Statistical Analysis of Ground Motions Estimated on the Basis of a Recipe for Strong-motion Prediction: Approach to Quantitative Evaluation of Average and Standard Deviation of Ground Motion Distribution

Abstract

For seismic hazard assessment, we study the variabilities of predicted ground motion on the basis of a “recipe for predicting strong ground motion” and propose approximations to evaluate spatial distributions of the standard deviation for PGV, R1.0, R2.0, and R5.0 in the estimated ground motions. For strong-motion prediction, we use a finite difference method for a long period range (>1.0 s). To estimate variabilities, a Monte Carlo simulation is used and we adopt the Latin Hypercube Sampling (LHS) technique to reduce computations. In this article, we consider only aleatory variabilities in source parameters among all possible variabilities, such as those in the source parameters, the propagation characteristics and site characteristics. Model sources are assumed for dip-slip fault and strike-slip fault, and the variabilities are considered for parameters such as asperity location, rupture starting point, average asperity slip contrast, stress drop and rupture velocity. On the target site, 100 instances of PGV, R1.0, R2.0 and R5.0 data are obtained for 100 sets of parameters and an average and a standard deviation of the log normal distribution, corresponding to the variability for ground motion estimation, are statistically analyzed. For all target sites uniformly distributed in the area around the faults, the average and the standard deviation are statistically analyzed and spread to spatial maps. It is found that the spatial distributions of standard deviation values for both the dip-slip and strike-slip faults are not uniform. Approximations are attempted to develop a quantitative evaluation for spatial distributions of the standard deviation of the log normal distribution for PGV, R1.0, R2.0, and R5.0. The spatial distributions by these approximations are considered to almost reconstruct the characteristics, which are statistically analyzed by the finite difference method.