In this article, a novel strategy to generate broadband earthquake ground motions from the results of 3D physics‐based numerical simulations (PBSs) is presented. Physics‐based simulated ground motions embody a rigorous seismic‐wave propagation model (i.e., including source, path, and site effects), which is however reliable only in the long‐period range (typically above 0.75–1 s), owing to the limitations posed both by computational constraints and by insufficient knowledge of the medium at short wavelengths. To cope with these limitations, the proposed approach makes use of Artificial Neural Networks (ANNs), trained on a set of strong‐motion records, to predict the response spectral ordinates at short periods. The essence of the procedure is, first, to use the trained ANN to estimate the short‐period response spectral ordinates using as input the long‐period ones obtained by the PBS, and, then, to enrich the PBS time histories at short periods by scaling iteratively their Fourier spectrum, with no phase change, until their response spectrum matches the ANN target spectrum. After several validation checks of the accuracy of the ANN predictions, the case study of the 29 May 2012 $Mw$ 6.0 Po Plain earthquake is illustrated as a comprehensive example of application of the proposed procedure. The capability of the proposed approach to reproduce in a realistic way the engineering features of earthquake ground motion, including the peak values and their spatial correlation structure, is successfully proven.

This article introduces a strong-motion dataset of near-source broadband earthquake ground motions from 3D physics-based numerical simulations—named BB-SPEEDset—obtained by the code SPEED (SPectral Elements in Elastodynamics with Discontinuous Galerkin)—developed at Politecnico di Milano, Italy. Taking advantage of the earthquake ground-motion scenarios produced so far by SPEED, in most cases validated against earthquake recordings, the main objective of this work is to construct and validate a dataset of simulated broadband waveforms to be used as a support for characterization and modeling of near-source earthquake ground motions. To pursue this objective, the following steps were necessary, namely: (1) the implementation of an effective workflow suitable to process in an homogeneous format various SPEED simulations; (2) the generation of broadband time histories using a technique based on artificial neural networks, trained on strong-motion records; (3) the creation of a flat file collecting, for each simulated scenario, the most relevant metadata (fault rupture scenario, site response proxies, source-to-site distances) as well as a comprehensive set of ground-motion intensity measures of the processed broadband waveforms (peak ground acceleration, velocity and displacement, spectral ordinates, duration, pulse period, etc.). Finally, a comprehensive set of consistency checks is made to verify the absence of any systematic bias in the trend of the BB-SPEEDset results with respect to the NEar-Source Strong-motion (NESS) version 2.0 near-source recorded ground-motion dataset. Indeed, the main features of near-source ground motion in BB-SPEEDset, ranging from the statistical distributions of peak and integral measures both at short and long periods, the ground-motion attenuation with distance, to the features of impulsive ground motions and directionality effects, are in substantial agreement with those from NESS.

This study assesses the spatial correlation of broadband earthquake ground motions from 3D physics‐based numerical simulations in near‐source conditions. State‐of‐the‐art models for predicting the spatial correlation are derived from wide datasets including densely recorded earthquakes in different areas worldwide and, therefore, they may be poorly representative of specific regions and near‐source effects. A large set of broadband ground motions simulated by the SPEED code, and enriched in the high‐frequency range with an Artificial Neural Network technique, is used to investigate the sensitivity of crucial parameters in geostatistical analysis (number of receivers), as well as of source, path, and site effects on spatial correlation, with a level of detail which could not be possible otherwise due to the paucity of recordings. First of all, the comparison of our results with those derived from earthquake recordings validates successfully the numerical approach in predicting the spatial correlation in a broad frequency range. Furthermore, the study points out that spatial correlation of response spectral accelerations is significantly affected by the magnitude, forward directivity effects, ground‐motion directionality (fault normal versus fault parallel), and relative position from the causative fault. These features may make critical the use of isotropic and stationary models especially in near‐fault conditions.

This work presents a novel empirical Ground Motion prediction Model (GMM) for vertical‐to‐horizontal (VH) response spectral amplitudes up to 10 s, peak ground acceleration and velocity for shallow crustal earthquakes in Italy. Being calibrated on the most up‐to‐date strong motion dataset for Italian crustal earthquakes (ITA18), the model is consistent with the ITA18 GMM for the horizontal ground motion. This property makes the model useful in probabilistic seismic hazard assessment for Italy to derive compatible vertical and horizontal response spectra. To account for the increase of VH ratios in the proximity of the seismic source, an adjustment term is introduced to improve the prediction capability of the model in near‐source conditions, relying on the worldwide NEar‐Source Strong motion dataset (NESS). The proposed model uses a simple functional form restricted to a limited number of predictor variables, namely, magnitude, source‐to‐site distance, focal mechanism, and site effects, and the variability associated with both VH and V models is provided.

We present a comprehensive review of Discontinuous Galerkin Spectral Element (DGSE) methods on hybrid hexahedral/tetrahedral grids for the numerical modeling of the ground motion induced by large earthquakes. DGSE methods combine the exibility of discontinuous Galerkin meth-ods to patch together, through a domain decomposition paradigm, Spectral Element blocks where high-order polynomials are used for the space discretization. This approach allows local adaptivity on discretization parameters, thus improving the quality of the solution without affecting the compu-tational costs. The theoretical properties of the semidiscrete formulation are also revised, including well-posedness, stability and error estimates. A discussion on the dissipation, dispersion and stability properties of the fully-discrete (in space and time) formulation is also presented. Here space dis-cretization is obtained based on employing the leap-frog time marching scheme. The capabilities of the present approach are demonstrated through a set of computations of realistic earthquake scenar-ios obtained using the code SPEED (http://speed.mox.polimi.it), an open-source code specifically designed for the numerical modeling of large-scale seismic events jointly developed at Politecnico di Milano by The Laboratory for Modeling and Scientific Computing MOX and by the Department of Civil and Environmental Engineering.