PARENT SESSION
Contributed Oral Session 20: Modeling Animal Populations
Monday, August 8, 1:30 PM - 5:00 PM, Meeting Room 513 C, Level 5, Palais des congrès de Montréal

The persistence of memory: Modeling the movements of foragers using memory-biased random walks.

Dalziel, Benjamin^*,1, Fryxell, John¹, ¹ University of Guelph, Guelph, Ontario, Canada

ABSTRACT- At the landscape scale, the distribution of foragers should be correlated with the distribution of their food. However, the individual movement decisions that allow this correlation must be made in the face of great uncertainty. This is in part because the perceptual range of an animal usually encompasses only a minuscule portion of the landscape over which its resources are arranged. There is considerable evidence that memory allows animals to revisit distant resource patches but the implications of this to broad-scale movement patterns remain ambiguous. We analyzed the trajectories of four elk (Cervus elaphus) fitted with GPS tracking collars and released into rural Central Ontario as part of a province-wide reintroduction program. Previous work suggests that elk exhibit a bi-phasic movement pattern, alternating between an encamped state, characterized by low rates of daily movement and high incidences of trajectory reversal, and an exploratory state, characterized by more directed trajectories traversed at a higher rate. Estimates of the daily magnitude and direction of displacement were used to parameterize a memory-biased random walk model. At each time step, the direction of movement was chosen from a composite distribution incorporating a theoretical force generated by each previously visited point in the landscape, proportional to the probability that the animal would attempt to return to that point. The contribution of each point to the composite was weighted by the time since the point was occupied, the distance between the point and the current location and the estimated resource value of the point. The model was parameterized using Markov Chain Monte Carlo techniques via the computer software WinBUGS. Model comparison using the Deviance Information Criterion (DIC) revealed that the memory model was at least as parsimonious for each of the four elk as a null model composed of a combination of correlated random walks. Comparison of the predictions of the memory model with the data suggests that elk may switch to an exploratory state as result of homogeneous angular distributions of information density, so that exploratory behaviour may serve to prevent elk from becoming trapped in areas of the landscape that constitute local maxima in resource value.

Key words: memory, movement models, Cervus elaphus, Bayesian