Community assembly, natural selection and maximum entropy models

Abstract

The papers in this Forum discussion debate various aspects of my maximum entropy model of community assembly. The questions raised centre around (1) the possible mechanisms generating the patterns predicted by my maxent model of community assembly, and (2) the appropriate statistical methods for testing the patterns. Here I briefly explain the proposed mechanistic basis of the model: natural selection occurring between individuals of different species. If trait differences are linked to differential demographic probabilities (i.e. fitness differences) then natural selection will constrain the average trait values found in the community and such average (‘community-aggregated’) traits will then possess information that is translated into the maximum entropy probabilities. If community assembly is strictly neutral then the maxent model will have no predictive ability. This also justifies the null model, and the permutation test, proposed by Roxburgh and Mokany. Which species, present in a larger geographical region, will be found at a local site? Which species will be dominant and which will be rare; in short, what will be the relative abundance of these species? If environmental conditions at the site change, how will the answers to these two questions change? These are some of the most basic questions in community ecology. Despite a century of hard work by both theorists and empiricists, we do not yet have general quantitative answers to these questions. I recently developed a mathematical model based on the maximum entropy formalism that attempts to provide quantitative answers to these questions and that is applicable to natural plant communities occurring in the field. This model was first introduced in Shipley et al. (2006) and has generated much debate. Three such papers (He 2010, McGill and Nekola 2010, Roxburgh and Mokany 2010) are published in this edition of Oikos and each raises important questions. The mathematical, biological, and methodological justifications for my model are developed in a recent book (Shipley 2009a). Here I will concentrate on the mechanistic basis of the model.