Modelling Earthquake Ground Motions in Seismically Active Regions Using Parametric Time Series Methods
G.W. Ellis and A.S. Cakmak
The application of the Autoregressive-Moving Average (ARMA) process is an efficient method to model strong motion accelerograms, after processing by a variance and frequency stabilizing transformation. This report presents two methods for modeling accelerograms. The first method was developed to model individual accelerogram components. From this modeling procedure, parameters describing the change in variance during the record, the change in the dominant frequency during this record, and the correlation structure of the stabilized series were estimated. This univariate procedure was used to calculate modeling parameters for 148 accelerogram components recorded in California. These parameters were related to physical variables, such as earthquake magnitude, epicentral distance, and site geology, allowing simulations to be generated for sites where no ground motion records are available. A second procedure was developed to model the three acceleration components together as a group. This multivariate procedure was used to calculate modeling parameters from accelerograms recorded in Mexico and Taiwan, with particular emphasis placed on the accelerograms recorded from the 1985 Michoacan earthquake. By relating the modeling parameters to physical parameters, it is possible to generate realistic three-dimensional simulations for sites in these regions.
Autoregressive-Moving Average (ARMA) Process, Accelerograms, Seismology, Seismic Modeling, and Earthquakes.