Classifying Carnivore Tracks Using Dimensions That Control for Snow Conditions

1. What they did

American martens and fishers live in the same snowy forests, eat similar food, use similar habitat, and leave tracks in snow that are hard to tell apart. Snow itself changes tracks: a track in deep, soft snow without a crust is deeper and can look wider than the same animal’s track on firm snow. The authors asked whether track measurements change with snow conditions, and whether marten and fisher tracks (and the sexes) could be classified more accurately using measurements that allow for the snow.

• Collaring. In northern Wisconsin, USA, they live-trapped martens and fishers and fitted adults with very high frequency (VHF) radio collars. They also followed juvenile and adult martens that had been translocated from Minnesota with collars as part of a population boost.
• Following trails. In December to March of 2008–2009 and 2009–2010, they located a collared animal by radio signal and backtracked its trail in snow for 500 m. If the animal was active, they found its trail behind it, followed it backward more than 100 m, then recorded 500 m.
• Measuring tracks. At fixed intervals (every 25 m in the first winter, every 100 m in the second) they measured stride length, step depth and straddle width, only on straight trails in level snow where the animal was using a 2×2 lope. Straddle width was measured by the minimum outline method, which leaves out disturbance from the body. Tracks blurred by wind-blown snow or melt were skipped.
• Measuring snow. Between sets of tracks they measured snow depth, crust depth, and snow compaction, the last with a penetrometer weighted like a female marten’s forefoot and dropped from 10 cm above the snow.
• Analysis. They tested how each track measurement related to each snow condition, then built four classification trees: two sorting tracks into female marten, male marten, female fisher and male fisher, and two sorting by species only. In each pair, one tree used track measurements alone, and the other also used snow conditions and track measurements divided by snow conditions.

2. What they found

The data

• 1658 track and 1738 snow measurements from 56 trails: 42 marten trails and 14 fisher trails.
• In 2008–2009 the trails came from 15 martens (8 female, 7 male) and 2 fishers (1 female, 1 male); in 2009–2010 from 16 martens (8 female, 8 male) and 6 fishers (3 female, 3 male).

Snow and track size

• Fisher step depth was strongly linked to all three snow conditions: steps were deeper when the crust lay deeper (r² = 0.82), when the snow was softer (r² = 0.65) and when the snow was deeper (r² = 0.62). Other fisher measurements were not linked to snow.
• Marten tracks were not strongly linked to snow conditions. Crust depth explained less than 35% of the variation in marten step depth, and snow depth and snow softness explained less than 25% of the variation in step depth and stride length.
• Straddle width and stride length stayed fairly steady across snow conditions in both species.

Telling tracks apart (species only)

• Track measurements alone: 98% of marten and 83% of fisher tracks classified correctly (mean 91%). Straddle width (10.8 cm or more points toward fisher) and stride length were the useful measurements.
• With snow conditions: 100% of marten (42 of 42) and 92% of fisher tracks (11 of 12) correct (mean 96%). Straddle width and step depth divided by penetrometer depth were the useful measurements.

Telling sexes and species apart

• Track measurements alone: female marten 68%, male marten 95%, female fisher 57%, male fisher 80% (mean 75%).
• With snow conditions: female marten 91%, male marten 95%, female fisher 71%, male fisher 100% (mean 89%). Stride length helped separate female from male martens, and step depth divided by penetrometer depth helped separate large-straddled male martens from fishers.
• Allowing for snow made the trees 5% more accurate for species and 14% more accurate for sex and species.

3. What they concluded

• Marten and fisher tracks in snow can be identified accurately by measuring them.
• Allowing for snow conditions improves accuracy. Because fishers carry about twice the weight per area of foot that martens do, step depth corrected for snow softness tracks foot-loading more closely and helps separate the two species.
• Snow-tracking that includes snow measurements could be used to study habitat use, distribution and numbers of each sex of martens and fishers, and allowing for snow would likely improve track classification for other carnivores too.
• Measuring snow on the trail itself, soon after the animal passed, matters, because snow conditions can change over a few meters or a few hours.
• The minimum outline method appears to keep straddle width steady across snow conditions.
• The trees may not work as well in other regions, because marten size varies widely across their range (fisher size much less). The straddle split of 10.8 cm, for example, might need to be smaller where martens are smaller.

4. How tracking was or was not used

Tracking was the core method. The researchers backtracked the snow trails of radio-collared martens and fishers for 500 m, so the species and sex of each trail were known. Along each trail they recognized the 2×2 lope, selected straight sections in level snow, and measured stride length, step depth and straddle width. Traps were also placed where signs of martens and fishers had been seen.

5. Tracker or observer certification

Not reported. The paper does not say who followed the trails or measured the tracks, or what their experience was. Because each trail belonged to a known collared animal, the classifications were tested against known identities rather than against trackers’ judgement. Several people are thanked for help with fieldwork.

6. Other technologies, skills or innovations

• Very high frequency (VHF) radio collars, some with activity switches, used to find animals and tell from the signal whether they were resting or active.
• A penetrometer scaled to a female marten’s forefoot-loading (24 g/cm²) to measure snow compaction.
• Classification trees presented as simple yes/no keys (Figures 3 and 4) that a tracker can follow with their own measurements.

7. Implications for conservation (from the paper)

• Misidentified tracks reduce the accuracy of snow-tracking studies; better classification improves estimates of habitat use, distribution and abundance.
• Snow-tracking can detect some carnivores better than track plates and cameras, including more than three times better for fishers in one cited study, and avoids the bias of baits and lures.
• Being able to tell the sexes apart from tracks would let snow-tracking studies look at males and females separately.