# Estimating wolverine Gulo gulo population size using quadrat sampling of tracks in snow

> How do you count an animal that is rarely seen and roams huge areas? Flying low over fresh snow in Alaska and Yukon, teams followed every wolverine track they found forward and backward through a grid of random squares. The result was the most precise wolverine population estimates reported at the time, from a method that rests on reading tracks well.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2007-golden-henry-estimating-wolverine-gulo-gulo/

## Citation and access

- **Authors:** Howard N. Golden, J. David Henry, Earl F. Becker, Michael I. Goldstein, John M. Morton, Dennis Frost, Sr. and Aaron J. Poe
- **Year published:** 2007
- **Journal or source:** Wildlife Biology
- **Full citation:** Golden, H. N., Henry, J. D., Becker, E. F., Goldstein, M. I., Morton, J. M., Frost, D., Sr. and Poe, A. J. 2007. Estimating wolverine Gulo gulo population size using quadrat sampling of tracks in snow. Wildlife Biology 13(Suppl. 2): 52–61.
- **DOI:** 10.2981/0909-6396(2007)13[52:EWGGPS]2.0.CO;2
- **Access status:** Hosted (freely available elsewhere)
- **PDF:** https://www.originalwisdom.com/wp-content/uploads/2026/11/2007_Golden_and_Henry_etal_estimating-wolverine-population-size-using-quadrat-sampling-of-tracks-in-snow.pdf

## Tags

- **Species:** Mammals > Wolverine
- **Region:** North America > Canada; North America > United States
- **Tracking:** Snow tracks; Tracking; Tracks; Tracks and signs identification; Trailing
- **Behaviour and ecology:** Behaviour; Denning; Movement
- **Methods and technology:** Aerial surveys; Global Positioning System (GPS); Population estimation; Simulation; Survey costs
- **Data type:** New field data
- **Observer qualification:** Reported but not validated
- **Discipline:** Wildlife tracking
- **Year published:** 2007

## Detailed keywords

- **Species:** Studied: wolverine (Gulo gulo). Food sources mentioned for the study areas: large ungulate carrion, smaller mammals and birds (Turnagain Arm and Kenai Mountains); caribou (Rangifer tarandus) carrion, small mammals and birds (Old Crow Flats). The survey method was first designed for wolves (Canis lupus).
- **Location and habitat:** Two survey areas. (1) Upper Turnagain Arm and northwestern Kenai Mountains (TAKM), south-central Alaska, United States (about 60°30’N, 149°30’W), 4,340 km²: steep, rugged mountains, deep valleys, glaciers and ice fields, sea level to 1990 m; coniferous and deciduous forest with light to moderately heavy canopy in the valleys, alpine tundra higher up; northern maritime climate; next to two large urban centres and two major highways, but mostly roadless. (2) Old Crow Flats (OCF), Old Crow Basin, northern Yukon, Canada, next to the Alaska border (about 68°N, 140°W), 3,375 km²: rolling hills and wetlands, mostly below 400 m, uplands 300 to 600 m; coniferous and deciduous forest with light to moderate canopy, tundra shrubs and grasses on the uplands; subarctic continental climate with dry, loose snow; remote and roadless, 50 to 80 km north of the village of Old Crow (Vuntut Gwitchin First Nation). Wolverines are harvested in both areas.
- **Topics:** population estimation from tracks in snow; sample-unit probability estimator (SUPE); network sampling; stratified random sampling of quadrats (sample units of about 25 km²); transect intercept probability sampling (TIPS) as the earlier method; aerial snow tracking from small aircraft; following track trails forward and backward; counting individuals from separate tracks in a trail; telling fresh tracks from old ones; boundary rule for tracks leaving the survey area; survey assumptions; sampling fractions and survey efficiency; denning females as a possible source of downward bias; survey costs and pilot and observer availability. Possible inconsistency: the methods say survey time per sample unit was 5 to 20 minutes, but the results give average flight times per sample unit of 22 minutes (TAKM) and 34 minutes (OCF); the averages may include time spent following tracks, but the paper does not say.
- **Statistics or models:** Sample-unit probability estimator (SUPE, Becker et al. 1998), calculated in the program SUPEPOP; population size, density, standard error (SE), 80% and 90% confidence intervals (CI), coefficient of variation (CV); simulations in SUPEPOP of different sampling fractions.
- **Data type:** New field data: aerial surveys in March 2004 (TAKM on 6 and 17 March; OCF on 16 to 18 and 21 to 22 March).

## Synopsis

### 1. What they did

Wolverines live at low densities and range widely, so they are hard to count. Earlier counts relied on radio-collared animals, natal dens, or an aerial track method called transect intercept probability sampling (TIPS), which had problems in steep terrain and under heavy forest canopy. The authors tested a newer method, the sample-unit probability estimator (SUPE), first designed for wolves, on wolverines in two very different places.

- Setting up the grid. Each survey area was divided into quadrats (sample units) of about 25 km². Using past knowledge of wolverines, their prey and harvest patterns, the units were sorted into a “high” stratum (more likely to hold wolverine tracks) and a “medium-low” stratum, and units were picked at random from each.

- Flying the surveys (March 2004). Pilot-observer teams in Super Cub or similar small aircraft started within 24 hours after fresh snowfall, flying at 90 to 130 km per hour and 35 to 150 m above the ground. They searched each chosen unit until they were confident no fresh wolverine tracks had been missed.

- Following the tracks. When a fresh track was found, the team followed it forward to its end and counted the wolverines, either by seeing them or from the number of separate tracks in the trail. They then backtracked it to its start, which might be a resting hole, a den, or where the last snowfall covered it. They recorded GPS points for the start and end and mapped every quadrat the trail crossed. A track that left the survey area was followed to its end, so only the share inside the area counted.

- Estimating numbers. The program SUPEPOP turned the track groups, the units each passed through, and the sampling effort into a population estimate with a measure of precision.

### 2. What they found

Turnagain Arm and Kenai Mountains, Alaska (4,340 km²)

- They surveyed 87 of 171 units (51%): 66% of the high stratum and 32% of the medium-low stratum, over two days with 32.5 hours of flying.

- They saw 11 individual wolverine tracks in 10 track groups, at all elevations, in both forest and tundra.

- Estimated population: 12.8 wolverines (80% CI 11.0 to 14.9), or 3.0 wolverines per 1,000 km², with a CV of 12.0%. The lower limit was set at 11 because more wolverines were actually seen than the calculated value of 10.7.

Old Crow Flats, Yukon (3,375 km²)

- They surveyed 96 of 135 units (71%): 72% of the high stratum and 67% of the medium-low stratum, over five days with 55 hours of flying.

- They found 31 track groups, all of single animals. Eight ran partly outside the boundary, so the share inside equalled 28.7 wolverines. Most tracks were in forested drainages.

- Estimated population: 32.9 wolverines (80% CI 30.1 to 35.7), or 9.7 per 1,000 km², with a CV of 6.5%.

Survey conditions and effort

- Snow, weather and light were fair to excellent. Only one track, blurred by wind and probably old, could not be backtracked and was left out.

- No tracks seemed to have moved from a unit already surveyed into one surveyed the next day.

- Old Crow Flats took about 50% more time per unit (34 versus 22 minutes) but about half the time per wolverine group found (106 versus 195 minutes).

- Simulations suggested that raising the Alaska sampling from 66% to 70% in the high stratum and from 32% to 50% in the medium-low stratum would add 10% to the effort and lower the CV by 2%.

### 3. What they concluded

- The CVs of 12% and 6.5% were the lowest yet reported for wolverine population estimates, even though the two areas differed three-fold in density. The Old Crow Flats density was the highest reported from either TIPS or SUPE surveys.

- SUPE gave precise estimates under very different terrain, canopy, weather and densities. Where practical, it may be less labour-intensive and cheaper than methods that do not use probability sampling of tracks.

- The authors recommend sampling 65 to 70% of units in high strata and 45 to 50% in medium-low strata, and restratifying areas for future surveys.

- Each estimate is a “snapshot” for one area and one short period, taken after the hunting and trapping season, when numbers are probably at their lowest; care is needed when comparing across times and places.

- Of the 10 survey assumptions, only one may have been broken: that all animals move during the survey. Some females may have stayed in their dens around the time of giving birth (mid-February to March), which would bias the estimates downward. Testing SUPE with radio-collared wolverines or double sampling could give correction factors.

- Drawbacks: the surveys must be flown to cover large areas, which is costly, and they depend on canopy, snow, weather and light. The availability of pilots and observers was sometimes the most important limit.

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

Tracking was the core of the method. From the air, teams identified fresh wolverine tracks in snow, told them apart from older tracks (which lose their sharp, crisp edges and look more diffuse), followed each trail forward and backward to its ends, and counted the animals from the number of separate tracks. Following the trail in both directions often showed quickly whether a track really was a wolverine’s. When a species could not be identified from the air, the team had to land and look closely or leave the track out. Where trails were confusing, for example around a resting hole or den, a trail had to be treated as one animal.

### 5. Tracker or observer certification

Reported but not validated. The authors state that at least one member of each pilot-observer team must be an experienced aerial observer and very skilled in track identification, and they thank the pilots for “expert tracking”. No certification or test of the observers’ tracking skill is reported.

### 6. Other technologies, skills or innovations

- Low, slow flying in small, manoeuvrable aircraft (Super Cub or similar), with teams in radio contact for safety.

- Global Positioning System (GPS) points for the start and end of each track, and mapping of each trail across quadrats, so tracks were not counted twice on different days.

- The free program SUPEPOP for the population calculations and for simulating different sampling efforts.

- Flexibility in how a quadrat is searched: straight lines under light canopy, overlapping circles under heavier canopy.

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

- Wolverines are valued for fur, cause some livestock losses, and are endangered in parts of their range, so their populations need monitoring, yet few methods exist for estimating their numbers or trends.

- SUPE gives a repeatable, scientifically designed estimate with a measure of precision, and can provide valuable baseline data on wolverine populations.

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

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

- The whole estimate rests on the observers’ tracks and signs skills from the air: recognising wolverine tracks, judging whether they are fresh, following them both ways and counting animals from separate tracks. Errors in any of these feed straight into the population number, yet the observers’ skill was not tested.

## Cite this summary

Lawrence, K. 2026. Summary of: Estimating wolverine Gulo gulo population size using quadrat sampling of tracks in snow (Golden and Henry et al., 2007). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2007-golden-henry-estimating-wolverine-gulo-gulo/