Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Geophysics

Supervisor

Dr. Kristy Tiampo

Abstract

In this study, different magnitude estimation methods were investigated for application to earthquake early warning (EEW) and tsunami early warning systems. This integrated study is divided into two main parts. First, I used strong motion accelerograms recorded by borehole and surface stations from the Kiban Kyoshin network (KiK-net) for Japanese earthquakes with moment magnitude (M) ≥ 5.0 in order to develop ground motion prediction equations (GMPEs). I developed new GMPEs for peak ground acceleration (PGA) and peak ground velocity (PGV) using two different catalogs. The first catalog included earthquakes with 5.0 ≤ M ≤ 8.1 from 1998-2010. In order to improve the determination of attenuation parameters and magnitude scaling, the second catalog included earthquakes with 5.0 ≤ M ≤ 9.0 from 1998-2011, which increased the time period by only one year but added approximately twice as much data to the first catalog. The GMPEs were used to estimate the magnitude from PGA values (Mpga) and from PGV values (Mpgv) for those events in the borehole and surface databases with at least 20 available records. The results confirmed that Mpga and Mpgv strongly correlate with the moment magnitude of the event. In addition, I studied the site effect terms in the GMPEs using the shear wave velocity in the uppermost 30 meters (VS30). It was found that correcting for VS30 improved the accuracy of magnitude estimates from surface recordings, particularly for Mpgv. Incorporation of this parameter into the GMPEs can provide a more accurate estimate of the earthquake magnitude in EEW systems. The GMPEs also were used to estimate the magnitude of the M9.0 Tohoku event and those estimates were compared with the magnitude estimates provided by the existing EEW system in Japan. I demonstrate that, unlike the estimates provided by the existing EEW system in Japan, the magnitude estimates from GMPEs do not saturate. The results demonstrate that Mpgv from borehole recordings had the smallest standard deviation among the estimated magnitudes and produced more stable and robust magnitude estimates. Based on this observation, I propose the incorporation of borehole recordings into EEW systems. This method can improve the existing EEW system in Japan or other regions that have a dense seismic network.

In the second part of this thesis, the displacement spectra of the strong ground motion recordings were used to directly estimate the magnitude of Japanese earthquakes with 4.5 ≤ M ≤ 9.0, 2000 to 2011, using the first available data provided by the KiK-net and Kyoshin network stations. The source parameters were determined using the inversion of displacement spectra for available P- and S-waves windows assuming the Brune source model. I tested the application of a fixed low-cut filter, and found that it decreases the accuracy of magnitude estimation for large events (M > 7.0). As a result, instead of a fixed low-cut filter I applied a frequency bandwidth cutoff based on a signal-to-noise ratio criterion. The results showed that magnitude estimation using the strong motion recordings from the closest station to the source of the event provides a good early estimate for the final size of the event, which can reduce the time required to calculate final magnitude and hence provides a longer warning time (from a few seconds to a few minutes). The results also indicated that the predicted magnitude based on the P-wave window (MP) provides a longer warning time, but with a larger uncertainty, in comparison to the estimation based on the S-wave window (MS). The magnitude estimate based on inversion of the displacement spectra is independent of magnitude scaling relationships, as is the case with magnitude vs. early P-wave parameter relationships or GMPEs, because it determines the moment magnitude from the estimated source parameters directly from the displacement spectra. Therefore, this method can be used in regions with sparse seismic networks where historic recordings of strong ground motion from potentially damaging earthquakes are not available to develop an empirical relationship, such as the Cascadia region of North America.

Download

Share

COinS