Agent-based movement models and landscape connectivity

Abstract

Human-caused changes in landscapes typically result in the loss, degradation, and fragmentation of animal habitats. One consequence of habitat fragmentation is changes in functional landscape connectivity, which, for animals, refers to their ability to move through a landscape among areas of suitable habitat. Our ability to anticipate the consequences of human-caused landscape change on connectivity depends in part on how well we are able to incorporate both animal movement behavior and landscape structure into predictive models for connectivity. In my dissertation research, I have explored various approaches to modeling animal movement and how to use such models to evaluate functional connectivity. Four chapters are presented. The first chapter presents a simple model for studying animal movement response to a single type of landscape feature. We demonstrate the model using data from a red diamond rattlesnake. The second chapter describes an approach for modeling movement and using individual-based movement models to evaluate functional connectivity. Model formulation, computer implementation, and application to connectivity evaluation are described and illustrated using a case study for puma in southern California. The third chapter presents a distance-weighted anisotropic detector model for perception that can be used in agent-based movement modeling. Modeling results suggest that increased anisotropy in the detection space leads to increased directional persistence and decreased use of the most suitable transition habitat. The fourth chapter describes a general approach to agent-based movement modeling and illustrates how to parameterize and evaluate these models using radio-tracking data. Forty-one models for three different puma were fit to data, and model selection was performed using AICc. The best models produced patterns that were consistent with observed data at the move level, but the patterns of nightly net displacement predicted by the best models were not consistent with the observed patterns at this scale; however, this nightly net displacement pattern may be produced by extending the current models to use spatial and temporal covariates. This work improves our ability to model individual-based movement and use such models to study functional landscape connectivity.