Rayleigh wave three-component beamforming: signed ellipticity assessment from high-resolution frequency-wavenumber processing of ambient vibration arrays. Issue 1 (16th July 2018)

• Record Type: Journal Article
• Authors: Wathelet, M; Guillier, B; Roux, P; Cornou, C; Ohrnberger, M
• Is Part Of: Geophysical journal international. Volume 215:Issue 1(2018:Oct.)
• Page Start: 507
• Page End: 523
• Publication Date: 2018-07-16
• Subjects: Fourier analysis -- Time-series analysis -- Site effects -- Surface waves and free oscillations -- Wave propagation; Geophysics -- Periodicals; 550
• DOI: 10.1093/gji/ggy286
• Languages: English
• ISSNs: 0956-540X

Abstract

SUMMARY: The variation of Rayleigh ellipticity versus frequency is gaining popularity in site characterization. It becomes a necessary observable to complement dispersion curves when inverting shear wave velocity profiles. Various methods have been proposed so far to extract polarization from ambient vibrations recorded on a single three-component station or with an array of three-component sensors. If only absolute values were recovered 10 yr ago, new array-based techniques were recently proposed with enhanced efficiencies providing also the ellipticity sign. With array processing, higher-order modes are often detected even in the ellipticity domain. We suggest to explore the properties of a high-resolution beamforming where radial and vertical components are explicitly included. If N is the number of three-component sensors, 2 N × 2 N cross-spectral density matrices are calculated for all presumed directions of propagation. They are built with N radial and N vertical channels. As a first approach, steering vectors are designed to fit with Rayleigh wave properties: the phase shift between radial and vertical components is either $-{\Pi/2}$ or ${\Pi/2}$ . We show that neglecting the ellipticity tilt due to attenuation has only minor effects on the results. Additionally, we prove analytically that it is possible to retrieve the ellipticity value from the usual maximization of the high-resolution beam power. The method is tested on synthetic data sets and on experimental data. Both are reference sites already analysed by several authors. A detailed comparison with previous results on these cases is provided.