Tree-Based and Bayesian Modeling of Food Web Collapse in the Permian Mass Extinction

• Activity Number: 478
• Type: Contributed
• Date/Time: Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Steve C. Wang*+ and Peter D. Roopnarine and Kenneth D. Angielczyk
• Companies: Swarthmore College and California Academy of Sciences and University of Bristol
• Keywords: paleontology ; CART ; MCMC ; simulation-based inference ; forward model ; calibration

Abstract

The Permian mass extinction was the most severe extinction event in the history of life, with up to 90% of species killed. Many causes have been proposed (asteroid impact, volcanism, etc.), but none is supported by incontrovertible evidence. To infer the true cause, it is important to determine the mode of extinction. E.g., did an initial extinction at the lowest level of the food web (plants) lead to the extinction of herbivores and then carnivores - a bottom-up mode of extinction? Or did animals at higher levels of the food web die first, causing extinctions at lower levels - a top-down mode? Or were all levels affected equally - a uniform mode? Here we use a stochastic forward model to simulate the effect of these modes of extinctions. Using simulated data from this model, we then attempt to infer the mode of extinction using (1) classification trees and (2) a Bayesian MCMC model.