← Back to Sessions

Removing the Poisson Assumption From Probabilistic Seismic Hazard Assessment Without Paying the Penalty

Session: Beyond Poisson: Seismic Hazards and Risk Assessment for the Real Earth

Type: Oral

Date: 4/19/2021

Presentation Time: 03:00 PM Pacific

Description: 

In his seminal paper on Probabilistic Seismic Hazard Assessment (PSHA), Allin Cornell wrote, “the assumption that the occurrences of earthquakes follow the behavior of the Poisson process model can be removed only at a great penalty” (Cornell, BSSA, 1968). Cornell suggested that the Poisson assumption was sufficient for the low exceedance probabilities over the long time-scales used for many engineering purposes (e.g., 2% in 50 years), even if it was inconsistent with both elastic rebound and the occurrence of aftershocks. That view was supported by Marzocchi and Taroni (BSSA, 2014). However, they also showed that the assumption breaks down for higher probabilities and shorter periods of time. Therefore, the Poisson assumption may not be sufficient for the higher exceedance probabilities currently used for infrastructure operating standards or for the shorter time periods and considerations of multiple exceedances used in the insurance industry. While simulations can be used to compute seismic hazard including behavior such as aftershocks (e.g. Field et al., BSSA, 2017), the computational requirements may be daunting for many applications. In fact, a clear goal of Cornell was an analytic model to allow users to explore the impact of a variety of parameters. We introduce an analytic method for PSHA that allows for arbitrary distributions of earthquake occurrence and that satisfies Cornell’s goal without an appreciable computational penalty. We apply this method to the highly clustered seismicity associated with caldera collapses at Kilauea and explore the implications of non-Poissonian behavior for PSHA as a function of the degree of clustering in seismicity. Understanding the difference between a non-Poissonian distribution and a Poisson distribution with the same mean rate requires examining the probability, rather than the commonly used rate, of exceedance. Thus, including non-Poissonian behavior in PSHA requires changing how we plot and talk about hazard.

Presenting Author: Andrew Michael

Student Presenter: No

Authors