Imaging and active faults properties

Understanding the occurrence and triggering processes of large earthquakes and seismic sequences is a hot topic in seismology. Different studies tried to gain insight into these complex aspects and suggest, for example, that the geometry of faults and their segmentation could have played a key role in the evolution of large earthquakes and seismic sequences. The fluid pressure in rock volume close to the fault was also evoked to explain the migration of the seismicity and the occurrence of large earthquakes. The imaging of geometrical and rheological heterogeneities on seismogenic faults is a good tool for inferring earthquake processes. The use of geophysical techniques requires a large amount of high-quality data and also for seismic sequences for which countinuous recorded data are available we can try to enhance model resolution. This PhD research aims to gain insight into the earthquake process and the role of fluids during the seismic sequence, using seismological methods and local earthquake tomography which allows for obtaining high-resolution velocity models of fault systems. The Apennines in Italy is an excellent site for studying earthquake-related phenomena. For this reason, I followed a multi-phase tour in seismic modeling from data acquisition, processing, earthquake detection/location and finally seismic tomography, to progressively improve the resolution of information. I applied Double-Difference (DD) 3D and time-lapse seismic tomography, which computes 3D-relative locations and Vp and Vp /Vs velocity models, on two well-known seismic sequences (i.e. 2012 Emilia and 2009 L’Aquila 2009 seismic sequences) simultaneously. The Vp and Vp /Vs models obtained recognized geometrical and lithological features of the fault structures and information about fluids located in the rock volume. In particular, the presence of high fluid pressure (high Vp /Vs ) along the fault system involved in the 2012 L’Emilia seismic sequence, justifies the rapid migration of seismicity after the first mainshock, suggesting also a dynamic interaction between the Ferrara and the Mirandola thrusts. The same methodology applied to the L’Aquila 2009 seismic sequence allows us to define the presence of a structural discontinuity located at about 3-5 km depth that justifies the lack of seismicity at such depth. The time-lapse tomography shows fluidmigration from the footwall to the hangingwall of the Paganica fault, inferring how the fluid could have played a key role in the seismic sequence evolution. These two seismic tomography applications have been obtained by inverting existing datasets. To increase the resolution further, I computed new machine learning (ML)- based seismic catalogs and verified the resolution power achievable by this data. For this reason, I applied these new techniques to an ongoing seismic sequence (i.e. the Mw 4.9 Mugello sequence) and then, on the L’ Aquila 2009 seismic sequence. The last application allowed me to build a new high-resolution seismic catalog and a new DD-seismic tomography using this new dataset. The new catalog shows improvement in both location errors and the number of data. The aid of these new data in tomography increases the resolution of velocity models, particularly Vs and Vp /Vs models. This last part aims to test the machine-learning capabilities to refine earthquake locations and local tomography, in terms of quality and time consumption, to apply these new seismology frontiers for real-time monitoring.