COS 95-6
Characterizing movement strategies of Galapagos giant tortoises using a Bayesian mixture distribution model and net squared displacement

Thursday, August 13, 2015: 9:50 AM
301, Baltimore Convention Center
Guillaume Bastille-Rousseau, College of Environmental and Forest Biology, State University of New York College of Environmental Science and Forestry, Syracuse, NY
Jonathan Potts, Sheffield University, Sheffield, England
Charles B. Yackulic, Southwest Biological Science Center, US Geological Survey, Flagstaff, AZ
Jacqueline L. Frair, Department of Environmental and Forest Biology, State University of New York College of Environmental Science and Forestry, Syracuse, NY
Hance Ellington, Trent University, ON, Canada
Stephen Blake, Max Plank Institute for Ornithology, Radolfzell, Germany

Movement is a temporally and spatially structured process that links animals with their environment. An important goal of movement ecology is to understand the causes and consequences of movement. Often this requires examining how different stereotypical movement strategies (e.g. migration, residency, nomadism) differ in terms of costs and benefits. Doing so requires objective, yet flexible, ways of characterizing movement strategies. Recent attempts to classify movement strategies have used a suite of models to fit the change in the net squared displacement (NSD) through time. We propose the use of mixture distribution models, analogous to a clustering approach, to characterize movement into homogenous bouts. Transition frequencies among bouts can then be used as an indication of movement strategy. Starting from a Gaussian mixture model, we extend the approach into a Bayesian framework that then allows flexible parametrization of different distributions and the integration of a matrix of switching transition probability among clusters. Our approach also allows the inclusion of other movement related variables and multiple years of movement data. We illustrate our approach by fitting it to simulated movement data and to giant tortoises (Chelonoidis spp.) GPS data inhabiting three of the Galapagos Islands. 


Using a set of simple rules, the mixture distribution model performs equivalently or better than previous approaches at identifying each simulated movement strategy. The mixture distribution model allows effective characterization of different movement strategies in giant tortoises. It also opens the door to characterization of intermediate strategies such as exploratory resident movement strategies among years. These were consistent for each individual, with few individuals transitioning from one strategy to another. Consistency in timing of migration was however more variable. Moreover, the predominance of each strategy differed among islands. These results indicate an important influence of environmental factors on movement patterns of giant tortoises. Overall, our approach offers a flexible characterization of movement strategies that will improve our understanding of drivers of animal movement.