Using Coverage-Based Rarefaction to Infer Non-Random Species Distributions

Understanding how species are non-randomly distributed in space, as well as how the resulting spatial structure of diversity responds to ecological, biogeographic and anthropogenic drivers is a critical piece of the biodiversity puzzle. However, most metrics that quantify the spatial structure of diversity (i.e., community differentiation), such as Whittaker’s classical β-diversity metric are influenced by sampling effects. As a result, these measures are influenced by species pool size, species abundance distributions and numbers of individuals. Null models have been proposed to evaluate the degree of differentiation among communities due to spatial structuring relative to that expected from sampling effects. However, to date, these null models do not accommodate the influence of sample completeness (i.e. the proportion of the species pool in the sample). Here, we develop an approach that makes use of individual- and coverage-based rarefaction and extrapolation. Using spatially explicit simulations, we show that our derived metric, βC, captures changes in intraspecific aggregation independently of changes in the species pool size. We then provide two case studies examining spatial structure in forest plots spanning latitudinal gradients: (1) a re-analysis of the “Gentry” plot dataset, and (2) comparing a high diversity plot in Barro Colorado Island, Panama with a low diversity plot in Harvard Forest, Massachusetts, USA. We find no evidence for systematic changes in spatial structure with latitude in these datasets. As it is rooted in biodiversity sampling theory and explicitly controls for sample completeness, our approach represents an important advance over existing null models for spatial aggregation. Potential applications of the approach range from better descriptors of biogeographic diversity patterns to the consolidation of local and regional diversity trends in the current biodiversity crisis. Open research statement The novel code for the calculation of βC can be found in supplementary material S4. Empirical data sets utilized for this research are as follows: Phillips & Miller (2002), Orwig et al. (2015), Condit et al. (2019). Our research repository including the novel code is also available at https://github.com/t-engel/betaC and will be uploaded to Zenodo upon acceptance of this manuscript.