Identifying stationary phases in multivariate time series for highlighting behavioural modes and home range...

Area of Science:

• Ecology
• Movement Ecology
• Bio-logging

Background:

• Biologging technology enables high-resolution tracking of animal movements.
• Animal movement data often exhibit piecewise stationary phases, reflecting different behaviors or home ranges.
• Identifying transitions between these phases is crucial for understanding movement ecology.

Purpose of the Study:

• Introduce segclust2d, a novel segmentation-clustering method for analyzing multivariate animal movement time series.
• Provide a user-friendly tool for identifying distinct movement phases and home range shifts.
• Compare the performance of segclust2d against existing complex methods.

Main Methods:

• Developed segclust2d, a segmentation-clustering algorithm for bivariate (or multivariate) time series.
• Applied the method to simulated and real animal movement data (zebra and elephant).
• Focused on time series of coordinates, speed, and turning angles to identify home ranges and behavioral modes.

Main Results:

• segclust2d accurately segments time series into stationary phases, corresponding to different movement behaviors or home ranges.
• Simulations show segclust2d rivals or outperforms existing methods, requiring fewer user-defined parameters.
• Demonstrated effectiveness on real data, identifying small-scale behavioral modes and large-scale home range shifts.

Conclusions:

• segclust2d is an effective and user-friendly tool for analyzing animal movement data.
• The method facilitates the study of animal behavior, home range dynamics, and movement ecology.
• It offers a robust alternative to more complex modeling approaches for movement data segmentation.