GNGTS 2023 - Atti del 41° Convegno Nazionale

Session 1.1 GNGTS 2023 Analysis of the Castelsaraceno microearthquake sequence through a semi-automated and machine learning-based approach S. Panebianco 1 , V. Serlenga 1 , C. Satriano 2 , F. Cavalcante 1 , T.A. Stabile 1 1 Istituto di Metodologie per l’Analisi Ambientale, Consiglio Nazionale delle Ricerche, Tito Scalo (PZ), Italia 2 University of Paris, Institute of the Globe Physics, Paris, France The accurate characterization of microearthquake sequences allows seismologists to gain insights into the physical processes involved in earthquake nucleation and rupture propagation and to image faults geometry at depth. In this context, the improvement of standard procedures for earthquake detection, phase-picking, and location – mainly based on manual approaches and requiring a large amount of work from expert seismologists – towards semi-automatic or full-automatic workflows would allow more efficient analysis of seismic sequences. In this work, we investigate the low-magnitude, foreshock-mainshock-aftershock sequence occurred on August 2020 close to Castelsaraceno village (High Agri Valley, Southern Apennines, Italy) through a semi-automated template matching and machine-learning workflow. Seismic data were mainly recorded by eight broadband stations belonging to the High Agri Valley geophysical Observatory (HAVO), located at a maximum epicentral distance of ~20 km from the seismicity cluster. First, we manually detected 65 microearthquakes in the period between 7 and 10 August, when most of the events of the sequence occurred. Then, more than twice the number of manually detected earthquakes were additionally identified through the semi-automated single-station template matching technique applied to the continuous data-stream recorded in the period between July and October 2020 by the two nearest stations to the cluster. The phase-picking was automatically performed through a deep-learning algorithm (Phasenet) on the 202 detected microearthquakes; then, an automatic multi-step absolute and relative earthquake location procedure was implemented. We identified a total of 76 events that occurred closely spaced in time (7-12 August) and in a narrow range of depths (10-12 km) as belonging to the Castelsaraceno sequence. The Ml 2.1 foreshock doublet followed a few hours later by the Ml 2.9 mainshock – both occurred on 7 August - ruptured the same patch, thus suggesting the main rupture process generating the sequence a persistent asperity. The estimated b-value (0.73± 0.04) also suggests negligible pore-fluid pressure conditions and the activation of the sequence in a relatively low-heterogeneity material. The analysis of aftershocks distribution and the focal mechanism of the mainshock, integrated with the