Premium
Characterizing change points and continuous transitions in movement behaviours using wavelet decomposition
Author(s) -
Soleymani Ali,
Pennekamp Frank,
Dodge Somayeh,
Weibel Robert
Publication year - 2017
Publication title -
methods in ecology and evolution
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.425
H-Index - 105
ISSN - 2041-210X
DOI - 10.1111/2041-210x.12755
Subject(s) - segmentation , pattern recognition (psychology) , computer science , temporal scales , scale (ratio) , spatial ecology , movement (music) , artificial intelligence , discrete wavelet transform , wavelet , wavelet transform , tracking (education) , biological system , ecology , geography , cartography , physics , biology , acoustics , psychology , pedagogy
Summary Individual behaviour, that is, the reaction of an organism to internal state, conspecifics and individuals of other species as well as the environment, is a crucial building block of their ecology. Modern tracking techniques produce high‐frequency observations of spatial positions of animals and accompanying speed and tortuosity measurements. However, inferring behavioural modes from movement trajectories remains a challenge. Changes in behavioural modes occur at different temporal and spatial scales and may take two forms: abrupt, representing distinct change points; or continuous, representing smooth transitions between movement modes. The multi‐scale nature of these behavioural changes necessitates development of methods that can pinpoint behavioural states across spatial and temporal scales. We propose a novel segmentation method based on the discrete wavelet transform (DWT), where the movement signal is decomposed into low‐frequency approximation and high‐frequency detail sub‐bands to screen for behavioural changes at multiple scales. Approximation sub‐bands characterizes broad changes by taking the continuous variations between behavioural modes into account, whereas detail sub‐bands are employed to detect abrupt, finer scale change points. We tested the ability of our method to identify behavioural modes in simulated trajectories by comparing it to three state‐of‐the‐art methods from the literature. We further validated the method using an annotated dataset of turkey vultures ( Cathartes aura ) relating extracted segments to the expert knowledge of migratory vs. non‐migratory patterns. Our results show that the proposed DWT segmentation is more versatile than other segmentation methods, as it can be applied to different movement parameters, performs better or equally well on the simulated data, and correctly identifies behavioural modes identified by the experts. It is hence a valuable addition to the toolbox of land managers and conservation practitioners to understand the behavioural patterns expressed by animals in natural and human‐dominated landscapes.