Molecular testing of observer identification of carnivore feces in the field

1. What they did

Many studies of carnivores rely on scats collected along roads and trails, and assume the people collecting them have identified the species correctly. That assumption is rarely tested. The authors tested how well observers could tell coyote scats from those of other similar-sized carnivores in the Alaska Range, where wolves, red foxes, lynx and dogs could all leave scats that look like a coyote’s.

• Collecting scats (2000 to 2002). In winter (January to March) they collected scats along 150 km of snowmobile trails and while snow-tracking coyotes on foot, 1,237 in all including 403 picked up opportunistically in summer. The project leader (Laura Prugh), who had previous experience identifying coyote scats, collected in all three years. An experienced technician helped in 2000, and a technician who was at first inexperienced helped in 2001 and 2002.
• Rating certainty (2001 and 2002). For every scat, the collector rated how sure they were that it came from a coyote: 100%, 95%, 90%, 70% or 50%. Scats below 50% were not collected. They judged from tracks in the snow, the scat’s diameter, volume and shape, and where it lay. For example, lynx scats tend to have bulbous segments and are often left at latrines in tree wells, while coyote scats tend to have twisted, tapering segments and are often left on trails.
• Measuring diameter (2000). In the first year they measured the maximum diameter of most scats instead of rating certainty.
• Testing the DNA. They tested 834 randomly chosen winter scats with a mitochondrial DNA (mtDNA) method that clearly separates coyotes from wolves, foxes, lynx and dogs, using known samples from captive and pet animals as checks.
• Counting tracks. Snow-track surveys after fresh snowfall each year gave an index of how common each carnivore was.

2. What they found

The DNA test

• DNA amplified from 647 of the 834 scats (78%). Of these, 596 (92%) were from coyotes.
• The test told coyotes apart from the other species but could not separate wolves, foxes and dogs from each other. Only one lynx scat was found.
• Laboratory materials cost about US$5.00 per sample, most of it ($3.25) for the DNA extraction kits.

How common the carnivores were

• Over three years the track surveys counted 259 coyote, 331 red fox, 126 wolf and 154 lynx tracks, so no species was much more abundant than the others. Pine marten (15) and wolverine (20) tracks were far less common, and scats that looked like mustelid scats were rarely seen.

Certainty ratings matched the DNA

• 100% of scats rated “100% certain” were coyote (129 of 129).
• 96% of scats rated “95% certain” were coyote (167 of 174).
• 89% of scats rated “90% certain” were coyote (55 of 62).
• At the low ratings, observers were conservative: 84% of “70% certain” scats (36 of 43) and 75% of “50% certain” scats (6 of 8) were coyote, more than expected.
• Overall the proportions matched the ratings closely, with no significant difference.

Observers and experience

• The two observers did not differ in accuracy, and there was no sign that experience changed accuracy. The technician who started out inexperienced in 2001 had 98% coyote scats among their collections that year. Both observers were slightly more accurate than their ratings predicted, in both years.

Seasons

• Certainty was higher in winter (average 89.8%) than in summer (78.3%). A scat was only rated “100% certain” when fresh coyote tracks were with it and no other animal’s tracks were present, and that never happened in summer.

Scat diameter

• Coyote scats were larger on average (21.7 mm) than noncoyote scats (18.4 mm), but the ranges overlapped widely (coyote 13 to 35 mm, noncoyote 12 to 26 mm).
• Diameter alone misclassified 73 of 183 coyote scats and 8 of 25 noncoyote scats. The authors suggest coyote scats were more likely to be confused with fox scats (smaller) than with wolf scats (larger).

3. What they concluded

• Trained observers identified coyote scats in the field with more than 90% accuracy, even with four similar-sized carnivores common in the area. This should be accurate enough for diet studies.
• Rating certainty, though subjective, gave reliable information about each sample. It was a more efficient and effective way to cut out nontarget scats than measuring diameter, which was time-consuming and could not reliably classify scats.
• They recommend that all scat surveyors rate their certainty when collecting, so researchers can later decide what level of uncertainty is acceptable and exclude samples accordingly.
• Ideally, observers would be trained with scats of known origin, which is possible where snow or mud shows whose tracks lead to the scat. Snow-tracking helped them calibrate their ratings by showing what coyote scats look like and how much they vary. Training with scats from captive animals is impractical: there are few captive carnivores, they eat different food (which can change scat appearance), and the scat’s location in the field is lost.
• Molecular testing is worth using when the target species is rare compared with other carnivores, when its scats look especially like those of other species, or when very high accuracy (over 90 to 95%) is needed. Its drawbacks are time and cost. Trained detection dogs are a promising, inexpensive alternative.
• In studies without molecular testing or dogs, observers should rate certainty systematically and, where there is snow, use tracks to improve accuracy.

4. How tracking was or was not used

Tracking was central to how the scats were found and judged. In winter the team collected scats while snow-tracking coyotes on foot, as well as along snowmobile trails. Tracks in the snow were one of the main clues to who left a scat: only scats with fresh coyote tracks and no other animal’s tracks were rated “100% certain”. Scat identification also drew on scat shape, size, segments and location (for example lynx latrines in tree wells). Snow-track surveys after fresh snowfall gave an index of each carnivore’s abundance. The DNA tests then checked the observers’ identifications.

5. Tracker or observer certification

Other experienced trackers. No certification is reported. The observers’ skill at identifying coyote scats was tested in this study against DNA. Observer #1 (the paper says the project leader, Laura Prugh, had previous experience; Table 2 gives Observer #1 two years of prior scat identification experience) and Observer #2 (a technician who was inexperienced at the start of 2001) were equally accurate. An experienced technician who helped in 2000 was not rated for certainty. T. Heakes and J. Bernier are thanked for field assistance.

6. Other technologies, skills or innovations

• A simple five-level certainty rating recorded for each scat at the moment of collection.
• Mitochondrial DNA test: polymerase chain reaction (PCR) with ScatID primers, then digestion with the TaqαI restriction enzyme, which cuts coyote DNA but not wolf, fox, dog or lynx DNA. Results were read on agarose gels.
• Positive controls from captive and pet animals (lynx, farmed red foxes, zoo wolves and a zoo coyote, and 18 purebred dogs from a dog show), and checks that prey DNA (snowshoe hare, Arctic ground squirrel) did not give a false result.
• Cost figures: about US$5.00 per sample in materials.

7. Implications for conservation (from the paper)

• Scat identification is a basic assumption of studies that use feces to learn about carnivore diet, abundance, parasites, hormones and individuals, and errors can bias the data.
• Rating certainty at collection lets researchers judge sample quality later and exclude doubtful samples.
• Molecular checks matter most for rare species, look-alike scats, or studies needing very high accuracy; other studies have found observers tend to over-count scats of rare species.