A re-evaluation of scaling techniques for natural records

The study focuses on the problem of reducing scatter in the response calculated from time history analysis using natural records by proper scaling of these records. Two ground motion data sets from two different seismotectonic environments are used, each one evenly distributed with respect to rock and alluvial sites. The first part of the study focuses on the effect of scaling on elastic and inelastic spectra for strength and displacement. It is found that, in the intermediate and long period range, any of the three velocity-related parameters studied are appropriate to use. In the case of inelastic spectra, use of SI scaling along the entire period range leads to a reasonably low amount of scatter in both strength and displacement spectra, with COV hardly exceeding 0.5 within any interval of practical interest. The problem is then addressed in the context of response variability in realistic multidegree-of-freedom systems, with a focus on multistorey frames. Narrow-band spectrum intensities calculated on the basis of elastic or inelastic pseudo-velocity spectra are suggested as an alternative method of scaling in structure-specific studies. Both proposals based on the narrow-band elastic SI concept lead to COV values between 10 and 40% in the calculated drifts and member ductilities, as well as a reasonably uniform distribution of scatter along the building height.