Phylogenetics Research Papers - Academia.edu (original) (raw)

A comprehensive analysis of early dinosaur relationships raised the possibility that the group may have originated in Laurasia (Northern Hemisphere), rather than Gondwana (Southern Hemisphere) as often thought. However, that study focused solely on morphology and phylogenetic relationships and did not quantitatively evaluate this issue. Here, we investigate dinosaur origins using a novel Bayesian framework uniting tip-dated phylogenetics with dynamic, time-sliced biogeographic methods, which explicitly account for the age and locality of fossils and the changing interconnections of areas through time due to tectonic and eustatic change. Our analysis finds strong support for a Gondwanan origin of Dinosauria, with 99 % probability for South America (83 % for southern South America). Parsimony analysis gives concordant results. Inclusion of time-sliced biogeographic information affects ancestral state reconstructions (e.g., high connectivity between two regions increases uncertainty over which is the ancestral area) and influences tree topology (disfavouring uniting fossil taxa from localities that were widely separated during the relevant time slice). Our approach directly integrates plate tectonics with phylogenetics and divergence dating, and in doing so reaffirms southern South America as the most likely area for the geographic origin of Dinosauria. Dinosaurs are one of the most charismatic and heavily studied groups of vertebrates (e.g., Sereno 1999), but despite ongoing work and many recent discoveries, certain aspects of early dinosaurian evolution remain unclear. A recent study of early dinosaur relationships presented a new phylogeny that departs radically from existing models, but forms a broader framework for inferring dinosaur evolutionary history than those that were previously available (Baron et al. 2017a). While that study was focused primarily on early dinosaurian anatomy and relationships, a general observation was made: based upon the geographic locations of key taxa in their phylogeny, a Laurasian setting for dinosaur origin may have been plausible. This proposal ran contrary to an established consensus based on earlier phylogenies that favoured a Gondwanan and, more specifically, South American setting for this event (Nesbitt et al. 2009; Brusatte et al. 2010; Langer et al. 2010; Sues et al. 2011; Boyd 2015). In particular, the European presence of silesaurids, a close outgroup to dinosaurs, and some early dinosaur taxa (e.g., Saltopus) hinted that northern Pangaea may have played an important role in the earliest stages of dinosaur diversification. However, given the space limitations associated with the original article, no quantitative analyses were carried out to evaluate this alternative hypothesis rigorously. Here, we test the alternative hypothesis implicit in Baron et al. (2017a) by analysing their early dinosaur dataset using a novel integration of time-sliced Bayesian biogeographic models (Landis 2016) and tip-dated Bayesian phylogenetics (Drummond et al. 2012; Bielejec et al. 2014; Gavryushkina et al. 2017). The current method simultaneously estimates phylogenetic relationships, divergence dates, evolutionary rates and biogeographic patterns. Biogeographical changes are inferred using models that account for the changing relationships of continents (and thus dispersal abilities) over time-a 'chronobiogeographical' paradigm (Hunn & Upchurch 2001)-and this information provides potentially critical information for improving estimates of tree topology and divergence dates. For comparison, we also perform analogous analyses using traditional biogeographic parsimony-based methods. Previous studies have advocated a similar approach to integrating phylogeny, stratigraphy, morphology and bio-geography in a parsimony framework (reviewed in Rossie & Seiffert 2006). However, that implementation was problematic: as parsimony trees lack an explicit time element, only very simple time-sliced biogeographic models were possible (e.g., three areas and three time slices) and tree topology and divergence dates had to be assessed manually. The analyses attempted here tackle this complex multidimensional problem more efficiently , using explicit Bayesian computational approaches.