Applying mixture models to derive activity states of large herbivores from movement rates obtained using GPS telemetry

Abstract

Context To interpret spatial utilisation distributions, there is a need to translate animal locations obtained from global positioning system (GPS) telemetry into the activities performed and, hence, benefits derived, from particular places and times of day. Derived activity patterns also reveal how animals cope in changing environmental conditions. Aim The aim of our research was to develop and test an objective, consistent and biologically faithful method for deriving activity states from movement rates between successive GPS locations. Methods The method entails fitting mixtures of component statistical distributions to the frequency distribution of hourly step displacements. Breakpoints indicating transitions between predominant movement modes were identified by fitting exponential segments. Breakpoints were incorporated as off-sets for gamma distributions, but not needed for log-normal distributions. This procedure was applied to movement data for three large grazing ungulates. Key results Models consistently distinguished four movement modes interpreted as representing resting, foraging, mixed movement and travelling activity. Breakpoints and parameter estimates were consistent among seasons and herds of each ungulate species. The exponential-segment model and both mixture models closely represented observed daily activity patterns. However, some adjustment of the derived time budgets was needed to be consistent with observations. Key conclusions Mixture models provide an objective, reliable and biologically meaningful procedure for assessing seasonal, annual and spatial variation in the activity patterns of large ungulates from GPS data. Implications The method can potentially be applied to other mobile foragers large enough to carry GPS collars.