Assessing the predictability of extinction risk using machine learning and pattern recognition

As global biodiversity continues to decline it is critical that we have robust tools to rapidly assess the extinction risk of species, particularly in the context of future environmental change. The International Union for Conservation of Nature's (IUCN) Red List of Threatened Species defines a series of objective criteria used to assess and monitor the extinction risk of, at present, over 90,000 species. However, for many taxonomic groups, assessments are lacking.
Approaches such as the Sampled approach to the Red List Index (SRLI) use smaller number of representative species in order to capture the overall extinction risk of lesser known taxonomic groups. However, even this approach can be logistically complicated and slow to implement.
In this project we aim to understand how new machine learning approaches can be used to predictively assess the extinction risk of species and how these predictions might vary (e.g. across species traits, taxonomic groups and biogeographical realms). Importantly, understanding the 'predictability' of different species can inform which species require full assessments and which species may usefully be assessed via machine learning tools. We will then use this approach to improve the assessment process and to investigate the how predicted assessments may change under future land use and climate change scenarios.