Empirical Map-Based Models of Nonergodic Site Response in the Greater Los Angeles Area

Grace A. Parker, & Annemarie S. Baltay

Submitted July 28, 2020, SCEC Contribution #10215, 2020 SCEC Annual Meeting Poster #TBD

We develop empirical estimates of nonergodic site response factors at seismic stations in the greater Los Angeles area using a combined dataset of recorded ground motions from 473 M 3-7.3 earthquakes in Southern California. The data are a combination of those from the NGA-West2 project (Ancheta et al. 2014), the 2019 Ridgecrest earthquake sequence (Rekoske et al. 2020), and more than 10,000 newly-processed records. We estimate site response using an iterative mixed-effects approach that takes into account azimuthal variations in anelastic attenuation and potential bias due to large spatial clusters of co-located earthquake sources. This process yields site response factors for peak ground acceleration (PGA), peak ground velocity, and pseudo-spectral acceleration (PSA) relative to a 760m/s reference condition as defined by the Boore et al. (2014; BSSA14) ground motion model. We employ regression kriging to interpolate the nonergodic site response factors to a densely-spaced grid between stations in the greater Los Angeles area. We use the linear site and basin terms from BSSA14 as the background model, with inputs of VS30 from Thompson (2018) and Z1.0 from Lee et al. (2014). We find our interpolated model resolution ranges from about 20 arcsec for PGA to about 38 arcsec for 10s PSA. We validate the model using site response calculated from four events at densely-spaced stations in the Community Seismic Network (Clayton et al. 2015; CSN) that were not considered in model development. We find good agreement between the interpolated model and CSN data, especially for long periods (R2 = 0.74 for 5.0 s PSA). We demonstrate two applications of our model, for the USGS ShakeAlert Earthquake Early Warning product, and the USGS National Seismic Hazard Maps. We find utilization of the model in ShakeAlert for the 2008 M5.4 Chino Hills and 2014 M5.1 La Habra earthquakes increases the accuracy of alerts. There is no substantial increase in computation time, as the model is formatted as a spatial lookup table similar to the current VS30-based table. Incorporating the model into nonergodic probabilistic seismic hazard analyses allows for a reduction of aleatory variability to the single-station sigma value and for the use of spatially varying epistemic uncertainty, thus improving hazard estimates especially at long return periods.

Parker, G. A., & Baltay, A. S. (2020, 07). Empirical Map-Based Models of Nonergodic Site Response in the Greater Los Angeles Area. Poster Presentation at 2020 SCEC Annual Meeting.

Related Projects & Working Groups
Ground Motions