Survey design sanity: fold, offset, azimuth

Every processing decision you will make later is constrained by what the acquisition team recorded. The best algorithm on Earth cannot produce an azimuth-dependent AVO result from a single-azimuth survey, and no deconvolution saves you from undersampling. Survey design is where many of the hardest processing problems are solved — or baked in permanently.

1. The three dimensions of trace coverage

For each CMP bin in your survey, three quantities decide what processing can do:

• Fold — the number of traces in the bin. Drives stacking SNR: amplitude gain ∝ √fold (from §0.8).
• Offset distribution — range of source-receiver distances in the bin. Wide range → can measure AVO (amplitude vs angle) and do velocity analysis; narrow range → can only stack.
• Azimuth distribution — spread of directions in the bin. Wide coverage → can measure anisotropy (HTI fractures, VTI shales), run wide-azimuth FWI, and detect stress orientation. Narrow azimuth → near-isotropic imaging only.

The widget lets you switch between three canonical designs and click any CMP bin to see what traces reached it.

2. What each preset tells you

• Narrow-azimuth 2D. One receiver line, one shot line. Every CMP is a single 1D collection of offsets. Azimuth is essentially zero. Can do 2D stacking and AVO. Cannot do anisotropy, wide-azimuth FWI, or anything cross-line.
• Wide-azimuth 3D (orthogonal). Shot lines and receiver lines cross at right angles. Every CMP in the interior sees trace coverage from every direction — ideal for isotropic imaging, AVO, anisotropy, and FWI. The expensive option, but the one that unlocks the most downstream work.
• Skinny 3D swath. A compromise: multiple shot/receiver lines, but narrow cross-line extent. Reasonable fold for imaging, adequate offsets for AVO, limited azimuth coverage. Use when the target is a known structural play and wide-azimuth is unaffordable.

3. Fold: the SNR multiplier

In the widget, maximum fold varies widely between designs. More fold = better SNR, but at a cost proportional to shots × receivers. The survey designer’s question is: what minimum fold is required for the target to appear above noise at the needed resolution?

A common rule of thumb:

• Fold 20–40 is typical for shallow land and marine imaging.
• Fold 60–120 is standard for deep-water or difficult imaging targets.
• Fold 200–500+ for tough subsalt and FWI-centric programs.

4. Offset: the AVO and depth-of-investigation lever

The longest offset in a CMP bin limits the maximum angle of incidence on a target reflector:

For a reservoir at 2500 m depth and a survey with 5 km maximum offset, θ_max ≈ 45° — enough for full AVO class discrimination. At 3 km max offset, θ_max ≈ 30° — marginal. Offsets shorter than reservoir depth leave you with near-offset only — no AVO.

5. Azimuth: the anisotropy lever

Fractured carbonates, stressed shales, and any HTI (horizontally transversely isotropic) medium have velocities that depend on which direction you look. Two surveys over the same reservoir, one wide-azimuth and one narrow, produce different images of a fractured zone: the wide one shows the fracture orientation as an azimuthal velocity or amplitude anomaly; the narrow one averages it out.

A standard quality metric is azimuth coverage: what fraction of the 360° of azimuth bins (typically 18 × 20° sectors) is populated with enough fold to analyze? The widget’s info strip reports this.

6. Edge effects

Any survey has edges. Near the edges, fold drops, offset range narrows, and azimuth coverage becomes lopsided. Processing on edge bins is unreliable; most surveys oversize the acquisition footprint so the image area sits comfortably inside the full-fold region.

7. What you cannot get back in processing

• Bandwidth you never recorded. A marine survey at 10 m tow has a notch at 75 Hz; no processing recovers those frequencies.
• Offsets beyond your maximum. Far offsets needed for AVO must be acquired; interpolation does not reach beyond the last trace.
• Azimuths never sampled. If every source-receiver pair was aligned E–W, processing cannot deduce N–S anisotropy.

What processing can do is minimize the penalty from irregular sampling (§2), fill small gaps with interpolation (§9.3), and optimize the imaging radius for what you have. But the ceiling is set in the field.

Where this goes next

Part 1 is complete. You now know what land and marine acquisition each record, how traces get sorted into CMP bins, what noise to recognize on a shot gather, and how survey design decisions cascade into every later step. Part 2 starts the actual processing: reformatting the data, repairing the amplitude path, applying the first statics corrections, and building the deconvolution operator.

References

• Sheriff, R. E., Geldart, L. P. (1995). Exploration Seismology (2nd ed.). Cambridge UP.
• Yilmaz, Ö. (2001). Seismic Data Analysis (2 vols.). SEG.
• Claerbout, J. F. (1976). Fundamentals of Geophysical Data Processing. McGraw-Hill.