Abstract
Quantification of reservoir properties and reservoir properties changes over time through the recording of geophysical monitoring data has been a long-standing challenge. This quantification step relies on accurate and flexible rock physics models and depends on the sensitivity of the geophysical data to the specific reservoir properties (Dupuy et al., 2021a). It is usually accepted that fluid saturation (mix of brine and supercritical CO2 in saline aquifer) and pore pressure are the main properties affected by the CO2 injection (Grude et al., 2013) if we neglect geochemical effects (which is acceptable in sandstone reservoirs). However, to build relationships between the geophysical observables and these reservoir properties, other parameters need to be estimated, such as porosity, fluid properties, rock frame properties, etc. This leads to solving an inverse problems with a minimum of 10 parameters. Using seismic data, we can carry out seismic inversion and derive at best 3 to 5 parameters. Despite possible approximations, it is usually not enough to derive estimates of CO2 saturation and pressure with good confidence.
We developed a two-step method to estimate rock physics properties by combining geophysical inversion and rock physics inversion. Both steps profit of a Bayesian formulation allowing to derive posterior distributions and estimate uncertainty of the final models (Dupuy et al., 2021b). Originally validated with full waveform inversion (FWI) of seismic data at Sleipner, we propose to test the added value of other geophysical methods. We build synthetic models representing Sleipner and Smeaheia reservoirs and changes of pressure and saturation over time due to CO2 injection and migration. We model changes in geophysical observables, i.e., changes in P- and S-wave velocities, in bulk density and in bulk resistivity. We assume we can derive these 4 geophysical observables with relevant inversion approaches e.g., acoustic or elastic FWI, gravimetry, CSEM (controlled source electromagnetic) inversion. Then, we estimate a relevant set of rock physics properties considering different prior distributions, either uniform or Gaussian distributions centered on the true value. Using different combination of input geophysical observables, we are able to determine which input data is required to estimate which rock physics property within a given uncertainty range.
Figure 1 gives an example of such results where P-wave velocity VP and S-wave velocity VS are extracted from seismic inversion, bulk density ρ is estimated by gravimetry and bulk resistivity Rt inverted from CSEM data. We show that combining P-wave velocity and resistivity and to some extent density is required to estimate correctly the CO2 saturation. Porosity estimates require VP and density while estimation of pore pressure and patchiness exponent is poorly constrained and mostly driven by the available prior information.
Deriving such conclusions is crucial to build monitoring plans and especially for conformance verification where modelled data (pressure and saturation) needs to be verified with geophysical observables.
We developed a two-step method to estimate rock physics properties by combining geophysical inversion and rock physics inversion. Both steps profit of a Bayesian formulation allowing to derive posterior distributions and estimate uncertainty of the final models (Dupuy et al., 2021b). Originally validated with full waveform inversion (FWI) of seismic data at Sleipner, we propose to test the added value of other geophysical methods. We build synthetic models representing Sleipner and Smeaheia reservoirs and changes of pressure and saturation over time due to CO2 injection and migration. We model changes in geophysical observables, i.e., changes in P- and S-wave velocities, in bulk density and in bulk resistivity. We assume we can derive these 4 geophysical observables with relevant inversion approaches e.g., acoustic or elastic FWI, gravimetry, CSEM (controlled source electromagnetic) inversion. Then, we estimate a relevant set of rock physics properties considering different prior distributions, either uniform or Gaussian distributions centered on the true value. Using different combination of input geophysical observables, we are able to determine which input data is required to estimate which rock physics property within a given uncertainty range.
Figure 1 gives an example of such results where P-wave velocity VP and S-wave velocity VS are extracted from seismic inversion, bulk density ρ is estimated by gravimetry and bulk resistivity Rt inverted from CSEM data. We show that combining P-wave velocity and resistivity and to some extent density is required to estimate correctly the CO2 saturation. Porosity estimates require VP and density while estimation of pore pressure and patchiness exponent is poorly constrained and mostly driven by the available prior information.
Deriving such conclusions is crucial to build monitoring plans and especially for conformance verification where modelled data (pressure and saturation) needs to be verified with geophysical observables.