Trait-dependent dispersal in rails (Aves: Rallidae): Historical biogeography of a cosmopolitan bird clade.

Abstract The ability of lineages to disperse over evolutionary timescales may be influenced by the gain or loss of traits after adaptation to new ecological conditions. For example, rails (Aves: Rallidae) have many cases of flightless insular endemic species that presumably evolved after flying ancestors dispersed over large ocean barriers and became isolated. Nonetheless, the details of how flying and its loss have influenced the clade’s historical biogeography are unknown, as is the importance of other predictors of dispersal such as the geographic distance between regions. Here, we used a dated phylogeny of 158 species of rails to compare trait-dependent and trait-independent biogeography models in BioGeoBEARS. We evaluated a probabilistic historical biogeographical model that allows geographic range and flight to co-evolve and influence dispersal ability on a phylogeny. The best-fitting dispersal model was a trait-dependent dispersal model (DEC+j +x+t21+m1) that accrued 85.2% of the corrected Akaike Information Criterion (AICc) model weight. The distance-dependence parameter, x was estimated at -0.54, ranging from -0.49 to -0.65 across models, suggesting that a doubling of dispersal distance results in an approximately 31% decrease in dispersal rate (2-0.54 = 0.69). The estimated rate of loss of flight (t21) was similar across all models (∼0.029 loss events per lineage per million years). The multiplier on dispersal rate when a lineage is non-flying, m1, is estimated to be 0.38 under this model. Surprisingly, the estimate of m1 was not 0.0, probably because the loss of flight is so common in the rails that entire clades of flightless species are found in the data, forcing the model to attribute some dispersal to flightless lineages. These results indicate that long-distance dispersal over macroevolutionary timespans can be modelled, rather than simply attributed to chance, allowing support for different hypotheses to be quantified and model limitations to be identified. Overall, by combining new analytical methods with a comprehensive phylogenomic dataset, we use a quantitative framework to show how traits influence dispersal capacity and eventually shape geographical distributions at a macroevolutionary scale.