Including Earthquake Sequences on National Seismic Hazard Models Through ETAS-Based Synthetic Catalogs: The Case of Italy

Probabilistic seismic hazard analysis (PSHA) assumes a time-invariant Poisson process on mainshocks, not accounting for aftershocks. The removal of these events is subjective, as it depends on the chosen declustering procedure. Moreover, declustering may also result in a significant underestimation of the seismic hazard because aftershocks can significantly contribute to ground shaking, as shown by recent earthquake sequences. The present study investigates the reliability and technical feasibility of a methodology based on epidemic-type aftershock sequence (ETAS) synthetic catalogs to include earthquake sequences in PSHA. Specifically, we apply the SimplETAS algorithm to create a large number of realistic synthetic catalogs, each spanning 50 yr. We assign a plausible seismogenic structure in the epicentral area to each earthquake in the synthetic catalog, enabling numerical hazard-curve generation without assuming a Poisson distribution. This methodology has been applied to four Italian cities: L’Aquila, Reggio Calabria, Firenze, and Milano. We compare our hazard curves with the ones from the National Seismic Hazard Model for Italy (MPS19), as well as with its sequence-corrected version, which accounts for earthquake sequences through a numerical adjustment of the Poisson earthquake rate (MPS19_cluster). The results indicate that: (1) explicitly modeling earthquake sequences substantially increases seismic hazard at all sites; (2) SimplETAS-based and MPS19_cluster hazard curves generally agree well, including at short-return periods where the numerical correction used in MPS19_cluster can be mathematically questionable; (3) in Milano—where seismic hazard is the lowest—the two curves diverge, underscoring the impact of the different spatial variability of the magnitude distribution in the SimplETAS and MPS19 models; and (4) notably, the ETAS-based approach permits flexible choice of any forecasting time window relevant to the risk analysis and naturally captures spatial correlation of risk.