Recursive multi‐frequency segmentation of movement trajectories (ReMuS)

Abstract

The quantity of GPS trajectories on the movement of individuals has far outstripped our ability to analyse them. Segmentation, a first step in trajectory analysis, is fundamental to imputing behaviour from patterns of movement. Behaviour occurs at multiple scales that are captured at different temporal frequencies. To identify behaviour from a complex trajectory it is essential to use a segmentation procedure that captures different frequency ranges. A limitation of existing segmentation algorithms is that they require training and parameterisation that can lead to special use segmentations that are not generalisable. This paper presents a new approach to the automated segmentation of movement trajectories at multiple scales. It has two distinguishing characteristics: it can capture movement patterns at multiple temporal frequencies through its recursive application; and requires no training or thresholding. The methodology is tested on two sets of simulated data and GPS tracking data of two species, the Turkey Vulture (Cathartes aura) and the Galapagos Albatross (Phoebastria irrorata). For the Turkey Vulture dataset, segments and transition points are evaluated using behavioural segmentations delineated by domain experts, and results are also compared with an existing approach. For the Albatross dataset, the segmentation results are evaluated using the geographic context of the movement tracks through visualisation. An Albatross trajectory is used to illustrate the application of recursive segmentation for extracting movement patterns at multiple temporal scales. The evaluation suggests that the segmentation technique is capable of extracting the right number of segments and in estimating transition points for most of the evaluated datasets. The proposed methodology has the potential to automate the segmentation process by enabling researchers to analyse a broad range of heterogeneous datasets without training. It will enable them to segment datasets recursively to derive multiple scales of movement patterns in a trajectory that may relate to different scales of behaviour.

A feature extraction based trajectory segmentation approach based on multiple movement parameters

“…They evaluated the performance of SeqScan using the ground truth which was including a set of animal trajectories. Ahearn and dodge (2018) introduced a new approach to segment the movement trajectories at multiple scales. They stated that to behavior identification in a complex trajectory it is important to use a segmentation method that captures (takes) different frequency ranges.…”

Ecological metrics and methods for GPS movement data

“…Path or trajectory segmentation methods are one area of research common to both GIScience and movement ecology, with the GIS community supporting significant research on pattern-oriented, cross-scale, and cross-type segmentation methods (Dodge et al 2009, Ahearn and Dodge 2018). In a study designed to explore the role of uncertainty in trajectory and segmentation analyses, Laube and Purves (2011) fitted 10 cows with GPS collars taking sub-second fixes to investigate questions of scale, granularity, and uncertainty when working with GPS data to assess movement parameters.…”

A Data Science Framework for Movement

“…quad-tree, KD-tree, R-tree) are established for fast data access and efficient retrieval and query of trajectory data (Pfoser, 2002). Trajectory segmentation is the process of decomposing a long and convoluted trajectory into parts of similar characteristics (Ahearn and Dodge, 2018). It is used to simplify trajectory data for further analyses and pattern detection.…”

“…Finally, data fusion and knowledge discovery should be informed and guided by the scale of observations, especially when integrating heterogeneous movement and context data sets from multiple sources. This is essential in movement data science in order to make relevant inferences from observed patterns at proper granularity and frequency levels in space and time (Ahearn and Dodge, 2018).…”