GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 1.1 GNGTS 2024 statstcs of associated events and phases as well as some examples that highlight the importance of fne-tuning the parameters, which becomes complex to do in the presence of very large datasets. This frst part of the workfow (detecton and associaton) resulted in 1.4M events. Fig. 1 – Map, cross sectons (1-3, width=5km) and longitudinal secton (4, width=20km). Cross sectons 1 and 2 clearly depict the Altotbernia Fault at depth and the shallower seismicity pertaining to the Pietralunga (1) and Gubbio (2) sequences. The lithological model on the background is from Latorre et al., 2016. To estmate the absolute hypocentral locatons we use an updated version of HypoSVI (Smith et al., 2022), a probabilistc locaton method in which the forward model is based on a physics informed neural network trained to solve the Eikonal equaton. The locaton procedure makes use of source- specifc staton terms (SSST) that can vary as a functon of source positon, allowing beter correcton for the unmodeled velocity structure with respect to the more classic statc staton correctons. During the locaton procedure, outlier picks are fltered out based on the statstcs of the residuals. We provide absolute locatons for more than 10 tmes the number of earthquakes recorded by the Italian seismic network, with a spatal distributon of seismicity that illuminates the expected crustal portons and agrees with independent informaton from a lithological model of the area (Fig. 1). We are currently in the stage of waveform similarity analysis (cross correlaton, CC). The CC refned traveltmes will be used to relocate the catalogue, providing a very detailed picture of the complex seismicity of the area. The similarity values will also be used to start investgatng the presence of repeatng earthquakes in order to beter defne the mixed distributon of locked and creeping