# Modeling Wolverine Occurrence Using Aerial Surveys of Tracks in Snow

> How do you find a rare, wide-ranging carnivore across 60,000 km² of northern forest? Flying low in a small ski plane, an experienced team read wolverine trails in the snow, following gait changes over long stretches to tell them from fisher tracks. Combined with a statistical model that allows for missed tracks, their 26 days of flying produced an objective map of where wolverines live in northwestern Ontario, Canada.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2007-magoun-ray-modeling-wolverine-occurrence-aerial/

## Citation and access

- **Authors:** Audrey J. Magoun, Justina C. Ray, Devin S. Johnson, Patrick Valkenburg, F. Neil Dawson and Jeff Bowman
- **Year published:** 2007
- **Journal or source:** The Journal of Wildlife Management
- **Full citation:** Magoun, A. J., Ray, J. C., Johnson, D. S., Valkenburg, P., Dawson, F. N. and Bowman, J. 2007. Modeling wolverine occurrence using aerial surveys of tracks in snow. Journal of Wildlife Management 71(7): 2221–2229.
- **DOI:** 10.2193/2006-372
- **Access status:** Publisher link only
- **Where to find it:** https://doi.org/10.2193/2006-372

## Tags

- **Species:** Mammals > Wolverine
- **Region:** North America > Canada
- **Tracking:** Gaits; Sign surveys; Snow tracks; Tracking; Tracks; Tracks and signs identification; Trailing
- **Behaviour and ecology:** Behaviour
- **Methods and technology:** Aerial surveys; Bayesian statistics; Detectability; Distribution mapping; Geographic Information System (GIS); Global Positioning System (GPS); Markov chain Monte Carlo (MCMC); Occupancy modelling; Radio telemetry; Simulation
- **Data type:** New field data
- **Observer qualification:** Reported but not validated
- **Discipline:** Behaviour and ecology; Wildlife tracking
- **Year published:** 2007

## Detailed keywords

- **Species:** Studied: wolverine (Gulo gulo). Other mammals whose tracks in snow could be seen from the survey aircraft (not studied): moose (Alces alces), caribou (Rangifer tarandus), white-tailed deer (Odocoileus virginianus), wolf (Canis lupus), coyote (Canis latrans), red fox (Vulpes vulpes), lynx (Lynx canadensis), river otter (Lontra canadensis), fisher (Martes pennanti), marten (Martes americana), mink (Neovison vison), beaver (Castor canadensis) and snowshoe hare (Lepus americanus). Fisher tracks were the most similar to wolverine tracks.
- **Location and habitat:** About 60,000 km² of the Boreal Shield Ecozone, northwestern Ontario, Canada (49 to 51° N, 90 to 96° W), mean elevation 390 m. Boreal forest of white and black spruce, jack pine, balsam fir, tamarack, birch, trembling aspen and balsam poplar; water bodies about 20% of the area; open fens and bogs, sparse forest, recent cutovers and burns; roads, logging and mining mostly in the southwest. Weather data from the Red Lake weather station.
- **Topics:** species distribution, extent of occurrence, area of occupancy, core areas, aerial track surveys, snow tracking from aircraft, track identification from the air (track pattern, gait changes, body prints, behaviour), telling wolverine tracks from fisher tracks, imperfect detection, repeated surveys, survey effort, seasonal change in detection, spatial autocorrelation, baseline for monitoring, threatened species, resource development, fur-trapping records.
- **Statistics or models:** Hierarchical spatial occupancy model in a Bayesian framework: occurrence modelled with a conditionally autoregressive (CAR) latent spatial process and a detection model with a covariate for surveys on or after 15 February. Markov chain Monte Carlo (MCMC) in OpenBUGS 2.2.0 through BRugs in R (50,000 iterations after a 10,000-iteration burn-in), 95% Bayesian credible intervals. A subset of the data (2 surveys per unit) was used to simulate reduced survey effort. OpenBUGS code is given in the appendix. Flight routes planned in ArcGIS 9.0. Possible inconsistency: the share of units with strong evidence of occurrence is given as 30% of 588 in the abstract and results, but as 29% (n = 170) in the reduced-effort comparison (170 of 588 is 28.9%). The core area (41% of the study area) is also slightly lower than the 42% implied by adding the strong (30%) and weak (12%) evidence of occurrence, probably rounding.
- **Data type:** New field data: aerial track surveys, January to March 2005 (26 survey days). Background data from a 2004 radio-telemetry study in the study area (unpublished) and Ontario Ministry of Natural Resources fur-trapping records, 1980 to 2005.

## Synopsis

### 1. What they did

In 2004 the Ontario Ministry of Natural Resources (OMNR) listed the wolverine as threatened in Ontario, Canada. Historic records, trapping records and unverified sightings can give a misleading picture of where a species lives, and resource development was expected to spread north. The authors wanted an efficient, affordable way to map where wolverines occur across large, remote forested areas, and to set a baseline for later comparison.

- Survey units. They divided about 60,000 km² of boreal forest in northwestern Ontario into 588 hexagons of 100 km² each, a size based on the smallest likely home range of a resident female wolverine.

- Aerial track surveys (January to March 2005). A pilot and one observer flew a small, two-seat ski plane (a Piper Super Cub) about 200 m above the ground, on sunny or bright overcast days. The pilot flew a winding path through each unit to pass over openings where the forest floor was visible. They waited at least 24 hours after snowfalls of 3 cm or more, or strong winds, and counted any wolverine tracks as evidence of occurrence, whatever their age. Track locations were recorded with Global Positioning System (GPS).

- Repeated surveys. Units were flown again and again, even after tracks had been found, and routes were spread across the study area in space and time so that areas with more and fewer wolverines were surveyed under similar conditions.

- Modelling. They used a hierarchical spatial model (a Bayesian statistical model that allows for tracks being missed and for neighbouring units being alike) to estimate the probability that wolverines occurred in each unit, including units never surveyed. They also reran the model with only the first 2 surveys per unit to see what less effort would give.

### 2. What they found

Survey effort

- They surveyed 575 of the 588 units (98%) over 26 days: 9 in January, 15 in February and 2 in March. The survey ended in late March because freeze and thaw spoiled tracking conditions.

- Most units were flown once (35%) or twice (45%); 16% were flown 3 times, 3% 4 times, and one unit 5 times. In total there were 1,079 survey flights, 905 of them (84%) by the main team.

- Flight time averaged 10 minutes per unit, including checking track identity.

Detections

- Wolverine tracks were found in 138 of 575 units (24%), all in the northern part of the study area.

- In 88% of those units, tracks were found within the first 2 surveys.

- Chance of detecting tracks in an occupied unit was 0.23 before 15 February and 0.54 afterwards.

- Possible wolverine tracks in 5 units could not be confirmed and were left out.

Where wolverines occur

- Strong evidence of occurrence (probability above 0.80) in 30% of the 588 units, weak evidence of occurrence (0.50 to 0.80) in 12%, weak evidence of absence (0.20 to 0.50) in 15%, and strong evidence of absence (below 0.20) in 43%. The highest probabilities were in the north-central part.

- Wolverine range (extent of occurrence) covered 59% of the study area, with a clear southern limit and no limit seen to the north, west or east.

- Area of occupancy (units above 0.20) was 33,400 km², 57% of the study area; the core area (above 0.50) was 24,300 km², 41%.

- The pattern broadly matched fur-trapping records from 1980 to 2005.

Less effort

- With only 2 surveys per unit, the southern limit was still clear, but occurrence probabilities dropped in 69% of units (by more than 0.20 in under 1%) and the share of units with strong evidence of occurrence fell from 29% to 23%.

### 3. What they concluded

- Aerial surveys of tracks in snow, combined with the model, can map wolverine distribution at a scale that suits a wide-ranging species living at low densities. The survey needed only 2 people and 26 days, and about 6 times less flying time than earlier aerial methods would have needed for this area.

- Three surveys per unit were needed to deal with missed tracks and show the general pattern well. The authors suggest that 70% of units should end up with either strong evidence of occurrence or strong evidence of absence.

- Large groups of neighbouring high-probability units (for example 10) likely show occupied home ranges. Small or isolated groups may show transient animals, residents in patchy habitat, animals colonizing new areas, or travel corridors.

- Differences in observer skill can bias results. They recommend one experienced team for most flights, or experienced pilots on all teams with skill spread evenly across the area, and flying over habitats where tracks can be seen.

### 4. How tracking was or was not used

Tracking was the core of the survey. From the air, the team identified wolverine tracks in snow by track size, shape, depth and gait, but above all by the track pattern, including changes in gaits and spacing with habitat and snow conditions. Body prints in deeper snow and the animal’s behaviour also helped. The 3×3 lope is the wolverine’s most common gait; when the team saw other gaits (a 2×2 pattern, walking, galloping, or bounding in deep soft snow) they followed the tracks until the 3×3 lope appeared. Fisher tracks were told apart by being considerably smaller, by more frequent changes in track pattern, by more winding travel and by a narrower track pattern.

To confirm tracks, the team circled them, followed them to watch changes in pattern or behaviour, followed fresh tracks until they saw the wolverine, or landed to look on the ground. The authors found wolverine tracks usually easier to identify from the air than from the ground, because long stretches of trail were visible and could be followed at more than 1 km a minute. Any doubtful track was left out to avoid false records. During the survey, melted fisher tracks on south-facing slopes were checked on the ground 3 times, and one trail was followed from the air until a wolverine was seen.

### 5. Tracker or observer certification

Reported but not validated. The pilot, Patrick Valkenburg, a professional wildlife biologist, had more than 25 years of experience snow-tracking wolverines and other wildlife from aircraft in Alaska and 3 months in northern Ontario. The two observers, Audrey Magoun and Justina Ray, had each tracked wolverines in the study area for 2 seasons. No certification or test of their skill is reported. Because one team flew most surveys, observer team was not included in the model.

### 6. Other technologies, skills or innovations

- Piper PA-18 Super Cub on wheel-skis, flown at 110 to 140 km an hour.

- GPS to record track locations and ArcGIS to plan flight routes through unit centres.

- A hexagon grid, giving each unit up to 6 equal neighbours.

- OpenBUGS software run from R, with the code published in the appendix.

### 7. Implications for conservation (from the paper)

- The 2005 survey gives a baseline: future surveys can show objectively whether wolverine distribution changes as development spreads.

- Comparing occupied and unoccupied units lets managers look at wolverine and habitat associations, and work with conservation groups, forest companies and trappers on management practices.

- The authors recommend repeating the survey at intervals that can inform forest management plans (for example every 5 years).

### 8. Implications for tracking (Original Wisdom's observations)

_These are Original Wisdom's observations, not the authors'._

- The paper sets out in detail how experienced trackers identify one species’ trails from the air: reading the whole track pattern and gait changes over long distances, rather than single prints, and following a trail until the typical gait appears.

- The authors note that differences in observer skill affect detection, and recommend experienced teams of similar skill. A tested standard of aerial track identification would let such surveys show how equal their teams’ skills are.

## Cite this summary

Lawrence, K. 2026. Summary of: Modeling Wolverine Occurrence Using Aerial Surveys of Tracks in Snow (Magoun and Ray et al., 2007). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2007-magoun-ray-modeling-wolverine-occurrence-aerial/