Graduate Fellow

Bayesian model testing of Bergmann’s Rule on mammalian biogeography

PI(s): Michael Landis (UC Berkeley (Berkeley,CA))
Start Date: 9-Sep-2013
End Date: 6-Dec-2013
Keywords: computational modeling, biogeography, macroevolution, phylogenetics, software

The global distribution of life results from many complex factors, with the phylogenetic covariance of distributions between species and the geographical distances between locales being most apparent. Species traits also play a role in dispersal processes; Bergmann’s Rule describes the tendency for large-bodied species (typically endotherms) to be found in colder climates. Ree et al. (2005) described a likelihood method to infer simple dispersal model parameters between a maximum of ten discrete areas using species occurrence data from extant taxa. Dispersal models requiring greater geographical resolution have remained unexplored due to the methodological limit upon the number of areas.
To model Bergmann’s Rule as a dispersal process, I will develop a statistical model of dispersal that conditions on body mass and latitudinal distance from the equator (as proxy for temperature) by introducing covariance in discrete and continuous trait evolution as described by Lartillot & Poujol (2010). Adapting a data augmentation method described by Robinson et al. (2003), my recent research produced an open-source C++ program called BayArea (Landis et al., in prep.) that allows thousands of areas to be included for analysis, which is necessary to discretize the global geography into fine latitudinal gradients.

Through simulation studies, I will verify my ability to select the correct model using Bayes factors, then test for the importance of Bergmann’s Rule in mammalian biogeography by using the time-calibrated phylogeny and body mass data reported by Nabholz et al. (2008) and the species occurrence data using the Global Biodiversity Information Facility (GBIF).

Related products