|
<< Click to Display Table of Contents >> Navigation: Velocity > ZOMF tomo create |
The ZOMF tomo create module builds a tomographic depth velocity model from Zero-Offset MultiFocusing (ZOMF) parameters. Starting from COMF (Common Offset MultiFocusing) geometry data — which includes the NIP (Normal Incidence Point) radii, semblance/correlation values, fold, and dip attributes derived from the MultiFocusing analysis — the module shoots rays through an initial depth velocity model, computes travel-time residuals between observed and modelled zero-offset times, and updates the velocity model by solving a least-squares linear system. The process iterates to convergence, producing a geologically consistent depth velocity model.
The module supports GPU, multi-threaded, and distributed cluster execution for large 3D surveys. Tikhonov regularisation in both lateral and depth directions ensures spatial smoothness and numerical stability of the solution. Optional per-iteration smoothing can be applied to the updated velocity model to prevent high-frequency artefacts from accumulating over multiple iterations.
A boolean flag indicating whether the input data is in COMF (Common Offset MultiFocusing) format. When enabled, the module reads COMF geometry from the connected trace headers item and derives ZOMF tomo parameters from the multi-offset COMF attributes. When disabled, the module uses a single-mode ZOMF approach without explicit multi-offset geometry.
The trace header collection containing the COMF source-receiver geometry. These headers define the spatial positions and offsets of the MultiFocusing traces used when Comf tomography is enabled.
The input CRE (Common Reflection Element) radii from the MultiFocusing analysis. The CRE radius characterises the curvature of the normal-incidence wavefront and is one of the key parameters used to constrain the depth velocity inversion. Connect the Rcre output from the MultiFocusing search module.
The semblance or correlation values from the MultiFocusing search, measuring the coherence of each COMF pick. High-semblance picks receive more weight in the tomography inversion. Picks below the Threshold correlation value are excluded entirely.
The fold values from the MultiFocusing search, indicating the number of traces contributing to each pick. Picks with fold below the Threshold fold value are excluded from the tomographic inversion.
The azimuth angle of the structural dip at each MultiFocusing pick location, in degrees from north. Together with Angle of dip, this defines the three-dimensional orientation of the reflector normal, which determines the direction of the normal-incidence ray used in the tomographic ray shooting.
The dip angle of the reflector at each MultiFocusing pick location, in degrees from horizontal. Used together with Azimuth of dip to determine the normal-ray takeoff direction for the tomographic forward modelling.
The initial depth-domain interval velocity model that serves as the starting point for the tomographic inversion. The module shoots rays through this model, computes travel-time residuals, and iteratively updates the model toward the ZOMF-constrained solution. Connect the output from a Dix conversion module or a previous ZOMF iteration. A good starting model accelerates convergence and prevents the inversion from falling into local minima.
When enabled, the near-surface reference velocity (V0) is automatically extracted from the top of the initial model at each CMP location, producing a spatially variable V0 map. When disabled (default), the single constant V0 value specified below is used everywhere. Enable this when the initial model contains spatially varying near-surface velocities that should be preserved in the inversion.
The near-surface (datum-level) velocity (in m/s) used as the reference velocity for the tomographic inversion. Default: 1500 m/s. This value is used when Construct V0 map from model is disabled. It anchors the ray shooting at the datum and is used to constrain the shallow part of the velocity model. Use the replacement velocity or near-surface velocity from the statics solution for the survey.
This container defines the spatial grid geometry for the ray-tracing computation and the quality thresholds used to filter input picks.
The horizontal spatial step of the ray-fan grid in both X and Y directions, in metres. Default: 500 m. This controls the lateral density of the ray shooting grid. Smaller values provide finer spatial resolution of the tomographic update but significantly increase computation time. Typical values are 200–1000 m depending on the survey bin size and required model resolution.
The time (or depth) step used to space the ray shooting origins along the zero-offset travel-time axis, in seconds. Default: 0.2 s. This determines the temporal density of the tomo picks contributed to the inversion. Reduce to 0.05–0.1 s for finer vertical resolution at the cost of more rays and longer compute time.
The minimum zero-offset two-way time at which rays are started, in seconds. Default: 0 s. Increase this to exclude very shallow picks where the velocity model is poorly constrained or where surface noise dominates the MultiFocusing picks.
The maximum zero-offset two-way time at which rays are traced, in seconds. Default: 4 s. Set this to the maximum time of the deepest target horizon. Rays beyond this time are not shot, reducing computation time for shallow models.
The maximum depth (in metres) of the output velocity model. Default: 2000 m. The tomographic update is computed from the datum down to this depth. Set this to encompass the full depth range of interest in the survey, allowing some margin beyond the deepest target.
The reference datum elevation (in metres) from which ray shooting begins. Default: 100 m. Set this to the flat processing datum of the survey. All rays are initiated at this elevation level and traced downward into the initial velocity model.
The minimum semblance (correlation) value required for a MultiFocusing pick to be used in the tomographic inversion. Default: 0.01. Picks with semblance below this value are excluded. Increasing this threshold (e.g. to 0.3–0.5) uses only the highest-quality, most coherent picks but may reduce the spatial coverage of the inversion. A value of 0.1–0.3 is a common starting point.
The minimum fold (number of contributing traces) required for a MultiFocusing pick to be included in the tomographic inversion. Default: 50. Picks with fold below this value are excluded as insufficiently constrained. Reduce this value in areas of low fold coverage to maintain adequate spatial sampling, but expect lower quality picks in those areas.
The lateral cell size (in metres) of the model update grid in the X and Y directions. Default: 500 m. This defines the resolution of the tomographic velocity update cells. It should be consistent with the Ray fan step X/Y and the available spatial density of the ZOMF picks.
The vertical cell size (in metres) of the model update grid. Default: 200 m. This defines the depth resolution of the tomographic velocity cells. Use a value consistent with the vertical resolution available from the ZOMF data and the Ray fan step ZT parameter.
The step size (in metres) used when shooting rays in the X and Y directions to sample the offset axis of the COMF data. Default: 100 m (X) and 100 m (Y). These parameters control how densely the ray fan covers the offset aperture in each direction. Smaller steps provide more complete ray coverage but increase computation time.
The maximum source-receiver offset (in metres) along the X direction over which rays are traced. Default: 2000 m. Set this to the maximum usable offset of the seismic data. The maximum offset in the Y direction defaults to 20 m, reflecting that the survey geometry is primarily 2D or has limited crossline extent for this ray component.
The time step used for numerical ray integration through the velocity model, in seconds. Default: 0.004 s (4 ms). A smaller ray time step produces a more accurate ray path but increases computation time. The value should be a fraction of the dominant wavelength in the velocity model. Leave at default unless working with very strong velocity gradients.
This container groups all parameters controlling the iterative linear solver used to update the velocity model from the travel-time residuals.
The number of outer tomographic iterations. Default: 10. Each global iteration re-traces rays through the updated model, recomputes travel-time residuals, and solves for a new velocity perturbation. More iterations refine the model further; convergence is typically reached within 5–20 iterations depending on the complexity of the velocity field.
The maximum number of iterations of the linear solver (LSCG or LSQR) within each global iteration. Default: 1000. The inner solver minimises the linear system relating velocity perturbations to travel-time residuals. The solver exits early if the residual falls below the solver's internal tolerance before reaching this limit.
The scaling factor (as a percentage) applied to the computed velocity perturbation before adding it to the current model. Default: 1%. A smaller perturbation ratio makes each step conservative and reduces oscillation, while a larger value accelerates convergence but risks overshooting. In practice, values of 0.5–5% are typical.
The fraction (as a percentage) of picks with the highest travel-time residuals that are downweighted or excluded from the solver. Default: 0.1%. This is a form of robust inversion that reduces the influence of outlier picks from poorly imaged zones or pick errors. Increase this value (e.g. to 1–5%) if the input picks contain significant noise.
The linear solver algorithm used to compute the velocity perturbation. Default: LSCG (Least-Squares Conjugate Gradient). LSCG and LSQR are iterative sparse solvers appropriate for large-scale 3D tomography. Mean value applies a simpler averaging approach. LSCG is recommended for most applications; LSQR may offer better convergence stability in some cases.
The Tikhonov regularisation weight applied in the X and Y (lateral) directions. Default: 1. This penalises lateral roughness in the velocity update, promoting spatial smoothness. Larger values produce a smoother (but possibly less accurate) model; smaller values allow more spatial variability. Adjust based on the lateral density and quality of the ZOMF picks.
The Tikhonov regularisation weight applied in the Z (depth) direction. Default: 1. This penalises rapid depth variation in the velocity update. Increase to suppress unrealistic high-frequency velocity changes with depth.
The weight applied to the normal-ray constraint equations in the solver system. Default: 1. The normal-ray equations enforce that the NIP normal rays are consistent with the velocity model. Increase this weight to strengthen the geometric constraint from the reflector dip information.
The weight applied to the V0 constraint equations in the solver. Default: 1. These equations pin the near-surface velocity to the specified V0 value, preventing the solver from updating the shallow velocity away from the known near-surface replacement velocity. Increase this weight if the inversion tends to produce unrealistic near-surface velocities.
When enabled, the residual norm at each local (inner) solver iteration is displayed on the convergence chart in addition to the global iteration residuals. Useful for diagnosing whether the inner solver is converging within each global step. Leave disabled for routine processing to reduce display overhead.
When enabled, includes a depth-derivative term in the local iteration solver equations to additionally constrain the vertical variation of the velocity update. May improve stability of the inversion in areas with poor vertical ray coverage.
When enabled, a spatial smoothing operator is applied to the velocity model after each global iteration before re-tracing rays. This prevents high-frequency artefacts from accumulating over multiple iterations and can stabilise convergence in areas with sparse pick coverage. Enable this when working with noisy or irregularly sampled MultiFocusing picks.
The smoothing radius applied along the inline direction after each iteration, in metres. Default: 300 m. Active only when Smooth velocity model after each iteration is enabled. Set to a value larger than the Variable step X/Y to ensure adequate smoothing of the velocity grid.
The smoothing radius applied along the crossline direction after each iteration, in metres. Default: 300 m.
The smoothing radius applied in the depth direction after each iteration, in metres. Default: 300 m.
The minimum depth (in metres) at which the tomographic solver updates the velocity model. Default: -9999 m (no shallow limit). Set to a positive value to protect the near-surface velocity from being modified by the inversion, for example when the near-surface is well-constrained by statics.
The maximum depth (in metres) at which the tomographic solver updates the velocity model. Default: 9999 m (no deep limit). Set to restrict the inversion update to a specific depth range, for example when only the shallow velocity needs updating.
The width of the ray corridor used to assign ray path contributions to velocity cells, expressed as a fraction of the cell size. Default: 0.2 (20%). Rays passing within this corridor of a velocity cell contribute to the tomographic update for that cell. A larger corridor widens the influence zone of each ray, improving spatial coverage but reducing spatial resolution.
This container defines the normalisation scales applied to the four types of equation in the solver system: travel-time arrival equations, normal-ray geometry equations, V0 constraint equations, and smoothness equations. Normalisation balances the influence of differently-scaled terms in the joint inversion. Leave at defaults unless the solver exhibits convergence problems related to scaling imbalance.
The normalisation scale (in seconds) for the travel-time arrival constraint equations. Default: 0.004 s (4 ms). This should be set to the approximate magnitude of the expected travel-time residuals.
The normalisation scale for the normal-ray geometry equations. Default: 0.1.
The normalisation scale (in m/s) for the V0 constraint equations. Default: 50 m/s. This should correspond to the acceptable uncertainty in the near-surface velocity.
The normalisation scale for the smoothness regularisation equations. Default: 1.
Parameters controlling the intermediate model display during processing, allowing the user to monitor the velocity model as it evolves across iterations.
The inline number of the 2D velocity slice displayed in the interactive monitoring view during processing. Default: 1. Set to an inline passing through the area of greatest interest for convergence monitoring.
The crossline number of the 2D velocity slice displayed during processing. Default: 1.
The half-width of the slab (in metres) extracted around the display inline/crossline for the intermediate visualisation. Default: 50 m. Traces within this distance of the display inline are averaged and shown.
When enabled, the velocity model is written to disk after each global iteration. This allows you to inspect the model evolution during the inversion and recover a partially converged model if processing is interrupted.
Contains the file path and naming prefix for the intermediate iteration output files.
The base file path for intermediate iteration output files when Save intermediate iterations is enabled. Iteration numbers and the prefix are appended to this path to form the individual filenames.
A text prefix prepended to each intermediate iteration filename to help identify files from this processing step.
Selects whether ray tracing and the solver run on CPU or GPU. GPU execution can significantly accelerate the ray-tracing computation for large 3D volumes.
When enabled, distributes the tomographic computation across multiple nodes in a cluster environment.
The number of model update blocks sent to each node per batch in distributed mode.
Sets the maximum number of CPU threads per compute node in distributed mode.
An optional text label appended to the distributed job name.
When enabled, allows manual specification of the CPU core affinity mask.
The CPU core affinity mask when Set custom affinity is enabled.
The number of parallel CPU threads used for ray tracing and the solver. Set to the number of available cores for best performance.
Optional shell scripts executed before or after the module runs, for example to set up data paths or copy output files to a project directory.
Path to a shell script executed before the module starts processing. Leave empty if not required.
Path to a shell script executed after the module finishes processing. Leave empty if not required.
When enabled, this module is bypassed and input data passes through unchanged. Use to temporarily disable the tomographic inversion without removing the module from the flow.
The output depth-domain interval velocity model produced by the ZOMF tomographic inversion. This model can be used directly as the velocity field for PSDM, as input to further tomographic refinement iterations, or for well-tie and depth conversion. The model is sampled on the grid defined by Variable step X/Y and Variable step Z parameters.