Identifying endangered species from footprints

1. What they did

The authors argue that conservation needs monitoring methods that are cheap, non-invasive and friendly to local communities. Fitting animals with telemetry devices can cause stress, change behaviour and reduce female fertility, and it is costly and rarely involves local people. Camera traps, vocal and coat patterns, and DNA from faeces and hair all help identify individuals, but each can be limited in range, accuracy or cost. Footprints, by contrast, are everywhere on suitable ground, cheap to collect and can be good biometric markers.

After many years working with expert trackers in Africa and seeing how accurately they identified individual animals from footprints along trails, the authors set out to turn those skills into a modern tool. In this short article they describe their footprint identification technique (FIT):

• Photographing. A series of digital images is taken along several trails, following a standardized photo protocol. A metric ruler is placed along the bottom and left side of the footprint, with a slip recording the Global Positioning System (GPS) position, date, photographer and track number.
• Measuring. Each image is imported into JMP software (SAS Institute) and landmarks are placed by hand at specific anatomical points. A script then takes more than 120 measurements of areas, lengths and angles.
• Analysing. The measurements that classify best are picked out by stepwise selection. Trails are compared with a cross-validated pairwise discriminant analysis and Ward’s clustering, which shows individuals as a tree diagram.
• Training and applying. For each species, an algorithm is first built from a training set of footprints from captive animals, then applied to a free-ranging population.

2. What they found

• Every species has its own footprint anatomy, and every individual of a species has its own footprint. Each footprint an individual makes also differs, with the substrate, gait, weather and terrain, so a model has to keep variation within one animal small and variation between animals large.
• FIT gives accuracy rates of greater than 90% in telling species, sex and age-class apart. The article gives no sample sizes; it refers to the authors’ earlier white rhino paper for details.
• Large, well-checked training sets, tested on subsets of known free-ranging individuals, show the technique can identify wild as well as captive animals.
• FIT has been adapted for black and white rhino, Amur and Bengal tiger, polar bear, cougar, cheetah, Baird’s and lowland tapir, and several small mammals at species level. A short video shows footprints being collected from captive and free-ranging Amur tigers in northeast China.
• Placing landmarks by hand, by an expert, is still the most accurate way to classify individuals, as in medical imaging. Automated methods (structured lighting, polynomial texture mapping and invariant moments) have yet to match it.

3. What they concluded

• The authors were writing the first general release of FIT software in JMP (FIT v.1), with variable extraction, data analysis and mapping, as a complete tool for biologists monitoring endangered species.
• They are still looking for automated feature extraction that is accurate, cost-effective and works in the field. With North Carolina State University and Tom Malzbender, they are testing a Kinect motion sensor camera, possibly with structured lighting, to make a depth map of footprints.
• Their aim is a fast, cheap, reliable and user-friendly tool for conservationists to monitor endangered species from footprints.
• They close by saying they remain humbled by the extraordinary tracking skills that evolved with our ancestors thousands of years ago.

4. How tracking was or was not used

FIT is built on footprints: photographing footprints along trails, often in snow, and measuring them to identify species, individuals, sex and age-class. The authors say the method was inspired by expert trackers in Africa whom they watched identify individuals from footprints along trails in the bush. No trailing (following tracks to find an animal) is described.

5. Tracker or observer certification

Not reported. The article does not say who found and photographed the footprints or who placed the landmarks, although it stresses that the accuracy of the method depends on expert landmark placement. The African trackers who inspired the work are called expert, but they are not named and their skill was not tested here.

6. Other technologies, skills or innovations

• Standardized footprint photo protocol with a metric ruler and an information slip (GPS position, date, photographer, track number).
• JMP software (SAS Institute) with a script that takes more than 120 measurements per footprint; FIT v.1 software in development.
• Possible automated methods: structured lighting, polynomial texture mapping, invariant moments, and a Kinect motion sensor camera making depth maps.

7. Implications for conservation (from the paper)

• Conservation needs low-cost, non-invasive, community-friendly monitoring, without the harm that collars and capture can cause.
• Footprints are everywhere, cheap to collect and can identify individuals, so they can give reliable data on endangered populations and individuals.
• Local communities’ commitment is essential to conservation, and invasive methods rarely involve them.