Predicting the individual identity of non-invasive faecal and hair samples using biotelemetry clusters.

Animal diet and health influence fitness, making individual variation in these markers essential for understanding how individuals and populations respond to their environments. Faecal and hair samples provide a record of this information and can be non-invasively collected from animals in the field. However, physiology, diet, and susceptibility to parasitic infections vary within individuals, requiring repeated samples from individuals. We developed a technique using biotelemetry data for individual identification of non-invasive faecal material and hair sampled from female elk (Cervus canadensis). We non-invasively collected individually genotyped faecal and hair samples from resting sites, then compared the accuracy of supervised machine learning models to predict the individual identities of the samples. We found both the tightness of global positioning system point clusters and activity level surrounding the sample allowed us to positively identify samples belonging to specific individuals with 77% accuracy. Our approach can be applied to other populations for which biotelemetry data are available and is potentially adaptable for other species. Furthermore, application of our approach will reduce the need for individual identification of non-invasive samples using genetic analysis, which is costly and prone to low recovery success. Increased access to physiological, dietary, and health information obtainable from individual non-invasive samples will strengthen our understanding of animal responses to their environments. [ABSTRACT FROM AUTHOR]