|
<< Click to Display Table of Contents >> Navigation: Tomography > Stereo tomography 3D(test) |
Stereo tomography 3D (test) updates a depth-domain velocity model using stereo tomographic inversion of locally coherent events extracted from prestack data. This module extends the 2D Stereo tomography solver to handle multi-line 2D surveys or full 3D geometries by operating on a multi-line tomo item container and an optional 3D tomo item. It can operate in three modes — 2D stand alone, 2D with intersection constraints, or full 3D — controlled by the Tomo type parameter. The module iteratively traces rays through the current velocity model, computes the misfit between predicted and measured travel times and slopes, and updates the model using LSQR inversion. This module is marked as a test version and may have limitations compared to the production Stereo tomography module.
Multi-line tomo dataset assembled by Set multi line tomo item. Contains the tomo parameters from all lines to be inverted jointly.
3D tomo item containing additional spatial constraints from a 3D dataset. Used when Tomo type is set to 3D.
Optional set of horizon picks used as hard constraints during inversion. Horizon-bounded layers prevent velocity updates from crossing geological boundaries.
Optional RMS velocity model used as a starting model or for initial velocity construction when the Dix conversion option is selected in the Create tomo params container.
Container grouping horizon-based layer constraint settings. When Use layer constraint is enabled, the inversion restricts velocity updates to the zone between the selected top and bottom horizons.
Top horizon of the constrained inversion zone. Select from the horizons loaded in the Horizons for constrains input.
Bottom horizon of the constrained inversion zone.
When enabled, the velocity update is restricted to the zone between Layer from and Layer to. Default: off.
Determines how multiple 2D lines and 3D data interact during inversion. Default: 3D. 2D stand alone inverts each line independently; 2D with intersection enforces velocity consistency at line crossing points; 3D jointly inverts all lines in a full 3D velocity volume.
Multi-line weighting factor that controls the relative contribution of inter-line constraints versus single-line inversion in the combined 2D/3D inversion. Default: 1.
Processing datum elevation (m) read from the input data. This value is not editable and is set automatically.
Velocity (m/s) used to fill the zone between topography and the processing datum in the output model. Default: 0 m/s (leave unfilled). Set to the air or water velocity if you need the datum-to-surface interval populated.
Number of outer tomographic update cycles. Default: 3. Each global iteration reduces the inversion grid spacing by the Sealing Factor, progressively resolving finer velocity detail. Three global iterations with the default Sealing Factor of 0.5 produce three resolution levels.
Ray tracing time step (s) used during forward modelling. Default: 0.01 s. Smaller values increase ray tracing accuracy at the cost of longer computation.
Number of localisation sub-iterations performed within each global iteration to refine the event-to-reflector association. Default: 10.
Number of LSQR inversion iterations within each global iteration. Default: 5. More iterations improve the quality of the velocity update per global cycle but increase computation time.
Weight of the spatial smoothness regularisation term in the inversion cost function. Default: 100. Range: 0.001–10000. Higher values produce smoother velocity models but may under-fit the data; lower values allow sharper velocity gradients but risk overfitting noise.
Number of standard deviations above the mean residual used to classify a tomo data point as an outlier. Default: 3. Range: 1–10. Points classified as outliers are down-weighted or excluded in subsequent iterations.
The global iteration number from which outlier rejection begins. Default: 1. Setting this to a value greater than 1 allows the first iterations to fit all data without rejection, which can help stabilise early convergence.
Maximum depth (m) of the output velocity model. Default: 4000 m. The model is truncated at this depth. Set to slightly below the deepest target reflector.
Minimum velocity (m/s) used as the starting model gradient value at the surface. Default: 1500 m/s. This is the V0 of the initial gradient velocity model used when Use Init Velocity is disabled.
Maximum velocity (m/s) at the base of the initial gradient model. Default: 5000 m/s. The starting model linearly interpolates between Initial velocity min at the surface and this value at Maximum depth.
Initial inversion grid cell size in the X (inline) direction (m) for the first global iteration. Default: 500 m. This grid is successively refined by the Sealing Factor in subsequent global iterations.
Initial inversion grid cell size in the crossline direction (m). Default: 500 m.
Initial inversion grid cell size in the depth (Z) direction (m). Default: 500 m.
Depth sampling interval (m) of the output velocity model. Default: 15 m. The inverted model is resampled to this interval before output. Smaller values produce finer output depth resolution.
When enabled, computes an alternative cost function alongside the main tomographic misfit for diagnostic purposes. Default: off.
Multiplicative factor applied to the inversion grid step between successive global iterations. Default: 0.5. Range: 0.1–5. A value of 0.5 halves the grid spacing at each global iteration, progressively resolving finer velocity structures. Larger values produce slower refinement; smaller values make more aggressive refinement.
Spatial smoothing applied to the velocity update at each iteration. Default: 1. Higher values produce a smoother model update.
Container grouping the LSQR iterative solver convergence parameters.
Condition number limit for the LSQR solver. Default: 10000. The solver stops if the estimated condition number of the system exceeds this value. Increase if the solver terminates prematurely on well-conditioned problems.
Percentage of the nominal LSQR iteration limit used as an early stopping criterion. Default: 70. The solver stops after completing this percentage of the maximum iterations if the residual is sufficiently small.
LSQR atol stopping tolerance — the relative change in the residual norm below which the solver is considered converged. Default: 0.000001.
LSQR btol stopping tolerance — the relative change in the right-hand side norm. Default: 0.000001.
Container grouping parameters that filter the tomo data points loaded from the tomo item before inversion.
Maximum number of tomo data points to use per line. Default: 5000. If the tomo item contains more points, they are decimated to this count. Reduce if memory errors occur; increase for denser coverage.
Minimum incidence angle (degrees) of events to include. Events with angles smaller than this value are excluded, which rejects near-horizontal ray paths that provide poor depth resolution.
Maximum incidence angle (degrees) of events to include. Default: 90 degrees. Exclude very steep rays by reducing this value.
Controls whether events within the Min angle/Max angle range are kept or removed. Default: None (no filtering). Leave retains only events in the angle range; Remove excludes events in the angle range.
Decimation factor applied when loading tomo data points. Default: 1 (no decimation). Set to 2 to use every second point, 3 for every third, and so on, to reduce computation time on very dense datasets.
Selects whether ray tracing and inversion run on CPU or GPU. GPU execution can significantly accelerate large 3D inversions.
Configuration for distributed (cluster) execution.
Minimum gathers per chunk during distributed execution.
When enabled, restricts CPU threads per node to the Number of threads value.
Optional text appended to the distributed job name.
When enabled, applies the Affinity mask to control CPU core assignment.
CPU affinity mask used when Set custom affinity is enabled.
Maximum CPU threads used during execution.
When enabled, this module is bypassed and data passes through unchanged.
Intermediate per-iteration depth velocity model gathers. Available for monitoring convergence during processing.
Final updated depth velocity model referenced to the common processing datum. Connect to Save seismic by gather to save for use in subsequent migration iterations.
Final updated velocity model in the original (topography) datum reference frame.
Depth velocity model for the currently active line, available for interactive display during processing.
Number of tomo data points actually used in the current inversion after all filtering and decimation steps.