Radial Processing

Key Takeaways

• The radial trace transform is a linear moveout (LMO) correction applied to each trace in a shot record: each trace is shifted in time by the amount t = x/v, where x is the source-to-receiver offset and v is the apparent velocity of the radial trace being defined, so that all events with a moveout velocity of v align horizontally across the shot record after the shift; after LMO correction, events that had velocity v are flat (horizontal), while events with other velocities are moved-out in the opposite direction from their original geometry; ground roll, which typically has apparent velocities of 300-1,200 m/s (depending on the near-surface shear velocity), occupies a specific range of radial traces corresponding to these velocities, while shallow reflections (apparent velocities of 1,500-3,000 m/s near the surface) and deeper reflections (apparent velocities exceeding 3,000 m/s) occupy different radial trace ranges; this velocity-based separation allows the ground roll to be muted in the radial domain without affecting the reflection data at those same frequencies and offsets, a selectivity that conventional f-k filtering often cannot achieve because ground roll and shallow reflection energy overlap in both frequency and wavenumber space on conventional shot records.
• Comparison with f-k (frequency-wavenumber) filtering reveals the key advantage of radial processing for ground roll attenuation: f-k filtering operates in the frequency-wavenumber domain and can separate ground roll from reflections when their apparent velocities differ sufficiently to produce non-overlapping dips in the f-k spectrum, but is compromised when the ground roll is dispersive (different frequencies travel at different velocities, as is typical for surface waves in layered near-surface geology), because the dispersive ground roll occupies a fan of dips in the f-k domain that may overlap with the reflection energy at certain frequencies; radial processing handles dispersive ground roll more effectively because it operates in the time-domain moveout space rather than the frequency-wavenumber space, allowing the noise to be identified from its velocity range regardless of its frequency content; however, radial processing requires careful velocity parameter selection (the minimum and maximum ground roll velocities to mute) and is most effective when the ground roll velocities are distinct from the shallow reflection velocities, which can be problematic in areas where the near-surface velocity is low (less than 800 m/s) and the ground roll overtakes shallow reflections in apparent velocity.
• Application to land seismic data in unconventional resource plays is a major use case for radial processing: in shale resource plays (Permian Basin, Anadarko, STACK/SCOOP, Montney, Duvernay), the acquisition requires wide-azimuth 3D surveys with dense receiver arrays on the surface, and high-amplitude ground roll generated by the vibroseis sources and surface waves is the dominant noise contaminating near-offset reflection data; the near-offset data is critical for AVO analysis used to discriminate gas-saturated from water-saturated shale intervals and to estimate anisotropy parameters for completion design; ground roll that is not removed degrades the intercept (A) and gradient (B) AVO attributes because it adds amplitude anomalies that mimic or mask fluid-related AVO anomalies; radial processing applied early in the processing sequence (before other noise attenuation steps) improves the input data quality for subsequent AVO analysis, pre-stack inversion, and anisotropy estimation, directly improving the rock and fluid property models used for completion optimization in horizontal wells.
• Radial processing for refraction noise removal addresses a different noise type from ground roll but uses the same transform principles: refractions (head waves traveling along a high-velocity interface and returning to the surface at a specific critical velocity determined by the refractor velocity) appear at a specific radial trace velocity equal to the refractor velocity; by identifying the refraction velocity from the first-break pattern on shot records and muting those radial traces, the refraction arrivals can be removed from the data without degrading the reflection signal; this is useful when refraction first breaks overlap with shallow reflections at near offsets (a problem in areas with a high-velocity near-surface layer such as limestone or permafrost) or when refraction multiples (head wave reverberations between refractors) contaminate the shallow reflection section; in areas with strong refractors (the Permian Basin Wolfcamp, for example, where the San Andres evaporite creates a strong refractor), radial refraction muting combined with radial ground roll muting substantially improves the signal-to-noise ratio of the near-surface reflection section.
• Integration of radial processing with other noise attenuation methods in a comprehensive noise removal sequence reflects the reality that no single method removes all coherent noise types from land seismic data: radial processing is typically applied first (on shot records before sorting) to remove ground roll and refractions whose velocity-moveout character is most clearly defined on shot records, followed by surface-consistent deconvolution to remove source and receiver coupling effects, surface wave attenuation (SWAT) or empirical mode decomposition for residual dispersive surface wave removal, and finally common-midpoint (CMP)-domain noise attenuation (high-resolution radon transform, SRME) for multiple removal; the order of operations matters because each step changes the apparent noise character for subsequent steps; radial processing performed incorrectly (wrong velocity bounds, excessive mute) can remove shallow reflection energy along with ground roll, degrading the very near-offset data it was intended to preserve, so quality control of the radial domain mute boundaries against first-break and ground roll velocity analysis is a mandatory step before applying radial processing to production data.