Censusing and monitoring black rhino (Diceros bicornis) using an objective spoor (footprint) identification technique

1. What they did

Black rhino numbers fell from about 65,000 across Africa in 1970 to about 2500. Zimbabwe set up Intensive Protection Zones to protect them, and monitoring relied on radio collars, transponders, ear-notching and de-horning, all of which mean darting and handling the animals. The authors wanted a non-invasive, low-cost way to count and monitor rhino, so they set out to identify individual animals from their footprints (spoor) in an objective, repeatable way.

• The animals. In the Sinamatella Intensive Protection Zone of Hwange National Park, Zimbabwe, they collected footprints from 15 rhino whose identity was known from ear notches or radio collars: 12 adults, 2 sub-adults and a calf. The rhino were found by tracking with scouts or by radio-telemetry.
• Photographing footprints. Only fresh, clear left hind footprints were used, because the hind foot usually steps on the front print and is less distorted by speed or posture. Each print was photographed from directly above, with a ruler along two sides, usually early in the morning when the light gave good contrast. An umbrella and reflector added contrast when needed. The location was taken with GPS. In total they had 290 footprints in 51 trails (a trail being one animal’s pathway on one occasion), with 12 to 29 footprints per rhino.
• Measuring. On each image, 13 landmark points were placed on the outer edges of the three toes and the heel. Custom software (NiSAS) added 13 more derived points and took 77 measurements (47 lengths and 30 angles), which were also combined into 113 measurements in all.
• Analysing. They used two statistical methods. Discriminant analysis assigns each footprint to one of the known rhino. Canonical centroid plots draw a confidence ellipse around a group of footprints, and whether two ellipses overlap shows whether two trails come from the same rhino; this method can also handle animals that are not yet known. They added a “reference centroid value”, built from the whole library of footprints, as a fixed reference point in each pairwise comparison.

2. What they found

Matching footprints to known rhino (discriminant analysis)

• Accuracy levelled out at about 15 measurements (92.8%) and rose to 95.2% with the 30 best measurements.
• With 30 measurements, 276 of 290 footprints (95.2%) were assigned to the right rhino. When each footprint was left out in turn and then tested as an “unknown”, accuracy was 87.2%.
• When whole trails were left out in turn, none of the 51 trails was matched to the wrong rhino, but six could not be assigned to any rhino (88.2% accuracy).
• Trails with fewer footprints were misclassified more often. The authors estimated that at least eight footprints in a trail are needed for 90% accuracy, and recommend 15 footprints per animal.

Comparing trails (canonical centroid plots)

• Comparing all 15 rhino at once did not always work: one trail overlapped with its own rhino but also with four others.
• Pairwise comparisons with the reference centroid value worked much better. A trail matched the rest of its own rhino’s footprints in 48 of 51 cases (94%), and never matched another rhino’s footprints (714 comparisons, 100%).
• Comparing trail against trail was harder. Individual comparisons were 86.4% correct for same-rhino pairs and 94.8% for different-rhino pairs, but only 18 of 51 trails (35.3%) matched all their own trails and none of the others.
• The more footprints in the two trails being compared, the more often they were matched correctly; with 15 or more footprints between them, correct matching was very likely.
• When small trails from the same rhino were pooled so that every trail had at least eight footprints (30 trails), accuracy was 94.4% for same-rhino pairs and 99.8% for different-rhino pairs.
• Used as a census of these 30 trails, the method gave 16 animals for a population of 15 (93.8% census accuracy).

What the measurements showed

Almost all the useful information came from the outline formed by the three toes.

3. What they concluded

• To their knowledge, this was the first footprint technique to give good accuracy with footprints photographed in the field, and to work for both monitoring and censusing.
• Standardized photography (digital cameras let them check each image on the spot), enough footprints per trail and per animal, the measuring software, pairwise comparisons and the reference centroid value were key to the accuracy.
• They suggest three ways to use it, with estimated accuracy (Fig. 9):
• All animals known (for example, under 30 rhino): collect 15 footprints from each animal; then even a single footprint can be matched with discriminant analysis (87% to 88%).
• At least five animals known: build a library from the known animals and compare each new trail of at least eight footprints against it with centroid plots (94%).
• Unknown population (census): collect as many trails as possible, each with at least eight footprints, and compare them in pairs (94%). Where a full census is not possible, two cow and calf pairs and one single animal, found far apart, could start the library.
• Advantages: non-invasive, so it does not disturb the animals or affect fertility; cheap (about US$10,000 in equipment to monitor 60 rhino, against about US$100,000 in the first year and US$75,000 a year after that for radio-collaring); useful in dense habitat or at low density; footprints can be collected at any time and show where rhino went at night; it uses the skills of the local people employed to protect rhino; and it can work alongside existing methods.
• Limitations: it cannot show where each animal is at any moment, and it needs ground that holds a footprint; very soft or constantly wet ground is unsuitable.
• Footprints might also be used to estimate age and sex, heel cracks might be used for pattern recognition, and the technique could be adapted to other endangered species that leave footprints, especially rare, nocturnal ones such as many carnivores. Field trials under more varied conditions and with other species were planned.

4. How tracking was or was not used

The 15 known rhino were regularly found by tracking them on the ground with scouts, or by radio-telemetry. Once a rhino was found, fresh footprints on its trail were photographed and added to its set. The authors note that cracks on the foot pad are often used by trackers in the field to recognize individuals, but this study used the shape and measurements of the footprint instead. They suggest that, in practice, scouts and field rangers could photograph footprints on routine law-enforcement patrols, that footprints could be collected at waterholes at dawn by backtracking from the water or walking a wide circle (about 100 m out) around it, and that rhino newly identified from footprints could be followed to see them and learn their sex.

5. Tracker or observer certification

Not reported. The scouts who tracked the rhino are not described, and nothing is said about their skill. Because the rhino were already known animals and identification was done by measurement, the results do not depend on the trackers’ identifications. The authors thank the scouts of the Sinamatella IPZ for explaining traditional tracking techniques to them.

6. Other technologies, skills or innovations

• Standardized footprint photography: Pentax 35 mm film cameras at first, then Agfa e-photo 1280 digital cameras, a wooden ruler for scale, and an umbrella and reflector for contrast.
• Trimble Ensign GPS for footprint locations.
• Adobe Photoshop to improve images and place landmark points with a cross-hair tool.
• NiSAS, footprint-measuring software developed by the authors with the SAS Institute, and JMP statistical software.
• A new “reference centroid value” to make pairwise comparisons stable, with a mathematical explanation in the appendix.
• Planned: a fully automated field system that compares each new trail with the known rhino of that area, then neighbouring areas. Footprint identities could be combined with GPS data and mapped in a Geographic Information System (GIS) such as ArcView.

7. Implications for conservation (from the paper)

• Footprint identification offers a non-invasive, cost-effective way to count and monitor endangered black rhino, without darting and handling them.
• Range maps from identified footprints could help place anti-poaching patrols. A few hours of footprint photography can locate several animals, while finding one radio-collared rhino can take a whole day.
• Footprints from accompanying animals can show births and the loss of calves, sub-adults growing up, and even mothers temporarily “exchanging” sub-adults.
• The authors see it as a scientific and ethical duty to keep intrusion into the lives of study animals to a minimum, and see non-invasive techniques as the way forward.