Model selection as a tool for phylogeographic inference: an example from the willow <i><scp>S</scp>alix melanopsis</i>

Phylogeographic inference has typically relied on analyses of data from one or a few genes to provide estimates of demography and population histories. While much has been learned from these studies, all phylogeographic analysis is conditioned on the data, and thus, inferences derived from data that represent a small sample of the genome are unavoidably tenuous. Here, we demonstrate one approach for moving beyond classic phylogeographic research. We use sequence capture probes and Illumina sequencing to generate data from >400 loci in order to infer the phylogeographic history of Salix melanopsis, a riparian willow with a disjunct distribution in coastal and the inland Pacific Northwest. We evaluate a priori phylogeographic hypotheses using coalescent models for parameter estimation, and the results support earlier findings that identified post-Pleistocene dispersal as the cause of the disjunction in S. melanopsis. We also conduct a series of model selection exercises using IMa2, Migrate-n and ∂a∂i. The resulting ranking of models indicates that refugial dynamics were complex, with multiple regions in the inland regions serving as the source for postglacial colonization. Our results demonstrate that new sources of data and new approaches to data analysis can rejuvenate phylogeographic research by allowing for the identification of complex models that enable researchers to both identify and estimate the most relevant parameters for a given system.