Abstract
Modern-day reservoir management and monitoring of geological carbon storage increasingly call for costly time-lapse seismic data collection. In this letter, we show how techniques from graph theory can be used to optimize acquisition geometries for low-cost sparse 4D seismic. Based on midpoint-offset domain connectivity arguments, the proposed algorithm automatically produces sparse non-replicated time-lapse acquisition geometries that favor wavefield recovery.
Introduction
Monitoring geological carbon storage and managing oil reservoirs requires time-lapse (4D) seismic data, which involves comparing "baseline" and "monitor" surveys over time. However, the high cost of full-density seismic data collection often necessitates sparse sampling. Traditionally, 4D seismic has relied on the "repeatability paradox"—the attempt to replicate exact source and receiver locations across surveys. This is not only difficult to achieve in the field but may also be suboptimal for modern joint inversion techniques that can leverage shared information between non-replicated surveys.
Methodology: Graph Theory and the Spectral Gap Ratio
To design efficient sparse surveys without the need for expensive wavefield simulations, the authors introduce a "simulation-free" approach based on graph theory. The core metric is the Spectral Gap Ratio (SGR), defined as the ratio of the first two singular values of a binary sampling mask ($\sigma_2/\sigma_1$).
In the midpoint-offset domain, a smaller SGR indicates better "connectivity" in the sampling graph. This connectivity serves as a mathematical proxy for the quality of wavefield recovery: a well-connected graph allows for more stable interpolation and reconstruction. By minimizing the SGR, the algorithm identifies sampling patterns that are inherently better suited for inversion.
Optimization via Simulated Annealing
The authors employ a Joint Recovery Model (JRM), which treats 4D data as a combination of a "common component" (shared information) and "innovations" (changes over time). The optimization objective is to minimize the SGR of the baseline survey, the monitor survey, and the union of both.
Using simulated annealing, the algorithm "wiggles" sampling points to find an optimal configuration that satisfies practical constraints, such as maximum gap size. As shown in the optimization trace, reducing the SGR objective directly correlates with a significant boost in the reconstructed Signal-to-Noise Ratio (SNR).
Key Results and Discussion
The methodology was validated using a realistic synthetic 2D dataset. The optimization successfully reduced the SGR from approximately 0.346 to 0.262. This mathematical improvement translated into a dramatic jump in wavefield recovery quality, with the SNR rising from 6.6 dB (using standard jittered sampling) to over 17 dB.
Diversity over Replication
A critical insight from this work is that non-replicated survey geometries are superior for joint inversion. By sampling different locations in the baseline and monitor surveys, the union of the two (the common component) becomes much denser. The optimized masks naturally reduced the overlap between source locations from 100% (replication) to approximately 22%, proving that diversity in sampling provides more information for the JRM to exploit.
Conclusion
By using the Spectral Gap Ratio as a proxy for recovery quality, this approach allows for the automated design of low-cost, high-performance 4D seismic surveys. The results challenge the traditional reliance on repeatability, suggesting that optimized, non-replicated sparse sampling is a more effective path forward for long-term reservoir and carbon storage monitoring. Future work will look to extend these graph-theoretic principles to 3D geometries and multi-monitor survey scenarios.
Summary updated