Modeling caprock bending stresses and their potential for induced seismicity during CO2 injection

The problem

Storing carbon dioxide underground is one proposed way to limit the rise in atmospheric greenhouse gases (IPCC, 2005). It has been suggested that, to be effective, a storage site should retain at least 99% of the injected CO2 for over 1000 years (Shaffer, 2010). Because the gas is buoyant, it is held in place by a caprock, a low-permeability layer above the storage reservoir. If an existing fault in the caprock slips, it could create a leakage pathway. Earlier models of caprock failure focused on high fluid pressure migrating upward out of the reservoir (e.g., Rutqvist and Tsang, 2002; Rohmer and Seyedi, 2010).

Some faults are critically stressed. They need only a small increase in stress to slip. In tectonically active regions, that increase is probably less than 5 MPa (e.g., Townend and Zoback, 2000). The megapascal (MPa) is a unit of stress. In this study, we asked whether an overpressured reservoir could also load caprock faults indirectly, by deforming the caprock.

What we did

We used the Springerville-St. Johns dome as an analogue site. It is a natural CO2 reservoir straddling the Arizona-New Mexico border, holding about 830 Mt (million tonnes) of CO2 (Stevens et al., 2001). The Coyote Wash normal fault confines a large part of it laterally. The reservoir is too shallow for practical storage, but its geometry gave a simplified layout for our model.

We simulated ten years of CO2 injection with FEHM, a code that couples fluid flow with rock deformation. The wells injected at a fixed overpressure, 5 MPa above the reservoir pressure in the benchmark run, and from 1 to 15 MPa in other runs. We gave the caprock a permeability of 10⁻¹⁹ m² and the reservoir 10⁻¹⁴ m², a ratio of 100,000. Our intent was to examine caprock stresses without a significant pressure rise reaching the caprock. The caprock had a stiffness (Young's modulus) of 20 GPa (gigapascals) against 5 GPa in the reservoir.

To measure the tendency for fault slip, we calculated the change in Coulomb failure stress (ΔCFS) on hypothetical faults parallel to the Coyote Wash Fault, dipping (inclined from the horizontal) at 60°. Stress magnitudes at Springerville have not been measured. We therefore used the change in CFS, which shows whether an operation promotes or inhibits fault reactivation. We also averaged ΔCFS over the model volume to give a single number, the integrated tendency toward fault failure (ITF).

What we found

In the benchmark run with a single well, the ground above the well rose by about 0.09 m (9 cm) over ten years. The expanding reservoir domed and spread the overlying rock. This produced horizontal extension near the surface. The resulting stress change has the same orientation and sense as the regional tectonic load, which was responsible for past slip on the Cedar Mesa and Coyote Wash faults.

After ten years, ΔCFS exceeded 0.5 MPa over a large part of the caprock above the well. It peaked at 1.2 MPa near the base of the caprock, and a region about 2 km wide with values above 0.5 MPa extended to the surface. This happens even though fluid pressure changes in a caprock this impermeable are small. This is because the deformation is driven by changes in total stress, the stress carried by the rock and its pore fluid together. These stresses are probably not large enough to break intact caprock. However, they would be sufficient to reactivate an undetected, well-oriented, critically stressed fault above or near the well.

Raising the overpressure from 5 to 15 MPa increased the peak ΔCFS from 1.2 to 3.5 MPa, about 23 to 24% of the applied overpressure. It also increased the ITF by almost a factor of 10 and the uplift from about 0.1 m to almost 0.3 m. Caprock stiffness had little effect on the peak ΔCFS. With a stiffer caprock, the ITF generally increased, because ΔCFS was raised over a larger volume. Uplift fell from 0.18 m at 5 GPa to about 0.05 m at 100 GPa.

The schematic redraws four of these simulations with the uplift exaggerated. In it, a softer caprock bulges more for the same overpressure, and raising the overpressure increases both the uplift and the peak ΔCFS.

With three to ten wells, ΔCFS increased in the area between the wells. We attribute this to interaction between the wells. Each time we added wells, the ΔCFS values grew larger. As a result, injecting "around but not on" a fault judged hazardous may not avoid loading it.

Limitations

Our models are linear elastic. At these shallow depths, the higher overpressures would be expected to exceed the minimum principal stress and cause tensile failure. This does not occur in our models. From our mesh tests, we estimate that the 500 m mesh overestimates ΔCFS magnitudes by about 15 to 30%. We also left out thermal stresses and the failure of faults within the reservoir or in the basement below.

Pressure management currently aims to keep reservoir pressure below the minimum principal stress by a safety margin, the threshold for hydraulic fracturing. We propose that limits on operating pressure should also consider how far the overpressured region could spread. Site-specific modelling is likely to be needed to support such decisions. We have not yet assessed how brine production wells, which extract formation water to relieve pressure, would change the induced caprock stresses.

Glossary

A caprock is a low-permeability rock layer that traps buoyant fluid in the reservoir beneath it.

Permeability measures how easily fluid flows through rock.

Overpressure is the fluid pressure above the reservoir's original pressure.

MPa (megapascal) and GPa (gigapascal) are units of pressure and stress, equal to a million and a billion pascals.

Mt is a million tonnes.

Young's modulus measures the stiffness of a rock, its resistance to stretching or squeezing.

A normal fault is a fault on which one block slides down relative to the other as the crust extends.

A fault's dip is its angle of inclination from the horizontal.

A well-oriented fault lies at an angle to the regional stresses that favours slip.

A critically stressed fault is one that needs only a small increase in shear stress to slip.

The Coulomb failure stress change (ΔCFS) is the change in stress that brings a fault closer to slipping (positive) or further from it (negative).

The integrated tendency toward fault failure (ITF) is ΔCFS averaged over the whole model volume.

The minimum principal stress is the smallest compressive stress acting at a point in the rock.

Tensile failure is the opening of new fractures when rock is pulled apart.

Total stress is the stress carried by the rock and its pore fluid together.

An analogue site is a natural setting used to study processes expected at an engineered one.