A comprehensive beamforming toolbox to characterise surface and body waves in three-component ambient noise wavefields

We give an overview of a new toolbox for easy and fast beamforming analysis of three-component ambient seismic noise and discuss examples from different seismic arrays to solve different application challenges. From only a couple of hours of array recordings, the beamformer provides estimates of surface wave dispersion curves, surface wave azimuthal anisotropy, frequency-dependent wavefield composition including surface and body waves, and the direction of arrival for different wave types and frequencies. The beamformer can be used with three-component arrays from the lab to the field scale, provided ambient noise is available in the corresponding frequency range. Compared to standard (single-component) beamforming analysis, our approach integrates all three components recorded at every seismometer. Considering the phase shifts across the components, it identifies wave-specific particle motion and hence discriminates different wave types on account of their polarisation. The new implementation of the beamformer does not use the cross-spectral density matrix of the data explicitly (as done, for example, by the MUSIC algorithm and Capon beamformer), which reduces computation times significantly and makes it feasible to compute beam responses for a full day of data recorded on 100s of stations on a standard laptop PC. The toolbox will be available on github for both MATLAB and Python.

In an example from Los Humeros geothermal field (Mexico) we show Rayleigh wave azimuthal anisotropy as a function of frequency, corresponding to varying fast directions as a function of depth. A good agreement between the observed anisotropy and stress data from well logs as well as geological information indicates that fast directions correlate with the orientation of major faults and dykes. Anisotropy analysis thus provides a means to assess fault properties at depth, giving information about potential secondary permeability – a vital parameter in deep geothermal plays. Beamforming analysis of noise recordings in the Groningen area (Netherlands) reveals dominant prograde motion in both fundamental and 1^st higher mode Rayleigh waves. This behaviour is indicative of a large impedance contrast between the very low shear-velocities in sedimentary basins and the underlying bedrock. The resolution of particle motion as a function of frequency allows us to observe the osculation frequency where fundamental and 1^st higher mode Rayleigh waves approach each other and both modes change particle motion from prograde to retrograde and vice versa. The osculation frequency can be used to estimate the depth of the major impedance contrast, that is, the depth of the sedimentary basin. While body wave observations must be interpreted with care, considering the resolution capabilities of the array with respect to the expected (larger) wavelengths, the examples show that body waves contribute to the ambient noise wavefield with varying degree as a function of frequency, challenging the assumption of surface wave dominance common in ambient noise studies. Overall, we demonstrate that our beamforming toolbox provides direct information about structural features as well as fundamental a-priori information on wavefield composition and source characteristics, valuable for further ambient noise methods.