Inferences about the evolutionary history of geographically structured populations: The intersection of population genetics, biogeography and systematics

A central goal of evolutionary biology is to understand the evolution of biological diversity in terms of population genetic processes. Since the youngest portions of lineages are characterised by geographical replacement of populations, this research, which lies at the intersection of population genetics, biogeography and systematics, involves using genetic and phenotypic data to make statistical inferences about the evolutionary history of geographically structured populations. I start by discussing this research program in Amazonian forest birds, and presenting new data on geographical genetic and phenotypic variation. These data are typical of those for many organisms, in that the Amazonian lowlands appear to be occupied by a spatial mosaic of geographically representative populations which in some regions appear to be reproductively isolated by intrinsic or extrinsic barriers to gene flow, or both. A common approach in population genetics is to assume that the data have been sampled from a stochastic process at stationarity and then to query the relative rates at which processes such as coalescence, mutation, recombination and dispersal are occurring. However, understanding recent diversification places greater emphasis on non-equilibrium models, and chapters 2 and 3 explore some issues in statistical inference for non-equilibrium models of the history of structured populations. In chapter 2 I investigate inferences based on completely linked data such as those from the mitochondrial genome and find that estimates of rates of gene flow are upwardly biased and that there is little power to distinguish a recent barrier to gene flow from the alternative of ongoing equilibrium gene flow. An outstanding problem is the lack of statistical methods for studying recombining data such as nearby SNPs and sequences from the nuclear genome. In chapter 3 I develop an approximate likelihood-based approach to inference from these types of data under a model in which a panmictic ancestral population has split into two panmictic daughter populations.