Deghosting vs Broadband Marine Seismic
The ‘broadband’ era of towed streamer seismic began with dual-sensor streamers (GeoStreamer), was followed by a variety of acquisition and processing solutions using hydrophone-only streamers in various configurations (deep flat tow, variable-depth tow, etc.), and has more recently returned to multi-sensor streamers. Apart from the dual-sensor streamer solution, all other published ‘broadband’ streamer solutions have used a flat sea-surface assumption, and each attempts to remove the free-surface ghost effects whilst recovering a frequency spectrum that meets some pre-conceived idealised output. Most broadband seismic results also incorporate spectral recovery solutions that primarily claim to compensate for the inescapable effects of anelastic attenuation and dispersion during seismic wavefield propagation throughout the earth. As such, removal of the free-surface ghost effects (or ‘deghosting’) constitutes only part of a broadband seismic processing flow.
Figure 1 models the frequency spectra that are recovered during various stages of broadband signal processing, assuming that full source and receiver deghosting works perfectly, as does attenuation compensation. Note the rapid amplitude decay below about 6 Hz for all stages of processing that is due to the fundamental physics of air gun arrays, and how the ‘reinforcing’ effect of the ghosts at mid frequencies is reduced by deghosting. No impulsive source solution exists that can output significantly stronger ultra-low frequency amplitudes than air guns. Overall, the full deghosting and spectral recovery sequence in Figure 1 improves both low and high frequency content, but the amplitudes below about 6 Hz decay rapidly, and the high frequency amplitudes shown in this simple modelling make no consideration for the increased noise that inevitably affects seismic data.
Viscoelastic imaging, often referred to as ‘Q migration’, has emerged in recent years as a powerful means to attenuate higher frequency noise within the migration kernel whilst simultaneously correcting for dispersion effects on pre-stack data. The low frequency broadband seismic story is more controversial as different vendors claim different success with ‘deghosting’, often recovering uniform amplitudes down to about 2 Hz. A quick examination of Figure 1 suggests that substantial spectral shaping must accompany deghosting to achieve such an outcome.
Fig.1. Superimposed amplitude spectra for a standard air gun array output at various stages of signal processing. The pale blue curve shows the recorded signal that is affected by both source and receiver ghost effects. The green curve shows the result after deghosting. The dark blue curve shows the result after compensation for recording filter effects. The grey curve shows the result after compensation for attenuation effects: A significant benefit at higher frequencies. (Click here to view enlargement)
There is plenty of contemporary evidence that aggressive spectral shaping to enhance the cosmetic appearance of stack seismic images can facilitate more robust interpretation of different geological facies—provided that high frequency resolution is not lost below the stronger low frequency texture of broadband seismic images.
The message here is that only part of that outcome is due to deghosting; the only physics-based component of the processing flow. The rest is subjective manipulation of the signal content.
In his next article, Andrew will discuss whether these simple principles also apply to pre-stack quantitative interpretation.