Stack

Key Takeaways

• Full stack versus partial stacks (near stack, mid stack, far stack) are a critical distinction in modern seismic processing and interpretation because AVO (amplitude versus offset) analysis requires knowledge of how the reflection amplitude changes with the source-receiver offset, which is lost when all offsets are summed together in the full stack: the full stack produced by summing all available offsets provides the best signal-to-noise ratio for structural mapping but averages over the offset-dependent amplitude variation that is the key indicator of Poisson's ratio contrasts associated with gas sands and other fluid effects; partial stacks are produced by summing subsets of offsets corresponding to specific angle ranges (near angles of 0-15 degrees, mid angles of 15-30 degrees, far angles of 30-45 degrees), preserving the angle-dependent amplitude information while still providing the noise attenuation benefit of averaging multiple traces within each angle range; the near and far partial stacks are used in intercept-gradient (AVO) analysis where the reflection amplitude as a function of angle (or offset) contains the information about the Poisson's ratio of the reservoir, with the near stack amplitude representing the zero-offset or acoustic impedance response and the far-near difference representing the gradient or shear wave contribution; full-offset stacks and partial stacks are complementary products of the processing sequence and modern 3D seismic datasets routinely deliver all three partial stacks plus the full stack as standard deliverables for interpretation.
• NMO velocity and stack optimization require careful velocity analysis to determine the stacking velocity that best aligns the reflection hyperbolas before stacking, because incorrect NMO correction leaves a residual moveout on the traces that reduces the quality of the stack by causing partial destructive interference rather than full constructive interference at the stack: the semblance analysis (computing the coherence of the traces in a CMP gather as a function of the velocity applied to correct the moveout, and displaying the result as a semblance panel) identifies the velocity that maximizes the coherence of each reflection, which is the optimal stacking velocity for that reflection at that CMP; picking the semblance maxima to define a velocity function that varies with time (depth) and space (CMP location) provides the velocity model needed for NMO correction and stacking; errors in the velocity picking cause the stack to sum traces that are not perfectly aligned, broadening the stacked wavelet (reducing temporal resolution) and reducing the amplitude of the stack relative to the optimal result; in anisotropic media (shale-rich sedimentary sections with transverse isotropy), the NMO curve is not exactly hyperbolic and higher-order moveout corrections or anisotropic NMO equations are needed to correctly align the traces before stacking, particularly at far offsets where the non-hyperbolic residual is largest.
• Pre-stack depth migration versus post-stack migration is a fundamental processing design choice that determines whether the stacking of multiple traces is performed before or after the migration (the processing step that repositions dipping reflectors to their true subsurface positions and collapses diffractions to their source points): in post-stack migration (the historical standard), the CMP stack is produced first and then the stacked section is migrated, which is computationally efficient but assumes that the stacking velocity equals the migration velocity, an assumption that breaks down in areas of strong lateral velocity variation (beneath salt bodies, in thrust belts); in pre-stack migration, the individual CMP traces are migrated to the common image point before stacking, allowing the correct depth-migration velocity to be used for imaging and the correct NMO velocity to be used for stacking independently, providing significantly improved image quality in complex velocity environments at substantially greater computational cost; pre-stack depth migration has become the standard processing approach for deepwater and subsalt exploration where the velocity variation is too severe for post-stack migration to produce accurate images, and the resulting pre-stack depth-migrated volume is the starting point for the AVO analysis, inversion, and interpretation workflows that characterize modern seismic exploration.
• Noise attenuation versus signal preservation trade-offs in the stacking process arise from the fact that not all signals in a seismic gather that are coherent (not random noise) should be preserved in the stack: multiples (seismic energy that has bounced multiple times between reflectors or between the surface and reflectors) are coherent signals that appear in the gather at a different NMO velocity than primary reflections, and they are attenuated in the stack when the NMO correction for primary reflections is applied (muting the traces at offsets where the multiple is so severely stretched by the primary NMO correction that it is cut by the stretch mute), but some multiple energy survives the stack if the primary and multiple NMO velocities are similar at short offsets; ground roll (the high-amplitude, low-frequency surface waves generated by the seismic source that propagate horizontally along the surface) is a coherent signal that is attenuated by stacking when the source-receiver geometry is designed so that the ground roll appears at different apparent velocities than the reflection signal (because ground roll has a low velocity of 200-500 m/s while reflections have apparent velocities of 1,500-6,000 m/s at near offsets), allowing frequency-wavenumber (F-K) filtering to attenuate the ground roll before stacking without damaging the reflection signal; the design of the acquisition geometry (shot spacing, receiver spacing, maximum offset, array length) directly determines the effectiveness of the stacking process at attenuating both incoherent noise (fold determines random noise attenuation) and coherent noise (offset range and geometry determine multiple and ground roll attenuation).
• Time-lapse (4D) seismic stacking for reservoir monitoring applies the stacking concept to repeated surveys acquired over the same reservoir at different times during production, and the comparison of the stacks from different vintages reveals changes in fluid saturation and pressure caused by production: a producing reservoir with water flooding will show amplitude changes between the base survey stack (before production) and the monitor survey stack (after water injection has replaced oil in part of the reservoir), with the water-swept zones showing amplitude changes consistent with the acoustic impedance change from oil to water saturation; the quality of the 4D signal (the amplitude difference between vintages) depends critically on the repeatability of the acquisition and processing between the two surveys, because non-repeatable differences in geometry, ambient noise, weather conditions, and ocean currents (for marine surveys) produce spurious amplitude changes that are indistinguishable from genuine fluid substitution changes in the stack comparison; the 4D seismic normalized root-mean-square (NRMS) metric quantifies the repeatability of successive stacks as a percentage of the signal amplitude, with NRMS below 5-10% considered excellent repeatability for production monitoring and NRMS above 30-40% indicating that genuine 4D signals may be obscured by acquisition noise differences between vintages.