Recent description counts predicted new vascular plant species per botanical country better than discovery-curve models in 1950-2020, and 2021-2025 counts passed darkspot totals in 106 of 360 areas

Question: can models of past species descriptions predict where new plant species will be described and how many remain? We used the World Checklist of Vascular Plants (WCVP) version 16 (358,113 accepted species; native ranges in 368 TDWG botanical countries). Prospective test: Ondo et al. (2024, New Phytologist) estimated the undescribed species in each botanical country from descriptions up to 2020. In only five years (2021-2025), the new descriptions were more than this estimate in 106 of 360 countries and more than its upper 95% limit in 59. In the 33 countries with the largest estimates, they were more than the estimate in 26 (sum of estimates 3,567; sum of new records 7,812). The estimates ranked the 2021-2025 counts worse than the 2001-2020 description count in WCVP v10 (2022) did (Spearman 0.82 against 0.91; difference -0.08, 95% CI -0.12 to -0.06). The model input data were almost complete to 2020, so missing records do not explain the low estimates. Hindcast: at origins 1950-2000, we fitted constant-rate depletion, logistic, taxonomic-effort and time-to-event (Weibull) models to each country's record up to the origin and forecast the next 20 years. The previous 20-year count ranked the countries better than every model at every origin (rho 0.90-0.94 against 0.76-0.89). The effort and time-to-event models forecast only 35-58% of the later descriptions. Each model either gave no finite total for many units (logistic: 51-56% of countries; constant-rate: 85-91% of families) or gave totals that the counts passed within 20 years (effort: 59-73% of countries; time-to-event: 8-17% of countries and 55-78% of families). The results do not change with the basionym year in place of the earliest-name year. Limitations: the tests use present taxonomy and present native ranges. They measure agreement with later descriptions, not the true number of undescribed species. The prospective test assumes that the darkspot value is the stock of undescribed species at the end of the data, as the authors' prediction code computes it.

Linnean shortfall · species discovery curves · darkspots · World Checklist of Vascular Plants · forecast validation · hindcast · taxonomic effort · botanical countries · undescribed species · biodiversity knowledge