From plant traits to plant communities: Towards a predictive community ecology

Background/Question/Methods Can we predict those species that will be present in a specified environment? Can we predict the abundances of these species as well as their presence? These two simple questions represent major goals of predictive community ecology. To answer these questions, we need a quantitative process-based model that emphasizes the aggregation of individual species into communities, while being applicable in the field. In the functional approach to assembly rules, the habitat operates as a set of filters that select species with suitable values of traits. Recently, Shipley and colleagues used this concept to develop a statistical mechanistic approach linking plant traits to community structure and explained 94% of the variance in species abundances in a chronosequence of secondary succession. However, their analysis was based on only 12 sites and with a relatively poor species pool (30 herbaceous species). Our primary goal was to test for repeatability of the method using 1337 quadrats from 4 habitat types and when prediction involved larger species pools (30, 60, 100 or 506 species). We also used this method to establish a trait hierarchy and determine if it was possible to predict community pattern (species richness, diversity, abundances) with a smaller set of traits.

Results/Conclusions We were able to predict species abundances, accurately with 10 traits and independently of the habitat considered (r² ranged form 0.72 to 0.92, and regression slope from 0.65 to 0.89). The only exception was for the largest species pool (506 species) where relationship was weak (r²=0.31). We also succeeded to predict species richness (r² ranged form 0.80 to 0.60), as well as diversity (measured by Shannon’s index, r² ranged form 0.93 to 0.76); however we tended to overestimate them as we increased the number of species in the pool. We observed that our ability to predict the abundance of rare species was relatively weak, and that we explained variance in species abundances better in the case of poor diversity plots, or when one or a few species dominated the community. Finally, we added the affinity of each species for a particular habitat to the model (an indirect way of accounting for “missing” traits) and improved the fit between observed and predicted abundances.