Using shape and size to quantify variation in footprints for individual identification: case study with white rhinoceros (Ceratotherium simum)

1. What they did

Recognising individual animals from their footprints could be a powerful noninvasive tool, but it raises a statistical question: every animal’s footprints vary, mostly because of the ground they walk on, and also because of gait and how old or disturbed a track is. Before trusting any identification method, you need to know whether footprints differ more between animals than within one animal’s own footprints. The authors, who developed the Footprint Identification Technique (FIT) with WildTrack, proposed a simple way to check this using two easily understood measures: footprint size and footprint shape.

They tested the idea on the white rhinoceros on Otjiwa Game Ranch, Namibia, in late 1999:

• Collecting footprints. Over 6 weeks in October and November, the team, together with anti-poaching units, intensively tracked and observed the rhinos. To keep variation down, they photographed only clear footprints made at a walk on level ground, and only left hind footprints (a walking white rhino usually places the hind foot over the front footprint). Every footprint came from a continuous trail, never a single isolated print.
• Knowing who made each trail. Cracks in the sole pad leave ridges in the footprint. These ridge patterns stayed the same through the season and led the researchers consistently to the same animal in the field, where animals were recognised by physical features and their stable social groups. In this way the team identified 26 rhinos from their footprints, and no trail was found that could not be assigned to a known animal.
• Measuring. On each of 961 footprint photos they placed 11 landmarks outlining the three toes, using a ruler in the photo for scale.
• Comparing. They compared footprint size between animals with standard tests, and footprint shape with a measure of shape difference and a permutation test (analysis of similarities, ANOSIM) for every one of the 325 possible pairs of rhinos.

2. What they found

Size

• The three smallest calves had clearly smaller footprints than all other animals.
• Among the rest, size alone often could not separate animals: Table 1 lists, for each rhino, up to eight others it could not be told from by footprint size.

Shape

• Overall, footprint shape differed clearly between rhinos (ANOSIM R = 0.74, P = 0.0001).
• Of 325 pairs, 324 were told apart by shape. Only one pair, the subadult females F1 and S1, could not be separated (R = 0.223). Table 1 does not list F1 and S1 as indistinguishable by size.
• Most pairs separated well: the mean R was 0.826 and the most common value was 0.996. Of the 92 pairs with R between 0.95 and 1, 61 included one of the three smallest calves and 20 included one of two larger calves, confirming that young animals’ footprints are shaped differently (calves’ prints are more rounded, older animals’ more elongated).
• Very clean separation was also found between various pairs of adults. Often it was spoiled only by a few outlying footprints.
• Of the 23 pairs with weaker shape separation (R below 0.5), only 3 could also not be told apart by size.

3. What they concluded

• For footprints represented by landmarks, well-defined measures of size and shape can show how much footprints vary within and between individuals. This is a relatively simple way to judge whether footprints from a population will be useful for identifying individuals, independent of any particular identification algorithm.
• The approach is also an exploratory tool: it can help choose landmarks when building an identification method, and help explain why a method fails to tell certain animals apart.
• It applies to any landmark-based footprint identification scheme, not only FIT.
• Results should improve with a minimum sample of about 25 footprints per individual, removal of outliers, and careful selection of footprints in the field once experience with a species is gained.
• How many individuals a method can tell apart is an open question, but in practice animals whose ranges do not overlap do not need to be distinguished.

4. How tracking was or was not used

The study rests on fieldwork with tracks. The team and anti-poaching units intensively tracked the rhinos, followed trails, and selected and photographed clear footprints. Each trail was assigned to an individual by reading the ridge patterns left by cracks in the sole pad; these trails led consistently to the same animal, which was then confirmed by its body features and social group. The authors say independent observers could reliably distinguish the ridge patterns in the field or in photographs, but give no data on this.

5. Tracker or observer certification

Not reported. The paper does not say who in the team or the anti-poaching units did the tracking and footprint selection, or describe their experience. This matters because the known identities of the 961 footprints, on which all the results depend, come from reading ridge patterns and following trails.

6. Other technologies, skills or innovations

• Digital photographs of footprints with a ruler for scale, and 11 landmarks placed by rule on each footprint (from the FIT method).
• Shape statistics (centroid size and Kendall shape distance) combined with nMDS plots and ANOSIM permutation tests, as a check of whether footprints carry enough information to identify individuals.

7. Implications for conservation (from the paper)

• Identifying individuals from their footprints, especially reclusive, nocturnal or rare species, promises a noninvasive, cost-effective technique for monitoring and research, without having to find or handle the animal.
• The main challenges are confirming that a species is suitable and developing a working identification method. Size and shape are easily understood features that can help with both.