Comparison of Nonergodic Ground Motion Components from CyberShake and NGA-West2 Datasets in California

Xiaofeng Meng, Christine A. Goulet, Kevin R. Milner, Robert W. Graves, & Scott Callaghan

Under Review 2022, SCEC Contribution #12616

In this study, we compare the Southern California Earthquake Center CyberShake platform against the NGA-West2 empirical datasets. Because the CyberShake and empirical datasets cover very different magnitude ranges and site conditions, we develop ground motion models for CyberShake datasets to compare trends with empirical ground motion models, and decompose the residuals for further analysis. We apply mixed effects regression to four CyberShake datasets in Southern, Central and Northern California at 2s, 3s, 5s and 10s periods and compare the results with the empirical datasets using the same approach. CyberShake captures the total variability of ground motions in the empirical datasets, but tends to predict larger median ground motions relative to the empirical ground motion models. We then calculate and compare the repeatable source-specific location, site and path effects between CyberShake and empirical datasets. We find that the correlations of site effects between the CyberShake and empirical datasets are generally satisfactory, and the variability of site effects is slightly smaller for CyberShake datasets. There is no apparent correlation of source-specific location effects between the CyberShake and empirical datasets. Comparison of path effects show a wide range of correlation coefficients. Finally, we investigate the source of observed differences between the CyberShake and empirical datasets. We attribute the larger median ground motion levels in CyberShake to a combination of the homogeneous slip patterns of the source, the lack of nonlinear site response, the low resolution of near-surface materials in the velocity models, and strong reflections at high contrast boundaries in the velocity models. These factors also impact the correlations of site and path effects between the CyberShake and empirical datasets. Moreover, the leakage from location terms into site and path terms further weakens the correlations. In summary, we find that CyberShake needs improvement for several components, but is still very useful to supplement empirical datasets for ground motion studies, especially to inform their non-ergodic components.

Citation
Meng, X., Goulet, C. A., Milner, K. R., Graves, R. W., & Callaghan, S. (2022). Comparison of Nonergodic Ground Motion Components from CyberShake and NGA-West2 Datasets in California. Bulletin of the Seismological Society of America, (under review).