Geometry assignment & QC — g-Platform
Geometry assignment / QC

Seismic acquisition QC software

Merge field navigation into the trace headers of land or marine 2D and 3D data, then prove it. g-Platform surrounds the assignment with connected tables, maps and gather views, header crossplots, first-break and azimuthal checks, and coverage QC after binning — and lets you correct what they find without leaving the module.

Download free trial How assignment works Watch Demos
Land & marine
2D and 3D, one module
Tables, maps, gathers
Every view stays in step
Fix in place
Move points, edit traces
A g-Platform source gather whose theoretical first-break curve does not follow the recorded arrivals, beside the source-receiver location map and the shot-offset QC map
The same g-Platform gather after the source was repositioned on the location map, with the theoretical curve now sitting on the first arrivals
Geometry, fold and offset QC maps in g-Platform, with a fold-of-coverage map, offset panels and an azimuth rose diagram
In view
Before geometry correction
Geometry is the one step nothing downstream can repair. Get it wrong and every module after it runs on trace headers that describe a survey which was never acquired — and the error arrives, invisibly, as a degraded image. So the work divides in three: read the headers off the field data, load the navigation that says where the sources and receivers actually were, and merge the two into the trace headers. The merge is the easy part. The reason g-Platform puts an interactive workspace around it is the fourth thing, which has no natural end: proving the result before the survey moves on.
Assignment

One workspace where every view knows what you clicked.

Assignment fills the trace headers from the navigation files; binning comes later as its own step. What makes the module worth sitting in is that its tables, maps and gathers are one interconnected display — select a point in any of them and the rest follow.

Three steps, then QC
Read the traces and their headers, load the source and receiver coordinates and the relation between them, then merge the two. The assignment module runs on its own rather than inside a seismic loop.
Land and marine alike
The same module assigns geometry on 2D and 3D, onshore and offshore. Land brings spread configurations and observer logs, sometimes handwritten; marine brings processed towed-streamer navigation with its own conventions.
Tables that show the mismatch
One table lists every disagreement between the field data and the navigation file. Others carry the spread, the raw field-file identifiers, and the source and receiver inventories — where a shot or a receiver can be killed and, later, revived.
Everything stays in step
Click a row in a table or a location on a map and that point is highlighted or displayed in every other view at once. It is the difference between hunting for a suspect shot and simply looking at it.
Move a point, watch it fit
Drag a source or receiver to a new position and the theoretical first-break curve on its gather updates as it moves. Line the curve's peak up with the nearest arrivals and the position is right; the old marker stays visible beside the new one.
Trace editing on the gather
Mark traces dead, reverse their polarity, or flag them as unused or auxiliary by dragging across them. Nothing is destroyed until the dataset is saved — and the edits survive only if the assignment module is not re-run afterwards.
Crossplot any three headers
Put one trace header on the horizontal axis, another on the vertical and a third on colour, with the matching columns highlighted in a table beside the plot. A large-data mode averages the points into a grid so a survey of millions of traces still responds.
Repair headers, keep the originals
Correct trace headers with expressions and equations while the original values are kept as a copy rather than overwritten. A companion step writes a per-gather minimum, maximum or average back onto every trace in the group.
QC checks

Six ways to make a wrong position look wrong.

A bad coordinate is invisible in a header table and obvious the moment the data is asked to agree with it. Each of these takes a different route to the same question — do the recorded traces behave as though the sources and receivers were where we say they were?

First breaks
Geometry by first-break arrivals

The first-break hyperbola expected from an approximate near-surface velocity is drawn against automatically picked arrivals in every common-point gather. Browse them one by one, or play the survey as a slide show at a delay you choose, watching for the gathers where the two disagree. A collection view shows a grid of neighbouring gathers together, which is how a locally confined error separates itself from a survey-wide one.

Misfit maps
Misfit statistics and channel fixes

The same disagreement, turned into numbers: peak and average misfit per source, regression misfits fitted per receiver line, counts of bad lines, and the total and maximum trace shift within a line — all coloured onto a map in survey or line-and-station coordinates. Sources can be repositioned by dragging, optionally snapping to a reference grid or the surveyed receiver points, and receiver channel mis-assignments corrected automatically against a threshold.

Azimuthal stack
Azimuthal coherency test

Traces at each acquisition point are grouped into azimuth sectors, moveout-corrected and stacked, giving one panel per point across the whole survey. Swapped coordinates and mis-assigned shot numbers show up as sectors that will not stack coherently.

Panel fold
Fold per offset-azimuth panel

Fold computed for each offset or azimuth panel rather than for the survey as a whole, so coverage can be confirmed tile by tile before migration. Click a bin on the fold map to see the sources and receivers that contributed to it.

Binning
Coverage after binning

Fold of coverage, minimum and maximum offset in a bin, bin centre to midpoint centroid distance, a rose diagram for azimuth, grid statistics, and a smoothed CMP topography surface written to the headers for velocity work and MultiFocusing.

Attributes
Amplitude & frequency maps

Attributes computed inside a signal window and a separate noise window — maximum, minimum and RMS amplitude, mean absolute amplitude, standard deviation, dominant frequency and spectrum width — coloured onto the location map with the amplitude spectrum beside it, per gather group or per trace.

Workflow

Six steps from field tape to geometry you can defend.

1
Read the field data
Load the traces and their headers from SEG-Y, or convert the field SEG-D first. Everything the assignment needs from the seismic side comes out of this step.
2
Load the navigation
Source and receiver coordinates and the relation between them — from SPS, SEG-P, UKOOA P1/90 or ASCII — plus the spread from an observer log where the field crew recorded it there.
3
Assign the geometry
Merge navigation and trace headers. Set an approximate moveout velocity for the QC curve, tell the module where the spread information lives, and flag the dead and auxiliary trace identifiers your recorder used.
4
Hunt the errors
Work the error table, walk the shots and receivers automatically or one at a time, crossplot the headers, and let the first-break and azimuthal checks find what the tables cannot.
5
Correct in place
Reposition points against the moveout curve, fix channel mis-assignments, kill what is unusable, edit the traces that need it — and do not re-run the assignment afterwards, or the edits go with it.
6
Bin, check coverage, save
Bin into CMPs, read the fold and offset maps and the rose diagram, then save the data with its corrected headers. From here the survey is ready for the processing sequence proper.
Controls

The settings that decide what the QC can see.

Most of these are about scope rather than algorithm: how much data a check reads, how wide a neighbourhood it compares against, and which traces it is allowed to believe. Set them too tight and a real error hides; too loose and everything looks anomalous.

Assignment settings
What the module does with your field data
QC velocity
An approximate moveout velocity for the theoretical first-break curve. It is a QC overlay only — it is never applied to the data.
Spread source
Take the spread from the trace headers when the navigation format already carries it, or from a separately loaded observer log when it does not.
Channel numbers
Channel numbering can be rebuilt from scratch where it is missing or wrong — deliberately a last resort, because it overwrites what the recorder wrote.
Trace identifiers
Tell the module which trace identifier marks a dead trace on this recorder, and which identifiers mark auxiliary channels, so neither is mistaken for signal.
Receiver tables
Optional, and worth thinking about on a large 3D: building them costs time and memory in proportion to the survey.
Point snapping
Constrain a relocated source to a reference grid, or a relocated receiver to the surveyed point list, so a correction stays consistent with the field deployment.
QC settings
How far each check looks
Crossplot axes
Horizontal, vertical and colour headers, with an optional grid-averaging mode and its own cell counts for datasets too large to scatter point by point.
Collection view
How many gathers appear side by side and stacked, and how far down the record the display runs — shorten it to zoom in on the first-break zone.
Offset window
A minimum and maximum offset for the misfit calculation, or a per-receiver-line regression mode with its own maximum offset spread.
Pick cleaning
Energy window length for the automatic first-break detection, plus relative and absolute median thresholds and the spatial apertures they are measured over.
Scan extent
How many consecutive shot gathers a QC pass reads — a small number for a fast test on a large 3D, the full count for the real check.
Display sort
Leave traces in file order or sort them by signed or absolute offset; sorting by offset is what makes the moveout trend immediately readable.
Modules in this group
The assignment and QC modules named on this page — not the full list
Geometry application QC trace geometry Binning 3D Check geometry by first breaks QC geometry by FB Geometry QC azimuthal QC panel fold QC Attributes QC attributes per trace Headers manipulation Header statistics
More g-Platform capabilities

Geometry is where the sequence either starts well or does not.

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.

g-Platform overview Data & geometry formats First-break picking Refraction statics Residual statics Velocity analysis Stack imaging Signal processing & denoise Multiple attenuation Deconvolution Time imaging Depth imaging FWI & RTM VSP processing
FAQ

Questions, answered.

Why is geometry assignment treated as the most important step?
Because nothing downstream can correct it. Assign the wrong geometry and the whole sequence runs on trace headers filled with inaccurate positions, which ends in an inaccurate subsurface image no amount of later processing recovers. That is why the assignment stage is worth spending real time on, and why the QC after it matters as much as the assignment itself — the point is not to apply geometry but to be able to show that the geometry you applied is right.
How does g-Platform find a geometry error?
Several ways, because different errors betray themselves differently. A table of errors lists outright mismatches between the SEG-Y and the navigation file. A theoretical first-break curve drawn from an approximate near-surface velocity shows, gather by gather, whether the arrivals fall where the assigned positions say they should. Stacking each acquisition point by azimuth sector after a moveout correction turns a swapped coordinate or a mis-assigned shot number into an incoherent panel. Crossplotting any three trace headers against each other exposes the patterns a table hides. And misfit statistics computed from first breaks — peak, average, per-receiver-line regression, bad-line counts — can be coloured straight onto a map.
Can positions be corrected without going back to the field data?
Yes. A source or receiver can be dragged to a new position on the location map, and the theoretical first-break curve on its gather updates as it moves, so you can position it until the curve sits on the real arrivals. Relocation can be constrained to snap to a reference grid or to the surveyed receiver points rather than left free, receiver channel mis-assignments can be corrected automatically against a threshold, and shots or receivers can be killed and revived from the tables. Old and new positions are both kept, and the changes are recorded in a table you can carry downstream. One rule matters: after editing, do not re-run the assignment module — it rebuilds the headers and the edits are lost.
What is checked after binning?
Coverage, mostly. Binning produces a fold map, the minimum and maximum offset in each bin, the distance from bin centre to the centroid of the midpoints that fell in it, a rose diagram for the azimuth distribution and summary statistics for the grid — plus a smoothed CMP topography surface written to the headers for velocity work and MultiFocusing. Before migration there is a second, finer check: fold computed per offset or azimuth panel rather than for the survey as a whole, so a tile that is short of traces is caught while it can still be fixed.
Get started

Prove the geometry before you process it.

Load the navigation, assign it, and put the survey through the checks until nothing is left to explain. Talk to Geomage about a demo, or take g-Platform for a trial run.

Download free trial Request pricing