The performance of ancestral state reconstruction versus the fossil record in recovering paleodiversity and paleogeography

Background/Question/Methods

A common motivation among ecologists and evolutionary biologists is measuring and explaining biological diversity. Increasingly, evolutionary biologists are turning to phylogenetic methods to address diversity-related questions. Meanwhile, paleontologists warn that methods using only modern data rely upon assumptions known through fossil data to be false. Unfortunately, the fossil record itself is highly incomplete for many groups, making empirical studies of paleodiversity and paleoecology difficult.

Paleontological researchers have observed that species geographic range sizes are heritable and likewise that some lineages have a higher propensity to speciate than others resulting in greater standing diversity. These tendencies often bear implications for the long-term survival of the lineages that possess them. Here, we use the current phylogenetic method of likelihood-based ancestral state reconstruction (ASR) on genera in the Rosales (Angiosperms: Eudicots) to investigate paleo-patterns of species richness and geographic range size. We then ground-truth our investigation against the known fossil record of these genera. The principal goal of this study is to evaluate the usefulness of ASR in measuring higher taxon-level characteristics like richness and geographic range size, but also note any new insights this combination of techniques provides into the ecological history of the Rosales.

Results/Conclusions

The outstanding fossil record of the Rosaceae allows us to compare ASR inferences against paleontological reality. Rosaceous lineages appear to have originated in subarctic North America during the Eocene and spread south and to the Old World. Most other Rosalean families appear to be older and temperate or tropical in origin. These overall trends are likely due to cooling climate over the Cenozoic driving environments at temperate latitudes from megathermal to cool or arid climates.

Preliminary analyses indicate that ASR tends to infer high paleodiversity in groups of genera (e.g. tribes) that have at least one member that is presently species-rich. Groups without a living, widespread or species-rich genus are more often inferred as less diverse than their fossils indicate. ASR less consistently estimates paleogeographic extent. This may be partially due to the aforementioned climate-driven expansion of the Rosaceae across the world. ASR correctly recovers extremely widespread genera (e.g. Prunus) with widespread fossil distributions. However ASR does not recover paleoendemics with widespread fossil distributions (Lyonothamnus) accurately. Thus, ASR appears to be useful in reconstructing paleodiversity and biogeography to a point, but its bias toward living, diverse tribes needs to be considered when interpreting its results.