GNGTS 2022 - Atti del 40° Convegno Nazionale

GNGTS 2022 Sessione 2.1 203 four magnitude classes M w 5.5 (M w = 5.4÷5.6), M w 6.0 (M w = 5.9÷6.1), M w 6.5 (M w = 6.4÷6.6), M w 7.0 class (M w = 6.9÷7.1). Note that the idea of this plot is not to give a ranking of the performance of GMMs, but to highlight any possible systematic misfit between GM models and observations of moderate-to-large magnitude earthquakes, in the near-source. AD03 and AMB96 models show a great consistency with respect to observed records at almost all magnitudes. This tendency may be explained by the great number of records of the calibration datasets of the GMMs, within the magnitude and distance ranges considered in this analysis. The overall positive values of δBe at M w 6.5 suggests a general tendency of GM models to underestimate data in this magnitude class. Conclusions. This analysis provided a quantitative evaluation of the possible misfit of some of the most used and recent GMMs against observations in a M w - R range dominating the hazard in the most seismically regions of Italy. By taking advantage of a newly published highly quality dataset such as NESS2, and by means of an extensive analysis of residuals and their components, our study provided some interesting insights into the predictive capability of models in such M w - R ranges, showing the good agreement as well of some of the oldest models with respect to records. We think that this kind of researchmay encourge the developing of newmodels constrained, as accurately as possible, in the near source of moderate to large magnitude earthquakes, typically poorly constrained by observations, taking possibly advantage of results from 3D physics-based numerical simulations, for the potential improvement of calibration datasets. Acknowledgements and funding. This work has been partially supported by swissnuclear within the research activity “Development of advanced numerical approaches for earthquake ground motion prediction”, in the framework of the Sigma2 project, and by the Department of Civil Protection within the ReLUIS project WP18 “Normative contributions related to seismic action”. The authors are sincerely grateful to Sreeram Reddy Kotha, Ezio Faccioli and Chiara Smerzini for their useful comments. Fig. 2 - Mean between event residuals ( δB e ) calculated for the selected set of GMMs. Results are shown for four magnitude classes separately: M w 5.5 (M w = 5.4÷5.6), M w 6.0 (M w = 5.9÷6.1), M w 6.5 (M w = 6.4÷6.6), M w 7.0 class (M w = 6.9÷7.1) and four different periods: 0s (light pink line), 0.5s (burgundy line), 1s (green line) and 2s (purple line). AD03 residuals for M w 5.5 class are shown in white, because these magnitudes are out of the range of applicability of the model. Positive mean δB e show an underprediction of observations, while negative values indicate an overprediction.