Tests of neutral theory predictions for the Barro Colorado Island tree community informed by regional abundance data

Background/Question/Methods

Neutral theory has the potential to serve as a quantitative process-based null model in ecology, whose rejection indicates the importance of processes of interest in assembling communities, such as niche differentiation and habitat filtering. So far tests of neutral theory have not been revealing however.  Its predictions provide a good fit to the relative abundance distribution of tree species in the tropical forest on Barro Colorado Island, as well as other data sets.  However, these predictions are based on using a neutral model for the regional or “meta” community as well as for the local community.  This neutral model for the regional community is problematic in its assumptions regarding speciation, and its key parameter is allowed to vary in generating neutral theory fits to local community observations.  Local community assembly processes might better be detected by testing a neutral model instead informed by observations of the state of the regional community.  Here we carry out such a test, examining how different the structure of the community of tree species in the 50 ha Barro Colorado Island plot looks than predictions of a spatially-implicit neutral model that bases each species’ regional abundance on data from over 50 plots that have been established around the Panama Canal Basin since 1994.

Results/Conclusions

We find that an unrealistically low immigration rate is needed for the species richness and species abundance distribution under neutrality to fit observations. We are able to detect the signature of selection for and against some species in the community.  Work currently in progress includes examining potential relationships between detected selection and biological parameters such as dispersal ability, abiotic associations, and life-history strategy. Our results suggest more refined versions of this approach, involving a more detailed neutral or dispersal-assembly model, may be fruitful.  Ultimately detected selection could be organized along trait axes to look for patterning indicative of niche differentiation or habitat filtering.