Non-Ergodic FAS Ground-Motion Model for California

Session: Forthcoming Updates of the USGS NSHMs: Hawaii, Conterminous U.S. and Alaska [Poster]
Type: Poster
Date: 4/28/2020
Time: 08:00 AM
Room: Ballroom
Description: 

A non-ergodic ground-motion model (GMM) for the effective Fourier amplitude spectrum (EAS) for California is developed. The GMM is developed for EAS rather than spectral acceleration (Sa) because the Sa scaling depends on the spectral shape making it more difficult to use ground-motion data from small magnitude events to constrain the path and site effects. The Bayless and Abrahamson (2019) EAS GMM (BA19) is used as the base GMM. In the non-ergodic GMM, the source and site constants, the inelastic attenuation, and the Vs30 scaling are modeled as spatially varying coefficients, whereas, the geometrical spreading (GS) and magnitude scaling are fixed to their BA19 (ergodic) values. The magnitude scaling was fixed because there are not enough large magnitude events to constrain region-specific large magnitude scaling. The GS was fixed to avoid possible over-saturation at close distances.

The GMM is developed as a varying coefficient model (VCM) using a Gaussian Process regression in which the coefficients depend on the coordinates of the source and the site. Using the VCM model leads to a 40% reduction in the total aleatory standard deviation compared to the BA19 GMM. Both the mean and epistemic uncertainty of the VCM coefficients are estimated. As the data for a geographic region becomes sparse, the mean values of the coefficients go to the BA19 values but with large epistemic uncertainty, so there is a smooth transition between the regions with data and regions without data.

For engineering applications, the EAS GMM will be converted to a GMM for Sa. The EAS GMM is used to compute the median EAS for a range of scenarios spanning the range of magnitudes, source and site coordinates, and site conditions of interest. For each scenario, the EAS is converted to Sa using random vibration theory. The data set of generated Sa values is then used to develop a nonergodic Sa GMM. In this approach, all of the extrapolation required in Sa GMM is constrained by the EAS model.

Presenting Author: Grigorios Lavrentiadis

Authors