Final Report: Protecting Drinking Water by Reducing Uncertainties Associated with Geologic Carbon Sequestration in Deep Saline Aquifers

EPA Grant Number: R834382
Title: Protecting Drinking Water by Reducing Uncertainties Associated with Geologic Carbon Sequestration in Deep Saline Aquifers
Investigators: Roy, William R. , Adams, Nathaniel , Askari-Khorasgani, Zohreh , Benson, Sally M. , Berger, Peter , Butler, Shane K , D'Alessio, Matteo , Freiburg, Jared T , Hackley, Keith C , Kelly, Walton R , Krothe, J. , Krothe, N.C. , Lin, Yu-Feng Forrest , Mehnert, Edward , Panno, Samuel V. , Ray, Chittaranjan , Rice, Richard J , Storsved, Brynne A , Strandli, Christin , Yoksoulian, Lois
Institution: University of Illinois at Urbana-Champaign , Hydrogeology, Inc. , Illinois State Geological Survey , Illinois State Water Survey , Isotech Laboratories , Stanford University , University of Hawaii at Honolulu
EPA Project Officer: Klieforth, Barbara I
Project Period: November 16, 2009 through November 15, 2014
Project Amount: $897,225
RFA: Integrated Design, Modeling, and Monitoring of Geologic Sequestration of Anthropogenic Carbon Dioxide to Safeguard Sources of Drinking Water (2009) RFA Text |  Recipients Lists
Research Category: Drinking Water , Water


Protecting Drinking Water by Reducing Uncertainties Associated With Geologic Carbon Sequestration in Deep Saline Aquifers was developed with an overarching goal of protecting underground sources of drinking water from potential threats from geological carbon sequestration (GCS). GCS is a process of permanently storing greenhouse gases in the subsurface rather than discharging them to the atmosphere. This technology is considered by scientists and policy makers to be a feasible approach to reducing greenhouse gas emissions and addressing global climate change (IPCC, 2005; Socolow and Pacala, 2006; IEA, 2013). For GCS projects, monitoring, verification and assessment (MVA) procedures are conducted to demonstrate that the sequestered carbon dioxide (CO2) is securely and permanently stored in the subsurface (USDOE, 2012). MVA procedures include atmospheric, hydrological, geochemical, and geophysical monitoring techniques, and generally include modeling of these data. Our research efforts were designed to reduce uncertainties associated with selected MVA data and associated modeling procedures. We organized our research effort into five major tasks, headed by one or two technical leaders:

  1. Monitoring at Natural Gas Storage Sites (Technical Leader: E. Mehnert, Illinois State Geological Survey [ISGS])

  2. Vertical Pressure Profiles for Monitoring CO2 and Brine Migration: Research and Validation of the Westbay System (Technical Leader: S. Benson, Stanford University)

  3. Enhancement of Regional Flow and Transport Models to Reduce Risk (Technical Leaders: Y.F. Lin, ISGS and C. Ray, University of Hawaii)

  4. Geochemical Investigations (Technical Leaders: W. Roy and L. Yoksoulian, ISGS)

  5. Saline Groundwater Discharge from the Illinois Basin (Technical Leader: S. Panno, ISGS)

This research project was conducted in parallel with a U.S. Department of Energy Phase III demonstration project conducted by the Midwest Geological Sequestration Consortium ( This demonstration project, known as the Illinois Basin - Decatur Project (IBDP), has a goal of safely and permanently sequestering 1 million tonnes of CO2 in a basal saline reservoir (Finley et al., 2013). IBDP began injecting CO2 into the Mt. Simon Sandstone on November 17, 2011, and had 622,000 tonnes sequestered by October 15, 2013, and 976,539 tonnes sequestered by November 4, 2014.

Task 1 was a data mining effort to compile geologic, hydrogeologic, and geochemical data mainly from natural gas storage fields completed in the Mt. Simon Sandstone in and near the Illinois Basin (Fig. 1). These storage fields have been developed in geologic structures in the top few hundred feet (100 to 150 m) of the Mt. Simon and have operated since the late 1950s. These data will be valuable for improving basin-scale modeling of geological carbon sequestration, as these fields are located between the ideal GCS sites in central Illinois and Indiana and groundwater resources in northern Illinois and southern Wisconsin. Data from wastewater injection wells and other wells were also included in this effort.

Task 2 was an effort to develop and demonstrate methods for monitoring migration and potential leakage of CO2 using multilevel (depth-discrete) pressure transient measurements. The methods were applied to the pressure transient data at IBDP, where the Westbay multilevel system was installed in a monitoring well to measure the pressure buildup during CO2 injection.

Task 3 was an effort to link a GCS model with an existing groundwater flow model for bedrock aquifers in Illinois. Procedures were developed to link the two different models to better understand the potential interactions of commercial-scale GCS in central Illinois and Indiana and current and future groundwater pumping from bedrock aquifers in northern Illinois and southern Wisconsin.

Task 4 was an effort to improve our understanding of GCS related geochemistry. Archived and fresh samples of the injection reservoir (Mt. Simon Sandstone) and the caprock (Eau Claire Formation) were exposed to CO2 at injection reservoir temperatures and pressures in the laboratory. Fresh samples of the Mt. Simon and Eau Claire were retrieved during drilling of IBDP's injection and verification wells (CCS1, VW1, and VW2). The results of the laboratory experiments were analyzed and modeled to evaluate the effects of GCS on the injection reservoir and the caprock.

Task 5 was an effort to improve our baseline data for naturally occurring saline springs in and around the Illinois Basin and to develop techniques to characterize these saline fluids. A common concern that the general public has regarding GCS is that CO2 injection will cause native brines to migrate from the injection reservoir into overlying formations and degrade water quality. The value of better baseline data is easily understood once this water quality concern is appreciated.

Summary/Accomplishments (Outputs/Outcomes):

This research project included five separate but related tasks. Two tasks focused on geochemistry and three focused on hydrogeology and hydrology. Through these tasks, we compiled available data, generated new data, improved existing techniques, and developed new techniques. Collectively, the results from these five tasks have resulted in a better understanding of the geochemistry and hydrogeology of the Mt. Simon and the Eau Claire Formations in the Illinois Basin, which will help protect underground sources of drinking water from potential threats from geological carbon sequestration. While some of the results presented here are applicable only to this specific saline reservoir and caprock, other results such as the new and modified techniques are easily transferable to other GCS reservoirs around the world. The collective result of these efforts has been to reduce the uncertainty associated with protecting underground sources of drinking water from potential threats posed by geological carbon sequestration.

For Task 1, the data mining efforts lead to the collection of a significant quantity of core porosity, core permeability and water quality data from natural gas storage sites in and around the Illinois Basin. These data were also collected from other types of wells such as Class I injection wells and water wells. A limited amount of two-phase flow data such as capillary pressure curves and relative permeability data also were collected.

During the 1950s through 1970s, developers of natural gas storage sites collected core and had it tested in commercial laboratories to determine porosity, horizontal permeability, and vertical permeability. Porosity data were the most common data collected, while vertical permeability was the least common data for reservoir rock and horizontal permeability was the least common data for the caprock. For this project, we digitized data for 12,652 core samples—9,383 Mt. Simon and 3,269 Eau Claire core. The quantity of core data varied across the 13 storage fields. Ancona had the largest collection of core (2,812 core samples) and Troy Grove had the smallest collection (339 core samples). Collectively, the core data demonstrate the considerable variability in the rock properties of the Eau Claire and Mt. Simon. In addition, the lithology of the Eau Claire is more variable than the Mt. Simon across the basin. More sandstone units are present in the Eau Claire in the northern part than the central part of the Illinois Basin, which contribute to some of the higher permeability values observed for the Eau Claire at the northern storage fields.

Aquifer tests were conducted at 10 gas storage fields and one waste injection well. Aquifer tests provide permeability data but at a larger scale than provided by core. Most aquifer tests were high quality tests with sufficient pumping to stress the aquifer, sufficient length to evaluate potential boundary effects and leaky confining layers, and sufficient observation wells to collect data. Like the core data, the variability of the aquifer test results demonstrated the variability of the hydraulic properties of the Mt. Simon. Data and analyses in this study indicated a scale dependence of hydraulic conductivity when comparing core and aquifer test data, which runs counter to Schulze-Makuch et al. (1999) who concluded that scale dependence of hydraulic conductivity was not observed in the Mt. Simon in the upper Midwest. This topic deserves additional research. Considerable effort was spent analyzing aquifer test data, but additional work could be done here. We used derivative analysis to analyze the aquifer test data at a few sites, but it could be applied for all sites. No groundbreaking results were uncovered using derivative analysis, but these aquifer test data are quite valuable and all efforts should be made to extract as much information as possible. The derivative analysis package, Hytool, needs to be expanded to include one or more of Hantush’s methods for leaky confined aquifers. To fully explain some aquifer test results, a groundwater model may be needed to account for the complexity of the site geology and three-dimensional flow arising from the test setup (pumping well with a 7.5 m [25 ft] screen in a 600 m [2000 ft] thick aquifer).

Additional geochemical data were digitized for this project. The primary variable of interest was total dissolved solids (TDS), which can be used to estimate brine density and brine viscosity. An improved map of TDS in the Mt. Simon in Illinois and Indiana (Fig. 2) was developed and includes the 10,000 mg/L TDS line, which is significant to regulators. This new map was developed using 169 TDS values while the existing map used 57 TDS values but included no Indiana data. This new map could be improved with additional data, especially in the southern portion of the Illinois Basin. More importantly, data are also needed to define the vertical distribution of TDS within the Mt. Simon, which is more than 600 m (2,000 ft) thick in some areas.

Collectively, the data compiled during Task 1 are essential data for modeling groundwater flow and GCS in the Illinois Basin. These data will be useful to basin-scale model developers (e.g., Zhou et al. [2010]). Better models are built using better data. Better models lead to better predictions, predictions with less uncertainty.

The objective for Task 2 was to develop and demonstrate methods for monitoring CO2 and displaced brine migration using multilevel (depth-discrete) pressure transient data. Through numerical studies using TOUGH2/ECO2N and analyses of simulated pressure data, diagnostics for reservoir structure (layering and anisotropy) and CO2 plume migration were identified, showing that information about CO2 plume migration is evident long before the CO2 reaches the monitoring well (Strandli and Benson, 2013). In particular, important insights about reservoir structure and CO2 plume migration can be obtained by (1) normalizing the pressure buildups to the pressure buildup at the depth of injection and (2) calculating vertical pressure gradients normalized to the initial hydrostatic pressure gradient (Fig. 3). Based on the simulated pressure data, we showed that pressure buildups normalized to the pressure buildup at the depth of injection and vertical pressure gradients normalized to the initial hydrostatic pressure gradient are diagnostic of reservoir structure soon after the start of injection and over time provide information on the height of the CO2 plume. The identified diagnostics were applied to the Illinois Basin - Decatur Project (IBDP), where the Westbay System was installed in verification well VW1 to record the pressure buildup at multiple depths within the Mt. Simon storage reservoir and above the Eau Claire Formation (primary seal) during CO2 injection. Using the diagnostic tools, we were able to correctly identify the height of the CO2 plume. Specifically, the multilevel pressure transient data alone indicated that the CO2 plume remained largely confined to the 23 to 24 m interval into which it was injected, and there was no indication of buoyancy driven flow towards the shallower portions of the Mt. Simon. This prediction was confirmed by Reservoir Saturation Tool logs and sampling carried out by the IBDP.

Moreover, a multilayered, radially symmetric model with TOUGH2/ECO2N was used to history match the pressure buildup at the injection well (CCS1) and VW1. Our overall excellent match with the pressure transient data from IBDP demonstrate that by history matching multilevel pressure transient data, a hydrogeological model can be developed that in turn can be used to predict future CO2 migration.

Overall, this project has successfully demonstrated that multilevel pressure measurements are very useful for monitoring pressure buildup and CO2 plume migration. By incorporating multilevel pressure monitoring into the monitoring program, additional information is available that can be used to minimize and manage risk associated with upward migration of CO2 and displaced brine.

Future work includes more rigorous sensitivity studies, including examination of other geologic model scenarios, and the application of formal inversion/optimization methods. A better understanding of the injection zone characteristics at the IBDP is needed to more accurately predict the lateral migration of the CO2.

In terms of recommendations for the use of multilevel pressure monitoring in future CCS projects, we found it important that at least one monitoring zone be located at the depth corresponding to the injection zone. The pressure buildup at this depth is likely to respond the quickest and is a useful reference for discerning the relative magnitudes of the pressure buildups at shallower depths, which in turn is important for determining the height of the CO2 plume within the storage reservoir. Overall, the number and placement of monitoring zones will affect the vertical resolution of reservoir heterogeneity and the ability to constrain a hydrogeological model with pressure history matching (including the ability to compensate for pressure sensors that might be offline for certain periods of time).

In Task 3, an improved understanding of the possible effects of potential future commercial-scale geological carbon sequestration projects on underground drinking water sources in the Illinois Basin was sought. This objective was accomplished using T2SW-Link, which is a new Python script that couples two codes—TOUGH2 and SEAWAT. TOUGH2 is suitable for simulating multiphase and variable density flow such as GCS, while SEAWAT-2000 is suitable for variable density, groundwater flow. Rather than using a single TOUGH2 model, using T2SW-Link to couple SEAWAT-2000 and TOUGH2 reduces the computational cost of modeling this system, allowing the impacts of the GCS project (increasing formation pressure) and groundwater pumping (reducing formation pressure) to be analyzed more efficiently. The use of the SEAWAT-2000 model, which is a MODFLOW based code, also allows us to incorporate current and projected freshwater pumping data developed for a previously calibrated MODFLOW model.

The goal of Task 3 was reached by providing a quantifiable understanding of the risk of groundwater contamination primarily from upgradient brine migration within the Mt. Simon. Risk can be lessened via reduction in parameter uncertainty by analyzing the parameter sensitivity and using numerical modeling for alternative scenarios. The parameter uncertainties of these models were analyzed using PEST with a parallel computing cluster. The results of the sensitivity analysis have identified model parameters that need to be improved and help prioritize future data collection on porosity, permeability, and storativity. The migration of native brine and its impact on freshwater drinking sources was investigated by passing pressure and salt concentration data between the models at specific time steps. Results show the method to be successful with pressure impacts of the simulated GCS activity reaching the Chicago, IL region, but rather limited brine migration. Simulations out to 300 years indicate that no impacts on the Mt. Simon aquifer at the Illinois-Wisconsin border and USDW should be expected for the hypothetical commercial-scale injection rates examined in this work (3 wells injecting 150 kg/s or 4.7 million tonnes/year each).

Several future studies are strongly suggested to improve this work for more insightful understanding and scientific basis for developing future monitoring strategies according to the simulation of formation pressure, brine movement and groundwater flow field. The coupling performance should be optimized by examining (1) the size of the overlapping region between TOUGH2 and SEAWAT-2000 modeling regions, and (2) total number of stress periods required to complete the simulation such as using longer stress periods for shorter run times without affecting model convergence and mass balance. A simple geologic model was used in this study. The geologic model could be improved in a number of ways by adding spatial variability of porosity and permeability, better vertical discretization of the geology, and low permeability zones found in the lower Mt. Simon that act as baffles inhibiting upward migration of CO2. After the success of this pilot-scale effort, we need to apply this technique using the full-scale Illinois Basin GCS model and higher CO2 injection rates (on the order of 100 million tonnes of CO2 per year).

For Task 4, laboratory experiments were conducted to evaluate how GCS, specifically exposure to carbon dioxide, might alter the injection reservoir rock and caprock. A variety of analytical techniques were utilized to determine physical, geochemical, and mineralogical changes for pre- and post-reaction rock samples, including transmitted light microscopy, scanning electron microscopy and energy-dispersive X-ray analysis, and X-ray diffraction analysis. Pre- and post-reaction brine samples were analyzed for inorganic anions and metals utilizing inductively coupled plasma-atomic emission spectrometry and ion chromatography. Experimental parameters were set to simulate in-situ conditions and the IBDP site.

The modeling program, Geochemist’s Workbench, was used to match geochemical data from the batch experiments. The original brine and mineral compositions were input into a batch model and run with React and Differential Evolution, then iteratively adjusted to the kinetic constraints of the silicate minerals to find the best fit to the changes observed in the post-experiment brine chemistry and sample mineralogy.

The most significant alteration observed in the post-reaction Mt. Simon Sandstone was the reduction in size of clay linings associated with quartz grains (Fig. 4) that occurred over the 6-month experimental duration (Yoksoulian et al., 2013). The mechanism for this reduction is not clear; however, it could be a result of dissolution of clay minerals or by mechanical force. This reduction in thickness of clay coatings could increase the porosity of the Mt. Simon Sandstone, allowing it to accept more CO2 than originally determined.

Post-reaction data from the Eau Claire Shale experiments indicated some degree of chemical reactivity observed as degradation of feldspars, clays, and pyrite over 3, 6, and 9 month experimental intervals, as seen in petrographic analyses. However, changes in pre-and post-reaction rock texture, mineralogy, and fluid chemistry were small and difficult to quantify with the analytical methods used. Geochemical modeling results indicate a rapid rate of dissolution of clay minerals, but results derived from the simulation experiments indicate that reactions reached equilibrium after approximately 6 months of exposure of the rock to the CO2-brine environment.

At the IBDP site, injection occurs at the base of the Mt. Simon Sandstone and contact of the CO2 with the Eau Claire Shale is not likely for many years. This would suggest that although an initially chemically reactive scenario is presented for the Eau Claire Shale-CO2-brine system at the project site, the effect of this reactivity may be significantly reduced within 6 months after CO2 contacts the Eau Claire due to buffering of CO2.

Post-experiment fluid analyses from this study indicate that if a breach in Eau Claire Shale seal were to occur, USEPA regulated components (Ba, Cu, NO3, Pb, and Tl from the Mt. Simon Sandstone and Cu, Pb, and Se the from Eau Claire Shale) could be released into overlying strata as a result of interactions between the rock, CO2, and brine. Results indicating concentrations of the USEPA regulated components As, Be, Cd, Pb, Se, and Tl that were below the method detection limit (MDL) are inconclusive due to the analytical method used; MDLs are upwards to 150 times greater than that of the USEPA maximum contaminant levels.

However, the shaley facies of the Eau Claire Formation is approximately 60 m (200 ft) thick in the project area of review. It is unlikely that dissolution of feldspar or clay minerals could cause a breach in the Eau Claire Shale (and overlying secondary (Maquoketa Shale) and tertiary (New Albany Shale) seals) before the dissolution reactions ceased.

For Task 5, natural groundwater seeps and springs in and around the Illinois Basin were sampled and their geochemistry was reviewed. The chemical composition of groundwater in the Illinois Basin, like most intracratonic basins in the midwestern United States, transitions from extremely fresh near surface (Cl- = 1 to 15 mg/) to concentrated brines at depth (Cl- > 120,000 mg/L). Naturally occurring NaCl-enriched groundwater anomalies (those exceeding background for Cl-) have been identified throughout the Illinois Basin (in Illinois, southern Indiana, northern Kentucky, northern Tennessee) as NaCl-enriched springs, or localized seeps into shallow freshwater aquifers (typically known from domestic wells). In the course of this investigation, 40 occurrences of upwelling NaCl-enriched, brackish and saline groundwater within and near the margins of the Illinois Basin were identified from historic accounts and investigated (Fig. 5). Each of these springs, seeps and anomalies were sampled during this investigation and analyzed for chemical and isotopic compositions with special emphasis on their halide chemistry (Cl and Br).