On the Selection of Ground-Motion Prediction Equations for Seismic Hazard Assessment in the South Iceland Seismic Zone

Abstract eng:
Ground-motion prediction equations (GMPEs) which quantify the attenuation of key earthquake strong-motion parameters with distance, have a major impact on seismic hazard analysis. In GMPEs, the random variability of amplitudes about a median prediction equation is considered as aleatory uncertainty whereas the uncertainty concerning the correct value of the median is epistemic. Epistemic uncertainties which arise from lack of knowledge about models and data should be considered in seismic hazard assessment to reach a reliable result for the region under study. Data-driven model selection would decrease epistemic uncertainties by reducing subjectivity and by guiding the selection process in a quantitative way. In this study, we review the likelihood-based (LH and LLH) and the Euclidean distance-based ranking (EDR) methods, then we introduce a new procedure on the basis of deviance information criterion (DIC), for selecting the proper GMPEs. To showcase the method, eight candidate GMPE models are ranked by LLH, EDR and DIC. The method is not only shown to optimize the selection of GMPEs for the given region in an unbiased way through the Bayesian statistics, but also solves the problem associated with the previous data-driven methods.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket