Session: Overall Prediction Accuracy and Simulation Validation for Real-World Applications [Poster]
Type: Poster
Date: 10/12/2023
Time: 07:00 AM
Room: Stanley Park Ballroom
A Framework for Incorporating 3D Simulation Data into Non-ergodic Ground Motion Models
We present a framework for incorporating results from 3D simulations into non-ergodic ground-motion models (NGMMs) that can be used in seismic hazard calculations and show results from two recent examples of incorporating results of 3D simulations into NGMMs for long-periods from the M9 project and CyberShake. The M9 simulations are for multiple realizations of a single scenario, whereas the CyberShake simulations are for a large number of different scenarios. The framework requires four modifications to ergodic GMMs to incorporate 3D simulation results: (1) modify the average basin scaling to be consistent the average basin scaling from 3D simulations; (2) use the varying coefficient model (VCM) to estimate the spatial distribution of the non-ergodic adjustment for the path and site effects in the region covered by the 3D velocity model; (3) reduce the aleatory variability to account for the more accurate modeling of site and path effects from the 3D simulations; and (4) estimate the epistemic uncertainty for non-ergodic terms. The currently available simulations can be used to address the first three items, but not the full epistemic uncertainty in non-ergodic terms. The epistemic uncertainty is due to the uncertainty in the 3D velocity model; however, 3D simulation results from multiple alternative 3D velocity models are not available. We know that the uncertainty is not zero, but without suites of 3D velocity structure to sample the uncertainty, we currently assume that the epistemic uncertainty due to the 3D velocity model is one-half of the standard deviation of the path+site terms as a reasonable value. An alternative would be to use comparisons of simulations with observe values as a proxy for uncertainty in the 3D velocity model. The proposed approach can also be used with empirical ground-motion data, which would allow a comparison of non-ergodic terms from simulations with non-ergodic terms from observed data. We expect that hybrid GMMs based on the combination of simulation data and empirical data will be the direction of future NGMM development.
Presenting Author: Chih-Hsuan Sung
Additional Authors
Chih-Hsuan Sung karensung@berkeley.edu University of California, Berkeley, Berkeley, California, United States Presenting Author
Corresponding Author
|
Norman A Abrahamson abrahamson@berkeley.edu University of California, Berkeley, Berkeley, California, United States |
Maxime Lacour maxlacour@berkeley.edu University of California, Berkeley, Berkeley, California, United States |
A Framework for Incorporating 3D Simulation Data into Non-ergodic Ground Motion Models
Category
Overall Prediction Accuracy and Simulation Validation For Real-World Applications
Description