New seismicity models for the next national Italian seismic hazard model

In 2015 the Seismic Hazard Centre of the National Institute of Geophysics and Volcanology started to promote and coordinate the activities of a project aimed at producing the new national probabilistic seismic hazard model for Italy. In this work, we present the results of the activities of the task 3: "seismicity models", focused on the definition of a set of seismicity models and on the analysis of their uncertainty. 12 working groups produced 11 seismicity models covering the entire Italian territory and 1 model is built ad hoc for the volcanic Etna area. A large number of models allows the exploration of uncertainty in the definition of seismic sources. The models will be combined according to an elicitation phase and a ranking with respect to the observations. Area Source Models: 4 seismicity models are based on seismogenic zonations defined according to the regional seismotectonic settings and past seismicity. Seismicity rates are computed from the declustered parametric catalog of Italian earthquakes (CPTI15, Rovida et al., 2016). 3 of the 4 area source models use macro-areas (groups of zones), to evaluate regional parameters of Gutenberg-Richter (GR) distributions, specifically the models A1, A3 and A4. Then, the seismicity rate for each zone is estimated from the rates of the macro areas using different approaches. Instead, A2 computes seismicity rates on each individual zone, by individuating the change-point between two Poisson processes: the rate of occurrence of part considered as complete of the data set of events is then assumed as representative of the long-term seismic rate occurrence. Fault-based models: A first model, F1, uses exclusively geological information taken from the DISS 3.2.1 (update of http://diss.rm.ingv.it/diss/). The seismic moment rates of a seismogenic fault is derived from the geologic moment rate. To evaluate off-fault seismicity rates F1 follows an empirical approach aimed to capture the natural distribution of observed earthquakes around faults and the inherent location uncertainty of both earthquakes and faults. The second model, F2, combines the seismic rates obtained for the faults, computed using the seismic-moment conservation principle that allows the estimation of the seismic moment rate from long-term slip rate and geometry of the fault source, with a background seismicity obtained using the smoothed-seismicity approach proposed by Frankel (1995). Smoothing seismicity models: A first model, G1, merges two different smoothed seismicity models following the well known and widely applied fixed (Frankel, 1995) and adaptive smoothing methods. A second model, G5 follows the Woo (1996) approach to propose a zone-free method solely based on the use of the earthquake catalog. Moreover, given the availability of some earthquake-fault associations in the DISS database, G5 also uses an anisotropic kernel function when there is a connection between faults and earthquakes. Geodetic models: A first model, G3, estimates the seismicity rates over the whole Italian territory using 919 GPS derived horizontal velocities. The strain rate tensor field is calculated on a regular grid taking into account the variable station spacing for the optimal smoothing parameters and finally applying a Gaussian filter. The model converts the strain rate in seismic moment rate and then to earthquake rate under the assumption (Ward, 1998) that earthquakes magnitudes follow a tapered GR distribution, where the b-value and corner magnitude are given. The second model, G4, follows the approach proposed by Bird et al. (2010) with some adaptations to Italy. To evaluate expected rate of seismicity: (1) the long-term strain-rate tensor is computed by GPS data; (2) for a given grid point a certain amount of strain-rate and slip