Divergence, gene flow, and the speciation continuum in trans-Beringian birds

Abstract

Understanding the processes of divergence and speciation, particularly in the presence of gene flow, is key to understanding the generation of biodiversity. I investigated divergence and gene flow in nine lineages of birds with a trans-Beringian distribution, including pairs of populations, subspecies, and species, using loci containing ultraconserved elements (UCEs). I found that although these lineages spanned conditions from panmixia to fully biologically isolated species, they were not smoothly distributed across this continuum, but formed two discontinuous groups: relatively shallow splits with gene flow between Asian and North American populations, no fixed SNPs, and lower divergence; and relatively deeply split lineages with multiple fixed SNPs, higher divergence, and relatively low rates of gene flow. All eight lineages in which two populations were distinguishable shared the same divergence model, one with gene flow without a prolonged period of isolation. This was despite the diversity of lineages included that might not have responded in the same ways to the glacial-interglacial cycles of connection and isolation in Beringia. Together, these results highlight the role of gene flow in influencing divergence in these Beringian lineages. Sample size is a critical aspect of study design in population genomics research, yet few empirical studies have examined the impacts of small sample sizes. Using split-migration models optimized with full datasets, I subsampled the datasets from Chapter 1 at sequentially smaller sample sizes from full datasets of 6 - 8 diploid individuals per population and then compared parameter estimates and their variances. Effective population size parameters (ν) tended to be underestimated at low sample sizes (fewer than 3 diploid individuals per population), migration (m) was fairly reliably estimated until under 2 individuals per population, and no trend of over- or underestimation was found in either time since divergence (T) or Θ (4Nrefμ) . Lineages that were split above the population level (subspecies and species pairs) tended to have lower variance at smaller sample sizes than population-level splits, with many parameters reliably estimated at levels as low as 3 diploid individuals per population, whereas shallower splits (i.e., populations) often required at least 5 individuals per population for reliable demographic inferences. Although divergence levels may be unknown at the outset of study design, my results provide a framework for planning appropriate sampling, and for interpreting results if smaller sample sizes must be used.