Incorporating uncertainty is essential to macroevolutionary inferences: Grass, grit, and the evolution of kangaroos [article]

Ian G Brennan
2019 bioRxiv   pre-print
Studying organismal ecology and evolution on deep timescales provides us opportunities to identify the processes driving patterns in diversity and forms. Macroecological and macroevolutionary studies of trait evolution however, often fail to account for sources of artifactual variation in the data---be it phylogenetic, temporal, or other. In some instances, this may not affect our evolutionary understanding, and accounting for sources of uncertainty may only subdue confidence in our inferences.
more » ... In more dramatic cases, narrow views of trait uncertainty may result in conclusions that are misleading. Because macroevolutionary analyses are built atop a number of preconceived hypotheses regarding the relationships between taxa, origination and divergence times, intraspecific variation, and environmental variables, it is important to incorporate and present this uncertainty. Here I use a dataset for Australian kangaroos to demonstrate the importance of incorporating uncertainty when testing patterns of diversification. After accounting for fossil age uncertainty, I provide evidence that a proposed Pliocene origin of Macropus kangaroos is at odds with combined evidence molecular and morphological dating methods. Depending on the estimated crown age of kangaroos, the evolution of hypsodonty is as likely caused by the continental expansion of C4 grasses as it is by increasing windborne dust levels or paleotemperature fluctuations. These results suggest that previous interpretations of the radiation of modern kangaroos are not as bulletproof as we believe, and that multiple factors have likely influenced their remarkable diversification across the Australian continent. More broadly, this demonstrates the importance of incorporating uncertainty in comparative ecological and evolutionary studies, and the value in testing the assumptions inherent in our data and the methods we employ.
doi:10.1101/772558 fatcat:mda33k34gbgtrbe4b4jvdk52xy