Uncertainty Quantification of Net‐to‐Gross and Frequency Effects on AVA Class III Discrimination in Thin‐Bedded Heterolithic Reservoirs

Decreasing net‐to‐gross (NTG) ratios systematically degrade Class III amplitude variation with angle (AVA) anomalies, creating a critical risk where gas‐saturated sands become seismically indistinguishable from background shale. This study quantifies this risk through a stochastic forward modelling workflow representing densely interbedded, heterolithic reservoirs such as distal turbidites and tidal heterolithics. The earth model uses a brine‐saturated sand baseline, with gas‐sand properties derived via Gassmann fluid substitution for physically consistent elastic parameterization. Fine‐scale (0.1 m), uncorrelated assignment of sand and shale mimics sub‐seismic and sub‐logging heterogeneity, a geologically plausible end‐member for parasequence architectures of thin beds whose stacking creates composite reservoir intervals below seismic tuning thickness. Using this model with a depth‐dependent compaction trend, we generated 5000 stochastic AVA realizations across five NTG levels (100%–20%). Intercept–gradient analysis confirms that as NTG decreases, the AVA response weakens and clusters converge toward the shale trend, with average reservoir porosity declining from 24.2% to 21.8%. To assess bandwidth effects, the analysis was repeated using Ricker wavelets at 15, 25, 35 and 45 Hz, yielding 20,000 total synthetics across four frequencies and five NTG levels. Detection reliability improves markedly with frequency: at 60% NTG, very‐low‐risk realizations increase from 66% at 15 Hz to 88% at 45 Hz. However, at 20% NTG, discrimination fails regardless of bandwidth, with Mahalanobis distance classification showing 76% of realizations falling within or near the background shale trend (14% high‐risk, 62% moderate‐risk). Jensen–Shannon divergence further confirms AVA‐based discrimination is effective only for large NTG contrasts. A complementary brine‐saturated ensemble across the same NTG range demonstrates that gas‐sand and brine‐sand populations overlap substantially at low NTG, identifying the fluid discrimination failure zone most relevant to exploration risk. Although specific thresholds are model‐dependent, the multi‐frequency analysis demonstrates how seismic bandwidth shifts detection limits, providing a transferable framework for site‐specific calibration in heterolithic, sub‐seismic‐scale reservoirs.