Identification of carnivore kill sites is improved by verified accelerometer data

  • June 8, 2020
  • by Tyler R. Petroelje, Jerrold L. Belant, Dean E. Beyer & Nathan J. Svoboda


Background: Quantifying kill rates is central to understanding predation ecology. However, estimating kill rates and
prey composition in carnivore diets is challenging due to their low densities and cryptic behaviors limiting direct
observations, especially when the prey is small (i.e., < 5 kg). Global positioning system (GPS) collars and use of collarmounted activity sensors linked with GPS data can provide insights into animal movements, behavior, and activity.

Methods: We verified activity thresholds for American black bears (Ursus americanus), a bobcat (Lynx rufus), and
wolves (Canis spp.) with GPS collars containing on-board accelerometers by visual observations of captive individuals’ behavior. We applied these activity threshold values to GPS location and accelerometer data from free-ranging carnivores at locations identified by a GPS cluster algorithm which we visited and described as kill sites or non-kill sites. We then assessed use of GPS, landscape, and activity data in a predictive model for improving detection of kill sites for free-ranging black bears, bobcats, coyotes (C. latrans), and wolves using logistic regression during May–August 2013–2015.

Results: Accelerometer values differed between active and inactive states for black bears (P < 0.01), the bobcat (P <0.01), and wolves (P < 0.01). Top-performing models of kill site identification for each carnivore species included activity data which improved correct assignment of kill sites by 5–38% above models that did not include activity. Though inclusion of activity data improved model performance, predictive power was less than 45% for all species.
Conclusions: Collar-mounted accelerometers can improve identification of predation sites for some carnivores as
compared to use of GPS and landscape informed covariates alone and increase our understanding of predator–prey