SCEC Award Number 16296 View PDF
Proposal Category Collaborative Proposal (Integration and Theory)
Proposal Title Stochastic Characterization of Crustal Structure for High-Frequency Ground Motion Prediction Using Dense-Array Observations
Investigator(s)
Name Organization
Greg Beroza Stanford University Nori Nakata Stanford University
Other Participants
SCEC Priorities 6c, 6d, 6e SCEC Groups Seismology, USR, GMP
Report Due Date 03/15/2017 Date Report Submitted 04/05/2017
Project Abstract
 Ground-motion prediction is a key component of seismic hazard analysis, which is typically carried out using ground-motion prediction equations. The standard deviation of the best-fit model characterizes the residual ground motion variability. These residuals are an admixture of random variability (aleatory un- certainty) and modeling error (epistemic uncertainty). When we estimate site-specific ground motion, the residuals can be reduced because effects from different paths and sites are not mixed. Recent devel- opment of dense, capable, and long-term seismometer networks allows us to estimate the site-, path-, and source-specific variability. In this study, we use a very dense array (Large-N array) at Long Beach, California, and demonstrate the potential to characterize the spatial variability of the vertical component of ground motion. Because the density of the array (2500 receivers with 100-m spacing), we can esti- mate very dense site information (i.e., detailed near-surface velocities) from the ambient seismic field, which we can use to estimate site amplification. The recording time is only a couple of months, but we observed more than 10 earthquakes, and hence we can estimate site-, path-, and source-specific varia- bility of ground motion. Since we do not have multiple events occurring in close proximity to one another, we cannot separate the source-location and source-parameter effects.
Intellectual Merit We estimate site- and path-specific variabilities using a very dense array at Long Beach, California and validate a vertical-component ground motion prediction equation (GMPE). We find that observed PGA’s correlate with hy- pocentral distances and Vs30, and the GMPE can correct these effects and reduce the random variability. However, the correction is limited (~5%), and hence to predict ground motion through this approach we need improved GMPEs. The very dense array provides an opportunity to estimate more site parameters, for example near-surface velocities in finer sampling than Vs30. From the detailed velocities, we can estimate the theoretical amplification. This type of information should be useful for improving GMPEs.
Broader Impacts Cheap and user-friendly seismic sensors are becoming widely available. These now include three-component sen- sors and dense arrays with such sensors are becoming common. Therefore, we have opportunities to validate and improve GMPEs at the sites where we are interested in using the technique proposed in this report. With structural image and improved GMPEs, we can estimate more accurate ground motion at the local scale.
Exemplary Figure Figure 4. Distribution of PGA without any correction (a), with correction of distance term (b), and correction of distance and near-surface velocity terms. The blue distributions on panels (b) and (c) are the same as the distribution shown in panel (a). Thick lines are the best-fit Gaussian distribution based on least-squares estimation and sigma indicates the standard deviation of the Gaussian distribution.