Sensitivity Analysis of Earthquake Loss Estimation Using the Space-Time ETAS Model for Seismicity Clustering

Earthquake occurrence in probabilistic seismic hazard analysis (PSHA) is routinely modeled by means of a Poisson process. The latter is a “memory-less” process, i.e. it maintains a constant rate in time and space. To this end, aftershock and triggered events, whose occurrence is strongly time- and space-dependent, are typically removed from the catalogues of past seismicity employed in most PSHA studies. This practice may be justifiable when applied to produce ground motion hazard maps for design but the notion that mainshock events can be considered representative of the damage potential of seismic sequences is questionable. Herein, we model seismicity as an epidemic-type aftershock sequence (ETAS) process which has been established as a standard tool for short-term seismicity forecasts. Regional loss estimates for the region of Umbria in Central Italy are calculated by means of ETAS-generated catalogues and compared with the results of a conventional, seismically fully consistent Poisson-based model. We further explore different forms and parameterizations of the ETAS model and investigate how some modelling choices, routinely applied in the literature, affect the final risk metrics. Preliminary results show that accounting for aftershock and triggered events in seismic risk assessment can lead to a substantial increase in loss estimates.