Automated assessments based on species distribution models can support regional Red Listing but need to be applied with caution: A case study from central Germany

Abstract

AbstractRegional Red Lists (RL) are a cornerstone for biodiversity monitoring and conservation legislation in many countries, including Germany. Yet, the effort to keep RL up-to-date by regularly re-assessing species often is a substantial burden for regional authorities and volunteers. Automation methods using species distribution models have proven promising to speed up assessments on the global RL, but their value for regional Red Lists remains unclear. Here, we use the central German state of Hesse as a model to test how an automated RL assessment based on modelled species distribution ranges at three time slices compares to the latest expert-based RL for > 1,100 plant species. We find the resulting assessments of extinction risk, current population situation and population trends to agree in roughly 50% of the cases. While the model-based assessments were simplistic in some cases, they more more adequate in others. In particular, the assessment of moderately common species was a particular strength of the model-based approach. By examining wrongly assessed species in detail, we identify six themes in which automation may be particularly useful to support expert based regional red listing in the future, including: population trend data, cultivated habitat, taxonomically problematic groups, rare and under collected species, and the quantification of uncertainty in the assessments.