Estimating clade-specific diversification rates and palaeodiversity dynamics from reconstructed phylogenies

Understanding palaeodiversity dynamics through time and space is a central goal of macroevolution. Estimating palaeodiversity dynamics has been historically addressed with fossil data because it directly reflects the past variations of biodiversity. Unfortunately, some groups or regions lack a good fossil record, and dated phylogenies can be useful to estimate diversification dynamics. Recent methodological developments have unlocked the possibility to investigate palaeodiversity dynamics by using phylogenetic birth-death models with non-homogeneous rates through time and across clades. One of them seems particularly promising to detect clades whose diversity has declined through time. However, empirical applications of the method have been hampered by the lack of a robust, accessible implementation of the whole procedure, therefore requiring users to conduct all the steps of the analysis by hand in a time-consuming and error-prone way. Here we propose an automation of Morlon et al. (2011) clade-shift model with additional features accounting for recent developments and we implement it in the R package RPANDA. We also test the approach with simulations focusing on its ability to detect shifts of diversification and to infer palaeodiversity dynamics. Finally, we illustrate the automation by investigating the palaeodiversity dynamics of Cetacea, Vangidae, Parnassiinae, and Cycadales. Simulations showed that we accurately detected shifts of diversification although false shift detections were higher for time-dependent diversification models with extinction. The median global error of palaeodiversity dynamics estimated with the automated model is low showing that the method can capture diversity declines. We detected shifts of diversification for three of the four empirical examples considered (Cetacea, Parnassiinae and Cycadales). Our analyses unveil a waxing-and-waning pattern due to a phase of negative net diversification rate embedded in the trees after isolating recent radiations. Our work makes possible to easily apply non-homogeneous models of diversification in which rates can vary through time and across clades to reconstruct palaeodiversity dynamics. By doing so, we detected palaeodiversity declines among three of the four groups tested, highlighting that such periods of negative net diversification might be common. We discuss the extent to which this approach might provide reliable estimates of extinction rates and we provide guidelines for users.