Spatial filtering in ambient noise interferometry

Area of Science:

• Geophysics
• Seismology
• Array Signal Processing

Background:

• Empirical Green's function retrieval from noise cross-correlations is theoretically sound but practically challenging in oceanic environments due to non-diffuse noise conditions.
• Directional sources in the ocean can introduce significant bias into the retrieved Green's function, limiting its accuracy for seismic imaging and monitoring.

Purpose of the Study:

• To develop and validate matrix-based spatial filtering techniques for removing unwanted contributions in ambient noise cross-correlations.
• To address the limitations of diffuse noise assumptions in oceanic seismic studies by mitigating directional source effects.
• To enable robust passive seismic wavefield separation and noise rejection using array processing.

Main Methods:

• Design and application of matrix-based spatial filters using both matrix least-squares with low-rank approximation and convex optimization approaches.
• Development of filter design solutions tailored for linear and planar seismic arrays.
• Testing filters on real-world ocean-bottom seismic data from a linear array and simulated data from a planar array.

Main Results:

• Successfully attenuated discrete interferer contributions in real-data cross-correlations from a linear array.
• Achieved passive upgoing/downgoing wavefield separation with planar arrays using simulated data.
• Demonstrated efficient rejection of horizontally propagating noise in planar array simulations.
• Investigated the influence of array size and frequency band on the performance of filtered cross-correlations.

Conclusions:

• Matrix-based spatial filtering offers a robust method for improving empirical Green's function retrieval in realistic oceanic noise conditions.
• These filtering techniques are effective in mitigating directional source biases and separating wavefields passively.
• The developed methods enhance the reliability of seismic data analysis from ocean-based arrays.