Modeling connectivity in landscape genetics: applications, optimization and assessing uncertainty

Connectivity modeling and corridor identification are an essential part of landscape genetics and important tools for the future of conservation biology. The previous decade has shown a steadily increasing interest and rise in publications in landscape genetics. This enthusiasm has led to advances in the methods and theoretical background of the field; however, there remain important, yet unresolved, challenges. Many of these are related to validation and uncertainty testing for resistance surfaces (hypotheses of connectivity). These fundamental issues need to be addressed before landscape genetics can gain the full recognition of a scientific discipline such as population genetics or landscape ecology. The results herein not only describe the application of traditional landscape genetic techniques to empirical data, but also explore two new major approaches to improving connectivity modeling and corridor identification. In the first new approach, general theory is advanced using resistant kernel modeling by assessing a wide range of potential resistance surfaces to broadly model species distribution, connectivity, and response to habitat fragmentation and loss. Resistant kernel models allow generality across several species based on abiotic (human footprint) and life-history traits (dispersal ability and population size) for the entire Western United States. The second approach is to introduce a genetic algorithm for optimizing the process of resistance map fitting to empirical data. Optimization has three benefits. The first is removing the potential bias of expert opinion. The second is making possible multimethod evaluations of model uncertainty using different statistical tests, genetic distance metrics, and connectivity models. Lastly, optimization allows one to compare a large number of models enabling sensitivity analysis testing (e.g. leave-one-out populations, loci, or individuals). Together optimization and sensitivity analysis provide better, and more consistent, identification of landscape corridors and illustrate where models fail due to sensitivity to noisy genetic data. Described herein is a more rigorous framework of resistance map fitting and testing to help alleviate drawing faulty inferences in landscape genetic studies