A novel semi-variance approach to extracting multiple movement modes from animal relocation data

Background/Question/Methods

Understanding the factors governing animal movement has long been a fundamental problem in ecology and conservation biology, and technological advances make it possible to explore this problem in ever more detail. Relocation data often consist of a complex mixture of different movement behaviors, and decomposing this mix into its component parts is a key challenge in movement ecology. Composite random walk models have been the main tools employed in analyses of multiple movement behaviors or “modes”. They can, however, be difficult to fit to data, are often parameter rich, and they require that the timescale(s) governing the movement process is (are) reasonably close to the data sampling rate. Here, we show how the semi-variance function (SVF) of a stochastic movement process offers both an alternative approach to identifying multiple movement modes, and a solution the sampling rate problem. We describe how a family of continuous-space, continuous-time stochastic movement models, representing a wide range of behaviors, can be expressed in terms of their SVFs. We then connect these SVFs to relocation data via variogram regression and compare them using standard model selection techniques.

Results/Conclusions

We illustrate our approach using relocation data from 28 Mongolian gazelles. The AIC-selected best model suggests that gazelle movement is characterized by a slow, ballistic foraging mode with a 10-hour timescale, a fast, diffusive patch-search mode with a 2-month timescale, and an asymptotic diffusion mode on longer timescales. The fitted movement model also allows us to calculate an average yearly individual home range size of 12,200 km².