|
<< Click to Display Table of Contents >> Navigation: Noise attenuation > Wave separation by gather (LPWD) |
The Wave separation by gather (LPWD) module separates coherent seismic wavefields within individual gathers using the Low-rank Plane-Wave Decomposition (LPWD) algorithm, which is based on the Berk antileakage Fourier transform. Unlike the HRWS module which operates on multi-gather volumes, LPWD processes one gather at a time in a standard sequential flow. Within each gather, the algorithm applies the Berk transform in overlapping time-trace windows, extracts the target wavefield from the frequency-wavenumber domain according to the defined slope range, and returns the residual (input minus extracted wavefield) as the output.
The LPWD module is typically used for ground-roll attenuation, guided-wave removal, or up/down wavefield separation in individual shot or CMP gathers. Because processing is gather-by-gather, it is straightforward to insert into a standard sequential processing flow without requiring the volume-level data access needed by HRWS.
The sorted index item that controls the order in which input gathers are processed. This is the standard sequence control input.
The input seismic gather (shot gather, CMP gather, or receiver gather) to be processed. The module applies the plane-wave decomposition within this gather to identify and separate the target wavefield.
The number of traces in each spatial processing window. The window slides along the trace axis of the gather with overlap. A larger window improves the spatial resolution of the plane-wave decomposition but increases computation time. Typical values are 20–60 traces depending on the size of the gather and the spatial coherence of the noise.
The length of the time processing window in seconds. Each time-trace block is transformed independently. Shorter windows give better time resolution for non-stationary data, while longer windows improve frequency resolution. Typical values are 50–200 ms.
The minimum frequency (in Hz) processed by the Berk transform. Set this to exclude very low frequencies where the separation algorithm becomes unstable or where no relevant wavefield content is expected.
The maximum frequency (in Hz) processed by the Berk transform. Set this to the highest meaningful signal frequency in the data (typically the Nyquist frequency or the band limit of the seismic data).
The number of dominant plane-wave components to identify in the Berk decomposition. A value of 1 extracts the single most dominant dip event (e.g. a strong ground-roll cone or a direct wave). Increase this value to separate multiple overlapping dip events simultaneously, but note that higher values increase the risk of leaking signal energy into the extracted wavefield.
Internal parameter. Leave at default.
When enabled, the plane-wave extraction is restricted to the dip range defined by From slope and To slope. This confines the extracted noise model to a specific apparent velocity range (for example, the surface-wave cone) and prevents signal energy outside that range from being removed. Enable this parameter whenever the noise has a well-defined dip range distinct from the signal.
The lower bound of the target dip (slope) range when Use constrains is enabled. Energy with apparent velocity below this slope will not be attributed to the extracted wavefield.
The upper bound of the target dip (slope) range when Use constrains is enabled. Energy with apparent velocity above this slope will not be attributed to the extracted wavefield.
When enabled, the module outputs the extracted wavefield (the noise model) rather than the cleaned residual. Enable this option to inspect the separated noise model and verify that the separation is correctly targeting the intended event type.
Controls the frequency-domain processing strategy. All frequency applies the Berk transform across the full frequency band within each time-trace window, treating all frequencies simultaneously. Each frequency applies the decomposition independently at each individual frequency slice, which can improve separation quality when the dominant noise dip varies with frequency (e.g. dispersive surface waves), but significantly increases computation time.