Bayesian hierarchical model of peak ground acceleration for the icelandic strong-motion arrays

Abstract

A reliable estimation of regional ground motion plays a critical role in probabilistic seismic hazard analysis (PSHA). The earthquake resistant design of structures within a region of a small spatial scale is often based on the assumption of relatively uniform form-factors which leads to the assumption of same station condition. However, for some small-scale regions this may not be the case. In this study, we propose a new Bayesian Hierarchical Model (BHM) for peak ground acceleration (PGA) records from two small-aperture Icelandic strong-motion arrays. The proposed BHM characterizes source effect, local station effect, source-station effect, and an error term that represents the measurement error and other unaccounted factors, separately. Posterior inference is based on a Markov chain Monte Carlo algorithm that uses the Metropolis algorithm. Uncertainty in unknown parameters is assessed through their joint posterior density. Analysis of PGA records based on the proposed BHM will improve the comprehensive understanding of the source effects, localized station conditions, and wave propagation.