Application of the conditional spectrum approach for the selection of ground motion time histories in vast areas: an example in the Lazio region

Selection of accelerometric time histories is a fundamental step in seismic microzonation studies as well as in structural and ground response analyses. In this study we apply the procedure of Mascandola et al. (2020) for record selection in vast areas to the Lazio region (Central Italy). Unlike the original approach, we apply a different unsupervised clustering algorithm to divide the study area into mesozones, defined as zones that are homogeneous in terms of seismic hazard. Moreover, we apply the conditional spectrum ( ) method for the selection of ground motion time histories. Concerning the zoning of the study area, two clustering algorithms are compared: K-means and S pectral clustering . We found that the latter provides a zonation that is more consistent with the spatial distribution of the seismic hazard as well as of hazard disaggregation and is therefore suggested for record selection in vast areas. For each mesozone, two conditional spectra (computed for two conditioning oscillator periods) are then defined and adopted as reference for the selection of real accelerometric records. The latter are selected from a large sample of accelerograms that include earthquakes having magnitude and distance consistent with the seismic scenarios controlling the hazard. The article ends with a comparison of different techniques for the selection of real acceleration time histories. We analyze the impact of different sets of accelerograms on the ground response of two soil profiles resonating at the same periods used for conditioning the spectra (i.e., 0.2s and 1.0s). The results confirm that the conditional spectrum approach is targeted at oscillatory systems with a well-specific vibration period, for which the conditional spectrum might provide more conservative ground motions than the standard uniform hazard spectrum approach. Conversely, the latter approach is preferable for dynamic analyses of systems where more than one response period is relevant. Furthermore, our findings show that dividing the area into mesozones and selecting accelerograms using a single reference spectrum for each mesozone produces results comparable to those obtained through conventional methods based on site-specific record selection. Therefore, our approach can significantly streamline the selection of accelerometric time histories, especially in the case of vast areas.