Spectral matrix analysis for detection of polarized wave arrivals and its application to seismic reflection studies using local earthquake data

Abstract

Local earthquakes observed at Sendai, Japan, were analyzed to confirm the validity of a method of polarization analysis using the spectral matrix of seismic wave and its application to seismic reflection studies of the crust using local earthquake data. Reflectors (Bright spots) are known below the Nagamachi-Rifu fault, which caused an M 5.0 class event in 1998. Polarization analysis was applied to earthquake data in and around the fault. Use of the Z -parameter, which is defined using the eigenvalues of the spectral matrix and a statistical value representing the confidence level for the detection of the arrival of polarized waves, allowed detection of linearly and elliptically polarized waves in coda waves. The Z -parameter was also used to image the reflectors by using a migration technique that assumes the P × P and S × S reflection waves travel through a multi-layered velocity structure. Distinct reflectors were detected at depths of around 10 km, 14 km, 17 km, 21-26 km, 35 km and 40 km, that is, from deeper than the fault and the Moho. This study demonstrated the feasibility of using the spectral matrix of three-component seismic signal to detect polarized waves and to image reflectors in the earth's crust and upper mantle.

Optimization and Quality Assessment of Arrival Time Picking for Downhole Microseismic Events

“…A characteristic function is generally constructed from the single receiver recording in single-level algorithms, and the maximum of the characteristic difference is chosen as the arrival time of the microseismic signal. The short-and long-time average ratio method (STA/LTA) [11,12], Akaike information criterion method (AIC) [13], polarization-based method [14], higher-order statistics such as skewness and the kurtosisbased method [15,16], and the time-frequency analysis method [17] are commonly used single-level methods. Hybrid algorithms have also been proposed to achieve more accurate and precise arrival-time results for low signal-to-noise ratio (S/N) events by combining information from one or more individual picking algorithms [18,19].…”