Field Demonstration of CO2 Leakage Detection in Potable Aquifers

Nov 10, 2014 - This study presents two field pulselike CO2-release tests to demonstrate CO2 leakage detection in a shallow aquifer by monitoring ...
1 downloads 0 Views 4MB Size
Download PDF
Article pubs.acs.org/est

Field Demonstration of CO2 Leakage Detection in Potable Aquifers with a Pulselike CO2‑Release Test Changbing Yang,*,† Susan D. Hovorka,† Jesus Delgado-Alonso,‡ Patrick J. Mickler,† Ramón H. Treviño,† and Straun Phillips‡ †

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 two field pulselike CO2release tests to demonstrate CO2 leakage detection in a shallow aquifer by monitoring groundwater pH, alkalinity, and dissolved inorganic carbon (DIC) using the periodic groundwater sampling method and a fiber-optic CO2 sensor for realtime in situ monitoring of dissolved CO2 in groundwater. Measurements of groundwater pH, alkalinity, DIC, and dissolved CO2 clearly deviated from their background values, showing responses to CO2 leakage. Dissolved CO2 observed in the tests was highly sensitive in comparison to groundwater pH, DIC, and alkalinity. Comparison of the pulselike CO2release tests to other field tests suggests that pulselike CO2release tests can provide reliable assessment of geochemical parameters indicative of CO2 leakage. Measurements by the fiber-optic CO2 sensor, showing obvious leakage signals, demonstrated the potential of real-time in situ monitoring of dissolved CO2 for leakage detection at a geologic carbon sequestration (GCS) site. Results of a two-dimensional reactive transport model reproduced the geochemical measurements and confirmed that the decrease in groundwater pH and the increases in DIC and dissolved CO2 observed in the pulselike CO2-release tests were caused by dissolution of CO2 whereas alkalinity was likely affected by carbonate dissolution.

■

INTRODUCTION Geologic carbon sequestration (GCS) has been proposed as a promising option for mitigation of climate change by reducing emission of anthropogenic CO2.1 One concern with GCS is unintended CO2 leakage from the storage formations into overlying potable aquifers through faults, fractures, and active or abandoned wells, potentially impacting underground sources of drinking water (USDW). Groundwater monitoring in potable aquifers has been proposed so as to provide valuable information for risk assessment as well as CO2 leakage detection.2−8 Various geochemical parameters have been proposed for groundwater monitoring at GCS sites, on the basis of chemical analysis of groundwater samples periodically collected from potable aquifers.9−14 These proposed parameters include groundwater pH, dissolved inorganic carbon (DIC), hydroelectric conductivity, total dissolved solids, alkalinity, carbon and oxygen stable isotopes, trace metal concentrations, and noble gases. A recent review of geochemical parameters and monitoring tools and approaches in potable aquifers for CO2 leakage detection was presented by Humez et al.15 Numerical models, laboratory experiments, field tests, and natural and © 2014 American Chemical Society

industrial analogs have been used to assess sensitivity and applicability of geochemical parameters for CO2 leakage detection.16 Because field tests can provide in situ information on responses of geochemical parameters to CO2 leakage, a growing number of field tests have been conducted and reported in the literature.16 Generally, these field tests can be classified on the basis of numbers of testing and monitoring wells used: Type I, single-well push−pull tests, and Type II, multiple-well tests. For Type I tests, groundwater charged with CO2 is injected into the target aquifer and incubated for a certain time period. Then the groundwater is pumped back for chemical analysis. Examples of Type I testing are those conducted at the Brackenridge Field Laboratory,17 at the Cranfield shallow aquifer,9 at the Permian Lovede basin,18 and at the Newark Basin.19 One advantage of Type I tests is that only one well is needed. Type II tests use multiple monitoring and injection wells, as in the ZERT test,2 the Plant Daniel test,6 Received: Revised: Accepted: Published: 14031

September 8, 2014 November 7, 2014 November 10, 2014 November 10, 2014 dx.doi.org/10.1021/es5044003 | Environ. Sci. Technol. 2014, 48, 14031−14040

Environmental Science & Technology

Article

Figure 1. Schematic of (a) configuration of testing well at Brackenridge Field Laboratory (During 2 h of the CO2 bubbling procedure, the well was open to air. After bubbling was stopped, the well was capped. The drawing is not to scale.) and (b) principle of operation of fiber-optic CO2 sensor: CO2 triggers a color change in the fiber cladding, resulting in a change in fiber attenuation at wavelengths related to color change.

the pilot-scale test at the Vrogum site,20 the test at the CO2 Field Lab,21 and the shallow-aquifer test in Northeast Germany.22 Type II testing can simulate migration of CO2 leakage from a leakage location (injection well) to a monitoring well in an aquifer. In this study, we present a novel field test, a pulselike CO2release test, to assess geochemical responses to CO2 leakage in potable aquifers. Similar to a single-well push−pull test, a pulselike CO2-release test needs just one testing well. However, unlike in a single-well push−pull test, CO2 gas is directly released into a testing well and geochemical responses are continuously monitored during the retreat phase of a pulselike CO2-release test. At the Brackenridge Field Laboratory, two pulselike CO2-release tests were conducted to assess responses of groundwater pH, alkalinity, DIC, and dissolved CO2 which were simulated with a geochemical model by Yang et al.16 The model results indicate that responses of groundwater pH and alkalinity to CO2 leakage depend on the presence of carbonates in aquifer sediments. Alkalinity is much more sensitive to CO2 leakage in carbonate-rich aquifers than in carbonate-poor ones; whereas groundwater pH is more sensitive to CO2 leakage in carbonate-poor aquifers than in carbonate-rich aquifers.16 Dissolved CO2 and DIC show high sensitivity to CO2 leakage regardless of the presence or absence of carbonates in aquifer sediments.16 Current groundwater monitoring of geochemical parameters for CO2 leakage detection relies mainly on periodic groundwater sampling from water wells in potable aquifers.23 A periodic sampling method for monitoring groundwater

chemistry in potable aquifers has been applied to various GCS demonstration projects,12,24−26 as well as to the field tests. Owing to advancements in technology and materials (especially optic fiber) during the past decade, various chemical sensors are currently available for real-time measurements of in situ dissolved CO2 in water.27,28 One type of chemical sensor, a fiber-optic CO2 sensor, has been proposed for detection of supercritical CO2 in brine.29 The current study aims also to test real-time in situ monitoring of dissolved CO2 in groundwater using a fiber-optic sensor for CO2 leakage detection in potable aquifers.

■

MATERIALS AND METHODS Brackenridge Field Laboratory. A siliclastic-dominated alluvial aquifer adjacent to the Colorado River in Austin, Texas, was selected for the field test.17,24 The shallow sand and gravel aquifer is on a river terrace at The University of Texas Brackenridge Field Laboratory (lat. 30° 16′ 15.27″ N, long. 97° 46′ 51.16″ W). Several shallow groundwater wells were completed with 2-in. polyvinyl chloride (PVC) pipe to the limestone bedrock at a depth of 6 m and screened in the lower 3 m (Figure 1a). A photo of the testing site is shown in Figure S1. The potentiometric surface at the testing well in the unconfined aquifer was measured at 2.4 m depth with a regional groundwater flow toward the river, dominated mainly by the topography. Core sediments recovered from the shallow aquifer show that the sediments vary in grain size from clay-rich layers to coarse sandy gravels, with most of the sediments in the medium- to coarse-grained sand range.24 Mineralogical analysis 14032

dx.doi.org/10.1021/es5044003 | Environ. Sci. Technol. 2014, 48, 14031−14040

Environmental Science & Technology

Article

were conducted with a Thermo Scientific Orion Star A121 pH portable meter. Groundwater alkalinity was titrated on site 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/). Groundwater samples were sent to the geochemical laboratories of the Jackson School of Geosciences, The University of Texas at Austin, for chemical analysis. Cation and anion concentrations of water samples were analyzed on two Dionex ICS-2000 ion chromatography systems equipped with autoeluent generators, an AS-HV autosampler, and an AD25 absorbance detector. Dissolved inorganic carbon was analyzed with the Torch Combustion TOC analyzer. During collection and preservation of these groundwater samples, one issue is the potential degassing of CO2,which would impact on DIC measurements.9 To reduce potential CO2 degassing, we stored groundwater samples in serum vials sealed with septa (Figure S3). In addition to the periodic groundwater sampling method for monitoring groundwater chemistry, a fiber-optic CO2 sensor, newly developed by Intelligent Optical Systems, Inc.,31 was installed in the lower 3 m of the testing well for real-time in situ monitoring of dissolved CO2 (Figure 1a). The sensor was configured to measure dissolved CO2 every 30 s. The fiberoptic sensor includes two main parts: the optic fiber and an optoelectronic subsystem (Figure 1a). The optic fiber consists of a light-guiding core with a cladding that has a lower refractive index, confining the propagating light by total internal reflection (Figure 1b). The cladding was made up of a polymer containing a small percentage of a colorimetric indicator (Figure 1b) that changes color in the presence of CO2.32 On contact with any segment of the fiber, dissolved CO2 in groundwater diffuses into the cladding and interacts with the indicator, causing the cladding to change color. The optoelectronic subsystem illuminates the fiber and detects the characteristic change in color, the extent of which depends on how much of the chemical indicator has interacted with dissolved CO2 in groundwater (Figure 1b). The output from the sensors represents a depth average over the segment in the testing well. The detection limit of fiber-optic sensors is ∼0.5 mg/L, and the limit of quantification is ∼2 mg/L (dissolved CO2). Detail information about fabrication, operation, and laboratory calibration of a fiber-optic CO2 sensor can be found in the work of Delgado-Alonso and Lieberman.31 Groundwater temperature in the testing well was continuously measured with a temperature sensor (model: RTDTemp101A by MADGETECH) every half hour from July 2 to 20. In addition, a weather station was installed at the site, about 50 m away from the testing well to collect information on air temperature and atmospheric pressure from July 2 to 20. Determination of CO2 Leakage Signal. It is of particular importance to determine CO2 leakage signal based on geochemical responses in a monitoring well. Determination of CO2 leakage may depend on two major factors: sensitivity of the geochemical parameter itself to CO2 leakage and spatial and temporal variations of the geochemical parameter at the study area. An example of the first factor could be that groundwater pH is more sensitive to CO2 leakage in a carbonate-poor aquifer than in a carbonate-rich aquifer. The second factor is site-specific, representative of background information on a geochemical parameter in potable aquifers. Sensitivity of a geochemical parameter can be expressed by a relative change in a geochemical parameter as the following equation:16

of the sedimentary samples indicates that the aquifer sediments contain (weight percent) 17% calcite, tD suggests CO2 leakage signal detected. tD, detection time, is the time moment that Sp exceeds α, revealing a CO2 leakage signal. Sp < α tells CO2 leakage signal has been not detected in the monitoring well. Note that detection time, tD, could be geochemical parameter dependent because response of each geochemical parameter varies to CO2 leakage. Furthermore, detection time could be different than the arrival time (tA) of a CO2 plume to a monitoring well if a CO2 plume arrives to the monitoring well but Sp < α. 14034

dx.doi.org/10.1021/es5044003 | Environ. Sci. Technol. 2014, 48, 14031−14040

Environmental Science & Technology

Article

Figure 3. Plots of (a) comparison of the direct measurements of dissolved CO2 (DMC) with the fiber-optic sensor and the indirect estimates of dissolved CO2 (IEC) from groundwater pH and alkalinity using PHREEQC and (b) histogram of differences of dissolved CO2 obtained from the two methods. (Note that dissolved CO2 shown in the plots were measured with the fiber-optic sensor at the same time that the groundwater samples were collected for groundwater pH and alkalinity measurements. About 56% of the differences of dissolved CO2 from the two methods are close to 0.)

CO2 releases (Table S4) with CO2 gas at one atmospheric pressure, calculated with PHREEQC,44 suggests little increase in alkalinity. Therefore, increases in alkalinity shown in Figure 2b were likely as a result of carbonate dissolution. As expected, after CO2 was released into the testing well, DIC showed a quick increase because of CO2 dissolution into the groundwater. After the CO2 releases were stopped, DIC gradually recovered from ∼0.04 mol/L to its background value at ∼0.01 mol/L (Figure 2c). Dissolved CO2 in groundwater was acquired with two methods: the direct measurements of dissolved CO2 (DMC) using the fiber-optic sensor, and the indirect estimates of dissolved CO2 (IEC) from groundwater pH and alkalinity calculated with PHREEQC.44 Once CO2 gas was bubbled into the testing well, dissolved CO2 increased quickly to ∼0.017 mol/L (Figure 2d). After CO2 release was stopped, DMC and IEC gradually decreased because of dissolved CO2 migration away from the testing well into the surrounding aquifer. A oneto-one comparison of DMC and IEC is shown in Figure 3a. It can be seen that most data points of DMC and IEC fall along the 1:1 line, except 3 of 27 data points that IEC is higher than DMC, possibly caused by measurement errors in alkalinity. Statistical analysis suggests that the 95% confidence interval for the mean difference between IEC and DMC fall within the range [−0.000556 mol/L, 0.000844 mol/L]. Distribution of the differences in IEC and DMC shows that only three data points that differences between IEC and DMC are greater than 0.003 mol/L (Figure 3b). A paired t test was then run to determine whether there is a statistically significant difference in dissolved CO2 acquired using the two methods. The paired t test shows that the t-value is 0.42 and p-value is 0.676 for the 27 pairs of IEC and DMC, suggesting that there is no statistically significant difference between IEC and DMC. Hence, the field tests confirm that the direct measurements of dissolved CO2 with the fiber-optic sensor was a good match to the indirect estimates of dissolved CO2 on the basis of measurements of groundwater pH and alkalinity using PHREEQC,44 suggesting that the fiber-optic sensor can provide reliable measurements of dissolved CO2 in groundwater.

discretized into 583 triangles and 1124 nodes (Figure S4). The testing well located at the center of the domain was represented with 102 triangular elements (Figure S4). The size of the triangular elements gradually increases from the center outward. The shallow aquifer was assumed to be homogeneous with a uniform thickness of 3.6 m. Aquifer porosity and hydraulic conductivity are 0.2 and 1 m/day, respectively, calibrated with the measurements of bromide concentrations in the second CO2-release test.24 Groundwater composition measurements of the samples collected from the testing well prior to the CO2 releases were used as the initial condition in the aquifer, and groundwater composition measurements of the samples collected prior to cease of the CO2 releases were used as the initial condition in the testing well (Table S4). The total simulation time in the model is 20 days after the first CO2 release was stopped. The simulator, INVERSE-BIOCORE2D was used in this study. That code and other CORE code series have been extensively verified and applied to various laboratory experiments and field cases.24,35,39−43

■

RESULTS AND DISCUSSION Results of the Field Tests. Measurements of groundwater pH, alkalinity, and DIC for the total 27 samples collected in the two pulselike CO2-release tests are listed in Table S5. Quantification of CO2 dissolved into the groundwater in the testing well is given in Section 1 of the Supporting Information. Groundwater pH, alkalinity, DIC, and dissolved CO2 as a function of elapsed time after the first CO2 release was stopped, are shown in Figure 2. The background value of groundwater pH is ∼7. Once CO2 was released into the testing well, groundwater pH showed a sharp decrease to ∼6 from its background value at the end of each CO2 release (Table S5). After CO2 release was stopped, groundwater pH gradually, over about a week, recovered to its background value (Figure 2a). The background value of alkalinity in the shallow aquifer was measured at ∼0.007 mol/L. During the CO2 releases, measurements of alkalinity showed a quick increase and reached ∼0.017 mol/L. After the CO2 releases were stopped, alkalinity gradually decreased to its background value (Figure 2b). Equilibrating the groundwater in the aquifer prior to the 14035

dx.doi.org/10.1021/es5044003 | Environ. Sci. Technol. 2014, 48, 14031−14040

Environmental Science & Technology

Article

Because the fiber-optic sensor measured dissolved CO2 in groundwater at 30-s intervals, it picked up the ambient fluctuations of dissolved CO2 in groundwater (Figure S5a). It can be seen that dissolved CO2 varied from ∼5 × 10−3 mol/L during day time to ∼1.5 × 10−3 mol/L during night time (Figure S5a). Such variations in dissolved CO2 could possibly be due to (1) impacts of daily groundwater temperature on CO2 solubility; (2) impacts of daily change in atmospheric pressure on CO2 solubility; and (3) impacts of daily temperature on microbial processes and root respiration, leading to further variations in dissolved CO2 in groundwater. Groundwater temperature, measured with the RTDTemp101A sensor, ranged from 25 to 25.7 °C, showing a pattern opposite to that of dissolved CO2 (Figure S5a). Measurements of atmospheric pressure varied from ∼992 to 1002 mbar from July 2 through 12 (Figure S5b). A model calculation with PHREEQC44 suggests that such small fluctuations in groundwater temperature and atmospheric pressure led to less than 0.3% change in CO2 solubility, much smaller than variations in dissolved CO2 observed (Figure S5a). Therefore, variations in dissolved CO2 were likely due to microbial activities and root respiration, which were dependent on temperature. This pattern is consistent with the finding of a soil gas study at this site in 2008, in which measurements of soil CO2 also showed daily fluctuations, attributed mainly to microbial activities and root respiration (Figure S5c).24 Measurements of groundwater pH, alkalinity, DIC and dissolved CO2 in terms of leakage time (eq 1) are shown in Figure S6. Time, tB, in eq 1, was set to the 20th day after the first CO2 release provided that groundwater chemistry was recovered to the background condition. Groundwater pH gradually decreased and reached a minimum value when CO2 was released (Figure S6a). Alkalinity, DIC, and dissolved CO2 gradually increased to ∼2 to 12 times their background values (Figure S6). Deviation of groundwater pH, alkalinity, DIC, and dissolved CO2 from their background values clearly showed CO2 leakage signals in the monitoring well, suggesting that measurements of groundwater pH, alkalinity, DIC, and dissolved CO2 could be used for CO2 leakage detection in the shallow aquifer of the testing site. Determination of the CO2 Leakage Signal. Relative changes in groundwater pH, alkalinity, DIC, and dissolved CO2 (eq 2) are shown in Figure 4. Apparently, among the four parameters, dissolved CO2 showed the greatest sensitivity to CO2 leakage, and pH exhibited the lowest sensitivity (Figure 4). Alkalinity and DIC showed a similar response to CO2 leakage (Figure 4). This result is consistent with the modelbased findings in a carbonate-rich aquifer reported by Yang et al.16 Because the number of measurements of groundwater pH, alkalinity, and DIC at the shallow aquifer at the Brackenridge Field Laboratory were not enough to estimate their coefficients of variation, as an illustrating example we used measurements of groundwater pH, alkalinity, and DIC from the SACROC shallow aquifer where groundwater samples were collected from 30 groundwater wells over an area of 450 km2 from June 2007 to August 2008.45 Calculated coefficients of variation of groundwater pH, alkalinity, and DIC are 0.064, 0.3, 0.3, respectively (Figure 4). The fiber-optic sensor provided over thousands of measurements of dissolved CO2 after the dissolved CO2 was recovered to its background condition (Figure S5), yielding an estimate of the coefficient of variation of dissolved CO2 at 0.5. By comparing relative changes in groundwater pH, alkalinity, DIC, and dissolved CO2, it appears

Figure 4. Plots of relative changes in pH, dissolved CO2, DIC, and alkalinity versus leakage time (tL = tB − tE, tE is elapsed time after first CO2 release and tB is 20th day after first CO2 release).

that dissolved CO2 shows the earliest signal of CO2 leakage to the monitoring well, and groundwater pH exhibits relatively a late signal (Figure 4). Modeling Results. The pulselike CO2 releases were simulated with the 2D reactive transport model. The model was calibrated to fit bromide concentrations measured during the retreat phase of the second CO2 release. The model result reproduces well Br concentration measurements (Figure S7a). Comparison of the modeling results with measurements of groundwater pH, alkalinity, DIC, and dissolved CO2 for the two CO2-release tests are shown in Figure 2 and Figure S6. It can be seen that the modeling results reproduced overall trends of the geochemical parameters observed during the retreat phases. The model confirms that increase in alkalinity (Figure S6b) was due mainly to carbonate dissolution, as evidenced by increases in concentrations of Ca (Figure S7b) and Mg (Figure S7c). Without considering carbonate dissolution, modeled values of alkalinity, Ca, and Mg are much lower than their measurements (Figure S8). The modeling results also suggest that the sharp decrease in pH (Figure S6a) and quick increase in dissolved CO2 (Figure S6d) were due mainly to CO2 dissolution during the period of CO2 releases while the increase in DIC was caused by both CO2 releases and carbonate dissolution (Figure S6c). Comparison of the Pulselike CO2-Release Tests with Other Field Tests. It is particularly interesting to compare how well the pulselike CO2 release tests for assessing responses of geochemical parameters performed in comparison to other field tests. Five field tests from the literature were selected because their data are publically accessible. PHREEQC44 was used to calculate CO2 pressure from the water chemistry data in the field tests. Comparison of groundwater pH, alkalinity, and DIC from the six field tests in terms of CO2 pressure is shown in Figure 5. It appears that the six field tests can be divided into 14036

dx.doi.org/10.1021/es5044003 | Environ. Sci. Technol. 2014, 48, 14031−14040

Environmental Science & Technology

Article

Figure 5. Plots of (a) groundwater pH; (b) alkalinity; and (c) DIC as a function of CO2 pressure. Data from six field tests (BFL_PLR: pulselike CO2-release tests at Brackenridge Field Laboratory, this study. BFL_PPT:17 single-well push−pull test conducted at Brackenridge Field Laboratory. Cranfield_PPT:9 single-well push−pull test at Cranfield shallow aquifer. ZERT_B22: 2008 CO2 injection at ZERT site. Lodeve_PPT:18 single-well push−pull test at Lodeve, southern France. Newark_PPT:19 2012 single-well push−pull test at Newark Basin). Note that CO2 pressure was estimated using modeling tool PHREEQC, on the basis of groundwater chemistry data.

tests (Figure S9c). Because aquifer sediments in the Newark Basin contain plagioclase (1.7%) in addition to calcite (0.8%) and dolomite (0.18%),19 introducing CO2 into the aquifer led to plagioclase dissolution, which contributed to increases in alkalinity and Ca. Alkalinity and Ca observed in the test conducted at the Newark Basin show more complexity than in the other five tests, especially when CO2 pressure is higher than 0.5 atm (Figures 5b and S9b). Comparison of the pulselike CO2 release tests to the other five field tests shows that the pulselike CO2-release can provide reliable assessment of responses of geochemical parameters to CO2 leakage. To further compare the selected field tests, we plotted concentrations of HCO3 versus SiO2 from the field tests (Figure S10). Because SiO2 concentrations were not measured in this study, the pulselike CO2-release tests were not included in Figure S10. As suggested by Hounslow,46 the molar ratio of bicarbonate to silica in water can reveal whether carbonate or silicate mineral weathering dominates groundwater chemistry. Groundwater chemistry is dominated by carbonate mineral weathering if the ratio of bicarbonate to silica >10 or by silicate mineral weathering if the ratio of bicarbonate to silica