Robustness of relative diversity metrics to inventory incompleteness: Could we estimate the near imponderable?

Aim Relative diversity metrics (RDMs), such as local : global or regional : global diversity ratios and percentage of endemics, are widely applied in biodiversity studies. This approach, however, is flawed because the real global diversity of many taxa is still largely unknown, especially for rare species. This potential inconsistency of RDMs has not been proved (or disproved) thus far. Here, I use simulation models to explore the efficiency of RDM estimations from incomplete inventories. Location Methods World-wide. Three simple models of species distribution were used: (a) the "everything is everywhere" model (unlimited dispersal, i.e., equal probability for a species to occur at every point); (b) the "environmental selection" model (each species can occupy only a portion of randomly distributed "suitable localities" according to its ecological preferences); and (c) the "endemicity" model (most species are restricted to a few regions each). To parameterize these models, the world-wide datasets on ciliates, flagellates and harpacticoid copepods were used. The RDMs were estimated by simulating equal-effort sampling and by varying the full number of species and number of sampled individuals per cell. Results Main conclusions Estimated species richness at any scale was heavily influenced by the sampling effort and was greatly underestimated with small sample sizes. The estimated RDM values also depended on the sampling effort (being either upward or downward biased) but were only slightly influenced by the unrecorded part of real global diversity (ignorance). The predictions of the first two models were generally close to each other but differed noticeably from those of the third model. The relative metrics, albeit influenced by undersampling, are nevertheless more robust to inventory incompleteness than absolute measures and can therefore be a reliable tool in comparative biodiversity studies, even if the true number of species cannot be estimated overall. Their estimated values, however, must be interpreted with caution.