From animal tracks to fine‐scale movement modes: a straightforward approach for identifying multiple spatial movement patterns

Abstract

Summary Thanks to developments in animal tracking technology, detailed data on the movement tracks of individual animals are now attainable for many species. However, straightforward methods to decompose individual tracks into high‐resolution, spatial modes are lacking but are essential to understand what an animal is doing. We developed an analytical approach that combines separately validated methods into a straightforward tool for converting animal GPS tracks into short‐range movement modes. Our three‐step analytical process comprises: (i) decomposing data into separate movement segments using behavioural change point analysis; (ii) defining candidate movement modes and translating them into nonlinear or linear equations between net squared displacement (NSD) and time and (iii) fitting each candidate equation to NSD segments and determining the best‐fitting modes using Concordance Criteria, Akaike's Information Criteria and other fine‐scale segment characteristics. We illustrate our approach for three sub‐adults, male wild boar Sus scrofa tracked at 15‐min intervals over 4 months using GPS collars. We defined five candidate movement modes based on previously published studies of short‐term movements: encamped, ranging, round trips (complete and partial) and wandering. Our approach successfully classified over 80% of the tracks into these movement modes lasting between 5 and 54 h and covering between 300 m to 20 km. Repeated analyses of GPS data resampled at different rates indicated that one positional fix every 3–4 h was sufficient for >70% classification success. Classified modes were consistent with published observations of wild boar movement, further validating our method. The proposed approach advances the status quo by permitting classification into multiple movement modes (where these are adequately discernable from spatial fixes) facilitating analyses at high temporal and spatial resolutions, and is straightforward, largely objective, and without restrictive assumptions, necessary parameterizations or visual interpretation. Thus, it should capture the complexity and variability of tracked animal movement mode for a variety of taxa across a wide range of spatial and temporal scales.

Mountain lions on the prairie: habitat selection by recolonizing mountain lions at the edge of their range

“…We used GPS locations and behavioral change‐point analysis to distinguish among movement states for individual mountain lions (Morelle et al ). To identify these states within each animal's movement dataset, we first calculated net‐squared displacement (NSD; Bunnefeld et al ), which is the linear distance between the initial location of a trajectory and all successive locations.…”

The spatial–temporal relationship of blue‐winged teal to domestic poultry: Movement state modelling of a highly mobile avian influenza host

“…Path segmentation techniques are routinely applied in movement ecology to help identify discrete behavioural states or canonical activity modes from animal telemetry (Getz & Saltz, 2008;Seidel et al, 2018). A few examples of segmentation techniques include k-means clustering (Van Moorter et al, 2010), change point analysis (Gurarie et al, 2009) or a combination of different segmentation methods (Morelle et al, 2017;Zhang et al, 2015). To classify initial movement states for training our spatial temporal model, we constructed hidden Markov models utilizing the moveHmm package (Michelot et al, 2016).…”

Using Piecewise Regression to Identify Biological Phenomena in Biotelemetry Datasets

“…the pruned exact linear time (PELT) algorithm or the penalized contrasts method; Lavielle 2005; Killick et al, 2012) or parametric state-space models such as hidden Markov models (Patterson et al, 2008). Often, these methods require restrictive assumptions, are computationally intensive, or are restricted to a single type of data stream (Morelle et al, 2017). Here, we demonstrate how piecewise regression can be used to analyze a wide range of biotelemetry data sets, offering a flexible and user-friendly approach with results that are easy to interpret.…”