Earth Sciences Division (ESD) Department of Energy (DOE) Lawrence Berkeley National Laboratory (LBNL)

Large-Scale Hydrological Impacts of Geological CO2 Storage

Project Leadership

Schematic showing different regions of influence related to CO2

Primary Contact

Key Personnel

Publications (click here)


If carbon dioxide capture and storage (CCS) technologies are implemented on a large scale, the amounts of CO2 injected and sequestered underground will be extremely large. The figure above shows schematically the large-scale subsurface impacts that will be experienced during and after industrial-scale injection of CO2. While the CO2 plume at depth may be safely trapped under a low-permeability caprock with anticlinal structure, the footprint area of the plume is smaller than the footprint area of the displaced brine, which in turn is much smaller than the footprint area of elevated pressure. The footprint area of displaced brine illustrates the approximate location of a displaced fluid volume that was originally located within the CO2 plume footprint. Of course, brine displacement occurs, to some degree, wherever a pressure gradient develops in response to injection, suggesting the possibility of water quality changes as brines or brackish water may migrate into freshwater regions. The footprint area of elevated pressure indicates the extremely large subsurface volumes where such pressure impacts might be expected.

Potential risks related to elevated formation pressure include geomechanical effects such as caprock fracturing and/or fault reactivation. Large-scale reservoir pressurization and brine displacement can also be a direct threat to groundwater resources if conductive pathways exist between deep storage formations and their overlying freshwater aquifers, because the pressure buildup at depth would provide a driving force for upward brine migration. This can be, for example, via local high-permeability flow paths such as faults and improperly plugged abandoned wells. In addition, seals may pinch out or have higher permeabilities locally, allowing for vertical inter-formation migration.

Issues related to large-scale pressure buildup and brine displacement may also cause operational and capacity problems. For example, if the same formation was used to store CO2 from multiple emitters, the operational schemes and the locations of the injection zone would have to be carefully planned to avoid unwanted feedback between neighboring sites. Such feedback can also occur with other activities involving exploitation of subsurface resources in the area, such as oil and gas or geothermal energy. Storage capacity may be a concern in compartmentalized formations, from which the displaced CO2 cannot easily escape laterally to make room for the injected CO2 (closed systems). When large volumes of CO2 are injected into a closed system, a significant pressure buildup will be produced, which can severely limit CO2 storage capacity, because overpressure and geomechanical damage need to be avoided., or may affect neighboring CO2 storage sites that reside in the same formation.

Starting in 2006, researchers at LBNL have conducted a systematic modeling assessment of industrial-scale CO2 injection and the potential for large-scale reservoir pressurization and brine displacement. Several research areas have been addressed:

Studies of Pressure Buildup and Brine Displacement in Idealized Subsurface Systems
Simulation of Hypothetical CO2 Storage Scenarios in Large North American Basins
Pressure Management via Targeted Brine Extraction

Studies of Pressure Buildup and Brine Displacement in Idealized Subsurface Systems

Our research started by categorizing subsurface reservoirs for CO2 storage as either “open”, “closed”, or “semi-closed”. In “open” formations, the native fluid can escape laterally and make room for the injected CO2. For such open formations, pressure buildup caused by CO2 injection may not be a limiting factor if the formation permeability is sufficiently high. In contrast, a “closed” system is a compartmentalized reservoir surrounded on all sides by barriers of very low permeability. When large volumes of CO2 are injected into a closed system, a significant pressure buildup can be produced, severely limiting the storage capacity. “Semi-closed” systems allow some fraction of the displaced brine to migrate into and through the overlying and underlying sealing units, which in turn would increase the storage capacity for CO2 .

LBNL scientists investigated the magnitude and extent of brine pressure buildup in closed and semi-closed formations and developed a method for quick assessment of CO2 storage capacity, complementing existing methods for capacity estimates in open systems. This method can be used to estimate the storage efficiency factor and the transient domain-averaged pressure buildup. One important finding of this research was the importance of upper- and lower-seal permeability on pressure buildup. Closed systems with impermeable seals allow CO2 storage only up to the point at which pressure in the storage formation approaches a sustainable threshold. In contrast, with small but non-zero seal permeability, brine leakage into and through the seals had a moderate to strong effect in reducing or limiting the pressure buildup in the storage formation, thus allowing for considerably higher storage efficiency, while CO2 was still safely trapped because of the combined capillary and permeability barriers. (Link to Zhou’s IJGGC 2008 paper).

Through numerical modeling, LBNL scientists also evaluated the possible impact of industrial-scale CO2 injection on multilayered groundwater systems with open boundaries. Considerable pressure buildup in the storage formation can occur more than 100 km away from the injection zone, while the lateral brine transport velocity and migration distance are generally much less significant. It was demonstrated again that seal permeability has a significant impact on pressure buildup and brine displacement behavior within the storage formation. Seals that are suitable for long-term trapping of CO2 but have a permeability higher than 0.1 millidarcy allow for considerable brine migration out of the storage formation vertically upward and/or downward. As a result, the pressure buildup in the storage formation can be strongly reduced compared to a perfect seal with zero or close-to-zero permeability (less than 0.01 millidarcy). In such cases, one needs to ensure that vertical pressure propagation and brine migration have no negative impact on freshwater aquifers. Modeling results, however, suggest that brine migration through a sequence of sealing units into shallow groundwater bodies is extremely unlikely because the migration velocity of displaced brine through these sealing units is extremely small. Note that no leaky faults and abandoned wells were considered in these studies. (Link to Birkholzer’s IJGGC 2009 paper).

LBNL scientists also developed two semi-analytical solutions for simplified calculation of pressure changes and rates of brine leakage in idealized multilayered system under single-phase brine flow conditions.  The first solution was developed to address the injection-induced pressure perturbation and vertical leakage in a “laterally bounded” system consisting of an aquifer and an overlying/underlying aquitard. Application to a large-scale injection-and-storage problem in a bounded system was demonstrated. (Link to Zhou’s TIPM 2009 paper).

The second solution accounts for the combined effect of diffuse leakage through aquitards and/or focused leakage through leaky abandoned wells in a multilayered system of any number of aquifers with alternating leaky aquitards in response to fluid injection/ extraction with any number of injection/pumping (active) wells, and passive leakage/recharge in any number of leaky wells (Link to Cihan’s WRR 2011 paper). Verification of this new solution was achieved by comparison with existing analytical solutions for diffuse leakage or for focused leakage, and against a numerical solution for combined diffuse and focused leakage. It was also demonstrated that this solution is a good approximation for pressure changes and leakage rates induced by CO2 injection in idealized multilayered systems (Link to Cihan et al., 2013, GROUND WATER). 

Simulation of Hypothetical CO2 Storage Scenarios in Large North American Basins

LBNL scientists investigated the potential pressure buildup and brine displacement induced by industrial-scale CO2 storage and assessing dynamic storage capacity (constrained by pressure buildup) in three large North American Basins: the Illinois Basin, the Southern San Joaquin Basin, and the Alberta-Williston Basin in the Northern Plains – Prairie Region of North America. The purpose of these investigations is to understand the impacts of possible future carbon sequestration scenarios in which a majority of CO2 from large stationary emitters in stored underground. Our modeling efforts honor the realistic complexity of basin hydrogeology and in situ pressure-temperature-salinity conditions by using detailed geologic models with rock properties provided by our project partners. The Cambrian-age Mount Simon sandstone formation in the Illinois Basin and the Cambrian-age Basal Aquifer in the Northern Plains – Prairie Region of North America were used as storage formations for multiple storage sites to assess the dynamic storage capacity for a CO2 storage scenario of full-scale deployment and its environmental impact. For the Southern San Joaquin basin, a single storage project at the proposed Kimberlina site for WESTCARB Phase III test was scaled up to a CO2 injection rate of five million metric tonnes per year (5 Mt/year) to assess pressure buildup and brine displacement. All simulations are conducted with very fast parallel simulation codes on multi-processor clusters.

For the Illinois Basin, an integrated modeling of basin- and plume-scale processes induced by full-scale deployment of CO2 storage was applied to the Mount Simon Aquifer. In the storage scenario, a total annual injection rate of 100 Mt CO2 around 20 injection sites of 30 km well spacing over 50 years was used. The CO2-brine flow at the plume scale and the single-phase flow at the basin scale were simulated using a unstructured 3D mesh of 1.25 million gridblocks and the TOUGH2-MP/ECO2N simulator. Simulation results show the overall shape of a CO2 plume consisting of a typical gravity-override subplume in the bottom injection zone of high injectivity and a pyramid-shaped subplume in the overlying multilayered Mt. Simon, indicating the important role of a secondary seal with relatively low-permeability and high-entry capillary pressure. The secondary-seal effect is manifested by retarded upward CO2 migration as a result of multiple secondary seals, coupled with lateral preferential CO2 viscous fingering through high-permeability layers. The plume width varies from 9.0 to 13.5 km at 200 years, indicating the slow CO2 migration and no plume interference between storage sites. On the basin scale, pressure perturbations propagate quickly away from injection centers, interfere after less than 1 year, and eventually reach basin margins. The simulated pressure buildup of 35 bar in the injection area is not expected to affect caprock geomechanical integrity. Moderate pressure buildup is observed in Mount Simon in northern Illinois. However, its impact on groundwater resources is less than the hydraulic drawdown induced by long-term extensive pumping from overlying freshwater aquifers. Our simulations showed that the pressure buildup may not be an  issue for storing large volumes of CO2 in the Mount Simon sandstone formation in the Illinois Basin, where the ongoing Illinois Basin–Decatur Project of The Midwest Geological Sequestration Consortium shows a moderate pressure change induced by 1000 t/day injection rate over the first year. (Link to Birkholzer & Zhou 2009, IJGGC and Zhou et al., 2010 Ground Water paper).

Figure 1. Overview of the Illinois Basin bounded by arches and domes, and the model domain: the flooded contour is the top elevation (m) of the Mt. Simon Sandstone, state borders are in thin black line, basin boundary in gray line, model boundary in red line, and penetration wells in red symbols. Relevant to the scenario of full-scale GCS deployment are the core injection area outlined in blue line, and the 20 injection sites in blue squares, as well as the ADM demonstration site in red circle.
  Figure 2. Simulated pressure buildup (bar) at the end of 50-year CO2 injection.

In the southern San Joaquin Basin in California, the deep Vedder Sand forms a partially closed storage system, with formation outcrop to the east along the edge of the Sierra Nevada mountain range and formation pinch-out towards all other directions, and the primary overlying seal by the Temblor-Freeman shale. A large-scale numerical model of 84 km by 112 km domain size was developed to cover the entire sequence of formations from the basement rock to the ground surface, as well as several key faults (Figure 3). The storage scenario assumed an injection rate of 5 Mt CO2/year at one well for a period of 50 years. The model accounted for pressure attenuation by diffuse water leakage through seals, by focused water leakage through the seal-pinchout area, and by water discharge into the outcrop area of the storage formation, and also represents the effect of fault zones on pressure-buildup propagation. The simulation results showed that the pressure perturbation in the Vedder Sand is confined by the southern, western, and northern closed boundaries of the storage formation, barred by the faults, and attenuated by the open eastern boundary which allows local resident water to flow into shallower formations Pressure buildup is also less significant in the northern region of the storage formation, because the local absence of the seal there allows water to migrate into overlying aquifers. (Link to Zhou & Birkholzer, 2011, GHG-S&T).

 
Figure 3. Three-dimensional view of numerical grid for Southern San Joaquin basin model. The model extent is about 80 km east-west and about 110 km north-south.

The Basal Aquifer in the Northern Plains – Prairie Region of North America covers an area of ~1500,000 km2 in the Alberta Basin and the Williston Basin (Figure 4), and has a porosity range from 1 % to 25 %, permeability range from 10 to 103 mD, and aquifer thickness up to 300 m. Sixteen storage sites were selected using CO2 source-sink mapping, with CO2 injection rate for each site ranges from 1.2 to 23 Mt/year and a total rate of 101.4 Mt/year. For each site, multiple injection wells were used to meet the requirement of maximum bottomhole injection pressure, leading to a total of 127 injection wells for the entire aquifer. LBNL scientists are developing a TOUGH2-MP/ECO2N model to simulate pressure buildup and dynamic storage capacity, as well as the distribution, migration, and long-term fate of CO2 plumes. An unstructured 3D grid was generated with fine grid resolution in the vicinity of the injection wells and progressive decrease with increasing distance from the injection sites, i.e. resolution ranging from 50 m and 20 km. This model covers the Basal Aquifer, the caprock, and the basement rock, without consideration of CO2/brine leakage through abandoned wells. The objective of our modeling efforts is to (1) quantify basin-scale pressure buildup and dynamic storage capacity, (2) predict the distribution, migration, and long-term fate of CO2 plumes in response to CO2 injection and storage, and (3) evaluate the environmental impact on fresh groundwater resources. This work is expected to be completed by October 2013.

  Figure 4. Top elevation (m) of the Basal Aquifer in the Northern Plains – Prairie Region of North America, with the aquifer boundary in red line, 16 storage sites in red circles, and penetration wells in black squares used in the geological model development.

Pressure Management via Targeted Brine Extraction

Our research also involves application of pressure management schemes for geologic sequestration by means of analytical and numerical simulations. The primary objective is to minimize pressure increases while maximizing CO2 storage in injection formations, reducing leakage risks into shallow aquifers, and minimizing costs associated with application of pressure management methods (e.g., injection/extraction rates and duration, number and location of injection/extraction wells).  Founded on these objectives, we introduced the concept of “impact-driven pressure management (IDPM)”, which involves optimization of fluid extraction to meet local performance criteria (i.e., the goal is to limit pressure increases primarily where environmental impact is a concern). To design and optimize “impact-driven” pressure management options, we developed an automatic optimization method for finding optimal well locations and selecting optimal pumping rates using the inverse modeling tool iTOUGH2/PEST framework and a suite of forward models varying from high fidelity multiphase simulators to low fidelity analytical solutions. Our initial scoping results suggested that strategic well placement and optimization of extraction may allow for a significant reduction in the brine extraction volumes needed to keep pressure increase in the storage formation below performance criteria such as critical pressure buildup for fault slippage and caprock damage or risk of upward brine leakage through possible nearby leaky wells (Link to Birkholzer et al., 2012).

The figure above is a schematic of generic application example - proof of concept studies. In the hypothetical example, a critically stressed fault with a critical pressure about 4 bar is present near to a CO2 injection site where CO2 is injected at a rate of 5 million tons per year over 50 years.  Two arrays of near-injection and near-impact (fault) wells (including active extraction and passive wells, where passive wells allow brine to flow upward to the surface or overlying/underlying layers through increased pressure gradient) are employed to control pressure while minimizing extraction volumes (Link to Birkholzer et al., 2012).

Specifically for problems with well placement optimization, objective functions can have multiple local optima and/or discontinuities in the solution space, and the gradient-based local optimization methods may fail to find the global optimum. In such cases, global optimization algorithms are needed. However, in general global optimization algorithms require a large number of forward model runs to find the optimum solution, and this appears to be a major limitation for the solution of practical optimization problems. LBNL scientists investigated combining the random selection property of the global optimization algorithms for variable well locations with the gradient-based solution strategy of the local methods for finding time-dependent injection and extraction rates. LBNL scientists are currently testing whether the coupled methodology with automated optimization schemes can solve 'real' pressure management problems during CO2 injection operations with multiple wells and multiple constraints in highly heterogeneous reservoir systems.