Identifying Eagle Behaviors Using Low-Cost Activity Indices Derived from High-Cost Accelerometry Data
TRICIA MILLER*, SHELDON BLACKSHIRE, ANDREW MCGANN, ROBERT FOGG, and MICHAEL LANZONE, Cellular Tracking Technologies, Rio Grande, NJ, U.S.A.
Recent advances in telemetry devices have made it possible to collect and recover acceleration data using cellular phone networks (GSM) that typically have larger coverage areas than traditional UHF base stations. In order for researchers to acquire acceleration datasets robust enough to identify different animal behaviors, monetary costs associated with cellular data transfer and costs to manage and interpret complex, large data sets can be significant. To address these cost issues yet provide important information to identify behaviors, we translated raw accelerometry data into an activity index. This activity index provides a summary of an animal’s activity at predefined time steps.
To test the utility of this activity index, we explored the relationship between raw accelerometry, GPS data, and our activity index rates recorded by GSM-GPS tracking devices from Cellular Tracking Technologies, LLC.
We deployed CTT BT3 1000 series telemetry units equipped with a GPS and an accelerometer on five Bald Eagles (Halieeatus leucocephalus) and five Golden Eagles (Aquila chrysaetos) in North America.
CTT’s on-board accelerometers recorded continuous sets of data at 40 Hz (forty points per second). We compared these data to an acceleration threshold that we specified.
Above that threshold, we categorized the data as activity events, with each activity event receiving a time stamp. These activity events were then aggregated over a discrete time window, with selection of a shorter time window corresponding to increased temporal resolution at the expense of a larger data set.
All activity calculations were done onboard the telemetry unit and all GPS, activity indices, and accelerometer data were transmitted via the GSM network.
Using a combination of the activity indices and GPS derived data, we were able to identify perching and roosting, take-off or landing events, powered flight events, and gliding/soaring events. We found that on average, eagles spent 50% of their time roosting, 37.5% perching, and 12.5% flying.
Our analyses illustrate how accelerometer data can be translated to low cost activity information that can be used in combination with GPS data to identify behaviors of eagles.