Sparse seismic acquisition offers cost and environmental benefits but often suffers from spatial aliasing. This paper investigates a deterministic, golden-ratio-based sampling strategy as a reproducible alternative to stochastic blue-noise designs. We evaluate this approach through a progression of numerical experiments: (i) an idealized acoustic twin-layer model for resolution testing, (ii) an intermediate visco-acoustic proxy with structural complexity, and (iii) a Marmousi benchmark with compressive sensing reconstruction. For the twin-layer configuration (1.5 m separation), golden-ratio layouts achieve a mean separation estimate of \(\bar{\Delta z}_{\text {gold}} = {1.30\,\textrm{m}}\) (relative error \(\approx 13\%\) ) compared to \(\bar{\Delta z}_{\text {reg}} = {2.00\,\textrm{m}}\) ( \(\approx 33\%\) ) for regular grids. In the visco-acoustic proxy at 30 % trace retention, golden sampling yields \(\textrm{SNR}\approx {6.1\,\textrm{dB}}\) and \(\textrm{SSIM}\approx 0.81\) , outperforming both regular ( \(\textrm{SNR}\approx {0.1\,\textrm{dB}}\) , \(\textrm{SSIM}\approx 0.30\) ) and jittered sampling ( \(\textrm{SNR}\approx {2.6\,\textrm{dB}}\) , \(\textrm{SSIM}\approx 0.55\) ). The Marmousi benchmark shows that golden-ratio layouts achieve statistically comparable performance to optimized jittered and Poisson-disk designs, with SSIM differences on the order of 0.003–0.007. The primary advantage lies in deterministic reproducibility and implementation simplicity rather than significantly superior reconstruction quality. These results suggest golden-ratio sampling as a practical alternative to stochastic designs, particularly for applications requiring repeatable acquisition patterns. Experimental Controls: All layouts were evaluated under identical modeling, noise, and reconstruction settings. Performance comparisons should be interpreted within the limitations of acoustic/visco-acoustic assumptions and stationary additive noise models.
