Multimode dispersion measurement of surface waves extracted by multicomponent ambient noise cross-correlation functions

Abstract

SUMMARY Cross-correlation functions of ambient seismic noise sometimes show multimode characteristics of surface waves, especially in observations in sedimentary areas and ocean areas. Multimode dispersion curves are useful for improving the depth resolution of subsurface imaging; nevertheless, measuring the multimode dispersion curves is not easy. Multimode interference of surface waves makes the cross-correlation functions complicated even without lateral heterogeneity of the subsurface structure, and the complex waveforms may result in unphysical dispersion measurement. We developed a method to determine multimode phase velocity dispersion curves based on the fitting of the synthetic cross-spectra to observed ones. The phase velocity in the synthetic cross-spectra is modelled as the function of a 1-D velocity structure, which achieves the measurement of physically realizable dispersion curves. The 1-D structures do not necessarily represent the Earth structure directly but act as model parameters of the dispersion curves within station pairs. The cross-spectral fitting has two steps, that is, array-based and single-pair fittings. The first step estimates the amplitude of each surface wave mode and the reference 1-D structure from the cross-spectral data within an array. The second step estimates the pair-dependent dispersion curves from the cross-spectra of a single station pair using the modal amplitudes and the reference structure estimated by the first step. The dispersion measurement based on the cross-spectral fitting can work even at short distances where the multimode inference is significant in the time-domain cross-correlation functions. We applied this method to synthetic and field data in seafloor observations. The synthetic and field applications show that the simultaneous use of multicomponent cross-correlation functions is effective to determine multimode dispersion curves. The multimode phase velocity dispersion curves in the ocean area are estimated stably even though the signal-to-noise ratio of cross-correlation functions is not high. The pair-dependent multimode dispersion curves estimated by the present method can serve as robust input data for high-resolution surface wave tomography.