An Improved Method to Extract Very-Broadband Empirical Green's Functions from Ambient Seismic Noise

The extraction of Green's functions by cross correlation of continuous seismic records at station pairs is best achieved in a diffuse wave field, where energies radiated by random sources have equal power. To partially satisfy the diffuse wave-field condition, original seismic records must be normalized as they are highly nonstationary and may have amplitude variations of orders of magnitude within the cross-correlation time window. We adapt a frequency-time normalization method to obtain seismograms with an even spectrum at all times within the data-processing unit. Compared with the commonly used one-bit normalization, the new method improves the signal-to-noise ratio of empirical Green's functions by a factor of similar to 2 and increases the effective data-recording duration by a factor of 4, assuming random local and instrument noise. It yields useful Rayleigh waves at periods up to 300 s for Program for Array Seismic Studies of the Continental Lithosphere (PASSCAL)-type deployments and 600 s for permanent stations with very-broadband sensors. Thus, the new method makes it possible to extend surface-wave signals to frequencies beyond those in traditional earthquake-based surface-wave tomography at both the high- and low-frequency ends.