# Monitoring lions (Panthera leo) using digital 3D models of their tracks

> Can a handful of photographs replace the plaster cast? In two KwaZulu-Natal parks, a researcher built 3D models of lion paws, tracks and whole trails from ordinary camera shots, in under a minute per track. The models told front from hind and left from right, and separated males from females, though telling individual lions apart from new tracks still needs work.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2017-marchal-monitoring-lions-panthera-leo/

## Citation and access

- **Authors:** Antoine F.J. Marchal
- **Year published:** 2017
- **Journal or source:** PhD thesis (joint degree: PhD Zoology, University of Pretoria, South Africa, and PhD Agricultural Sciences and Bio-engineering, University of Liège, Gembloux Agro-Bio Tech, Belgium)
- **Full citation:** Marchal, A.F.J. 2017. Monitoring lions (Panthera leo) using digital 3D models of their tracks. PhD thesis, University of Pretoria, Pretoria, South Africa, and University of Liège, Gembloux, Belgium. 151 pp.
- **Access status:** Hosted (freely available elsewhere)
- **PDF:** https://www.originalwisdom.com/wp-content/uploads/2026/11/2017_Marchal_monitoring-lions-using-digital-3d-models-of-their-tracks.pdf

## Tags

- **Species:** Mammals > Lion
- **Region:** Africa > South Africa
- **Tracking:** Gaits; Individual identification; Track measurements; Track photographs; Tracking; Tracks; Tracks and signs identification
- **Behaviour and ecology:** Behaviour
- **Methods and technology:** Analysis of variance (ANOVA); Camera traps; Direct observation; Mann–Whitney U test; Wilcoxon matched pairs test
- **Data type:** New field data
- **Observer qualification:** Not reported
- **Discipline:** Wildlife tracking
- **Year published:** 2017

## Detailed keywords

- **Species:** Studied: lion (Panthera leo), paws of captured lions and tracks and trails of wild lions of known and unknown age, sex and identity. Mentioned for comparison only: tiger (pugmark census method in India), mountain lion, snow leopard, leopard, cheetah, African wild dog, black and white rhinoceros and other species whose tracks have been measured in earlier studies; spotted hyena and stray domestic dogs (livestock losses wrongly blamed on African wild dogs near Hluhluwe-iMfolozi Park).
- **Location and habitat:** Hluhluwe-iMfolozi Park (about 900 km², hilly woodland savanna with shrub thicket, dry riverbeds; 40 to 560 m above sea level) and Tembe Elephant Park (about 300 km², sandy plains with ancient coastal dunes, sand forest and mixed woodland, sandy roads), KwaZulu-Natal, South Africa. Both parks are fenced and managed by Ezemvelo KZN Wildlife. Tracks were recorded on flat sandy riverbeds and roads, a few on substrate prepared with a rake in front of camera traps.
- **Topics:** non-invasive monitoring; close-range digital photogrammetry (“virtual plaster cast”); digital three-dimensional (3D) models of paws, tracks and trails; track and trail measurements (pace, stride, straddle and 40 trail variables); gait (direct register, overstep, mix); front or hind and left or right foot identification; age, sex and individual identification from tracks; traditional and geometric morphometrics (fixed landmarks, curve and surface semi-landmarks); manipulator (observer) bias; substrate variation; criticism of the tiger pugmark census method; possible uses in occupancy, capture-recapture and total counts, human-wildlife conflict (identifying problem animals), community and citizen science monitoring, trophy hunting (age identification) and recording poachers’ footprints; the art of tracking and Liebenberg’s levels of tracking (introduction).
- **Statistics or models:** Mann-Whitney U test; Wilcoxon signed ranks test; Spearman’s rank-order correlation; Kruskal-Wallis test; t-test; Principal Component Analysis (PCA); Linear Discriminant Analysis (LDA) with and without jack-knifed (leave-one-out) prediction; Generalised Procrustes Analysis; Procrustes ANOVA and repeated-measures nested ANOVA (landmark repeatability); multivariate analysis of variance (MANOVA); thin-plate spline deformation grids. Software: Agisoft PhotoScan, CloudCompare, Photoshop, R (Geomorph package). Inconsistencies in the thesis: in Chapter 5, Table 5.1 lists 47 paw rows, with HiP618 and HiP628 each listed twice and HiP614 listed as both “HiP 614” and “HIP614”, while the text says 45 individuals; the Chapter 5 discussion gives the 97.41% individual identification result as Figure 5.7a, but the figure caption puts individual identification in Figure 5.7b; Table 4.1 reads “Direction observation” (for direct observation).
- **Data type:** New field data, collected by the author in Hluhluwe-iMfolozi Park and Tembe Elephant Park between January 2013 and April 2016 (trails: June to December 2015). Paws were sampled opportunistically during captures for park management. Chapters 2 and 4 were also published as journal papers (Journal of Zoology 2016; African Journal of Wildlife Research 2017).

## Synopsis

### 1. What they did

Tracks offer a low-cost, non-invasive way to study animals that are hard to see, but using them to monitor populations has been controversial. Earlier methods recorded tracks only in two dimensions (drawings, plaster casts, photographs), depended on the person recording them, ignored the effect of the ground, often confused which foot made a track, and judged age, sex and identity subjectively. This thesis asked whether lions could be monitored from digital three-dimensional (3D) models of their tracks.

The author worked in Hluhluwe-iMfolozi Park and Tembe Elephant Park, South Africa, from 2013 to 2016. Lions there are closely managed, many are known individuals (collars, brands and scar patterns), and they often walk in sandy riverbeds and on roads. Paws were photographed while lions were immobilised for management purposes. Tracks under 24 hours old, of walking lions, were photographed after the lion was seen, after it walked past a camera trap, or after the tracks were identified as lion with a tracking book.

The thesis has four studies:

- Virtual plaster cast (Chapter 2). Using 20 paws and 20 tracks, the author tested close-range photogrammetry (building a 3D model from 10 to 15 ordinary photographs taken from different angles, with a ruler for scale). Digital measurements were compared with calliper measurements, and the minimum number of photographs was worked out.

- 3D trails (Chapter 3). 26 trails (4 to 33 track sets each) were photographed by walking along both sides of the trail, as in aerial photogrammetry. Pace, stride and straddle, plus 40 distances, angles and areas, were extracted and used to tell sexes and individuals apart.

- Which foot (Chapter 4). Twelve landmarks were placed on 3D models of 24 paws (all four feet of six lions) and 170 tracks from 10 trails, to identify front or hind and left or right from shape alone.

- Age, sex and individual (Chapter 5). 81 paws (front left and hind left) from 45 lions and 116 good-quality hind left tracks from 25 trails were analysed using size and shape, with three levels of detail: fixed landmarks only, landmarks plus curves, and landmarks plus curves and surfaces.

### 2. What they found

Recording paws and tracks in 3D

- Taking the photographs took less than a minute per paw or track, with only a digital camera and a ruler.

- After a correction factor, the 3D models matched calliper measurements to within about half a millimetre on average (for example -0.01 ± 0.39 cm for tracks that were not masked).

- At least seven photographs were needed for a paw and eight for a track; the author advises taking 14 to 16, because about 7% of photographs are of poor quality.

- On tracks, the edge between the track and the surrounding ground is hard to see in photographs. Contour lines drawn from the 3D depth gave a more objective outline than tracing the edge by hand.

Trails

- The 3D trail measurements matched tape measurements closely (mean difference 0.06 ± 2.38 cm). Photographing the longest trail (66 tracks, over 20 m) took 7 minutes; building its model took over 7 hours on a computer.

- As one lioness sped up from a direct register walk to an overstep walk, her pace increased while stride and straddle decreased.

- Among overstepping adults, males had a significantly longer stride (mean 153.2 cm against 122.44 cm) and wider straddle than females.

- Using pace, stride and straddle alone, the stricter (jack-knifed) test identified sex correctly in 96.9% of cases and individuals in 77.4%. Using all 40 trail variables raised this to 100% for sex and 90.3% for six individual adults.

Which foot made the track

- Shape alone identified the foot correctly in up to 100% of paws (95.8% in the jack-knifed test) and 95.3% of tracks (92.9% jack-knifed).

- The small outer toe is shorter and closer to the main pad, and the middle toe reaches highest. Front paws and tracks are larger and rounder than hind ones.

- More front tracks than hind tracks had to be discarded as poor quality, which the author links to lions carrying most of their weight on the front legs.

Age, sex and individual

- Paw size increased with age, and males had bigger paws than females at every age. Paws of juvenile males overlapped in size with those of subadult and adult females, so size or shape alone was not enough; together they classified age and sex best.

- Hind paws carried more age and sex information than front paws (73.2% against 42.5% in the jack-knifed test).

- Adding curves and surfaces did not help with paws, but did help with tracks. From hind left tracks, age and sex were identified correctly in 85.1% of jack-knifed cases.

- Individual identification from tracks reached 97.4% when tested on the same data used to build the model, but only 45.7% in the jack-knifed test for 25 individuals.

### 3. What they concluded

- Digital close-range photogrammetry is a rapid, practical, low-cost and reliable way to record lion paws, tracks and whole trails in 3D in the field, and the digital models allow far more measurements than can be taken by hand.

- Trails, not just single tracks, carry information on the sex and identity of lions walking at the same gait, and could also show body condition (limping, snares, pregnancy, a full stomach) and behaviour.

- The foot that made each track can be identified objectively from shape, which allows like-for-like comparisons (for example hind left tracks only).

- Combining size and shape gives the best identification of age and sex. Individual identification from tracks is not yet reliable for new tracks; the author expects larger samples and methods such as neural networks to improve it.

- This is a first attempt. Future work should measure manipulator bias, the effects of light, cameras, substrate, terrain and body condition, automate outlining and landmarking, compare 2D with 3D, combine track and trail variables, and test the method against established monitoring methods.

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

Lion tracks and trails were the subject of the study. Tracks were identified as lion after a sighting, from camera trap photographs, or with tracking books (Liebenberg 1990; Gutteridge and Liebenberg 2013). The foot that made each track was identified from its position in the trail, and gaits were read from the placement of tracks (direct register as a normal walk, overstep as a fast walk). Track age (under 24 hours) was set from the time of the sighting or camera trap photograph. Where lions could not be identified, trails of lions walking side by side in the same direction were treated as different individuals. Two trails in Chapter 5 were made by courting and mating lions. No trailing to find lions is reported as part of the method. The introduction reviews the art of tracking, Liebenberg’s simple, systematic and speculative tracking, and the accuracy of San trackers reported by Stander et al. (1997).

### 5. Tracker or observer certification

Not reported. The author recorded all tracks and trails himself (paws with the help of field biologists). His own tracking experience is not described; species identification relied on sightings, camera traps or tracking books. The thesis names the recording technique and manipulator bias as sources of error and recommends measuring manipulator bias in future, but no observer test was done apart from the repeatability of placing landmarks on the 3D models (97.6% for paws, 92.5% for tracks, one person).

### 6. Other technologies, skills or innovations

- Close-range photogrammetry with consumer digital single-lens reflex cameras (Nikon D7100 and D80) and Agisoft PhotoScan, described as the first use of photogrammetry to record paws, tracks and trails of living animals in 3D.

- A trail sampling protocol borrowed from aerial photogrammetry: four passes along the trail and clusters of photographs at each end, about one photograph every 30 cm.

- 3D landmark-based geometric morphometrics in R (Geomorph), with curve and surface semi-landmarks; depth contours and colour gradients to help outline tracks.

- Cuddeback camera traps to identify the lions that made tracks.

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

- Identifying species, age, sex, body condition and individuals from tracks could reduce the need for invasive methods such as capture, branding and collaring, though it will not replace them (for example GPS collars that warn when a lion leaves a fenced park).

- Tracks could feed into occupancy surveys, relative abundance indices, pride-by-pride total counts and capture-recapture estimates.

- Because only a camera and a scale are needed in the field, park staff, local community members, traditional trackers and citizen scientists could collect tracks, with 3D processing and analysis done elsewhere and models shared online.

- In human-wildlife conflict, tracks are often the only evidence left by problem animals; correct identification could avoid killing the wrong animal or species and help with compensation claims.

- Ageing lions from tracks could support sustainable trophy hunting, and the method could extend to other species and to poachers’ footprints.

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

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

- The gait findings echo what trackers read in the field: the same lioness left very different pace, stride and straddle measurements as she sped up, so trail measurements only compare fairly between animals moving at the same gait.

- The jack-knifed results are the ones to watch: individual identification from tracks fell from 97.4% on the training data to 45.7% on new tracks, showing how much a test on new data matters before any track-based method is used for counts.

## Cite this summary

Lawrence, K. 2026. Summary of: Monitoring lions (Panthera leo) using digital 3D models of their tracks (Marchal, 2017). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2017-marchal-monitoring-lions-panthera-leo/