An evaluation of alternative approaches for simulating animal movement in spatially-explicit individual-based models

• Main Author: Watkins, Katherine Shepard
• Other Authors: Twilley, Robert
• Format: Others
• Language: en
• Published: LSU 2012
• Subjects: Oceanography & Coastal Sciences
• Online Access: http://etd.lsu.edu/docs/available/etd-06052012-163626/

Simulating animal movement in spatially-explicit individual-based models (IBMs) is both challenging and critically important to accurately estimating population dynamics. I compared four distinct movement approaches or sub-models (restricted-area search, kinesis, event-based, and run and tumble) in a series of simulation experiments. I used an IBM loosely based on a small pelagic fish that simulated growth, mortality, and movement of a cohort on a 2-dimensional grid. First, I tested the sub-models calibrated (i.e., trained) with a genetic algorithm in one set of environmental conditions in three other novel environments. The sub-models performed well, except restricted-area search and event-based that needed to be trained in environments with gradients similar to the test environment. Also, run and tumble only trained in steep habitat quality gradients. The sub-models were then trained and tested across a range of spatio-temporal resolutions (cell size and time step). The sub-models generally performed well across resolutions, but the sub-models did not perform equally well at all resolutions. Kinesis and run and tumble performed better at coarser resolutions, and restricted-area and event-based performed better at finer resolutions. I attributed the trends across resolution to differences in how the habitat quality individuals experienced changed at each time step. Finally, I trained and tested the sub-models in an IBM with dynamic prey and predator fields. I trained and tested the sub-models in dynamic and static versions of the environment. Sub-models trained in the dynamic environment performed well in both dynamic and static test environments; however, sub-models trained in static environment did not perform consistently well in dynamic test environment. Overall, restricted-area search, kinesis, and event-based were robust across the range of conditions in which I tested them, but run and tumble only performed well in environments with very steep habitat quality gradients. In selecting a movement sub-model, researchers should consider the assumptions of potential sub-models, the observed movement patterns of the species of interest, the shape and steepness of the underlying habitat quality gradient, and the spatio-temporal resolution of the model. Sub-models that will be applied in dynamic conditions should be calibrated in comparable dynamic conditions.