Gentili, Stefania

none

We investigated the high frequency attenuation of S-waves in Southeastern Alps and Northern External Dinarides using waveforms from 331 earthquakes (3.0< Mw< 6.5). The spectral decay parameter, k, was computed using 1345 three component high quality records, collected by the Italian Strong Motion Network (RAN) and by the Short-Period Seismometric Network of North-Eastern Italy (NEI) in the period 1976-2007. Weak motion data from 11 stations of the NEI network and strong motion data collected by 5 accelerometers of the RAN were analyzed. The k parameter was estimated in the 0-250 Km distance range, in a frequency band extending from the corner frequency of the event up to 25 or 45 Hz, using the amplitude acceleration Fourier spectra of S-waves. The observed record-to-record variability of k was modeled by applying a generalized inversion procedure, using both parametric and non-parametric approaches. Our results evidence that k is independent on earthquake size, while it shows both site and distance dependence. Stations of the NEI network present the same increase of k with epicentral distance, Re, and show values of the zero-distance k parameter, k0(S), between 0.017 and 0.053 s. For the whole region, the k increase with distance can be described through a linear model with slope dk/dRe = (1.4±0.1)x10^(-4) s/Km. Assuming an average S-wave velocity, =3.34 Km/s between 5 and 15 Km depth, we estimate an average frequency independent quality factor, =2140, for the corresponding crustal layer. The non-parametric approach evidences a weak positive concavity of the curve that describes the k increase with Re at about 90 Km distance. This result can be approximated through a piecewise linear function with slopes of 1.0x10^(-4) s/Km and 1.7x10^(-4) s/Km, in accordance with a three layers model where moving from the intermediate to the bottom layer both and decrease. Two regional dependences were found: data from earthquakes located westward to the NEI network evidence weaker attenuation properties, probably because of S-wave reflections from different part of the Moho discontinuity under the eastern Po Plain, at about 25-30 Km depth, while earthquakes located eastward (in western Slovenia), where the Moho deepens up to 45-50 Km, evidence a higher attenuation. Moreover, the k estimates obtained with data from earthquakes located in the area of the 1998 (Mw=5.7) and 2004 (Mw=5.2) Kobarid events are 0.017 s higher than the values predicted for the whole region, probably because of the high level of fracturing that characterizes fault zones. The comparison between measured and theoretical values of k, computed at a few stations with available S-wave velocity profiles, reveals that the major contribution to the total k0(S) is due to the sedimentary column (from surface to 800 m depth). The hard rock section contribution is limited to 0.005 s, in accordance with a maximum contribution of 0.010 s predicted by the non-parametric inversion.

The Kobarid area (western Slovenia) was struck by two seismic sequences in 1998 and 2004. Corresponding mainshocks, occurred on April 12, 1998 and July 12, 2004, had magnitude MW =5.7 and MW =5.2, respectively and were located 2.6 km away from each other. Both of them were recorded by the Friuli-Venezia Giulia Seismometric Network, managed by the Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS), installed in north-eastern Italy (http://www.crs.inogs.it). During the sequences, the network recorded hundreds of aftershocks and both standard locations and duration magnitude (MD) estimations were performed. The seismic activity lasted 198 and 135 days, respectively and consisted of 700 events with MD in the range [1.5, 4.6] and 300 events with MD in the range [1.1, 3.6], respectively (Bressan et al., 2009). In this work, we estimate the seismic moment, M0, the corner frequency, fC, the Brune stress drop, ΔσB, the apparent stress, σa and the radiated energy, ES, of the mainshock and of 164 aftershocks of the Kobarid (2004) seismic sequence. The obtained results are compared with two previous analysis performed in the same region, concerning the source parameter scaling of the background seismicity (Franceschina et al., 2006) and of other seismic sequences (Bressan et al., 2007) occurred in the area.

In recent years, there have been significant advances in the understanding of seismicity scaling laws, the study of spatiotemporal correlations, and earthquake clustering, with direct implications for time-dependent seismic hazard assessment. New models based on seismicity patterns, considering their physical and statistical significance, have shed light on the preparation process before large earthquakes and the evolution of clustered seismicity in time and space. On the other hand, the increasing amount of seismic data available at both local and global scales, together with accurate assessments of the reliability of the catalogs, offers new opportunities for model verification. This Special Issue brings together eight peer-reviewed articles. The articles represent a collection of innovative applications of earthquake forecasting, including the earthquake preparation process, seismic hazard assessment, statistical analysis of seismicity, synthetic catalogs, and cluster identification. It is therefore invaluable to seismologists, statistical seismologists, research students, government agencies, and academics. We are especially grateful to all the authors as without them this Special Issue would not have become a reality. As guest editors, we would like to thank the reviewers for their careful evaluation and valuable contributions. Special thanks go to Assistant Editors Carlos Sanchez and Jill Fang for their dedication to this project and their invaluable collaboration in setting up, promoting, and managing the Special Issue

complex multiple mainshocks sequence (24/08/2016, Mw 6.0; 26/10/2016, Mw 5.4 and 5.9; 30/10/2016, Mw 6.5) occurred in central Italy, causing the death of nearly 300 people and widespread destruction of entire villages. The Region-Time-Length (RTL) method is used to analyze the seismicity preceding the first Mw 6.0 Amatrice mainshock. This analysis is performed using the earthquake catalogue maintained by the Istituto Nazionale di Geofisica e Vulcanologia (INGV) after a preprocessing, which includes the declustering of aftershocks. A well-evident quiescence that preceded the sequence was detected. The quiescence extended throughout a broad region north of the epicenter. The largest event of the sequence and the aftershocks covered most of the quiescence region, except for a small area to the west. The quiescence started from the beginning of September 2015 and lasted for approximately 1 year up to the Amatrice mainshock.

Source parameters of the mainshock (ML=5.3) and of 165 aftershocks (0.8 < ML < 3.5) of the 2004 Kobarid (Western Slovenia) seismic sequence are investigated in order to determine the corresponding source scaling relations. Data recorded from July to December 2004 by the Friuli and Veneto seismic network (FV), managed by the Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS) and installed in Northeastern Italy, are employed to obtain the SH-wave amplitude Fourier spectra of the selected earthquakes. For source spectra computation, we consider only records with significant values of the signal-to-noise ratio and, to account for local amplifications, we compute standard H/V spectral ratios (HVSR) for all the stations of the network. After correction for attenuation effects, source spectra obtained at stations with negligible site effects show a good fit with a ω-square model. We adopt different approaches to compute the source parameters and final results are chosen based on the obtained misfits between observed and theoretical source spectra. For 21 earthquakes of the sequence the obtained results are confirmed by the Empirical Green Function (EGF) technique, applied by estimating the spectral ratios of couples of events with hypocentral distance differences smaller than 500 m and magnitude differences greater than 1. The mainshock of the sequence is characterized by a seismic moment of 3.5x10^16 Nm and a corner frequency of 0.8 Hz, corresponding, in the Brune’s model (1970), to a fault radius of 1465 m and a stress drop of 4.9 MPa. Aftershocks have seismic moments in the range [3.3x10^11, 1.8x10^14] Nm, corner frequencies between 1.9 and 12.4 Hz (Brune radii between 95 and 638 m) and stress drops in the range [0.03, 1.55] MPa. The observed scaling of seismic moment (M0) with the local magnitude (ML) is consistent with the trend: Log M0 = 1.06 ML + 10.56. The Brune radius (rB) increases with the seismic moment according to: Log rB = 0.22 Log M0 - 0.40. Moreover, in spite of the high dispersion that characterizes the estimates of the Brune stress drop (ΔσB), we observe also an increase of ΔσB with M0. The mainshock is characterized by 2.4 x10^12 J radiated energy (ES) and 1.9 MPa apparent stress (σa). Aftershocks have energies between 2.0 x10^5 and 7.4 x10^8 J and apparent stress values increasing with the seismic moment in the range [0.01, 0.48] MPa. Radiated energies increase with seismic moments according to the empirical relationship: Log ES = 1.53 Log M0 - 12.47. The scaling of both ΔσB and σa with M0 in the range of magnitude between 0.8 to 5.3, evidences the non-self-similarity characteristics of the 2004 Kobarid seismic sequence. Similar results have been obtained by previous studies concerning the source parameter scaling of the background seismicity and of other seismic sequences of the area.

This article presents the first publicly available version of the NExt STrOng Related Earthquake (NESTORE) software (NESTOREv1.0) designed for the statistical analysis of earthquake clusters. NESTOREv1.0 is a MATLAB (www.mathworks.com/products/ matlab , last accessed August 2022) package capable of forecasting strong aftershocks starting from the first hours after the mainshocks. It is based on the NESTORE algorithm, which has already been successfully applied retrospectively to Italian and California seismicity. The code evaluates a set of features and uses a supervised machine learning approach to provide probability estimates for a subsequent large earthquake during a seismic sequence. By analyzing an earthquake catalog, the software identifies clusters and trains the algorithm on them. It then uses the training results to obtain forecasting for a test set of independent data to estimate training performance. After appropriate testing, the software can be used as an Operational Earthquake Forecasting (OEF) method for the next stronger earthquake. For ongoing clusters, it provides near-real-time forecasting of a strong aftershock through a traffic light classification aimed at assessing the level of concern. This article provides information about the NESTOREv1.0 algorithm and a guide to the software, detailing its structure and main functions and showing the application to recent seismic sequences in California. By making the NESTOREv1.0 software available, we hope to extend the impact of the NESTORE algorithm and further advance research on forecasting the strongest earthquakes during seismicity clusters.

In this study, we investigate the occurrence of large aftershocks following the most significant earthquakes that occurred in Italy after 1980. In accordance with previous studies (Vorobieva and Panza, 1993; Vorobieva, 1999), we group clusters associated with mainshocks into two categories: “type A” if, given a main shock of magnitude M, the subsequent strongest earthquake in the cluster has magnitude ≥M − 1 or type B otherwise. In this paper, we apply a pattern recognition approach using statistical features to foresee the class of the analysed clusters. The classification of the two categories is based on some features of the time, space, and magnitude distribution of the aftershocks. Specifically, we analyse the temporal evolution of the radiated energy at different elapsed times after the mainshock, the spatio-temporal evolution of the aftershocks occurring within a few days, and the probability of a strong earthquake. An attempt is made to classify the studied region into smaller seismic zones with a prevalence of type A and B clusters. We demonstrate that the two types of clusters have distinct preferred geographic locations inside the Italian territory that likely reflected key properties of the deforming regions, different crustal domains and faulting style. We use decision trees as classifiers of single features to characterize the features depending on the cluster type. The performance of the classification is tested by the Leave-One-Out method. The analysis is performed on different time-spans after the mainshock to simulate the dependence of the accuracy on the information available as data increased over a longer period with increasing time after the mainshock.

In this paper, we propose an innovative machine learning approach called NESTORE, which analyses seismic clusters to forecast strong earthquakes of magnitudes similar or greater to those of the mainshock. The method analyzes the seismicity in the first hours/days after the mainshock and provides the probability of having a strong subsequent earthquake. The analysis is conducted at various stages of time to simulate the increase in knowledge over time. We address the main problem of statistics and machine learning when applied to spatiotemporal variation of seismicity: the small datasets available, on the order of tens or fewer instances, need a more accurate analysis with respect to the classical testing procedures, where hundreds or thousands of data are available. In addition, we develop a more robust NESTORE method based on a jackknife approach (rNESTORE), and we successfully apply it to California seismicity.

The large amount of digital data recorded by permanent and temporary seismic networks makes automatic analysis of seismograms and automatic wave onset time picking schemes of great importance for timely and accurate earthquake locations. Since 2002 the Centro di Ricerche Sismologiche (CRS, Seismological Research Center, http://www.crs.inogs.it/) of the Istituto Nazionale di Oceanografia e di Geofisica Sperimentale (OGS, Italian National Institute for Oceanography and Experimental Geophysics) is involved in the EU Interreg IIIA project Trans-national seismological networks in the South-Eastern Alps together with other four institutions: the Earth Science Department of the Trieste University in Italy, the Civil Protection Department of the Friuli-Venezia Giulia Autonomous Region (DPCFVG) in Italy, the Environmental Agency of the Republic of Slovenia (ARSO), and the Austrian Central Institute for Meteorology and Geodynamics (ZAMG). The Antelope software suite has been chosen as the common basis for real-time data exchange, rapid location of earthquakes and alerting. Each institution has an instance of Antelope running at its data center and acquires data in real-time from its seismic stations and those of the other partners. Antelope detects events by STA/LTA algorithm and the association is based on location by grid search. The actual set up for fast location capabilities uses only P arrivals. The location is performed by grid search over 87x81 nodes for an extension of 7x6.4 degrees (corresponding to cells of 8.9 km in longitude and 8.7 km in latitude) centered in Lat=46.26o, Lon=13.28o with depth steps at 0, 2, 4, 6, 8, 10, 12, 14, 16, 20 and 24 km, using the 1D uniform velocity model IASPEI91. Recently the CRS acquired a new SUN cluster hardware: consequently a new set up of the Antelope software suite has been tested to improve location accuracy using a denser grid and also S-phases arrivals. The results of the performances of the new configuration will be shown; in particular, we compute the variance of the differences between the location data sets of the two different configurations, inferring the precision of each data set by comparing them with the reference OGS bulletin database. We adopt the recall, precision and accuracy estimators to appraise objectively the results and compare them with those of the other datasets.