A probabilistic algorithm to process geolocation data.

Area of Science:

• Biologging and animal tracking
• Ecological modeling
• Wildlife movement ecology

Background:

• Light level loggers (geolocators) are increasingly used to study animal movements.
• Inferring precise locations from geolocator data presents challenges due to assumptions and subjectivity.
• Accurate tracking is crucial for understanding species distribution, especially during non-breeding periods.

Purpose of the Study:

• To develop an intuitive and accurate framework for computing animal locations from geolocator data.
• To improve upon existing methods for position estimation from light-level recorders.
• To provide a robust tool for analyzing animal migration and distribution patterns.

Main Methods:

• Developed an iterative forward step selection framework to process twilight data from geolocators.
• Weighted potential locations using analysis-specific parameters.
• Tested the method on black-browed albatross and wandering albatross using data from solstice and equinox periods.
• Integrated travel speed, sea surface temperature, and land avoidance data.

Main Results:

• The new framework significantly reduced location errors compared to the traditional threshold method.
• Median errors were 185 km for solstice and 145 km for equinox periods.
• The approach demonstrated higher accuracy by incorporating additional behavioral and environmental data.

Conclusions:

• The algorithm provides stable location estimates with uncertainty, even around equinoxes, without needing solar angle calibration.
• Integrating travel speed, behavior, and environmental data further enhances accuracy.
• The framework is available as the open-source R package probGLS for broad application in biologging studies.