A closer look at hazard-consistent ground motion record selection for building-specific risk assessment: Effect of soil characteristics and accelerograms’ scaling

Abstract

Conditional spectrum (CS) record selection is a standard tool for the selection of ground motion recordings consistent with the hazard at a site of interest. For many important applications the seismic hazard is computed for reference rock conditions and the effects of local site conditions are investigated downstream via soil dynamics. Ideally, when selecting for rock conditions, the recordings should be real, recorded on rock sites, and strong enough to push the structure into its severe nonlinear range. However, rock-recorded ground motions are often scarce and the analyst is usually left with the suboptimal option of augmenting the set with scaled motions and motions recorded at soil stations. We investigated how this practice can affect the structural response estimates by means of nonlinear dynamic analyses of a comprehensive set of single-degree-of-freedom (SDOF) systems. We first performed site and structure-specific CS record selection using ground motion databases of only soil-recorded motions versus databases of only rock recorded ones. In addition, we performed the CS based on extreme cases of using ground motions scaled by either only small or only large factors. Our focus was on the statistical comparison of the common intensity measures (IMs) of the selected records and, more importantly, of the structural response estimates when these extreme opposite selection choices are made. We show no significant statistical differences in the distributions of the IMs and structural responses, while the fragility and response hazard remain almost identical, for all practical purposes. These results imply that, as long as hazard consistency is accurately enforced in the ground motion record selection, using CS, selecting soil-recorded motions or record scaled up to 10, will not bias the response estimates. Finally, repeating the above procedure for response prediction of one multiple degrees of freedom (MDOF) system suggests that our conclusions are applicable also to more complex structural systems.