Search and foraging behaviors from movement data: A comparison of methods

Area of Science:

• Animal behavior
• Movement ecology
• Foraging energetics

Background:

• Search behavior is a proxy for foraging effort, but prey capture is also critical.
• Many studies infer prey capture from movement data alone, with limited validation.
• The correlation between identified search behavior and actual prey capture is largely untested.

Purpose of the Study:

• To evaluate seven behavioral annotation methods for identifying northern gannet search behavior.
• To compare method outputs with dive data from time-depth recorders to assess prey capture correlation.
• To guide the selection of appropriate methods for movement studies lacking direct prey capture validation.

Main Methods:

• Applied seven distinct behavioral annotation algorithms to GPS tracking data of northern gannets.
• Utilized simultaneous time-depth recorder data to identify prey capture events (dives).
• Quantified the proportion of dives within algorithm-identified search areas and assessed false positives/negatives.

Main Results:

• Significant variation existed in how well methods identified search behavior linked to dives.
• Hidden Markov models (HMMs) were most successful, with 81% of dives occurring within HMM-identified search areas.
• k-Means clustering and first passage time methods showed higher rates of dives outside identified search areas.

Conclusions:

• No single method perfectly balances accuracy, analytical complexity, and prey capture correlation.
• Hidden Markov models offer a robust approach for linking movement-derived search behavior to prey capture.
• Careful selection of behavioral annotation methods is crucial for accurate ecological inference and effective species management.