Removal of coherent noise from electroseismic data

The electroseismic method offers the possibility of imaging thin (much smaller than the seismic wavelength) layers in the subsurface. Removal of coherent source-generated noise is an essential step in the processing of electroseismic data. We present a signal-noise separation technique that begins with the determination of signal and noise prediction error filters (PEF’s) from windows of the original data and uses these PEF’s in an iterative inversion for signal and noise models. The noise residual is weighted during the inversion to prevent the comparatively weaker (by about two orders of magnitude) signal from being obscured by the coherent noise. Application of this processing sequence to real data, in which synthetic signal arrivals are obscured, demonstrates the effectiveness of the technique.