WP1. Enhancements of Existing Crustal Velocity Models and Seismic Datasets
WP1.1 Crustal Velocity Models
The crustal velocity models (CVM) have been developed for seismic hazard-related and ground motion simulation purposes in different regions, for example, through 3D waveform tomography, to velocity models (Small et al. 2017, Agostinetti et al. 2022, Chirabba and Amato 2003). However, the potential for these models to contribute to new advances in seismic hazard analysis depends on their accuracy and flexibility (e.g., meshing capabilities).
This WP aims to compute high-resolution tomographic models of the three target regions by using a high-quality pool of local earthquakes' P and S travel times. This process will combine shallow near-surface structures (basins and topography) from other geological and geophysical investigations. We will perform a graded tomographic inversion technique following the approach Gunasekera et al. (2003). Starting from the published tomographic models characterized by nodal spacing in 5-10 km in horizontal directions and 3-5 km along with the depth, we will determine new velocity models with a finer grid spacing. This description will be achieved by sensibly increasing the number of seismic stations in the tomographic inversions. These additional stations that will increase the resolution of our models have been mainly deployed on the alluvial plains to study amplification effects. Our investigations will be carried out by exploiting a large number of datasets for the three selected areas (AREA 1, 2, and 3 in Figure 1), related to the main seismic sequences (2009 L’Aquila, 2016-17 Amatrice-Norcia) and background seismicity (Bagh et al. 2017, Frepoli et al., 2017). We will use high-density network data consisting of i) permanent and temporary seismic stations deployed in the epicentral region to record the 2009 and 2016-2017 seismic sequences (Chiaraluce et al. 2017); ii) seismic stations deployed on the Amatrice, L’Aquila, and Fucino plains to study the amplification effects (EMERSITO, Cara et al. 2019, 2011; Margheriti et al. 2011, Milana et al. 2011. In addition, for AERA 1, the 3D geological model of the crustal volume where the 2016-2017 seismic sequence occurred (RETRACE project, Di Bucci et al. 2021) will be used to detail the shallower layers of the tomographic reconstruction.
WP1.2 Strong motion and macroseismic dataset
The conversion of Ground Motion Parameters (GMPs), both recorded and simulated, into Macroseismic Intensity (MI) and vice-versa allows for testing the results of ground motion scenarios and finite-fault simulations of earthquakes that occurred before the existence of strong motions recordings. Italy, particularly the central Apennines region, presents a unique and rich availability of both kinds of data.
For this purpose, the dataset of macroseismic intensity data and ground motion parameters built by Gomez Capera et al. (2020) to derive empirical conversion relations for the Italian area will be updated and improved. This dataset will be updated with the recently published versions of the two input databases, DBMI15 version 4.0 (Locati et al. 2022), ITACA version 3.2 (Russo et al. 2022), and possibly enriched with new data and further GMPs.
Another MI dataset and earthquake source parameters will be derived from the latest versions of CPTI15, and DBMI15 will be used within WP3.1 to calibrate the updated Intensity Prediction Equations (IPEs). The dataset will be linked to the seismogenic sources defined in WP1.3 to include further source-to-site distance measures for selected case studies.
Other MI datasets potentially used within the project for different purposes, i.e., macroseismic intensity distributions of earthquakes of any magnitude and period, are available through the web services of DBMI15. WP1.2 will perform sanity checks and quality assessments of the MI datasets used within the different WPs and Tasks within the project to ensure a robust and accurate evaluation of the results. The flat file of the ITACAext (Brunelli et al. 2022) dataset, integrated with velocimetric data in Central Italy and classified in terms of aftershock/mainshock (based on updated event localizations), will also be employed to develop and test ground motion prediction models within the WP3.1 and WP4.
WP1.3 to include further source-to-site distance measures for selected case studies
Other MI datasets potentially used within the project for different purposes, i.e., macroseismic intensity distributions of earthquakes of any magnitude and period, are available through the web services of DBMI15. WP1.2 will perform sanity checks and quality assessments of the MI datasets used within the different WPs and Tasks within the project to ensure a robust and accurate evaluation of the results. The flat file of the ITACAext (Brunelli et al. 2022) dataset, integrated with velocimetric data in Central Italy and classified in terms of aftershock/mainshock (based on updated event localizations), will also be employed to develop and test ground motion prediction models within the WP3.1 and WP4.
WP1.3 Geological and seismotectonic input data
Subsurface geological and structural data at various scales is essential for constraining and developing crustal velocity models. In contrast, the geometry of the shallow geological structures, i.e., the thickness of the recent deposits and depth of the bedrock interface of the Quaternary intermountain basins, may serve for studying local amplification effects. Seismotectonic input data, the geometry, and kinematics of the seismogenic sources are necessary to calculate synthetic ground motions and generate kinematic rupture models with the slip distribution on the fault planes.
WP2. Characterization of Earthquake Source and Seismic Wave Propagation in the Central Apennines
WP2.1 Kinematic and Dynamic Earthquake Rupture Modeling
Analyses of recent large earthquakes indicate that multi-scale earthquake rupture models that combine large-scale, large-slip features with smaller but effective high-frequency seismic energy generation areas better reproduce overall characteristics of ground motion on a broad frequency range (Graves and Pitarka 2010, Pitarka et al. 2021). A critical element capturing important aspects of the earthquake source physics is the stress drop parameter since it is directly linked to the spectral level of ground motion. Empirical and theoretical studies (Bindi et al., 2019, Gallovic and Velentova, 2020) show that this parameter is highly variable, and its features are difficult to determine.
This WP will determine the stress drop parameter with empirical methods: spectral fitting, spectral ratios (Calderoni et al. 2019), and the coda method (Mayeda et al. 2003), using seismic records from 2009.
We will also study the source rupture directivity effects (Colavitti et al. 2022) of the shaking scenarios simulations in Central Italy. This way, the empirical information on directivity effects can indicate a more precise nucleation point (Yoshida, 2019; Ross et al., 2020) or bilateral (Boatwright, 2007).
Since the evolution of earthquake rupture models and a description of kinematic and dynamic rupture parameterization is constrained by simulations, in this WP, we will generate stochastic/hybrid deterministic rupture models for the seismogenic sources (WP1.3) using Graves and Pitarka (2016) approach that combines deterministic large slip patches with the randomized spatial field. Deterministic fault rupture models will be retrieved from a global nonlinear inversion procedure for earthquakes M>6.0 in the study area (Cirella et al., 2012, 2018, Scognamiglio et al., 2018, Spudich et al., 2019). To build a database of source rupture scenarios in central Italy, we will employ an innovative approach (Valentová et al. (2021) based on dynamic simulations (FD3D_TSN code, Premus et al., 2020). Moreover, the open-source Pylith code (Aagaard et al. 2017) will be utilized to calculate the final slip and stress on the fault surface for the dynamic modeling in the study areas.
WP2.2 Seismic Wave Attenuation Models
It has long been understood that seismic attenuation has the potential to be a relevant source of information about the Earth's interior. It is an important parameter that could significantly improve our understanding of subsurface processes, integrating the seismic velocity and conceding more detailed outcomes (Cormier, 2011). Moreover, the attenuation of seismic waves is strongly affected by rock permeability, pore fluids, and saturation levels (Gabrielli et al., 2022, Akinci et al., 2020, 2022, Malagnini et al. 2019). It can provide important physical information about crustal medium and its heterogeneities, giving us a description of its property and structure (e.g., fractures, fluids). Under the framework of the previous Dynamic Earth project 2020-2022, Gabrielli et al. (2022) observed that the coda attenuation tomography demonstrated an evident variation between the pre-sequence and the sequence over a series of time windows before and after the three main shocks of the Amatrice-Visso-Norcia, AVN, 2016-2017 seismic sequence and possible influence of deep-CO2-bearing fluids migration in the evolution of the Central Apennines seismic sequence. However, the scattering attenuation indicated the influence of tectonic structures with the strongest contrasts in scattering marked by the Monti Sibillini thrust and the highest scattering delineating carbonate formations. Moreover, Castro et al. (2022a) showed that the spatial variability of the high-frequency attenuation is probably related to different local geology, faulting activity, and tectonic structures in different directions concerning the Apennine orientation.
This WP investigates the temporal variability of the high-frequency attenuation parameter κ (the spectral decay parameter) before and after two main earthquakes, the M6.0 Amatrice earthquake and the M6.5 Norcia earthquake 2016. This approach can help us understand the factors affecting the variability of seismic attenuation and give an indicator to monitor the earthquake cycle in Central Italy.
This WP will also investigate the anelastic attenuation of the S-wave in 2D space and examine how the Q(f) variations modify the earthquake-induced ground motions in the central Apennines. To this end, we will measure the total attenuation of S waves (Qs) using the coda-normalization (CN) approach (Yoshimoto et al. 1993) and a large number of high-quality, three-component time histories (fore- and after-shocks of the L'Aquila 2019 and AVN 2016-17 sequences).
WP3. Ground Motion Predictions using Innovative Approaches
WP3.1 Empirical Non-Ergodic Approaches
Ground motion models (GMM) are key elements in applied seismology to predict the shaking at sites for a given scenario event defined by, at least, magnitude and distance. Recently, the research community has made a strong effort to reduce the variability associated with the prediction and move from global to regionalized GMMs. The most advanced strategy for this aim is to relax the ergodic assumption through a statistical decomposition technique (Al-Atik et al. 2010). The systematic residuals of a nonergodic GMM can also be spatially interpolated on a grid to obtain the empirical spatially-correlated maps (Sgobba et al. 2021) and related to seismological parameters (Bindi et al. 2017; Picozzi et al. 2019a,b and 2022) describing the source (e.g., stress drop, apparent stress, magnitude scale), the propagation (e.g., quality-factor) and the site effects (e.g., high-frequency attenuation). Recent empirical observations and numerical modeling demonstrated that such parameters vary not only with space but also with time (Bindi et al. 2018, Castro et al. 2022b, Picozzi et al. 2021 and 2022), especially during the seismic sequences, due to the modification of the properties in the earth's crust after the occurrence of strong events.
In this WP, we will study the characteristics of the ground shaking pattern evolution during seismic sequences in Central Italy by exploiting a very dense dataset of ground motion parameters extensively investigated by several authors (Sgobba et al. 2022, Spallarossa et al. 2021) and updated in WP1 and WP2. The main goal of WP 3.1 is to answer the following questions: do aftershocks generate different ground shaking concerning mainshocks (in terms of both median and variability)? How should the explanatory variables be modified to predict the ground motions during a seismic sequence? Which magnitude scale should be used to improve the ground motion intensity prediction (Mw vs. Me vs. ML)?
From the datasets developed within WP1.2, this WP will also calibrate updated Intensity Prediction Equations (IPEs) and GMMs, at the national scale. The spatial variations of the attenuation of macroseismic intensity and instrumental recordings, with special reference to Central Italy, will be investigated, compared, and possibly quantified through the analysis and the decomposition of the residuals into the source and path components. This analysis will allow us to determine regional correctives to define regional attenuation models specific to the study area.
WP3.2 Physics-based Numerical/ /Hybrid Ground Motion Simulations
Seismic ground motion hazard assessments depend mainly on empirical observations of earthquake shaking intensity. This latter is highly variable and depends on earthquake- and region-dependent properties. Yet, we have few recorded motions from large earthquakes in the near-fault region where damage is significant. However, advanced simulation techniques offer the possibility of generating many seismograms and providing a better understanding of earthquake physics.
This WP will generate physics-based ground shaking scenarios to investigate ground motion variability induced by the large earthquake ruptures and elastic wave propagation through a complex 3D earth structure. We will calculate ground motion parameters and engineering interest time series relative to some strongest events that occurred in central Italy; the 2009 L'Aquila M6.2; the 2016 Norcia M6.5, Amatrice M6.0, and the 1915 Avezzano M7.0 earthquakes.
We will use consolidated and innovative stochastic, numerical, and hybrid ground motion simulation methods. In the hybrid approach, the synthetics' low-frequency portion (f<1 Hz) will be calculated using a 3D wave propagation method. A stochastic finite-fault simulation model will attain the high-frequency part of seismograms accounting for physics-based source models, spatially varied seismic wave attenuation, and site-related parameters (derived in WP2 and WP3.1). We will also use a physics-based deterministic approach to calculate fully deterministic and 3D earthquake ground-motion simulations. We will use SW4, a finite-difference simulation code with a conforming curvilinear mesh to model surface topography with high numerical accuracy (Sjogreen and Petersson, 2012). In this way, we will generate more realistic shaking scenarios up to a larger frequency (0-5 Hz) (Pitarka et al. 2022) by including the coupled effects of i) a fault rupture (hybrid deterministic and stochastic kinematic/dynamic physics-based rupture models derived in WP2) and ii) path- and site-specific wave propagation through 3D Earth models (e.g., diffraction, scattering, focusing, site and basin amplification, and topographic effects) (well-constrained local 3D velocity models derived in WP1). The simulations will be performed by running the parallelized central processing unit (CPU) version of the code on the Quartz supercomputer at Lawrence Livermore National Laboratory. We will also benefit from the HCP structure that will be established in INGV in the coming year.
WP4 Testing the Performance of Ground Motion Models and Empirical Relationships
It has been demonstrated that the uncertainty corresponding to the selection of the GMMs influences the hazard results more than other aspects of seismicity modeling. This epistemic uncertainty is often treated by adopting an expert opinion approach through a logic tree/ensemble model framework where the weight assigned to each model (i.e., logic tree branch weights) corresponds to the degree of belief of experts in different prediction models. Therefore, selecting appropriate GMMs and their treatment in a logic tree is a fundamental task in any seismic hazard analysis (Stewart et al. 2015). The ranking of GMMs, based on the comparison of GMMs against observations, is a critical step in selecting appropriate GMMs for the region of interest (Scherbaum et al. 2009, Roselli et al. 2016). The present task aims at identifying GMMs and simulation methods suitable for predicting ground motions in central Italy. In particular, we intend to investigate the performance of empirical/semi-empirical models and numerical ground-motion simulations developed within WP3 against observed data from WP1 in Central Italy. Such a task will be performed using classical and standardized residual analyses (Scherbaum et al. 2004). These results will help identify the most suitable model/s to predict ground motion for future events in Central Italy. Furthermore, a prototype method based on a weighted likelihood approach will be applied to rank the models. Classical ranking schemes assign the same weight to each observation independently of its intensity (Scherbaum et al., 2009). Here, we plan to implement a strategy to give different weights to each data point. Such weights will be assigned according to engineering metrics, like vulnerability functions of structures and infrastructure systems. Such an approach will be based on fragility curves for different structural typologies and damage states as a function of ground motion intensity measure (IM) types. This is particularly relevant for engineering applications. Finally, this task aims at finding a proper statistical approach to test the reliability of the empirical relationships between macroseismic intensity and the observed ground motion IMs. Such a test is a novelty in this field, and it could accurately validate the use of these important relationships.