Time series modelling of strong ground motion from multiple event earthquakes

Abstract

Large earthquakes are often made up of several subevents. Thus the cumulative damage is higher than for single event earthquakes. Many procedures have been developed to simulate earthquake ground motion occurring from a single energy release; however, procedures to model accelerograms with several periods of strong shaking and to relate the modelling parameters to physical variables have not been developed. In this research, a database of strong motion accelerograms from multiple event earthquakes (including the 1978 Miyagiken-Oki earthquake) the 1968 Tokachi-Oki earthquake, the 1983 Nihonkai-Chubu earthquake, and the 1985 Michoacan earthquake) was modelled by an ARMA process, after first processing the records with multivariate variance and frequency stabilizing transformations. The modelling parameters were related to the time, magnitude and location of each subevent and to the site conditions using a regression analysis.