Molecular and morphological inference of the phylogeny, origin, and evolution of Aesculus l. (Sapindales)

by 1978- Harris, AJ

Reconstructing the biogeographic history of a taxon and developing hypotheses regarding the events responsible for its current distribution typically requires a well-supported, bifurcating phylogenetic tree. Biogeographic reconstructions normally involve some method of mapping ancestral distribution areas as characters across all nodes of a single phylogenetic tree. In the real world of systematics, phylogenies are not always wellsupported nor does the data always produce perfectly bifurcating trees. We propose that biogeographic information about taxa or lineages of interest need not be restricted to reconstruction of a single scenario across a tree topology. Here, biogeographic analysis using DIVA (Dispersal Vicariance Analysis) software is repeated for a set of randomly sampled Bayesian trees generated using MrBayes. We define a node as the hypothesized ancestor of a specific lineage and its unspecified sister (x). For each node, the probability of each possible ancestral range is calculated as follows: [PT1 +…PT100] / N, where P is the frequency of a range set estimated for a node and Tn is one of N randomly sampled trees.Thus, ranges for each node are reported as a set of probabilities from which biogeographic hypotheses may be developed and tested. This approach obviates the need for use of a single tree topology with well-supported relationships between groups for biogeographic reconstruction. The utility of this approach was tested using Aesculus L. 91