How well can we estimate diversity dynamics for clades in diversity decline?

The fossil record shows that the vast majority of all species that ever existed are extinct and that most lineages go through an expansion and decline in diversity. However, macroevolutionary analyses based upon molecular phylogenies have difficulty inferring extinction dynamics, raising questions about whether the neontological record can contribute to an understanding of the decline phenomenon. Two recently developed diversification methods for molecular phylogenies (RPANDA and BAMM) incorporate models that theoretically have the capacity to capture decline dynamics by allowing extinction to be higher than speciation. However, the performance of these frameworks over a wide range of decline scenarios has not been studied. Here we investigate the behavior of these methods under decline scenarios caused by decreasing speciation and increasing extinction through time on simulated trees at fixed intervals over diversity trajectories with expansion and decline phases. We also compared method performance over a comprehensive dataset of 214 empirical trees. Our results show that both methods perform equally well when varying speciation rates control decline. When decline was only caused by an increase in extinction rates both methods wrongly assign the variation in net diversification to a drop in speciation, even though the positive gamma values of those trees would suggest otherwise. We also found a tendency for RPANDA to favor increasing extinction and BAMM to favor decreasing speciation as the most common cause of decline in empirical trees. Overall our results shed light on the limitations of both methods, encouraging researchers to carefully interpret the results from diversification studies.