DeepDive: estimating global biodiversity patterns through time using deep learning

This paper introduces DeepDive, a novel approach for estimating global biodiversity changes through time using deep learning and stochastic simulations. DeepDive overcomes challenges posed by the incompleteness of the fossil record by incorporating spatial, temporal, and taxonomic sampling variations. The method outperforms existing approaches, particularly at large spatial scales, delivering robust palaeodiversity estimates. Applications to Permian-Triassic marine animals and proboscideans reveal revised assessments of mass extinctions and significant diversity fluctuations.

Rebecca B. Cooper, Joseph T. Flannery-Sutherland, Daniele Silvestro