Machine learning identifies ecological selectivity patterns across the end-Permian mass extinction

Abstract The end-Permian mass extinction occurred alongside a large swath of environmental changes that are often invoked as mechanisms, even when direct link is lacking. One way to elucidate the cause(s) investigate selectivity, it can reveal critical information on organismic traits key determinants and survival. Here we show machine learning algorithms, specifically gradient boosted decision trees, be used identify well predict risk. To understand which factors led during an extreme global warming event, quantified ecological selectivity marine extinctions in well-studied South China region. We find varies between different groups organisms synergy multiple stressors best explains overall pattern. Extinction risk was greater for genera had low species richness, narrow bathymetric ranges limited deep-water habitats, stationary mode life, siliceous skeleton, or, less critically, calcitic skeletons. These selective losses directly effects rapid injections carbon dioxide into ocean–atmosphere system, combined expanded oxygen minimum zones, warming, potentially ocean acidification.