Article pubs.acs.org/est
Integrated Framework for Assessing Impacts of CO2 Leakage on Groundwater Quality and Monitoring-Network Efficiency: Case Study at a CO2 Enhanced Oil Recovery Site Changbing Yang,*,† Susan D. Hovorka,† Ramón H. Treviño,† and Jesus Delgado-Alonso‡ †
Bureau of Economic Geology, The University of Texas at Austin, Austin, Texas 78759, United States Intelligent Optical Systems Inc., Torrance, California 90505, United States
‡
S Supporting Information *
ABSTRACT: This study presents a combined use of site characterization, laboratory experiments, single-well push−pull tests (PPTs), and reactive transport modeling to assess potential impacts of CO2 leakage on groundwater quality and leakage-detection ability of a groundwater monitoring network (GMN) in a potable aquifer at a CO2 enhanced oil recovery (CO2 EOR) site. Site characterization indicates that failures of plugged and abandoned wells are possible CO2 leakage pathways. Groundwater chemistry in the shallow aquifer is dominated mainly by silicate mineral weathering, and no CO2 leakage signals have been detected in the shallow aquifer. Results of the laboratory experiments and the field test show no obvious damage to groundwater chemistry should CO2 leakage occur and further were confirmed with a regional-scale reactive transport model (RSRTM) that was built upon the batch experiments and validated with the single-well PPT. Results of the RSRTM indicate that dissolved CO2 as an indicator for CO2 leakage detection works better than dissolved inorganic carbon, pH, and alkalinity at the CO2 EOR site. The detection ability of a GMN was assessed with monitoring efficiency, depending on various factors, including the natural hydraulic gradient, the leakage rate, the number of monitoring wells, the aquifer heterogeneity, and the time for a CO2 plume traveling to the monitoring well.
■
INTRODUCTION Carbon capture and geological sequestration is recognized as a viable technology for reducing greenhouse gas emissions into the atmosphere and also mitigating the effects of global climate change.1 Geological carbon sequestration (GCS) requires careful characterization of storage formations and overlying sealing units, in order to decrease the risk of unintended leakage. Ensuring protection of underground sources of drinking water (USDW) from unintended leakage of CO2 and/or brine through potential leakage pathways such as faults, fractures, and plugged and abandoned (P/A) wells is a focus for regulatory agencies because such leakage may pose risks to groundwater quality.2−8 Because no leakage has been detected during field observations for the past and existing GCS projects, various approaches, including laboratory experiments, 9−11 field tests,3,12−17 and numerical modeling,2,4,18−24 have been used to model impacts of CO2 leakage (should it occur in a site) on USDW by providing information not only about risks to groundwater quality in the event of CO2 leakage but also about the appropriateness of groundwater geochemistry to detect CO2 leakage signals. Laboratory experiments can provide direct observations of change in groundwater chemistry caused by CO2 leakage. However, one disadvantage of laboratory experiments is that small amounts of aquifer sediments are © 2015 American Chemical Society
exposed to water and CO2 under disturbed conditions that differ from actual field conditions in an aquifer, which may reduce their usefulness as a primary means of studying impacts of CO2 leakage on groundwater quality.3,13 Field tests can directly reveal impacts of CO2 leakage on groundwater quality, but they are much more expensive than laboratory experiments. Reactive transport modeling (RTM) simulates various hypothetical CO2 leakage scenarios that may potentially impact groundwater quality; the reliability of modeling results depends on simplifications of natural systems and uncertainties in model parameters.20 A combination of different approaches to the assessment of impacts of CO2 leakage on USDW can constrain uncertainties associated with individual approaches and increase the reliability of results.25 While various studies on potential impacts of CO2 leakage on USDW have been presented in the literature, it is not yet clear how a groundwater monitoring network (GMN) at GCS sites should be designed.7 Various solution methodologies for designing a GMN for hazardous waste sites, solid waste landfills, and other sites have been reported in the Received: Revised: Accepted: Published: 8887
March 28, 2015 June 5, 2015 June 8, 2015 June 8, 2015 DOI: 10.1021/acs.est.5b01574 Environ. Sci. Technol. 2015, 49, 8887−8898
Article
Environmental Science & Technology literature.26−28 Because there are differences between GCS and other waste disposal,29,30 a methodology to assess a geochemical monitoring network design for CO2 leakage detection is needed. The United States Environmental Protection Agency (USEPA) provides general guidance on the design of a GMN for Class VI wells; the design of a network is a key component of a monitoring system that serves to detect any leakage through the confining zone that may endanger USDW.31 Sun et al.32 present an optimal method for designing a pressure-based monitoring network for CO2 leakage detection. The objectives of this study are (1) to present a comprehensive case study of impacts of CO2 leakage on groundwater quality using an integrated framework that combines site characterization, laboratory experiments, single-well push−pull tests (PPTs), and RTM; and (2) to assess the detection ability of GMNs for CO2 leakage detection at a GCS site in terms of monitoring efficiency (ME).
was discovered in the 1940s and actively produced oil until the 1960s; all production wells were plugged and abandoned before 1966.36 As of 2012, a total of 4 million metric tons of CO2 for enhanced oil recovery (EOR) have been injected into the reservoir, which is hosted by basal sandstones and conglomerates of the lower Tuscaloosa Formation at a depth of ∼3000 m and capped by local fluvial mudrocks.37 Dark mudstones and fine-grained sandstones of the middle marine Tuscaloosa overlie on the lower Tuscaloosa Formation, providing the base of the confining system for CO2 storage during EOR. There are likely no natural pathways for CO2 to leak upward because (1) the hydrocarbon reservoir for CO2 injection is isolated by the multiple mudrocks and fine-grained sandstone from the USDW; and (2) the nontransmissive fault, though cutting through the northeast section of the oil field, cannot be mapped in the thick Midway formation.38 Therefore, the leakage pathways of concern are possible failure of the P/A wells at the site penetrated by the Tuscaloosa Formation (Figure S1, Supporting Information).
■
SITE CHARACTERIZATION Hydrogeological Characterization. The study area is located at about 25 km east of the city of Natchez, Adams County, Mississippi. The area is moderately hilly and covered in forest. The average annual precipitation is ∼137 cm.33 The base of USDW having a total dissolved solids (TDS) value of less than 10 000 mg/L ranges 400−600 m in depth.33−35 The local freshwater-bearing formation (TDS < 1000 mg/L) is primarily the Miocene Catahoula sands, predominantly fluvial to marginal-deltaic in terms of depositional setting and fluvial to marginal-deltaic in depositional origin, which overlie the mostly confining Jackson−Vicksburg Group.33 The Catahoula sands, ∼700 m thick, consist of three subunits: the upper unconfined− confined unit and the middle and lower confined units.33,35 A total of 13 groundwater wells, shown in Figure S1 (Supporting Information), were drilled into the upper subunit of the Catahoula aquifer. Twelve were for groundwater supply for oilproduction-well drilling and one well (UM-1) was drilled to support research and was continuously cored for analysis of geological stratigraphy and aquifer sediments. Regional groundwater flows generally westward (or southwestward locally), with an average regional hydraulic gradient of ∼0.5% in the study area.33,35 Groundwater Chemistry Characterization. From 2008 through 2014, a total of 12 field campaigns for collecting groundwater samples from the 13 groundwater wells have been completed (Figure S1, Supporting Information). All groundwater samples were collected and filtered through 0.45 μm filters after groundwater wells were purged for three to five casing volumes, preserved at 4 °C, and then, sent to chemical laboratories for analysis of concentrations of major ions and trace metals in the groundwater. Onsite measurements of groundwater pH were conducted with a Thermo Scientific Orion Star A121 pH portable meter. Groundwater alkalinity was titrated onsite with a Hach alkalinity digital titrator (model AL-DT) and was further calculated using the inflection-point method on the USGS alkalinity calculator (http://or.water. usgs.gov/alk/). A Hanna multiparameter meter (model HI 9828) was used onsite to measure groundwater temperature and TDS. A small number of aquifer sedimentary samples were analyzed using random-powder X-ray diffraction (XRD) on a Bruker AXS D8 diffractometer. CO2 Leakage Pathway Characterization. The Cranfield oil and gas field, shown in Figure S1 (Supporting Information)
■
LABORATORY EXPERIMENTS AND FIELD TESTS
Detailed information about the laboratory batch experiments and the single-well PPT is described by Yang et al.3 A brief description of the laboratory experiments and the field test is provided here. Batch Experiments of Water−Rock−CO2 Interactions. Two batch experiments were conducted at room temperature (24 °C): a reactive batch with about 106 g of aquifer sediments, and a control batch without aquifer sediments. Instead of distilled water used in the batch experiments reported in the literature,9,39 a total of 410 mL of groundwater collected from the shallow aquifer was added to each of the batches. Both batches were then sealed with a rubber stopper with holes for gas inlet and outlet and connected to a gas tank. CO2 gas was injected for 1 min every 20 min. The CO2 partial pressure in the head space of the batches was maintained at 1 × 105 Pa. Approximately 3−5 mL of water was periodically collected from the batches using a pipet. An aliquot was taken for pH measurement. The remaining sample was filtered through a 0.45 μm filter and split for analysis of major ion and trace metals. After CO2 gas was introduced, a total of 13 samples were collected from each batch over 650 h and sent to chemical laboratories for analysis of concentrations of major and trace elements. Single-Well PPT. A total of 3825 L of groundwater pumped from the aquifer, stored in water tanks, and then, equilibrated with CO2 was reinjected into the aquifer over 10 h. NaBr solution was added to the injected water as a tracer. During the incubation phase (∼55 h), the injected groundwater mixed with natural groundwater and reacted with aquifer sediments. During the pulling phase, the injected water/background water mixture was pumped back through the submersible pump and flowed through a measuring cell with a Hanna multiparameter meter, installed for continuously monitoring pH, conductivity, and temperature. The elapsed time and pumping rate were automatically recorded through an in-line flow meter. A total of 38 water samples were collected at incremental water volumes extracted from the discharge stream, filtered with a 0.45 μm cartridge filter, and sent to chemical laboratories for analysis of concentrations of major ions and trace metals. 8888
DOI: 10.1021/acs.est.5b01574 Environ. Sci. Technol. 2015, 49, 8887−8898
Article
Environmental Science & Technology
Figure 1. Plots of nine groundwater monitoring networks (GMNs) assessed with the regional-scale reactive transport model. (Solid triangles are the existing groundwater wells, blue crosses represent the hypothetical groundwater wells, and gray circles are the plugged and abandoned wells.) Note that GMN1 includes existing wells and GMN9 includes additional wells. GMN2−8 consist of entirely new monitoring wells. (Well density is 0.322 wells/km2 for GMN1, 0.124 wells/km2 for GMN2, 0.173 wells/km2 for GMN3, 0.223 wells/km2 for GMN4, 0.223 wells/km2 for GMN5, 0.371 wells/km2 for GMN6, 0.371 wells/km2 for GMN7, 0.866 wells/km2 for GMN8, and 0.742 wells/km2 for GMN9.)
■
REACTIVE TRANSPORT MODELING
dissolution rate constants of minerals, are listed in Table S1 (Supporting Information). Of the various geochemical processes capable of affecting the mobilization of trace metals from aquifer sediments to groundwater caused by CO2 leakage, three were considered in the model: dissolution of carbonates, cation-exchange reactions, and adsorption/desorption from the clay mineral surfaces. The presence of trace metals in carbonates as impurities (such as manganese and barium in crystal structures) is well documented in the literature.40 Adsorption/desorption processes can be influenced by various factors such as pH and the type of sorbents. In this study, two main clay minerals, kaolinite and illite, were considered principal adsorbents for As and Pb, mainly because the minerals are abundant in the aquifer sediments and have strong lead- and arsenic-adsorption capacities. A nonelectrostatic surface complex model was used for simulating the desorption/adsorption of arsenic and lead. Thermal dynamic constants of arsenic and lead to kaolinite and illite are listed in Table S1 (Supporting Information).
Identification of Major Geochemical Processes in the Batch Experiments. One challenge to applying RTM to predict potential impacts of CO2 leakage on groundwater chemistry is selection of geochemical processes that are related to CO2 intrusion in the aquifer. Although batch experiments have limitations,5,13,17 the results provide an opportunity for identifying major geochemical processes that can be used to set up the preliminary geochemical model. The preliminary geochemical model considers aqueous complexation, mineral precipitation/dissolution, cation exchange reactions, and adsorption/desorption. A total of 44 aqueous complexes that have concentrations high enough to impact the results were selected and listed in Table S1 (Supporting Information. Selection of minerals was based on the XRD analysis of the sedimentary samples. Mineral precipitation/dissolution was modeled under kinetic constraints. Equilibrium constants for aqueous complexation, mineral dissolution/precipitation, and log K, as well as the 8889
DOI: 10.1021/acs.est.5b01574 Environ. Sci. Technol. 2015, 49, 8887−8898
Article
Environmental Science & Technology
Figure 2. Plots of time evolutions of measurements of (a) pH, (b) Ca, (c) Mg, (d) Si, (e) Na, (f) K, (g) Ba, and (h) As, after CO2 gas was introduced into the flasks with adding aquifer sediments (orange circles) and without (blue triangles). Dashed lines indicate results with the geochemical model.
transport in the aquifer at the regional scale and the geochemical model validated by the field test. The model domain covers an area of 127.6 km2, ∼3 times the area of the CO2 EOR site (∼40 km2, Figure S2b, Supporting Information), to minimize impacts of boundary conditions on numerical solutions. The shallow aquifer is simplified to have a thickness of 7 m, though thickness may vary spatially in the field. Regional groundwater flows from right to left through the domain, with fixed hydraulic heads on both the right and left boundaries. Sensitivity runs of regional groundwater flow directions were also conducted. CO2 leaked through failed P/ A wells is assumed to be a dissolved gas in either fresh groundwater or brine (Table S2, Supporting Information); no gas phases are modeled in the aquifer. These wells were all considered in the model as equally likely to provide potential leakage pathways. Different mass rates of CO2 leakage, ranging ∼0.9−100 t/year, were simulated. The lower bound, reported by Nicot et al.,38 was estimated using a semianalytical solution of well leakage. The upper bound was estimated by assuming that 1% of 10 million tons of CO2 stored in the deep reservoir is leaked over 1000 years.42 The RSRTM simplified upward migration CO2 leakage through failed P/A wells from the deep reservoir to the depth of the aquifer. Numerical Simulator. The numerical code for simulating nonisothermal flow and reactive solute transport, CORE2D V4, coupled with biogeochemical processes in subsurface environments, was used in this study.43−46 Geochemical processes can be modeled under both local chemical equilibrium and kinetic
Cation-exchange reactions can also impact the mobilization of trace metals as reported by Zheng et al.,41 who conducted geochemical modeling of the CO2 release test at the ZERT site. In this study, cation-exchange reactions were considered as an alternative for the mobilization of trace metals caused by the codissolution of dolomite. The selectivity coefficients of the cation-exchange reactions are listed in Table S1 (Supporting Information). Initial groundwater chemistry in the geochemical model for simulating batch experiments is listed in Table S2 (Supporting Information). Validation of Geochemical Processes with the SingleWell PPT. The RTM of the single-well PPT includes the geochemical model tested with the batch experiments and the groundwater flow/transport to further validate the geochemical processes and calibrate the model parameters in the field. The cylindrical domain has a radius of 25 m and a thickness of 7 m (Figure S2a, Supporting Information). The top and bottom boundaries are no-flow boundaries. The testing well is located at the inner boundary. A constant head was imposed upon the outer boundary. The regional hydraulic gradient was neglected. The initial value of hydraulic conductivity was estimated on the basis of the pumping tests, compared with values reported in the literature.33,35 Porosity was estimated from sedimentary samples. Hydraulic conductivity, porosity, and the dispersion coefficient were further calibrated to match the tracer concentration measurements during the pulling phase. Regional-Scale RTM (RSRTM) of CO2 Leakage Scenarios through Failed P/A Wells. The RSRTM consists of flow/ 8890
DOI: 10.1021/acs.est.5b01574 Environ. Sci. Technol. 2015, 49, 8887−8898
Article
Environmental Science & Technology
Figure 3. Plots of (a) Br, (b) pH, (c) alkalinity, (d) Ca, (e) Mg, (f) Si, (g) K, (h) Ba, (i) As, and (j) Pb observed during the pulling phase of the single-well push−pull test (symbols) and the reactive transport modeling results. Dashed lines represent modeling results without considering heterogeneous reactions (mineral dissolution/precipitation and adsorption/desorption).
one of the wells of the GMN that is higher than one standard deviation of the groundwater chemistry data in the shallow aquifer.
conditions. Model parameters can be estimated using the Gauss−Newton/Levenberg−Marquardt method.47 This numerical code and its various components have been verified extensively against other codes and analytical solutions and also applied to various laboratory and field experimental results.48−50 Monitoring Network Efficiency. The RSRTM was used to assess the ME of 9 GMNs (Figure 1) and well density ranges from 0.12 to 0.87 wells/km2. ME is defined by the following equation: ME =
Wd WT
■
RESULTS AND DISCUSSION Groundwater Chemistry Characterization. A piper diagram of the groundwater chemistry data in the shallow aquifer (Figure S4, Supporting Information) suggests that groundwater types are mainly Na−Cl, Ca−Na−HCO3, and Ca−Na−HCO3−Cl.51 Statistics of groundwater geochemistry data collected from the 13 groundwater wells in the shallow aquifer are listed in Table S3 (Supporting Information). TDS of groundwater range 45−200 mg/L. Groundwater pH ranges 5.4−7.5, and alkalinity ranges 30−154 mg calcite/L. Concentrations of most trace metals are lower than the USEPA maximum contamination levels (MCLs) of drinking-water standards, except Mn and a few As samples having concentrations above the USEPA MCLs in the shallow aquifer (Table S3, Supporting Information). XRD analysis of the sedimentary samples shows that the aquifer sediments contain
(1)
where WT is the total number of P/A wells and Wd is the number of P/A wells where CO2 leakage can be detected by the GMN. CO2 leakage through a failed P/A well being detected with an indicator of dissolved CO2 by a monitoring well is shown in Figure S3 (Supporting Information). CO2 leakage from a P/A well is detected by a GMN if there is a change in dissolved inorganic carbon (DIC), dissolved CO2, or pH in any 8891
DOI: 10.1021/acs.est.5b01574 Environ. Sci. Technol. 2015, 49, 8887−8898
Article
Environmental Science & Technology (weight percent) ∼56% quartz, ∼16% microcline, ∼20% kaolinite, ∼6% illite, and
