Interpretable, accurate predictions of species distributions and community composition: Making the most of prior information

This talk from ESA 2012 demonstrates an early approach I took to modeling species assemblages.  It has two parts.

First, it uses a technique called MARS to identify useful basis functions of environmental variables that have an important role in determining community composition.

Second, it determines each species' response to these variables and to the other species in the assemblage by optimizing a fully-observed Markov random field.