Predicting species richness from species discovery rates – and where it fails
Аннотация
Abstract The number of described species underestimates true diversity in virtually any clade, yet robust species richness estimates are essential for inferring total diversity, enabling cross‐taxon comparisons, and informing conservation. Here, I evaluate the use of species discovery rates to estimate unknown diversity in fossil clades by projecting future descriptions from historical trends to the year 2300. Two predictive methods are employed: parametric cumulative distribution functions (CDFs) and Bayesian dynamic generalized additive models (DGAMs). Both are applied to three fossil mollusc groups with well‐documented records: non‐marine molluscs (freshwater bivalves and freshwater and terrestrial gastropods) globally; freshwater gastropods from the European fossil record; and Mesozoic–Cenozoic scallops (Pectinoidea). For the European gastropod record, comprehensive occurrence data additionally allow abundance‐based diversity estimation (iChao1) for comparison. To assess model robustness, I simulate data incompleteness through subsampling at varying ratios across selected time periods and by removing recent records. Results show that the best‐fitting distribution model is highly sensitive to the degree and timing of subsampling: depending on which historical period is most affected, the preferred model type may shift dramatically, yielding richness estimates that range from imminent saturation to gains exceeding 200% by 2300. DGAMs perform more reliably, producing more conservative and subsampling‐robust projections on average. Together, the results urge caution when predicting species richness from discovery rates and highlight the critical importance of complete, well‐curated fossil databases for meaningful diversity estimates.