Depth imaging
Kirchhoff Prestack Depth
An interval velocity model in depth, eikonal travel-time tables through it, Kirchhoff prestack depth migration into common image gathers — and an interactive session that updates the model from the gathers the last migration produced. Anisotropy is carried in the model, where it belongs.
Constant datum
The depth chain's reference surface
VTI & HTI
Thomsen parameters from the data
CPU or GPU
Distributed across the cluster
In view
Depth gathers & semblance
In depth imaging the velocity model is the product; the image is the evidence. Time imaging maps the subsurface well enough to work with, but its limits are structural, and depth migration is the step that puts reflectors where they physically are — at the cost of compute, of better algorithms, and of the iterations it takes to earn the model. The loop is short and it closes: build an interval velocity model in depth, compute travel times through it, migrate into common image gathers, read the residual moveout left in those gathers as the model's own error, correct the model and go round again. Where time migration starts from topography, the depth chain works from a constant datum — a difference to settle in the geometry QC before the first travel time is computed.
Velocity model building
Everything downstream rests on this model.
The depth model starts as a conversion of what time processing already knows, and is then refined by tomography, by picking on migrated gathers, and by the anisotropy the data itself reveals.
An initial model from time
The Vrms field from time processing becomes an interval velocity model in depth — through a tomographic inversion, or by a Dix-type conversion — optionally starting from a constant velocity gradient over a defined depth range.
Travel-time tables
Travel times are solved through the interval model — the eikonal equation, so lateral velocity change is honoured — and stored per source and per receiver. The table aperture is set wider than the migration aperture, and a depth step factor keeps 3D tables to a workable size.
Grid tomography
A regular grid of knots through the volume and an iterative least-squares inversion at each of them, with global iterations that refine a coarse grid into a fine one — so the large trends are settled before the detail is resolved. 2D in the X-Z plane, or the full volume in 3D.
Stereo tomography
Locally coherent events, each carrying two slopes as well as a travel time, are interpreted as pairs of ray segments that constrain velocity independently of continuous horizons. The events come from the imaging session as tomo items.
Anisotropy from the data
Thomsen epsilon is picked at the PSDM stage from migrations run at a range of velocity percentages; HTI delta is computed layer by layer from isotropic against check-shot thicknesses; and delta and epsilon together can be fitted to the residual moveout on depth-migrated gathers by least squares.
Salt and sharp contrasts
Picked horizons and a salt body collection constrain the inversion layer by layer, and the depth grid can be refined to resolve contrasts as sharp as a salt top or an unconformity, rather than smoothing across them.
Migration
Kirchhoff in depth, three ways to run it.
The integral solution to the wave equation, evaluated with Green's functions built from the travel-time tables: amplitudes are summed along the operator and mapped to a reflectivity model in depth. What changes between runs is the size of the job and what it is for.
Production
PreSDM with travel-time tables
The main pass: PreSTM or enhanced gathers plus the tables and a stack for header mapping, producing depth common image gathers. Aperture constant or varying with depth, an anti-alias coefficient and maximum frequency, a maximum angle aperture, and the option to image near offsets only from a dedicated near-offset table.
Offset mode
Offset-panel migration
The same algorithm executed panel by panel, each offset cube migrated independently. It keeps very large surveys tractable and gives offset-separated output for amplitude work. CPU or GPU, distributed across the cluster, with each node writing to local storage for better I/O.
Analysis
On-the-fly migration
For velocity work the raw shots are migrated on demand at the bin under analysis with the current model, so a layer's semblance can be recomputed without re-migrating the survey.
Velocity update & products
Read the gathers, correct the model.
Residual moveout on a depth gather is the model's error made visible. The imaging session is built around picking it, applying it and seeing what the correction did — and around producing everything the interpreter needs from the same run.
In the imaging session
What you do with the migrated gathers
Delta Vrms picking
Pick the residual on a semblance panel beside the depth gather, manually or automatically, at a chosen interval, then correct the velocity model from the picks and display the updated model against the old one.
Layer stripping
Update one layer at a time between picked horizons, with depth-moveout semblance computed for that layer and super-gathers to steady it where fold is thin, from the shallowest layer down.
Gather conditioning
Internal and external mute picking, residual moveout applied to flatten the gathers, and preprocessing performed inside the module — which keeps intermediate datasets off disk entirely.
Interpretation in place
Horizon picking, geobody modelling and well-log display in the same session, so the image is checked against the wells while the model is still being built.
Tomography hand-off
The session creates the tomo items the stereo-tomography inversion consumes, so picking and inversion are two stages of one loop rather than two projects.
Arbitrary lines
Draw a line across the location map and get its stack, at a bin spacing you choose — for a section along a structure rather than along the acquisition grid.
What the run produces
Images, gathers and conversions
Velocity
The updated depth interval velocity model — the run's real output, and the input to the next iteration or to depth-domain inversion work.
Stacks
A depth stack, a time stack and a time-to-depth converted stack from the same session, so the depth result can be compared directly against the time image it replaces.
Gathers
Processed common image gathers in depth and in time, plus their residual-moveout versions, for QC, further conditioning and amplitude work.
Angle products
Depth gathers converted to angle gathers by ray-parameter integration through the interval model, then stacked into common-angle volumes for AVO and AVA.
Domain conversion
Time to depth on an interval model with one-way travel-time tables, depth back to time by stretching, and a straightforward post-stack conversion on an RMS field when that is all that is needed.
Execution
CPU or GPU per run, distributed over the cluster; on GPU a large depth migration turns round of the order of thirty times faster than the same job on CPU.
Modules in this group
The depth chain in the g-Platform module tree
PSDM Imaging
Kirchhoff PreSDM (offset mode) — migration TT
Depth velocity updater
Time tables calculation
Grid tomography 2D/3D
Stereo tomography
VTI epsilon estimation
HTI delta estimation
Anisotropy Thomsen parameter estimation
V depth to angle
Angle stack
Convert time to depth data
Stretch depth to time
Velocity editor
Workflow
Six steps, and then round again.
1
Settle the datum
QC that sources and receivers sit on one constant datum and that the bin elevations are real values below it, with the elevation and datum headers annotated. Depth imaging runs from that datum, not from topography.
2
Convert the time model
Turn the Vrms field into an interval velocity model in depth, smooth and geologically plausible, over the depth range and sample interval the project needs.
3
Compute travel times
Solve travel times through that model for every source and receiver, with an aperture set wider than the migration aperture — and with delta and epsilon connected if the medium is anisotropic.
4
Migrate the first iteration
Run PreSDM with the tables to produce depth common image gathers. A decimation factor keeps the parameter tests quick before the full run is committed.
5
Pick the residual, update
Take the gathers into the imaging session, pick delta Vrms or strip layer by layer, apply the correction and inspect the updated model against the previous one — or feed the picks to grid or stereo tomography.
6
Iterate, then deliver
Repeat until the gathers are flat, then produce the final depth and time stacks, the conditioned gathers, the angle stacks, and the depth-to-time conversion the interpretation team works in.
More g-Platform capabilities
Depth is the last step of a long chain.
The same project, the same batch processing and the same cluster carry a survey from field tape through to depth. These are the other topics in the g-Platform workflow.
FAQ
Questions, answered.
What does depth imaging give me that time imaging cannot?
Position. Time imaging maps structure and stratigraphy well enough for most of exploration, but it carries limits that no amount of parameter work removes, and depth imaging is the step that gets closest to a true subsurface picture. The price is compute, better algorithms and time. One practical difference to plan for: time migration in g-Platform runs from topography, while the depth chain runs from a constant datum, so sources and receivers must be QC'd onto that datum and the bin elevations must sit below it before travel times are computed.
Where does anisotropy enter the depth chain?
In the velocity model, not in a separate migration algorithm. Travel-time tables are computed for an anisotropic medium as soon as delta and epsilon volumes are supplied. Those volumes come from the data: Thomsen epsilon is picked at the PSDM stage by migrating analysis points at a range of velocity percentages, HTI delta is derived layer by layer from isotropic against check-shot thicknesses, and delta and epsilon together can be estimated from depth-migrated gathers by least-squares fitting of the residual moveout.
How is the depth velocity model updated?
From the gathers the previous migration produced. In the PSDM imaging session you pick delta Vrms on a semblance panel beside the depth gather and apply the correction to the model, and you can drive a layer-stripping update that computes depth-moveout semblance for one layer at a time between picked horizons. Tomography carries the same information: grid tomography inverts on a knot grid, and stereo tomography inverts ray-segment pairs built from locally coherent events, using the tomo items the imaging session creates.
What products come out of a PSDM run?
An updated velocity model plus a family of images: the depth stack and the time stack, a time-to-depth converted stack, common image gathers processed in depth and in time, their residual-moveout versions, and angle stacks. Separate conversion modules move data the other way as well — time to depth on an interval model with one-way travel-time tables, depth back to time by stretching, and a simple post-stack conversion on an RMS field.
Get started
Build the depth model on your own survey.
From a converted time model through travel-time tables and PreSDM to gathers flat enough to stop iterating — take g-Platform for a trial run, or talk to Geomage about a depth project.