NATIONAL PARK SERVICE Wolf Ecology and Prey Relationships on Isle Royale

Opportunities to study the large, native predators of North America in an undisturbed situation are quite rare today, for these animals have been eliminated from most of their former range, at least in the continental United States. Persecution of the gray wolf (Canis lupus) has accompanied the spread of modern civilization into every frontier on the continent. Today, wolves are found in the United States only in Alaska, northern Minnesota, and Isle Royale, with perhaps a handful in upper Michigan and in remote areas of the mountainous West. In Canada, wolves still inhabit most of the provinces and territories where the land has not been altered for agriculture or livestock production.

Isle Royale has not always been known for its wolf population. The wolf is a relative newcomer to this island, having arrived in the late 1940s (Fig. 1). For decades prior to their arrival, the island was famous for its moose herd (Alces alces), the increase and decline of which attracted nationwide attention (Fig. 2).

Fig. 1. Isle Royale has supported wolves since the late 1940s.

Fig. 2. Moose colonized Isle Royale in the early 1900s.

In addition to concern for the welfare of the moose population, there was great public interest in the other natural features of the island. Although the island had experienced periodic attempts at copper mining and associated disturbances in the form of man-caused fires and logging in the 1800s, by the third decade of this century forest regrowth had restored the landscape. The need to preserve this unique island was clear by the late 1920s, and Isle Royale was established as a national park in 1940.

The island has been the focus of many geological, botanical, zoological, and archaeological studies since the early part of this century. After wolves colonized the island, biologists were presented with a predator-prey system uncomplicated by a great number of interacting species and, importantly, an ecosystem set aside by man for preservation of its natural features.

A long-term study of wolf-moose relationships on Isle Royale began in 1958 under the direction of Durward L. Allen of Purdue University. Individuals with principal responsibility for field studies in wolf-moose ecology were L. David Mech, 1958-62; Philip C. Shelton, 1962-63; Peter A. Jordan, 1964-66; and Michael L. Wolfe, 1967-70. My work began in 1970.

In this monograph, previous published work on the Isle Royale wolves is briefly reviewed. Data gathered on the wolf population from 1971 through 1974 are presented in detail. I also have attempted a synthesis of moose mortality data recorded in systematic fashion from 1958 through the winter study of 1974 to provide a composite picture of mortality patterns.

Study Area

Isle Royale is actually an archipelago, distinguished by a single large island 72 km long and 14 km at the widest point. Many small islands extend off the peninsulas of the main island. The closest Canadian island is 20 km from Isle Royale, and the Canadian mainland is 24 km distant. Although Isle Royale is part of Michigan, it is closer to the northeastern tip of Minnesota (Figs. 3, 4).

Fig. 3. Map showing location of Isle Royale. (click on image for a PDF version)

Fig. 4. Map of Isle Royale, showing locations mentioned in text. (click on image for a PDF version)

Geology and Physiography

Recent geological investigations of Isle Royale (Huber 1973a, b, c; Wolff and Huber 1973) provided the basis for this brief account of the formation and subsequent reworking of the bedrock strata that form this island.

The oldest rocks found on Isle Royale are Keweenawan lavas, which probably erupted a little over a billion years ago from what is now the middle of the Lake Superior basin. Occasionally, sedimentary deposits washed in from surrounding regions as the areas slowly subsided, resulting in interbedded volcanic and less-resistant sedimentary strata. Continued subsidence created the basin which now contains Lake Superior and tilted the bedrock strata that form Isle Royale toward the southeast (Fig. 5).

Fig. 5. Cross-section through Lake superior, Isle Royale, and Keweenaw Peninsula (Huber 1975). (click on image for a PDF version)

Preglacial stream erosion removed portions of the softer strata running the length of present-day Isle Royale. Glacial ice further scoured these valleys, resulting in the longitudinal ridge-and-valley topography that is characteristic of Isle Royale (Fig. 6). The final glacial retreat occurred during the Valderan period, about 11,000 years ago. The ice withdrew in a northeasterly direction, exposing the western end of the island first, then halted for a period of time just west of Lake Desor. Consequently, significant deposits of glacial debris can now be found at the southwestern end of the island, while the more rapid retreat of ice from the northeastern end left little material behind. Further retreats of the glacier opened up successively lower outlets for the large, postglacial lakes that formed at the edge of the ice, and lake levels dropped, exposing more of Isle Royale. Coupled with the fall of lake levels has been a slow postglacial rise of the bedrock. The highest elevation on the island is 238 m above Lake Superior, which lies 181 m above sea level.

Fig. 6. Characteristic ridge and valley topography of Isle Royale.

The soil mantling the northeastern two-thirds of the island is, for the most part, thin and azonal, with many sloping ridges devoid of soil except in depressions. Exposure of bedrock on the southwestern third of the island is much less frequent, and soils are somewhat deeper and more developed, primarily because glacial deposits at this end were greater.

Climate

Surrounded by the world's largest body of fresh water, Isle Royale's climate is greatly modified relative to nearby mainland areas. The principal effects of Lake Superior are to buffer temperature changes and increase precipitation and humidity over the island (Fig. 7).

Fig. 7. Lake superior is responsible for high humidity and frequent fog.

TEMPERATURE

Isle Royale temperatures are generally cooler in summer and warmer in winter than those of the mainland, with less range in temperature extremes. Temperatures of -34°C (-30°F) have been recorded only three times in the last 14 years, with the record -37°C (-35°F) recorded in 1972. The mean number of days per year with a maximum temperature over 32°C (90°F) is only two along the north shore of Lake Superior (U.S. Department of Commerce 1968); however, further inland in Minnesota, extreme temperatures over 32°C and below -34°C are not uncommon.

Temperatuers during the annual 7-week winter study are generally above -18°C (0°F) (Fig. 8), and thaws in midwinter are frequent. The years 1964 and 1973 stand out as unusually warm, while 1967 and 1972 were, on the average, the coldest recorded since measurements began in 1962.

Fig. 8. Winter temperature records for Windigo, Isle Royale National Park.

PRECIPITATION

Most of the weather systems that move over Isle Royale come from the northwest, and moisture is picked up as they cross Lake Superior. Those storms which do come from the east bring more moisture toward Isle Royale and the Minnesota shore of Lake Superior. Precipitation in northeastern Minnesota, closest to Lake Superior, is considerably above that of northwestern Minnesota. Precipitation on the south side of the lake, in upper Michigan, is even greater, since this area receives moisture picked up by prevailing winds (Table 1). Annual precipitation for Isle Royale has not been measured on a year-round basis but is probably intermediate between northeastern Minnesota and upper Michigan.

TABLE 1. Mean annual precipitation in mainland areas in Minnesota and Michigan (1931-55)^a

Snowfall. Mean annual snowfall on the north shore of Lake Superior (1931-60) was approximately 150 cm, while the Keweenaw Peninsula in upper Michigan received in excess of 250 cm. The average snowfall for Houghton, Michigan, was 452cm from 1931 to 1960 (U.S. Department of Commerce 1968). In recent years, when water content of snow was measured on Isle Royale during the annual study period, precipitation (ca. January—March) on the island was 55-60% more than the nearest mainland stations in Minnesota, and about 40% less than Houghton, Michigan, on the south side of the lake (Table 2).

Table 2. Winter precipitation (cm) on Isle Royale and nearby mainland weather stations.

Peek (1971a), working in northeastern Minnesota, found that snow depths were greater and crusts stronger in the part of his study area closest to Lake Superior. On Isle Royale, mild temperatures may produce crusting conditions that do not exist in colder, mainland areas. In all three winters that I have spent on the island, there were well-developed crusts either within or on top of the snowpack for a sufficient length of time to affect either wolf mobility or moose habitat preference.

During the span of wolf-moose studies on Isle Royale, trends in snowfall and snow depth have been generally upward, though not without great year-to-year variation. Measurements of snow depth on Isle Royale during the winter period have not been taken consistently, so Grand Marais weather data were used to show trends. These data indicate above-average snow depths for the mid-1960s and for 1969 through 1972 (Fig. 9). A snow-depth index was plotted for this station, utilizing maximum depths recorded in each month from November through May. These depths were used to construct a graph of maximum snow depths for each winter since 1958-59 (Appendix A). The area under the curve was determined using a grid overlay (Bishop and Rausch 1974). While not corresponding precisely to conditions on Isle Royale, this index can be used to indicate year-to-year variations in snow depth.

Fig. 9. Snowfall and index of snow depth at Grand Marais, Minnesota, 1959-74.

Isle Royale snow studies. In addition to snow depth, other physical characteristics of snow were measured to interpret more adequately the effects of snow on moose and wolf mobility and behavior. The support quality of the snow was measured in 1972 and 1973 with a compaction gauge similar to that developed by Verme (1968), and in 1974 with a Swiss Rammsonde penetrometer. The Ram penetrometer gave more satisfactory results when crusts were present on or near the snow surface. Both instruments rely on vertical force to penetrate the snowpack although they differ somewhat in design and use (Appendix B). Snow depths and penetrability readings were taken frequently at three sites differing in canopy coverage. Density and hardness within the snow profile were measured from one to three times during each winter study according to the method of Klein et al. (1950).

Snow depths varied considerably according to the type of overhead canopy and exposure to the wind, and were greatest in open areas and least under thick conifer canopies. Snow depths in an open area protected from the wind were consistently at least 20-25 cm greater than a nearby site beneath a thick canopy (Fig. 10).

Fig. 10. Snow depth at three sites on Isle Royale, 1972-74, illustrating annual variations in snow levels and the effect of overhead canopy.

Overhead canopy also affects maturation of the snow. The effects of sun and wind, principally responsible for the formation of crusts and settling of snow, are reduced considerably by a dense overhead canopy. Usually, crusts beneath conifer canopies are less well developed, and vertical hardness is almost always lower in canopied areas (Kelsall and Prescott 1971:134-137). The support quality of such snow, of course, is considerably less (Figs. 11 and 12).

Fig. 11. Penetrability (determined by compaction gauge) of snow at three sites on Isle Royale, 1972-73.

Fig. 12. Ram hardness number (R) (Appendix B) of snow at three sites on Isle Royale, 1974.

An important exception to the general rule of softer snow beneath conifers occurs during thaws, when snow remaining on conifer canopies (called "qali", Pruitt 1958) melts and drips onto the snow surface below. This occurred in early March 1973 and 1974, resulting in a strong surface crust and increased snow density over the entire island, especially below conifers. Crust strength in these situations was quite variable throughout the day, depending on the temperature.

A small number of snow profiles were studied by digging a pit to ground level. Physical characteristics of each distinct layer of snow were recorded. These snow-profile analyses revealed that the density of fresh snow on Isle Royale was usually between 0.06 and 0.10 g/cm³, and as the snow matured, its density increased accordingly. Density of well-aged snow on Isle Royale ranged from 0.30 to 0.35 g/cm³. These changes in snow density are evident in the top layers of the snow profiles shown in Appendix C.

Vegetation

Isle Royale is almost entirely forested. Fire history, glacial deposits, and a "washboard" topography all contribute to a rich mosaic of tree species. Because the island lies in a transition zone between the boreal and the northern hardwood forests, elements of both grow here. Typical boreal tree species are found near the Lake Superior shoreline, where atmospheric moisture is greater and temperatures during the growing season are lower and less variable than further inland (Linn 1957). These include white spruce (Picea glauca), balsam fir (Abies balsamea), white birch (Betula papyitfera), and aspen (Populus tremuloides). Coniferous vegetation normally succeeds deciduous species in this forest type, leading to extensive stands of spruce and fir (Fig. 13).

Fig. 13. Spruce-fir forest is typical of Isle Royale shorelines.

At higher elevations in the interior of the island, the principal tree species are sugar maple (Acer saccharum) and yellow birch (Betula allegheniensis), common elements of the northern hardwood forest (Fig. 14). In areas between the two principal climax types, Linn (1957) found hardwoods dominant on dryer sites and conifers dominant wherever soil moisture was sufficient for their existence.

Fig. 14. Yellow birch—sugar maple forest in island's interior.

Hansen et al. (1973) described stand composition in detail for the principal forest types found on Isle Royale. Krefting et al. (1970) delineated these in a vegetation map (Fig. 15).

Fig. 15. Vegetation map for Isle Royale, modified principally from Kefting et al. 1970. (click on image for a PDF version)

The forest designated as aspen-birch-fir-spruce is typically composed of old-growth, deciduous stands with ample coniferous reproduction. Spruce and fir dominate such stands, particularly at the northeast end. This forest type, usually found along the lakeshores of the island, now provides the most important winter habitat for moose on the island.

Areas designated as aspen-birch are also old-growth stands that emerged after fires in the previous century. Coniferous reproduction is relatively sparse both here and in sugar maple-yellow birch stands.

About 20% of the island (10,400 ha) burned in 1936; 576 ha within the 1936 burn, south of Lake Desor, burned again in 1948 (Krefting 1974). Regeneration, mainly aspen and birch, followed these burns, although today there are many other species resulting from different intensities of burn and perhaps remaining seed sources (Fig. 16). Cedar (Thuja occidentalis) is common within areas of the 1936 burn in lowlands west of Siskiwit Bay. In the center of the island, solid stands of aspen and birch now have grown out of reach of moose, and there is little coniferous reproduction. White spruce is common in the 1936 burn east and south of Lake Richie, with stands interspersed with large areas of bare rock outcrops.

Fig. 16. Fires and subsequent erosion create thin soils on ridgetops.

Mammalian Fauna

Isle Royale is sufficiently far from the mainland to isolate it effectively from colonization by many terrestrial species. Sixteen species of mammals occur presently on Isle Royale (Table 3); Johnsson and Shelton (1960) list an additional 30 mainland mammals that have never been recorded on the island (Figs. 17, 18, 19). Coyotes (Canis latrans) and woodland caribou (Rangifer tarandus) inhabited the island within the current century but are not present now. The Norway rat (Rattus norvegicus) and white-tailed deer (Odocoileus virginianus) were introduced but did not persist on the island. The long-tailed weasel (Mustela frenata) and red-backed vole (Clethrionomys gapperi) were recorded on the island in the early 1900s, but Johnsson and Shelton (1960) considered these records doubtful. Marten (Martes americana) were also said to occur in the early 1900s, but evidence is sparse.

TABLE 3. Mammals currently present on Isle Royale.

Fig. 17. Woodland deermouse, the smallest rodent species on Isle Royale.

Fig. 18. Snowshoe hare, primary prey of the red fox on Isle Royale.

Fig. 19. A unique subspecies of red squirrel inhabits Isle Royale.

Of the four bat species listed for the island, I identified only the little brown bat (Myotis lucifugus) between 1970 and 1974. Johnson and Coble (1967) found positive evidence of the red bat (Lasiurus borealis) in castings of pigeon hawks in 1966, providing the only island record for this species. Of the remaining 12 species of mammals, I observed all but the lynx (Lynx canadensis) and short-tailed weasel (Mustela erminea). Weasel tracks were seen each winterfrom 1971 to 1974. A cat, judged by its tracks to be a lynx, was seen by Donald E. Murray and Ranger Zeb McKinney in winter 1970 (Murray, pers. comm.). If lynx are still present on the island, they are very rare.

The status of the otter (Lutra canadensis) was uncertain in the early 1960s (Mech 1966), but well-distributed signs of otter were seen regularly each winter 1972-74. This species appears to be a recent colonizer of Isle Royale that has increased steadily.

Methods of Study

Field work was concentrated in two periods: 7 weeks in midwinter and roughly 6 months from May to October (Appendix D). While field techniques varied seasonally, the emphasis at all times centered about increasing our knowledge of wolves and wolf-moose relationships on Isle Royale. Wolf study methods were basically observational, with heavy reliance on aerial tracking in winter and reading wolf sign in both summer and winter (Fig. 20). Carcasses and skeletons of moose were examined whenever possible. Size and composition of the moose population were estimated by aerial methods.

Fig. 20. Pilot Donald E. Murray.

Winter Wolf Study

The winter study extended from late January into mid-March, which is normally the period when ice in the harbors and lakes is sufficient to land a ski-plane. Headquarters for the annual winter study were at Windigo, at the southwest end of the island. Knowledge of wolf movements came only from aerial tracking. Shorelines of the island and inland lakes (preferred wolf travel routes) were followed from the air whenever possible to locate wolf tracks. We tried to maintain a complete travel record for the main packs by backtracking to the location of the previous observation. Small groups and single wolves were located whenever possible by checking old kills of the large packs.

When time permitted, we circled wolves at an elevation of several hundred feet and observed them with binoculars, recording data on a portable tape recorder. From 1972 through 1974, I observed wolves from the air for 30 hours. Ground observations were possible at carcasses of necropsied moose at Windigo. Identification of individual wolves was based on body markings; the relative fullness of the tail and variations in the dorsal spot and tip of the tail were the most useful characteristics. The number of individual wolves identified varied from year to year and depended on the distinctiveness of wolf body markings and the amount of time available for observation. Usually, identification from one day to the next, and especially from one year to the next, was limited to the alpha or dominant male and female in each pack. Photographs proved helpful in year-to-year identification of alpha wolves.

Captive wolves were observed at the Chicago Zoological Garden, Brookfield, Illinois, to aid interpretations of wolf behavior in the wild. Behavior patterns were the sole means of sex determination for Isle Royale wolves. Both males and females often exhibit the same or only slightly different behavior, so one must be cautious when drawing conclusions. For example, while males may lift a hind leg when urinating, females occasionally do likewise. The only urination posture that is unambiguous is the male's leaning forward, with back legs outstretched (Fig. 21). If the anal region of a wolf is examined frequently by other wolves during the breeding season, it is safe to assume that the animal examined is a female. The examining wolf is not necessarily a male however, since female pups at Brookfield Zoo exhibited the same behavior. In each pack,the dominant, or alpha, male and female were distinguished by the fact that they never displayed submissive behavior toward other wolves, while other wolves were submissive in their presence.

Fig. 21. East Pack alpha male (1974) urinating in posture unique to males; alpha female stands behind him.

Identification of pups in midwinter presented a problem. Sometimes pups could be distinguished by their behavior. Pups seemed to be more playful, were sometimes hesitant in situations such as crossing glare ice, and behaved inconsistently toward dominant wolves during group greetings. Sometimes pups were distinguished by a lighter, uniform color, slim body, and guard hairs that stick up along their back, creating a scraggly outline (Fig. 22). At other times, however, pups and adults are indistinguishable. For example, in 1974 we examined the carcass of a wolf that appeared to be an adult. Unworn canine teeth and lack of epiphyseal closure in the radius (Rausch 1961), however, showed that it was actually a pup. Underdeveloped pups may be more pup-like in appearance (Van Ballenberghe and Mech 1975), leading one to conclude that pups are plentiful and show high survival in years when actually they were identified only because they were underdeveloped. Thus, the number of pups distinguished by appearance and/or behavior could not be used as an annual index to pup production and survival.

Fig. 22. Pups may appear scraggly, as three wolves on left, or may be indistinguishable from adults.

Winter Moose Census

The present phase of the Isle Royale studies was initiated in 1970 with a firm foundation of aerial moose-inventory techniques which have been steadily refined. The first aerial censuses on Isle Royale used a transect, or strip, method, in which all the moose seen in midwinter during many parallel flight lines were tallied (Aldous and Krefting 1946) (Fig. 23). Cole (1957) and Mech (1966) employed a variation of this strip method by "buzzing" moose that were spotted in order to flush any nearby animals. Mech's 1960 census was the last attempt at a complete count of the Isle Royale herd. Weather factors and moose-distribution patterns in subsequent years were such that stripwise coverage of the island would have produced a far less meaningful estimate. In addition, confidence intervals cannot be constructed for such an inventory.

Fig. 23. Aerial counts were used to determine moose population size and composition.

Although Jordan (unpubl. data) employed a strip census method in 1964, it was discontinued thereafter in favor of intensive aerial searching of small plots randomly distributed over the island. After 1965, the island was divided into several zones of relative moose densities. The plots occurring in each zone were used to calculate mean moose density and variance for that zone (Wolfe and Jordan, unpubl. data).

I used a similar aerial-plot method. Since the population estimate depends on the area included in each zone of moose density, it is important that the zone assignments accurately reflect current moose distribution. For this reason I made some modifications in zones in both 1972 and 1974.

Winter moose censuses in both years were flown in a 90-hp Aeronca Champion. We flew overlapping circles a few hundred yards wide over a given plot until it had been covered completely. Most plots were flown at an altitude of 100-200 m; the "open" character of burned-over areas permitted a censusing altitude of about 250 m. Moose densities were highest in coniferous cover where visibility was poor, so these areas were circled most intensively. Counting in both years continued throughout February. Although the counting period was long, no noticeable variations in moose distribution invalidated the zone assignments. Within each zone, we attempted to sample plots randomly, but in practice this plan was modified considerably by the weather and a need to minimize time spent flying from one plot to another. Plots were counted in areas where flying and counting conditions were near optimum: little or no wind (less than 10 mph), and a high overcast to eliminate shadows. Allocation of sampling among the zones was such that sampling error would be minimized at whatever point the work was terminated. Censusing in both years was discontinued early in March after moose were confined to areas of conifer cover by strong crusts on the snow surface.

ACCURACY OF AERIAL MOOSE CENSUS

Studies in Alaska show that observer experience, number of observers, snow conditions, time of day, terrain, and type of aircraft could all affect moose counts (LeResche 1970; LeResche and Davis 1971; LeResche and Rausch 1974). Inexperienced observers saw less than half of the moose in experimental enclosures, even under good conditions. In nine counts by an experienced observer, with the pilot participating, the average proportion of moose seen was 70%; the highest accuracy attained in such a count was 87%. The flight patterns were either narrow transect or concentric circles of ever-decreasing radii.

Compared to the relatively open cover of the Alaska study area, Isle Royale has large areas of dense conifer cover which make moose counting more difficult. In addition, moose concentrate in these conifer areas, further reducing accuracy. To compensate somewhat, we spent a greater amount of time circling each plot. The Isle Royale plots were kept small (1-2 km²) to reduce observer fatigue and resultant errors.

An accuracy of 80% was assumed for the 1972 and 1974 counts. Factors responsible for the assumed high level of accuracy are the flight pattern of overlapping circles, the intense circling, and the long experience of the pilot.

Aerial moose counts probably can do little more than provide a rough index of trends in the population. In addition to the above problems, the uneven nature of moose distribution contributes to a high sampling error; 95% confidence intervals are generally in excess of 20% of the estimate (Evans et al. 1966; Mantle 1972:124-137; Peek 1971a).

Summer Ecology of the Wolf

Most of the spring-fall field work consisted of ground searches for moose remains and wolf sign (Fig. 24). Since wolves made extensive use of the 270km of hiking trails during spring and fall, when visitation is lowest, we also walked the trails at these times monitoring wolf sign and checking for the occasional kill made on a trail. When human activity caused the wolves to abandon the trails, we generally did likewise. Open ridges, shorelines, creek beds, swamp edges, and animal trails then became our travel routes.

Fig. 24. Summer field headquarters on Rock Harbor.

Howling responses were sometimes useful in locating wolves. Human imitations of howls were frequently broadcast over the island in 1971 and 1972 through a portable megaphone with electronic amplification. However, success was limited because we could move about only on foot. By checking known travel routes and reports of howling, we finally located wolf activity areas (den and rendezvous sites) in the summer of 1973. The presence of wolves at summer homesites was then monitored by camping near enough to hear spontaneous howling. In order to minimize disturbance to the wolves, human howling was employed to elicit wolf responses only when the location of wolves was uncertain.

We collected a large sample of wolf droppings (scats) in summer 1973, to determine food habits at this time of year. All samples were autoclaved to avoid contamination by Echinococcus, a tapeworm whose eggs are passed with feces. Scats were pulled apart under a binocular scope, and the incidence of prey remains in each scat recorded. Hair was examined under reflected light and magnification. Identification was aided by photographs and descriptions provided by Adorjan and Kolenosky (1969).

Moose Mortality Patterns

The collection of skeletal remains of Isle Royale moose over a 16-year period (1958-74) provided a unique opportunity to investigate mortality patterns in a naturally regulated moose herd and to document changes in the type of moose killed by wolves from one year to the next. In addition to the winter-spotted kills, other carcasses and skeletal remains located at random during summer field work provided data on year-round mortality (Fig. 25). Information on date and cause of death, age, sex, skeletal abnormalities, and marrow condition was recorded whenever possible.

Fig. 25. Extensive carcass examinations revealed moose mortality patterns.

Since 1959, a major effort has been made each winter to locate wolf kills, primarily by following wolf tracks. Sometimes this has led to carcasses of old kills or moose that died of causes other than wolf predation. Almost all of these dead moose were subsequently ground-checked, either in winter or as soon as possible the following spring. Of 141 carcasses located in winter (1971-74), 136 were examined from the ground. An additional 292 carcasses or skeletal remains were examined in summer (1970-73).

The autopsy record file was the primary source of material for the discussion of mortality patterns. Information from the entire collection (836 known-age moose) was reviewed and coded to utilize computer-sorting. During the most recent phase of the project (1970-74), age was determined for 404 moose (of the 428 remains examined).

An approximate date of death was assigned to each dead moose according to the degree of decomposition or weathering. The presence of hair, rumen contents, hide, and various decomposing organisms indicated death within a year, usually the previous winter. Bits of dried flesh rarely persist for more than 2 years, especially in wet sites where decomposition proceeds at a more rapid rate. Bones often remain on the ground on Isle Royale for a decade or more, since none of the rodent species on the island gnaw bones other than antlers to any great extent, and weathering is very gradual (Fig. 26). Remains older than 2 years were placed in one of three groups: 2-5 years, 5-10 years, or more than 10 years, and in most cases an approximate 5-year period covering the estimated time of death; e.g. 1960-64, 1965-69, 1970-74. Since I often found bones from moose that had been examined by project personnel in previous years, there was ample opportunity to check this method of estimating time of death against old remains with a known date of death.

Fig. 26. Skeletal remains of moose may remain for decades.

When possible, a cause of death was assigned. Bones that were heavily chewed and widely scattered indicated that wolves had fed on the carcass. Such remains were classed as probable wolf kills because winter observations indicated that wolves on Isle Royale killed most of the moose upon which they later fed.

Calves were aged according to tooth eruption criteria; all adults were aged by counting cementum annuli in a section of polished upper or lower molar. Wolfe (1969) believed that a conservative estimate of the reliability of this technique would be ±2 years. But, if cementum annulations are consistent and clearly visible, accuracy is much improved (Fig. 27). I have found some variation in the clarity of annulations in a few specimens; for such moose age estimates probably should be ±2 years, as Wolfe suggested. Cementum annulations do not seem to deteriorate under field conditions; I have had little difficulty counting lines in teeth from moose that were estimated to have died more than 20 years ago.

Fig. 27. Annual cementum deposits in moose teeth indicate their age. Cross-section of molar appears in (a), while magnified view of cementum is show in (b).

Skeletal remains were sexed on the basis of the presence or absence of antlers or antler pedicels. The stage of antler development also was used to determine the season of death for males. All available bones were checked for abnormalities, and all of the teeth collected. Since 1971, a metatarsus has been collected for use as in Index to body size and development.

Fat content of bone marrow is an indicator of the condition of an animal, since the fat stored in marrow of the leg bones is usually the last to be utilized by an animal suffering from malnutrition (Cheatum 1949; Bischoff 1954) (Fig. 28). Because many of the bones from moose killed on Isle Royale were dehydrated when collected, visual descriptions of marrow were used instead of actual fat content (percent dry weight). Visual estimates were compared to actual fat content (expressed as grams of fat/cm³) for 143 leg bones from 49 moose; such estimates were only accurate enough to assign marrow samples to three broad categories (Table 4). Bone marrow, examined before the present study, was assigned to one of three classes based on marrow descriptions recorded on autopsy cards. Serious fat depletion probably was present in marrow samples from class 3 (lowest fat content), with an occasional specimen from class 2.

TABLE 4. Classification of fat content of moose bone marrow.

Fig. 28. Bone marrow is one of the last fat reserves to be utilized by malnourished moose. This is a fat-depleted example.

Moose Population Structure

Aerial counts were conducted after leaf fall in October 1972-74 determine the sex and age structure of the moose herd. An additional index of herd composition was obtained using records of moose seen during summer ground work.

During autumn aerial counts a 75-hp Piper J-3 was used. We did not use fixed plots, nor did we attempt complete coverage of the area flown. A high overcast produced the best conditions for observing moose, since sun created shading problems and also seemed to reduce moose activity. Peek (1971a) suggested that fall composition counts could be biased by the tendency of bulls to congregate in the rut and postrut periods, and sometimes by the preference of cows with calves for heavier cover. I attempted to minimize these biases by sampling a wide variety of habitats and by flying overlapping circles.

Moose were sexed by the presence of antlers in males or a white vulva patch in females (cf. Mitchell 1970) (Fig. 29). Moose that did not have visible antlers were inspected on repeated low passes until the presence or absence of a vulva patch was confirmed. Calves were distinguished by relative head and body size (Mech 1966). Bulls with spikes or small forks were considered yearlings. The proportion of yearlings may be somewhat underestimated if there are many yearlings with antlers larger than forks, although this would be offset by 2.5-year-old moose with small antlers.

Fig. 29. A cow showing white vulva patch characteristic of females, plus mounting marks of bulls during rut.

During ground coverage of the island in the spring-fall period, a record was kept of all moose observed to provide an index of calf abundance. Calves probably are underestimated by this method because cows tend to hide young calves (Pimlott 1959). However, these observations provided both comparative information for fall aerial counts and an index of moose productivity (through occurrence of twins).

Statistical Analysis

Unless otherwise noted, statistical procedures followed Sokal and Rohlf (1969). The significance level for rejection of the null hypothesis was 5%. The P value provided refers to the value obtained in specific cases. A "heterogeneity G-test" was used to detect differences in age distributions of subsamples in the moose-mortality data. In this test, a calculated G value is compared to a critical chi-square value to determine statistical significance. Differences between two percentages were analyzed with the aid of a test statistic (t_s).