Classifying behavior from short-interval biologging data: An example with GPS tracking of birds.

Area of Science:

• Ecology
• Bioinformatics
• Animal Behavior

Background:

• Digital data collection generates vast datasets, posing analytical challenges for ecological studies.
• Traditional methods for classifying bird movement behaviors are often unsuitable for large-scale biologging data.

Purpose of the Study:

• To apply K-means clustering for classifying bird behaviors using high-frequency GPS track data.
• To demonstrate the utility of K-means clustering in identifying behavioral variations related to life stage and age.

Main Methods:

• Utilized K-means clustering on six focal variables derived from GPS data collected at 1-11 second intervals.
• Applied the algorithm to over 2 million GPS telemetry data points from free-flying bald eagles.

Main Results:

• Identified four distinct clusters corresponding to ascending, flapping, and gliding flight, and perching.
• Mapped behavioral states, confirming alignment with natural history observations of flight patterns and behaviors.

Conclusions:

• K-means clustering is an efficient and effective method for classifying short-interval biologging data to understand bird movement behaviors.
• This approach provides insights into small-scale behavioral variations previously unattainable with other analytical methods.