The Effect of Seismic Sequences in Probabilistic Seismic Hazard Analysis

Abstract

ABSTRACTTraditionally, probabilistic seismic hazard analysis (PSHA) considers only mainshock events and models their temporal occurrence through a homogeneous Poisson process. Thus, it disregards foreshocks and aftershocks, assuming they have a minor effect on PSHA. However, recent earthquake sequences, such as those in 2016–2017 in Central Italy and 2010–2011 in Christchurch, New Zealand, exposed the shortcomings of such a universally used but unconservative approach. Our efforts to quantify the bias in seismic hazard and risk estimates follow from these considerations. Herein, we investigate the epidemic-type aftershock sequence (ETAS) model’s ability to reproduce the statistical features of long-term historical seismicity in Italy in two different regions.In addition, we calculate and compare the seismic hazard at two sites in Central Italy using different approaches: (1) with seismicity clustering modeled using the ETAS model; (2) with only mainshocks modeled by means of the Poissonian approach; and (3) with seismicity clustering modeled via a combination of Poisson and modified Omori law. We consider two cases: (1) the “unconditional case,” which uses years of varying seismicity as initial conditions and, therefore, can be considered as a tool for predicting the long-term average hazard, and (2) “conditional case,” in which the hazard is estimated after a specific period, in our case higher than average seismicity. We scrutinize the different modeling assumptions during the process and investigate the effect of using different declustering methods in Poisson-based models. As expected, we find that using the mainshock-only seismicity models yields lower hazard estimates compared to those obtained with the Omori and ETAS model. In addition, we show that Omori and ETAS model predict similar results in the unconditional case, but the Omori model considerably underpredicts the hazard in the conditional case, for a site close to the sequences, when temporal variations in seismic hazard are accounted for.