Elise Loggers, Montana State University

Telemetry units with a global positioning system (GPS) have transformed our understanding of animal movement and behavior. We typically assume that animal behavior does not affect GPS location fix success, but this assumption may be violated when body position affects orientation of the antennae, compromising satellite acquisition. Accounting for this potential behavioral bias may be important for reliable inferences. Modern radio collars often incorporate onboard sensors, providing auxiliary data streams (e.g., activity) that can help reconstruct behavior and improve location estimates for failed GPS fixes. We used telemetry and activity data from GPS-collared grizzly bears (Ursus arctos) to test whether a two-stage hidden-Markov and state-space modeling framework improved interpolation of failed fix locations. We then investigated how including estimates of failed GPS locations in analyses that rely on location data affects parameter estimates (e.g., movement metrics, behavioral classifications, resource selection).
Fix attempts were 2.29 (CI =2.28–2.30) times more likely to fail during inactivity than activity and were likely the result of bears resting in positions that compromised satellite communication. Our methods accurately identified 97% of inactive fixes, and estimated locations were within 16.5 m (SD=63.4 m) of true locations. Including interpolated inactive locations reduced median movement rates by 2.3 times (CI=2.2–2.5) and increased the proportion of time that behavioral models classified bears as resting. Variation in resource selection coefficients indicated that including estimated inactive locations changed inferences about bear-environment relationships: coefficient magnitudes shifted and confidence intervals did not overlap. Characterizing the relationship between failed fixes and behavior helps reduce the risk of conflating artifacts of data collection with biological patterns, enhancing inference and improving conservation decisions.