Impact of Rock Frame Models on Time-Lapse Seismic Attributes During CO₂ Injection into Depleted Gas Reservoirs

A well-calibrated petroelastic model is critical for time-lapse quantitative interpretation, linking dynamic reservoir property changes to seismic attributes such as impedance, AVO gradient, and intercept. This study assesses the sensitivity of these attributes to saturation and pressure changes using several common dry rock frame models. Although all models fit static sonic logs reasonably well, their 4D seismic responses differ significantly. For CO₂ displacing brine in a depleted gas reservoir, the Stiff sand model generates the highest 4D responses, while the 95th percentile of the patchy cement model derived via probabilistic calibration produces the lowest. In some models, the gradient shifts from negative to positive at certain saturation levels during injection, deviating from the typical negative gradients expected for softening effects, highlighting the complex behaviour of CO₂ and hydrocarbon gas in the reservoir. Cementation type in the patchy cement model minimally affects saturation-induced elastic changes but strongly influences pressure-related responses alongside hyperparameter variations. Additionally, stress-sensitive models based on compliance theory produce notably different pressure-induced 4D responses compared to Hertzian theory-based models. These findings emphasize the importance of selecting appropriate models and accounting for model uncertainties to ensure reliable 4D seismic interpretations for reservoir monitoring in carbon capture and storage projects.