SCEC Award Number 16246 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title EPISTEMIC UNCERTAINTIES IN GROUND MOTION PREDICTION FROM VIRTUAL EARTHQUAKES
Investigator(s)
Name Organization
Marine Denolle Harvard University
Other Participants
SCEC Priorities 6e, 6c, 6a SCEC Groups Seismology, GMP, GMSV
Report Due Date 03/15/2017 Date Report Submitted 03/08/2017
Project Abstract
 Our goal is to evaluate the epistemic uncertainties that arise from correcting the surface-to-surface ambient seismic noise impulse response for source depth and focal mechanism in the Virtual Earthquake Approach (VEA). Our strategy is to evaluate semi-analytically and numerically the uncertainties in PGD, PGV, PGA, and duration in a 1D laterally homogeneous medium and in a locally 1D medium. We have greatly progressed in the first task and are currently addressing the second task. In addition to the depth and mechanism correction, we realized that the calibration required to bring the ambient noise impulse response to the dimensions of a displacement surface-wave Green’s function may also bring bias and uncertainties in the prediction of ground motion. We address this problem using prediction and validation of ground motions from the M5.2 2016 Borrego Springs, California, Earthquake.
Intellectual Merit This project addresses key elements to ground motion prediction as it establishes epistemic uncertainty of the Virtual Earthquake Approach. Errors in our best knowledge of the elastic structure, used as a local 1D velocity profile, likely results in ground motion uncertainties. The uncertainties simply scale between velocity perturbation and PGD and there is a large increase in the uncertainties at high frequencies (or for PGV and PGA). Duration uncertainties remain small (<5%). However, by virtue of surface-wave dispersion, characterizing epistemic uncertainties due to the 1D approximation of the elastic structure nearby the source may not be established without a full numerical wave propagation.
In concurrence with a ground motion project that aims to cross-validate ground motion prediction methods (GMPEs, wave propagation simulations, and VEA), we are establishing another source of epistemic uncertainties for the VEA. The calibration of the amplitudes using a small earthquake, if performed as a single scalar over a broad frequency band, yields a frequency-dependent bias. We plan on establishing a new frequency-dependent calibration factor.
Together, the analysis of epistemic uncertainty will improve the reliability of our ground motion predictions.
Broader Impacts This research has involved strong collaborations between PI Denolle with another early career scientist (Pierre Boué, Université Grenoble-Alpes), and with other institutions (Gregory Beroza, Stanford University; Frank Vernon, University of California – San Diego; Naoshi Hirata, Tokyo University). The project also involves the senior thesis research of an undergraduate student (Leore Lavin, Harvard University). The research has been presented at the SCEC annual meeting, AGU meeting, and will be submitted for peer reviewing in two separate publications that are in preparation.
Exemplary Figure Figure 2: Ratios of peak ground motion (displacement PGD, velocity PGV, acceleration PGA) of the predictions over the observations. Top panels (a-c) for transverse component (Love waves), medium panels (d-f) for the radial component (Rayleigh waves), bottom panels (g-i) for the vertical component (Rayleigh waves).