EVALUATING SELECTIVITY PATTERNS DURING A MASS EXTINCTION USING AN INVERSE MODELING FRAMEWORK

Selectivity patterns are a critical line of evidence for constraining the drivers of past mass extinction events. However, inferences from extinction selectivity patterns are often based on a posteriori interpretation of extinction/survival patterns rather than on a prior expectation of the extinction/survival patterns predicted by different scenarios. Mass extinction events typically coincide with multiple correlated environmental changes and the timing and nature of these changes is not always clear from limited proxy data, and consequently it is not always clear which potential drivers are most implicated by a given selectivity pattern. To address this, we turn to inverse methods, which are particularly useful for determining which of a large number of potential scenarios are best supported by observations. Our approach is grounded in empirical observations of species occurrences prior to an extinction event. We create probabilistic ecological niche models (ENMs) based on observed occurrences of each species and a hypothetical climatic/environmental starting state simulated using the GENIE Earth system model. These ENMs are then used to predict expected extinction/survival and geographic distributions of surviving species in a hypothetical transitional or final climatic/environmental state. By comparing predictions from a large number of potential climatic/environmental change scenarios to observed patterns, we can provide insight into which hypothetical sequence of climatic/environmental changes are most consistent with the known fossil record. We will present preliminary results of this methodological approach focused on the Late Ordovician Mass Extinction (LOME) record of graptolites. Graptolites are particularly useful for this question, as they are widespread, diverse, and undergo large macroevolutionary changes temporally linked to climatic changes. In addition, many graptolite species had particular habitat preferences that we can reconstruct independently of the ENM’s and that thus provide a means to assess their fit to alternative scenarios of Hirnantian environmental change.