Coarse-Scale Distribution Surveys and Occurrence Probability Modeling for Wolverine in Interior Alaska

1. What they did

Wolverines live throughout Interior Alaska, but little was known about where exactly they occur. Management relied on trapping harvest records and chance sightings, which are unreliable: some harvest goes unreported, locations are not checked, and trapper effort is not measured. Meanwhile trapping, roads and industrial development are spreading into remote country that once served as refuges.

The researchers adapted an aerial snow-track method first used in Ontario, Canada, to a much larger area (180,000 km²):

• Sample units. They divided the study area into 180 hexagons of 1,000 km² each, about the size of a male wolverine’s home range.
• Flying the survey. In February and March 2006, teams in small Piper Super Cub aircraft flew low (100 to 250 m) and slow (about 120 km per hour) along 32 km routes through each hexagon, looking for wolverine tracks in the snow. They flew one route, and if they saw no wolverine track, flew more routes the same day, up to four. They did not fly within about 24 hours of snowfall of 3 cm or more or strong winds.
• Checking tracks. Teams took as long as needed to confirm each track. They followed any questionable track until they saw the wolverine’s diagnostic 3-track lope or ruled out a wolverine, and sometimes landed to inspect tracks on the ground. Tracks that stayed questionable were left out.
• Modelling. They built a statistical model that estimates, for each hexagon, the probability that a wolverine track was really present, allowing for tracks that were missed. It took into account the survey date, the number of routes flown, the share of lowland (305 m or lower), human influences (roads, trails, buildings, mines) and whether neighbouring hexagons had tracks. They also used simulations to work out how many routes are needed to be confident a hexagon is empty.

2. What they found

The survey

• They surveyed 149 of the 180 hexagons (83%); weather and funding stopped the rest. Most (79%) needed only one route.
• Wolverine tracks were seen in 99 hexagons (66%). Tracks were found in 58% of hexagons on the first route, rising to 64% and 66% with a second and third; a fourth route found no new ones.
• Average flight time per route was 22.3 minutes, with no difference between hexagons with and without wolverines.

Detection

• On the mean survey date (20 February), the chance of seeing a track on one route, when one was there, was 0.81. It rose from 69% on 8 February to 91% on 12 March: detection improved later in the season.
• Detection approached 100% with three or more routes.

Where wolverines were

• Strong evidence of wolverines in 72% of the 180 hexagons, strong evidence of absence in 12%, and weak evidence either way in 16%. Their core area of occupancy covered 83% of the study area.
• Wolverines were more likely to be present at higher elevations and less likely where there were more human influences. Elevation and human influences were not correlated with each other.
• Clear absences were around Fairbanks and North Pole, along the Parks Highway from Nenana to Healy, and around Circle and Central.
• Harvest records were not a good guide: areas of low, medium and high harvest were scattered with no clear link to where wolverines were or were not found.

Survey effort

• Hexagons with uncertain results needed 2 to 4 routes without a track to give strong evidence of absence. Favourable hexagons (high elevation, little human influence, surrounded by occupied units) needed the most routes.
• The survey covered 149,000 km² for US$0.11 per km² (2006 rates).

3. What they concluded

• The method gave managers a baseline map of wolverine distribution across Interior Alaska and an efficient, lower-cost way to detect coarse-scale changes.
• Managing wolverines over large areas from harvest data alone can mislead: areas of low harvest may have no wolverines left, while areas of high harvest may still hold residents.
• Wolverines in Interior Alaska can persist even in lower-quality lowland habitat, except where harvest and human influences are high. The authors think the human effect is most likely a mix of development and trapping.
• A conservative design is 4 routes per hexagon, but the model allows effort to be focused where wolverines are more likely. Surveying in March may be more efficient if snow stays good.
• Because detection is so high, the method shows large-scale range contraction but may miss it until an area is almost empty; for fine-scale change, smaller sample units or population surveys are needed.
• The method should work across most of the boreal range of wolverines where tracks can be seen from the air, but tracks must be identified correctly and training may be needed.

4. How tracking was or was not used

The whole survey rests on spotting and identifying wolverine tracks in snow from a low-flying aircraft. To avoid false records, teams followed any doubtful track (trailing from the air) until they saw the wolverine’s typical 3-track lope pattern, or until they could rule out a wolverine, and sometimes landed to check tracks on the ground. Doubtful tracks were not used. The authors describe these as evidentiary standards for confirming a detection. Snowmobile trails and signs of human development were also recorded along each route.

5. Tracker or observer certification

Reported but not validated. The three pilots each had more than 20 years of experience snow-tracking wolverines, wolves and other furbearers from the air (two had worked on the Ontario study), and the two observers (authors C. Gardner and J. Lawler) each had at least 3 years of experience. No test of their identification skill is reported. Pilot-observer pairs were tested as a factor affecting detection and dropped from the model. The authors stress that the technique depends on correct track identification and that training may be needed.

6. Other technologies, skills or innovations

• Low, slow aerial survey in Piper Super Cub aircraft, with hexagon-shaped units giving several route choices.
• A removal design: extra routes flown only until a track is found, so detection can be estimated from a single visit.
• Bayesian spatial occupancy modelling that borrows information from neighbouring units, including unsampled ones.
• Simulations to plan how many routes each unit needs.

7. Implications for conservation (from the paper)

• Trapping and industrial growth are the main concerns for wolverine range contraction in Interior Alaska; the map gives a baseline to watch for change.
• Harvest data alone can give a wrong picture of where wolverines live.
• Where unsampled units cannot be avoided, spreading them across the study area makes best use of information from neighbours.
• If contraction in one specific place is a concern, managers need a method that estimates numbers, not just presence.