2011 Progress Report: Risk-Based Decision Making for Assessing Potential Impacts of Geologic CO2 Sequestration on Drinking-Water SourcesEPA Grant Number: R834387
Title: Risk-Based Decision Making for Assessing Potential Impacts of Geologic CO2 Sequestration on Drinking-Water Sources
Investigators: McCray, John , Kaszuba, John , Marcon, Virginia , Maxwell, Reed , Sitchler, Alexis
Current Investigators: McCray, John , Maxwell, Reed , Sitchler, Alexis
Institution: Colorado School of Mines , University of Wyoming
Current Institution: Colorado School of Mines
EPA Project Officer: Klieforth, Barbara I
Project Period: February 1, 2010 through January 31, 2013 (Extended to January 31, 2014)
Project Period Covered by this Report: November 1, 2010 through October 31,2011
Project Amount: $899,987
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
4.1 Literature Review
- Current potential CO2 sequestration sites
- Current research on leakage of CO2 from sequestration sites
4.2 Geochemical modeling and experiments
- Intermediate scale experiments
- Geochemical modeling
4.3 High T and P laboratory geochemical experiments
4.4 Risk Modeling
- Basin scale hydrologic modeling
- Streamline modeling
4.5 Evaluation of potential CO2 injection projects
4.1 Literature Review
4.1.1 Current potential CO2 sequestration sites
A review of current potential CO2 sequestration sites was performed.
4.1.2 Current research on leakage of CO2 from sequestration sites
In addition to the literature review performed on CO2 leakage rates from sequestration sites, we undertook a literature review related to brine leakage rates and found that little research has been performed on the impacts of brine leakage on shallow aquifers. We, therefore, performed an analysis (detailed below) on the impacts of brine leakage on shallow aquifers.
Analysis of brine geochemistry: injection of large amounts of CO2 into a deep saline formation will lead to competitions of CO2 with brine over pore space. This will cause pressure increases in the target formation, and propagation of pressure farther and faster than the extent of the CO2 plume itself. The displacement of brine and the pressure increase raise the concern of brine leaking from the target formation, mostly in the far-field region (away from the injection well), following mechanisms similar to CO2 leakage (i.e., leaky wells, fractures, faults, etc.). The expected geochemical changes to USDWs from brine leakage are inherently different from ones expected from CO2 leakage. However, little is known about the geochemical composition of deep brines because there is usually no economic incentive to study them. To shed light on the geochemistry of deep brines, we utilized the NATCARB database, which is maintained by the National Energy Technology Laboratory (NETL). The database contains more than 120,000 entries of brine geochemical data, mostly provided by the oil industry, although the data are not complete for each entry. We applied a rigorous data assurance methodology to eliminate entries that may not be representative of the native brine (i.e., not associated with contamination, not sampled from locations other than wells, etc.), and that contain less than 10,000 mg/L total dissolved solids (TDS; regulatory constraint for minimum salinity of target formations), which reduced the database to about 67,000 entries. We then applied simple statistics on the remaining data, to produce expected concentrations and overall distributions of various geochemical parameters, which were then placed against EPA regulatory limits (MCLs) and secondary standards for drinking water. It was found that the median or log-normal mean of most elements do not exceed any EPA regulatory limit, which indicates that their concentrations are not expected to surpass the limit in a USDW even in the extreme (and unlikely) case of full displacement of USDW waters by brine. However, several dissolved constituents did exceed the regulatory levels, including chloride, iron, manganese, nitrate, sulfate and overall TDS (the latter is expected, given our data screening criteria). It also was found that the concentration distribution of several constituents was not narrow around the median or log-normal mean (i.e., the concentrations spanned several orders of magnitude above the regulatory level). This means that while the probability of encountering above-limit concentrations in brine is low, the severity of these low-probability events is very high. Still, the overall analysis suggests that in case of brine leakage, direct salinity (in terms of chloride content or overall TDS) will probably be the first parameter to deem a water in a USDW not suitable for drinking. In addition, we found that the expected pH of brines revolves quite narrowly around neutral acidity, which means that brine leakage will probably not be detected through monitoring of pH, as was suggested by some researchers for CO2 leakage. It is apparent from our analysis that a straightforward electrical conductivity measurement may be more appropriate than pH measurement in detecting brine leakage. In our work, we also analyzed the impact of brine leakage into a USDW in an agricultural context, that is – if the impacted waters are used for irrigation. The analysis was comprised of theoretical mixing of two salinity end-members – brine and USDW – at different brine fractions in the mixture, corresponding to different levels of brine contamination. The TDS resulting from these mixtures were compared to crop salinity sensitivities of different crops grown in the United States. Based on this analysis, the yields of most crops are expected to be affected at a brine fraction of 0.1. The most sensitive crop, strawberry, would show a decrease in yield if irrigated with USDW water containing 0.6% brine.
4.2 Geochemical modeling and experiments
4.2.1 Intermediate scale experiments
Laboratory experiments of carbonate rocks under elevated CO2 partial-pressures: several experimental studies exist, in which the researchers examined the geochemical effects of high-CO2 environments on natural rock material. These studies have been divided into high-pressure/high-temperature experiments (analogues for target formation), and studies in which CO2 was bubbled through beakers containing rock and water, with no attempt to control the overall pressure (beakers open to the atmosphere). In most works, for practical reasons, a single CO2 bubbling rate, or single CO2 pressure, are applied in the system. In our work, we addressed three gaps found in the literature:
- Effects of intermediate CO2 partial-pressures on natural rock material, which are more realistically representative of shallow aquifers.
- Effect of elevated CO2 partial-pressures on carbonate rocks. To our knowledge carbonate aquifers have not been studied in this context, other than one study which examined an oolitic limestone under storage conditions (Sterpenich et al., 2009).
- Application of several partial-pressure of CO2, at increasing orders of magnitude, as analogues for different leakage rates into a USDW.
The overarching goals of these experiments are identifying metals that may be released into solution from natural rocks under elevated partial-pressures of CO2, and to identify the minerals and processes that control their release.
The main portion of the experimental work was construction of beakers containing natural carbonate rocks (limestones and dolomites) and water, under overall pressure of 1 bar. The partial pressure of CO2 in the beakers was increased in stages, from 0 (pure N2) through 1% (99% N2), 10% (90% N2) and 100% CO2. Aqueous samples were drawn from the beakers at increasing time intervals, and pH was measured in situ. Results from the metal analysis of the aqueous samples showed that minor and trace metals were released into solution during the experiment, sometimes in accordance with calcium release, and sometimes not. Stabilization of pH, due to dissolution of carbonate minerals, did not necessarily correspond to stabilization of the aqueous concentrations of these metals. Minor and trace elements that were found to be released into solution include lead, arsenic, nickel, cobalt, uranium and chromium. Some elements were detected in certain beakers, and not in others. Among the elements that were detected consistently, the release pattern of these metals into solution was overall similar among different rock samples, but the degree of increase in solution varied between rocks. Most elements did not exceed their EPA-mandated MCL for drinking water under our experimental constraints. In specific cases, however, arsenic, chromium and nickel did exceed the EPA-mandated MCLs (State of California-mandated MCL for nickel), under 1 bar of CO2.
Supporting analyses included high-resolution energy-dispersive spectroscopy (EDS), to identify and quantify the distribution of minerals present in the rock samples, and X-ray diffraction (XRD) analysis of the whole rocks and clay fraction.
Future work includes laser-ablation-inductively-coupled-plasma-mass-spectrometry (LA-ICP-MS) analysis to quantify the elemental distribution in specific mineral phases and better understand the source of elements seen in the pressurized experiments. The processes controlling the release of these elements will be evaluated through geochemical modeling of the experimental system.
4.2.3 Geochemical modeling
A reactive transport simulation study was performed to evaluate the release of lead from different minerals (galena and calcite) in a shallow aquifer impacted by CO2 leakage. Sets of simulations were performed where monitoring well pumping rate, aquifer mineralogy, and permeability anisotropy were varied. Five realizations of both a high and low anisotropy permeability field were generated using geostatistical methods. Each realization was simulated using the reactive transport simulator PFLOTRAN. Results from the different realizations were compared by evaluating lead concentrations at a pumping well downstream.
Even with relatively coarse grid spacing (9 m x 9 m x 0.9 m) and simple aquifer mineralogy, the simulations were composed of 5,470,524 grid cells and 11 chemical species with a total number of degrees of freedom of > 60 million. Simulations were run on Jaguar, the Cray XT5 at Oak Ridge National Laboratory, utilizing 2,048 processors and had wall clock times of approximately 1.5 hours (> 3,000 processor hours per simulation). Each ensemble of 5 realizations, therefore, required > 15,000 processor hours. For the 8 different ensembles of different aquifer scenarios a total of > 120,000 processor hours (> 14 years) were required. It is immediately obvious that without high performance computing these simulations would not be feasible.
Increased anisotropy resulted in more lateral and vertical spreading of the plume of impacted water, leading to increased Pb2+ concentrations and lower pH at a well down gradient of the CO2 leak. Pb2+ concentrations were higher in simulations where calcite was the source of Pb2+ compared to galena. The low solubility of galena effectively buffered the Pb2+ concentrations as galena reached saturation and precipitated along the flow path. In all of the cases Pb2+ concentrations remained below the maximum contaminant level set by the EPA. Results from this study suggest that bicarbonate concentrations may be a better geochemical indicator of a CO2 leak under the conditions simulated here.
4.3 High T and P laboratory geochemical experiments
Laboratory experiments are designed to evaluate metal mobilization as a consequence of injection of supercritical CO2 into a storage reservoir. Our experiments emulate an actual CO2 sequestration reservoir, the Aneth Field. Initial experiments evaluated reactivity and metals mobilization in Gothic Shale reacted with synthetic reservoir brine and brine + supercritical CO2. Information regarding trace metal mobilization was discussed in the previous progress report. The Appendix to this report presents a detailed description of experimental methods and preliminary experimental findings for major element chemistry.
Subsequent experiments evaluated reactivity and metals mobilization within the complete reservoir and caprock system (Desert Creek Limestone and the Gothic Shale) in brine and brine + supercritical CO2. All of these experiments have been completed, and the data are being analyzed. An abstract describing preliminary findings was submitted to and accepted by the Eleventh Annual Conference on Carbon Capture & Sequestration. A brief description of these preliminary findings follows.
Experiments react brine, limestone and/or shale caprock (water:rock = 20:1), and supercritical CO2 at 250 bars and 1600C. This temperature was selected to accelerate kinetics without changing in situ water-rock reactions. The rocks and minerals are 75wt% powder (<45 microns) to enhance reactivity and 25wt% fragments for textural analysis. The limestone consists of calcite and dolomite in equal proportions with trace amounts of pyrite. After a steady state is reached (~28 days), CO2 is injected to simulate sequestration into a deep aquifer. Samples are collected on a logarithmic time scale to evaluate changes in major and trace element water chemistry, dissolved carbon and sulfide, and pH. In these experiments, Pb, Fe, and Ba concentrations increase immediately after CO2 injection. Concentrations subsequently decrease to approximate steady-state values, likely due to mineral precipitation. In experiments that emulate the caprock-reservoir boundary, final Fe (0.7mg/kg) and Pb (0.05mg/kg) values exceed EPA limits, whereas Ba (0.140 mg/kg) values remain below EPA limits. In experiments that simulate deeper reservoir conditions, away from the caprock boundary, final Fe (3.5mg/kg) and Pb (0.017mg/kg) values indicate less mobilization than seen near the caprock, but values still exceed EPA limits. Barium concentrations always remain below the EPA limit of 2 mg/kg. Arsenic analyses are under evaluation. If these brines leak from a storage reservoir and mix with a potable aquifer, the results suggest that Ba will not be a contaminant of concern. However, Pb and Fe in brine may exceed the threshold allowed for human consumption and require careful attention in a sequestration scenario.
4.4 Risk Modeling
4.4.1 Basin scale hydrologic modeling
The parallel integrated hydrologic model ParFlow was used to simulate basin scale hydrology. This model was expanded into a streamline approach for reactive transport modeling as described in section 4.4.2.
4.4.2 Streamline modeling
A model represented complex hydrological fully kinetic geochemical reactions was developed. This model is a Lagrangian streamline approach where a large, heterogeneous three-dimensional flow field is reduced to a number of one-dimensional transport simulations. Each of these deconvolved, one-dimensional reactive transport problems are solved with an aqueous geochemical model (CrunchFlow) along a streamline, with the aim of representing complex geochemical reactive transport over a large domain in three-dimensional space. The streamline modeling approach allows for the mapping of these one-dimensional reactive transport simulations back onto a three-dimensional flow field, thus accounting for spatial heterogeneity within the aquifer and the complex aqueous geochemical processes. The efficiency of the streamline approach allows for a stochastic representation of uncertainty and variability inherent in groundwater systems.
A CO2 leakage scenario from a hypothetical CCS site is used where a resulting plume of CO2 lowers the groundwater pH and mobilizes metals from an existing mineral host-rock distribution. The plume migration and related pH buffering and kinetic dissolution/precipitation processes within the aquifer are simulated under varying degrees of hydraulic variation using this streamline-geochemical modeling approach. Metal concentrations and pH at groundwater pumping wells are then evaluated for responses to physical heterogeneity geochemical processes. Ensembles of 100 realizations capturing system uncertainty were performed to evaluate probable outcomes. Results show that as hydraulic heterogeneity increases, pH among source to well streamlines drops along with overall observed pumping well pH compared to domains with little hydraulic heterogeneity. Thus, through this novel model effort, a link is established between hydraulic heterogeneity and geochemical response. Furthermore, the complex interactions between geochemical processes and dynamic flow paths have implications for understanding the risks and observable outcomes to CCS implementation.
4.5 Evaluation of potential CO2 injection projects
This task has not yet started.
High Temperature-Pressure Laboratory Experiments
Analysis of experimental data will be completed with the goal of identifying specific mineralogy and metals mobilization of critical interest. These findings will be used to design and conduct the final phase of experiments.
Completion of batch experiments with carbonate rocks at elevated CO2 partial pressures will be completed by May 2012. Experimental design for upscaled (intermediate) scale experiments is underway. These experiments will be started in the summer of 2012.
Results from the batch experiments will be modeled using a Geochemist Workbench and/or PHREEQ-C to quantitatively assess the contribution of mineral dissolution and metal desorption to solution composition.
The compilation of an efficient streamline modeling approach that couples complex hydrological flow to a fully kinetic geochemical representation allows for a stochastic human health risk assessment. The next stages of research are to apply the streamline model to a human health risk assessment framework that accounts for exposure variability in human populations. Here the influence of including a fully kinetic geochemical representation in risk assessment will be evaluated. Preliminary work suggests that the geochemical processes may reduce the uncertainty in human health risk assessment, thereby providing valuable information to policymakers.
Journal Articles on this Report : 3 Displayed | Download in RIS Format
|Other project views:||All 70 publications||14 publications in selected types||All 14 journal articles|
||Bearup LA, Navarre-Sitchler AK, Maxwell RM, McCray JE. Kinetic metal release from competing processes in aquifers. Environmental Science & Technology 2012;46(12):6539-6547.||
||Navarre-Sitchler AK, Maxwell RM, Siirila ER, Hammond GE, Lichtner PC. Elucidating geochemical response of shallow heterogeneous aquifers to CO2 leakage using high-performance computing: implications for monitoring of CO2 sequestration. Advances in Water Resources 2013;53:45-55.||
||Siirila ER, Navarre-Sitchler AK, Maxwell RM, McCray JE. A quantative methodology to assess the risks to human health from CO2 leakage into groundwater. Advances in Water Resources 2012;36:146-164.||