Biogeographical Network Analysis of Cretaceous Terrestrial Tetrapods: A Phylogeny-Based Approach.

Network methods are widely used to represent and analyze biogeography. It is difficult, however, to convert occurrence data of fossil vertebrates to a biogeographical network, as most species were known from a single locality. A new method for creating a biogeographical network that can incorporate phylogenetic information is proposed in this study, which increases the number of edges in the network of fossil vertebrates and enables the application of various network methods. Using ancestral state reconstruction via maximum parsimony, the method first estimates the biogeographical regions of all internal nodes of a given phylogeny using biogeographical information on the terminal taxa. Then, each internal node in the phylogenetic tree is converted to an edge in the biogeographical network that connects the region(s), if unambiguously estimated, of its two descendants. The new method was applied to phylogenetic trees generated by a birth-death model. Under all conditions tested, an average of $CDATA[$CDATA[$>$$70% of the internal nodes in phylogenetic trees were converted into edges. Three network indices-link density, average link weight, and endemism index (EI)-were evaluated for their usefulness in comparing different biogeographical networks. The EI reflects the rate of dispersal; the other indices reflect nonbiogeographical parameters, the number of taxa and regions, which highlights the importance of evaluating network indices before applying them to biogeographical studies. Multiple Cretaceous biogeographical networks were constructed from the phylogenies of five tetrapod taxa: terrestrial crocodyliforms, terrestrial turtles, nonavian dinosaurs, avians, and pterosaurs. The networks of avians and pterosaurs showed similar topologies and a strong correlation, and unexpectedly high endemism indices. These similarities were probably a result of shared taphonomic biases (i.e., the Lagerstätten effect) for volant taxa with fragile skeletons. The crocodyliform network was partitioned into the Gondwanan and Laurasian continents. The dinosaur network was partitioned into three groups of continents: 1) North America, Asia, and Australia; 2) Europe and Africa; and 3) India, Madagascar, and South America. When Early and Late Cretaceous dinosaurs were analyzed separately, the dinosaur networks were divided into 1) North America, Asia, and Australia; and 2) Europe, Africa, India, and South America for the Early Cretaceous and 1) North America, Asia, and Europe; and 2) India, Madagascar, and South America for the Late Cretaceous. This partitioning of dinosaur and crocodyliform networks corroborates the results of previous biogeographical studies and indicates that the method introduced here can retrieve biogeographical signals from a source phylogeny when sufficient data are available for most targeted biogeographical regions.