Effectiveness of Scat Detection Dogs for Detecting Forest Carnivores

1. What they did

Carnivores are hard to study: they range widely, live at low densities and keep out of sight. Dogs trained to find scat had been tested on very few species, and never on fishers or bobcats. The authors set out to learn how well detection dogs could find the scats of three very different forest carnivores at once: black bears, fishers and bobcats, across the forests of Vermont. As far as they knew, it was the first use of scat detection dogs in the forests of eastern North America.

• Training. Professional trainers (Packleader Dog Training, Washington State) trained each dog for 4 to 6 weeks, adding one target species at a time, with at least 20 scats per species from many individuals and diets, so dogs learned the species’ scent rather than one animal or one food. Dogs were rewarded with a tennis ball and taught to sit or stay at a scat (“indicating”). Each handler trained for 10 days to read and assist their dog.
• Surveys (May to August 2003 and 2004). Five teams (each a dog, a handler and an orienteer who kept the team on course) searched a 2 km, diamond-shaped transect at 168 randomly placed sites, about 5 km apart. Some sites were searched two or three times (220 visits in all). Scats up to 100 m outside the transect were included.
• Recording each scat. The team named the likely species from its shape, size and contents, gave that call a confidence level (high, medium or low) based on the dog’s behaviour, the scat’s appearance, its age and how degraded it was, aged bear scats where possible, and judged how likely they would have been to find it without the dog.
• Checking by DNA. 347 scats were sent for DNA species identification. Because fresh black bear scats are unmistakable in the region, high-confidence bear scats were usually not tested.
• Analysis. From DNA results they worked out how often each species and confidence level was named correctly, used those rates to decide whether a site where DNA failed could still count as a detection (a probability of correct identification of 0.90 or more), and used occupancy models to estimate the chance of detecting each species on a single visit and what affected it.

2. What they found

Scats found

• The teams found 1,596 scats: 747 putative black bear, 665 fisher and 153 bobcat; 31 more were too degraded or ambiguous to assign. That averaged 7.2 scats per transect.
• 62% of the identifications were rated high confidence, 25% medium and 13% low.
• The teams rated 83% of scats very unlikely to have been found without the dog, 11% moderately likely and 6% highly likely.

How accurate the field identifications were

• Of 347 scats tested, 38% gave no usable DNA, 21 gave a mixed or unknown result, and 16 gave only prey DNA (such as white-tailed deer, cow or red squirrel).
• Correct identification rates: black bear 1.00 (high confidence) and 0.50 (medium); fisher 0.93, 0.94 and 0.77 (high, medium, low); bobcat 0.90, 0.29 and 0.13.
• Other species identified by DNA were coyote, red fox, northern raccoon, house cat and long-tailed weasel. Coyote and bobcat scats were difficult or impossible to tell apart by shape alone, and raccoon scats could look like small bear scats.

Where each species was detected

• Black bears at 96 of 168 sites (57.1%), fishers at 103 (61.3%) and bobcats at 21 (12.5%). For 70 of these sites DNA could not confirm the species, and the probability method was used instead.

Chance of detection on one visit (when the species was present)

• Black bear 0.86, fisher 0.95, bobcat 0.40.
• Terrain ruggedness, vegetation density and weather (temperature, humidity, wind, rain) made little or no difference for bears or fishers.
• For fishers, detection differed by team: one team’s chance was 0.41, against 0.78 to 0.95 for the others. That team’s handler was the only one with no previous detection dog field experience, and the dog was easily distracted by chipmunks and red squirrels.
• There were too few bobcat detections to test what affected them.
• A survey took 4.1 hours on average.

3. What they concluded

• Detection dogs were highly effective at finding scats of all three carnivores while largely ignoring other species, and gave an efficient and accurate way to collect detection and non-detection data on several species at once.
• The authors suggest the dogs may have increased the number of scats found 5 to 15 times over people searching without a dog (based on the teams’ own subjective ratings).
• Differences between teams can affect results, so dog and handler training needs careful attention, and teams should be assigned to sites at random.
• Bobcats were hard to detect, perhaps because they live at low densities, defecate less, leave scats in hard-to-search places or clumps, or sometimes bury them. Low detectability does not rule out surveys, as long as it can be estimated; 2 to 3 visits per site are recommended.
• Field identifications can be wrong, so DNA should confirm at least a subset of scats, and DNA failure rates should be planned for: the chance of confirming fisher presence from one high-confidence scat on one visit was about 0.55 (0.95 detection × 0.58 DNA success).
• A handler who mistakenly rewards a dog at the wrong scat can teach it to indicate on that species.
• Dogs may not be the best choice where another proven method exists, where scats are not needed, in extreme climates or dangerous terrain, where dogs risk injury from wildlife, or in busy public areas.

4. How tracking was or was not used

The study is a scat survey. Dogs found the scats by scent, but people did the sign identification: the teams named each scat’s species from its shape, size and contents, rated their confidence, estimated the age of bear scats (from moistness, odour, texture, mould, insects, contents and comparison with scats of known age) and judged whether scats close together came from one animal at one time. DNA tests then measured how accurate those identifications were, giving a rare test of field scat identification against genetics. No trailing was reported.

5. Tracker or observer certification

Not reported. The paper describes the handlers’ 10-day dog-handling training and says one handler had no previous detection dog field experience, but says nothing about the teams’ training or experience in identifying scats. This matters because the confidence ratings and identifications rest on their judgment; the DNA results give a measure of how accurate they were.

6. Other technologies, skills or innovations

• Professionally trained scat detection dogs, each paired with a trained handler and an orienteer.
• DNA species identification (16S ribosomal RNA mitochondrial gene) by Wildlife Genetics International.
• A new way to decide whether a site counts as “detected” when DNA fails: combining correct identification rates by confidence level with the number of scats found.
• Garmin Etrex handheld Global Positioning System (GPS) receivers; random site selection in a Geographic Information System (GIS, ArcGIS with Hawth’s Analysis Tools); density boards for vegetation openness; a topographic ruggedness index.

7. Implications for conservation (from the paper)

• Scat holds valuable information on genetics, diet, habitat use, dispersal and movement, stress and reproduction, and dogs can find large numbers of carnivore scats.
• The authors suggest researchers consider detection dogs especially when target species are at low density, when finding scats will benefit the project, or when avoiding the biases of trail-based surveys or baits and lures is a high priority.
• Their method for judging confidence that a site is occupied could be used wherever only some samples can be confirmed by DNA.