The Mexico Earthquake of September 19, 1985—Estimation of Response Spectra in the Lake Bed Zone of the Valley of Mexico

We estimate pseudo acceleration response spectra with 5% damping (Sa) for the great Michoacán earthquake at 22 lake bed sites (in addition to five sites where the event was well recorded) in the valley of Mexico. The method used consists of (a) estimation of Fourier acceleration amplitude spectra (FAS) at the sites via observed FAS at Ciudad Universitaria (CU) and compiled spectral ratios of these sites with respect to CU obtained from recorded earthquakes, (b) estimation of the duration of strong ground motion during the Michoacán earthquake at the sites, and (c) application of random vibration theory to predict Sa. A constant duration of 71 sec at all sites is found to be satisfactory. The main uncertainties in estimated Sa come from (a) possible nonlinear behaviour of the clay during the great earthquake (the spectral ratios used in the computation are mostly for moderate earthquakes, when the clay behaves linearly), and (b) variations in site's spectral ratios due to randomness and/or its dependence on earthquake characteristics which are presently unknown. Assuming that the nonlinear behaviour was neglegible and the spectral ratios estimated from one or few recordings are representative, we find (a) Sa ≥ 600 gals over a wide area of the lake bed zone, (b) Sa ≥ 1g at 14 out of the 27 sites, (c) regions of high Sa in the center of the city and east of CU, and (d) a good correlation between severe damage and regions with Sa ≥ 800 gals. Although in general our results agree with reported effects of the earthquake, because of the uncertainties mentioned above direct use of the computed response spectra for code or design purposes is discouraged.