Ground roll dispersion analysis at Springbank, Alberta

David G. Schieck, Robert R. Stewart

A three-component shallow reflection seismic line from Springbank, Alberta demonstrates large amplitude, surface wave noise due to the use of a surface energy source. Changes in phase velocity versus frequency, referred to as dispersion, of surface waves is primarily dependent on the shear wave velocities and thicknesses of the wave channels.

Shear wave refraction first breaks are picked from the radial or in-line shear wave shot records. These picks are inverted with for near-surface shear wave lithology. This enabled a dispersion model to be calculated which is used to identify and interpret surface wave noise on the shot record. By summing the -p wavefield transform of the vertical and radial component on a shot consistent basis the dispersion curve can be observed directly. The change in phase velocity with frequency can be approximated by a linear frequency modulated (LFM) wave over the frequency bandwidth of the Rayleigh waves.

An offset varying compression operator is cross-correlated with each trace in the shot gather before multi-channel filtering to remove linear dispersion. This reduces dispersed surface waves to a single aliased dip which is more easily removed by the 2-D median f-k filter. The output is then uncompressed by flipping the operator in time and again cross-correlating.

This hybrid multichannel filter is applied to a synthetic shot record contaminated with LFM noise at a bandwidth of 8-30hz with velocity bounds of 230- 800 m/s. The f-k domain representation of the synthetic shot record has a single focused dip of 520 m/s after LFM compression. This enables better removal of dispersive noise by the multichannel filters such as the velocity f-k or median f-k.

This new filter is applied to the real three-component data set. The results indicate that the best way to minimize surface wave noise in shallow reflection studies is through careful choice of the acquisition parameters to ensure the target events are within the optimum offset window. In this example, the ground roll filtering is unable to recover the underlying reflection signals within the surface wave noise. The best CDP stack obtained for the P-wave data is due to an inside or near offset mute to completely remove the Rayleigh waves from the shot records before stacking.