Coherent noise attenuation methods for low-fold seismic data

Abstract

A shot interval equal to the group interval is required to satisfy the stack-array criterion for split spread seismic data. If the shot interval is larger than the group interval, much of the coherent noise energy may leak through to the stack. This is because coherent noise, although probably unaliased in the shot domain, is highly aliased in the CMP domain. Prestack velocity filtering may be used to attenuate coherent noise, but can result in signal distortion and smoothing effects. A method of using a wavenumber-filter (k-filter) in the shot domain to attenuate coherent noise is suitable for use with modern low-fold data, i.e., data shot with small group intervals, many receivers, but a large shot interval. The k-filter operates by notching energy at wavenumbers in the shot domain that correspond to alias peaks of the CMP array response. To overcome notch shifting caused by NMO correction, a shot domain NMO correction is applied before k-filtering the data. The k-filter design is based on the Remez exchange method. This method allows for different passband and stopband error levels. Signal distortion and smoothing are reduced by keeping the error level near k = 0 small, and by minimising the equivalent area in the f-k domain that is removed by the k-filter. A simple model was used to estimate the theoretical attenuation level of the k -filter. A level of approximately 13 dB was achieved with the model data. The k-filter was also applied to field data from the Surat Basin, Australia. The results were not as good as the model results, probably because of variations in coherent noise characteristics across CMP gathers. Coherent noise characteristics have to be uniform across the whole length of each CMP gather for the method to work well. However, the k-filter did reduce significantly the level of coherent noise energy on the stacked data.