Common-midpoint spatial autocorrelation analysis of seismic ambient noise obtained from a spatially unaliased sensor distribution

Abstract

We have introduced a passive surface-wave method using seismic ambient noise obtained from dozens of receivers forming spatially unaliased 2D arrays. The method delineates 2D or 3D S-wave velocity ([Formula: see text]) models to depths of several hundreds of meters, without using any sources. Typical data acquisition uses 50–100 vertical-component 2 Hz geophones on the surface with 5–30 m receiver spacing. Cableless seismographs with GPS record 20–60 min of ambient noise. We establish a 2D grid covering the investigation area and use a common-midpoint spatial autocorrelation (CMP-SPAC) method to calculate phase velocities, resulting in a dispersion curve for each grid point. The method provides dozens of ispersion curves in the investigation area. We use a 1D nonlinear inversion to estimate a 1D [Formula: see text] profile for each grid point, and then we construct pseudo-2D or pseudo-3D [Formula: see text] models from the 1D [Formula: see text] profiles. The precision and accuracy of the CMP-SPAC method were tested with a numerical simulation using a 3D finite-difference method. The results of the simulation demonstrated the applicability of the method to complex velocity structures. We applied the method to an active fault investigation in China. Sixty-four cableless seismographs were deployed in an investigation area of 330 × 660 m (217,800 m2) with 5 and 30 m receiver spacings for dense and sparse grids, respectively. A 3D [Formula: see text] model was obtained to a depth of 150 m from CMP-SPAC analysis. The resultant 3D [Formula: see text] model indicates approximately 50 m of vertical displacement on a known fault.