Modeling interactions between many species in diverse communities can be prohibited by the number of parameters required, especially if these interactions differ across environments. We developed a Bayesian sparse modeling approach to narrow parameters down, by grouping most interacting species together in a generic term and using the data to select just those species that have non-generic interactions. We tested our method on simulated datasets and have applied it to empirical data from the species-rich York gum-jam woodland understory.

This project grew out of the sToration working group sponsored by iDiv, and is now available as a preprint here, as a recorded talk here, and has recently been accepted for publication. We’re really excited to have this method implemented in ecology, and I’m happy to chat with anyone interested in learning more!