Uniquely identifying individual animals within a population is essential to the application of many techniques in research and management of wildlife. A variety of methods are available (e.g., artificial markings, radiocollars), but they can be expensive, invasive, and in some cases alter normal behavior and jeopardize animal health.
With the advancement of data management and photograph technology, non-invasive methods such as the use of photographs to identify individuals and monitor populations have received increased use. Most of the identification software available is species or genera specific and based on the comparison of shapes and patterns (e.g., stripes on zebras).
Comparing measurements of morphological traits (e.g., horn length) however, has received little attention and a software that can deal with a variety of morphological measurements would find broad applications in ecological and behavioural research.
In order to address my research objectives, we developed and tested computer-assisted photograph-identification software capable of incorporating measurement data and metadata (i.e., other un-measureable information specified by the user) from photographs, multiple sources of measurement error, and an interface that allows users to make the final decision about a potential match. This software, called L-PIC (Likelihood-based Photo Identification Code; available as an R package at http://lpic.r-forge.r-project.org/), uses a likelihood-based algorithm to calculate a matching score between pairs of photos.
We tested L-PIC to identify the false rejection rate, using 91 photographs which represent 33 tagged bison (Bison bison). We then used the software to estimate the 2011 adult population size of bison in Prince Albert National Park, Canada. The manuscript reporting these results is available in the Wildlife Society Bulletin (click here for publication list).