COS 52-5
Estimating species associations and trait-environment relationship for specious ecological communities using Bayesian hierarchical modelling

Tuesday, August 11, 2015: 2:50 PM
339, Baltimore Convention Center
F. Guillaume Blanchet, Mathematics and Statistics, McMaster University, Hamilton, ON, Canada

In community ecology, building models is often a complex task for a few reasons. First, the multivariate nature of community data is technically challenging to handle, resulting in difficulties in making inferences and predictions. Second, because most species in a community are rare, obtaining reliable inferences when constructing species-specific models is a difficult task. Third, to better understand the complexity of nature, ecologists are using an increasing diversity of data (e.g. habitat characteristics or species traits); linking these different data types in an ecologically meaningful way require technical developments beyond that of traditional statistics. Fourth, some of the questions asked by ecologists, such as “how to account for species association for specious communities?”, cannot be answered without the development of new methodologies. In this study, I present a framework that models species co-occurrence by estimating the positive and negative correlations among species within a large community and that also accounts for species traits, habitat characteristics and random effect.


Using simulations, I showed that the modelling framework is efficient in capturing many types of species association structures (from simple to complex ones). In addition, the modelling framework can also be used to illustrate even complex species association in a fairly intuitive way. To illustrate how this framework can be used on real ecological data, I analysed a diatom dataset of 499 species gathered from 105 sites in 7 streams found in a latitudinal gradient across Finland. This dataset also included 12 habitat characteristics and one species trait (diatom size). In this analysis, I showed that the proposed modelling framework can be used to make prediction about the relationship between species traits and the environment for species known to be present in the sampled streams but that were not sampled at the 105 sites. Moreover, I also show that when taking into account environment and traits, the association between species seems to be strong only for a small but specific number of species.