Seismic modeling to monitor CO2 geological storage: The Atzbach-Schwanenstadt gas field

Abstract

We develop a petro-elastical numerical methodology to compute realistic synthetic seismograms and analyze the sensitivity of the seismic response when injecting carbon dioxide (CO2) in a depleted gas reservoir. The petro-elastical model describes the seismic properties of the reservoir rock saturated with CO2, methane and brine, and allows us to estimate the distribution and saturation of CO2 during the injection process. The gas properties, as a function of the in-situ pressure and temperature conditions, are computed with the Peng-Robinson equation of state, taking into account the absorption of gas by brine. Wave attenuation and velocity dispersion are based on the mesoscopic loss mechanism, which is simulated by an upscaling procedure to obtain an equivalent viscoelastic medium corresponding to partial saturation at the mesoscopic scale. Having the equivalent complex and frequency-dependent bulk (dilatational) modulus, we include shear attenuation and perform numerical simulations of wave propagation at the macroscale by solving the viscoelastic differential equations using the memory-variable approach. The pseudo-spectral modeling method allows general material variability and provides a complete and accurate characterization of the reservoir. The methodology is used to assess the sensitivity of the seismic method for monitoring the CO2 geological storage at the Atzbach-Schwanestadt depleted gas-field in Austria. The objective of monitoring is the detection of the CO2 plume in the reservoir and possible leakages of CO2. The leakages are located at different depths, where the CO2 is present as gaseous, liquid and supercritical phases. Even though the differences can be very subtle, this work shows that seismic monitoring of CO2 from the surface is possible. While the identification of shallow leakages is feasible, the detection of the plume and deep leakages, located in the caprock just above the injection formation, is more difficult, but possible by using repeatability metrics, such as the normalized RMS (NRMS) images. Considering real-data conditions, affected by random noise, a reference detection threshold for deep leakages and the CO 2 plume in the reservoir corresponds to a signal-to-noise ratio of about 10 dB.