Attenuate: Seismic Noise Suppression, Multiple Removal, and Q-Based Wave Energy Loss

Key Takeaways

• Dual meaning in geophysics: processing action and physical phenomenon: In seismic processing, to attenuate means to suppress noise, multiples, or coherent interference in recorded data in order to preserve primary reflection energy with maximum fidelity. In rock physics and wave propagation, to attenuate means the physical loss of wave energy along the propagation path through geometrical spreading and intrinsic absorption. A seismic processor attenuates multiples using SRME or Radon transforms; the subsurface simultaneously attenuates the wave energy through Q-controlled absorption. Both senses of the word are used routinely in exploration and production technical documents, and the context (data processing versus physical earth) almost always clarifies which meaning is intended. Confusion between the two meanings can arise in discussions of gas reservoir identification, where low-Q physical attenuation produces an observable frequency-dependent amplitude signature in the data that might be confused with processing artifacts from imperfect multiple attenuation, requiring careful integration of processing quality control and rock physics interpretation to distinguish the two effects.
• Surface-related multiple elimination (SRME) for marine seismic multiples: Surface-related multiple elimination (SRME) is a data-driven prediction method that exploits the recorded seismic wavefield itself as the prediction operator for surface-related multiples. The algorithm exploits the fact that any surface-related multiple can be expressed as a cross-correlation of the full recorded wavefield with itself, predicting the multiple by convolving appropriate trace pairs. SRME requires no subsurface velocity model, making it robust in complex geological settings such as deepwater Gulf of Mexico and Norwegian North Sea where model-based methods fail due to lateral velocity variation. In favorable acquisition geometries with dense, regular receiver sampling, SRME can reduce water-bottom multiple energy by 20 to 40 dB (a factor of 100 to 10,000 reduction in energy) before adaptive subtraction further refines the result. Three-dimensional SRME (3D SRME) extends the method to account for the full azimuthal multiple geometry in wide-azimuth towed-streamer and ocean-bottom node surveys; the added computational cost of 3D versus 2D SRME is substantial but necessary for accurate multiple prediction beneath complex seafloor topography and in areas where out-of-plane multiple energy is significant.
• Radon-domain multiple attenuation and f-k filtering for ground roll: The Radon transform (parabolic Radon or tau-p Radon) attenuates multiples by exploiting the difference in moveout velocity between primaries and multiples. In the Radon domain, events with different apparent slownesses (reciprocal velocity) map to distinct parabolic trajectories; a mute or weighting function applied in tau-p space suppresses energy traveling at the lower velocity characteristic of multiples before inverse transformation back to the offset-time domain. Radon attenuation is most effective for short-period interbed multiples that SRME does not fully predict, and it is routinely applied as a second pass after SRME to address residual multiple energy. For land seismic data, ground roll (Rayleigh surface waves with high amplitude, low frequency, and low velocity) is attenuated by frequency-wavenumber (f-k) filtering: in the f-k domain, ground roll maps to a low-velocity cone separable from primary reflections by muting below a velocity threshold. F-k filtering is a mandatory early step in processing seismic data from land environments including the Western Canadian Sedimentary Basin foothills, the Permian Basin, and Middle Eastern desert surveys where ground roll energy dominates near-offset traces and overwrites shallow reflections.
• Physical Q factor: gas sands (5 to 20), brine sands (30 to 80), and frequency-dependent decay: The seismic quality factor Q quantifies intrinsic (physical) attenuation as Q = 2 pi times (peak strain energy stored) divided by (energy dissipated per wave cycle), with low Q indicating high attenuation and rapid amplitude decay. Gas-saturated porous sands exhibit Q values as low as 5 to 20 because the compressibility contrast between gas and brine in the pore space drives vigorous local fluid flow (squirt flow) during wave-induced pressure oscillation, dissipating energy viscously. Brine-saturated sandstones have Q of 30 to 80; dry consolidated rocks and carbonates have Q of 80 to 200. The amplitude decay equation for intrinsic attenuation is A = A0 times e to the power of (negative pi times f times t divided by Q), where f is frequency (Hz) and t is two-way travel time (seconds). This exponential relationship means that higher frequencies are attenuated faster than lower frequencies: in a formation with Q = 50, the 100 Hz component of a wave is attenuated 10 times more than the 10 Hz component over the same travel time, progressively shifting the dominant frequency toward lower values with depth and degrading the vertical resolution available for thin-bed identification at deep targets.
• Inverse Q filtering (Q compensation) to restore lost bandwidth at depth: Inverse Q filtering, also called Q compensation or absorption compensation, is the seismic processing step that corrects for the frequency-dependent amplitude decay caused by intrinsic absorption, partially restoring the high-frequency content lost during propagation to improve resolution at depth. The algorithm applies a time-variant, frequency-dependent gain function that is the inverse of the forward absorption operator, boosting attenuated high frequencies in proportion to their estimated loss along the wave propagation path through the estimated Q model. Q models for the correction are derived from VSP spectral ratio analysis (comparing amplitude spectra at successive depth levels on the downgoing wave), from borehole array sonic Q measurements, or from surface seismic Q tomography. Stabilization is essential: any inverse filter also amplifies noise at high frequencies where signal-to-noise ratio is inherently low, so a water-level regularization or frequency-dependent gain cap prevents noise amplification beyond usable levels. In practice, Q compensation can improve the dominant frequency of deep events by 15 to 30 Hz, yielding sharper reflections and enabling identification of thin reservoir packages in the 10 to 30 metre range that would otherwise fall below tuning thickness on non-compensated data.