INFERRING TRUE NUMBER OF DINOSAUR SPECIES BY ESTIMATING FOSSIL SAMPLING PROBABILITIES

One of the prominent goals of paleobiology is to infer the dynamics of species diversity over time and space, while recognizing that collections of fossil specimens yield a potentially incomplete picture of past biodiversity. In the last decades, several approaches have been developed to account for the bias(es) of the fossil record while estimating past diversity; subsampling techniques and proxy-modeling approaches are most frequently used to (hopefully) extract reliable signals of the rise and falls of diversity by positing comparative diversity metrics across geological bins. While these are valuable for estimating the relative dynamics of species diversity over time, they suffer from two serious drawbacks. Firstly, they can only yield relative species richness, i.e. the absolute number of species cannot be properly inferred. Secondly, they both suffer from a range of conceptual and computational difficulties and while the variables presented might be indicative of changes in diversity, they are not always easily interpreted. Here we present a simple scheme where we see the whole process from fossilization to detection as a Poisson process and use observation data to estimate sampling rates for different taxonomic groups across different geological stages. By using such estimated sampling rates we can predict past biodiversity in terms of true species richness, a major improvement to other methods that only yield relative diversities over time. We apply our approach to the three major groups of dinosaurs and compare our findings with earlier approaches and highlight potential future improvements and applications.