GNGTS 2022 - Atti del 40° Convegno Nazionale

230 GNGTS 2022 Sessione 2.1 NESTOREV1.0 APPLICATION TO ITALIAN SEISMICITY P. Brondi 1 , S. Gentili 1 , R. Di Giovambattista 2 1 CRS, National Institute of Oceanography and Applied Geophysics (OGS), Udine, Italy 2 National Institute of Geophysics and Volcanology (INGV), Rome, Italy Abstract. The Italian territory is one of the most seismically active areas in Europe, where large earthquakes are often followed by strong aftershocks. These intense subsequent events can lead to the collapse of already weakened buildings and to further loss of lives. Thus, their forecasting would be important in terms of seismic risk reduction and civil protection. Recently, the machine learning-based algorithm NESTORE (Next STrOng Related Earthquake) [Gentili and Di Giovambattista 2017, 2020, 2022] was proposed to forecast clusters in which the mainshock is followed by an aftershock of similar magnitude. Specifically, if the mainshock- aftershock magnitude difference is less than or equal to 1 the clusters are identified as type A, otherwise as type B. After the occurrence of a strong mainshock of magnitude M m , the NESTORE most recent version of the software, called NESTOREv1.0 (Gentili et al. , 2022), computes nine parameters (features) related to the evolution of the number of events with M>M m -2 and their spatial-magnitude distribution over time in the first hours after the mainshock. The values of these features are used to train nine independent decision trees in order to predict the A-type clusters as long as the strong following earthquake occurs at least 6 hours after the mainshock. NESTOREv1.0 is composed of three modules that respectively (1) identify clusters from a seismic catalog, (2) find appropriate thresholds for features to distinguish cluster types on a training dataset, and (3) use the result of the previous module to provide the probability of being type A for an independent test set. In our analysis, we considered for the Italian territory the seismicity recorded in the last 40 years by the ISIDE and OGS catalogs, the latter of which benefits from the dense OGS network in Northeast Italy. We selected clusters with local mainshock magnitude greater than 3.7 in Northeast Italy and greater than 4 in the remaining territory. We obtained 32 clusters for northeastern Italy and 45 for the rest of the national area, which is dominated by the central Apennines seismicity. Analysing the training results on the two catalogues, the NE area appears quite different from most of the national one, showing a smaller productivity both in terms of number of events and energy of the earthquakes. For this reason, two separate trainings of NESTORE are necessary for future correct forecasting, in good agreement with previous results on earlier versions on NESTORE (Gentili and Di Giovambattista, 2017, 2020). We detected another anomalous area in a smaller region between north-western Tuscany and the south-western part of Emilia-Romagna, with a seismicity characterized by bursts of earthquakes of short duration with anomalous productivity. Since the available training set for that region is almost exclusively populated by type B clusters, no separate training is possible, and we excluded the area from our analysis. By using ISIDE catalogue, we trained NESTOREv1.0 on the time-span 1980-2009 (corresponding to 22 clusters) and tested the forecasting capability on the clusters occurring between 2010 and 2021 (14 cases). The software test provides a correct forecasting for 86% of the cases, showing encouraging results for the application of NESTOREv1.0 in Italy. Acknowledgements . Funded by a grant from the Italian Ministry of Foreign Affairs and International Cooperation References Gentili S. and Di Giovambattista R.; 2017: Pattern recognition approach to the subsequent event of damaging earthquakes in Italy. Physics of the Earth and Planetary Interiors, 266 , pp. 1-17. Gentili S. and Di Giovambattista R.; 2020: Forecasting strong aftershocks in earthquake clusters from northeastern Italy and western Slovenia . Physics of the Earth and Planetary Interiors, 303 , 106483. Gentili, S. and Di Giovambattista, R.; 2022: Forecasting strong subsequent earthquakes in California clusters by machine learning , Physics of the Earth and Planetary Interiors, accepted. Gentili, S., Brondi, P. and Di Giovambattista, R.; 2022: NESTOREv1.0: a machine learning-based Matlab package for forecasting strong aftershocks in seismic clusters Submitted to GNGTS 2022.