Compressive Seismic Acquisition - Random vs Optimized

Compressive Seismic Acquisition (CS-A) is a non-uniform survey design technique that solves spatial aliasing issues on typically designed uniform seismic surveys. This article compares two different CS-A design implementations: random and optimized.

Random

The random CS-A design used in this comparison (from a popular survey design software package) tackles spatial aliasing by moving each source or receiver point by a random constrained distance from the original location. This increases the probability of greater wavenumber coverage and reduced spatial aliasing for a given bin size. In this case, the maximum radial distance to move a point was constrained to 20 m.

Optimized

The optimized solution is In-Depth's Regular Indexed CS-A design (US Patent No. 10317542). This design increases wavenumber coverage and minimizes spatial aliasing for a given bin size. Given the survey constraints, CS-A uses an objective function and many iterations to determine which realization will provide the best possible reconstruction. This optimization considers the shape and position of obstacles beforehand, so it will cluster the survey nodes around obstacles to maximize reconstruction quality in the affected zones.

Survey setup

Both random and optimized designs start from the same conventionally designed uniform seismic survey. The number of sources and receivers was kept constant for a fair comparison. Several versions of the random design were produced. Different random results can be obtained for each realization since the random design is not optimized to fit an objective function.

To judge the efficacy of each design, In-Depth Mutual Coherence (MC-) maps (US Patent No. 10156648) or "MC-maps" were generated separately for each result, source, and receiver grid.

For an MC-map QC, the smaller the number (0~1), the better it will be reconstructed in processing (1 or red is the worst, 0 or blue is the best).

Results

Conclusion

The optimized design is clearly superior to the random design. When you think about it, each realization of a random design is simply a starting point for an optimized design. With that optimization, the probability of obtaining an ideal design by chance is much higher. Simply put, without the benefit of an objective function and numerous iterations, a random design cannot compete with an optimized design.

If you want to see this MC-map QC and In-Depth's patented CS Survey Conversion in action, these tools are now available in TesserACT® and MESA®.