Bias of NGA-East GMMs and Site Amplification Models Relative to Supplemented Cena Ground Motion Database
Description:
In this study we investigate bias in the median ground motion model (GMM) and linear site amplification model created as part of the Next Generation Attenuation project for stable continental regions (NGA-East), and present a bias adjustment term that can be used in forward applications. Bias is anticipated because these models were not developed in an iterative manner and the ground motion data used to calibrate the GMMs were modified to a reference site condition using ergodic site amplification models appropriate for active tectonic regions. Moreover, the F760 term in the site amplification model (amplification of sites with VS30=760 m/s relative to the reference of 3000 m/s) carries significant parametric uncertainty in κ0 (high-frequency spectral decay parameter) due to the lack of empirical data at the reference site condition. Using an expanded version of the NGA-East database (Ramos-Sepulveda et al. 2023) in which potentially triggered events in the Texas-Oklahoma-Kansas region were excluded, we compute residuals from the median versions of the GMM and site amplification models. The residuals were partitioned using mixed-effects analysis to quantify bias, event terms, and within-event residuals. With over 1500 recordings with periods from 0.01 to 3 sec and over 900 recordings from 3 to 10 sec, we find period-dependent bias in 5% damped response spectral acceleration across a wide period range. Results depict a negative bias (indicating over-prediction) at short periods (0.01 to 0.5 sec) and null to positive bias at longer periods. We do not find strong regional trends in event terms, suggesting this bias is not isolated to one region and may be widespread throughout central and eastern North America (CENA). A bias correction factor is proposed for short-term utilization in applications including the National Seismic Hazard Model. In the future, modifications to the GMMs or the site amplification models will be developed to remove this bias.
Session: USGS National Seismic Hazard Models: 2023 and Beyond
Type: Oral
Date: 4/18/2023
Presentation Time: 02:00 PM (local time)
Presenting Author: Maria E. Ramos-Sepulveda
Student Presenter: Yes
Invited Presentation:
Authors
Maria Ramos-Sepulveda Presenting Author Corresponding Author mariaramos1@g.ucla.edu University of California, Los Angeles |
Grace Parker gparker@usgs.gov U.S. Geological Survey |
Tristan Buckreis tristanbuckreis@ucla.edu University of California, Los Angeles |
Morgan Moschetti mmoschetti@usgs.gov U.S. Geological Survey |
Eric Thompson emthompson@usgs.gov U.S. Geological Survey |
Meibai Li mli@utexas.edu University of Texas at Austin |
Okan Ilhan okan.ilhan.uiuc@gmail.com Ankara Yildirim Beyazit University |
Ellen Rathje e.rathje@mail.utexas.edu University of Texas at Austin |
Youssef Hashash hashash@illinois.edu University of Illinois Urbana-Champaign |
Jonathan P Stewart jstewart@seas.ucla.edu University of California, Los Angeles, Los Angeles, California, United States |
Bias of NGA-East GMMs and Site Amplification Models Relative to Supplemented Cena Ground Motion Database
Category
USGS National Seismic Hazard Models: 2023 and Beyond