Demographic history and isolation with migration model caveats
Estimating population sizes and timing demographic events depends on accurate estimates of mutation rate per generation. As such, an important caveat to the demographic and isolation with migration results relates to assumptions about generation times and mutation rates in these bird species. Generation times are reliable when information about age of sexual maturity, adult survival, and offspring survival is known in a species. In tropical bird species, little of this information is known, and indeed none of the focal species studied here have known generation times. Additionally, reproductive effort is known to vary among temperate and tropical species (Hau et al., 2010), and even along elevational gradients in tropical regions (Boyce et al., 2015).
The mutation rates we used here were calculated from time-calibrated avian phylogenies, and implicitly rely on accurate timing of nodes from the estimated phylogeny (Oliveros et al., 2019). These mutation rate estimates are in numbers of mutations per year (Table 1) and are generally consistent in magnitude with other Passeriformes mutation rate estimates (Nam et al., 2010). As such, in the context of estimating mutation rate per generation, we likely have more certainty in our absolute mutation rates (per year) and less certainty in mutation rate per generation because of unknown generation times. Regardless, we expect that our two-year generation time assumptions are within a reasonable amount of error from real values, e.g., we don’t expect these values to be off by an order of magnitude.
Overall, given the uncertainties with both mutation rate and generation times, we should expect that any population sizes and timings estimated in demographic histories and isolation with migration models may vary. Absolute timing events will be inaccurate proportional to error in our absolute mutation rate estimates (mutations per year). Because our population size estimates (i.e., 4Nμ) directly depend on mutation rates per generation, the error in population size estimates will directly scale with our error in estimating how the mutation rate relates to generation times. For example, if the true generation times of our focal species range from a year to four years (and we used an estimate of two years), estimated effective population sizes will be doubled or halved, respectively. Because we see a strong correlation between genomic diversity and harmonic mean of population sizes including all individuals across species (Fig. 3), even with potential error in estimating generation times, we can still interpret relative differences between species’ genomic diversities and population size trends.