COMPARISON OF METHODS TO PREDICT RESPONSE SPECTRA AT INSTRUMENTED SITES GIVEN THE MAGNITUDE AND DISTANCE OF AN EARTHQUAKE

Abstract

We compare four methods, all based on Bayesian regressions, to predict response spectra at instrumented sites, located at the hill and lakebed zones in Mexico City, that have recorded several recent earthquakes. The regressions are built as functions of magnitude and closest distance to the rupture area of the earthquake, using more than 20 subduction earthquakes (thrust-faulting mechanism) recorded since 1964. The four methods are: (1) a regression on observed response spectra; (2) a regression on observed Fourier amplitude spectra and use of random vibration theory to estimate the response spectra, (3) a regression on observed response spectra at the reference hill-zone station multiplied by response spectral ratios; and (4) a regression on Fourier amplitude spectra at the reference hill zone, multiplied by Fourier spectral ratios and use of the random vibration theory to estimate response spectra.