Nick Matzke: Model choice in phylogenetic biogeography - workshop and seminar

Wednesday, November 13, 2013 - 12:00pm to 1:00pm
OML 400 See map

Historical biogeography has been characterized by unresolved debates about which processes, such as dispersal or vicariance, are most important for explaining geographic distributions. Biogeography programs typically make fixed assumptions about which processes are allowed, and these assumptions dominate inference. A new R package, BioGeoBEARS (, implements the most popular models (LAGRANGE DEC, DIVA, BayArea) in a common likelihood framework, and adds new “+J” models that allow jump dispersal/founder-event speciation. Statistical model choice shows that founder-event speciation is a crucial process in almost every clade, although it is measurably weaker in non-island systems. Simulations indicate that better models estimate ancestral ranges with dramatically greater confidence, and higher accuracy.  All of these models are special cases of a supermodel implemented in BioGeoBEARS, enabling estimation of the relative importance of different speciational processes given sufficient data.  At the conclusion of my talk, I outline a program to expand this research by linking phylogenetic biogeography with species distribution modeling (SDM). Standard SDM is done one-species-at-a-time, ignoring phylogenetic autocorrelation, and typically ignores non-climatic causes of distributions such as dispersal limitation.  As a result, all of these factors are confounded in current environmental niche models. With a joint estimation approach we may be able to untangle these factors.

BioGeoBEARS Workshop - Modeling diversification in space

5:00pm - 7:00pm, OML 400

I will offer a workshop on my new R package for historical biogeography, “BioGeoBEARS”, available on CRAN.

The main goals of the workshop are for participants to (1) understand the different historical biogeography methods, and understand what assumptions they are making, (2) to learn how to run BioGeoBEARS and do statistical model comparison and ancestral range estimation, and hopefully (3) try it out on their own data.

A basic demo script, and installation instructions are available at: , including notes on current issues (currently, BioGeoBEARS needs the 2012 version of optimx, not the 2013 version).  It seems to work fine on all platforms (email me if you have issues).  I am planning a major update this week which will change some of the details before the workshop.

Nick Matze, new postdoc fellow at NIMBios, will be visiting for the week to discuss his research and give a workshop on the new R package BioGeoBEARS for modelling diversification in geographic space.