A parsimonious approach to modeling animal movement data

Area of Science:

• Ecology
• Animal Tracking
• Biologging

Background:

• Traditional methods for improving animal tracking data, like speed filtering, are insufficient.
• Complex state-space models for location estimation are time-consuming and difficult to apply.
• Improving the accuracy of animal tracking location estimates is crucial for ecological research.

Purpose of the Study:

• To present and validate an intuitive and efficient alternative model for enhancing animal tracking location estimates.
• To assess the model's performance using high-quality GPS-validated ARGOS and geolocation data from elephant seals.
• To demonstrate the model's flexibility in incorporating additional environmental data and solving real-world tracking challenges.

Main Methods:

• Developed a novel approach using bootstrapping random walks biased by forward particles.
• Integrated data accuracy estimates and assimilated external data like sea-surface temperature, bathymetry, and physical boundaries.
• Validated the model with elephant seal tracking data (ARGOS, PTTs, geolocation, and GPS).

Main Results:

• The model provided location estimates within 4.0-12.0 km (50% of the time) and 9.0-20.0 km (90% of the time) for ARGOS tracks of varying quality.
• Geolocation data yielded location estimates with 50% error <104.8 km and 90% error <199.8 km.
• Assimilating high-resolution coastline data reduced invalid on-land locations by nearly an order of magnitude.

Conclusions:

• The developed model offers an intuitive, flexible, and efficient alternative to existing animal tracking data processing methods.
• It significantly improves location accuracy, even with low-quality data from challenging species like elephant seals.
• The model's ability to assimilate diverse data sources and address issues like obstacle avoidance suggests broad applicability in ecological studies.