# Traditional Versus Multivariate Methods for Identifying Jaguar, Puma, and Large Canid Tracks

> Is that a jaguar, a puma or a big dog? Using 167 tracks of known animals from 18 zoos, researchers found that people who felt confident identifying these tracks scored anywhere from 37% to 87%, and most published measurement rules got more than one in five wrong. Their new measurement-based keys and models told cats from canids, and jaguars from pumas, with over 95% accuracy on test tracks.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2010-de-angelo-paviolo-traditional-versus-multivariate-methods/

## Citation and access

- **Authors:** Carlos De Angelo, Agustín Paviolo and Mario S. Di Bitetti
- **Year published:** 2010
- **Journal or source:** The Journal of Wildlife Management
- **Full citation:** De Angelo, C., Paviolo, A. and Di Bitetti, M. S. 2010. Traditional versus multivariate methods for identifying jaguar, puma, and large canid tracks. Journal of Wildlife Management 74(5): 1141–1153.
- **DOI:** 10.2193/2009-293
- **Access status:** Hosted (freely available elsewhere)
- **PDF:** https://www.originalwisdom.com/wp-content/uploads/2026/11/2010_De-Angelo_and_Paviolo_etal_traditional-versus-multivariate-methods-for-identifying-jaguar-puma-and-large-canid-tracks.pdf

## Tags

- **Species:** Mammals > Coyote; Mammals > Domestic dog; Mammals > Grey wolf; Humans; Mammals > Jaguar; Mammals > Maned wolf; Mammals > Mountain lion
- **Region:** South America > Argentina; South America > Brazil; South America > Colombia; South America > Paraguay; North America > United States; South America > Venezuela
- **Tracking:** Track casts; Track measurements; Track photographs; Tracking; Tracks; Tracks and signs identification
- **Methods and technology:** Mann–Whitney U test; Observer reliability; Pearson correlation
- **Data type:** New field data
- **Observer qualification:** Other experienced trackers
- **Discipline:** Wildlife tracking
- **Year published:** 2010

## Detailed keywords

- **Species:** Studied: jaguar (Panthera onca), puma (Puma concolor), large domestic dogs and maned wolf (Chrysocyon brachyurus). Tracks came from 28 jaguars, 29 pumas and 8 maned wolves in zoos (mainly adults of both sexes, plus 2 tracks of a young jaguar and 3 tracks of 2 young pumas of about 1.5 years old) and from 35 large dogs, most of them stray dogs, in urban areas. Also tested, in a preliminary check: 6 coyote (Canis latrans) and 8 gray wolf (Canis lupus) tracks taken from field guides. Other species mentioned for comparison: American marten, fisher, mink, polecat, Asiatic black bear, sun bear and tiger.
- **Location and habitat:** Captive animals in 18 zoos in 6 countries: Argentina (Zoo Batán, Parque Ecológico Urbano de Río Cuarto, Buenos Aires City Zoo, Temaikèn, Roque Sáenz Peña Zoo, Tatú Carreta Zoo in La Cumbre, Córdoba City Zoo), Brazil (Zoológico Bosque Guaraní and Refugio Biológico Bela Vista, Foz do Iguaçu), Paraguay (Zoológico Itaipú, Hernandarias), Colombia (Zoológico Santa Fe, Medellín), Venezuela (Parque Zoológico Caricuao, Caracas; Parque Zoológico Las Delicias, Maracay) and the United States (Lincoln Park Zoo, Chicago; Sedgwick County Zoo, Wichita; Caldwell Zoo, Tyler; Philadelphia Zoo; Little Rock Zoo). Dog tracks from urban areas in Argentina. Substrates included mud, dust and wet or dry sand; 78% of zoo tracks were on unprepared ground, the rest on smoothed surfaces. The volunteer identification exercise took place at a jaguar monitoring workshop in Puerto Iguazú, Misiones, Argentina. The authors work in the Atlantic Forest of Argentina, where jaguars and pumas live together.
- **Topics:** telling felid from canid tracks; telling jaguar from puma tracks; qualitative (shape) identification traits such as claw marks, heel pad shape and toe position; testing published measurement criteria (Belden 1978; Smallwood and Fitzhugh 1989; Aranda 1994; Childs 1998; Shaw et al. 2007); observer accuracy and variation between observers; heel pad width and track size; size-independent shape measures; identification keys; zoo tracks versus field tracks; track age class (juveniles and dispersing young animals); human and predator conflict; noninvasive sign surveys.
- **Statistics or models:** Stepwise discriminant function analysis (Wilks’ lambda), with leave-one-out cross-validation and independent validation tracks; posterior probabilities of group membership; Mann–Whitney test and F test of variances (volunteers versus random classification); box plots and independent-sample t-tests to preselect variables; Pearson correlations to remove redundant variables; Shapiro–Wilk, Levene’s and Box’s M tests; data transformations. SPSS for Windows. Tracks digitised and measured in AutoCAD 2004. Inconsistencies noted: in Table 3, model B1’s score ranges leave gaps (below -1.2 is “>99% puma” but the next band starts at -0.769; above 1.2 is “>99% jaguar” but the band below ends at 0.769). The text says model B2 misclassified 1 jaguar and 3 puma tracks (4), while Table 3 lists 4 misclassified training tracks (one of them “undetermined”) plus 1 misclassified independent track. In Key 2, step B says DS above 1.95 is jaguar but step B3 runs to 1.93, leaving a small gap.
- **Data type:** New field data: tracks collected from captive animals and urban dogs, 2004 to 2008, as plaster casts (98), photographs with a scale (57) and tracings on acetate (12), 167 tracks in all. Volunteer identification exercise, 5 to 6 October 2005. Existing published identification criteria were tested against these tracks.

## Synopsis

### 1. What they did

Jaguars and pumas are the largest cats in the Americas and share most of their range. Researchers and managers often rely on their tracks, but the two are hard to tell apart and both can be confused with the tracks of large canids such as dogs and maned wolves. The authors set out to test the traditional ways of identifying these tracks and to build more accurate measurement-based methods.

- Collecting known tracks (2004 to 2008). With 18 zoos in 6 countries, they collected tracks from 28 jaguars, 29 pumas and 8 maned wolves, and from 35 large dogs in urban areas. Animals walked freely; tracks were not pressed by hand. Collectors followed general advice rather than a fixed protocol, so that the tracks would vary the way field tracks collected by different people do. They used 167 tracks: front and hind, left and right, on many substrates and of mixed quality. They set aside 35 tracks for independent testing only.

- Testing people (October 2005). At a jaguar monitoring workshop, 67 participants from Argentina, Brazil and Paraguay, who considered themselves able to identify these tracks, first reviewed the traditional shape traits together. They then classified 30 track photographs (10 each of jaguar, puma and big canid) in a slide show. Their scores were compared with 67 random classifications.

- Testing published methods. They checked how often 4 shape traits and 5 published measurement criteria got the tracks right.

- Building new methods. Each track was digitised and measured, giving 155 measurements and ratios. They narrowed these down to 35 useful variables and used discriminant function analysis (a statistical method that finds the combination of measurements that best separates groups) to build one model for canid versus felid tracks and two for jaguar versus puma tracks: one that uses track size, and one that does not. From simpler versions of the models they built three step-by-step identification keys.

### 2. What they found

People identifying tracks from photographs

- Volunteers got 61.5% of tracks right on average, better than random guessing (35.1%), but nobody got them all right.

- Scores ranged widely, from 37% to 87%, and varied more between volunteers than random guessing did.

- A group of 10 hard tracks (canid, puma and jaguar) was misidentified by more than half the volunteers.

Shape traits

- Claw marks were the most useful: absent in 93.5% of puma tracks and present in 97% of dog tracks.

- The front of the heel pad was flatter or concave in 84.8% of puma tracks, and rounded or pointed in 72.7% of dog tracks.

- The other two traits did poorly. Against expectation, 63.6% of dog tracks had a three-lobed heel pad base (a trait usually given for cats), and 72.7% had rounded inner edges on the outer toes. These traits were also hard to judge.

- No trait separated 100% of tracks, and maned wolf tracks looked like dog tracks.

Published measurement criteria

- Most misclassified more than 20% of tracks. Belden’s toe width ratio misclassified 45% of dog and 24% of puma tracks. Smallwood and Fitzhugh’s toe angle misclassified 21% of dog and 17% of puma tracks.

- Aranda’s toe shape ratio for jaguar versus puma misclassified 87% of puma toes.

- All puma and jaguar tracks had heel pads wider than 3.5 cm, as Shaw et al. described for pumas.

- Childs’ heel pad to track area ratio did better on hind tracks, but only 30 tracks could be tested, and 23% fell in its undecided middle range. Hind heel pad widths overlapped between the species (puma 4.07 to 5.50 cm, jaguar 5.18 to 8.99 cm).

- Jaguar tracks were mostly bigger: average heel pad width 7.14 cm for jaguars and 5.06 cm for pumas.

The new models

- Canids versus felids (model A1): 3 measurements, more than 98% accurate in every check and 100% on the independent tracks. Only one puma and one jaguar track were wrongly called canid.

- Jaguar versus puma, with size (model B1): 7 variables. To stop size dominating, it was trained on the bigger half of puma tracks and the smaller half of jaguar tracks. No training, excluded or independent tracks were misclassified, and 90% were classified with more than 95% probability.

- Jaguar versus puma, without size (model B2): 6 shape variables. Slightly less accurate (95% on independent tracks), with its few errors at medium to high probabilities (58% to 95%).

- Keys: Key 1 (canid versus felid) needs 6 measurements and left about 20% of tracks for the full model. Key 2 (jaguar versus puma, starting with heel pad width: below 4.5 cm puma, above 7.9 cm jaguar) left 41% of training tracks for the full model. Key 3 (no size) left about half.

- In a preliminary test, model A1 and Key 1 correctly called 6 coyote and 8 gray wolf tracks from field guides canids.

- Young animals’ tracks in the sample were classified correctly by Key 2 and model B1.

### 3. What they concluded

- Shape traits are useful but subjective, and people vary widely in how well they apply them. Field experience does not always mean skill in identifying signs. The authors think differences in tracking experience are the main reason for the spread of scores.

- Published single-measure criteria performed poorly on varied tracks, probably because they were built from few known tracks, or from field tracks that were themselves identified by shape, often in one region.

- The new models are robust and could be used on unknown field tracks collected by different people in different ways. The canid versus felid model does not depend on size, so it should work across regions and with any large canid.

- Zoo and urban tracks may differ from wild tracks in substrate, paws and gait, which the authors could not account for.

- Their advice for field use: (1) only use the keys and models on tracks with heel pads wider than about 4 cm, to avoid confusion with young animals and smaller species; (2) start with the keys; (3) use the full models when the keys cannot decide; (4) compare the score with Table 3 to judge confidence. Use model B2 where pumas are large or jaguars small, or when a photo has no scale. When an identification matters a lot (a new locality, or identifying the predator in a conflict), have more than 2 tracks classified with more than 80% probability, classify each track separately rather than averaging, and treat poor tracks with caution.

- Limitations: some measurements are hard and slow to take, need digitising, and the calculations need mathematical skill. The keys and models are provided in a spreadsheet as online supplemental material.

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

The study is about tracks and signs identification of footprints. Tracks of known animals were collected as plaster casts, scaled photographs and tracings, then measured. Volunteers identified track photographs in a classroom exercise, which the authors note was not a real field situation. The paper also compares shape traits of single tracks with traits of a whole trail (pattern of tracks), but did not test trail traits. No trailing was done.

### 5. Tracker or observer certification

Tracks were collected by the authors and zoo staff from animals of known species, so no identification was needed. The 67 volunteers in the identification exercise (mainly biologists and park rangers) considered themselves able to identify these tracks and had field experience, but not all had the same tracking expertise. No certification is reported. Their skill was tested in the study, and their scores ranged from 37% to 87%.

### 6. Other technologies, skills or innovations

- Tracks digitised with the spline tool in AutoCAD 2004, scaled and lined up on a base line between the outer heel lobes, and measured to 0.01 cm.

- Detailed measurement scheme for heel pad, toes and whole track, including areas, perimeters and angles (Figure 1 and Appendix C).

- Step-by-step identification keys and a spreadsheet that calculates the scores (online supplemental material).

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

- The method gives a more systematic way to use jaguar and puma tracks with confidence to survey presence, relative abundance and habitat use, without relying on the opinion of experts in track identification.

- Tracks can then serve more confidently as a complementary source of information, for example in placing camera traps or detecting use of wildlife corridors.

- Tracks are often evidence for identifying the predator in attacks on livestock or people, or for the presence of potentially dangerous animals near towns or recreation areas. Management actions may differ for puma, jaguar or big canid, so correct identification is extremely important.

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

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

- The volunteers’ scores (37% to 87% on the same photographs) show why an observer’s tracks and signs identification skill needs to be tested rather than assumed from field experience, which is what tracker evaluations set out to do.

- The test used single photographs without trail context, and the authors note that trail patterns are also used in the field. How well experienced trackers do with whole trails and real substrates was not tested.

## Cite this summary

Lawrence, K. 2026. Summary of: Traditional Versus Multivariate Methods for Identifying Jaguar, Puma, and Large Canid Tracks (De Angelo and Paviolo et al., 2010). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2010-de-angelo-paviolo-traditional-versus-multivariate-methods/