated mean winter density of moose during the study period was 1.5/km², means ranging from 1.2 to 1.8 animals/km² (Fig. 4). The corresponding roe deer density was considerably lower, 0.4/km², means ranging from 0.2 to 0.7 animals/km² (see Fig. 4). Consequently, assuming that the wolf territory was 600 km² (including the Norwegian side), the mean number of moose and roe deer in winter was 900 (range: 720–1,080) and 240 (range: 120–420), respectively.

Our moose pellet data did not differ significantly from the negative binomial distribution in any of the five years (see goodness-of-fit probabilities in Appendix Ia, Model 1). For roe deer, no significant differences from the negative binomial distribution were found in four of the five years, but a significant difference (P = 0.036) was found in 1989/90 (see goodness-of-fit probabilities in Appendix Ib, Model 1). Despite this deviation in one year, we treated pellet data from all years, and for both species, as if negatively binomially distributed in further analyses.

Figure 3.

Age distribution of moose killed by wolves and hunters, in south-central Scandinavia. As the wolves did not kill male moose older than two years of age, males of this age category were not included in the sample of hunter-killed moose.

Figure 4.

Mean winter densities of moose and roe deer in a 523 km² area corresponding to the wolves' territory. Data are derived from pellet group counts. For a measure of variance, based on mean number of pellet groups each year, see Model 2 in Appendix Ia for moose, and Model 1 in Appendix Ib for roe deer. For statistical analysis of differences between years within both species, see the text, and Appendix IIa and IIb for moose and roe deer, respectively.

We found significant differences between years in mean number of pellet groups for both moose and roe deer. For moose, there was a significant variation between years in mean (m) values of number of pellet groups, but constant k-values; i.e. best fit to Model 2 (Appendix IIa). For roe deer, there were significant variations between years in both mean (m) values and k-values; i.e. best fit to Model 1 (Appendix IIb). Consequently, the biological conclusion is that the winter density of moose decreased during the study period, at least there was a significantly lower density the last winter (1991/92; see Fig 4). On the other hand, the winter density of roe deer increased significantly during the study period, approximately by a factor of 3 (see Fig. 4).

Preferred ungulate prey

Several studies conducted in North America, where the general ecological conditions and the body size of wolves are similar to those in Scandinavia, have shown that moose are not preferred when smaller- sized ungulates are available, e.g. white-tailed deer Odocoileus virginianus or elk Cervus elaphus (Murie 1944, Pimlott, Shannon & Kolenosky 1969, Mech & Frenzel 1971, Voigt, Kolenosky & Pimlott 1976, Carbyn 1983). In our study, the ungulates available for wolves to prey upon were moose and roe deer. Also here the smaller species, the roe deer, was the one most commonly taken. Our data suggest that roe deer were killed about twice as often as moose despite the fact that moose density was about three times as high as that of roe deer. This suggests (i) that roe deer was preferred, and perhaps a more optimal prey, and (ii) that the roe deer population was potentially more affected by the wolves. Still, our calculations suggest that the moose was clearly the most important prey, constituting about two thirds of the total mass ingested, whereas roe deer only constituted about one quarter. The reason for this inverted relationship is the much larger body mass of the moose.

Summer and winter prey

Some studies have reported a change in species composition in the diet of wolf between summer and winter because of easier access to some, often smaller, prey species during the summer, e.g. beaver at Isle Royale (Peterson 1977), whereas other studies have found no notable changes in species composition between seasons, e.g. Peterson et al. (1984). Despite the fact that both beaver and badger, which both are less active during winter, occurred in the territory, we could not detect any significant differences between summer and winter in their occurrence in the diet. Neither could we find any substantial seasonal difference in the proportion of moose and roe deer consumed, although there were probably more roe deer in the territory during summer, due to more pronounced migrations of roe deer in winter towards areas of lower altitude outside the territory. This lack of response may be ascribed to aggregation of overwintering roe deer, e.g. around settlements, thus perhaps making them easier to find despite lower numbers, and that the snow made them comparatively easy to kill in winter compared to moose.

Badgers in the diet

To our knowledge, this is the first study showing that badger constituted a significant part of the diet of wolves; about 8% by mass and 23% by number, corresponding to approximately 45–90 badgers taken by five wolves annually. The relatively large proportion of hair from badgers in wolf scats also in winter did not correspond to the kills found when snow-tracking the wolves; in fact we had only a few indications of wolves hunting for badgers, but no kills were found. The most important reason for this is probably that badgers hibernate when conditions for snow tracking are good, which was only in parts of the mild winters during this study, but another reason could be its smaller size (see next section). We have no estimates on the density of badgers in the territory, and therefore no conclusions on the effect of wolf predation on the population of badgers can be drawn. However, the size and distribution of the badger population in Scandinavia have increased dramatically during the last 150 years, and one reason for this, besides possible climatic changes, may be the corresponding decrease in wolf numbers (Bevanger & Lindström 1995).

Prey found in scats vs tracking

When comparing number of kills found by tracking and the proportions suggested by the scat analysis, a discrepancy between small prey species and moose was consistent for all species. For example, 65 kills of moose but only 7 kills of roe deer were found even though roe deer constituted the largest proportion by number according to our scat analysis. We suggest that this was due to the following methodological reasons: (i) little, or nothing, is left of a kill of a small species, which therefore is more difficult to detect, especially after a snow fall, (ii) the wolves probably stayed for a shorter time at a smaller kill, resulting in fewer tracks in that area, (iii) the killing of small prey species does not attract ravens Corvus corax, which were used as indicators when searching for kills, to the same degree as the killing of a large prey species, (iv) we probably tracked wolves in groups proportionally more often than loners, and the former may kill a higher proportion of moose, and (v) to some extent we avoided tracking wolves around settlements, where roe deer seemed to aggregate in winter, and therefore we probably were less likely to find kills of roe deer.

Sex and age of moose killed by wolves vs hunters

Our comparisons of the age and sex of wolf-killed moose to those of hunter-killed moose showed no big differences, even though these mainly refer to the winter period because most kills were found during this season. The only deviation was that the wolves killed a higher proportion of females and no males older than two years. One must, however, bear in mind that both sex ratio and age distribution of the moose population were changed by hunting on beforehand, which influenced what was available for both wolves and hunters, and that the sample killed by hunters probably was biased due to recommendations and preferences by the hunters.

Moose population trend and effects of predation

The number of 900 moose in the territory indicated by our census is a minimum estimate of the annual mean number because it reflects the mean number during winter, thus mainly after the autumn moose hunt (roe deer hunting occurred much less extensively). Hence, when comparing the number of moose in the territory with the estimated number of 45–90 moose killed annually by five wolves, our conclusion is that they killed less than 10%, probably closer to 5%, of the moose population each year. A large proportion of the wolf-killed ungulates were young-of- the-year; more than 50% in moose, probably somewhat less in roe deer.

Hence, the slight decrease in the moose population, found during the last year of our census, is more likely to be explained by other factors, such as hunting quotas, food availability, and/or reproductive success, than by wolf predation. Nevertheless, to be able to maintain a wolf population with the present territory size (see below) in a long-term perspective without a decrease in the density of moose, a slight reduction in hunting quotas, probably around 10–20%, would be appropriate.

It is important to note, however, that there were no other wolf territories in the region. If the wolf population becomes saturated it is likely that the size of the territories will decrease considerably (see e.g. Fritts & Mech 1981). This will result in higher densities of wolves and higher predation on moose. Consequently, there is no serious present conflict with the moose hunters, but there is a possibility of a future conflict.

Roe deer population trend and effects of predation

The tendency of our census to underestimate ungulate density was probably more pronounced for roe deer than for moose. The reasons are (i) that roe deer to a greater extent than moose migrated from the wolf territory during the winters because they are more negatively affected even by moderate snow cover, and (ii) we probably missed areas with aggregations of roe deer because the census was based on relatively few sampling areas, whereas moose were more evenly distributed in winter and therefore more correctly estimated. Moreover, the fact that the sampling area for roe deer pellets covered 40% of the area for moose pellets (see Methods), made the estimate of roe deer density less robust compared to that of moose. Our conclusion is, therefore, that the annual mean number of roe deer was probably significantly higher than the mean number in winter of 240 roe deer estimated in our census. Hence, it is difficult to assess the impact of wolves, which are estimated to kill 90–180 roe deer annually, on the roe deer population. However, it seems obvious that the annual roe deer harvest of wolves constitutes a larger proportion of the roe deer population than the annual moose harvest of the moose population.

Despite this probably significant predation pressure, our ungulate census indicated a threefold increase in the roe deer winter density. One might claim that the wolf territory was a ‘sink area' for the roe deer population and that neighbouring areas were ‘source areas', supplying the territory with roe deer. However, the situation was complicated by an increasing population of lynx in the region, its density at the end of the study perhaps being as high as 0.5/100 km², and their main prey appeared to be roe deer as well (Liberg & Glöersen 1995, pers. obs.). This suggests that predation by wolves and lynx was more than compensated for by positive factors for the roe deer population, such as the absence of the red fox (due to sarcoptic mange epidemic), and a series of mild winters.

Moreover, we expected to find a functional response from the wolves, i.e. an increasing proportion of roe deer in the diet due to the population increase. However, the proportions of hair from ungulates in wolf scats were relatively stable throughout the study period. We have no explanation for this lack of response, but it may be ascribed to the 'hunting conservatism' of wolves which, has been proposed in other studies (see e.g. Okarma 1995, p. 375).