Probability in PSHA: Reply to "Comment on 'PSHA Validated by Quasi-Observational Means' by Z. Wang"

Abstract

One of the attractions of the Monte Carlo simulation approach to seismic hazard is its conceptual simplicity and directness. What a site operator wants to know is what type of earthquake shaking his plant might possibly have to endure in its lifetime. In response, the seismologist takes all the available information on regional seismogenesis, distils it into a numerical model, and then uses that model to investigate all the possible earthquake ground motions that might occur at the site. Given a sufficient number of simulations, every possible outcome will be sampled, and the most likely outcomes will be observed more often than the less likely outcomes. Wang (2012), in commenting on the description of this method in Musson (2012a), states that randomization of an earthquake catalog is an insufficient approach to assessing seismic hazard, as it ignores the tectonic basis for seismicity, and cannot include possible future events not represented in the historical catalog, such as the 2011 Tohoku event. Such a catalog randomization approach has been suggested in the past, for instance by Ebel and Kafka (1999). But this is not what is described in Musson (2012a). Simulated catalogs used for hazard analysis are not randomizations of the historical catalog, but realizations of possible future events based on what is known about regional seismicity. Any individual simulation may have a different number of events from the historical catalog, but it will be consistent with what is known about local seismotectonics. So it may include larger earthquakes than those observed historically, up to a limit inferred from the tectonic situation. Indeed, a catalog randomization approach would not be able to reproduce Cornell-type results as shown in Musson (2012a), since the model used as an example in that paper includes a fault which is assessed on the basis that …