A Likelihood Framework for Estimating Phylogeographic History on a Continuous Landscape

Due to lack of an adequate statistical framework, biologists studying phylogeography are abandoning traditional methods of estimating phylogeographic history in favor of statistical methods designed to test a priori hypotheses. These new methods may, however, have limited descriptive utility. Here, we develop a new statistical framework that can be used to both test a priori hypotheses and estimate phylogeographic history of a gene (and the statistical confidence in that history) in the absence of such hypotheses. The statistical approach concentrates on estimation of geographic locations of the ancestors of a set of sampled organisms. Assuming a spatially explicit random walk model of migration, we derive an equation describing the likelihood of the geographic coordinates of the individuals represented by internal nodes on a tree (the parameters of interest) and the mean per-generation dispersal distance (which can be estimated as a nuisance parameter), given the geographic coordinates of the individuals represented by tips on the tree (topology and branch lengths are assumed to be known). Using a maximum likelihood approach, which is implemented in the new program PhyloMapper, we apply this statistical framework to a 246-taxon mitochondrial genealogy of North American chorus frogs, focusing in detail on one of these species. We demonstrate three lines of evidence for recent northward expansion of the mitochondrion of the coastal clade of Pseudacris feriarum: higher per-generation dispersal distance in the recently colonized region, a noncentral ancestral location, and directional migration. After illustrating one method of accommodating phylogenetic uncertainty, we conclude by discussing how extensions of this framework could function to incorporate a priori ecological and geological information into phylogeographic analyses.