Extended Data Fig. 2: Minimum set analysis on a subsample of the data in which 30 individuals were randomly drawn from each different year within a site.

We compared the results (dashed blue line) to a null model (solid red line) in which subsamples of 30 individuals were all drawn from the same year. Thus, confidence intervals include both variation across sites in the number of bee species needed, and uncertainty from the sampling process. Specifically, shaded areas represent the 95% confidence intervals across 100 replicates of the sampling process. Each point represents the mean across all replicates for a single site. The difference between the endpoints of the observed and null accumulation curves represents the increase in number of species needed that is due to turnover in species composition across years.