Abnormal Events: Definition, Seismic Noise, and Data Interpretation

In reflection seismic acquisition and processing, abnormal events are any coherent or incoherent signals recorded on a seismic trace that are not primary reflections from subsurface horizons. The term encompasses diffractions, multiples, refractions, surface waves, guided waves, direct arrivals, air waves, and various forms of ambient or coherent noise. Despite the label suggesting they are rare, abnormal events routinely dominate raw seismic records and constitute one of the primary challenges in producing a clean, interpretable seismic image. A thorough understanding of each event type, its physical origin, its characteristic move-out, and the processing tools designed to suppress it is essential for any geoscientist or geophysicist working with seismic data.

How Abnormal Events Are Generated

Every seismic source, whether a marine air gun array or a land vibroseis fleet, generates a wavefield that travels simultaneously in multiple directions. The dominant downgoing wavefield illuminates subsurface horizons and returns to the surface as primary reflections, which carry the geological information interpreters seek. However, a significant portion of the source energy travels along alternative paths, interacts with boundaries in ways that do not follow the simple reflection law, or undergoes multiple reflections before returning to the receiver array. These alternative pathways generate the full suite of abnormal events recorded on every seismic trace.

Diffractions originate wherever the subsurface contains a geometric discontinuity, such as a fault tip, the edge of a reef, a channel margin, or an angular unconformity. When the downgoing wavefield strikes a sharp edge, energy is scattered in all directions according to Huygens' principle. On a seismic section, this appears as a characteristic hyperbolic tail, with the apex located at the scatterer and the limbs fanning outward with increasing offset from the apex. In unmigrated data, fault-plane diffractions frequently interfere with adjacent reflectors and blur structural interpretation. Refractions, by contrast, travel as head waves along high-velocity boundaries, typically the base of the weathering layer or a hard intra-formation contact, and arrive at receivers before the direct wave beyond a crossover offset defined by the velocity contrast. Surface waves, commonly called ground roll in land seismic, are low-frequency, low-velocity dispersive waves that propagate along the free surface with very high amplitude and can mask reflections across a broad frequency band.

Multiples are reflections that have bounced more than once before being recorded. Surface multiples bounce off the free surface or the sea floor; interbed multiples reflect between two subsurface interfaces without involving the free surface; and peg-leg multiples involve one downward reflection from the surface and a second from a subsurface interface. In deepwater marine surveys, the ocean-bottom multiple arrives shortly after the deepest target reflections and is often the single largest coherent noise source on the record. In carbonate sequences on land, interbed multiples generated by the high-impedance contrast between salt or anhydrite layers and surrounding carbonates can overwhelm primary reflections at target depths.

Classification of Abnormal Event Types

A systematic classification helps processors choose the correct suppression strategy for each event type. Diffractions are kinematically distinct from reflections: both have hyperbolic move-out in offset-time space, but a diffraction's move-out velocity equals the true medium velocity at the scatterer, whereas a reflection's normal-moveout (NMO) velocity is a root-mean-square average of interval velocities. This difference allows migration to collapse diffractions to their apex points, converting what appeared as noise into structural information. In practice, pre-stack depth migration achieves this collapse most accurately when a reliable velocity model derived from full-waveform inversion (FWI) or tomography is available.

Refractions and head waves travel with the apparent velocity of the refracting layer and exhibit linear move-out on shot gathers rather than hyperbolic move-out. In refraction statics workflows, this linearity is exploited to estimate near-surface velocity models and compute static corrections that improve the coherence of reflections in stacked sections. In the absence of refraction statics, weathering-layer velocity variations create residual statics errors that degrade stack quality across the entire survey area. Surface waves on land records are characterized by low group velocities, typically 200 to 800 metres per second (650 to 2,600 ft/s), low frequencies of 2 to 15 Hz, and very high amplitudes that can be 30 to 50 dB above primary reflections at near offsets. Their dispersive nature, where different frequencies travel at different velocities, produces the fanning move-out pattern visible on shot records. Frequency-wavenumber (f-k) filtering and tau-p (intercept time versus ray-parameter) transforms are classical tools for suppressing surface waves because these events occupy a narrow triangular region in the f-k domain that does not overlap significantly with primary reflection energy.

Multiple Suppression Techniques

Surface Related Multiple Elimination (SRME) is the modern industry standard for attenuating surface-related multiples in marine data. The method is based on the prediction that any surface-related multiple can be expressed as a convolution of two primary reflections, and that the autocorrelation of the upgoing wavefield recorded at the surface contains the full multiple model. SRME operates entirely in the data domain without requiring any subsurface model, which makes it robust even when velocity information is uncertain. The method requires a dense and well-sampled near-offset trace distribution because missing near-offset traces degrade the multiple prediction. In practice, SRME is followed by an adaptive subtraction step that matches the predicted multiples to the observed multiples using a least-squares filter, minimising any damage to primary energy during subtraction.

Parabolic Radon demultiple transforms common-midpoint (CMP) gathers from offset-time space into the tau-p domain, where primaries and multiples separate based on their different move-out curvatures. Primaries after NMO correction appear flat or have small residual move-out, while multiples retain significant residual move-out because their stacking velocity is lower than that of the equivalent primary at the same two-way time. A mute applied in the Radon domain retains high-slowness events (multiples) in a model that is subtracted from the data in the offset domain. The Radon approach is particularly effective for interbed multiples and peg-leg multiples that SRME does not predict. High-resolution Radon transforms that employ L1-norm or iterative reweighted least-squares algorithms outperform conventional L2-norm approaches in resolving closely spaced move-out differences between primaries and multiples.

Extended SRME (ESRME) and 3D SRME extend the original 2D method to full 3D acquisition geometries, accounting for out-of-plane multiple paths that 2D SRME misses. 3D SRME requires 3D-regularised data on a dense, regular grid, making 5D interpolation (which reconstructs missing traces in five dimensions: time, inline, crossline, azimuth, and offset) a near-mandatory preprocessing step before 3D SRME in modern deepwater surveys. FWI-based demultiple, where the FWI model is used to generate synthetic multiples that are then adaptively subtracted, is an emerging technique that can handle complex prismatic multiples and other events not easily handled by conventional methods.

Diffractions as Structural and Stratigraphic Indicators

The geophysics community has progressively shifted its view of diffractions from pure noise toward a valuable signal source. Diffraction imaging workflows explicitly separate the diffractive component of the wavefield from the specular (reflective) component using plane-wave destruction filters or dip-consistent filters, then migrate only the diffractive component to produce high-resolution images of edges, fractures, and small-scale heterogeneities. Because diffractions are generated wherever there is a sub-wavelength discontinuity, diffraction images can locate fault tips and fracture corridors with horizontal resolution approaching the dominant wavelength of the seismic data, approximately 20 to 50 m (65 to 165 ft) at typical exploration frequencies. This resolution is two to five times finer than the conventional Fresnel-zone horizontal resolution of standard reflection imaging.

In stratigraphic imaging, diffractions from channel edges, reef flanks, and the termination of high-impedance beds against angular unconformities provide geometric constraints on reservoir geometry that complement conventional amplitude analysis. A diffraction event from a channel edge, for example, precisely locates the lateral termination of the reservoir sand body and allows geometrically accurate modelling of the channel width. In fractured carbonate reservoirs common in the Middle East, diffraction images from the inter-well region can identify fracture corridors that are not sampled by wellbores and that control fluid flow, directly improving reservoir simulation grid design. These applications have elevated diffraction imaging from an academic curiosity to a routine workflow step in high-resolution reservoir characterisation projects, particularly in fields where reservoir complexity drives well-placement risk.