Seismic monitoring of CO 2 geo‐sequestration: realistic capabilities and limitations

Abstract

Summary Seismic is useful for monitoring and verification of subsurface geo-sequestration CO2 injection and storage projects. The physical properties of CO2-saturated rocks can vary strongly, thus the resulting seismic wavefield can be rich and complex, posing significant challenges to obtain accurate CO2 images and quantitative inversion results. We discuss the seismic rock and fluid properties of CO2-saturated rocks under various realistic pressure and temperature conditions, show the effects of CO2 injection with realistic 3D models and finite-difference simulated seismic data, and compare simulation data images and inversion results to real seismic data at the Sleipner CO2 sequestration site. Rock and fluid physics In this paper we focus on the geo-sequestration application of injecting CO2 into subsurface porous and permeable brine-saturated rock formations for long-term storage. For many geo-sequestration sites and timescales of interest, the depth, pressure and temperature of the reservoir storage rocks will imply that the bulk of injected CO2 will be in an immiscible supercritical phase above the critical point (Pc = 7.38 MPa, Tc = 31.1 o C) in the phase diagram, thus having physical properties of both a gas and fluid. Figure 1 shows our calculations of the bulk modulus (incompressibility) and density of CO2 at various pressure and temperature conditions. CO2 is a gas below the red line, a fluid above the red line, and supercritical to the right of the red line in the phase diagram. The (P,T) conditions for Sleipner are plotted on the diagrams (red circle) and clearly fall within the supercritical region, as most sequestration projects are likely to do (dashed yellow box). Note that the compressibility and density of CO2 can vary as much as one order of magnitude across the geo-sequestration (P,T) range, especially at higher pressures (depths). Figure 2 shows seismic P- and S-wave velocity (Vp, Vs) and density curves for saturated Sleipner-type sandstone rocks as a function of the CO2/brine fluid mixture, when the CO2 behaves as a dense supercritical “fluid” (at 37 o C, 10MPa) and as a light supercritical “gas” (at 44 o C, 7.5 MPa). In the supercritical “fluid” case, the bulk density of the saturated rock is a weak linear function of the CO2 saturation, and Vp is a nonlinear function showing a strong decrease for CO2 saturations up to about 30% but little or no change for larger CO2 saturations. In the supercritical “gas” case, the bulk density of the rock is a strong measurable and linear function of the CO2 saturation, and Vp is a nearly binary function showing a strong decrease for small amounts of CO2 < 5-10% but little or no change for larger CO2 saturations. In both supercritical cases, Vs is fairly insensitive to CO2 saturation. 3D models and seismic simulations We built a 3D earth model loosely based on Sleipner logs, rock and fluid properties, and geologic/seismic structure, as shown in Figure 3. We introduced porosity heterogeneity using spatial statistics from another reservoir to examine its effects on seismic imaging and inversion. We created three fluid and pressure distributions in the model: (i) brine only, i.e. before CO2 injection (T0), (ii) after limited injection has created one layer of CO2 just below the impermeable cap rock (T1), and (iii) after further injection has created two vertically stratified layers of CO2, one beneath the cap rock as in T1 but with stronger saturation, and a second weaker saturated layer trapped below a deeper shale (T2). We generated synthetic seismic shot-gather data sets for each of the three CO2 injection scenarios acquired along a 2D line through the 3D model, using a visco-elastic finite-difference (FD) algorithm running on our parallel cluster. Figure 3 shows the z-component data (mainly P-waves) and the x-component data (mainly S-waves) of the elastic wavefield for a single shot gather. Note that a single layer of CO2 (at T1) generates a complex coda of many strong events in the shot-gather difference data (T1-T0) of Figure 3! The wavefield differences are even more complex at T2 for two stratified layers of CO2, as we show in the presentation. Seismic imaging We performed industry-standard prestack depth migration velocity analysis and image processing of the data sets for all three scenarios. Aside from the changes in CO2 saturation and pressure in the geo-sequestration layers, these data sets are perfectly repeatable from a time-lapse perspective. Figure 4 shows the P-wave depth image before and after CO2 injection (T0 and T1), and the time-lapse image (T1-T0). Figure 5 shows a zoom of the time-lapse difference image (T1-T0) for P-waves (left), converted PS-waves (right) and a comparison to a real time-migrated Pwave difference image from Sleipner (center). Note that even with our best velocity analysis and depth imaging in