Meeting Summary: Communities of interacting species can have widely different origins. Consider Darwin’s finches, all of which evolved from a single common ancestor that arrived on the Galapagos Islands and diversified over the past 2–3 million years. This is an example of a community on isolated oceanic islands that is made up of a set of interacting species that evolved in situ. By contrast, one might look to the plant communities of small islands in the coastal northwest of North America, which have been assembled from mainland species whose diversification happened in the distant past. These plant communities, and others like them, are assembled via immigration from a large regional species pool.
Communities that are evolved versus assembled might have fundamental differences; unfortunately, we do not know, at least partially because these two extremes of community formation are currently the domain of different fields of study. In situ radiations include some of the most spectacular examples of adaptive radiation, and have been studied through the lens of phylogenetics and trait evolution. By contrast, local communities that have been assembled (by immigration) from a regional species pool have been studied extensively through the lens of both neutral and non-neutral models of community ecology (MacArthur & Wilson 1967; Hubbell 2001; Snyder & Chesson 2003; Rosindell & Harmon 2013; Haegeman & Etienne 2017). We still do not know whether these two types of communities have distinct and consistent differences. Do assembled versus evolved communities conform to the ends of a spectrum, with adaptive radiation at one end and immigration-extinction equilibrium among dispersers at the other? And does this spectrum also predict other aspects of the species in a local community?
Through ongoing collaborations, initiated with a working group at the Santa Fe Institute in November of 2017, we have nearly completed development of a massive simulation based model the incorporates six key eco-evolutionary processes: demographic drift, mutation, immigration, speciation, competition, and environmental filtering. The simulation framework is near completion, however, our ability to fit this model to real data depends on new machine learning approaches that we have yet to perfect. Our new proposed working group will develop these techniques and also prepare our entire inference pipeline (model parameterization, simulation, and machine learning fit to data) for release via open source software and manuscripts to be submitted for publication.