Quantifying song behavior in a free-living, light-weight, mobile bird using accelerometers.

Area of Science:

• Bioacoustics
• Animal Behavior
• Ecology

Background:

• Understanding animal communication requires continuous, individual-level observation in natural habitats.
• Animal-borne acoustic recorders present challenges for vocal studies.
• Light-weight accelerometers offer a potential method to detect vocalizations through associated body vibrations.

Purpose of the Study:

• To develop and validate a classification model using accelerometer data to identify behaviors in free-living European Nightjars (Caprimulgus europaeus).
• Specifically, to validate the model's ability to accurately classify song behavior.
• To assess the utility of accelerometers for studying individual-based variation in bird vocalizations.

Main Methods:

• Collected one-dimensional accelerometer data from light-weight tags on free-living European Nightjars.
• Developed a classification model to identify four behaviors: rest, sing, fly, and leap.
• Validated song behavior classification using simultaneous GPS tracking and stationary audio recordings.

Main Results:

• The classification model achieved 92% accuracy in detecting song activity.
• Song activity duration from acceleration data strongly correlated with simultaneously recorded song bouts (correlation coefficient = 0.87).
• Audio recorders detected only 8% of classified song bouts beyond a 20m threshold, highlighting limitations.

Conclusions:

• Accelerometer-based identification of vocalizations is a promising tool for studying communication in free-living, small birds.
• This method can overcome limitations of traditional audio recorders in capturing individual-based song variation.
• Further research can refine accelerometer use for detailed bioacoustic studies.