Using accelerometer data to refine timing of predation events of white-tailed deer fawns (Odocoileus virginianus)

Deploying global positioning system (GPS) collars equipped with accelerometers on large mammals provides fine-scale activity data in addition to traditional GPS-collar location data. While GPS collars use activity sensors to detect mortality events of collared animals, these mortality alerts from collared prey species are frequently delayed as predators move collars and carcasses after the actual predation event. We aimed to explore a technique to improve accuracy of identifying mortality timing in neonate ungulates. Using data collected from 171 GPS-collared white-tailed deer ( Odocoileus virginianus ) fawns that died by coyote ( Canis latrans ) predation, we used piecewise regression to detect activity that may indicate a flee and subsequent predation event. We used the model changepoints to estimate the true time of mortality. We found two-segment and three-segment piecewise regression models that represented abrupt shifts in activity around mortality events to be successful at identifying a short burst of activity likely associated with a predation event. The piecewise regression selected a three-segment model for 17.5% ( n = 30) of the individuals and a two-segment model for the remaining 82.5% ( n = 141) of individuals. The models detected a changepoint in the activity data on average 13.5 ± 10.1 (SD) hours (range = 0.0–44.5 h, n = 171) earlier than the collar-detected mortality. Many neonatal ungulate survival studies may be overestimating the time of death determined from GPS-collar mortality alerts alone. We found that applying simple piecewise regression to activity data surrounding a mortality event identified abrupt changes in activity that likely corresponded with predation events and provided refined estimates of mortality timing in neonate ungulates. We suggest this approach can help account for temporal error associated with collar mortality alerts, improve survival estimates during biologically relevant periods, and provide additional evidence when assigning cause-specific mortality.