# A morphometric modeling approach to distinguishing among bobcat, coyote and gray fox scats

> Can you tell a bobcat’s scat from a coyote’s or a gray fox’s just by measuring it? Using 110 scats confirmed by DNA in coastal California, this study found that size and shape, even with a chemical clue added, got the species right no more than three times in four. A few clues stood out: every scat beside a scrape belonged to a bobcat, and light, unsegmented, plant-rich scats pointed to gray fox.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2015-reid-morphometric-modeling-approach-distinguishing/

## Citation and access

- **Authors:** Rachel E. B. Reid
- **Year published:** 2015
- **Journal or source:** Wildlife Biology
- **Full citation:** Reid, R. E. B. 2015. A morphometric modeling approach to distinguishing among bobcat, coyote and gray fox scats. Wildlife Biology 21(5): 254–262.
- **DOI:** 10.2981/wlb.00105
- **Access status:** Open access (CC BY)
- **PDF:** https://www.originalwisdom.com/wp-content/uploads/2026/11/2015_Reid_a-morphometric-modeling-approach-to-distinguishing-among-bobcat-coyote-and-gray-fox-scats.pdf

## Tags

- **Species:** Mammals > Bobcat; Mammals > Coyote; Mammals > Domestic dog; Mammals > Grey fox; Mammals > Spotted skunk
- **Region:** North America > United States
- **Tracking:** Scat; Sign surveys; Tracking; Tracks and signs identification
- **Behaviour and ecology:** Behaviour
- **Methods and technology:** Analysis of variance (ANOVA); DNA; Global Positioning System (GPS); Machine learning
- **Data type:** New field data
- **Observer qualification:** Not reported
- **Discipline:** Wildlife tracking
- **Year published:** 2015

## Detailed keywords

- **Species:** Studied: bobcat (Lynx rufus), coyote (Canis latrans) and gray fox (Urocyon cinereoargenteus), through their scats. Also identified by DNA: one spotted skunk scat (no scientific name given), left out of the analysis. Present but avoided: domestic dog scats, recognised in the field by their uniform texture and colour. Comparison from an earlier study: Mexican gray wolf (Canis lupus baileyi) and coyote scats (Reed et al. 2004).
- **Location and habitat:** Central coast of California, United States: Año Nuevo State Park and Reserve (San Mateo County), with native dunes, coastal terrace prairie and mixed evergreen forest; Younger Lagoon Natural Reserve (University of California Natural Reserve System, Santa Cruz County), with dense coastal shrub, willow thickets, coastal prairie, and salt and freshwater marsh; and Moore Creek Preserve (Santa Cruz County), managed coastal prairie that is periodically grazed. Scats were collected along roads and trails on transects of about 5 km running from the coast inland.
- **Topics:** scat identification to species; scat morphology and morphometrics (diameter, length, mass, number of pieces, taper, segmented, ropey, flat); scrapes next to scats; scat position on the trail or road; carbon to nitrogen (C:N) ratio of scat as a measure of how carnivorous an animal’s diet is; DNA-verified reference scats; mesopredators living side by side (sympatric); “end member morphologies”; cost of DNA versus cheaper first-pass identification; misidentification in earlier diet and space-use studies.
- **Statistics or models:** One-way analysis of variance (ANOVA) with Tukey post hoc tests; Shapiro–Wilk test (normality) and Bartlett test (equal variances); log transformation of mass and C:N ratio; linear and quadratic discriminant function analysis (LDA and QDA, in the R package MASS), with a 70% training set, a 30% test set and jackknife (leave-one-out) cross-validation; random forests (cforest in the R package party) with out-of-bag error and variable importance; a single conditional inference classification tree (ctree). Note on inconsistencies in the paper: (1) the text gives mean scat diameters of 13.3 mm (bobcat), 16.45 mm (coyote) and 11.5 mm (gray fox), but the boxplot in Figure 1 shows median diameters of roughly 18, 21 and 16 mm; (2) the text says the best morphology-only LDA model used number of scat pieces, diameter, taper length and log mass, while Table 2 lists length instead of number of pieces (the same for the model with C:N ratio); (3) the text gives identical centroid test results (F2,57 = 47.99 and 16.79) for both LDA models; (4) the random forest misclassified 35 of 110 scats (1 bobcat, 23 coyote, 11 gray fox), which would be about 68% correct, while the out-of-bag correct rate given is 62%; (5) taper length is measured in mm in Table 1 and the Discussion, but Figure 1E is labelled in cm.
- **Data type:** New field data: scats collected every three months (quarterly) from 2011 to 2013. Transects were first cleared of all scats, and scats deposited over the following week were collected. 122 scats were sent for DNA identification; 110 were confirmed as bobcat (57), coyote (28) or gray fox (25). The detailed protocol and the R code for the models are in the paper’s supplementary material.

## Synopsis

### 1. What they did

Scats are widely used to study elusive carnivores: where they live, how many there are, and what they eat. Scats are often identified to species by their shape and size, but scats of coyotes, bobcats and gray foxes can look very alike, and DNA testing is expensive. The author asked whether measurements of scats could reliably tell these three species apart.

- Collecting scats. On the central coast of California, the author cleared roads and trails of scats and came back a week later to collect the new ones. This was done every three months from 2011 to 2013, to cover seasonal changes in diet and weather. Each scat’s location was recorded with a Global Positioning System (GPS) unit.

- Measuring them. In the field: diameter, length, number of pieces, taper length and degree of taper, and whether the scat was segmented, ropey (twisted), both or neither. Also noted: where the scat lay (middle, edge or off the edge of the trail or road) and whether a scrape was next to it. In the lab: dry mass, and the carbon to nitrogen (C:N) ratio, which is lower in the scats of animals that eat more meat. Smell was left out because it cannot easily be measured.

- Confirming the species. 122 scats were swabbed and identified by mitochondrial DNA at Wildlife Genetics International.

- Building models. The author compared the three species’ scats with statistical tests, then built models to predict the species from the measurements: discriminant function analysis (linear and quadratic), random forests (a machine learning method) and a single classification tree.

### 2. What they found

The scats

- Of the 122 scats, 9 failed DNA identification, 2 had mixed DNA and 1 was from a spotted skunk. That left 57 bobcat, 28 coyote and 25 gray fox scats.

- Of the 12 traits, only three differed between species: diameter, mass and C:N ratio.

- Coyote scats were wider than bobcat and gray fox scats, but bobcat and gray fox scats did not differ in width (see the note on figures under Detailed keywords).

- Gray fox scats were lighter (6.6 g on average) than coyote (17.9 g) and bobcat (12.4 g) scats. Coyote and bobcat scats could not be told apart by mass.

- C:N ratio differed between all three: lowest in bobcat scats (6.7), then coyote (8.6), and highest in gray fox (12.1).

- 84% of bobcat scats were segmented.

- Five scats were found next to a scrape, and all five were bobcat.

- Eight scats were irregular, flat “puddles”; five of these were gray fox.

The models

- Measurements of shape and size alone could not reliably separate the three species. The best linear model got 60% right in the training set, 67% in the test set and 62% overall.

- Adding C:N ratio helped a little: 74%, 75% and 71%. Overall accuracy never went above 75% in any model.

- Random forests got 57% right with shape and size alone, and 62% with all traits. Mass was the most important trait, followed by C:N ratio and diameter. The random forest was best at identifying bobcat scats; the quadratic models were better for coyote scats.

- Most wrong identifications happened when the model was unsure (50 to 60% probability). Confident wrong answers were rare (under 5%).

- Some combinations were telling. Two to four gray fox scats were identified with over 90% confidence: they had high C:N ratios, were not segmented and were very light. Three coyote scats were identified with over 87% confidence: they were wide, with long tapers and middling C:N ratios. Scats with a mix of “bobcat” and “coyote” traits were hard to assign.

Cost

C:N analysis costs about $6 to $10 per sample, compared with about $25 per sample for DNA analysis in this project.

### 3. What they concluded

- Bobcat, coyote and gray fox scats are very similar in shape and size. Context such as position on the trail or a nearby scrape adds only a little to correct identification.

- There do seem to be “end member morphologies”: sets of traits typical of one species. A segmented scat with a low C:N ratio is more likely a bobcat’s, and a scrape next to a scat is more likely a sign of bobcat.

- The models can be used as a first-pass tool. Researchers can choose a confidence cut-off (for example 85 or 90%), accept identifications above it, and check the rest with DNA or scat-detecting dogs. This saves some cost.

- Researchers can use these models or build their own for other species, using a subset of DNA-verified scats. The method may work better for species whose scats differ more.

- Fresh tracks near a scat could improve identifications, but this information is often missing when many scats are collected at once.

- Some earlier studies that relied on scat shape alone may have misidentified a larger share of scats than thought, and so misread the diets and space use of these animals.

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

Scats were found and collected by searching roads and trails along set transects, after first clearing them. Scrapes next to scats were recorded, and dog scats were left out because they were recognised by their texture and colour. Species were not identified in the field: every scat in the analysis was identified by DNA, and the study tested whether scat measurements could match those DNA results. Tracks are mentioned as other signs that could help, but they were not used in the models. No trailing was reported.

### 5. Tracker or observer certification

Not reported. The author and several field assistants collected the scats; their experience is not described. Because species were confirmed by DNA, the main results do not depend on the collectors’ skill, although finding scats, measuring them and spotting scrapes did.

### 6. Other technologies, skills or innovations

- Mitochondrial DNA (16S rRNA gene) species tests from swabs of each scat, by Wildlife Genetics International.

- C:N ratio from an elemental analyser and isotope ratio mass spectrometer, as a cheaper measure of how carnivorous the diet was.

- Machine learning (random forests) alongside discriminant function analysis, all in R.

- A step-by-step measuring protocol and ready-to-use R code in the supplementary material, so others can run their own scat measurements through the models.

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

- Identifying scats by shape and size alone can misrepresent the diets and space use of bobcats, coyotes and gray foxes. Researchers and wildlife managers studying these animals need to check such identifications with other methods.

- A model that gives each scat a probability lets researchers decide which scats need costly DNA checks, saving money.

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

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

- For these three species, a scat on its own is often not enough for a confident identification. The paper notes that field identifications are more reliable alongside other signs such as tracks, which supports reading the whole scene rather than the scat alone.

- Context helped in a small way: in this study, every scat beside a scrape was a bobcat’s, and most bobcat scats were segmented.

- The study shows the value of DNA-verified reference scats for checking how well people and models identify scats.

## Cite this summary

Lawrence, K. 2026. Summary of: A morphometric modeling approach to distinguishing among bobcat, coyote and gray fox scats (Reid, 2015). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2015-reid-morphometric-modeling-approach-distinguishing/