LNA Radon 3D slow

Description

LNA Radon 3D slow attenuates coherent, linearly moving-out noise such as ground roll, guided waves and refracted or direct arrivals in 3D seismic data with irregular source/receiver geometry, using a local, iterative linear tau-p (Radon) transform. For every output bin it gathers nearby traces using independent source-side and receiver-side spatial search windows, then iteratively identifies and accumulates the dominant, coherent linear-moveout events that fall within a chosen apparent-velocity band and frequency range. The output trace at each bin is rebuilt entirely from this accumulated linear-moveout model, so energy that does not fit a consistent linear moveout within the chosen velocity window is excluded from the result. Depending on how the velocity window is tuned, the module can be used either to build a clean, noise-suppressed volume (aiming the window at the wanted primary trend) or to build a standalone linear-noise model volume (aiming the window at ground-roll or guided-wave velocities) for later subtraction elsewhere in the flow. It writes a brand-new output data volume rather than modifying the input in place, and the 'Create maps' action lets you preview the fold obtainable with the current aperture settings before committing to the full, comparatively slow, processing run.

Input data

Input data

The input seismic dataset the module reads gathers from.

Trace headers

The input trace geometry and headers - source, receiver and bin coordinates plus offsets - used to build the spatial index and drive the source/receiver aperture searches.

Parameters

Output file name

The path and name of the new internal data volume (.gsd) the module writes the processed result to. The module always creates a new file rather than overwriting the input.

Aperture Source

How far, around the source position of each trace being processed, the module searches for other traces to include in the local group of traces used to build the tau-p model. Tune this with the source-line spacing in mind: too small starves the model of traces, giving unstable or poorly resolved velocity picks, while too large mixes traces from different source patches, degrading moveout consistency and slowing the run. Minimum value is 1 m.

Default: 50 m.

Aperture Receiver

The equivalent of Aperture Source but for the receiver position of each trace. Together the two apertures form a source/receiver search box around each trace, and the larger of the two also sets the bin aperture and reference offset used by the transform. Increase it for sparser receiver geometries or when more fold is needed for a stable model; keep it tight where azimuth or offset variation should stay localized. Minimum value is 1 m.

Default: 350 m.

Window selection

A temporal smoothing/search window used when picking the dominant linear-moveout event at each time sample during the iterative reconstruction, and for the taper applied around each pick. Widen it to make picks more stable and continuous across nearby time samples, reducing jitter; narrow it to allow finer time resolution between closely spaced distinct events. Minimum value is 0.

Default: 20 ms.

Low Frequency1 / LowFrequency2

The lower and upper corners of the low-cut ramp that band-limits the data before the Radon transform - below Low Frequency1 the data is rejected, and above LowFrequency2 it fully passes. With both left at their matching default the low-cut is effectively an abrupt cut at 1 Hz; widen the gap between them for a gentler taper and to avoid edge artifacts.

Default: 1 Hz / 1 Hz.

HighFrequency1 / High Frequency2

The lower and upper corners of the high-cut ramp - HighFrequency1 is the last frequency that fully passes, and above High Frequency2 the data is fully rejected. Together with Low Frequency1/LowFrequency2 these define the working frequency band for the tau-p decomposition; narrow this band toward the target noise's spectrum (typically low frequency for ground roll) to focus the model on that noise, or keep it broad if trying to preserve broadband primary energy.

Default: 90 Hz / 90 Hz.

V min

The lower bound of the apparent-velocity (moveout) window the Radon transform searches. Set this near the slowest linear event you want to capture, for example near ground-roll velocity if building a noise model, or just below the slowest primary trend if isolating signal. Important: V min and V max must define a genuinely different (wider) band - if left equal, as they are by default, the module will fail to run, so always widen this range before executing. Valid range is 1 to 50,000 m/s.

Default: 1500 m/s.

V max

The upper bound of the apparent-velocity window, paired with V min. It must be set higher than V min (the module errors out immediately if they are equal, as they are by default) to define a real velocity band for the transform to search. Valid range is 1 to 50,000 m/s.

Default: 1500 m/s.

Delta P

The step size of the slope/velocity grid scanned between V min and V max. Smaller values give finer discrimination between apparent velocities that are close together, better separating signal from noise of similar moveout, at the cost of more parameters to solve and longer runtime; larger values are faster but coarser.

Default: 2 ms.

Number of iteration

The number of refinement passes the algorithm performs: each pass finds the strongest remaining linear-moveout event, adds it to the accumulated model, and re-searches what is left. More iterations resolve more, and weaker, overlapping linear events (sharper separation) but increase runtime and risk picking noise itself if pushed too far; too few leaves genuine linear energy unmodeled or leaking into the residual.

Default: 10.

Penalty Factor

A damping weight that suppresses model contributions at larger absolute slopes relative to near-zero slope, stabilizing the reconstruction and reducing spurious picks at the edges of the velocity grid. Raise it to further suppress steep or extreme-slope picks for a more conservative, flatter-biased model; lower it toward zero to let the model fit steep events more freely. Minimum value is 0.

Default: 0.01.

Calculation area

First inline number(-1 no limit)

Restricts processing to inlines at or after this number. Leave at -1 (no limit) to process the whole volume; set an explicit value to restrict a run to a QC subset, split a large volume into manageable batch jobs, or re-run only a specific area after a parameter change. Minimum value is -1.

Default: -1 (no limit).

Last inline number(-1 no limit)

Restricts processing to inlines at or before this number, paired with First inline number to bound the inline range processed. Leave at -1 (no limit) to process to the end of the volume. Minimum value is -1.

Default: -1 (no limit).

First crossline number(-1 no limit)

Restricts processing to crosslines at or after this number, with the same purpose as First inline number but on the crossline axis. Leave at -1 (no limit) to process the full crossline range. Minimum value is -1.

Default: -1 (no limit).

Last xLine number(-1 no limit)

Restricts processing to crosslines at or before this number, paired with First crossline number to bound the crossline range processed. Leave at -1 (no limit) to process to the end of the volume. Minimum value is -1.

Default: -1 (no limit).

Output data

Output data file

The new data volume written at the Output file name path. Each output trace is reconstructed purely from the accumulated linear tau-p model for its bin, so only coherent linear-moveout content found within the Aperture, V min/V max and frequency-band settings is carried through, while other energy is excluded from the written result.

References

Sacchi, M. D. and Ulrych, T. J., 1995, High-resolution velocity gathers and offset space reconstruction: Geophysics, 60, 1169-1177.

Trad, D., Ulrych, T. and Sacchi, M., 2003, Latest views of the sparse Radon transform: Geophysics, 68, 386-399.