GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 2.1 GNGTS 2024 Figure 2: Residual for the horizontal DSR(5-75) for the Three Models: a) Bommer et al. 2009, b) Afshari and Stewart 2016, and c) Tafreshi and Bora 2023 The residuals were subsequently decomposed by applying the random effects approach (Al Atik et al. 2010) into offset (a0), between event ( δ Be), site-to-site ( δ S2Ss), and event-and-site corrected ( δ W0es) terms to check that the scaling with magnitude, distance, and V S30 (which are the main explanatory variables) of the considered models are compatible with those observed for ITA18. The results showed that all the models did not capture the magnitude scaling of ITA18 data: Bommer et al. (2009) and Tafreshi and Bora (2023) have similar trends, underestimating the low magnitude durations and overestimating the high magnitude ones. Afshari & Stewart (2016) GMM has a better median performance, but the duration observed for high magnitude earthquakes are underestimated (Figure 3-a,d & g). The other residuals did not exhibit significant bias, except for the event and site corrected residuals of Tafreshi and Bora (2023), which showed a different scaling with distance, probably because the model mimics the regional attenuation characteristics of Iran, for which the model was calibrated.