DEVELOPMENT OF AN INVERSE MODELLING FRAMEWORK FOR DETERMINING THE DRIVERS OF EXTINCTION EVENTS

Extinction events are typically associated with complex global environmental perturbations, making it difficult to determine which specific changes were most important in driving extinctions. Extinction selectivity patterns can provide critical insight into the drivers of extinction, but most selectivity analyses consider only a single global change scenario and do not account for the potential effects of incomplete sampling. We present an inverse modelling framework for evaluating multiple hypothetical global change scenarios and determining the range of scenarios that are most consistent with paleogeographic occurrence and extinction/survival patterns observed in the fossil record. Our approach begins by simulating a broad range of potential global states with the cGENIE Earth system model, using constant paleogeography but varying boundary conditions such as atmospheric CO₂ and O₂. For each model we then use the paleogeographic distribution of observed pre-extinction fossil occurrences to generate probabilistic ecological niche models (ENMs) for all species. Randomly selecting a new model to represent the perturbed global state, we use these ENMs to predict where, if anywhere, each species would be expected to persist. Controlling for the geographic distribution of sampling, we evaluate model-data agreement by comparing these predictions to the observed record of extinction, survival, and range shifts. Considering hundreds of potential global change scenarios allows for broad exploration of the model-data agreement landscape. We apply this approach to multiple extinction events of varying magnitudes and durations, particularly the Late Ordovician Mass Extinction (LOME) record of graptolites and the Eocene/Oligocene record of foraminifera. These events broadly coincide with widespread glaciation and changes in climate but differ in magnitude and duration as well as community consensus on the drivers of extinction.