Landscape connectivity models: comparing landscape-centric versus movement-informed approaches

In actively dispersing organisms, movement is a key component of the dispersal process, and landscape structure can affect the movement path of dispersing individuals. Functional connectivity is an emergent property of the interactions between the dispersal abilities, and the structural connectivity that facilitates or impedes such movement. The direct study of dispersal movement remains challenging. Hence numerous indirect techniques have been developed to model putative dispersal paths among patches based on the fundamental concept of minimizing the cumulated resistance along the path across the landscape between the start and the end points.
We cannot expect that an animal dispersing over large distances has knowledge of the endpoint of its dispersal path, and we hypothesize that actual dispersal movement is driven by local habitat quality in a manner congruent with habitat cueing (natal-habitat biased dispersal), independent of the final settling location. In the context of dispersal, we predict that individual-based models informed with empirical movement data more realistically assess functional connectivity than landscape-centric models that ignore the movement process.
We compared three modeling approaches to assess connectivity between suitable habitat patches based on empirical data. We used the little owl Athene noctua as our case study, and modeled connectivity between a source population in southwest Germany to suitable patches in northwest Switzerland using least-cost path models (LCP), circuitscape models (CS), and individual-based models (IBM). We developed a habitat suitability model based on telemetry data of adult, residents owls, and used the inverse as our resistance-landscape assuming more resistance to movement for LCP and CS modelling where habitat suitability is poor. We empirically parameterized IBM’s based on dispersing owl movements and the resistance landscape, to simulate thousands dispersal paths from the source population. We measured the time (number of generations) and the number of individuals which would successfully reach the targeted destination patch. We also measured the deviation of the simulated paths from the ideal LCP and CS corridors, and discuss the conservation implication of using one or the other approach.