Open AccessFrequency‐dependent phase coherence for noise suppression in seismic array data
Author(s) -
Schimmel M.,
Gallart J.
Publication year - 2007
Publication title -
journal of geophysical research: solid earth
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.67
H-Index - 298
eISSN - 2156-2202
pISSN - 0148-0227
DOI - 10.1029/2006jb004680
Subject(s) - slowness , coherence (philosophical gambling strategy) , seismometer , attenuation , filter (signal processing) , classification of discontinuities , acoustics , coherence time , instantaneous phase , microseism , noise (video) , computer science , optics , seismology , geology , physics , mathematics , artificial intelligence , mathematical analysis , quantum mechanics , image (mathematics) , computer vision
We introduce a coherence measure for seismic signal enhancement through incoherent noise attenuation. Our processing tool is designed for densely spaced arrays and identifies signals by their coherent appearance. The approach is based on the determination of the lateral phase coherence as function of distance, time, frequency, and slowness. The coherence is derived from the local phases of neighboring stations which we obtain from analytic signals through the S‐transform. The coherence is used to attenuate incoherent components in the time‐frequency representations of the seismic records. No waveforms are averaged in our approach to maintain local amplitude information. This way we construct a data‐adaptive filter which enhances coherent signals using the frequency‐dependent and slowness‐dependent phase coherence. We explain the method and show its abilities and limitations with theoretical test data. Furthermore, we have selected an ocean bottom seismometer (OBS) record section from NW‐Spain and a teleseismic event registered at Spanish broadband stations to show the filter performance on real array data. Incoherent noise has been attenuated in all cases to enable a less ambiguous signal detection. In our last example, the filter also reveals weak conversions/reflections at the 410‐km and 660‐km discontinuities which are hardly visible in the unfiltered input data.