BAYESIAN MODELING OF THE DURATION OF A MULTI-STAGE MASS EXTINCTION EVENT

Determining the time and shape of an extinction event from stratigraphic positions of fossil finds is an important step in inferring the cause(s) of a mass extinction. Marshall (Geology, 1995) introduced a method for testing whether the fossil record of ammonites on Seymour Island, Antarctica, is consistent with a simultaneous extinction at the position of a known iridium anomaly. In previous work, I have developed methods for testing whether the fossil record is consistent with multiple hypothesized extinction pulses, and for estimating the stratigraphic distance between these pulses, using a classical (frequentist) statistical paradigm. Here, I expand on previous work by proposing a method for estimating the distance between multiple extinction pulses using a Bayesian approach. This approach has several advantages: (1) it results in clear and directly interpretable interval estimates for the duration of the extinction event; (2) it naturally incorporates relevant prior knowledge, such as geochemical or geophysical evidence; and (3) it reflects how humans intuitively draw conclusions from observed data. I demonstrate the method using data from the Permo-Triassic and Cenomanian-Turonian extinctions.