Application of activity sensors for estimating behavioral patterns

The increasing use of Global Positioning System (GPS) collars in habitat selection studies provides large numbers of precise location data points with reduced field effort. However, inclusion of activity sensors in many GPS collars also grants the potential to remotely estimate behavioral state. Thus, only using GPS collars to collect location data belies their full capabilities. Coupling behavioral state with location data would allow researchers and managers to refine habitat selection models by using diel behavioral state changes to partition fine-scale temporal shifts in habitat selection.We tested the capability of relatively unsophisticated GPS-collar activity sensors to estimate behavior throughout diel periods using free-ranging female elk (Cervus canadensis) in the Jemez Mountains of north-central New Mexico, USA, 2013–2014. Collars recorded cumulative number of movements (hits) per 15-min recording period immediately preceding GPS fixes at 0000, 0600, 1200, and 1800 hr.We measured diel behavioral patterns of focal elk, categorizing active (i.e., foraging, traveling, vigilant, grooming) and inactive (i.e., resting) states. Active behaviors (foraging, traveling) produced more average hits (0.87 ± 0.69 hits/min, 4.0 ± 2.2 hits/min, respectively; 95% CI) and inactive (resting) behavior fewer hits (−1.1 ± 0.61 95% CI). We differentiated active and inactive behavioral states with a bootstrapped threshold of 5.9 ± 3.9 hits/15-min recording period. Mean cumulative activity-sensor hits corresponded with observed diel behavioral patterns: hits increased during crepuscular (0600, 1800 hr) observations when elk were most active (0000–0600 hr: d = 0.19; 1200–1800 hr: d = 0.64) and decreased during midday and night (0000 hr, 1200 hr) when elk were least active (1800–0000 hr: d = −0.39; 0600–1200 hr: d = −0.43). Even using relatively unsophisticated GPS-collar activity sensors, managers can remotely estimate behavioral states, approximate diel behavioral patterns, and potentially complement location data in developing habitat selection models.