Species that dominate spatial turnover can be of (almost) any abundance

An ongoing quest in ecology is understanding how species commonness influences compositional change. While each species’ contribution to beta diversity (SCBD) depends both on its abundance and how widespread it is (e.g., occupancy) a general expectation for these influences is lacking. Using published data for 9924 species across 177 metacommunities, we modeled relative SCBD as a function of abundance and occupancy using both correlative and mechanistic regression models (the latter derived from population demographic theory). Although the correlative model provided a superior fit to the data, both results suggest it is infrequent (high abundance and mid-high occupancy) species that make the dominant contribution to beta diversity. The nature of their interaction is most apparent when depicted in abundance-occupancy sample space, which shows the probability of making a dominant contribution to beta diversity is a concave-up function of abundance. Species found in an intermediate number of sites (0.56) required the smallest share of total abundance (0.05) to make a top-decile contribution. The abundance-occupancy sample space illustrates how empirical abundance-SCBD relationships can be linear or unimodal and provides a general framework to understand global change processes. To preserve compositional turnover, species of infrequent abundance and occupancy should be prioritized.

Methods
The dataset used for analysis was calculated from 177 different datasets in 117 different study systems collated from 3 published databases: (i) The metaCommunity Ecology:
Species, Traits, Environment and Space (CESTES) database (Jeliazkov et al. 2020); (ii) Ulrich and Gotelli (2010), and, (iii) Deane et al. (2020). 

Each sites x species abundance dataset was analysed separately by calculating each species contribution to beta diversity (SCBD; Legendre and De Caceres 2013), which was the response variable. Raw SCBD scores were converted to normalised SCBD rank by dividing the rank (highest observed SCBD being rank 1) by the number of species in the metacommunity. Thus, SCBD.rnk was on the interval [0, 1). Explanatory variables extracted from the raw data were the number of individuals for all species across all sites (relative abundance) and the number of sites that each species was observed (occupancy).

Jeliazkov, A., D. Mijatovic, S. Chantepie, N. Andrew, R. Arlettaz, L. Barbaro, N. Barsoum et al. 2020. A global database for metacommunity ecology, integrating species, traits, environment and space. Scientific Data 7:e6.

Ulrich, W., and N. J. Gotelli. 2010. Null model analysis of species associations using abundance data. Ecology 91:3384-3397.

Deane, D. C., P. Nozohourmehrabad, S. S. D. Boyce, and F. L. He. 2020. Quantifying factors for understanding why several small patches host more species than a single large patch. Biological Conservation 249:e108711.

Legendre, P., and M. De Caceres. 2013. Beta diversity as the variance of community data: dissimilarity coefficients and partitioning. Ecology Letters 16:951-963.