To Scale or Not to Scale Seismic Ground-Acceleration Records

Current estimates of seismic structural fragilities are commonly made on the basis of finite collections of actual or virtual ground-acceleration records that are scaled to have the same scalar intensity measure, for example, peak ground acceleration or pseudospectral acceleration. This paper models seismic ground-acceleration records by samples of Gaussian processes X(t) and constructs scaled versions X˜(t) of X(t) by following current procedures. This analysis shows that X˜(t) and X(t) have different probability laws, so that fragilities on the basis of X˜(t) provide limited if any information on the seismic performance of structural systems, that is, fragilities on the basis of X(t). The usefulness of current fragility estimates on the basis of scaled seismic ground-acceleration records is questionable, and scaling ground motions is not recommended.