Group A, Poster #167, Ground Motions

Calibration of near-surface velocities in the greater Los Angeles area

Te-Yang Yeh, & Kim B. Olsen
Poster Image: 

Poster Presentation

2023 SCEC Annual Meeting, Poster #167, SCEC Contribution #13164 VIEW PDF
The near-surface seismic structure (surface to about 1000 m depth), in particular the shear-wave speeds (Vs), has important implications for prediction of strong ground motions. In the SCEC CVM version CVM-S4.26.M01 (CVM-S hereafter), the shallow Vs values are unrealistically high outside the major basins in Southern California due to the lack of geotechnical constraints, resulting in significant underprediction of ground motions. Hu et al. (2022) showed using 3D simulations of the 2014 Mw 5.1 La Habra earthquake that a geotechnical layer (GTL) using a Vs30-based generic overlay method to an extended depth (700 - 1000 m) can effectively reduce the misfit between data and simulation. Here, we... build on these results to test the sensitivity of depth extent of GTL to waves from different azimuth. Specifically, we compare results from 3D finite difference simulations of 0 - 1 Hz wave propagation using AWP-ODC to seismic data from for 7 well-recorded events with moment magnitudes between 4.4 and 5.4. Without implementing the GTL, our simulations of all 7 events reproduced the general underprediction trend at ‘rock’ sites (surface Vs > 1000 m/s) from Hu et al. (2022). With the GTL implemented for these sites, we found an average optimal range for the tapering depth of 500 - 700 m which generates mostly unbiased spectral fit to the observed amplitude spectra from all the events. We also find that a spatially-variable GTL taper depth may improve the fit to data.

We also used the 7 events to test whether the high-resolution Vs model of the San Gabriel, San Bernardino, and Chino Basins by Li et al. (2023) obtained from ambient noise tomography improves the fit to 0-1 Hz waveforms for the La Habra event. Vp and density values were obtained from the empirical relations by Brocher (2008), and the model was incorporated into CVM-S using a cosine-taper method proposed by Ajala and Persaud (2021). We find up to 50% improvement in FAS bias at frequencies between 0.2-0.5 Hz for sites within the updated basin model, which introduces higher surface velocities and more complexity, as compared to CVM-S. The improved fit to data leaves the possibility that the model update may have altered the waveguide effects due to interconnected basins along the northern edge of the greater Los Angeles basins (e.g., Olsen et al., 2008, 2009; Graves et al., 2008).

SHOW MORE