Irregular acquisition geometry causes discontinuities in the appearance of surface wave events, and a large offset causes seismic records to appear as aliased surface waves. The conventional method of sampling data affects the accuracy of the dispersion spectrum and reduces the resolution of surface waves. At the same time, mode kissing of the low-velocity layer and inhomogeneous scatterers requires a high-resolution method for calculating surface wave dispersion. This study tested the use of the multiple signal classification (MUSIC) algorithm in 3D multichannel and aliased wavefield separation. Azimuthal MUSIC is a useful method to estimate the phase velocity spectrum of aliased surface wave data, and it represent the dispersion spectra of low-velocity and inhomogeneous models. The results of this study demonstrate that mode-kissing affects dispersion imaging, and inhomogeneous scatterers change the direction of surface-wave propagation. Surface waves generated from the new propagation directions are also dispersive. The scattered surface wave has a new dispersion pattern different to that of the entire record. Diagonal loading was introduced to improve the robustness of azimuthal MUSIC, and numerical experiments demonstrate the resultant effectiveness of imaging aliasing surface waves. A phase-matched filter was applied to the results of azimuthal MUSIC, and phase iterations were unwrapped in a fast and stable manner. Aliased surface waves and body waves were separated during this process. Overall, field data demonstrate that azimuthal MUSIC and phase-matched filters can successfully separate aliased surface waves.
DeepMUSIC: Multiple Signal Classification via Deep Learning
