3D Seismic Survey Design, Spatial Sampling, and Wavefield Modelling

1. The Subsurface Bin Geometry and Partitioning Grid

In three-dimensional (3D) reflection seismic design, the surface topography overlying a geological asset is divided into a systematic matrix of small, discrete cells known as **bins**. These bins serve as the fundamental horizontal pixels of the final processed 3D image volume. Every active acoustic channel that records a seismic reflection trace assigns that data point to a specific bin based on where the reflection raypath's midpoint sits geographically.

Due to the fundamental physical principles of midpoint geometry, the natural size of a subsurface bin is precisely half the distance of the surface equipment layouts. The spacing chosen between inline detector channels defines the **Inline Bin Size**, while the spacing between parallel source execution lines sets the **Crossline Bin Size**.

Altering these dimensions directly scales your horizontal spatial resolution. Tighter intervals produce finer bin grids capable of distinguishing small, complex fault blocks, stratigraphic traps, or narrow channel sands. However, altering this grid density immediately restructures the data properties across the entire survey area.

2. Mechanics of Fold of Coverage and Trace Redistribution

**Fold of Coverage** describes the data redundancy within a survey design. It signifies the number of individual seismic trace raypaths that sample the exact same subsurface bin location from different source points and surface offsets. Redundancy is the primary defense against data ambiguity; higher fold yields multiple look-angles at a geological reservoir, which helps resolve depth structures.

3. Stacking Laws, Destructive Interference, and Desired S/N Targets

The reduction of fold is not merely a statistical issue; it deeply undermines the raw image quality through processing mechanics. Raw seismic records are heavily contaminated by ambient environment noise, including wind shear, marine swell action, cultural traffic, ground roll, and scattered wave multiples. To uncover weak geological boundaries beneath this noise, processing engines rely on **CMP Stacking**.

During the stacking phase, all traces allocated to a single bin are corrected for timing offsets and summed together. Because true geological reflections are coherent, they align and amplify constructively. Conversely, random ambient noise carries uncorrelated phase alignments and cancels out destructively.

The mathematical physics governing this attenuation state that random noise is reduced by the square root of the fold count. If a target exploration layer is buried beneath a highly scattering, low-velocity overburden (such as thick volcanic basalt sheets, glacial tills, or shifting desert dunes), achieving your **Desired S/N** threshold requires a very high stacking fold. Shrinking the bin dimensions without adjusting field effort starves the cells of necessary traces, resulting in a noisy, uninterpretable final volume where structural event paths are masked.

4. Maximum Offset and Maximum Minimum Offset (LM_os)

**Maximum Offset** represents the distance between the source station and the furthest active receiver channel utilized in the stack. Correctly selecting this length is paramount to resolving deeper horizons, tracking sound velocity variations, and providing reliable amplitude information for advanced rock-physics processing.

Conversely, the **Maximum Minimum Offset ($LM_{os}$)**—or Largest Minimum Offset—defines the shallowest horizon your survey can successfully illuminate. While in a perfect 2D linear seismic survey near-offsets are close to zero, 3D layouts feature widely spaced receiver lines. If a source is discharged directly in the center of the gap between lines, the distance to the nearest active sensor creates an initial tracking blind zone.

During processing, wide-angle distortions are discarded via a **Shallow Mute**. If your 3D survey lines are laid out too far apart, the resulting $LM_{os}$ becomes larger than the depth of your shallow target ($LM_{os} > Depth_{shallow}$). Once muted, **zero data traces are left** within the inner bins, leaving shallow horizons completely unmappable.

5. Temporal Sampling, Recording Length, and the Nyquist Frequency

Seismic instruments convert returning analog acoustic waveforms into digital information by reading amplitude levels at precise time increments called the **Sampling Rate ($\Delta t$)**, usually measured in milliseconds (1ms, 2ms, or 4ms).

To reconstruct a wave without distortion, the sampling rate must satisfy the **Nyquist-Shannon Sampling Theorem**. The absolute highest frequency that a digital system can reconstruct without aliasing is the **Nyquist Frequency**:

If your incoming signal contains frequencies higher than $f_N$, those waveforms wrap around and masquerade as low frequencies (**Temporal Aliasing**). Simultaneously, the total **Recording Length ($T$)** represents the listening window duration. It must remain open long enough to capture the two-way travel time of the deepest target, plus any extra time added by far-offset geometric travel paths.

6. Migration Aperture Fringe Boundaries

When imaging complex subsurface geology, reflections do not always return from directly beneath the surface position. For dipping layers or truncated fault blocks, the reflection points are laterally displaced. Modern migration processing focuses scattered wave energy back to its true coordinate origin, but it cannot position data that was never recorded in the field.

To record these outward-scattering wave paths, the surface recording grid must be padded out past the structural boundaries of the reservoir asset. This extra safety margin is the **Migration Aperture ($W_a$)**:

Where *Depth* equals the true vertical depth of the target horizon layer, and *θ* represents the maximum structural dip angle. Truncating this perimeter fringe to save costs leaves processing algorithms starved of wide-angle input, introducing severe edge smears or "migration smiles."

7. Numerical Ray Tracing and Elastic Wave Modeling

Before deploying expensive hardware in the field, geophysicists run advanced pre-survey numerical simulations to stress-test their design metrics.

**Asymptotic Ray Tracing** is a high-frequency approximation based on optical principles. It models sound energy traveling as thin directional pathways ("rays"). By tracing thousands of rays through a digital model, designers run illumination workflows to locate structural shadows and adjust surface layouts before acquisition begins.

Conversely, **Elastic Wave Equation Modeling** uses finite-difference calculations across a cellular grid to compute full wavefield propagation without shortcuts. Because it treats the earth as an elastic solid rather than an acoustic fluid, it tracks both compressional waves (**P-waves**) and shear waves (**S-waves**). This high-fidelity simulation produces realistic synthetic data streams that resolve complex effects like surface ground roll, fault-tip diffractions, and mode conversions.

8. Case Study: The Sinopec "I" Geophysical Experiment

The real-world implementation of these integrated principles is clearly visible in landmark industrial tests like the **Sinopec "I" Geophysical Experiment**. Deployed over highly challenging geological basins (such as the Tarim or Sichuan basins), this project targeted ultra-deep reservoirs located 7,000 to 8,000 meters down, masked by highly scattering near-surface overburdens.

To resolve fine fractures and karst systems, Sinopec deployed an extreme high-density spatial framework. They pushed natural bin sizes down to $12.5\text{m} \times 12.5\text{m}$ (using $25\text{m}$ receiver station spacing) to combat spatial aliasing along deep faults. They offset the resulting fold dilution by scaling up trace redundancy into the hundreds, boosting the signal-to-noise profile via stacking laws. Long-offset profiles exceeding $8,000\text{m}$ were matched with narrow crossline line paths, balancing maximum deep velocity analysis with near-offset density ($LM_{os}$) to completely preserve shallow structural data.