Modeling Community Structure across Hierarchical Scales: A Case Study Using Late Ordovician Deep-Subtidal Assemblages from the Cincinnati Arch

Are communities random associations of species or are they structured? If the latter, what ecological rules cause this structure? Here I summarize four models of community assembly that can be used to evaluate the degree of community structure. I define structure quantitatively in terms of the life habits occupied by species within a community. The model of redundancy specifies that communities are composed of clusters of ecologically similar species. Partitioning specifies that community ecospace is subdivided by species along life habit gradients. Expansion specifies that community ecospace is enlarged by the emergence of species with novel life habits. These models are general because they can occur equally through competition or a wide variety of other agencies; however, they all yield distinctly structured communities. A fourth, neutral model also exists in which life habits are not involved in community membership. Model simulations demonstrate that three community metrics (life habit richness and two measures of disparity) are sufficient to distinguish each model as a factor of community species richness. The models can be evaluated using maximum likelihood-based model selection.

As a case study, I focus on the Kope and Waynesville Formations from the Type Cincinnatian (Late Ordovician). Both units represent deep subtidal habitats and contain parautochthonous assemblages. Because the 216 samples hierarchically span multiple localities and stratigraphic levels, they allow testing of whether models are scale-sensitive. Results demonstrate that individual samples are frequently well fit by structured models (especially redundancy and partitioning). The two formations also tend to display different community structures, with Kope best fit by partitioning and Waynesville best fit by the null or weak redundancy models. These results are maintained when individual samples are aggregated hierarchically at larger regional and temporal scales, suggesting the models might prove generally useful for broader-scale, synoptic analyses.