Improving ambient noise cross-correlations using a polarization-based azimuth filter

Abstract

Seismic interferometry is widely used to image and monitor earth structures at different scales by extracting an empirical Green’s function (EGF) from the cross-correlation of ambient noise under the assumption of diffuse wavefields or energy equipartitioning. The EGFs, however, may be incorrectly estimated and lead to a misunderstanding of subsurface structures, because the distribution of ambient noise sources is neither isotropic nor stationary. To extract reliable EGFs from the non-ideal ambient noise data, we propose a polarization-based azimuth filter (PAF) to attenuate the ambient noise energy of non-stationary sources to improve the quality of cross-correlation functions. The PAF is a time-frequency domain azimuth filter for three-component seismic data, which uses time-frequency polarization analysis to measure azimuths of ambient noise signals, and then filters the three-component data to obtain signals from desired directions. Based on the stationary-phase theory, we use the PAF to extract the ambient noise signals from the stationary-phase zones (SPZs), which improves the quality of the cross-correlation function and gets rid of the requirement for long observations. A synthetic test and two short-time engineering examples demonstrate that uneven source distributions might cause serious spurious signals in EGFs and we show the improvement in EGFs using our method. In addition, the PAF may allow for the complete separation of the fundamental and higher mode of Rayleigh waves, which uncovers more information implicit in the ambient noise data.