Diversity Estimation
To account for potential sampling issues in both datasets, diversity was estimated using three methods. Richness was estimated using the S2/m equation of Alroy (2020). This gives similar results (Fig. S1) to that of Chao 1 (Chao, 1984) in our analyses but is more aggressive and a consistent lower bound (Alroy, 2020), so we emphasise it in the main text.
To supplement this, we used the analytical version (Chao and Jost, 2012) of shareholder quorum subsampling (SQS - Alroy [2010a, 2010b, 2010c]). We note that SQS is routinely referred to as coverage-based rarefaction (CBR) in the ecological literature and that the distinction between SQS and CBR is not conceptual but operational, as with the distinction between the original formulation of rarefaction (Sanders, 1968) and the analytical formulation (Hurlbert, 1971) that is now widely used.
As these methods are influenced by sample size in our data (Fig. S2), we use Simpson’s D (Simpson, 1949) as a diversity metric. When computed using the modern formula (Hurlbert, 1971), Simpson’s D is almost independent of sample size, and it has long been used in the analysis of ecological field data (Magurran, 2013, Morris et al. , 2014). Additionally, we use Fisher’s alpha, which is often used in the study of high diversity systems (Hubbell, 2015).
All analyses were done in R (version 4.0.2; R Core Team [2020]), and the functions for generating diversity and richness estimates have been included in the Supplementary Information.