VSP processing

<< Click to Display Table of Contents >>

Navigation:  VSP >

VSP processing

 

Description

VSP processing is a comprehensive zero-offset Vertical Seismic Profile (VSP) processing module. It ingests raw VSP field records acquired at a borehole, performs first-break picking, up/down wavefield separation, spiking deconvolution on the up-going waves, NMO correction, and optionally constructs a near-well interval velocity model by tomographic inversion of the first-break times. The processed up-going wavefield is stacked into a well stack trace that ties the borehole to surface seismic, and optionally a full CDP stack volume in time or depth can be generated around the well.

This module is designed for zero-offset or near-offset VSP surveys where the source is positioned at or very close to the wellhead. First-break picks can be imported from external files, picked automatically, or edited interactively. The resulting well stack trace and velocity model can be directly compared to surface seismic sections within g-Platform.

Input data

Input DataItem

The standard seismic data container carrying the raw VSP field records. Connect the output of an upstream data reader module.

Input SEG-Y data handle

SEG-Y file handle used for reading auxiliary data such as checkshot times and well marker information.

Input trace headers

Trace header vector providing geometry information (receiver depths, coordinates) required for first-break picking and velocity model construction.

Top mute picking item

Optional corridor mute pick defining the upper boundary of the up-going wave window. Connect a mute picking item or leave disconnected to apply no top mute.

Bottom mute picking item

Optional corridor mute pick defining the lower boundary of the up-going wave window. Connect a mute picking item or leave disconnected to apply no bottom mute.

Parameters

General

Core survey and borehole reference parameters used throughout the processing flow.

 V0

Near-surface reference velocity used as the initial velocity for ray tracing and depth conversion. Default: 1500 m/s. Set this to the average velocity of the shallow section near the well, typically derived from uphole surveys or refraction statics.

 KB

Kelly bushing elevation (m) — the height of the rotary table above mean sea level or above the reference datum. Default: 0 m. Set this to the measured KB value from the well header to ensure correct depth-to-time conversion.

First break auto picking

Controls the automatic first-break (direct arrival) picker used when running the Auto pick FB picks on current source or Auto pick FB picks on all sources custom actions. Accurate first-break picks are essential for up/down wavefield separation and velocity model building.

 Window

Search window length (s) centred on the expected first-break time within which the picker looks for the onset. Default: 5 s. Reduce this value to restrict picking to a narrow band around the predicted arrival, which helps avoid noise picks when the direct wave is well-defined.

 Threshold

Signal-to-noise ratio threshold that a sample must exceed to be accepted as a first-break pick. Default: 3. Increase this value on noisy data to avoid false picks; decrease it on high-quality data where the first break is subtle.

 Pick positive phase

When enabled (default), the picker identifies the first positive amplitude peak as the first break. Disable for surveys where the direct arrival has a negative polarity due to instrument or acquisition convention.

 Magnet type

Snap-to method applied after the initial pick to refine its position. Default: Nearest. Nearest snaps to the nearest zero-crossing or onset; Max energy snaps to the peak amplitude within the magnet window.

 Magnet window

Half-length (s) of the snap window used by the magnet to refine the first-break pick. Default: 0.2 s. Reduce this if the magnet snaps to an incorrect nearby peak.

NMO

Normal Moveout correction applied to the VSP gathers prior to stacking, using the velocity model derived from first-break times.

 Stretch factor

Maximum allowable NMO stretch expressed as a fraction (0–1). Default: 0.5 (50%). Samples stretched beyond this threshold are muted to avoid low-frequency NMO artefacts at shallow times. Reduce to 0.3–0.4 for cleaner near-offset stacking.

Up/down separation

Controls the separation of down-going (direct and multiply-reflected) waves from up-going reflections. Effective up/down separation is the most critical step in VSP processing — the up-going wavefield contains the reflectivity information that ties to surface seismic.

 Separation method

Algorithm used to separate up-going from down-going wavefields. Default: FK domain. FK domain applies an F-K fan filter that passes only up-going apparent velocities (positive wavenumbers); it requires accurate first-break picks to orient the filter correctly. Median filtering subtracts the median trace (dominated by the down-going wave) from each receiver trace and is more robust when the receiver array is short or picks are uncertain.

 Taper

Width of the transition zone (fraction, 0–0.25) at the boundary of the F-K rejection fan filter. Default: 0.05 (5%). Increase to suppress Gibbs ringing at the filter edge; decrease for a sharper filter boundary.

 Use FK filtering when no FB picks available

When enabled (default), the FK filter is still applied to sources that have no first-break picks, using a fixed apparent velocity boundary. Disable to skip separation entirely for unpicked sources and avoid potentially incorrect filtering.

 Filter window

Number of traces used in the sliding median filter window when Separation method is Median filtering. Default: 10 traces. Larger windows give smoother separation but may over-attenuate laterally varying up-going events.

Deconvolution

Spiking deconvolution applied to the up-going wavefield to broaden the frequency bandwidth and remove the borehole source wavelet. The operator is designed trace-by-trace from the autocorrelation of the separated up-going wave.

 Apply deconvolution to up-going waves

When enabled (default), spiking deconvolution is applied to the separated up-going wavefield. Disable when the source wavelet has already been removed by an earlier processing step or when deconvolution artefacts are undesirable.

 Operator length

Length (s) of the autocorrelation-based deconvolution operator. Default: 0.5 s. Set this to cover at least one full source wavelet length plus any reverberations that need removing. Typical VSP values range from 0.2 s to 1.0 s.

 Lag time

Starting lag (s) for the autocorrelation used to design the deconvolution operator. Default: 0 s. A non-zero lag can be used when the main energy of the source wavelet does not begin at zero time.

 Noise

Pre-whitening level added to the zero-lag autocorrelation value before inverting the deconvolution operator, expressed as a fraction of the maximum autocorrelation. Default: 0.001 (0.1%). Increasing this stabilises the inversion on noisy data at the cost of reduced whitening efficiency. Values between 0.001 and 0.01 are typical.

 Convert wavelet to min phase

When enabled, the deconvolution operator converts the source wavelet to a minimum-phase equivalent before whitening. Default: off. Enable when subsequent processing steps (e.g., match filter to surface seismic) assume a minimum-phase source wavelet.

Well stack trace

Controls how the processed up-going wavefield from adjacent receiver levels is stacked to form the final well stack trace output used for seismic-to-well tie.

 Number of shown traces

Number of receiver traces summed (stacked) around each depth level to produce the well stack. Default: 6 traces. Increasing this improves signal-to-noise ratio but widens the effective aperture of the stack.

 Taper

Cosine taper length (s) applied at the top and bottom of each up-going trace before stacking. Default: 0.05 s. The taper suppresses edge artefacts from the corridor mute boundaries.

Velocity model

Parameters controlling the construction of the near-well interval velocity model extrapolated from the first-break times. This model is used for NMO correction and for building the CDP stack.

 Velocity model grid step

Lateral grid cell size (m) of the output velocity model. Default: 25 m. Use a step consistent with the expected lateral velocity variation — smaller values give finer resolution but increase computation time.

 Additional aperture

Distance (m) by which the velocity model is extended laterally beyond the receiver spread footprint. Default: 500 m. Increase this value to ensure the model covers the full CDP stack aperture.

 Velocity model DZ

Vertical sample interval (m) of the output velocity model. Default: 5 m. Set to match or be finer than the receiver spacing to capture the vertical velocity detail resolvable from the VSP data.

 Correlation radius

Number of neighbouring receiver traces included in the correlation window when matching first-break picks across levels. Default: 5 traces. Increase this on noisy data to improve pick continuity.

 Velocity smooth radius

Lateral smoothing radius (m) applied to the velocity model after building it from first-break picks. Default: 50 m. Increase to suppress high-frequency lateral velocity noise; reduce to preserve real lateral heterogeneity.

Tomography

Parameters for the tomographic first-break inversion that refines the velocity model beyond the simple extrapolation approach. Tomography is triggered by the Build full velocity model run option and iteratively updates the slowness model to fit the first-break traveltimes.

 Number of global iterations

Number of outer tomographic iterations in which the entire velocity model is updated. Default: 5. More iterations improve convergence but increase runtime. Typical values are 3–10.

 Number of local iterations

Number of inner conjugate-gradient solver iterations within each global update step. Default: 5. Increase for more accurate model updates per outer iteration.

 Frequency

Dominant frequency (Hz) of the VSP data, used to compute the Fresnel zone radius that controls tomographic resolution. Default: 20 Hz. Set this to match the peak frequency of the first-break arrivals in your data.

 Resolution

Minimum spatial scale (m) of velocity anomalies the tomography is allowed to recover. Default: 5 m. This acts as a regularisation length — larger values produce smoother velocity models and are more appropriate for sparse acquisition geometries.

 Use velocity constrain

When enabled (default), a spatial constraint is applied to keep the tomographic velocity model close to the initial velocity extrapolated from the well. This stabilises the inversion in areas with poor ray coverage. Disable only when the initial model is known to be poor.

 Velocity constrain window XY

Lateral smoothing window (m) for the velocity constraint. Default: 500 m. Larger values apply a stronger lateral constraint, keeping the model close to the background velocity trend.

 Velocity constrain window Z

Vertical smoothing window (m) for the velocity constraint. Default: 200 m. Increase when the vertical velocity gradient is expected to be smooth (e.g., deep sedimentary basins).

 Use velocity smooth

When enabled (default), a Gaussian smoothing filter is applied to the velocity model after each global iteration to suppress short-wavelength artefacts. Disable for a higher-resolution but noisier model.

 Velocity smooth window XY

Lateral radius (m) of the Gaussian smoothing filter applied to the velocity model. Default: 200 m. Set to roughly half the expected minimum lateral scale of velocity anomalies.

 Velocity smooth window Z

Vertical radius (m) of the Gaussian smoothing filter. Default: 200 m. Reduce if thin high-velocity layers are present and need to be preserved.

 Store time tables on disk

When enabled, eikonal traveltime tables are written to and read from disk rather than kept in memory. Default: off. Enable on workstations with limited RAM when processing large models.

 Temporary storage path

File system path where traveltime tables are stored when Store time tables on disk is enabled. Specify a directory on a fast local disk to minimise I/O overhead.

 Threshold on update

Convergence threshold expressed as a fraction (0–0.25) of the model update norm. Default: 0.05 (5%). Iterations stop early if the relative velocity update falls below this value, saving computation time when the model has converged.

CDP stack

Parameters for the optional volumetric CDP stack generated from the processed VSP data around the borehole. This output can be directly compared to surface seismic and used for near-well structural interpretation.

 Bin grid step X

Inline bin spacing (m) for the CDP stack grid. Default: 25 m. Set to match the desired lateral resolution of the CDP image.

 Bin grid step Y

Crossline bin spacing (m) for the CDP stack grid. Default: 25 m.

 Sample ratio

Vertical sample interval (m) of the CDP stack output volume. Default: 5 m.

 Number of samples

Total number of depth samples in the CDP stack output. Default: 1000 samples. The total output depth range equals Number of samples × Sample rate.

 Aperture

Maximum offset (m) from the well used to gather traces for the CDP stack migration. Default: 5000 m. Larger apertures improve illumination at depth but include more distant (and noisier) receiver groups.

Ray casting

Controls the ray visualisation displayed on the location map in the Graphics panel, showing how rays propagate from sources through the velocity model to the borehole receivers.

 Display rays

When enabled (default), ray paths are rendered on the location map for the currently selected source. Disable to improve display performance with large source arrays.

 Ray casting step

Step size (m) along each ray path during visualisation. Default: 5 m. Reduce for smoother ray curves in highly refracting velocity models; increase to improve display speed.

 Receiver decimation

Displays only every N-th receiver ray to reduce visual clutter. Default: 1 (show rays for all receivers). Increase when there are many closely spaced receivers and the display becomes crowded.

 

Output data

The module produces the following outputs: the Well stack trace is a stacked corridor gather representing the up-going reflectivity as a function of two-way time and is used directly for seismic-to-well tie; Well stack trace with multiples is the same output but without multiple suppression; NMO gather converted to depth is the NMO-corrected gather mapped to the depth axis; CDP stack and CDP stack in time are the optional full volumetric stacks in depth and time respectively; Well velocity model is the 1D interval velocity log at the borehole; Velocity model is the full 2D/3D velocity volume built by extrapolation and tomography.

Custom actions

Import FB picks from binary/text file

Loads first-break picks from an external binary or text file, replacing any existing picks. Use when picks were made in an external application or transferred from a previous project.

Export FB picks to binary/text file

Saves the current first-break picks to a file for archiving or use in external applications.

Auto pick FB picks on current source

Runs the automatic first-break picker on the currently displayed source gather only. Use for interactive QC before running on the full dataset.

Auto pick FB picks on all sources

Runs the automatic first-break picker over all source gathers in the dataset.

Clear picking on current source

Removes all first-break picks for the currently selected source gather. Use before re-picking a problematic gather.

Clear all FB picks

Removes all first-break picks from the entire dataset. Use before re-running autopicking with updated parameters.

--------------------

Export checkshot to text file

Exports the first-break derived checkshot table (depth–time pairs at each receiver level) to a text file for use in well-log depth conversion or seismic-to-well tie workflows.

Export well markers to text file

Exports well formation-top markers (interpreted at the borehole) to a text file.

--------------------

Select closest source

Automatically selects and displays the source gather closest to the wellhead. Useful to start interactive processing from the best-quality near-offset gather.

Clear corridor picking

Clears the top and bottom corridor mute picks used to isolate the up-going wavefield.

Load corridor picking

Loads corridor mute picks from a previously saved file.

Save corridor picking

Saves the current corridor mute picks to a file for reuse in future processing sessions.

Zero offset processing

Runs the complete zero-offset VSP processing sequence (up/down separation, deconvolution, NMO, stacking) on all gathered source data and updates all output displays.

--------------------

Rebuild CDP stack

Rebuilds the CDP stack output volume from the already-processed up-going wavefield, using the current CDP stack parameters. Use after adjusting bin size, aperture, or sample interval without re-running the full processing chain.