Identifying individual mountain lions Felis concolor by their tracks: refinement of an innovative technique

1. What they did

Large carnivores such as mountain lions are rare, secretive and spread over huge areas, so counting them by radio-telemetry is costly and slow. Tracks last longer than the animals stay in one place, so a reliable way to tell individuals apart by their footprints could help count them over large areas at lower cost. In 1993, Smallwood and Fitzhugh used measurements of rear tracks and a statistical method called discriminant analysis to separate mountain lions, but they could only assume that tracks from different regions came from different animals.

This study tested the idea on animals whose identity was known. In Round Valley, California, nearly every mountain lion had been captured and radio-collared for a study of predation on mule deer. The researchers asked four questions:

• Are tracks from the same mountain lion, made in different places and conditions, recognised as the same animal?
• Are tracks from different mountain lions, made in similar conditions, recognised as different animals?
• Is one track set, photographed at different times of day and under different cloud cover, recognised as one animal?
• How does the age of a track set affect the result?

Photographing tracks. A photographer with a tripod-mounted camera photographed tracks with a scale beside each one, in snow and soils from sandy loam to clay, on slopes under 35%, at different times of day and with filters, flashes and umbrellas. A second person used telemetry to confirm which collared mountain lion made each track set. The first author was not told the identities until the analysis was finished (the authors call this double-blind). One track set was photographed from a few hours old to one week old.

Measuring tracks. From each photo they took three lengths and one angle (as in the 1993 study) and added the areas of the four toes and the heel pad. They measured tracks both by hand (projecting the slides and tracing them onto paper) and by computer (digitising them in ArcInfo, a Geographic Information System (GIS)). They then compared the track sets with Fisher’s discriminant analysis, which groups track sets with each other rather than sorting them into a fixed number of animals decided in advance.

2. What they found

• Sample: 12 track sets were studied: nine from three known mountain lions and three whose identity was uncertain because two mountain lions were close together. Nine sets, with 4 to 16 tracks each, went into the main analysis.
• Left and right feet: left and right rear tracks of the same animal did not differ, so they were combined. Front tracks were dropped because there were too few; mountain lions often step into their front tracks with their rear feet.
• Most useful measurements: the toe areas (outer toe best) and the outer toe spread varied most between track sets. Heel width and heel pad area were the least useful.
• Manual versus computer: the two kinds of area measurement agreed closely (correlations 0.81 to 0.95) and gave the same groupings, so hand measurements were used.
• Correct groupings: two sets from one mountain lion grouped together, and six sets from a second mountain lion grouped together.
• Incorrect grouping: one set from a third mountain lion overlapped with the second animal’s group. The analysis would have counted two mountain lions where there were three.
• Track age and time of day: of one track set photographed on three days (fresh at 07:30, several days old at 17:30, one week old at 07:30), the photo taken in the evening did not overlap the other two, so it looked like a different animal. Time of day seemed to add more variation than track age.
• More dimensions: the first two discriminant functions explained 75% of the differences between track sets, three explained about 91% and four about 96%. Using three improved separation.
• Photo enhancement: it did not improve most tracks, and the computer struggled to separate track from surrounding soil or snow. Computer measuring took a long time and did not improve the groupings.

3. What they concluded

• The method does not give complete reliability. Errors come from different substrates, different times of day (shadows make tracks look bigger or smaller in photos), plotting in only two dimensions, and measurement error.
• Fine soils, clays and soils with a dusty top layer show clear track edges; pebbly, sandy soils do not. Fresh snow is good but melts quickly and distorts tracks. Snow tracks should not be combined with soil tracks, nor clay with sand.
• Large numbers of tracks per set (preferably more than five), on similar substrates and at similar times of day, are needed. Getting enough tracks was the main difficulty.
• The method is cheap, needs only a camera and accessories, can be done by one person, and hand measuring works as well as computer software.
• Heel width, commonly used by mountain lion trackers to identify an animal, had little power to tell individuals apart compared with toe areas and outer toe spread, and should be used with great caution.
• With good samples on suitable substrates, the technique may help count mountain lions across a landscape over a limited period.

4. How tracking was or was not used

Tracks were the data. The team found, photographed and measured footprints of known mountain lions in soil and snow, and assessed how substrate, slope, light and track age changed the measurements. Telemetry, not trailing, confirmed which animal made each track set. The paper also tests a common tracker practice, using heel width to tell individual mountain lions apart, and finds it weak.

5. Tracker or observer certification

Not reported. The paper does not describe who found and photographed the tracks or their tracking experience. People who helped radio-locate the mountain lions and prepare tracks for digitising are thanked by name.

6. Other technologies, skills or innovations

• 35 mm camera on a tripod with colour slide film and a scale beside each track; filters, flashes and umbrellas to control light.
• Radio-telemetry to confirm the identity of the animal making each track set, with the identities hidden from the analyst until the end.
• Tracks digitised and measured in ArcInfo GIS after adjustment in Adobe Photoshop and Aldus Freehand, compared with hand tracings from projected slides.
• Toe and heel pad areas added to the linear measurements used in the earlier study.
• Fisher’s discriminant analysis with 95% confidence ellipsoids, which does not need the number of animals to be known in advance.

7. Implications for conservation (from the paper)

• Indirect ways of counting large carnivores, such as this one, could be useful in wildlife conservation, particularly in developing countries, where funds are limited and large areas must be surveyed quickly.
• The technique could answer general questions such as how many individuals are in an area, or presence and absence, and could be tested on other rare species such as snow leopards, jaguars, sloth bears and giant pandas.
• Regularly checked, well-placed track transects may offer a way to monitor uncommon species over time.