Estimating wolf densities in forested areas using network sampling of tracks in snow

1. What they did

Counting large carnivores is hard, especially in forest, where animals are rarely seen from the air, and where no animals carry radio collars. Becker and colleagues (1998) developed a way to estimate wolf numbers from tracks in snow, called the sampling unit probability estimator (SUPE), in open country in Alaska. This study tested it for the first time in dense forest, amid concern that wolves might be declining in Algonquin Park, Ontario.

• Dividing the area. The team split a 3,425 km² area in western Algonquin Park into 137 sample units of 5 × 5 km. Before the survey, they rated each unit as having a high (61 units) or low (76 units) chance of holding fresh wolf tracks, based on its rivers, lakeshores and trails (where tracks are easy to see) and its hemlock cover (where moose gather in winter).
• Searching for tracks (February 2002). Starting about 24 hours after a 5 cm snowfall on a deep snow base, a four-person crew in a helicopter searched the watercourses, roads and trails in a sample of units for fresh wolf tracks (made since the last snowfall or windstorm). They landed to check tracks when the species was uncertain, and watched for carcasses and ravens as signs of a wolf kill.
• Following the tracks. When they found fresh tracks, they followed them forward to the wolves, then backward until the tracks were no longer fresh, to map every sample unit the group had passed through. They counted pack size by seeing the wolves, or by counting where the pack’s trail split into separate trails, or the beds where the pack had rested.
• Estimating numbers. The chance of finding each group’s tracks depends on how many sample units their tracks cross, and this gives each group’s contribution to the total estimate.
• Checking with radio collars (2003). They then radio-collared 46 wolves, located them from the air from January to April 2003, and estimated density in an overlapping area the “traditional” way, from pack territories and pack sizes, to compare with the track survey.

2. What they found

The 2002 track survey

• Over 11 flying days they searched 45 sample units: 28 high (46% of high units) and 17 low (22% of low units).
• They found 17 fresh track networks: 3 single wolves, 2 pairs, 2 packs of 3, 3 packs of 4, 5 packs of 5 and 2 packs of 6.
• This gave an estimate of 87 wolves in the study area, a density of 2.5 wolves per 100 km². The abstract gives the 90% confidence interval as ± 0.3 and the Results as ± 0.5.
• Average pack size (not counting single wolves) was 4.2. They estimated about 16 packs and 4 single wolves in the study area.

The 2003 telemetry estimate

• Ten packs in the census area held 42 wolves, giving 1.8 resident wolves per 100 km². Adding an estimated 10 lone wolves (based on the ratio of lone to pack wolves trapped in 2002) gave 2.3 wolves per 100 km².
• Average pack size from telemetry was also 4.2, the same as from the track survey.

Did the survey meet its assumptions?

• Wolves move: even wolves at or near fresh deer or moose carcasses had moved more than 1 km in the previous day.
• Tracks are continuous: tracks were sometimes lost in thick conifer cover. The team then circled the patch until they found the trail again or knew the wolves had not left it, so every track segment could be matched to the wolves that made it.
• No fresh tracks missed: with no collared wolves in the area at the time, they could not prove this. But they never found a trail that led back into a unit they had already searched without seeing it, and the estimate matched telemetry estimates from 1988 to 1992, 2003 and 2004.
• No double counting: they searched from south to north, searched nearby units close together in time, and checked the origin of any tracks within 10 km of a group already counted.
• Larger packs had a higher chance of being found (a sign the high and low rating worked well).

Time and cost

• Each unit took 23 to 60 minutes to search, not counting time spent following tracks; with following, the average was 46 minutes (range 23 to 132). This was far longer than the 12 to 33 minutes per larger unit from a fixed-wing plane in Alaska, and more than twice as long as the team had expected.
• The 2002 survey cost about 40,000 Canadian dollars (excluding staff time). Collaring and radio-tracking wolves in 2003 cost about 25% more (about 50,000 dollars).

Wolves and deer yards

In winter 2003, 5 of 6 packs monitored in eastern Algonquin made repeated trips to deer yards outside the park, but only 1 of 10 packs in the western census area left their territory. Wolf density in western Algonquin seems to have changed little between the late 1980s and 2002 to 2003.

3. What they concluded

• The SUPE gives an objective, repeatable and apparently accurate estimate of wolf density in forest, with a measure of precision. It also avoids having to guess the share of lone wolves, a problem for telemetry estimates.
• It will be particularly useful where wolves are not radio-collared, or where radio telemetry is not practical or socially acceptable.
• Pack sizes from a single count of tracks are minimums: pack members sometimes travel apart, and counting from tracks in snow alone may have undercounted some packs.
• Following tracks in forest is slow, so the survey costs much more per area than in open country. Because collared wolves also serve other research aims, the authors could not recommend the SUPE on cost savings alone.
• Inexperienced observers will probably need a helicopter to follow wolf tracks in heavy forest. Professional wolf trackers in the Yukon and Alaska have followed wolf tracks through dense cover from small fixed-wing planes, and using such trackers should allow a larger area to be surveyed for the same money.

4. How tracking was or was not used

Tracking from the air is the heart of this method:

• Tracks and signs identification: telling wolf tracks from those of foxes, otters, fishers, deer and moose, landing to look more closely when unsure; and judging whether tracks were fresh (made since the last snowfall or windstorm).
• Trailing: following each fresh trail forward to the wolves and backward until it was no longer fresh; circling thick conifer patches to pick up lost trails; searching nearby unsampled units for tracks entering or leaving.
• Counting animals from sign: pack size was counted where the trail split into separate trails, or by counting beds at a resting site, landing when a good count could not be made from the air.
• Kill sites: carcasses and ravens or other scavengers were used as clues that a wolf kill might be nearby.

5. Tracker or observer certification

Not reported. The survey crew is named in the acknowledgements (two helicopter pilots and five survey observers), but their tracking experience is not described or tested. This matters, because the estimate depends on not missing fresh tracks, telling wolf tracks from others, and counting pack members from tracks and beds. The authors mention professional wolf trackers in the Yukon and Alaska only as a possible way to cut costs.

6. Other technologies, skills or innovations

• A Bell 206B Jet Ranger helicopter for searching and following tracks, with set flight lines over each sample unit.
• The program SUPEPOP for the population and variance estimates.
• Very high frequency (VHF) radio collars, foot-hold traps and helicopter net-gunning for the telemetry comparison.

7. Implications for conservation (from the paper)

• Reliable counts are central to conserving wolves and other large carnivores, and the SUPE can provide them in forest where collared wolves are not available.
• Many wolves from eastern Algonquin are trapped or shot outside the park while visiting deer yards in winter, which led to a 30-month moratorium on hunting and trapping wolves within 10 km of the park boundary from December 2001. Fewer wolves from western Algonquin left the park in winter, so they may be less exposed to hunting and trapping.
• Wolf density in western Algonquin seemed stable from the late 1980s to 2003, even though it may have declined in the east during the 1990s.