Description

This module processes multicomponent Ocean Bottom Node (OBN) gathers that combine a hydrophone (pressure) sensor with a geophone package (vertical and, where available, transverse/radial horizontal components). It first denoises the vertical geophone component, removing coupling and tilt-related noise that does not appear on the hydrophone, using a choice of algorithms. It then separates the denoised wavefield into its up-going and down-going parts by combining the hydrophone and adjusted geophone data, which is the standard way to remove the water-layer receiver-side ghost from OBN data. Use this module as the core denoise-and-separate step of an OBN processing sequence, before further imaging or deghosting-dependent steps.

Input data

Seismic gather

A multicomponent OBN gather containing the traces to be processed. The module identifies each trace's component (hydrophone/pressure, vertical geophone, and, where the denoise method requires them, transverse and radial geophone components) from the trace type header, so this header must be set correctly for every component before running the module.

Denoise

Denoise method

Selects the algorithm used to remove noise from the vertical geophone component before wavefield separation. "JSR" (Joint Sparsity Recovery) jointly inverts the hydrophone and vertical data to directly solve for the up-going wave, down-going wave and noise, and can optionally work in a transform domain for data with significant dip. "MF" (Masking filter with adaptive subtraction) removes noise that is correlated with the horizontal (transverse and radial) geophone components, which is typically low-frequency tilt or rocking noise on ocean-bottom sensors. "Adaptive" (Adaptive subtraction) estimates and subtracts a best-fit filtered version of the hydrophone data from the vertical component, and is a good general-purpose choice when the noise on the vertical component is coherent with the hydrophone signal.

Default: Adaptive. "MF" requires transverse and radial components to be present in the input gather.

Use transform

Available when Denoise method is "JSR". Chooses whether the joint sparsity inversion runs directly on the time-domain data ("None") or first transforms the data into a domain that flattens locally coherent dipping events ("Seislet") or into a wavelet domain ("Wavelet"). Working in a transform domain can improve the separation on data with significant lateral dip.

Default: None.

Noise amplitude coefficient

Available when Denoise method is "JSR". Controls how much relative weight the joint inversion gives to the noise term, effectively setting how much amplitude can be attributed to noise rather than to the up- and down-going wavefields. Increase it for more aggressive noise removal; decrease it to keep more energy in the wavefield outputs if genuine signal is being pulled into the noise output.

Default: 1.0 (minimum 1e-6).

Number of global iterations

Available when Denoise method is "JSR". Number of outer iterations of the sparse inversion used to jointly solve for the wavefield components and noise. Increasing it can improve convergence and separation quality at the cost of longer run time.

Default: 3 (minimum 1).

Number of local iterations

Available when Denoise method is "JSR". Number of inner iterations performed within each global iteration of the inversion. Increase for a more precise solution at each outer step, at the cost of run time.

Default: 5 (minimum 1).

Tolerance

Available when Denoise method is "JSR". Convergence tolerance that allows the inversion to stop early once the solution stabilizes within this threshold. Lower values require tighter convergence and can increase run time.

Default: 1e-6 (minimum 0).

Mu

Available when Denoise method is "JSR". Internal penalty weight used by the inversion solver to enforce agreement with the input data. Larger values enforce the data fit more strictly; this is an advanced tuning parameter and in most cases should be left at its default.

Default: 1e4 (minimum 1).

Weighting coefficient

Available when Denoise method is "MF". Sets the sensitivity of the masking filter that separates horizontal-component-correlated noise from the vertical geophone signal, effectively acting as the amplitude-ratio threshold used to classify a sample as noise. Increase it if too much genuine signal is being removed; decrease it if noise is not being sufficiently suppressed.

Default: 1.0 (minimum 1e-8).

Smooth coefficient

Available when Denoise method is "MF". Controls how sharp the transition is between samples classified as "signal" and "noise" by the masking filter. Higher values give a sharper, more binary transition; lower values give a softer, more gradual mask.

Default: 1 (minimum 1).

Number of iterations (MF)

Available when Denoise method is "MF". Number of iterations used by the solver that estimates how each horizontal component couples into the vertical geophone. Increasing it can improve the fit at the cost of longer run time.

Default: 100 (minimum 1).

Trace window (Adaptive)

Available when Denoise method is "Adaptive". Number of neighboring traces on either side of each trace used when estimating the adaptive matching filter between the hydrophone and vertical geophone data. Larger windows give a more stable filter estimate by averaging over more traces, but assume the noise character is similar across that spatial range.

Default: 5 traces (minimum 1).

Time window (Adaptive)

Available when Denoise method is "Adaptive". Length of the sliding time window used to estimate the adaptive matching filter between the hydrophone and vertical geophone data. Shorter windows adapt faster to noise that changes quickly with time; longer windows give a more stable, averaged estimate.

Default: 0.3 ms (minimum 0.00001 ms).

Min lag

Available when Denoise method is "Adaptive". Earliest (most negative) time lag searched when building the adaptive matching filter between hydrophone and geophone data. Together with Max lag it sets the aperture of the matching filter; widen it if the noise on the two components is not perfectly time-aligned, for example due to sensor coupling delays.

Default: 0 ms. Must not exceed Max lag.

Max lag

Available when Denoise method is "Adaptive". Latest (most positive) time lag searched when building the adaptive matching filter. Must be greater than or equal to Min lag.

Default: 0 ms.

Step lag

Available when Denoise method is "Adaptive". Increment, between Min lag and Max lag, used when scanning for the best matching-filter lag. Smaller steps give finer lag resolution at the cost of more computation.

Default: 0.004 ms (minimum 0.00001 ms).

Solver type

Available when Denoise method is "Adaptive". Numerical method used to solve for the adaptive matching filter: "LSQR" is a standard least-squares solver, "LSQR advance" is an enhanced variant of it, and "FISTA" is a sparsity-promoting iterative solver that tends to produce a more compact filter. Try FISTA if you want a shorter, sparser filter, or if LSQR-based results look unstable.

Default: LSQR.

Lambda

Available when Denoise method is "Adaptive". Regularization strength, as a percentage, applied while solving for the adaptive matching filter. Higher values produce a smoother, more stable but less exact filter; lower values fit the data more closely but can become unstable or overfit noisy traces.

Default: 0.0001% (minimum 0).

Use raw model

Available when Denoise method is "Adaptive". Includes the raw (unmodified) hydrophone trace as one of the building blocks used to construct the adaptive matching filter. This should normally stay enabled, as it represents the direct coupling between the hydrophone and the noise on the vertical component.

Default: on (enabled).

Use Hilbert model

Available when Denoise method is "Adaptive". Also includes a 90-degree phase-shifted (Hilbert transform) version of the hydrophone trace as a building block for the matching filter, which helps capture noise that is coherent with the hydrophone signal but phase-shifted relative to it.

Default: on (enabled).

Use diff model

Available when Denoise method is "Adaptive". Also includes the time derivative of the hydrophone trace as a building block for the matching filter, useful when the noise coupling mechanism responds to the rate of change of the pressure signal rather than to its raw amplitude.

Default: on (enabled).

Use Hilbert diff model

Available when Denoise method is "Adaptive". Also includes the time derivative of the Hilbert-transformed hydrophone trace, combining the phase-shift and rate-of-change effects of the two options above. Enabling more of the four model options gives the adaptive filter more flexibility to match complex noise, at some risk of also removing genuine signal; disable options if the filter appears too aggressive.

Default: on (enabled).

Hydrophone scale coefficient

Multiplies the hydrophone (pressure) component amplitude before it is used in denoising and wavefield separation, so its amplitude and units can be matched to the vertical geophone component. Applies regardless of the selected Denoise method. Adjust it if the hydrophone and geophone sensors have different absolute calibration so the two signals are on a comparable scale before being combined.

Default: 1.0 (minimum 1e-10).

Adaptive separation

Find operator by

Chooses the numerical method used to estimate the operator that adjusts the geophone amplitude and phase to match the hydrophone before the final up/down wavefield separation. This step runs after denoising for the MF and Adaptive methods (and for JSR when a transform is used). "LSCG" uses a least-squares conjugate-gradient solver; "FISTA" uses a sparsity-promoting iterative solver that can produce a more compact operator.

Default: LSCG.

Number of iterations (Adaptive separation)

Number of iterations used by the chosen solver (LSCG or FISTA) when estimating the separation operator. More iterations can improve accuracy at the cost of run time.

Default: 100 (minimum 1).

FISTA epsilon

Available when Find operator by is "FISTA". Convergence and sparsity threshold used by the FISTA solver when estimating the separation operator.

Default: 0.1 (minimum 0).

Time window (Adaptive separation)

Length of the time window, centered on each sample, used to estimate the local separation operator. Larger windows use more temporal context to produce a stable operator, but assume the operator does not change quickly with time; smaller windows adapt faster to time-varying conditions.

Default: 0.4 ms (minimum 0).

Trace window (Adaptive separation)

Number of neighboring traces on either side included when estimating the local separation operator. Larger windows give a spatially smoother, more stable operator at the cost of adapting less to lateral changes; smaller windows adapt more locally but produce a noisier operator estimate.

Default: 5 traces (minimum 0).

Conversion operator aperture

Half-length, in time, of the operator (filter) used to adjust the vertical geophone component before wavefield separation. A longer aperture lets the operator correct for larger time-lag mismatches between the hydrophone and geophone, for example from sensor coupling or calibration differences, at the cost of a longer, less well-constrained filter.

Default: 0.2 ms (minimum 0).

Output data

Up waves gather

The up-going wavefield recovered by combining the hydrophone and denoised, gain-matched vertical geophone data. This is the component of primary interest for imaging, with the down-going, water-layer-reflected energy removed.

Down waves gather

The down-going wavefield component (energy traveling downward at the seabed, including the water-layer ghost), separated from the up-going wavefield.

Vz noise

The noise component removed from the vertical geophone by the selected Denoise method. Reviewing this gather is a useful quality-control step to confirm genuine signal is not being removed along with the noise.

Operators

A quality-control gather showing, per trace, the local operator estimated during adaptive wavefield separation, plotted against operator time (lag). Reviewing this can help confirm the separation operator is well behaved and not showing unexpected shifts across the survey.

Vertical component

Quality-control output showing the original, raw vertical geophone component as extracted from the input gather (populated when Denoise method is "MF").

Denoised vertical component

Quality-control output showing the vertical geophone component after denoising, for comparison against the Vertical component output (populated when Denoise method is "MF").

Hydrophone autocorr

Quality-control autocorrelation of the hydrophone trace against itself, versus time lag (populated when Denoise method is "MF").

Hydrophone and geophone crosscorr

Quality-control cross-correlation between the hydrophone trace and the original (pre-denoise) vertical geophone trace, versus time lag (populated when Denoise method is "MF").

Hydrophone and denoised geophone crosscorr

Quality-control cross-correlation between the hydrophone trace and the denoised vertical geophone trace, versus time lag, useful for confirming the denoise step improved the hydrophone/geophone match (populated when Denoise method is "MF").

References

Barr, F. J. and Sanders, J. I., 1989, Attenuation of water-column reverberations using pressure and velocity detectors in a water-bottom cable: 59th SEG Annual International Meeting, Expanded Abstracts, 653-656.