Prediction of Ground-Motion Time-Series at an arbitrary location using Gaussian Process Interpolation: Application to the Ridgecrest Earthquake

Aidin Tamhidi, Nicolas Kuehn, Yousef Bozorgnia, Ertugrul Taciroglu, & Tadahiro Kishida

Published August 15, 2019, SCEC Contribution #9724, 2019 SCEC Annual Meeting Poster #247 (PDF)

Every seismic event is recorded only at a finite number of locations. In many engineering applications a variety of seismic intensity measures (IMs), typically peak ground acceleration or response spectrum, are required as input, that are estimated either through ground motion prediction equations (GMPEs) or spatial interpolations of their recorded values. To understand dynamic responses of the structures, estimation of the entire ground motion (GM) time-series at specific geographic locations is required as well. In the present study, we examine the accuracy of the Kriging, which is Gaussian Process (GP) interpolation method, by predicting GM time series at an arbitrary location using available spatially sparse records. To achieve a quantitative assessment and objective verification of this approach, we used simulated ground motion time series of the 1906 San Francisco earthquake. These time series are transformed into the frequency domain, and both the real and imaginary parts are interpolated from the neighboring stations at the desired locations for predicting the GM time-series. The GP interpolation for the simulated 1906 San Francisco earthquake results in the optimized various model hyperparameters such as length scale. Our current work is aimed at applying the developed and verified procedure on ground motion time-series recorded during the recent Ridgecrest 2019, M6.4 and M7.1 earthquakes. The specific objective is to predict the GM time-series at the location of Ridgecrest Regional hospital for which extensive structural damage was observed, but there is no recorded ground motion from the earthquakes mentioned above.

Citation
Tamhidi, A., Kuehn, N., Bozorgnia, Y., Taciroglu, E., & Kishida, T. (2019, 08). Prediction of Ground-Motion Time-Series at an arbitrary location using Gaussian Process Interpolation: Application to the Ridgecrest Earthquake. Poster Presentation at 2019 SCEC Annual Meeting.

Related Projects & Working Groups
Ridgecrest Earthquakes