Issues with species occurrence data and their impact on extinction risk assessments

Species extinction risk status is critical to support conservation actions. However, full assessments published on the Red List are slow and resource intensive. To tackle assessments for mega-diverse groups, gains can be made through preliminary assessments that can help prioritize efforts toward full assessments. Here, we quantified how incomplete data collation and errors in the taxonomic, spatial, and temporal dimensions of species-occurrence data translate into misclassifications of extinction risk. Using a dataset of >30 million records of terrestrial plants occurring in Brazil compiled from nine databases we conducted preliminary risk assessments for ~94 % of the 6046 species assessed by the Brazilian Red List authority. We found that no unique database contained data sufficient to perform extinction risk assessment of all species; e.g., the risk of 78 % of species can be assessed using data from GBIF. The overall accuracy (66–75 %) and specificity (89–98 %, correct prediction of non-threatened species) were less affected by incomplete data collation and issues in species-occurrence records. Sensitivity rates (correct prediction of threatened species) were commonly low to moderate and strongly affected by incomplete data collation (13–47 %) and spatial issues (38 %). Our results demonstrate that species' preliminary risk assessments have high accuracy in identifying non-threatened species, even when data collection is low and in the presence of issues in species occurrence data highlighting that such an approach can be used to efficiently prioritize species for full Red List assessments. In addition, caution is needed before declaring a species as threatened without considering data collation intensity and quality.