Joint estimation over multiple individuals improves behavioural state inference from animal movement data.

Area of Science:

• Ecology
• Movement ecology
• Quantitative biology

Background:

• State-space models (SSMs) are crucial for inferring animal movement behaviors across populations.
• Scaling SSM inference requires methods that effectively integrate data from multiple individuals.
• Accurate behavioral state estimation is fundamental to understanding ecological processes.

Purpose of the Study:

• To evaluate a joint estimation approach for state-space models in animal movement analysis.
• To compare the performance of joint versus nonhierarchical estimation formulations.
• To assess the impact of data quality, specifically location error, on behavioral state inference.

Main Methods:

• Simulated movement paths with known behavioral states were used to compare estimation error.
• State-space models were formulated using both nonhierarchical and joint estimation approaches.
• The models were applied to Argos telemetry data from Weddell seals.

Main Results:

• Joint estimation significantly reduced behavioral state estimation error, particularly when movement patterns between states were dissimilar.
• The joint model outperformed the nonhierarchical model with simulated data exhibiting heavy-tailed location errors.
• Analysis of Weddell seal data revealed substantially reduced uncertainty in state switching probabilities and improved resolution of behavioral state sequences with joint estimation.

Conclusions:

• Joint estimation in state-space models offers superior performance for inferring animal movement behaviors.
• This approach is particularly advantageous when dealing with noisy or error-prone location data.
• Hierarchical or joint estimation models are recommended for robust behavioral state analysis in animal telemetry studies.