Physics-based simulation of sequences with multiple main shocks in Central Italy

The application of a physics-based earthquake simulator to Central Italy allowed the compilation of a synthetic seismic catalogue spanning 100000 yr, containing more than 300000 M >= 4.0 simulated earthquakes, without the limitations that real catalogues suffer in terms of completeness, homogeneity and time duration. The seismogenic model upon which we applied the simulator code was derived from version 3.2.1 of the Database of Individual Seismogenic Sources (DISS; http://diss.rm.ingv.it/diss/), selecting, and modifying where appropriate, all the fault systems that are recognized in the portion of Central Italy considered in this study, with a total of 54 faults. Besides tectonic stress loading and static stress transfer as in the previous versions, the physical model on which the latest version of our simulation algorithm is based also includes the Rate and State constitutive law that helps to reproduce Omori's law. One further improvement in our code was also the introduction of trapezoidal-shaped faults that perform better than known faults. The resulting synthetic seismic catalogue exhibits typical magnitude, space and time features which are comparable to those in real observations. These features include the total seismic moment rate, the earthquake magnitude distribution, and the short- and medium-term earthquake clustering. A typical aspect of the observed seismicity in Central Italy, as well as across the whole Italian landmass and elsewhere, is the occurrence of earthquake sequences characterized by multiple main shocks of similar magnitude. These sequences are different from the usual earthquake clusters and aftershock sequences, since they have at least two main shocks of similar magnitude. Therefore, special attention was devoted to verifying whether the simulated catalogue includes this notable aspect. For this purpose, we developed a computer code especially for this work to count the number of multiple events contained in a seismic catalogue under a quantitative definition. We found that the last version of the simulator code produces a slightly larger number of multiple events than the previous versions, but not as large as in the real catalogue. A possible reason for this drawback is the lack of components such as pore-pressure changes due to fluid-diffusion in the adopted physical model.