The missing link in biogeographic reconstruction: Accounting for lineage extinction rewrites history

Aim <p>In the most widely used family of methods for ancestral range estimation (ARE), dispersal, speciation and extirpation events are estimated from information on extant lineages. However, this approach fails to consider the geographic distribution of extinct species and their position on the phylogenetic tree, an omission that could compromise reconstruction. Here, we present a method that models the geographic distribution of extinct species and we quantify the potential inaccuracy in ancestral range estimation when extinction rates are above zero.</p> Location <p>Global applications, with an example from the Americas.</p> Taxon <p>All taxa, with an example from hummingbirds (<em>Amazilia</em>).</p> Methods <p>Methods capable of explicitly modelling extinct branches along with their reconstructed geographic information (GeoSSE) have been overlooked in ARE analysis, perhaps due to the inherent complexity of implementation. We develop a user-friendly platform, which we term LEMAD (Lineage Extinction Model of Ancestral Distribution) that generalizes the likelihood described in GeoSSE for any number of areas and under several sets of geographic assumptions. We compare LEMAD and extinction-free approaches using extensive simulations under different macroevolutionary scenarios. We apply our method to revisit the historical biogeography of <em>Amazilia</em> hummingbirds.</p> Results <p>We find that accounting for the lineages removed from a tree by extinction improves reconstructions of ancestral distributions, especially when rates of vicariant speciation are higher than rates of in situ speciation, and when rates of extinction and range evolution are high. Rates of in situ and vicariant speciation are accurately estimated by LEMAD in all scenarios. North America as the most likely region for the common ancestor of hummingbirds.</p> Main conclusions <p>Methods that neglect lineage extinction are less likely to accurately reconstruct true biogeographic histories of extant clades. Our findings on an empirical dataset reconcile the Eurasian origin of <em>Amazilia</em> with biogeographic reconstructions when lineage extinction is considered.</p>