Module 06 · Horizontal resolution in practice

Where does my channel really end?

Module 04 shrank two points down until they became one. Here is the same limit acting on something you would actually map: a channel that narrows downstream and splits into two distributaries. Watch the point where it appears to stop, and the point where the two branches stop looking like one. Neither of those places is in the geology — they are in the acquisition.

Two ways a feature goes missing

Module 04 slid two point features together until they imaged as one. The same limit applies to a mapped body in two separate ways.

Too narrow. A channel narrower than the resolution length does not image as a blurred channel. It disappears, and the map shows it stopping.

Too close together. Two branches nearer than the resolution image as one channel. The map shows a single channel that forks late, when it actually forked much earlier.

Why both failures are hard to spot

Neither failure leaves a visible sign on the map. There is no blurred patch and no marked edge of the data. A channel that drops below the resolution length looks the same as a channel that ends.

The boundary drawn is therefore partly a property of the survey rather than of the rock. Reprocess with more bandwidth or a wider aperture and the same channel gets longer.

The page opens on unmigrated data, which is where every survey starts. Read what can be said from it before pressing 3D.

The channel

Trunk width 900 m

Width at the branch tips 0 m

How far the branches spread 700 m

The data

Two-way time to the channel 1.80 s

Peak frequency 22 Hz

Velocity 3000 m/s

Migration

Migration aperture 1200 m

Mark where it appears to end Outline the true channel

Display

Units

The channel, and the channel you would map

Click either map to move the section line

1 The channel that is there sand, everywhere inside the outline

2 The channel the seismic shows after imaging

Across the channel at your section line

Grey is the truth, crimson is what you would see

Wavelength λ = V / f— m

Fresnel zone across— m

Resolution after imaging— m

Channel appears to end at—

True channel continues to—

Branches only separate beyond—

At the section line: true width—

At the section line: separation—

Verdict—

Try this — and what to watch for

Look at the map before changing anything. The point: it is the starting point for every survey, before processing. The page opens on unmigrated data. The channel is a broad smudge. It supports a statement that a channel runs through here, and roughly which way, and little else — not where it ends, and not that it forks at all. Press 3D and compare.
With 3D on, compare the two maps. The point: the gap between them is reservoir you would never post. Where does the channel stop in each? The imaged channel fades out well before the real one does. Everything downstream of that point is still sand, still the same thickness, and completely absent from your map. If you were mapping the extent of a reservoir from this, you would draw the boundary in the wrong place and never know.
Find where the branches become two. The point: it is how a distributary system gets mapped as a single channel. Drag the section line upstream from the far end and watch the profile. Far downstream the two branches sit well apart and read as two. Move upstream and they close until the trough between them fills in, and from there back to the split they image as one wide channel. The map shows a single channel that forks once, when in fact it forks much earlier.
Take it deeper. The point: the same body gets shorter on the map purely by being deeper. Move the channel from 1.2 s to 3 s. Resolution worsens and the apparent end of the channel marches upstream. The same body, unchanged, gets shorter on the map as it gets deeper. Comparing feature sizes between a shallow field and a deep one is not comparing like with like.
Raise the frequency. The point: bandwidth is the only control here that improves the imaged extent. Push the peak frequency up and watch the apparent end move. The apparent end moves downstream. The migrated resolution length improves in proportion to frequency, while the unmigrated Fresnel zone shrinks only as its square root.
Reduce the aperture. The point: features near the edge of a survey are imaged less reliably than those in the middle. Pull the aperture down to 500 m. The image degrades toward the unmigrated case. Near the edge of a survey the migration has energy from only one side to gather, so features that are well resolved in the middle of the survey are lost at the margin.
Switch to 2D. The point: on a 2D grid the shape you map depends on which way the boat sailed. Resolution is good along the line and poor across it, so the channel smears in one direction only. Whether you can see the branches now depends on which way they happen to trend relative to your line — a geological conclusion decided by the survey orientation.

If you remember nothing else

Features narrower than the resolution do not go blurry, they go missing. A narrowing channel appears to end, and the ending looks convincing.
Two things closer together than the resolution image as one thing. Branches, stacked channels, closely spaced faults — all merge.
The limit is roughly λ/4 after good 3D migration, and the width of the Fresnel zone without it, which is often ten times worse.
Boundaries you map are partly a property of the survey. Reprocess with more bandwidth or a wider aperture and the channel gets longer, with no new rock involved.
Absence of evidence is weak evidence here. "The channel ends" and "the channel narrows below what we can see" look identical on a map.

Method, assumptions, and where to read more

The channel is drawn as a map of amplitude: uniform inside the body, zero outside, so that thickness and tuning are held out of it and only horizontal resolution is on trial. Module 07 does the opposite and holds width fixed while thickness varies.

Imaging is done as a low-pass filter in the wavenumber domain, using the two-dimensional FFT: spatial detail finer than the resolution length is removed, everything coarser is kept, with a raised-cosine taper across the cutoff. That is what a band-limited image is. The cutoff is calibrated so that two features separated by exactly the stated resolution are on the point of separating — checked against the model across a factor of eight in scale, where the split occurs between 0.96 and 1.02 times the quoted number.

The resolution length itself is the same as module 04: the Fresnel zone radius (V/2)·√(t/f) without migration, and λ/(4 sin θ) with it, where θ is the largest angle the aperture allows. In 2D the filter is applied along the line only, so the cross-line direction keeps the Fresnel limit.

Left out: noise, tuning, velocity error, dip, migration artifacts, and the fact that a real channel has a fill whose properties change along it. Every one of them makes the edge harder to find, not easier.

Brown, A.R., 2011, Interpretation of Three-Dimensional Seismic Data, 7th ed.: AAPG Memoir 42 / SEG Investigations in Geophysics No. 9.
Chen, J., and Schuster, G.T., 1999, Resolution limits of migrated images: Geophysics, 64, 1046–1053.
Lindsey, J.P., 1989, The Fresnel zone and its interpretive significance: The Leading Edge, 8, 33–39.
Zeng, H., and Hentz, T.F., 2004, High-frequency sequence stratigraphy from seismic sedimentology: applied to Miocene, Vermilion Block 50, Tiger Shoal area, offshore Louisiana: AAPG Bulletin, 88, 153–174.
For anything unfamiliar, the SEG Wiki hosts Sheriff's Encyclopedic Dictionary of Applied Geophysics free at wiki.seg.org/wiki/Dictionary.

Next up

07 · Does bright mean thick — or does it mean gas?

You have just seen what happens to the shape of a feature in map view. The next module goes after its brightness: a sand thinning onto a salt dome, a bright rim around the flank, and the oldest question in amplitude interpretation — hydrocarbon, or just tuning?

Open module 07 → All modules