Species richness correlates of raw and standardized co-occurrence metrics

Measuring beta-diversity and changes in species composition across multiple sites and environments is a major research focus in macroecology, and a variety of metrics have been proposed to quantify species co-occurrence patterns in a species x site occurrence matrix. However, indices of beta-diversity and species co-occurrence are often statistically dependent on the number of species in an assemblage. We compared the results of several common co-occurrence metrics with patterns generated by a spatially explicit neutral model simulation. We found that all measures of co-occurrence and beta-diversity, whether raw, rescaled or standardized by a null model expectation, were highly correlated with the total species richness of the landscape. The one important exception were the effect sizes of the fixed-fixed null model algorithm, which preserves row and column sums of the original matrix during matrix randomization. Our results call for a careful interpretation of meta-analyses of assemblages that differ widely in species richness. At a minimum, observed species richness should be used as a statistical covariate in regression analyses, and results of the fixed-fixed algorithm should be compared carefully with the results of other randomization tests.