GNGTS 2022 - Atti del 40° Convegno Nazionale

312 GNGTS 2022 Sessione 2.2 parametrized in terms of the V S30 , about 40% of data corresponds to soil conditions with V S30 < 800 m/s, and about 60% to rock conditions, with dominance for hard rock sites with V S30 larger than 1000 m/s. Within the soil classes, the majority of waveforms is on stiff soil with V S30 > 400 m/s, but an appreciable number of data is on very soft sites with V S30 as low as 150 m/s. In addition to the accelerograms, a flat-file provides, for each scenario and each selected receiver, a list of source metadata, post-processing metadata, receiver metadata, site response proxies, source-to-site distances and a broad spectrum of intensity measures (IM). The fields of the flat-file are consistent with the ones adopted in state-of-the-art near-source recorded flat- files, such as the NESS2 dataset (Sgobba et al., 2021). In addition to standard peak IMs, such as Peak Ground Acceleration (PGA), Peak Ground Velocity (PGV), Peak Ground Displacement (PGD) and response spectral accelerations (SA), a variety of integral and frequency-related IMs is included, such as the Housner Intensity (HI), the Cumulative Absolute Velocity (CAV), the Arias Intensity (IA), the IA-based duration (e.g., time interval between 5% and 95% of the total IA, Ds595). Furthermore, in the flat-file compilation, special care was given to the characterization of pulse-like waveforms according to the algorithm proposed by Shahi and Baker (2014), which are of particular interest in earthquake engineering applications owing to their increased damage potential. Fig. 1 – Overview of BB-SPEEDset: distribution of data with respect to M w -R jb (left) and Vs30 (right). In the M w -R jb plot, the NESS2 dataset is also shown for comparison. Adapted from Paolucci et al. (2021b). A wide set of consistency checks were made to validate the BB-SPEEDset for its future engineering use, by comparing the statistical distributions of different IMs, their attenuation with distance, the features of directional and impulsive near-source accelerograms, with those obtained from recorded ground motions. As an illustrative example, Fig. 2a shows the cumulative distribution function of selected IMs, namely, PGA and PGV (RotD50 component) as computed from the entire BB-SPEEDset flat-file, in comparison with those obtained from NESS dataset, for the same Mw-Rjb range of data. In Fig. 2b, the trend of the period of the impulsive waveform T p (according to Shahi and Baker, 2014) as a function of PGD/PGV, for NESS (left), and BB-SPEEDset (right), is shown. Closed-form analytical relationships between T p and PGD/PGV ratios are also shown in Fig. 2b for Ricker waveleft and double-impulse functions. The comparisons of Fig. 2 (along with other checks not shown here for brevity) point out that the IMs and ground motion features from BB-SPEEDset are consistent, on statistical basis, with those from near-source recorded dataset, as no systematic bias is found. Finally, to appreciate the level of detail of information stored in BB-SPEEDset, Fig. 3 shows the results which can be extracted by the dataset for a specific earthquake, namely the 2009