COS 11-6
The view from a caribou: A collar + GPS + accelerometer + on-board video = extensive data on an elusive species

Monday, August 5, 2013: 3:20 PM
L100B, Minneapolis Convention Center
Anna A. Mosser, Integrative Biology, University of Guelph, Guelph, ON, Canada
Tal Avgar, University of Alberta
Arthur R. Rodgers, Centre for Northern Forest Ecosystem Research, Ontario Ministry of Natural Resources
John M. Fryxell, Department of Integrative Biology, University of Guelph, Guelph, ON, Canada
Ian D. Thompson, Canadian Forest Service
Background/Question/Methods

Collecting detailed information on animals that are difficult to track in largely inaccessible environments poses a barrier to ecological research. We deployed tracking collars for 18-36 weeks on 46 individual female woodland caribou (Rangifer tarandus caribou) in northern Ontario, Canada. Each collar incorporates a Global Positioning (GPS) unit, providing accurate spatial information each hour, and an accelerometer, recording continuous measures of animal activity and fine-scale body movement. The collars also house a high-resolution video camera, which records 10 second video clips every 5 minutes for 5 hours each day. The camera, mounted at the base of the neck, provides a wide range of information, from forage selection and behavioral patterns to parturition dates, grouping patterns, and insect harassment. Here we focus on behavioral patterns and matching accelerometer signatures with each behavior category. This information will allow us to determine behavioral patterns and estimate energetic expenditure throughout the day, as well as for tracked animals that have collars without video.

Results/Conclusions

We report on data from the first 16 collars, those retrieved from the field so far. Using a classification tree, we were able to isolate three behaviors (lying down, feeding, and walking) with distinct accelerometer measurements. When the classification was applied to the test dataset, 78% of the behaviors were correctly identified. As additional data become available, we expect that other primary behaviors (standing and running) will be able to be identified based on accelerometer information. Accelerometer scores were positively correlated with behaviors of higher energetic cost, thus allowing for an energetic calibration of the accelerometer measures and continuous accounting of energy expenditure during the deployment period.