ON THE IMPORTANCE OF INCOMPLETE SAMPLING IN PALAEONTOLOGICAL AND PHYLOGENETIC APPROACHES TO ESTIMATING SPECIATION AND EXTINCTION RATES

Estimating speciation and extinction rates is essential for understanding the past, present and future of biodiversity, but is challenging given the incomplete and non-uniform nature of the rock and fossil records. Critical interest in this topic has led to a divergent suite of independent methods – paleontological estimates are based on sampled stratigraphic occurrences, while phylogenetic estimates are based on the observed branching times in a given phylogeny, typically derived using the genes of living species. The fossilized birth-death (FBD) process is a model that explicitly recognizes that the branching events in a phylogenetic tree and the fossils sampled from the rock record were generated by the same underlying diversification process. A crucial advantage of this modeling framework is that it explicitly incorporates the possibility that some species may never be sampled. Here, we present a new extension of the FBD model that allows diversification rates to be estimated from stratigraphic range data, in the absence of any phylogenetic character data. This eliminates entirely the requirement for anything about the underlying phylogeny to be known, such that the method can be applied when only occurrence data is available, but still allows the inherent, incomplete phylogenetic structure of the data to be considered. We tested this new method using simulations, and here we focus on the impact of non-uniform preservation and taxonomic uncertainty. We compared this method to fossil-based alternatives, including the widely implemented boundary-crosser method, and to a phylogenetic model that does not incorporate incomplete species sampling. We show that our new method is both accurate and precise, but that the performance of all approaches is sensitive to the underlying model of preservation. Furthermore, we demonstrate that results will be sensitive to taxonomy, specifically the way in which species are defined with respect to the true underlying branching process. Finally, we demonstrate that our approach outperforms other methods, especially when fossil data is sparse, by explicitly incorporating incomplete species sampling.