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

1. Recent advances in bio-logging open promising perspectives in the study
of animal movements at numerous scales. It is now possible to record
time-series of animal locations and ancillary data (e.g. activity level
derived from on-board accelerometers) over extended areas and long
durations with a high spatial and temporal resolution. Such time-series
are often piecewise stationary, as the animal may alternate between
different stationary phases (i.e. characterised by a specific mean and
variance of some key parameter for limited periods). Identifying when
these phases start and end is a critical first step to understand the
dynamics of the underlying movement processes. 2. We introduce a new
segmentation-clustering method we called segclust2d (available as a R
package at cran.r-project.org/package=segclust2d). It can segment bi- (or
more generally multi-) variate time-series and possibly cluster the
various segments obtained, corresponding to different phases assumed to be
stationary. This method is easy to use, as it only requires specifying a
minimum segment length (to prevent over-segmentation), based on biological
rather than statistical considerations. 3. This method can be applied to
bivariate piecewise time-series of any nature. We focus here on two types
of time-series related to animal movement, corresponding to (i) at large
scale, series of bivariate coordinates of relocations, to highlight
temporary home ranges, and (ii) at smaller scale, bivariate series derived
from relocations data, such as speed and turning angle, to highlight
different behavioural modes such as transit, feeding and resting. 4. Using
computer simulations, we show that segclust2d can rival and even
outperform previous, more complex methods, which were specifically
developed to highlight changes in movement modes or home range shifts
(based on Hidden Markov and Ornstein-Uhlenbeck modelling), which, contrary
to our method, usually require the user to provide relevant initial
guesses to be efficient. Furthermore we demonstrate it on actual examples
involving a zebra's small scale movements and an elephant's
large scale movements, to illustrate how various movement modes and home
range shifts, respectively, can be identified. 15-Aug-2019