Meeting Abstract
Tri-axial accelerometers have been used to remotely identify the behaviors of a wide range of cursorial mammals. Assigning behavior to the accelerometer data often involves the use of captive animals or surrogate species. While it is assumed that accelerometer signatures of captives or surrogates are similar to those of their instrumented wild counterparts, this has rarely been tested. Here we developed accelerometer-behavior signature libraries for two species of ursid that use a plantigrade style of locomotion to remotely distinguish walking, swimming, eating, and resting. Data from captive and wild animals were compared by video-taping captive polar bears (Ursus maritimus; n=3) and brown bears (U. arctos; n=2) while they wore accelerometer-equipped collars in their enclosures and by collecting video from accelerometer-equipped camera collars deployed on wild polar bears on the spring sea ice of the Beaufort Sea (n=5). A subset of data from all bears was used in random forest models to predict behaviors. With the remaining subset we discriminated walking and resting with >95% accuracy and eating and swimming with >50% accuracy. We assess the causes of this discriminatory variation and evaluate the ability of captive data from polar bears or brown bears to predict wild polar bear behavior. By validating the use of accelerometer signatures in free-ranging ursids, this study also provides a framework for understanding how habitat conditions such as declines in Arctic sea ice may influence activity patterns and energy demands in the polar bear.
