A novel method to predict dark diversity using unconstrained ordination analysis

Questions Species pools are the product of complex ecological and evolutionary mechanisms, operating over a range of spatial scales. Here, we focus on species absent from local sites but with the potential to establish within communities - known as dark diversity. Methods for estimating dark diversity are still being developed and need to be compared, as well as tested for the type, and amount, of reference data needed to calibrate these methods. Location South Bohemia (48 degrees 58 ' N, 14 degrees 28 ' E) and Zelezne Hory (49 degrees 52 ' N, 15 degrees 34 ' E), Czech Republic. Method We compared a widely accepted algorithm to estimate species pools (Beals smoothing index, based on species co-occurrence) against a novel method based on an unconstrained ordination (UNO). Following previous work, we used spatially nested sampling for target plots, with the dark diversity estimates computed from smaller plots validated against additional species present in larger plots, and a reference dataset (Czech National Phytosociological Database of >30,000 plots as global reference data). We determined which method provides the best estimate of dark diversity with an index termed the Success Rate Index. Results When using the whole reference dataset (national scale), both UNO and Beals provided comparable predictions of dark diversity that were better than null expectations based on species frequency. However, when predicting from regionally restricted spatial scales, UNO performed significantly better than Beals. UNO also tended to detect less common species better than Beals. The success rate of combining UNO and Beals slightly outperformed the results obtained from the single methods, but only with the largest reference dataset. Conclusions The UNO method provides a consistently reliable estimate of dark diversity, particularly when the reference dataset is size-limited. For future calculations, we urge caution regarding the choice of dark diversity methods with respect to the reference data available, and how different methods handle species of high, and low, occurrence frequency.