Using simulation modeling to demonstrate the performance of graph theory metrics and connectivity algorithms

Models and metrics to measure ecological connectivity are now well-developed and widely used in research and applications to mitigate the ecological impacts of climate change and anthropogenic habitat loss. Despite the prevalent application of connectivity models, however, relatively little is known about the performance of these methods in predicting functional connectivity patterns and organism movement. Our goal in this paper was to compare different connectivity models in their abilities to predict a wide range of simulated animal movement patterns. We used the Pathwalker software to evaluate the performance of several connectivity model predictions based on graph theory, resistant kernels, and factorial least-cost paths. In addition, we assessed the efficacy of synoptic and patch-based approaches to defining source points for analysis. In total, we produced 28 different simulations of animal movement. As we expected, we found that the choice of connectivity model used was the variable that most influenced prediction accuracy. Moreover, we found that the resistant kernels approach consistently provided the strongest correlations to the simulated underlying movement processes. The results also suggested that the agent-based simulation approach itself can often be the best analytical framework to map functional connectivity for ecological research and conservation applications, given its biological realism and flexibility to implement combinations of movement mechanism, dispersal threshold, directional bias, destination bias and spatial composition of source locations for analysis. In doing so, we provide novel insights to guide future functional connectivity analyses. In future research, we could use the same model for several different species groups and see how this reliability depends on the species analyzed. This could bring to light other elements that play an essential role in predicting connectivity.