Landscape genetics aims to understand how genetic heterogeneity arises based on connectivity among landscape elements and a spatial mosaic of environmental and ecological influences. We present a novel approach that uses genetic data for improved inference of landscape connectivity. The approach is centered on new metrics we devised to summarize neutral and adaptive genetic variation. We used an amalgam of spatial and Bayesian statistics techniques to model the spatial distribution of gene flow. We borrowed techniques from spatial statistics and used multivariate ordination to obtain environmental factors that were used to model genetic variation. We used recent advances in Bayesian modeling to perform spatial interpolation of genetic variation. Specifically, for Bayesian inference in latent Gaussian modeling, we used integrated nested Laplace approximation, a computationally faster alternative to Markov chain Monte Carlo.
Results/Conclusions
To test our approach, we used simulations and empirical data. We modeled neutral and adaptive genetic variation and interpolated 1-km resolution genetic variation maps. For the empirical data, we used two different systems (a structural pest and a vector of sleeping sickness): the eastern subterranean termite (Reticulitermes flavipes) in the Appalachian Mountains, and tsetse (Glossina fuscipes fuscipes) in Uganda. Seasonal fluctuations in precipitation and temperature had an effect on both systems. We identified loci under selection and used these to produce adaptive (in addition to neutral) genetic variation maps. Our approach provided predictions of strength and directionality of gene flow and can be used to identify areas of priority for pest management and vector control. Similarly, this approach could provide climate-change-dependent genetic forecasting capability in a diversity of fields, including conservation and invasion biology.