# Validation of a rigorous track classification technique: identifying individual mountain lions

> Can you tell one mountain lion from another just by their footprints? Using silicone molds of 13 real lion feet, researchers made hundreds of tracks in the lab, had five people trace them, and found six measurements that matched 96% of tracings to the right lion. The method also counted the right number of lions in 9 of 10 tests on real field tracks.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2001-lewison-fitzhugh-validation-rigorous-track-classification/

## Citation and access

- **Authors:** Rebecca Lewison, E.L. Fitzhugh and Steve P. Galentine
- **Year published:** 2001
- **Journal or source:** Biological Conservation
- **Full citation:** Lewison, R., Fitzhugh, E.L. and Galentine, S.P. 2001. Validation of a rigorous track classification technique: identifying individual mountain lions. Biological Conservation 99(3): 313–321.
- **DOI:** 10.1016/S0006-3207(00)00197-X
- **Access status:** Publisher link only
- **Where to find it:** https://doi.org/10.1016/S0006-3207(00)00197-X

## Tags

- **Species:** Mammals > Mountain lion
- **Region:** North America > United States
- **Tracking:** Individual identification; Track measurements; Tracking; Tracks
- **Methods and technology:** Observer agreement
- **Data type:** Existing data; New field data
- **Observer qualification:** Other experienced trackers
- **Discipline:** Wildlife tracking
- **Year published:** 2001

## Detailed keywords

- **Species:** Studied: mountain lion (Puma concolor; also called puma or cougar; Felis concolor in older papers). Track identification work on other species mentioned: tiger (Panthera tigris), snow leopard (Panthera uncia), pine marten (Martes martes) and black rhinoceros (Diceros bicornis).
- **Location and habitat:** California, United States: casts from 11 lions killed under depredation permits at six locations in two California counties (described in the Discussion as three areas about 80 to 100 miles apart, each about 200 square miles); one mold from a track in another California county and one from a track in Florida. Field data: five track sets from five locations in Northern California (1987). Laboratory work at the University of California, Davis, in a sandy loam soil (67% clay, 23% silt, 10% sand) at two depths, deep (7 mm or more) and shallow (1 to 3 mm).
- **Topics:** identifying individual animals from their tracks; artificial tracks made from molds of real feet; track tracing on acetate; effects of different track recorders and soil depth on track measurements; heel pad shape; choosing the fewest, most powerful track measurements; number of tracings per track set; double blind validation; testing on independent field data; computer measurement of digitized tracings (SigmaScan Pro 4); track photographs in a standardized “black box”; track surveys and density estimates for large carnivores.
- **Statistics or models:** discriminant function analysis (DFA), with standardized discriminant function coefficients, squared Mahalanobis distances, Wilks’ lambda and classification matrices; multivariate analysis of variance (MANOVA); effect size from coefficients of determination (R²) using Euclidean distances between canonical vectors; Box’s M test; STATISTICA software. Possible inconsistencies or unclear points in the paper: (1) 13 molds traced by five recorders, seven times in deep soil and eight in shallow soil, would give 975 tracings, but the paper reports 797, without explaining the difference. (2) The lab validation used 177 unused tracings to build 72 groups of four track sets each with four, six or eight tracings, which needs far more than 177 tracings, so tracings must have been reused between groups; this is not explained. Likewise, 20 field tracings formed 10 test groups. (3) Table 1 gives F with “d.f. 1, 2” for five recorders. (4) The abstract’s “96% of track tracings” (Table 4, 96.2% of 239 tracings) and the “96% of lab track groups” in the validation (69 of 72) are two different results that share the same figure.
- **Data type:** New field data (laboratory-generated tracks from molds of 13 mountain lion feet, made over about two months; the year is not given) and existing data (field tracings collected by S. Smallwood in Northern California in 1987).

## Synopsis

### 1. What they did

Track surveys are a quick and inexpensive way to monitor large carnivores such as mountain lions, but counts are much more useful if tracks can be linked to individual animals. Earlier work (Smallwood and Fitzhugh 1993) suggested individual lions could be told apart by measuring their tracks, but nobody could check the results, because the identity of the lions in the field was unknown and there were too few tracks from known animals.

The authors solved this by making tracks in the laboratory from known individuals:

- Making the feet. They took plaster casts of the rear right foot of 11 mountain lions killed under depredation permits in California (10 males and one female, aged 2 to 7 years) and turned them into silicone molds. Two more molds came directly from tracks in soil, one from California and one from Florida, giving 13 lions.

- Making the tracks. They pressed each mold into sandy loam soil at two depths, deep (7 mm or more) and shallow (1 to 3 mm), seven times in deep soil and eight in shallow soil, over about two months.

- Tracing the tracks. Five track recorders, trained by Fitzhugh in a standard measuring protocol, traced every track on acetate sheets held on glass just above it, giving 797 tracings of the whole track (heel pad and four toes).

- Measuring. From each tracing they took seven linear and angle measurements of the whole paw by hand, and ten shape measurements of the heel pad with image software.

- Testing. They checked how much the recorder and the soil depth changed the measurements, chose the smallest set of measurements that best separated individual lions, and then tested that set on unused lab tracings and on real field tracings, where the analyst did not know how many lions were present.

### 2. What they found

Effect of recorders and soil depth

- Different recorders and soil depths did change some single measurements significantly.

- But when the measurements were combined, the models explained almost the same amount of variation with or without recorder or soil depth (for linear and shape measurements combined, 96.4% for lion alone, 96.7% with recorders, 96.6% with soil depth). Their overall effect was small.

- Soil depth mattered more than the recorder, especially for heel pad shape.

- Two extra people who digitized the tracings on computer made no difference to the measurements.

The best measurements

- Six measurements were chosen: three of the whole paw (angle between toes, outer toe spread, heel to lead toe length) and three of the heel pad (perimeter, shape factor, major axis length).

- Correct matching of tracings to known lions: 89.9% with the seven linear measurements, 75.3% with the ten heel pad shape measurements, and 96.2% with the six combined.

- Combined, they worked a little better in deep soil (97.4%) than in shallow soil (93.3%). Shape measurements did better in shallow soil, and linear ones in deep soil.

Validation

- On unused lab tracings, in a double blind test where only the person preparing the data knew how many lions were present, the method found the correct number of lions in 69 of 72 groups (96%).

- Using four, six or eight tracings per track set made no difference to the errors.

- On the 1987 field data (20 tracings from five track sets), it found the correct number of lions in 9 of 10 groups.

### 3. What they concluded

- The six-measurement method identified individual lions from their tracks better than any technique so far, and it worked even when the number of lions was not known beforehand.

- The measurements describe real, separate dimensions of the paw and heel pad, and they hold up despite different recorders and soil depths.

- The lab test was harder than real fieldwork in one way: each track set mixed tracings from several recorders, which is very unlikely in a field survey.

- At least four tracings per track set are needed.

- Field data will add other variation not tested here: the flexing of a live foot, uneven ground, speed of travel, other soils, and left and right feet mixed in a track set. The authors expect only small losses in accuracy.

- Because 11 of the lions came from fairly small areas, many were probably related, so the method seems able to separate related individuals.

- Making artificial tracks from known individuals could be used to test measurements and field conditions for other rare or hard-to-study carnivores.

- Image measurement software made the work faster and more accurate, and photographs of tracks could one day replace tracing.

### 4. How tracking was or was not used

The study is about reading individual identity from footprints. Tracks were made in the laboratory from molds of real feet, then traced and measured by hand and by computer. The field test used tracings that S. Smallwood made directly over wild mountain lion tracks in 1987, using the same protocol. No trailing or other signs were used.

### 5. Tracker or observer certification

Not reported as tracker certification. The five track recorders were trained by co-author Fitzhugh in a published track measurement protocol (Fjelline and Mansfield 1989). Before each session, they all traced the same practice track and compared results. Their consistency was then tested in the study itself: differences between recorders changed some single measurements but had little effect on the combined method. Their tracking experience is not described.

### 6. Other technologies, skills or innovations

- Silicone molds made from plaster casts of real feet, to make repeatable tracks from known individuals.

- Tracing on acetate sheets over glass plates raised about 8 mm above the track.

- SigmaScan Pro 4 image software to measure scanned heel pad tracings.

- A standardized “black box” for photographing tracks (Galentine and Fitzhugh 1996) was being tested as a way to measure tracks without tracing.

### 7. Implications for conservation (from the paper)

- Counting mountain lions from small study areas and scaling up to whole regions is unreliable; random track surveys across a whole region would help, but need a good way to tell individuals apart.

- A reliable method to identify individual lions from tracks could make track surveys more precise and lead to more accurate population estimates than are now possible.

- The same approach may help monitoring of other rare or cryptic carnivores.

### 8. Implications for tracking (Original Wisdom's observations)

_These are Original Wisdom's observations, not the authors'._

- The study separates two questions that are hard to untangle in the field: is the track really different, or did the person or the ground make it look different? Using molds of known feet answers this, and the answer here was that recorders and soil depth changed single measurements but not the combined picture.

- The method depends on careful, repeatable tracing by trained people, and on several clear tracks from the same animal (at least four).

## Cite this summary

Lawrence, K. 2026. Summary of: Validation of a rigorous track classification technique: identifying individual mountain lions (Lewison and Fitzhugh et al., 2001). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2001-lewison-fitzhugh-validation-rigorous-track-classification/