The ERA-ACT funded Pre-ACT project presents findings from its Work Package 2, focusing on estimating relevant CO₂ storage monitoring parameters that can be used in conformance and decision-making studies. Discrimination between competitive effects (saturation and pore pressure) on geophysical observables will be discussed and quantified with different approaches.

Wolfgang Weinzierl (Research Scientist, GFZ) will present a deep neural network approach to rock physics inversion. The latent vector in this case is a subset of the viscoelastic attributes VP, VS, ρ, QP and QS. After an initial investigation of the performance of different neural network architectures for predicting rock physics properties, tests inverting for a range of parameters are presented. The developed methodology is subsequently applied to a scenario-based simulation of CO₂ injection targeting a shallow aquifer at the Svelvik ridge, Norway.

Bastien Dupuy (research scientist, SINTEF) will then present a complimentary approach based on Bayesian inversion. Quantitative monitoring workflow will be described, by combining geophysical imaging with uncertainty assessment with Bayesian rock physics inversion in a second stage. This process allows for thorough estimation and propagation of uncertainty and provides estimates of relevant monitoring properties (saturation, fluid distribution, pore pressure) with confidence intervals. Such approach is crucial for proper conformance monitoring and associated decision-making. Examples will be shown using Sleipner, Snøhvit, Smeaheia or Svelvik datasets.